Abstract
Accurate prediction of in vivo hepatic drug clearance using in vitro assays is important to properly estimate clinical dosing regimens. Clearance of low-turnover compounds is especially difficult to predict using short-lived suspensions of unpooled primary human hepatocytes (PHHs) and functionally declining PHH monolayers. Micropatterned cocultures (MPCCs) of PHHs and 3T3-J2 fibroblasts have been shown previously to display major liver functions for several weeks in vitro. In this study, we first characterized long-term activities of major cytochrome P450 enzymes in MPCCs created from unpooled cryopreserved PHH donors. MPCCs were then used to predict the clearance of 26 drugs that exhibit a wide range of turnover rates in vivo (0.05–19.5 ml/min per kilogram). MPCCs predicted 73, 92, and 96% of drug clearance values for all tested drugs within 2-fold, 3-fold, and 4-fold of in vivo values, respectively. There was good correlation (R2 = 0.94, slope = 1.05) of predictions between the two PHH donors. On the other hand, suspension hepatocytes and conventional monolayers created from the same donor had significantly reduced predictive capacity (i.e., 30–50% clearance values within 4-fold of in vivo), and were not able to metabolize several drugs. Finally, we modulated drug clearance in MPCCs by inducing or inhibiting P450s. Rifampin-mediated CYP3A4 induction increased midazolam clearance by 73%, while CYP3A4 inhibition with ritonavir decreased midazolam clearance by 79%. Similarly, quinidine-mediated CYP2D6 inhibition reduced clearance of dextromethorphan and desipramine by 71 and 22%, respectively. In conclusion, MPCCs created using cryopreserved unpooled PHHs can be used for drug clearance predictions and to model drug-drug interactions.
Introduction
Metabolism by the liver accounts for the overall clearance of ∼70% of marketed drugs (Wienkers and Heath, 2005). Thus, accurate prediction of in vivo human hepatic clearance using preclinical models is important to set drug doses in the clinic (Ring et al., 2011). Significant species-specific differences in liver pathways can lead to inaccuracies in the predictions of human drug clearance when using animals (Shih et al., 1999). Therefore, in vitro human liver models are now used increasingly to predict human drug clearance (Di and Obach, 2015). Whereas human liver microsomes are useful for evaluating cytochrome P450 (P450)-mediated drug clearance in a high-throughput screening format, the lack of phase-II enzymes and membrane-bound transporters limits their utility for predicting clearances of different drug types. On the other hand, while retaining the complete architecture and cell types of the liver, liver slices are not amenable to high-throughput screening. Cancerous hepatic cell lines can be expanded cheaply and nearly indefinitely; however, they suffer from abnormal P450 levels and only represent single donors (Wilkening et al., 2003). Thus, cryopreserved primary human hepatocytes (PHHs), which can be sourced from multiple donors, are ideal for on-demand assessment of drug disposition since they integrate all of the relevant metabolic pathways of the liver (Godoy et al., 2013).
PHHs can be kept viable in suspension for 4–6 hours or plated in confluent monolayers on adsorbed collagen for a few days. For suspension PHHs, pooling can mitigate the large functional variability inherent across PHH donor lots. However, the limited incubation time using suspension hepatocytes often does not allow low-turnover drugs to deplete sufficiently to predict in vivo clearance (Brown et al., 2007; Ring et al., 2011). Low-turnover drugs are being developed increasingly for one-pill-a-day dosing regimens, and often rank ordering of candidate compounds in a chemical series by clearance rates is necessary to progress with development. The relay method has addressed this limitation by transferring the drug-laden supernatant from 4-hour pooled PHH suspension incubations to freshly thawed PHHs to allow active enzymes to metabolize the drugs for prolonged times (∼20 hours) (Di et al., 2012). However, this method requires at least 5-fold more PHHs (10+ pooled donors) than a single incubation, thereby depleting a limited lot faster and necessitating screening and large-scale banking of newer pooled lots. Additionally, suspension hepatocytes do not properly polarize with appropriate localization of transporters to the apical and basolateral domains, which is limiting for predicting clearance of drugs that are transporter substrates. Whereas plated monolayers prolong PHH viability for a few days and show polarized phenotype when overlaid with an extracellular matrix gel (Bi et al., 2006), the P450 activities rapidly decline to <10% of levels observed in freshly isolated PHHs (Lecluyse, 2001; Khetani and Bhatia, 2008).
