Prediction of Human Hepatic Clearance for Cytochrome P450 Substrates via a New Culture Method Using the Collagen Vitrigel Membrane Chamber and Fresh Hepatocytes Isolated from Liver Humanized Mice

Abstract

In drug discovery, hepatocytes have been widely utilized as in vitro tools for predicting the in vivo hepatic clearance (CL) of drug candidates. However, conventional hepatocyte models do not always reproduce in vivo physiological function, and CYP activities in particular decrease quite rapidly during culture. Furthermore, conventional in vitro assays have limitations in their ability to predict hepatic CL of metabolically stable drug candidates. In order to accurately predict hepatic CL of candidate drugs, a new method of culturing hepatocytes that activates their functional properties, including CYP activities, is in high demand. In the previous study, we established a novel long-term culture method for PXB-cells^® using a collagen vitrigel membrane (CVM) chamber, which can maintain CYP activity and liver specific functions at high levels for several weeks. In this study, the vitrigel culture method was applied to predictions of hepatic CL for 22 CYP typical substrates with low to middle CL, and the prediction accuracy by this method was assessed by comparing CL data between predicted (in vitro intrinsic CL using the dispersion model) and observed (in vivo clinical data) values. The results of this study showed that in vitro CL values for approximately 60% (13/22) and 80% (18/22) of the compounds were predicted within a 2- and 3-fold difference with in vivo CL, respectively. These results suggest that the new culture method using the CVM chamber and PXB-cells is a promising in vitro system for predicting human hepatic CL with high accuracy for CYP substrates, including metabolically stable drug candidates.

INTRODUCTION

Improving accuracy in the prediction of human pharmacokinetics (PK) is indispensable in drug development, and it is essential to determine the dynamic characteristics of drug candidates before clinical introduction. In recent years, though fewer drugs have been withdrawn from the market due to pharmacokinetic-related problems,^1,2) it is still important to appropriately predict human PK and side effects when estimating drug efficacy and safety before clinical introduction. In order to predict human PK, clearance (CL) and distribution volumes are key parameters that must be accurately estimated. For CL prediction, a method using a human biological sample—in particular, human hepatocytes—is considered the gold standard. However, when hepatocytes are used in suspension assay, metabolic activity or hepatic function are drastically reduced within several hours. Further, the number of slowly metabolized drug candidates, which are difficult to evaluate in conventional assay systems, are increasing in the field of drug discovery, thus, evaluation methods for low CL drug candidates such as the Relay method³⁾ and co-culture system^4,5) have been developed. In addition, human PK prediction using chimeric mice has attracted attention, and their use in drug discovery has increased.^6,7) However, in chimeric mice, since only a small number of mouse hepatocytes remain, the human effectiveness of some drug candidates which can be highly metabolized by mice tends to be overestimated. Therefore, it is important to recalibrate the contribution of mouse hepatocytes in human PK prediction when chimeric mice are used. Additionally, PXB-cells are fresh chimeric hepatocytes isolated from liver humanized mice (PXB-mice), and these have been used as an alternative tool comparable to human cryopreserved hepatocytes.^8–11) Although such PXB cells generally come from a single donor, PXB-cells can be stably supplied, allowing for reproducible cell assays using the same donor. On the other hand, PXB-cells have been reported to decrease the mRNA expression level of mouse CYP to 1/1000 or less upon culturing for 1 week or more after cell seeding.¹²⁾ It is attractive that the contribution of mice would be minimized by culturing PXB-cells long term, and that PXB-cells may be used as an alternative tool for cryopreserved human hepatocytes in terms of the stable supply of cells and at a low cost. In a previous study, we established the vitrigel culture method that can improve both metabolic activity and hepatic function.¹³⁾ Therefore, in this study, we have focused on the possibility of constructing a highly accurate evaluation system for human prediction using a collagen vitrigel membrane (CVM) chamber with fresh hepatocytes isolated from chimeric mice PXB-cells. For this purpose, we calculated the CL_int per the disappearance of unchanged forms using 22 typical CYP substrates in this new culture method, compared it with in vivo human CL, and evaluated the accuracy of the in vitro in vivo correlation (IVIVC).

