Abstract
Accurate prediction of pharmacokinetics (PK) parameters in humans from animal data is difficult for various reasons, including species differences. However, chimeric mice with humanized liver (PXB mice; urokinase-type plasminogen activator/severe combined immunodeficiency mice repopulated with approximately 80% human hepatocytes) have been developed. The expression levels and metabolic activities of cytochrome P450 (P450) and non-P450 enzymes in the livers of PXB mice are similar to those in humans. In this study, we examined the predictability for human PK parameters from data obtained in PXB mice. Elimination of selected drugs involves multiple metabolic pathways mediated not only by P450 but also by non-P450 enzymes, such as UDP-glucuronosyltransferase, sulfotransferase, and aldehyde oxidase in liver. Direct comparison between in vitro intrinsic clearance (CLint,in vitro) in PXB mice hepatocytes and in vivo intrinsic clearance (CLint,in vivo) in humans, calculated based on a well stirred model, showed a moderate correlation (r2 = 0.475, p = 0.009). However, when CLint,in vivo values in humans and PXB mice were compared similarly, there was a good correlation (r2 = 0.754, p = 1.174 × 10−4). Elimination half-life (t1/2) after intravenous administration also showed a good correlation (r2 = 0.886, p = 1.506 × 10−4) between humans and PXB mice. The rank order of CL and t1/2 in human could be predicted at least, although it may not be possible to predict absolute values due to rather large prediction errors. Our results indicate that in vitro and in vivo experiments with PXB mice should be useful at least for semiquantitative prediction of the PK characteristics of candidate drugs in humans.
Introduction
It is important to predict human pharmacokinetics (PK) and metabolism of drug candidates in the preclinical stage of pharmaceutical development. Various approaches to predict human clearance (CL) with in vitro metabolic systems, such as human liver microsomes and hepatocytes, have been reported (Nagilla et al., 2006; Brown et al., 2007; Fagerholm, 2007; Stringer et al., 2008; Chiba et al., 2009; Hallifax et al., 2010) but with limited success. One of the reasons for the discrepancy between predicted and observed CL may be that the preparation, storage, and experimental treatment of hepatocytes alter the normal function of metabolic enzymes (Wang et al., 2005). Although this might be ameliorated by using fresh hepatocytes immediately after isolation from the liver, these are not readily available and in any case show considerable interindividual differences.
It has become possible recently to predict CL and half-life (t1/2) by means of computational approaches and physiologically based modeling (Ekins and Obach, 2000; De Buck et al., 2007; Tabata et al., 2009; Paixão et al., 2010). Accurate prediction of human PK is a key issue for the development of new drugs, because many new drug candidates with diverse chemical structures are metabolized not only by cytochrome P450 (P450) but also by non-P450 enzymes, such as UDP-glucuronosyltransferase (UGT) and sulfotransferase (SULT). It is also necessary to take into account the effects of cell permeability, transporter-mediated uptake, and excretion (Chiba et al., 2009; Huang et al., 2010).
Chimeric mice with humanized liver (PXB mice; PhoenixBio Co., Ltd., Hiroshima, Japan) have been generated from urokinase-type plasminogen activator/severe combined immunodeficiency mice transplanted with human hepatocytes (Tateno et al., 2004). In these mice, approximately 80% of the hepatocytes are human. The expression levels and metabolic activities of P450 and non-P450 enzymes in livers of PXB mice with a high replacement index (RI) are similar to those of humans (Katoh et al., 2004, 2005), and human-specific metabolites are formed in PXB mice (Inoue et al., 2009; Kamimura et al., 2010; Yamazaki et al., 2010; De Serres et al., 2011). Thus, PXB mice could be a good in vivo model for predicting drug metabolism in humans.
However, quantitative methods for predicting PK parameters of humans from data in PXB mice have not been established yet. Therefore, we selected 13 model compounds that are metabolized by P450 and/or non-P450 enzymes in liver and compared the PK parameters in humans and PXB mice, using both in vitro and in vivo approaches, to evaluate the utility of this animal model for the prediction of human PK.
Materials and Methods
Chemicals.