Organizing hepatocytes using engineering tools and coculture with stromal cells can help maintain hepatic functions for prolonged times and at levels higher than is possible with conventional monolayers, which has improved predictive capacities for drug studies (Khetani et al., 2015). Khetani and Bhatia (2008) developed a micropatterned coculture (MPCC) model in which PHHs are organized onto collagen-coated domains of empirically optimized dimensions and subsequently surrounded by 3T3-J2 murine embryonic fibroblasts. Major hepatic functions (i.e., drug metabolism enzymes, transporters) are stable in MPCCs for ∼4 weeks. Owing to the reduced number of PHHs (∼10% of confluent monolayers), MPCCs were incubated for up to 7 days without a medium change, which led to detection of a greater number of clinically relevant metabolites than was possible with suspension PHHs (Wang et al., 2010). Chan et al. (2013) also used the 7-day drug incubations in MPCCs to predict clearance of low-turnover compounds. However, drugs with high clearance rates were not tested, nor was MPCC performance compared to suspension hepatocytes and plated monolayers using the same donor. Furthermore, it remains unclear whether MPCCs created using cryopreserved PHHs from multiple donors can maintain high levels of major P450s for several weeks, which could allow initiation of drug incubation at different culture ages. Therefore, here we sought to determine levels and longevity of major P450 enzymes in MPCCs created from cryopreserved PHH donors (unpooled) in a 96-well plate format. We then used MPCCs to predict clearance rates of 26 drugs with a wide range of in vivo turnover rates, and compared results for a subset of these drugs across MPCCs, suspension PHHs, and plated monolayers created from the same donor. Finally, we assessed the effects of drug-mediated P450 modulation on drug clearance rates to mimic drug-drug interaction (DDI) scenarios.
Materials and Methods
Culture of Primary Human Hepatocytes.
Cryopreserved PHHs were purchased from vendors permitted to sell products derived from human organs procured in the United States by federally designated Organ Procurement Organizations (BioreclamationIVT, Baltimore, MD; Triangle Research Laboratories, Research Triangle Park, NC; Life Technologies, Carlsbad, CA). Information (lot classifier, age, sex, ethnicity, cause of death, available medical history) on the PHH lots is provided in Supplemental Table 1. PHH vials were thawed at 37°C for 120 seconds and diluted with 25 ml of prewarmed KryoThaw I (SciKon, Chapel Hill, NC). The cell suspension was then spun at 50g for 10 minutes, the supernatant was discarded, and the cells were resuspended in hepatocyte seeding medium, the formulation of which was described previously (Khetani et al., 2013). Hepatocyte viability was assessed using Trypan blue exclusion (typically 80–95%). Liver-derived nonparenchymal cells were consistently found to be less than 1% of all the cells.
MPCCs were created as previously described (Berger et al., 2014). Briefly, adsorbed collagen was lithographically patterned in each well of a multiwell plate to create 500-μm diameter circular domains spaced 1200 μm apart, center-to-center. Hepatocytes selectively attached to the collagen domains leaving ∼4500 attached hepatocytes on ∼13 collagen-coated islands within each well of a 96-well plate. 3T3-J2 murine embryonic fibroblasts were seeded 18–24 hours later in each well to create MPCCs. Serum-supplemented culture medium, the formulation of which has been described previously (Ramsden et al., 2014), was replaced on cultures every 2 days (∼64 μl/well).
To create suspension cultures, 5.6 × 104 hepatocytes were dispensed into each well (32 μl serum-free culture medium per well) of an uncoated 96-well plate. To create conventional confluent monolayers, 5.6 × 104 hepatocytes were seeded into each well (64 μl serum-supplemented culture medium per well) of a 96-well plate coated with rat-tail collagen type I as described previously (Khetani and Bhatia, 2008). Serum-free culture medium for both suspension cultures and conventional monolayers was composed of William’s E base (Sigma-Aldrich, St. Louis, MO), 15 mM HEPES buffer (Corning Cellgro, Manassas, VA), 1% vol/vol ITS+ supplement (Corning Life Sciences, Tewksbury, MA), 1% vol/vol penicillin-streptomycin (Corning Cellgro), 100 nM dexamethasone (Sigma-Aldrich), 0.2% vol/vol amphotericin B (Life Technologies), and 0.01% vol/vol gentamycin (Life Technologies). Fetal bovine serum (Life Technologies) was added at 10% vol/vol for seeding conventional monolayers for 24 hours and then cultures were switched to the serum-free formulation above for drug dosing studies.