MATERIALS AND METHODS

Chemicals and Reagents

Alprazolam was purchased from Toronto Research Chemicals (North York, ON, Canada). Atomoxetine and risperidone were purchased from AK Scientific (Union City, CA, U.S.A.). Clozapine, desipramine, diazepam, diclofenac, flecainide, lidocaine, metoprolol, mexiletine, omeprazole, propranolol, quinidine, tolbutamide, zolpidem, fetal bovine serum (FBS), Hanks’ balanced salt solution (HBSS), Williams’ medium E, and dimethyl sulfoxide (DMSO) were obtained from Sigma-Aldrich (St. Louis, MO, U.S.A.). Glimepiride, midazolam, prednisolone, and riluzole were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). Sildenafil was purchased from LKT Laboratories (St. Paul, MN, U.S.A.). Voriconazole was purchased from Tokyo Chemical Industry (Tokyo, Japan). (S)-(−)-Warfarin was purchased from Cayman Chemical (Ann Arbor, MI, U.S.A.). Dulbecco’s modified Eagle’s medium (DMEM), 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), and antibiotics (penicillin and streptomycin) were purchased from Life Technologies Corp. (Grand Island, NY, U.S.A.). All other materials and chemicals not specified above were of the highest grade.

Preparation of Collagen Vitrigel Membrane Chambers

A CVM chamber is currently commercially available under the product name ad-MED Vitrigel^®. ad-MED Vitrigel^® was purchased from Kanto Chemical Co., Inc. (Tokyo, Japan). The CVM chambers were prepared according to the previous protocol.¹³⁾

Ethics Statement

All in vivo experiments and protocols for animal studies were approved by the Laboratory Animal Ethics Committee at PhoenixBio Co., Ltd. (Hiroshima, Japan).

Preparation and Culture of PXB-Cells

Cryopreserved human hepatocytes (Hispanic, 2-year old, female: Lot. BD195) were purchased from BD Bioscience (San Jose, CA, U.S.A.). The hepatocytes were thawed and transplanted into cDNA-urokinase-type plasminogen activator-transgenic/severely combined immunodeficient mice to prepare PXB-mice^® as described previously.¹⁴⁾ The PXB-cells were isolated from PXB-mice^® with a blood human albumin level higher than 12 mg/mL by a two-step collagenase perfusion method.¹¹⁾ The culture of PXB-cells in CVM chambers was conducted in accordance with the vitrigel culture procedure⁹⁾ for 14 d at 37°C in a humidified atmosphere of 5% CO₂ and 95% air. The media both inside and outside the chamber were changed on Days 1, 3, 7, and 10, replaced with fresh culture medium (dHCGM medium: DMEM supplemented with 10% FBS, 20 mmol/L HEPES, 15 µg/mL L-proline, 0.25 µg/mL insulin, 50 nmol/L dexamethasone, 44 mmol/L NaHCO₃, 5 ng/mL EGF, 0.1 mmol/L ascorbic acid 2-phosphate, 100 IU/mL penicillin G, 100 µg/mL streptomycin, and 2% DMSO).

Metabolic Assay on the CVM

Test compounds (22 typical CYP substrates) were individually added to triplicate wells at a final concentration of 1 µmol/L, except for (S)-(−)-warfarin (0.1 µmol/L). The standard solution for each substrate was diluted in Williams’ Medium E to prepare the assay medium, which was then preheated to 37°C before the initiation of incubation. In the vitrigel culture, prior to assay, each vitrigel membrane was inserted in an option ring (Kanto Chemical), and the dHCGM medium was removed from the PXB-cells, followed by washing with 400 µL of preheated PBS. Then, PBS was removed from the PXB-cells, and a second washing was conducted with 125 µL of Williams’ Medium E preheated at 37°C. After the removal of Williams’ Medium E from the PXB-cells, 125 µL of Williams’ Medium E preheated at 37°C was added to the PXB-cells, and a metabolic assay was initiated by adding 125 µL of assay medium to each vitrigel chamber (n = 3/substrate). Assay time points (0, 5, 15, 30, 60, 120, 240, 1440 min) were appropriately selected for each substrate, according to their disappearance, to obtain CL_int (5 time points/substrate). All cultures and metabolic assays were conducted at 37°C in a humidified atmosphere containing 95% air and 5% CO₂. Samples (10 µL) were taken from each vitrigel chamber sequentially (n = 1/well) and kept in the freezer (−20°C) until analysis by liquid chromatography with tandem mass spectrometry (LC-MS/MS).

