Abstract
In the present study, the beagle dog was evaluated as a preclinical model to investigate organic anion transporting polypeptide (OATP)–mediated hepatic clearance. In vitro studies were performed with nine OATP substrates in three lots of plated male dog hepatocytes ± OATP inhibitor cocktail to determine total uptake clearance (CLuptake) and total and unbound cell-to-medium concentration ratio (Kpuu). In vivo intrinsic hepatic clearances (CLint,H) were determined following intravenous drug administration (0.1 mg/kg) in male beagle dogs. The in vitro parameters were compared with those previously reported in plated human, monkey, and rat hepatocytes; the ability of cross-species scaling factors to improve prediction of human in vivo clearance was assessed. CLuptake in dog hepatocytes ranged from 9.4 to 135 µl/min/106 cells for fexofenadine and telmisartan, respectively. Active process contributed >75% to CLuptake for 5/9 drugs. Rosuvastatin and valsartan showed Kpuu > 10, whereas cerivastatin, pitavastatin, repaglinide, and telmisartan had Kpuu < 5. The extent of hepatocellular binding in dog was consistent with other preclinical species and humans. The bias (2.73-fold) obtained from comparison of predicted versus in vivo dog CLint,H was applied as an average empirical scaling factor (ESFav) for in vitro–in vivo extrapolation of human CLint,H. The ESFav based on dog reduced underprediction of human CLint,H for the same data set (geometric mean fold error = 2.1), highlighting its utility as a preclinical model to investigate OATP-mediated uptake. The ESFav from all preclinical species resulted in comparable improvement of human clearance prediction, in contrast to drug-specific empirical scalars, rationalized by species differences in expression and/or relative contribution of particular transporters to drug hepatic uptake.
Introduction
For many acidic or zwitterionic drugs, transporter-mediated uptake clearance is an important contributor to hepatic disposition and can be a rate-determining process for drug hepatic clearance and corresponding drug-drug interactions (DDIs) (Gertz et al., 2013; Shitara et al., 2013; Zamek-Gliszczynski et al., 2013; Varma et al., 2015). The value of in vitro–derived transporter kinetic data within a physiologically based pharmacokinetic (PBPK) paradigm is now widely appreciated (Jones et al., 2015; Galetin et al., 2017; Yoshida et al., 2017; Guo et al., 2018), yet prediction success is still limited. Robust studies using preclinical animals to increase confidence in the subsequent application of in vitro–in vivo extrapolation (IVIVE) of human in vitro transporter data (De Bruyn et al., 2018) would help alleviate this shortcoming.
Recently, we demonstrated the utility of the cynomolgus monkey as a preclinical species for evaluation of organic anion transporting polypeptide (OATP)–mediated hepatic clearance and DDIs (De Bruyn et al., 2018; Ufuk et al., 2018). Despite an overall good relationship between in vitro–derived clearance in cynomolgus monkey hepatocytes and in vivo clearance in the same species, the underprediction trend was apparent. The bias correction noted for the cynomolgus monkey clearance prediction was subsequently applied as an empirical scaling factor (ESF) to improve the prediction of human hepatic clearance for the same OATP substrates from human hepatocytes. The success of the cross-species scaling approach based on cynomolgus monkey may be linked to the excellent agreement with humans in OATP protein sequence homology. This work was extended to include two widely used preclinical species, the beagle dog and Sprague-Dawley rat, for which OATP proteomics is less well defined but believed to show poor homology with humans.