Hepatocyte Functionality Assays.
Urea concentration in supernatants was assayed using a colorimetric endpoint assay utilizing diacetyl monoxime with acid and heat (Stanbio Laboratories, Boerne, TX). Albumin levels were measured using an enzyme-linked immunosorbent assay (MP Biomedicals, Irvine, CA) with horseradish peroxidase detection and 3,3′,5,5′-tetramethylbenzidine (TMB; Fitzgerald Industries, Concord, MA) as the substrate. CYP2C9 activity in cultures was measured using a luminescence-based assay (CYP2C9-glo, luciferin-H) by Promega (Madison, WI). Following incubation for 3 hours with the CYP2C9-glo substrate in serum-free dosing culture medium, culture supernatants were processed according to manufacturer instructions and luminescence was measured using a luminometer (BioTek, Winooski, VT). In addition, activities of major P450s in MPCCs were assessed using substrates shown in Supplemental Table 2. Cultures were incubated with these substrates for 1 hour in serum-free dosing culture medium. Supernatants from cultures were frozen at –80°C prior to further analysis.
Drug Dosing.
MPCCs were allowed 7–9 days to functionally stabilize and then dosed in serum-free culture medium with a set of 26 drugs (Table 1); their known in vivo turnover rates ranged between 0.05 ml/min per kilogram and 19.5 ml/min per kilogram. Conventional PHH monolayers were allowed 24 hours to acclimate before dosing in serum-free culture medium with a subset (10 total) of the drugs (Table 2). Suspension hepatocytes were dosed immediately after dispensing in wells with the same subset of drugs as those tested on conventional monolayers (Table 3). All drug solutions were prepared in serum-free culture medium at 1 μM and placed on the cells (64 μl total volume per well). MPCCs were incubated with drug solutions for up to 7 days without a medium change; conventional monolayers were dosed for up to 4 days; and, suspension cultures were dosed for up to 4 hours. Supernatants (50 μl) from representative wells (single time point per well of a 96-well plate) were collected at six to seven time points spread across the time series for each type of culture model. For MPCCs and conventional monolayers, removal of the supernatants was sufficient to stop the reaction with the cells. For suspension hepatocytes, however, mixing the cell suspension with 100 μl of acetonitrile was necessary to quench the reaction. All samples were immediately frozen at –80°C prior to further analysis.
For DDI studies, MPCCs stabilized for 7 days were treated with serum-supplemented culture medium containing P450 inducer (12.5 μM rifampin for CYP3A4) for 3 days or P450 inhibitor (0.5 μM ritonavir for CYP3A4 and 4 μM quinidine for CYP2D6) for 18 hours. Inducer and inhibitor concentrations were chosen on the basis of preliminary experiments that evaluated effects on the pertinent P450s (data not shown). Cultures were subsequently dosed in serum-free culture medium with 1 μM P450 substrates (midazolam for CYP3A4, and desipramine and dextromethorphan for CYP2D6) with or without P450 inducer/inhibitor. Control cultures were treated with dimethyl sulfoxide (Corning Cellgro) alone (0.1% vol/vol) for the aforementioned time periods before dosing with P450 substrates. Sample collection was carried out as described above prior to further analysis.
Liquid Chromatography–Mass Spectrometry Analysis.
Liquid chromatography–mass spectrometry analysis on culture supernatants (crashed with acetonitrile) was carried out by Integrated Analytical Solutions (Berkeley, CA). The amount of substrate metabolite or parent compound was measured using an Applied Biosystems/MDS Sciex API 3000 mass spectrometer (Foster City, CA) coupled to a Shimadzu VP System (Columbia, MD). The liquid chromatography mobile phases consisted of: A, water containing 0.2% formic acid; and B, methanol containing 0.2% formic acid. Samples were eluted through a Luna Mercury C8(2) column (2 × 30 mm; Phenomenex Inc., Torrance, CA), a Luna Mercury Hydro-RP column (2 × 3 mm), a Duragel G C18 column (2 × 10 mm; Peeke Scientific, Redwood City, CA), or a Titan 200 C18 column (2.1 × 30 mm; Sigma-Aldrich, St. Louis, MO). Solvent gradients from 0% (B) or 5% (B) to 95% (B) over the course of 1–2 minutes at flow rates between 0.4 ml/min and 0.8 ml/min were used to elute the compounds from the columns. Injection volumes ranged from 2 to100 μl for analysis.