6-Deoxypenciclovir and mirtazapine were obtained from Toronto Research Chemicals Inc. (North York, ON, Canada). Dapsone, lamotrigine, salbutamol, and sulindac were purchased from Sigma-Aldrich (St. Louis, MO). Diclofenac was purchased from Tokyo Chemical Industry Co. Ltd. (Tokyo, Japan). Fasudil was obtained from Tocris Bioscience (Bristol, UK). (S)-Naproxen was purchased from Cayman Chemical (Ann Arbor, MI). Ibuprofen, ketoprofen, and (S)-warfarin were purchased from Wako Pure Chemicals (Osaka, Japan). Zaleplon was kindly provided by King Pharm. Inc. (Bristol, UK). All of the other regents and solvents were commercial products of the highest available grade or analytical grade.
Animals.
The present study was approved by the animal ethics committee and was conducted in accordance with the regulations on the use of living modified organisms of Hiroshima University. PXB mice (10–14 weeks of age) with human hepatocytes were prepared by the reported method (Tateno et al., 2004). Human hepatocytes of a donor (African-American boy, 5 years old) were obtained from BD Biosciences (San Jose, CA). PXB mice were housed in a temperature- and humidity-controlled environment under a 12-h light/dark cycle.
The RI was determined by the measurement of human albumin in blood collected from the tail vein. The RI was estimated by the correlation curve between the human albumin levels in mouse blood and determined by using human-specific cytokeratin 8/18-immunostained liver sections (Tateno et al., 2004). The RI values of PXB mice used in this study ranged from 73.4 to 93.4%.
Administration.
Drug solution (5 ml/kg) was administered intravenously to PXB mice at 0.3 to 5 mg/kg body weight. Solutions of dapsone, diclofenac, 6-deoxypenciclovir, fasudil, ketoprofen, ibuprofen, mirtazapine, naproxen, salbutamol, and sulindac were prepared in saline. In the cases of ketoprofen, ibuprofen, naproxen, and sulindac, equivalent amounts of alkali were added. Dapsone solutions contained 10% dimethyl sulfoxide (DMSO), and mirtazapine solutions were prepared with 10% DMSO and equivalent amounts of hydrochloric acid. Lamotrigine, and zaleplon solutions were prepared with 10% DMSO and 10% polyethylene glycol 400 in saline. Equivalent amounts of hydrochloric acid also were added to the solutions of lamotrigine and zaleplon. Warfarin was formulated in 3% DMSO and 97% saline with an equivalent amount of sodium hydroxide.
Blood samples after dosing were collected from orbital veins of PXB mice at predetermined times using heparinized glass. These samples were centrifuged, and the plasma was stored at −30°C.
Determination of Drug Concentrations in Plasma.
A 10 μl aliquot of plasma was added to 40 μl of acetonitrile or methanol containing an internal standard (carbamazepine, ketoprofen, or ibuprofen). The mixtures were centrifuged at 14,000g for 5 min, and the supernatant was subjected to liquid chromatography tandem mass spectrometry (LC/MS/MS).
Isolation and Purification of Hepatocytes from PXB Mice.
Fresh hepatocytes were isolated from PXB mice (13–15 weeks of age) by means of the in situ collagenase perfusion method and purified as described previously (Yamasaki et al., 2010). PXB mouse hepatocytes (h-hepatocytes) contained approximately 7% mouse hepatocytes. We used h-hepatocytes purified by the use of 66Z rat IgG and magnetic beads bearing anti-rat IgG antibodies. The magnetic removal of mouse hepatocytes reduced the level of mouse hepatocytes to approximately 2% (in this study, the purity of human hepatocytes from PXB mouse hepatocytes ranged from 96.6 to 99.7% after purification). Cell viability of the hepatocytes used in the experiments ranged from 79 to 91%, as determined by means of the trypan blue exclusion test.
In Vitro Metabolic Studies Using h-Hepatocytes.
The h-hepatocyte suspension (1 × 106 cells/ml) was incubated in Krebs-Henseleit buffer without serum in the presence of 10 μM of the test drug at 37°C under an atmosphere of 5% CO2/95% O2. The final concentration of acetonitrile was 0.5% (v/v) in the reaction mixture. The plates (24 wells) were shaken gently with an orbital shaker. The incubation mixture was sampled at 0, 0.25, 0.5, 1, and 2 h after treatment, and reactions were stopped by freezing the mixture in liquid nitrogen. When required, the samples were thawed, spiked with two volumes of acetonitrile or methanol containing an internal standard, and centrifuged. Aliquots of the supernatants were subjected to LC/MS/MS.