An increasing number of studies have recently investigated hepatic uptake transporter–mediated clearance and DDIs in cynomolgus monkeys, both in vitro and in vivo (Shen et al., 2013; Chu et al., 2015; De Bruyn et al., 2018; Ufuk et al., 2018). In contrast, there is minimal information on or systematic evaluation of activity of hepatic transporters in beagle dogs (Wilby et al., 2011). Studies in rats have been limited to hepatocyte investigations, and predictions of in vivo clearance have been inconsistent (Huang et al., 2010; Wood et al., 2017). Recent protein quantification by mass spectrometry has revealed interspecies differences in absolute expression levels of hepatobiliary transporters in the liver and drug-metabolizing enzymes in various species (Heikkinen et al., 2015; Wang et al., 2015). Canine Oatp1b4 is the most abundant transporter and represents approximately half of the total abundance of hepatobiliary transporters in dog liver, in contrast to human OATP1B1 and OATP1B3, which contribute only approximately 29% to the total abundance of hepatic transporters expressed in human liver (Wang et al., 2015). Oatp1a1, Oatp1a4, and Oatp1b2 are major Oatp transporters expressed on the sinusoidal membrane of rat hepatocytes (Wang et al., 2015). In cynomolgus monkeys, Oatp1b1 and Oatp1b3 are mainly expressed in the liver (Wang et al., 2015), showing a high level of amino acid homology (>90%) between human and cynomolgus monkey transporter counterparts (Shen et al., 2013). In the case of beagle dogs, Oatp1b4 shows 68% and 71% homology at protein level to human OATP1B1 and OATP1B3, respectively (Gui and Hagenbuch, 2010).
To date, limited in vitro and in vivo data have been reported in the beagle dog for OATP-mediated hepatic clearances, and examples of IVIVE of transporter-mediated hepatic clearance in this preclinical species are few. Here, we report the in vitro characterization of hepatic uptake for nine known OATP substrates (namely, atorvastatin, cerivastatin, fexofenadine, pitavastatin, pravastatin, repaglinide, rosuvastatin, telmisartan, and valsartan) in plated male dog hepatocytes +/- cocktail of OATP inhibitors. Based on data in other preclinical species and humans, selected OATP substrates cover a range of low to high clearance drugs with different contributions of active versus passive processes to the overall hepatic clearance. In addition to in vitro studies, in vivo pharmacokinetic studies were conducted in beagle dogs following intravenous drug administration to evaluate hepatic clearance prediction by IVIVE. The use of dog ESFs to refine prediction of human clearance was explored, as recently reported for data obtained in cynomolgus monkey hepatocytes (De Bruyn et al., 2018). Dataset average and individual drug specific ESFs (ESFav and ESFsd, respectively) were derived from the relationship between in vivo and in vitro predicted clearance in beagle dog, were determined. A similar exercise was performed for the rat using previously published hepatocyte kinetic data derived under experimental conditions (Ménochet et al., 2012a; Cantrill and Houston, 2017) similar to those reported here for the dog studies. Furthermore, in vitro–derived parameters were compared across all preclinical species investigated. The extent of improvement in human clearance prediction based on application of ESFs from beagle dog was compared with the application of ESFs from monkey and rat data for the same data set of OATP substrates.
Materials and Methods
Chemicals.
1-Aminobenzotriazole (ABT), atorvastatin, fexofenadine, pravastatin, repaglinide, rifamycin SV, and sulfobromophthalein (BSP) were purchased from Sigma-Aldrich (Poole, UK). Atorvastatin lactone, repaglinide acyl-β-d-glucuronide, and telmisartan acyl-β-d-glucuronide were purchased from Toronto Research Chemicals Inc. (North York, Canada). Cerivastatin, pitavastatin, rosuvastatin, and valsartan were purchased from Sequoia Research Products (Pangbourne, UK). For in vivo studies, atorvastatin, telmisartan, and rosuvastatin were purchased from Thermo Fisher Scientific (Waltham, MA). Pitavastatin and repaglinide were purchased from Selleck Chemicals (Houston, TX). Fexofenadine and pravastatin were purchased from TCI America (Portland, OR). Valsartan was purchased from Sigma-Aldrich (St. Louis, MO). Cerivastatin was purchased from Ochem Incorporation (Des Plaines, IL).
Hepatocyte Uptake Studies.