Data Analysis.
The natural logarithm of the concentration of drug remaining in culture supernatants was first plotted against time of incubation. Further analysis of the data as described below was conducted only if the correlation coefficient of the linear fit to the natural log–transformed data was greater than 0.8. If the correlation coefficient was less than 0.8, the data set was deemed not usable for prediction of clearance. The in vitro depletion half-lives of the drugs in culture supernatants were calculated by eq. 1, where “slope” was derived from the natural logarithm of the concentration of drug remaining plotted against time. For most of the compounds, the calculated half-life was within the maximal incubation time (i.e., no extrapolation) for both MPCCs and conventional monolayers, except for theophylline (both donors ∼5-fold higher half-life than maximal incubation time), zolmitriptan (one of the two donors ∼2-fold), and meloxicam (one of the two donors ∼2-fold). For suspension hepatocytes, half-life needed to be extrapolated beyond the incubation time for 6 of 10 drugs (∼2- to 3-fold).(1)Next, the in vitro drug half-lives were used to calculate the intrinsic clearance (CLint) using scaling factors (eq. 2), where 21 g of liver weight per 1 kg of body weight, and 120 × 106 hepatocytes per 1 g of human liver were used as standard parameters (Obach et al., 1997). PHH numbers per well were 4500 for MPCCs and 56,000 for suspension and conventional monolayers. All cultures had incubation volumes of 64 μl.(2)The hepatic clearance (CLh) was calculated from CLint using the well-stirred model (eq. 3), with liver blood flow (Q) being 21 ml/min per kilogram and protein binding (fu) set to 1 if no protein binding correction was made or to its respective known in vivo value (Table 1).(3)Error bars on graphs represent standard errors of the means. Microsoft Excel and GraphPad Prism 5.0 (La Jolla, CA) were used for data analysis, while GraphPad Prism 5.0 was used for plotting data.
Results
Long-Term Functional Characterization of MPCCs.
MPCCs were created in an industry-standard 96-well plate format using two cryopreserved PHH donors (lots: RTM and JNB). Prototypical hepatic morphology (polygonal shape, distinct nuclei and nucleoli, presence of bile canaliculi) was maintained for both donors in MPCCs for ∼4 weeks (Fig. 1A and Supplemental Fig. 1A). As observed previously (Khetani and Bhatia, 2008), albumin secretion in MPCCs took ∼7–9 days to reach higher steady-state levels than in the first few days of culture (Fig. 1B and Supplemental Fig. 1B). Urea secretion, on the other hand, either remained stable from the very beginning of the culture period or showed some downregulation initially followed by stabilization for the remainder of the time-series (Fig. 1C and Supplemental Fig. 1C). Although donor-dependent differences in albumin and urea secretion (as well as other functional markers as described below) were observed, overall trends were similar.
Activities of CYP1A2, 2C9, 2D6, and 3A4 were measured in both donors over time by quantifying metabolites of prototypical substrates, whereas the activities of CYP2A6, 2B6, 2C8, 2C19, and 2E1 were measured only in donor RTM (Supplemental Table 2) owing to limitations in number of vials available for donor JNB. While enzyme activities were detected for ∼4 weeks in both donors, there were differences in both the kinetics and magnitude of time course for specific P450 activities (Fig. 2 and Supplemental Fig. 1). CYP1A2 activities in RTM- and JNB-MPCCs at ∼4 weeks of culture were 90 and 72% of week 1 activities, respectively (Fig. 2A and Supplemental Fig. 1D). CYP2C9 activities in RTM- and JNB-MPCCs at ∼4 weeks were 176 and 91% of week 1 activities, respectively (Fig. 2C and Supplemental Fig. 1D). CYP2D6 activities in RTM- and JNB-MPCCs at ∼4 weeks were 73 and 85% of week 1 levels, respectively (Fig. 2D and Supplemental Fig. 1E). CYP3A4 activity in RTM at ∼4 weeks was 131% of week 1 activity, and activity in JNB at 4 weeks had gradually declined to 58% of week 1 levels (Fig. 2E and Supplemental Fig. 1E). By ∼4 weeks in RTM culture, CYP2A6 and CYP2B6 activities were downregulated to 24 and 47% of week 1 levels, respectively. However, most of the decline occurred after 19 days in culture (Figs. 2, A and B). CYP2C8 activity, on the other hand, was upregulated by day 9 of culture to 658% of week 1 levels and then remained fairly stable until day 30 (Fig. 2B). Use of CYP2C9-glo as a substrate showed similar relative stability as the measurement of 4-OH-tolbutamide in the same donor (Fig. 2C). CYP2C19 activity was only stable for 15 days followed by downregulation to 22–33% of week 1 levels for the remainder of the time-series (Fig. 2D). CYP2E1 activity at week 4 of culture declined to 12% of week 1 levels (Fig. 2E). Nonetheless, activities of all major P450s tested were detected out to at least 4 weeks in MPCCs created using both RTM and JNB donors. Lastly, we measured glucuronidation and sulfation (phase-II) activities in MPCCs created using both RTM and JNB donors (Fig. 2F and Supplemental Fig. 1F). By ∼4 weeks in culture, RTM maintained phase-II activities to 90–116% of week 1 activities, while JNB at 4 weeks maintained activities to 77–105% of week 1 activities.