LC-MS/MS Analysis

For analyses, the frozen samples were thawed at room temperature, and 200 µL of acetonitrile–methanol (7 : 3 (v/v)) containing internal standard (IS) was added. The samples were mixed with a vortex and centrifuged at 2095 × g for 10 min, and the supernatants were then analyzed by LC-MS/MS. Propranolol was used as an IS in LC-MS/MS analyses for all assays. Ultra performance liquid chromatography (UPLC) was performed using the Waters ACQUITY liquid handling system (Waters, Milford, CT, U.S.A.). For the UPLC system, the mobile phase was a gradient with 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B) at a flow rate of 0.3 mL/min. The initial composition of the mobile phase was 10% of B for 0.5 min, followed by a linear gradient to 99% of B over 2.5 min, where it was maintained at 99% of B for 1.0 min, then back to 10% B in 0.1 min, after which it was allowed to re-equilibrate to the initial conditions. The total run time was 6.0 min. Separation was carried out using an Atlantis T3 column (2.1 × 50 mm) (Nihon Waters, Tokyo, Japan) or an L-Column ODS (2.1 × 150 mm) (Chemicals Evaluation and Research Institute, Tokyo, Japan), and eluted fractions were directly passed through a Waters Micromass Quattro Premier Mass Spectrometer (Waters). The ionization mode, precursor ion, product ion, cone voltage (CV), collision energy (CE), and retention time (RT) are described in Table 1.

Table 1. LC-MS/MS Parameters of the 22 Substrates and IS Substance

Compound name	Responsible metabolic enzymes	LC-MS/MS parameters
		Ionization mode	Precursor ion (m/z)	Product ion (m/z)	CV (V)	CE (eV)	RT* (min)	IS RT* (min)
Clozapine	CYP1A2	+	327.38	191.78	60	40	1.93a)	2.01a)
Lidocaine	CYP1A2	+	235.22	85.76	50	15	1.71a)	2.00a)
Mexiletine	CYP1A2	+	179.98	57.73	30	10	2.83a)	2.95a)
Riluzole	CYP1A2	+	234.97	137.61	50	35	3.62b)	2.80b)
Diclofenac	CYP2C9	+	295.94	213.72	20	30	4.16b)	2.82b)
Glimepiride	CYP2C9	+	491.21	125.63	40	30	4.13b)	2.81b)
Tolbutamide	CYP2C9	+	271.11	90.56	30	30	2.74a)	2.04a)
(S)-(−)-Warfarin	CYP2C9	+	309.13	162.71	30	15	2.96a)	2.01a)
Diazepam	CYP2C19	+	285.09	192.76	40	30	2.91a)	2.04a)
Omeprazole	CYP2C19	+	346.15	197.79	20	15	1.93a)	2.01a)
Voriconazole	CYP2C19	+	350.12	126.50	30	35	3.63b)	2.78b)
Atomoxetine	CYP2D6	+	256.21	147.29	20	10	2.90b)	2.81b)
Desipramine	CYP2D6	+	267.47	71.71	20	15	2.17a)	2.01a)
Flecainide	CYP2D6	+	414.90	300.74	40	40	2.88b)	2.81b)
Metoprolol	CYP2D6	+	268.23	115.68	40	20	1.83a)	2.05a)
Risperidone	CYP2D6	+	411.02	190.54	50	35	1.88a)	2.01a)
Alprazolam	CYP3A4	+	309.16	204.74	60	40	2.55a)	2.01a)
Midazolam	CYP3A4	+	326.04	290.89	45	30	2.07a)	2.04a)
Prednisolone	CYP3A4	+	361.29	146.74	30	20	2.30a)	2.01a)
Quinidine	CYP3A4	+	325.25	171.61	50	40	1.65a)	2.05a)
Sildenafil	CYP3A4	+	475.36	99.20	30	25	2.81b)	2.79b)
Zolpidem	CYP3A4	+	308.15	234.88	60	50	1.90a)	2.01a)
Propranolol (IS)	—	+	260.11	115.73	30	20	—	—

CV: cone voltage, CE: collision energy, IS: internal standard, LC-MS/MS: liquid chromatography with tandem mass spectrometry, RT: retention time. a) Atlantis T3 column was used. b) L-Column ODS was used. * Each RT was described relative to the value of a typical RT.