LC/MS/MS Conditions.
Aliquots (10 μl) of plasma and h-hepatocyte suspension were introduced into the high-performance liquid chromatography system with an autosampler (Agilent Technologies, Santa Clara, CA). Several mobile phase conditions were used. Mobile phase condition 1 consisted of 10 mM ammonium acetate (A) and acetonitrile (B) on an Inertsil ODS-3 column (3 μm, 50 × 2.1 mm; GL Sciences Inc., Tokyo, Japan) at 40°C for the analysis of diclofenac, ibuprofen, ketoprofen, mirtazapine, (S)-naproxen, sulindac, and (S)-warfarin. The flow rate was set at 0.2 ml/min. The starting condition for the high-performance liquid chromatography gradient was 90:10 (A/B). From 0 to 5 min, the mobile phase composition was changed linearly to 10:90 (A/B), and this was held until 8 min. The gradient then was returned to 90:10 (A/B) linearly from 8 to 8.1 min, and the column was re-equilibrated to the initial condition.
Mobile phase condition 2 consisted of 0.1% formic acid (A) and methanol (B) on a YMC-Triart C18 column (3 μm, 50 × 2.1 mm; YMC Co., Ltd., Kyoto, Japan) for the analysis of dapsone, 6-deoxypenciclovir, fasudil, lamotrigine, salbutamol, and zaleplon. The starting condition was 90:10 (A/B). From 0 to 5 min, the mobile phase composition was changed linearly to 10:90 (A/B), and this was maintained until 8 min, then the column was re-equilibrated to the initial condition.
The MS/MS experiments were conducted by using API2000 LC/MS/MS systems (Applied Biosystems, Foster, CA). Mass number of the ionization mode, molecular ion, and product ion for the model compounds were as follows: dapsone m/z = 248.99 [M + H]+ to 92.18, 6-deoxypenciclovir m/z = 238.05 [M + H]+ to 210.95, diclofenac m/z = 294.14 [M + H]− to 249.53, fasudil m/z = 292.07 [M + H]+ to 99.09, ibuprofen m/z = 204.88 [M + H]− to 158.52, ketoprofen m/z = 253.16 [M + H]− to 208.73, lamotrigine m/z = 256.03 [M + H]+ to 210.96 mirtazapine m/z = 266.14 [M + H]+ to 194.97, (S)-naproxen m/z = 228.68 [M + H]− to 168.55, salbutamol m/z = 240.18 [M + H]+ to148.03, sulindac m/z = 357.07 [M + H]+ to 232.96, (S)-warfarin m/z = 309.06 [M + H]+ to162.97, zaleplon m/z = 306.08 [M + H]+ to 236.12.
Determination of PK Parameters.
Pharmacokinetic parameters were determined by noncompartmental methods using the concentration-time curve profile. The total clearances (CLt) after intravenous administration were calculated as dose/AUCiv. AUCiv values were estimated from the time course using the trapezoidal method with extrapolation from the last quantifiable point to infinity. The terminal elimination t1/2 was estimated as ln 2/ke, where ke is that of the plot of the terminal elimination phase on a logarithmic scale.
Calculation of In Vitro Intrinsic Clearance.
In vitro intrinsic clearance (CLint,in vitro) was calculated from the time course of the disappearance of the test drug during incubation with h-hepatocytes. Each plot was fitted to the first-order elimination rate constant as C(t) = C0*exp(−ke*t), where C(t) and C0 are the concentration of unchanged test drug at incubation time t and that at preincubation and ke is the disappearance rate constant of the unchanged drug.
Subsequently, CLint,in vitro (μl · min−1 · 106 cells−1) values were converted to CLint,in vitro (ml · min−1 · kg−1) for the whole body. CLint,in vitro data were scaled up using physiological parameters, human liver weight 26 g/kg (Davies and Morris, 1993) and PXB mouse liver weight 140 g/kg, and the hepatocellularity (132 × 106 cells/g liver) of PXB mice. These parameters were taken from the average of observed data in PXB mice (RI = 80%).
Calculation of In Vivo Intrinsic Clearance.