Drug Clearance Predictions in MPCCs.
MPCCs were incubated for up to 7 days with 26 drugs listed in Table 1 (0.05–19.5 ml/min per kilogram in vivo clearance). Prototypical depletion of three drugs in MPCC supernatants is shown in Supplemental Fig. 2. Drug clearance from in vitro MPCC data (and other models below) was predicted using the drug half-life, scaling parameters, and well-stirred model as described in the Materials and Methods. Predicted drug clearance rates in MPCCs with or without incorporation of protein binding into the analysis are shown in Table 1. Clearance data in two donors (JNB and RTM) was acquired for all drugs except timolol, imipramine, and diclofenac, for which only a single donor was used owing to PHH sourcing limitations. On average, with protein-binding correction, MPCCs predicted 31, 58 and 69% of the drug clearance values within 2-fold, 3-fold, and 4-fold of in vivo clearance rates, respectively. When no protein-binding correction was incorporated, MPCCs predicted on average 62, 73, and 77% within 2-fold, 3-fold, and 4-fold of in vivo clearance rates, respectively. We found that for compounds with in vivo reported clearance values less than or equal to ∼1 ml/min per kilogram, correction for protein binding significantly improved MPCC predictive capacity, which is consistent with a previous study (Chan et al., 2013). When utilizing a “mixed analysis” approach in which correction for protein binding was only incorporated for compounds with reported clearance rates less than or equal to 1 ml/min per kilogram (low to very low turnover) and fu = 1 for all other compounds, MPCCs predicted 73, 92, and 96% of the drug clearance values within 2-fold, 3-fold, and 4-fold of in vivo clearance rates, respectively (Table 1 and Fig. 3). Additionally, MPCCs were able to correctly rank-order analog drugs on the basis of predicted clearance rates (sumatriptan/zolmitriptan, methylprednisolone/prednisolone, and lorazepam/diazepam). Lastly, the predicted drug clearance values in MPCCs created from two donors for drugs with clearance rates greater than 5 ml/min per kilogram (Fig. 4A) and less than 5 ml/min per kilogram (Fig. 4B) were compared against each other using linear regression analysis. Both donors provided similar predictions of drug clearance rates across all drugs (average R2 = 0.94, slope = 1.05) despite the functional differences observed in Figs. 1 and 2, and Supplemental Fig. 1.
Comparison of Predicted Drug Clearance Rates across Different Culture Models.
A subset of the drug set, 10 total drugs in particular across a wide range of in vivo turnover rates (0.19–19.5 ml/min per kilogram), were also tested in conventional PHH monolayers (Table 2) and suspension PHH cultures (Table 3) created from one of the PHH donors (RTM) used for MPCCs in Table 1. Prototypical depletion of three drugs in supernatants of conventional monolayers and suspension cultures is shown in Supplemental Figs. 3 and 4, respectively. Conventional monolayers were useful for predicting clearance rates for 9 of 10 compounds, except for naproxen, which did not metabolize sufficiently in the monolayers to make a clearance prediction. Without protein-binding correction, monolayers predicted 50, 60, and 80% and with protein-binding correction, 10, 20, and 20% within 2-fold, 3-fold, and 4-fold of in vivo clearance rates, respectively. When utilizing the mixed analysis approach as described above, monolayers predicted 40, 40, and 50% of the clearance rates within 2-fold, 3-fold, and 4-fold of in vivo clearance rates, respectively, which was in contrast to the data obtained in MPCCs (80, 100, and 100% within 2-fold, 3-fold, and 4-fold, respectively, for the 10-drug subset). Most of the clearance rates obtained from monolayers were 12–86% lower than those predicted using MPCCs, with the exception of erythromycin (3% higher in monolayers relative to MPCCs) and theophylline (26% higher). For suspension cultures, 70% of the drugs with in vivo clearance rates less than or equal to 6.1 ml/min per kilogram demonstrated little to no metabolism over the time course of 4 hours. For the three compounds that demonstrated metabolism in suspension cultures (verapamil, naloxone, and timolol), clearance values were predicted within 3-fold of in vivo clearance values. Overall, 20, 30, and 30% of the drug clearances were predicted within 2-fold, 3-fold, and 4-fold of in vivo clearance rates, respectively, in suspension PHHs.