Calculations of CL_int

Calculation of CL_int was based on the substrate disappearance rate. The remaining percentage of the substrate during the incubation was calculated by dividing the mass spectrometry peak area ratio at a certain incubation time by that at time zero. The slope of the remaining percentage was calculated for each data set, and the half-life (T_1/2) and CL_int were calculated using the slope, number of hepatocytes (2.1 × 10⁵ cells) seeded to each well, and the incubation volume of 250 µL (using Eqs. 1 and 2, below).

• 1) T_1/2 (min) = (ln 2)/ − slope
• 2) In vitro CL_int (µL/min/million cells) = ln 2 × 1000/[T_1/2 × cell conc. (million cells/mL)]

As a quality criterion to effectively distinguish the CL_int value from no measurable turnover, a comparison of fits was conducted using GraphPad Prism 7.0 (GraphPad Software, Inc., Suite, U.S.A.). Hepatic clearance (CL_h) was calculated from CL_int using the dispersion model (Eqs. 3 and 4). The physiological human scaling factors were 120 million cells/g liver, 25.7 g liver/kg body weight and 21 mL/min/kg of liver blood flow (Q_h). Fraction unbound in plasma (f_u,p) and the blood to plasma ratio (R_b) were obtained from the literature,^4,15–18) and each fraction unbound in blood (f_b) was calculated by dividing the value of f_u,p by that of R_b. The protein binding rate in metabolic culture medium or adherent cells (f_u,inc) was assumed to be 1.

• 3) Scaled CL_int = in vitro CL_int × (120 × 10⁶ (cells/g liver) × 25.7 (g liver/kg) × f_b)/f_u,inc
• 4) In vitro predicted CL_h = Q_h × [1 − (4a/((1 + a)² × exp[(a − 1)/2D_N] − (1 − a)² × exp[−(a + 1)/2D_N]))] where the dispersion number (D_N) = 0.17,¹⁹⁾ a = (1 + 4 × R_N × D_N)^0.5, and R_N = f_b × scaled CL_int/Q_h

In vivo human nonrenal plasma clearance (in vivo CL) for the 22 substrates was calculated by human plasma CL and renal excretion (Eq. 5).

• 5) In vivo observed CL = human plasma CL − F_renal where F_renal is the fraction of drug excreted by renal elimination

Then, the in vitro predicted CL_h was compared with in vivo observed CL, and the accuracy of each prediction was assessed using the average fold error (AFE), the absolute average fold error (AAFE), and the root mean squared error (RMSE). AFE, AAFE, and RMSE were calculated as follows:

• 6) AFE = 10^{[(∑ log(in vitro predicted CL_h/in vivo observed CL))/n]}
• 7) AAFE = 10^{[(∑ ∣log(in vitro predicted CL_h/in vivo observed CL)∣)/n]}
• 8) RMSE = [(∑ (in vitro predicted CL_h−in vivo observed CL)²)/n]^0.5

RESULTS

Vitrigel Culture Method to Estimate Hepatic Clearance for CYP Substrates

Previous studies have shown that the activity of major CYP isoforms was increased by the vitrigel culture method and peaked on Day 14 of culture,¹³⁾ and therefore, the metabolic activity of each CYP substrate was assessed on Day 14 of culture in this study. Twenty-two compounds were selected as typical substrates for 5 major CYP isoforms as the test compounds (Table 2). In order to estimate the in vitro CL_int, the elimination rate constant was calculated from the disappearance of the unchanged form for each substrate. Then, the predicted in vitro CL_h was calculated by a dispersion model using f_u,p and R_b. For the sake of convenience, the number of cells after culturing for 14 d was adapted to the initial number of seeds and used for the calculation of in vitro CL_int. Then, the predicted in vitro CL_h values were plotted against the observed in vivo CL values (Fig. 1).