Cryopreserved male beagle dog hepatocytes (lots XVD, XZG, and YHF, all single donors) were purchased from BioIVT (Baltimore, MD). Cryopreserved dog hepatocytes were thawed according to the manufacturer’s standard protocol, and 0.5 ml of suspended hepatocytes (0.7 × 106 viable cells/ml) was added to each well of collagen I-coated BioCoat 24-well plates (BD Biosciences, Bedford, MA). After 4-hour culturing in a CO2 incubator, the medium was discarded, and the cell monolayers were preincubated with Dulbecco’s phosphate-buffered saline (DPBS) containing 1 mM ABT +/- OATP inhibitor cocktail (100 µM rifamycin SV and 50 µM BSP) for 30 minutes. ABT was used as a pan-inhibitor to inactivate cytochrome P450 (P450) activities in dog hepatocytes. Uptake data in the presence of OATP inhibitor cocktail were used to determine the passive diffusion clearance (CLpassive) value. This approach was based on preliminary data that showed more pronounced inhibition of uptake of the prototypical OATP probes pitavastatin and repaglinide in dog hepatocytes compared with the use of rifamycin SV or BSP alone (data not shown). Subsequently, uptake was started by adding fresh DPBS containing OATP substrate at a concentration of 0.5 µM (with the exception of pravastatin, 5 µM) +/- OATP inhibitor cocktail and over 2 min at 37°C. The buffer was collected, and 200 µl of water was added to lyse the cells after washing the cells with ice-cooled DPBS three times. All uptake experiments were performed in triplicate. In addition, extended uptake studies over 90 minutes (with up to eight time points) were performed to reach equilibrium and determine the total cell-to-medium drug concentration ratio (Kp); the same low drug concentration was used as in shorter incubations. Any potential cell loss was accounted for by measuring protein concentrations in cell lysates using the bicinchoninic acid assay according to the manufacturer’s protocol (Life Technologies Ltd., Paisley, UK). Drug concentrations in cell lysate and medium were quantified by liquid chromatography with tandem mass spectrometry (LC-MS/MS); conditions are detailed in Supplemental Table S1. In addition, the metabolism of atorvastatin to atorvastatin lactone (Prueksaritanont et al., 2002) and repaglinide and telmisartan to their respective acyl-glucuronides (Gill et al., 2012; Säll et al., 2012) was monitored.
Determination of Plasma Protein Binding.
Fraction unbound in dog plasma (fup) was determined for all compounds via equilibrium dialysis at a concentration of 1 μM. In brief, presoaked dialysis membranes were placed in a 96-well microequilibrium dialysis device, then 125 μl of plasma spiked with compound was loaded opposite to 100 μl of 100 mM sodium phosphate buffer (pH 7.4) in triplicate. Plates were sealed and incubated at 37°C with 5% CO2 for 4 hours, shaking at 200 rpm, after which 25 μl of plasma/buffer was taken and mixed with 25 μl of buffer/plasma. The 50 μl samples were then quenched with 150 μl of acetonitrile containing internal standard for the individual compound, and concentrations were quantified via LC-MS/MS, as described in Supplemental Table S2. Recovery over the 4-hour period was tested for all compounds and ranged from 97% to 108%.
In Vivo Studies.
In vivo study protocol was reviewed and approved by the Institutional Animal Care and Use Committee. Male beagle dogs (n = 3/study arm) weighed 8–15 kg at the time of study conduct and were fasted overnight prior to being administered compound. All compounds were intravenously administered at 0.1 mg/kg as a 20% captisol solution and were given over 30 minutes via a temporary catheter inserted into the cephalic vein. Blood samples were taken from the jugular vein at 0.25, 0.42, 0.58, 0.75, 1, 1.5, 2.5, 4.5, 8.5, 12.5, and 24.5 hours following administration. Blood samples were collected in tubes containing EDTA and centrifuged to collect plasma. Studies were carried out in cages with plexiglass surroundings for collection of urine from 0 to 4, 4 to 8, 8 to 12, and 12 to 24 hours. Additional blood samples were taken at predose, 4, 8, 12, and 24 hours for measurement of plasma creatinine. Creatinine was also measured in an aliquot from each urine collection interval. Plasma and urine samples for determination of compound concentrations were stored at ≤−20°C until analysis via LC-MS/MS, as described in Supplemental Table S2.
Analysis of In Vitro Hepatocyte Data.