Effects of DDIs on Drug Clearance Rates.
CYP3A4 activity in MPCCs was induced ∼4-fold relative to vehicle control by a 3-day treatment with rifampin, or CYP3A4 was inhibited down to ∼4% relative to vehicle controls by an 18-hour treatment with ritonavir (data not shown). Inducing CYP3A4 levels by ∼4-fold led to significantly more turnover of midazolam (56.4% depletion in vehicle control versus 99.8% depletion in rifampin-treated cultures) over 24 hours of incubation (Fig. 5A). On the other hand, incubation with ritonavir significantly inhibited midazolam turnover (8.2% depletion) relative to vehicle controls. When the turnover of midazolam over time was converted to predicted clearance rates, the vehicle control produced a rate within 2-fold of in vivo clearance (vehicle control: 10.7 ml/min per kilogram; in vivo: 8.7 ml/min per kilogram), whereas rates in both induced and inhibited cultures were outside the 2-fold window (induced: 18.5 ml/min per kilogram; inhibited: 2.3 ml/min per kilogram) owing to the DDI (Fig. 5B).
When we inhibited CYP2D6 in MPCCs using quinidine, turnover of dextromethorphan over a 96-hour incubation was significantly inhibited relative to vehicle control cultures (85% turnover in vehicle controls versus 13.5% turnover in inhibited cultures) (Fig. 6A). Predicted dextromethorphan clearance from the turnover data were within 2-fold of in vivo clearance in vehicle control cultures (vehicle control: 7.6 ml/min per kilogram; in vivo: 8.6 ml/min per kilogram), but outside 2-fold in inhibited cultures (2.2 ml/min per kilogram) (Fig. 6B). On the other hand, effects of quinidine-mediated CYP2D6 inhibition were not as pronounced for desipramine turnover as for dextromethorphan (Fig. 6, C and D). In particular, 68% of desipramine was depleted over 48 hours in vehicle control cultures as opposed to 54% in inhibited cultures (Fig. 6C). Such turnover corresponded to predicted desipramine clearance rates within 2-fold of in vivo clearance for both vehicle control and quinidine-inhibited cultures (vehicle control: 8.5 ml/min per kilogram; quinidine-inhibited: 6.6 ml/min per kilogram; in vivo: 10.3 ml/min per kilogram) (Fig. 6D).
Discussion
Prediction of in vivo human drug clearance using in vitro hepatic clearance data can help identify compounds with poor pharmacokinetic characteristics. An ideal hepatocyte culture platform for such purposes uses as few as possible limited PHHs in a reproducible/miniaturized format, maintains high levels of drug metabolism enzymes with proper hepatocyte polarity to allow incubations with drugs that interact with multiple pathways, is compatible with multiple cryopreserved PHH donors for on-demand screening, and can be used to predict clearance of compounds with a wide range of turnover rates, including slowly metabolized compounds. Additionally, the ability to interrogate effects of drug incubations on PHH enzyme levels and subsequently victim drug disposition is important for modeling clinical DDIs (Khetani et al., 2015). Toward approximating such features, we show that MPCCs created in a 96-well plate format display high levels of P450 and phase-II activities for ∼4 weeks. Such activities coupled with the ability to dose drugs for 7 days without a medium change led to better overall prediction of drug clearance rates than with suspension cultures or conventional monolayers created from the same donor. Finally, modulating P450 activities via perpetrator drugs altered the clearance of victim drugs in MPCCs.