Table 2. In Vitro and in Vivo CL Data of the 22 Substrates

Compound name	Responsible metabolic enzymes	fu,p	Rb	Human CL (mL/min/kg)	Frenal	In vivo observed CL (mL/min/kg)	In vitro CLint (µL/min/million cells)	In vitro predicted CLh (mL/min/kg)	Ratio*
Clozapine	CYP1A2	0.044b)	0.70b)	6.1b)	0.005c,g)	6.07	23.2	2.77	0.46
Lidocaine	CYP1A2	0.3a)	0.84a)	9.2a)	0.02a)	9.02	17.4	9.86	1.1
Mexiletine	CYP1A2	0.36d)	1f)	8.3d)	0.095c,g)	7.51	14.8	10.7	1.4
Riluzole	CYP1A2	0.02a)	1.7a)	3.52a)	<0.01a)	3.48	56.2	3.32	0.95
Diclofenac	CYP2C9	0.005a)	0.55a)	4.22a)	<0.01a)	4.18	22.6	0.339	0.081
Glimepiride	CYP2C9	0.005a)	0.55a)	0.62a)	<0.005a)	0.617	7.88	0.120	0.19
Tolbutamide	CYP2C9	0.058a)	0.55a)	0.17a)	<0.01a)	0.168	0.580	0.103	0.61
(S)-(−)-Warfarin	CYP2C9	0.01a)	0.55a)	0.045a)	<0.02a)	0.0441	0.638	0.0195	0.44
Diazepam	CYP2C19	0.015a)	0.71a)	0.38a)	<0.01a)	0.376	17.8	0.791	2.1
Omeprazole	CYP2C19	0.05d)	0.59e)	8.4d)	0c,g)	8.40	25.5	3.26	0.39
Voriconazole	CYP2C19	0.42a)	1f)	3.80a)	<0.02a)	3.72	22.4	14.6	3.9
Atomoxetine	CYP2D6	0.013a)	0.55a)	3.92a)	0.01a)	3.88	59.1	2.12	0.55
Desipramine	CYP2D6	0.184b)	0.89b)	7.0b)	0.02c)	6.86	19.8	8.00	1.2
Flecainide	CYP2D6	0.39a)	0.89a)	9.1a)	0.43a)	5.19	12.5	9.67	1.9
Metoprolol	CYP2D6	0.88b)	1.2b)	12.1b)	0.10c,g)	10.9	4.62	9.40	0.86
Risperidone	CYP2D6	0.11a)	0.67a)	3.56a)	0.03a)	3.45	14.7	4.03	1.2
Alprazolam	CYP3A4	0.29a)	0.78a)	0.59a)	0.20a)	0.472	4.38	3.39	7.2
Midazolam	CYP3A4	0.02a)	0.69a)	4.6a)	0.01a)	4.55	38.3	2.14	0.47
Prednisolone	CYP3A4	0.28a)	1f)	1.73a)	0.17a)	1.44	1.58	1.31	0.91
Quinidine	CYP3A4	0.25b)	0.92b)	4.7b)	0.18c,g)	3.85	8.15	5.18	1.3
Sildenafil	CYP3A4	0.0941b)	1f)	6.0b)	0.075c,g)	5.55	36.7	7.94	1.4
Zolpidem	CYP3A4	0.13b)	0.76b)	4.3b)	0.005c,g)	4.28	10.8	3.68	0.86

f_u,p: fraction unbound in plasma, R_b: blood to plasma ratio, CL: clearance, CL_int: intrinsic clearance, CL_h: hepatic clearance, F_renal: the fraction of drug excreted by renal elimination. a) Chan et al.⁴⁾ b) Stringer et al.¹⁵⁾ c) Dave et al.¹⁶⁾ d) Obach et al.¹⁷⁾ e) Uchimura et al.¹⁸⁾ f) R_b was assumed as 1. g) These values were obtained by dividing the reference value (% Urinary Excretion) by one hundred. * Ratio = in vitro Predicted CL_h/in vivo Observed CL.

Fig. 1. In Vitro in Vivo Correlation from the Vitrigel Culture Method Using the CVM Chambers and PXB-Cells

The graph shows a log plot of the calculated hepatic clearance values from the vitrigel culture method compared to the observed in vivo values from the clinical data. The solid line represents a perfect correlation. The dashed lines represent boundaries between a 2-fold underprediction and a 2-fold overprediction. The dotted lines represent boundaries between a 3-fold underprediction and a 3-fold overprediction. (Color figure can be accessed in the online version.)

Of the 22 compounds, prediction accuracy within 2 and 3 times were 59% and 82%, respectively. In addition, the AFE was 0.878, AAFE was 1.99, and RMSE was 3.24. On the other hand, the substrates that were overestimated or underestimated more than 5 times were diclofenac and glimepiride of the CYP2C9 substrate and alprazolam of the CYP3A4 substrate.

