Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

Hepatic clearance for the metabolism of compounds kinetically consists of two major determinants: intrinsic (metabolic) clearance of the unbound compound and unbound fraction of compound in the blood (or plasma when corrected by the blood-to-plasma partition). Generally, the intrinsic clearance for the unbound compound is measured in vitro by the incubation of isolated hepatocytes or subcellular fractions such as S-9 and microsomes in the protein-free medium. The in vitro metabolic parameters thus obtained are extrapolated by using anatomical parameters such as cell numbers and protein content in the intact liver for the prediction of in vivo metabolic activity (Houston and Carlile, 1997; Iwatsubo et al., 1997; Obach, 1999). Separate experiments necessarily are further carried out to measure the unbound fraction in the plasma. Many technical problems including adsorption of compounds to the apparatus during the equilibrium dialysis and ultra-filtration often hamper the accuracy of the evaluated values (Bertilsson et al., 1979; Desoye, 1988). To improve the accuracy and avoid complexity for predicting in vivo metabolic clearance from in vitro experiments, we have recently developed a novel and convenient in vitro method for predicting in vivo metabolic clearance by using freshly isolated rat hepatocytes suspended in rat serum (Shibata et al., 2000). Oral bioavailability and hepatic clearance for 16 widely used compounds were well predicted directly from the in vitro metabolic clearance values obtained from a single incubation without separate evaluation of unbound fraction in the plasma. The purposes of the present study were to 1) determine whether the same methodology was applicable to the prediction of in vivo metabolic activity in humans by using cryopreserved human hepatocytes and 2) establish the in vitro-in vivo scaling-up paradigm to calibrate the interindividual variation of the metabolic activities among cryopreserved preparations from different subjects for more reliable and quantitative prediction in humans.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Naloxone and lidocaine were purchased from Nacalai Tesque (Kyoto, Japan) and Tokyo Kasei Kogyo Co., Ltd. (Tokyo, Japan), respectively. Buspirone, metoprolol, phenacetin, propranolol, quinidine, timolol, and verapamil were obtained from Sigma-Aldrich (St. Louis, MO). Antipyrine, caffeine, diazepam, diclofenac, and imipramine were obtained from Wako Pure Chemical Industries (Osaka, Japan). Blood was collected from three healthy male volunteers aged 25 to 40 years old and allowed to coagulate for 3 h at room temperature. The blood was later centrifuged (15 min, 1800g) to obtain serum. The serum was stored at 80°C until use. The pH of human serum was adjusted to 7.4 at 37°C by adding 1N-HCl solution before use. Cryopreserved human hepatocytes (lot numbers 56, 57, 64, 70, 73, 83, 97, 100, 106, and 120) were purchased from In Vitro Technologies, Inc. (Baltimore, MD). Cell viability was assessed using the 0.4% trypan blue exclusion test, and the count of living cells was started 5 min after mixing the pigment. Cell viabilities were between 45 and 60%. Hepatocytes were resuspended in 100% human serum at an ice-cold temperature at the following densities: 1 × 10⁶ cells/ml for naloxone, buspirone, verapamil, lidocaine, imipramine, metoprolol, and timolol; 2 × 10⁶ cells/ml for quinidine, caffeine, propranolol, diclofenac, and phenacetin; and 5 × 10⁶ cells/ml for antipyrine and diazepam. Suspensions of hepatocytes (370 µl) were pipetted into 1.5-ml tubes, and an aliquot (3.7 µl) of 100 µM compound in a water (or 50% CH₃CN for quinidine and phenacetin) was added to obtain the final concentration of 1 µM (or 50 µM for antipyrine and 10 µM for caffeine). Each sample (50 µl) was transferred to two 96-well plates with flat bottoms (n = 3), each of which was used for the incubation or for the control. Ninety-six well plates were incubated at 37°C with shaking at 150 rpm under 95% O₂/5% CO₂ in the water bath incubator. After the onset of incubation, the plates were placed on ice at the designated time point, and the reaction was terminated by the addition of a 150-µl ice-cold ethanol solution containing the internal standard (no internal standard for antipyrine and caffeine). The sample was centrifuged (10,000g × 10 min), and the amount of compound remaining in the supernatant was measured by HPLC¹-UV (antipyrine and caffeine) or LC-MS/MS (others) as described below.