In vitro uptake rates (pmol/min/mg protein) were calculated from the slopes of the initial uptake rate-time profile in the absence and presence of OATP inhibitor cocktail; the rates were then divided by the initial substrate concentration to determine total uptake clearance (CLuptake, µl/min/mg protein) and CLpassive (µl/min/mg protein), respectively (Yabe et al., 2011). Passive influx diffusion clearance was assumed to be equal to the passive efflux, and effect of membrane potential (Yoshikado et al., 2017) was not considered. Adsorption of the drug to the plate in the absence of cells was <7% across all substrates investigated, regardless of the presence of OATP inhibitor cocktail. In vitro hepatic active uptake clearance (CLactive) was calculated by subtracting the CLpassive from the CLuptake (Yabe et al., 2011). The conversion of CLuptake, CLpassive, and CLactive to µL/min/106cells was carried out using 1.2 mg of protein/106 dog hepatocytes based on in-house data generated using the bicinchoninic acid assay (data not shown). The CLpassive was log-transformed and compared with the log D7.4 values of the drugs investigated, taken from a previous study (Yabe et al., 2011). The Kp parameter, which represents intracellular binding in addition to active uptake processes, was calculated from eq. 1:
(1)where Ccell and Cmedium represent drug concentrations in cells and medium, respectively. The hepatocyte volume was set to 3.9 µl/106 hepatocytes with an assumption of the same volume as reported in rats (Reinoso et al., 2001). The cell-to-medium concentration ratio for unbound drug (Kpuu) was calculated from eq. 2 (Yabe et al., 2011; Shitara et al., 2013):
(2)where CLint represents intrinsic clearance for either metabolism and/or biliary excretion. Cerivastatin, fexofenadine, pitavastatin, pravastatin, rosuvastatin, and valsartan are metabolically stable or metabolized mainly by P450. Metabolic clearance was assumed negligible for these drugs because ABT, a pan-P450 inhibitor, was included in the incubations. The biliary excretion of all the drugs investigated was considered negligible under the current experimental setup due to internalization of efflux transporters following short-term culturing (Bow et al., 2008). The formation of atorvastatin lactone, repaglinide, and telmisartan glucuronides was not negligible under prolonged incubation times used to determine the Kp. Therefore, intrinsic metabolic clearance (CLint,met) was calculated from the slope of the linear phase of metabolite formation over time, and these CLint,met values were used to calculate the Kpuu of atorvastatin, repaglinide, and telmisartan using eq. 2. The fraction unbound in the cell (fucell) was calculated indirectly using eq. 3 (Yabe et al., 2011):
(3)In addition to the analysis of the initial uptake rate data, the fucell, CLactive, and CLpassive were determined by simultaneous fitting of uptake data over an extended time course in the absence and presence of OATP inhibitor cocktail. An adaptation of the mechanistic two-compartment model (Ménochet et al., 2012b) in MATLAB (version 8.5.1, 2015; MathWorks, Natick, MA) was applied to estimate CLactive, CLpassive, and fucell under the assumption that CLactive in eqs. 4 and 5 approaches zero in the presence of transporter inhibitors. As a proof of concept, this method was only applied for six drugs that do not undergo metabolism under the experimental conditions used (cerivastatin, fexofenadine, pitavastatin, pravastatin, rosuvastatin, and valsartan) and using lot XVD of dog hepatocytes:
(4)
(5)where Scell and Smed represent the intracellular and media drug concentrations, respectively; and Vcell and Vmed are the intracellular and media volumes, which were set at 3.9 and 400 µl, respectively.
Analysis of In Vivo Dog Data.
In vivo data were analyzed using Watson LIMS 7.5 (Thermo Fisher Scientific). Systemic in vivo plasma clearance (CLtotal) was determined using eq. 6:
(6)where AUC0–∞ represents the extrapolated area under the plasma concentration-time curve. For compounds with measurable urinary excretion, renal clearance (CLR) was determined with eq. 7:
(7)where Ae0–24 and AUC0-24 represent the amount excreted in the urine and the area under the plasma concentration-time curve from 0 to 24 hours, respectively. To account for possible incomplete urine collection, CLR was corrected for recovery of creatinine in the urine. CLR for creatinine was calculated as shown in eq. 7, and corrected CLR for drugs investigated was obtained as shown in eq. 8:
(8)where GFR represents reported glomerular filtration rate in dogs of 3.2 ml/min/kg (Mahmood, 1998).