DISCUSSION

This study aims to improve the prediction accuracy of human hepatic CL for drug candidates by the vitrigel culture method using PXB-cells. As test compounds, 22 typical substrates of 5 major CYP isoforms with low to medium CL were selected. The vitrigel culture method using the CVM chambers and PXB-cells showed that in vitro CL values for approximate 60% (13/22) and 80% (18/22) of compounds were predicted within a 2-fold and 3-fold difference from the observed in vivo CL, respectively. In particular, among 5 major CYP isoforms, high prediction accuracy was observed in the compounds CYP1A2, CYP2D6, and CYP3A4. In addition, good prediction accuracy was obtained in (S)-(−)-warfarin and tolbutamide, which are known to be slowly metabolized compounds. These results suggest that the vitrigel culture method would be useful for CL prediction for drug candidates that are mainly metabolized by CYP isoforms, even in the case of metabolically slow (low CL) compounds. On the other hand, there are three compounds which were overestimated or underestimated 5-fold or more: alprazolam, diclofenac, and glimepiride.

Diclofenac and glimepiride would bind to PXB-cells, since both compounds showed very high plasma protein binding (f_u,p = 0.005), and the concentration of unbound drug became low, resulting in underestimation by the vitrigel culture method. Although their nonspecific bindings to CVM or PXB-cells were not evaluated in this study, the same phenomenon has been confirmed by the HepatoPac method.⁴⁾ Additionally, there is a possibility of the involvement of other CYP and/or non-CYP enzymes, even when the test substrates are selective for one CYP. For example, diclofenac is metabolized by CYP2C9 and uridine 5′-diphospho-glucuronosyltransferase (UGT),²⁰⁾ resulting in a large in vivo and in vitro variation. Regarding alprazolam, a CYP3A substrate, CL was overestimated 7.2 times in the vitrigel culture method. It was reported that alprazolam caused enterohepatic circulation in rodents.²¹⁾ Although there was no parallel report of enterohepatic circulation in humans, the actual CL_h in human may be high if enterohepatic circulation occurred in humans, which would be a possible reason for overestimation by the vitrigel culture method. Nonspecific binding and enterohepatic circulation are general problems for CL prediction in all assay systems, not specific to the vitrigel culture method, and therefore, these points should be always considered when CL for drug candidates is predicted in drug discovery.

In this study, PXB-cells were selected as an alternative to cryopreserved human hepatocytes. The reason we did not select cryopreserved human hepatocytes was that cryopreserved human hepatocytes vary largely in lot difference, so the same lot cannot be stably supplied. In contrast, PXB-cells can be stably supplied for many years and have high reproducibility (data not shown). It is extremely advantageous to be able to produce stable results from the same lot in order to select better/best final drug candidate from among many candidates, with high prediction accuracy and assessment at any time. There are two major concerns regarding PXB-cells: the single donor factor, and the contamination of mouse hepatocytes. Since PXB-cells are prepared from single donor, it is difficult to cover various biochemical variations that naturally derive from human diversity. Therefore, we are planning to investigate whether the vitrigel culture method can give us the appropriate data with sufficiently high prediction accuracy, even in cryopreserved human hepatocytes. When good data are obtained, PXB-cells in vitrigel culture will be used for selecting a final candidate from among many candidates, and cryopreserved human hepatocytes in vitrigel culture will later be utilized to examine the variability of human PK to determine a final candidate, which would be an adequate system to predict human CL and its variation with high accuracy in the process of drug discovery. Regarding the concern about the contamination of mouse hepatocytes, PXB-cells have been reported to decrease the mRNA expression level of mouse CYP to 1/1000 or less upon culturing for 1 week or more after cell seeding.¹²⁾ Since our vitrigel culture method employs a 2-week culture period to maximize CYP activity, the influence of mouse hepatocytes may be as small as possible. In fact, the prediction accuracy of diazepam was improved within a 3-fold difference in this study, even though diazepam is easily metabolized by mouse hepatocytes, supporting our contention that PXB-cells cultured using CVM chambers for 2 weeks would minimize or diminish the contribution of mouse hepatocytes.