Antipyrine was measured at 254 nm by the Alliance 2690-2487 HPLC-UV system (Waters Corp., Milford, MA) that was connected to Inertsil ODS-3 4.6 × 250 mm (GL-Sciences, Tokyo, Japan), which acted as an analytical column. The HPLC method involved the isocratic elution with acetonitrile/water (25:75) containing 10 mM ammonium acetate at the flow rate of 1 ml/min. The retention time of antipyrine was 7 min. Caffeine was measured by the UV wavelength of 280 nm at a flow rate of 1 ml/min of acetonitrile/water (10:90) containing 0.1% trifluoroacetic acid. The retention time of caffeine was 16 min. Other compounds were measured by the Alliance HT 2790 HPLC (Waters Corp.), PU-1580 HPLC pump (Jasco, Tokyo, Japan), NANOSPACE SI-2 switching valves (Shiseido, Osaka, Japan), and API-3000 LC-MS/MS detector (PerkinElmerSciex Instruments, Boston, MA) with a turbo ionspray interface. Multiple reaction monitoring of positive-ion mode was used for all analyses. Analytical methods including the selection of ions and parameters for multiple reaction monitoring were automatically obtained for each compound by the application software, Analyst (PerkinElmerSciex Instruments). Mass number of molecular ions and product ions for each compound was identified as follows (molecular > product): naloxone 328.4 > 310.4, buspirone 386.2 > 122.2, verapamil 455.3 > 165.2, lidocaine 235.2 > 86.2, imipramine 281.1 > 86.0, metoprolol 268.3 > 116.0, timolol 317.1 > 261.1, diazepam 285.1 > 193.3, quinidine 325.1 > 307.3, propranolol 260.2 > 116.3, diclophenac 295.9 > 215.2, and phenacetin 180.1 > 138.0. A fast-gradient condition using two switching valves and pumps (3.5 min/cycle) was used for the analysis. Capcell Pak UG-120 4.0 × 10 mm (Shiseido) was used as an analytical column, and the flow rate of 1 ml/min of acetonitrile/water (10:90) containing 10 mM ammonium acetate was the initial condition used. After the injection of a sample (5 µl), the ratio of acetonitrile/water was changed to 90:10 linearly for 1 min and maintained for the next 0.5 min. The column was then washed with acetonitrile/water (90:10) containing 10 mM ammonium acetate at a back flow rate of 1 ml/min. The effluent was split with 0.2 ml/min, and only the effluent from 0.5 to 1.5 min after the injection was introduced into the LC-MS/MS detector. Modified conditions were used for metoprolol, timolol, and phenacetin. In the case of metoprolol and timolol, Symmetry Shield RP18/3.5 µM 2.1 × 10 mm (Waters Corp.) was used as the analytical column. After the injection of a sample, the ratio of acetonitrile/water was changed linearly to 66:34 for 1.4 min, and the effluent from 0.4 to 1.5 min after the injection was introduced into the LC-MS/MS detector. In the case of phenacetin, Inertsil ODS-3 2.1 × 150 mm (GL-Sciences) was used as the analytical column. After the injection of a sample, the ratio of acetonitrile/water was changed linearly to 34:66 for 3 min and then changed linearly to 90:10 for the next 2 min. The effluent from 0.5 to 5 min after the injection was introduced into the LC-MS/MS detector. Diazepam was commonly used as the internal standard. When diazepam was the analyte, quinidine was used as the internal standard.