In addition to creatinine, plasma iohexol clearance was measured, and values obtained were comparable to creatinine clearance (data not shown). In vivo hepatic clearance (CLH) was calculated by subtracting CLR from CLtotal.
Extrapolation of In Vitro Hepatic Uptake Clearance to In Vivo.
The IVIVE method was based on the overall uptake clearance parameter, and therefore, net effect of multiple processes was captured. Dog in vitro CLuptake µL/min/106cells was scaled by a hepatocellularity value of 175 × 106 cells/g of liver and average dog liver weight of 32 g of liver/body weight. A median value of hepatocellularity values reported in the literature for beagle dogs was used (details of data collation in Supplemental Table S3). In vivo intrinsic hepatic clearance (CLint,H; mL/min/kg) was obtained by applying the well stirred model (eq. 9):
(9)where QH represents average hepatic blood flow in beagle dog of 40 ml/min/kg (summary of individual studies reported in the literature is in Supplemental Table S3), and RB represents blood-to-plasma ratio. For atorvastatin, the CLH corrected by the RB value was greater than QH; therefore, the blood-based CLH,B (CLH/RB) value was capped at 90% of canine hepatic blood flow (36 ml/min/kg). Differences in CLint,H predicted from in vitro data relative to in vivo CLint,H and the corresponding precision of the prediction were assessed by geometric mean fold error (gmfe, eq. 10) and the root mean square error (eq. 11) (Gertz et al., 2010), respectively:
(10)
(11)where N indicates the number of drugs included in the analysis.
To evaluate whether dog as a preclinical species can improve prediction of hepatic uptake transporter–mediated clearance in humans, the gmfe obtained from dog IVIVE was applied as an ESFav in human predictions. In addition to ESFav, the performance of human clearance prediction was assessed by using drug-specific ESF obtained from the ratio of observed to predicted clearance in dog for each individual drug (ESFsd) (Naritomi et al., 2001; Ito and Houston, 2005; De Bruyn et al., 2018). In addition to dog, the same strategy was applied using scalars obtained from the IVIVE of monkey and rat in vitro data obtained in plated hepatocytes and for the same set of OATP substrates. For monkey, the ESF values were taken from a previous publication (De Bruyn et al., 2018). In the case of rat, the ESFs were calculated from previously reported in vitro data in rat hepatocytes (Ménochet et al., 2012a; Cantrill and Houston, 2017) and literature-collated in vivo clearance values (details listed in Supplemental Table S4). The direct IVIVE of human clearance was performed using the mean of in vitro data from four donors of human hepatocytes for which OATP1B1 c.521T>C genotype information was not known (Ménochet et al., 2012b; De Bruyn et al., 2018). Considering variability in transporter expression and/or activity in different lots of human hepatocytes, in vitro data from multiple donors were included in the evaluation. To assess the ability of preclinical species to refine prediction of human CLint,H, the predicted CLint,H values in humans for nine OATP substrates were multiplied by either ESFav or ESFsd obtained from rat, dog (excluding bosentan, as data were not available), and monkey.
Results
Uptake Parameters in Dog Hepatocytes.
A time-dependent increase in intracellular accumulation was observed for all nine drugs investigated. The mean uptake parameters obtained for individual drugs in three lots of dog hepatocytes are listed in Table 1; parameters obtained in each individual lot are summarized in Supplemental Table S5. A 15-fold range in CLuptake was observed, with cerivastatin showing the highest CLuptake (143 ± 57 µl/min/106 cells), followed by repaglinide, telmisartan, pitavastatin, and atorvastatin (CLuptake > 100 µl/min/106 cells). In contrast, uptake was more than 10-fold lower in the case of fexofenadine and pravastatin (CLuptake < 10 µl/min/106 cells). Drugs in the current data set showed a 35-fold range in CLpassive; cerivastatin and pravastatin showed the highest and lowest CLpassive, respectively. The formation of non-P450 metabolites was minimal over a short incubation time, and rates of metabolism represented 0.4%, 6.6%, and 9.9% of uptake rates of parent drugs for atorvastatin lactone, repaglinide glucuronide, and telmisartan glucuronide, respectively. Therefore, CLint,met was not accounted for in the calculation of CLactive. Overall, OATP substrates investigated in the current study showed >65% contribution of the active transport to total uptake, with 5/9 drugs having >75% of active contribution (Table 1). The CLactive values of pitavastatin, pravastatin, rosuvastatin, and valsartan were previously reported using freshly isolated male dog hepatocytes in suspension with the temperature method (37°C vs. 4°C) (Wilby et al., 2011). Their CLactive values were similar to those obtained in the present study, suggesting only a marginal effect of cryopreservation and plated format on hepatic uptake transporter activities in dog hepatocytes despite a difference in donors. Comparison of the CLuptake, CLpassive, and CLactive values obtained in the individual donors of dog hepatocytes resulted in the overall good agreement (Supplemental Table S5), with most of the parameter values within 2-fold between individual donors with the exception of fexofenadine, where CLuptake, CLpassive, and CLactive showed differences across donors resulting in large CV on those parameters (58%–93%). In addition to fexofenadine, pravastatin CLpassive showed a more than 3-fold difference among donors.