In recent years, in evaluating systems considered to have high predictability of low CL compounds, the Relay,³⁾ HepatoPac⁴⁾ and HµREL⁵⁾ methods have been developed. With respect to the Relay method, it is an attempt to solve the problem of activity reduction in conventional hepatocyte assays by replacing cryopreserved human hepatocytes every 4 h, and it is claimed that prediction accuracy using this method is higher than conventional methods. Evaluation methods that use highly active hepatocytes, one after another in a short time, are extremely advantageous. However, it is a laborious and expensive method as it requires the exchange of expensive hepatocytes every 4 h in one assay.

Regarding the HepatoPac and HµREL methods, a co-culture system is adopted. Therefore, when conducting CL prediction, it is necessary to show that there is no metabolic reaction in the cells used for co-culture with hepatocytes. In addition, a complex system like co-culture makes evaluation more difficult.

Compared to these new evaluation systems, the merits of the vitrigel culture method are described below.

1. Since it is a monolayer culture system, the culture method is simple. Although it requires 2 weeks of cultivation, it may be maintained by replacing the medium only twice a week.

2. By simple seeding of hepatocytes on CVM, liver functions including CYP activity can be maintained, and IVIVC accuracy is high.

3. It can predict the CL of drug candidates with low to middle CL, whereas the HepatoPac method showed relatively low prediction accuracy for intermediate to high CL compounds.

In recent drug discovery, the use of drug candidates that are metabolized by non-CYP enzymes, such as UGT, aldehyde oxidase (AO), carboxylesterase (CES), flavin-containing monooxygenase (FMO) and so on, is increasing.²²⁾ The prediction accuracy of CL for such non-CYP substrates is still low, however, so an evaluation system with high prediction accuracy is desired to diminish the attrition rate of drug candidates. In future research, we will apply non-CYP compounds to our culture system.

Additionally, since the vitrigel culture can sustain liver function for a long period, the detection and identification of human-specific metabolites, which are difficult to detect in short-term culture systems, can be expected. Furthermore, the vitrigel culture system could be applied to the study of hepatic toxicity caused by reactive metabolites and/or long culture treatment.

In summary, we have demonstrated that the vitrigel culture method using the CVM chamber and PXB-cells is a promising in vitro system for predicting human hepatic CL with high accuracy for CYP substrates, including metabolically stable drug candidates.

Acknowledgments

This work was supported by a Grant to Mr. T. Matsumi and Ms. T. Watadani from PhoenixBio, for the supply of PXB-cells.

Conflict of Interest

The authors declare that this work was aided, in part, by the Japan Agency for Medical Research and Development (AMED) foundation.

Supplementary Materials

The online version of this article contains supplementary materials.