For standard compounds, the following assumptions were reasonably applied to the prediction, 1) the hepatic metabolism is the major route of elimination, 2) all metabolic enzymes in the cryopreserved preparation of human hepatocytes remain active comparably to in vivo, and 3) the absorption is complete for all standard compounds. The in vitro intrinsic clearance (CL_{int, in vitro}) was calculated from the following equation by using cell density (D), incubation time (T, 120 min for the calculation of CL_{int, in vitro}), and the ratio (R) of unchanged compound concentration at time T to that at time 0 when the unbound drug concentration was much lower than its K_m value (Shibata et al., 2000); CL_{int, in vitro =} (log_e R)/(D × T). To extrapolate the in vitro clearance to the in vivo value, the empirical scaling factor (average SF₇₀₊₇₃) for the optimized cryopreserved preparation pooled from equal volumes of human hepatocytes (lot numbers 70 and 73) was calculated according to the method described under the Results section. The value of average SF_70 + 73 was 10.8 × 10⁹ cells/kg of body weight and used for the extrapolation as follows: CL_{H, int, in vitro, 70+73} = CL_{int, in vitro, 70+73} × average SF₇₀₊₇₃ where CL_{H, int, in vitro, 70+73} and CL_{int, in vitro, 70+73} represent the in vitro hepatic intrinsic clearance and in vitro intrinsic clearance, respectively, measured in the pooled preparation of lot 70 and 73. We chose the dispersion model as a liver model because a good predictability of hepatic availability (F_H) for high clearance drugs was previously reported (Iwatsubo et al., 1997). The hepatic clearance (CL_{H, predicted, 70+73}) was predicted from the obtained in vitro hepatic intrinsic clearance (CL_{H, int, in vitro, 70+73}) by using the following equation (eq. 1) with the dispersion model (Iwatsubo et al., 1997);

F_H was further calculated from F_H = 1  hepatic extraction ratio (E_H) = 1  CL_H/(Q_H × R_B). In these equations, the liver blood flow rate (Q_H) and dispersion number (D_N) for humans were assumed to be 20.7 ml/min/kg (Davies and Morris, 1993) and 0.17 (Roberts and Rowland, 1986), respectively. The blood-to-plasma concentration ratio (R_B) was used as reported or assumed to be unity if the value was not available.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Tables 1 and 2 summarize the pharmacokinetic profiles in humans and the results of extrapolations from the in vitro data for the standard compounds tested in the present study. These compounds were chosen to represent a wide range of oral bioavailability (2-96%) and plasma clearance (0.3-28.3 ml/min/kg). These standard compounds are reported to have complete absorption, negligible urinary excretion (<20% of dose), and the major route of elimination by hepatic metabolism. Therefore, it was reasonably assumed that the in vivo plasma clearance (CL_{P, in vivo}) and oral bioavailability (F_{PO, in vivo}) are equal to the hepatic metabolic clearance (CL_H) and F_H, respectively. In vivo values for hepatic intrinsic clearance (CL_{H, int, in vivo}) of standard compounds were calculated from F_{PO, in vivo} by the dispersion model using the iterative calculation method (Goal Seek method in Microsoft Excel). To calculate empirical scaling factor from the comparison of in vivo and in vitro hepatic intrinsic clearance, CL_{int, in vitro} values were evaluated in the cryopreserved preparations of human hepatocytes obtained from 10 different subjects. An approximately 3- to 5-fold variation in the CL_{int, in vitro} was observed among 10 cryopreserved preparations of human hepatocytes. The prediction of CL_{H, int, in vivo} resulted in a marked underestimation when the biologically based scaling factor [3.1 × 10⁹ cells/kg, calculated from the assumption that each gram of human liver contains 120 × 10⁶ cells/g liver (Iwatsubo et al., 1997), and an average human has 1800 g of liver (Davies and Morris, 1993)] was used to extrapolate CL_{int, in vitro} to CL_{H, int, in vivo} (Fig. 1). Mean values for the empirical scaling factor (SF_mean) were calculated by averaging the ratio of CL_{H, int, in vivo} to the corresponding in vitro values (CL_{int, in vitro}) for each standard compound (Table 1). The average SF_mean value among seven standard compounds was found to be 8.5 × 10⁹ cells/kg, which was approximately 3 times larger than that of biologically based value (3.1 × 10⁹ cells/kg). Reasonably accurate predictions were achieved (Table 1; Fig. 1) when the scaling factors, thus empirically obtained, were used for the extrapolation.