CLt of PXB mice was calculated from the plasma concentrations after dosing using noncompartmental methods as described. CLt was assumed to be equal to the hepatic clearance.
Values of CLt, plasma unbound fraction (fu), and blood/plasma concentration ratio (Rb) in humans were taken from the literature.
In vivo intrinsic clearance (CLint,in vivo) was calculated from the in vivo CLt, fu, Rb, and average hepatic blood flow (Q) based on a well stirred model as CLint,in vivo = CLt/{(fu/Rb) × (1 − CLt/Q)} (Pang and Rowland, 1977). These CLt values were converted to blood clearance using Rb values.
The Q values of humans and PXB mice were set at 21 and 90 ml · min−1 · kg−1 (same as in normal mice), respectively (Davies and Morris, 1993). In addition, Rb and fu of human were assumed to be equivalent to those of PXB mice. If CLt of drugs exceeded liver blood flow, then the hepatic clearance was taken as 90% of liver blood flow. CLint,in vivo of 6-deoxypenciclovir, lamotrigine, and sulindac were evaluated from oral clearance (CLoral) as CLoral/fu/Rb.
Results
Selection of the Model Compounds for Analysis.
In this study, we selected 13 compounds with diverse chemical structures (Fig. 1); Elimination of these selected drugs involves multiple metabolic pathways mediated not only by P450 but also by non-P450 enzymes, such as UGT, SULT, and aldehyde oxidase (AO) in liver. Mirtazapine and warfarin were known to be mainly metabolized by P450. Diclofenac, ibuprofen, and naproxen were metabolized by both UGT and P450. Furthermore, the model compounds metabolized by AO, such as 6-deoxypenciclovir, fasudil, sulindac, and zaleplon, were added in this study. These were reflected in the data set that spanned a wide range of PK parameter characteristics. CLt and t1/2 after intravenous administration of selected model drugs to humans were obtained from the literature. If CLt after intravenous administration was not available from the literature, we used the value of CLt/F after oral administration. The PK parameters and the major enzymes responsible for drug metabolism in humans are shown in Table 1. The spreadsheet containing these values with the literature references is included as an attachment in the supplemental data (Supplemental Tables 1 and 2).
Disappearance of Parent Drugs after Incubation.
Remaining amounts of all of the compounds decreased linearly for 2 h on incubation with h-hepatocytes. The values of CLint,in vitro in hepatocytes, calculated using scaling factors to humans and PXB mice whole body as described under Materials and Methods, are listed in Table 2. These CLint,in vitro values covered a wide range. Fasudil showed high clearance, whereas warfarin was very stable.
PK Study of the Model Compounds in PXB Mice.
Plasma concentrations and PK parameters in PXB mice after intravenous administration of drug solutions at 0.3 to 5 mg/kg are shown in Fig. 2 and Tables 3 and 4. Each RI value in PXB mice was 73.4 to 93.4%.
CLt values of warfarin and lamotrigine were relatively low, whereas those of fasudil and salbutamol were much higher; the range of CLt was 0.2 to 198 ml · min−1 · kg−1. The t1/2 value of lamotrigine was the longest, and those of 6-deoxypenciclovir and fasudil were short, as shown in Table 3.
Comparison of Intrinsic CL between h-Hepatocytes and Humans.
Direct comparison between CLint,in vitro from h-hepatocytes and CLint,in vivo calculated for a well stirred model in humans showed a moderate correlation (r2 = 0.475, p = 0.009) (Fig. 3). For 2 of 13 (15%) compounds, observed CLint,in vivo was predicted within a 3-fold error from hepatocyte CLint, in vitro. However, for 8 of 13 (62%) compounds, observed CLint,in vivo was predicted with a 3- to 10-fold error.
Figure 4 shows the relationship between CLint,in vivo and CLint,in vitro for PXB mice; again, the correlation was moderate (r2 = 0.435, p = 0.014). For 6 of 13 (46%) compounds, CLint,in vivo of PXB mice was predicted within a 3-fold error from h-hepatocyte CLint,in vitro. For 5 of 13 (38%) compounds, CLint,in vivo was predicted within a 3- to 10-fold error.
Relationship between Intrinsic Clearance in Humans and PXB Mice In Vivo.