Uptake parameters of nine drugs investigated in dog hepatocytes
The in vitro uptake parameters of eight drugs (cerivastatin, fexofenadine, pravastatin, pitavastatin, repaglinide, rosuvastatin, telmisartan, and valsartan) in dog hepatocytes were compared with those reported in human hepatocytes under similar experimental conditions (De Bruyn et al., 2018) (Fig. 1). In each case, a comparison was made between scaled parameters (expressed per gram of liver) due to differences in hepatocellularity between dog and human. The CLuptake values obtained in dog hepatocytes were in good agreement with values in human hepatocytes, whereas the correlation of the CLpassive and CLactive values was less marked. Cerivastatin and telmisartan CLpassive values in human hepatocytes were approximately 3-fold greater than values obtained in dog hepatocytes, whereas opposite trends were seen for pravastatin and valsartan (Table 1). In the case of CLactive, values for pravastatin, cerivastatin, fexofenadine, and valsartan were approximately 2- to 5-fold greater in dog hepatocytes.
Comparison of uptake parameters between humans and dogs. CLuptake (A), CLpassive (B), and CLactive (C) in plated dog hepatocytes scaled to per gram of liver were compared with those in plated human hepatocytes (Ménochet et al., 2012b; De Bruyn et al., 2018). Data represent the mean ± S.D. of n = 3 dog hepatocyte donors. In human hepatocytes, data represent the mean ± S.D. of n = 4 donors (details in Supplemental Table S9 footnote). The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, cerivastatin; 2, fexofenadine; 3, pitavastatin; 4, pravastatin; 5, repaglinide; 6, rosuvastatin; 7, telmisartan; 8, valsartan.
Kp Parameters in Dog Hepatocytes.
The Kp profiles over 90 minutes were investigated for all nine drugs in three donors of plated dog hepatocytes. Most of the drugs investigated reached steady state within 30 minutes except for fexofenadine and valsartan (Supplemental Fig. S1). The mean Kp parameters from three lots are shown in Table 2, and values obtained in each individual lot are summarized in Supplemental Table S6. A 28-fold range in mean Kp values was observed for the current data set, with values ranging from 13.6 ± 2.4 (pravastatin) to 375 ± 117 (atorvastatin). For six of the nine drugs, the Kp values were >100, with the exception of rosuvastatin, fexofenadine, and pravastatin. Analogous to the trends in uptake parameters, an overall good agreement was seen in Kp values obtained in different donors (within 3-fold difference); fexofenadine and pravastatin were again the outliers (Supplemental Table S6).
Kp parameters of nine drugs investigated in dog hepatocytes
Data represent the mean ± S.D. from three lots of dog hepatocytes. The CLint,met was determined from the slope of metabolite formation over time and was assumed to be equal to the CLint in eq. 2.