View this table: [in this window] [in a new window]   TABLE 1 In vitro-in vivo correlation of seven compounds using 10 individually prepared cryopreserved human hepatocytes and key parameters for in vivo prediction All these values were quoted from the literature as follows. Naloxone [Asali and Brown (1984); Holford (1998)]; buspirone [Gammans et al. (1986)]; verapamil [Gross et al. (1988); McAllister and Kirsten (1982); Obach (1999)]; lidocaine; [Wing et al. (1984), Remmel et al. (1991)]; imipramine [the pharmacokinetics were mean value of the report of Nagy and Johansson (1975) and Ciraulo et al. (1988)]; metoprolol [Johansson et al. (1974); Regardh et al. (1974); Regardh et al. (1981)]; timolol [Wilson et al. (1982), [Holford (1998].

View this table: [in this window] [in a new window]   TABLE 2 Profiles of 14 compounds tested with the pooled preparation of human hepatocytes lot 70 and 73 and the key parameters for the in vivo prediction All these values were quoted from the literature as follows. Naloxone [Asali and Brown (1984); Holford (1998)]; buspirone [Gammans et al. (1986)]; verapamil [Gross et al. (1988); McAllister and Kirsten (1982); Obach (1999)]; lidocaine [Remmel et al. (1991); Wing et al. (1984)]; imipramine [the pharmacokinetics were mean value of the report of Ciraulo et al. (1988) and Nagy and Johansson (1975)]; metoprolol [Johansson et al. (1974); Regardh et al. (1974); Regardh et al. (1981)]; timolol [Wilson et al. (1982), Holford (1998)]; antipyrine [Elfstrom and Lindgren (1978); Vesell et al. (1975)]; diazepam [Divoll et al. (1983); Greenblatt et al. (1980); Maguire et al. (1980)]; quinidine [Guentert et al. (1979); Hardy and Schentag (1988); Hughes et al. (1975)]; caffeine [Blanchard (1982); Newton et al. (1981)]; propranolol [Walle et al. (1989); Wilson et al. (1982)]; diclofenac [Chan et al. (1987); Obach (1999); Willis et al. (1980)]; phenacetin [Raaflaub and Dubach (1975); Vesell et al. (1975)].

View larger version (23K): [in this window] [in a new window]   Fig. 1.   CLH, int, in vivo versus CLint, in vitro in 10 cryopreserved preparations of human hepatocytes. Each symbol represents the mean of three determinations. Data obtained from lot 70 and 73 are shown by open triangles and squares, respectively, whereas those in the individual preparations are shown by closed circles. Solid and broken line represent the predictions by the empirical scaling factor (average SFmean in Table 1) and biologically based scaling factor (SFbiol, see Text), respectively, from 10 cryopreserved preparation.

The averaged results from 10 or more preparations of human hepatocytes appeared to provide more reliable predictions for the human liver metabolism, whereas it was less convenient and cost-effective. It was found that the pooled preparation of two lots (lot 70 and 73) achieved the same extent of predictability for all seven standard compounds as the averaged results from 10 individual preparations (Tables 1 and 2). Pooled preparation demonstrated that the metabolic activity was constant during a 2-h incubation time period for standard compounds (Fig. 2). The in vitro- in vivo-correlation study was further extended to another seven compounds by using pooled cryopreserved preparation of human hepatocytes (Table 2). The predictions of CL_{H, predicted, 70+73} and F_{H, predicted, 70+73} were carried out with the average value of scaling factor (average SF₇₀₊₇₃, 10.8 × 10⁹ cells/kg) for total 14 compounds, which was obtained empirically as described for 7 standard compounds in the pooled hepatocyte preparation of lot 70 and 73. Reasonably good correlations were obtained for both oral bioavailability (Fig. 3, panel A) and hepatic clearance (Fig. 3, panel B). These data demonstrated that the in vitro metabolic clearance obtained in the pooled preparation from cryopreserved human hepatocytes reasonably well predicted in vivo hepatic clearance and availability with the aid of empirical scaling factor.