We quantified albumin secretion, urea synthesis, and enzyme activities over several weeks in MPCCs created from two cryopreserved PHH donors. The fibroblasts used in MPCCs do not metabolize drugs that primarily undergo hepatic metabolism (Khetani and Bhatia, 2008; Chan et al., 2013). Albumin and urea secretion rates in MPCCs were relatively stable across both donors for 3–4 weeks. CYP1A2, 2C9, 2D6, glucuronidation/sulfation activities were relatively stable between 1 and 4 weeks of culture across both donors; however, CYP3A4 declined by ∼40% in one donor. We also measured the activities of CYP2A6, 2B6, 2C8, 2C19, and 2E1 in a single donor. CYP2A6 and 2B6 were relatively stable for ∼3 weeks; 2C19 activity was relatively stable for ∼2 weeks, and CYP2E1 levels showed a decline between 1 and 4 weeks of culture. Nonetheless, activities of all major enzymes tested were detected for ∼4 weeks in MPCCs with good stability for 2–3 weeks for most P450s. Culture medium formulations that can improve the stability of all major P450s in MPCCs for at least 4 weeks should prove useful for drug dosing at later time points.
The 26 drugs chosen span a larger range of in vivo turnover rates (0.05–19.5 ml/min per kilogram) than previously tested in MPCCs (Chan et al., 2013). These drugs can undergo metabolism via P450s (i.e., tolbutamide, diclofenac) and phase-II enzymes (naloxone, diazepam), and some are transporter substrates (i.e., desipramine, furosemide) (McGinnity et al., 2004; Koepsell et al., 2007). The well-stirred model was used to predict clearance from drug depletion in supernatants, an approach well suited for screening large numbers of compounds. Marginal differences have been observed across well-stirred, parallel-tube, and dispersion models for clearance prediction, except for very high turnover compounds (Hallifax et al., 2010). However, the well-stirred model produced good predictions for the high-turnover compounds here. On average across two donors and when using a “mixed analysis” approach (i.e. correction for protein binding was only incorporated for compounds with reported clearance rates less than or equal to 1 ml/min per kilogram), MPCCs predicted 19 of 26 compounds (73%) within 2-fold of the known in vivo clearance rates, 24 of 26 compounds (92%) within 3-fold, and 25 of 26 compounds (96%) within 4-fold, with meloxicam’s predicted clearance rate at a 6-fold deviation. In another study with MPCCs and in another engineered liver platform, meloxicam was under-predicted and its turnover was highly dependent on the donor (Dash et al., 2009; Chan et al., 2013).
The accuracy of predicted clearance rates for high- and medium-turnover compounds was improved when plasma protein binding was not incorporated, as also observed previously (Hallifax et al., 2010; Ring et al., 2011). On the other hand, MPCCs metabolized low-turnover drugs (less than or equal to 1 ml/min per kilogram) significantly faster in the serum-free medium than in vivo, and thus use of reported fu values significantly improved the accuracy of clearance predictions, as also observed previously (Blanchard et al., 2005; Smith et al., 2012). Although the mechanism is not known, others have speculated that since slowly metabolized compounds have more time to bind to proteins in vivo than higher turnover compounds, the unbound fraction available for metabolism for slowly metabolized compounds may be lower than for higher turnover compounds (Atkinson and Kushner, 1979). Therefore, incorporation of protein binding correction in the analysis for low-turnover compounds becomes important for more accurate clearance predictions. It is possible that inclusion of human albumin, alpha-1-acid glycoprotein, and lipoproteins in culture medium at concentrations found in human blood may allow a more consistent analysis scheme for the entire range of drug-turnover rates (Chao et al., 2009).
The predicted drug clearance rates across two PHH donors used in MPCCs were strongly correlated (average R2 = 0.94, slope = 1.05 for all the compounds) despite differences in P450 activities. Furthermore, in comparison with another study that tested 7 of 26 compounds used here with the same two donors in MPCCs (Chan et al., 2013), we found good agreement with the rank ordering of drugs by their predicted clearance rates, thereby showing the reproducibility of the platform. With their use of individual PHH donors and ∼40–50% fewer cells for seeding compared with suspension and conventional confluent cultures, MPCCs can be used with limited donor lots for a larger number (∼2- to 3-fold) of screening studies. Pooled plateable PHH lots could provide an “average” human response in plated culture formats; however, plating efficiencies need to be uniform across the various PHH donor lots to ensure that monolayers are composed of similar numbers of each donor’s PHHs. Nonetheless, individual PHH lots with specific polymorphisms can provide useful information on population-specific differences in drug clearance.