UDP-glucuronosyltransferase–mediated CLint,met values of atorvastatin, repaglinide, and telmisartan determined in the extended Kp experiments were at least 15% of their respective CLpassive values. Therefore, CLint,met was considered when calculating the Kpuu for these three drugs (eq. 2). A 16-fold range was observed in Kpuu values among nine drugs. The Kpuu values of rosuvastatin and valsartan were >10, whereas cerivastatin, pitavastatin, repaglinide, and telmisartan showed Kpuu values <5. There was a large range in fucell (122-fold) in dog hepatocytes; high intracellular binding was observed for atorvastatin, cerivastatin, pitavastatin, repaglinide, and telmisartan (fucell < 0.05). In addition, CLactive, CLpassive, and fucell for six drugs were estimated by simultaneous fitting of uptake data (single low drug concentration) over an extended time course ± OATP inhibitor cocktail. The estimates obtained by the mechanistic two-compartment model (Supplemental Table S7) were comparable to those calculated in the two-step analysis of data from short incubation and Kp experiments. The fitting of the mechanistic model to drug cell concentrations versus time (± inhibitor) as well as goodness-of-fit plots are illustrated for rosuvastatin as a representative drug in Supplemental Fig. S2. The CLpassive and fucell values for the nine drugs investigated were strongly correlated with the respective log D7.4 (Supplemental Fig. S3, A and B); a relationship was also noted between the extent of intracellular binding in dog hepatocytes and fup for the drugs investigated (Supplemental Fig. S3C).
Species Differences in Uptake Parameters.
The CLuptake, CLpassive, and CLactive in dog hepatocytes were compared with the previously reported values in Sprague-Dawley rat, cynomolgus monkey, and human (Ménochet et al., 2012a,b; Cantrill and Houston, 2017; De Bruyn et al., 2018) (Fig. 2). Rat hepatocyte CLuptake and CLactive values were generally in good agreement with those obtained in dogs, with rosuvastatin being an outlier in both cases (Fig. 2, A and E). Rat hepatocyte CLpassive values were similar to those in dog hepatocytes for cerivastatin, rosuvastatin, and telmisartan but were up to 6-fold smaller for the remaining drugs (Fig. 2C). Monkey hepatocyte total CLuptake data showed a good agreement with dog hepatocyte data (Fig. 2B). In contrast to rat, CLpassive values obtained in monkey hepatocytes were either similar to the data in dog (fexofenadine, pitavastatin, and pravastatin) or almost 6-fold greater for the remaining drugs (Fig. 2D). An opposite trend was seen for CLactive in which data were either comparable between the two species (rosuvastatin, telmisartan, and valsartan) or up to approximately 6-fold smaller for the remaining drugs in monkey hepatocytes (Fig. 2F).
Comparison of uptake parameters among species. CLuptake (A and B), CLpassive (C and D), and CLactive (E and F) in plated dog hepatocytes were compared with those in rats (A, C, and E) (Ménochet et al., 2012a) and cynomolgus monkeys (B, D, and F) (De Bruyn et al., 2018). The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, cerivastatin; 2, fexofenadine; 3, pitavastatin; 4, pravastatin; 5, repaglinide; 6, rosuvastatin; 7, telmisartan; 8, valsartan.
In Vivo Studies and Extrapolation of Dog In Vitro Transporter Data to In Vivo.
The pharmacokinetic studies for the nine drugs investigated were conducted following a single intravenous infusion over 30 minutes to three male beagle dogs. The in vivo parameters obtained are shown in Table 3. The CLtotal values ranged from 1.48 ± 0.47 ml/min/kg (repaglinide) to 48.4 ± 13.3 ml/min/kg (atorvastatin). Hepatic clearance was the major elimination mechanism in beagle dog for the drugs investigated, as CLR contributed <40% to the CLtotal. Six drugs (cerivastatin, fexofenadine, pitavastatin, pravastatin, repaglinide, and valsartan) were classified as low clearance drugs in beagle dogs, as their hepatic blood clearance (CLH,B) was <30% of hepatic blood flow.
Pharmacokinetic parameters of nine drugs investigated following intravenous administration in dogs
Data represent the mean ± S.D. from three male dogs intravenously infused. CLH was determined by subtracting CLR from CLtotal.