We directly compared CLint,in vivo calculated based on a well stirred model in humans and PXB mice. As shown in Fig. 5, there was a good correlation (r2 = 0.754, p = 1.174 × 10−4) between literature CLint,in vivo in human and measured CLint,in vivo of PXB mice for these compounds. For 4 of 13 (31%) compounds, observed CLint in vivo in humans was predicted within a 3-fold error from PXB mouse CLint,in vivo. For 7 of 13 (54%) compounds, human CLint,in vivo was predicted within a 3- to 10-fold error.
Relationship of Elimination t1/2 between Humans and PXB mice.
Figure 6 shows the relationship of t1/2 after intravenous administration between humans and PXB mice. Compounds for which literature data were not available were excluded from this figure. A good correlation (r2 = 0.886, p = 1.506 × 10−4) was found. For 6 of 9 (67%) compounds, human observed t1/2 was predicted within a 3-fold error from PXB mice t1/2. For 3 of 9 (33%) compounds, the error was in the range of 3- to 10-fold.
Discussion
The prediction of human PK parameters is an important step during the preclinical development of pharmaceuticals to reduce costs by enabling the early elimination of candidates with unsuitable properties. However, species differences make it difficult to predict human PK from animal data; monkey data may lead to underprediction (Chiou and Buehler, 2002; Akabane et al., 2010), whereas dog data may cause overestimation (Chiou et al., 2000). In vitro-in vivo scaling from data obtained with human hepatic microsomes and hepatocytes is a widely used approach but often results in the underprediction of in vivo CL (Obach, 1999). We considered the possibility that PXB mice, in which hepatocytes are replaced with human hepatocytes to the extent of approximately 80% (Tateno et al., 2004), may have superior predictive utility, because the expression levels and activities of both P450 and non-P450 enzymes well reflect those of the donor hepatocytes (Yoshitsugu et al., 2006; Yamasaki et al., 2010). In this study, we checked metabolic activities (CYP2C9, CYP2D6, UGT, SULT, and AO) using probe substrates between donor hepatocytes and h-hepatocytes purified from PXB mice (Supplemental Table 3) as well as the expression of drug transporters and blood albumin (Tateno et al., 2004; Nishimura et al., 2005).
For the present study, we selected 13 model compounds with diverse chemical structures (Fig. 1), which are metabolized through multiple pathways by P450 and non-P450 enzymes, such as UGT, SULT, and AO. Their values of CL cover a wide range from 0.055 to 118 ml · min−1 · kg−1 (Table 1).
First, we performed an in vitro metabolic study using fresh h-hepatocytes isolated from PXB mice. We calculated CLint,in vitro using fresh h-hepatocytes and compared the results with human CLint,in vivo estimated by use of a well stirred model (Pang and Rowland, 1977). These results using a parallel tube model and a dispersion model were also similar to those of a well stirred model (S. Sanoh, unpublished observations). A moderate correlation (r2 = 0.475, p = 0.009) was found, but this approach was not superior to prediction using other methods.
CLint,in vitro values of diclofenac, ibuprofen, warfarin, and zaleplon were approximately similar to reported values using cryopreserved hepatocytes (Ekins and Obach, 2000; Nagilla et al., 2006; Stringer et al., 2008), supporting the idea that CLint,in vitro values are similar in fresh hepatocytes and cryopreserved hepatocytes (Naritomi et al., 2003; McGinnity et al., 2004).
A similar correlation (r2 = 0.435, p = 0.014) was observed between CLint,in vitro and CLint,in vivo in PXB mice (Fig. 4). In both cases, the numbers of compounds for which absolute values of CLint were predicted within a 3-fold error were insufficient.
Therefore, we next evaluated the predictability of hepatic clearance and t1/2 from in vivo data in PXB mice. The values of CLint,in vivo estimated by intravenous administration in PXB mice were well correlated (r2 = 0.754, p = 1.174 × 10−4) with observed CLint,in vivo in human. Surprisingly, we also found a good correlation (r2 = 0.886, p = 1.506 × 10−4) between t1/2 values in PXB mice and humans. However, although the rank order was the same, there were rather large prediction errors, so it may not be possible to predict absolute values. This is consistent with the findings of Xiao et al. (2010) in PXB mice.