View larger version (22K): [in this window] [in a new window]   Fig. 2.   Disappearance curves of seven standard compounds in the pooled human hepatocytes preparation (lot 70 and 73) suspended in serum. Each symbol represents the mean of three determinations. The numbers correspond to the compound numbers listed in Table 1.

View larger version (15K): [in this window] [in a new window]   Fig. 3.   Prediction of oral bioavailability (A) and hepatic clearance (B) using the cryopreserved preparation of human hepatocytes pooled from two separate lots. The hepatocyte suspension was prepared by combining equal volumes of cryopreserved preparations of human hepatocytes from lot 70 and 73. Each symbol represents the mean ± S.D. of three determinations of predicted values from cryopreserved preparations (x-axis) and those of observed values taken from the literature (y-axis). Numbers correspond to the compound numbers listed in Table 2. Predictions (FPO, predicted, 70+73 and CLH, predicted, 70+73) were carried out as described under the Materials and Methods section by using the average of empirical scaling factor (average SF70+73) of 10.8 × 109 cells/kg (Table 2).

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Isolated hepatocytes have been recognized as more in vivo relevant in vitro systems than the subcellular fractions such as liver S-9 and microsomes for the prediction of in vivo metabolism. Although the freshly isolated human hepatocytes appeared to be one of the best preparations for the prediction of in vivo metabolism in humans (Lavé et al., 1999), the cryopreserved human hepatocytes instead became more prevalent and widely used for the routine analysis (Li, 2001). Cryopreserved human hepatocytes have been reported to quantitatively retain most of the phase I metabolic activities observed in the fresh liver, whereas some phase II metabolic activities to certain substrates were lower in the cryopreserved preparation than the intact human liver (Li et al., 1999; Steinberg et al., 1999; Hengstler et al., 2000; Rialland et al., 2000). In addition, consistent with the fact that each drug metabolizing enzyme activity in the human liver is known to individually vary between subjects, an approximately 3- to 5-fold variation was found in the in vitro metabolic clearance for the standard compounds among preparations from different human subjects (Table 1; Fig. 1). The interindividual variation in the metabolic capacity in the liver appears to reflect the observed large variation in the clearance in humans.

The empirical scaling factors for the in vitro-to-in vivo extrapolation (8.5 × 10⁹ cells/kg from 10 individual preparations in Table 1; 10.8 × 10⁹ cells/kg from pooled preparation in Table 2) were approximately 3 to 4 times larger than the anatomically calculated value (3.1 × 10⁹ cells/kg). In addition, the variation of scaling factor obtained from 10 individual preparations (SF_mean in Table 1) between different compounds was much larger than that obtained from the pooled-cryopreserved preparation of human hepatocytes (SF₇₀₊₇₃ in Table 2). These data suggested that the empirical scaling factor applicable only to the preparation used in the prediction was critically important for more reliable and rational predictions, which might compensate the inherent interindividual variation and/or loss of metabolic activities among different cryopreserved preparations.

In summary, the present study demonstrates that the direct evaluation of metabolic clearance in cryopreserved human hepatocytes in the presence of human serum was a convenient and useful tool for the prediction of hepatic clearance and availability. The calibration paradigm described in this report minimized the interindividual variation of metabolic activities among different subjects and improved the predictability of the in vitro data for the in vivo metabolic clearance with the aid of empirical scaling factor. The present method could be helpful at the early discovery stage to identify more promising can- didates for further development that have lower hepatic clearance and higher oral bioavailability in humans.