Suspension PHHs, created from the same donor as that used in MPCCs, did not sufficiently deplete medium- and low-turnover drugs (7 of 10) within 4 hours to allow prediction of clearance rates, whereas clearance rates of higher turnover drugs were predicted within 2- to 3-fold of in vivo levels. Thus, although single donors can be used in MPCCs, pooled lots of 10 or more carefully selected donors, along with the relay method, are necessary for applying suspension PHHs to the prediction of medium- and low-turnover compounds (Di et al., 2012). On the other hand, conventional PHH monolayers predicted the clearance rates for 9 of 10 drugs. Even though conventional monolayers display a rapid decline in functionality within the first 4–24 hours (Khetani et al., 2015), continued metabolism of some low-turnover compounds (i.e., diazepam) was observed over 4 days. Naproxen, however, did not turn over in conventional monolayers even after 4 days of incubation even though it was depleted within 3 days in MPCCs, which contained ∼10-fold fewer attached PHHs. Overall, when using the “mixed analysis” approach, conventional monolayers predicted 40, 40, and 50% of the compounds within 2-, 3-, and 4-fold of in vivo turnover rates, respectively, whereas MPCCs predicted 80 and 100% within 2- and 3-fold, respectively. Furthermore, conventional monolayers predicted lower clearance rates than in MPCCs for 8 of 10 drugs, probably because of the lower enzyme activity per cell in conventional monolayers (Khetani and Bhatia, 2008).
The short lifetime (<7 days) of conventional PHH monolayers coupled with a significant decline in P450 activities limits their utility in evaluating effects of P450 modulation on drug disposition, especially for those P450s (i.e., 2D6) that are not as abundant as 3A4 (Khetani et al., 2015). Here, we hypothesized that the greater longevity and higher functionality of MPCCs could help mitigate such a limitation. Inducing CYP3A4 in MPCCs via rifampin for 3 days led to an ∼73% increase in subsequent midazolam clearance, while inhibiting CYP3A4 via ritonavir for 18 hours led to a ∼79% decrease in midazolam clearance relative to vehicle controls. The ∼1.7-fold increase in midazolam clearance in MPCCs with rifampin pretreatment is in line with the ∼2-fold increase observed in the clinic, albeit live patients were pretreated with rifampin for 7 days (Gorski et al., 2003). Furthermore, our use of serum with inducers in MPCC culture medium coupled with higher baseline P450 activities typically leads to lower fold induction values (2- to 8-fold) than can be observed with declining conventional monolayers incubated with inducers in serum-free medium (up to 80- to100-fold) (Rae et al., 2001; Hariparsad et al., 2004; Williamson et al., 2013). Inhibition of CYP2D6 via quinidine led to ∼71 and ∼22% reduction in clearance of dextromethorphan and desipramine, respectively, relative to controls. Such a difference across the compounds highlights the need for evaluating the effects of DDI on drug clearance in vitro.
Incorporation of liver stromal cells in MPCCs may allow prediction of drug clearance rates under disease states such as fibrosis and inflammation (Nguyen et al., 2015). Furthermore, miniaturization into a 384-well format should enable higher throughput screening in MPCCs. In conclusion, we show that MPCCs with unpooled cryopreserved PHHs can predict the clearance rates of drugs with a wide range of in vivo turnover rates, including slowly metabolized drugs. The accuracy of drug clearance prediction in MPCCs was significantly better than that observed in suspension and conventional monolayers created from the same donor. Furthermore, the longevity of MPCCs allowed evaluation of the effects of DDI on drug clearance, which should prove useful for better modeling of clinical scenarios.
Acknowledgments
We are grateful to Dustin Berger and Brenton Ware for assistance with cell culture.
Authorship Contributions
Participated in research design: Lin, Shi, Moore, Khetani.
Conducted experiments: Lin, Shi, Moore.
Performed data analysis: Lin, Khetani.
Wrote or contributed to the writing of the manuscript: Lin, Khetani.
Footnotes
- Received June 18, 2015.
- Accepted October 8, 2015.
Funding was provided by the National Science Foundation (CAREER Award CBET 1351909 to S.R.K.) and Colorado State University. S.R.K. is an equity holder in Hepregen Corporation, which has exclusively licensed the MPCC technology from Massachusetts Institute of Technology for drug development applications.
↵This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- DDI
- drug-drug interaction
- P450
- cytochrome P450
- PHH
- primary human hepatocytes
- Copyright © 2015 by The American Society for Pharmacology and Experimental Therapeutics