We used PXB mice that showed that the average RI values were approximately 80%. It was a concern that the contribution of the remaining approximately 20% mice hepatocytes may be reflected on clearance. CLint,in vitro values of these model compounds in host mice hepatocytes (severe combined immunodeficiency mice) were almost higher than those of h-hepatocytes within a 10-fold range (Supplemental Fig. 1). The extent of the difference may not influence the predictability of CLint,in vivo.
For the estimation of CLint,in vivo in PXB mice, the fu values of those model compounds is the same as those in humans because human albumin is expressed in the blood of PXB mice. Inoue et al. (2009) reported fu value of warfarin in PXB mice was similar to that in humans. Furthermore, fu values of some compounds (dapsone, diclofenac, ketoprofen, salbutamol, and zaleplon) in this study were also approximately similar to those in humans (S. Sanoh, unpublished observations).
We assumed that the Rb values of those model compounds is also the same as those in humans, because Rb values of some compounds (dapsone, diclofenac, ketoprofen, salbutamol, and zaleplon) in this study were also approximately similar to those in humans (S. Sanoh, unpublished observations).
Q values were assumed to be 90 ml · min−1 · kg−1, respectively, corresponding to the values of normal mice (Davies and Morris, 1993). In further work, it would be desirable to examine whether these values are appropriate.
In this study, we selected model compounds metabolized not only by P450, but also by non-P450 enzymes, including AO. 6-Deoxypenciclovir, fasudil, sulindac, and zaleplon are metabolized mainly by AO in humans. It has been reported that human CL for drugs metabolized by AO may be underpredicted from data obtained with human liver cytosol and S9 due to the loss or deactivation of AO during preparation, homogenization, storage, and experimental procedures (Zientek et al., 2010). PXB mice have high AO activity, being similar to humans (Kitamura et al., 2008), and may be a useful source of fresh h-hepatocytes.
Our results indicate that PXB mice can be used at least for semiquantitative prediction of not only CLt but also t1/2 in humans. PXB mice also would be useful for in vitro estimation and comparison of PK of various candidate compounds, because large amounts of fresh, identical hepatocytes (1.1 × 108 cells/mouse) are available by transplantation of donor hepatocytes (2.5 × 105 cells/mouse). The combination of in vitro study in PXB mice and in vitro study using PXB hepatocytes may prove to be particularly effective.
Authorship Contributions
Participated in research design: Sanoh, Sugihara, Kotake, Tayama, Horie, Kitamura, and Ohta.
Conducted experiments: Sanoh and Horiguchi.
Contributed new reagents or analytic tools: Sugihara, Ohshita, and Tateno.
Performed data analysis: Sanoh and Horiguchi.
Wrote or contributed to the writing of the manuscript: Sanoh, Kotake, Tateno, and Ohta.
Acknowledgments
We thank members in PhoenixBio Co., Ltd. for the isolation of hepatocytes from PXB mice.
Footnotes
This work was supported by a Grant-in-Aid for Young Scientists (B) from Japan Society for the Promotion of Science [Grant 22790109]; and PhoenixBio, Co., Ltd.
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
↵ The online version of this article (available at http://dmd.aspetjournals.org) contains supplemental material.
-
ABBREVIATIONS:
- PK
- pharmacokinetics
- CL
- clearance
- AO
- aldehyde oxidase
- CLint,in vitro
- in vitro intrinsic clearance
- CLint,in vivo
- in vitro intrinsic clearance
- CLoral
- oral clearance
- CLt
- total clearance
- P450
- cytochrome P450
- DMSO
- dimethyl sulfoxide
- fu
- plasma unbound fraction
- h-hepatocytes
- PXB mice hepatocytes
- LC/MS/MS
- liquid chromatography tandem mass spectrometry
- NAT
- N-acetyltransferase
- PXB mice
- chimeric mice with humanized liver
- Q
- hepatic blood flow
- Rb
- blood/plasma concentration ratio
- RI
- replacement index
- SULT
- sulfotransferase
- t1/2
- half-life
- UGT
- UDP-glucuronosyltransferase
- AUCiv
- area under the concentration versus time curve by intravenous administration.
- Received August 5, 2011.
- Accepted November 2, 2011.
- Copyright © 2012 by The American Society for Pharmacology and Experimental Therapeutics