REFERENCES

• 1) Kola I, Landis J. Can the pharmaceutical industry reduce attrition rates? Nat. Rev. Drug Discov., 3, 711–715 (2004).
• 2) Arrowsmith J, Miller P. Trial watch: phase II and phase III attrition rates 2011–2012. Nat. Rev. Drug Discov., 12, 569 (2013).
• 3) Di L, Trapa P, Obach RS, Atkinson K, Bi YA, Wolford AC, Tan B, McDonald TS, Lai Y, Tremaine LM. A novel relay method for determining low-clearance values. Drug Metab. Dispos., 40, 1860–1865 (2012).
• 4) Chan TS, Yu H, Moore A, Khetani SR, Tweedie D. Meeting the challenge of predicting hepatic clearance of compounds slowly metabolized by cytochrome P450 using a novel hepatocyte model, HepatoPac. Drug Metab. Dispos., 41, 2024–2032 (2013).
• 5) Hultman I, Vedin C, Abrahamsson A, Winiwarter S, Darnell M. Use of HμREL human coculture system for prediction of intrinsic clearance and metabolite formation for slowly metabolized compounds. Mol. Pharm., 13, 2796–2807 (2016).
• 6) Sanoh S, Naritomi Y, Fujimoto M, Sato K, Kawamura A, Horiguchi A, Sugihara K, Kotake Y, Ohshita H, Tateno C, Horie T, Kitamura S, Ohta S. Predictability of plasma concentration–time curves in humans using single-species allometric scaling of chimeric mice with humanized liver. Xenobiotica, 45, 605–614 (2015).
• 7) Miyamoto M, Iwasaki S, Chisaki I, Nakagawa S, Amano N, Hirabayashi H. Comparison of predictability for human pharmacokinetics parameters among monkeys, rats, and chimeric mice with humanised liver. Xenobiotica, 47, 1052–1063 (2017).
• 8) Ishida Y, Yamasaki C, Yanagi A, Yoshizane Y, Fujikawa K, Watashi K, Abe H, Wakita T, Hayes CN, Chayama K, Tateno C. Novel robust in vitro hepatitis B virus infection model using fresh human hepatocytes isolated from humanized mice. Am. J. Pathol., 185, 1275–1285 (2015).
• 9) Kanamori T, Togawa-Iwata Y, Segawa H, Yamamuro T, Kuwayama K, Tsujikawa K, Inoue H. Use of hepatocytes isolated from a liver-humanized mouse for studies on the metabolism of drugs: application to the metabolism of fentanyl and acetylfentanyl. Forensic Toxicology, 36, 467–475 (2018).
• 10) Umetsu T, Inoue J, Kogure T, Kakazu E, Ninomiya M, Iwata T, Takai S, Nakamura T, Sano A, Shimosegawa T. Inhibitory effect of silibinin on hepatitis B virus entry. Biochemistry and Biophysics Reports, 14, 20–25 (2018).
• 11) Yamasaki C, Kataoka M, Kato Y, Kakuni M, Usuda S, Ohzone Y, Matsuda S, Adachi Y, Ninomiya S, Itamoto T, Asahara T, Yoshizato K, Tateno C. In vitro evaluation of cytochrome P450 and glucuronidation activities in hepatocytes isolated from liver-humanized mice. Drug Metab. Pharmacokinet., 25, 539–550 (2010).
• 12) Yamasaki C, Yanagi M, Kageyama Y, Iwasaki Y, Ishida Y, Tateno C. The 42nd Annual Meeting of the Japanese Society of Toxicology, Kanazawa (2015).
• 13) Watari R, Kakiki M, Oshikata A, Takezawa T, Yamasaki C, Ishida Y, Tateno C, Kuroda Y, Ishida S, Kusano K. A long-term culture system based on a collagen vitrigel membrane chamber that supports liver-specific functions of hepatocytes isolated from mice with humanized livers. J. Toxicol. Sci., 43, 521–529 (2018).
• 14) Tateno C, Kawase Y, Tobita Y, Hamamura S, Ohshita H, Yokomichi H, Sanada H, Kakuni M, Shiota A, Kojima Y, Ishida Y, Shitara H, Wada NA, Tateishi H, Sudoh M, Nagatsuka S, Jishage K, Kohara M. Generation of novel chimeric mice with humanized livers by using hemizygous cDNA-uPA/SCID mice. PLOS ONE, 10, e0142145 (2015).
• 15) Stringer R, Nicklin PL, Houston JB. Reliability of human cryopreserved hepatocytes and liver microsomes as in vitro systems to predict metabolic clearance. Xenobiotica, 38, 1313–1329 (2008).
• 16) Dave RA, Morris ME. A quantitative threshold for high/low extent of urinary excretion of compounds in humans. Biopharm. Drug Dispos., 37, 287–309 (2016).
• 17) Obach RS, Lombardo F, Waters NJ. Trend analysis of a database of intravenous pharmacokinetic parameters in humans for 670 drug compounds. Drug Metab. Dispos., 36, 1385–1405 (2008).
• 18) Uchimura T, Kato M, Saito T, Kinoshita H. Prediction of human blood-to-plasma drug concentration ratio. Biopharm. Drug Dispos., 31, 286–297 (2010).
• 19) Roberts MS, Rowland M. Correlation between in vitro microsomal enzyme activity and whole organ hepatic elimination kinetics: analysis with a dispersion model. J. Pharm. Pharmacol., 38, 177–181 (1986).
• 20) Lazarska KE, Dekker SJ, Vermeulen NPE, Commandeur JNM. Effect of UGT2B7*2 and CYP2C8*4 polymorphisms on diclofenac metabolism. Toxicol. Lett., 284, 70–78 (2018).
• 21) Wang Y, Roy A, Sun L, Lau CE. A double-peak phenomenon in the pharmacokinetics of alprazolam after oral administration. Drug Metab. Dispos., 27, 855–859 (1999).
• 22) Cerny MA. Prevalence of non-cytochrome P450-mediated metabolism in food and drug administration-approved oral and intravenous drugs: 2006–2015. Drug Metab. Dispos., 44, 1246–1252 (2016).