The prediction of hepatic clearance in dogs is shown in Fig. 3. Good agreement between predicted and observed CLint,H was observed, with a gmfe of 2.73 and 55% of values predicted within 2-fold of the observed data. Subsequently, the gmfe obtained in dog IVIVE for this data set was applied to human IVIVE as an ESFav with the aim to assess whether information obtained in dog as a preclinical species can improve prediction of hepatic uptake transporter–mediated clearance in humans. In addition to dog, a similar exercise was carried out using ESFav obtained from IVIVE of rat and monkey in vitro data obtained for the same set of OATP substrates (Table 4). The in vitro and in vivo data of nine drugs investigated in rats, monkeys, and humans are summarized in Supplemental Tables S4, S8, and S9, respectively.
Correlation of predicted and observed CLint,H in dogs. Predicted CLint,H values were compared with in vivo CLint,H in dogs. Predicted and observed data represent the mean ± S.D. of n = 3. The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, atorvastatin; 2, cerivastatin; 3, fexofenadine; 4, pitavastatin; 5, pravastatin; 6, repaglinide; 7, rosuvastatin; 8, telmisartan; 9, valsartan.
Statistical data comparing the accuracy and precision of the use of species-related ESFs to predict human CLint,H
Direct method involved no use of ESFs, whereas ESFav and ESFsd indicate the use of drug set average and individual drug specific ESFs, respectively.
The direct prediction of human clearance from in vitro data (no ESF) resulted in underprediction of in vivo CLint,H (3.22-fold bias, Fig. 4; Table 4). The improvement in prediction of human CLint,H observed using ESFav was comparable across the three preclinical species, resulting in approximately 2-fold bias and increased precision (Fig. 5, A, C, and E; Table 4). This lack of difference in the prediction success among the ESFav from the three preclinical species is also reflected in the residual plots (Figs. 4C and 6). In contrast, the use of individual drug-specific scaling factors (ESFsd) resulted in species-selective effects on the predictive performance (Fig. 4D). There was no improvement in the prediction bias of human clearance using rat ESFsd (gmfe 3.41), with <25% of the drugs falling within the 2-fold error (Fig. 5D; Table 4). Use of ESFsd obtained from dog and monkey data resulted in comparable bias (3.23- and 2.96-fold, respectively) with minor improvement in human clearance prediction. A higher proportion of drugs predicted within the 2-fold error was evident when monkey ESFsd was applied (67%) relative to dog ESFsd (25%) (Fig. 5B and 5F, Fig. 6; Table 4).
Correlation of predicted and observed CLint,H in humans in the absence of an empirical scaling factor (A, B), and direct comparison of the performance of ESFs from rat, dog, and monkey in improving human CLint,H prediction (C, D). Predicted CLint,H values were compared with observed human CLint,H (A) and to precision error expressed as the log of predicted/observed CLint,H ratio either in the absence (B) or presence of drug set average (ESFav) (C) and individual drug-specific (ESFsd) (D) empirical scaling factors. Human hepatocyte data represent the mean ± S.D. of n = 4 donors (details in Supplemental Table S9 footnote). Error bars in (C and D) were excluded for clarity and distinction of preclinical species. The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, cerivastatin; 2, fexofenadine; 3, pitavastatin; 4, pravastatin; 5, repaglinide; 6, rosuvastatin; 7, telmisartan; 8, valsartan; 9, bosentan.
Correlation of predicted and observed CLint,H in humans using average and drug-specific empirical scaling factors obtained from rat, dog, and monkey. Predicted human CLint,H values were compared with in vivo human CLint,H following application of drug set average empirical scaling factors (ESFav) (A, C, and E), or individual drug-specific empirical scaling factors (ESFsd) (B, D, and F) from dog (A and B), rat (C and D), and monkey (E and F). Predicted and observed CLint,H values in rats, monkeys, and humans were previously reported (Supplemental Tables S4, S8 and S9, respectively). Human hepatocyte data represent the mean ± SD of n = 4 donors (details in Supplemental Table S9 footnote). The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, cerivastatin; 2, fexofenadine; 3, pitavastatin; 4, pravastatin; 5, repaglinide; 6, rosuvastatin; 7, telmisartan; 8, valsartan; 9, bosentan.
Residual plots for the prediction of in vivo CLint,H (mL/min/kg) in humans. Predicted CLint,H values were plotted against the precision error (log of predicted/observed CLint,H ratio) when using data set average empirical scaling factors (ESFav) (A, C, and E) and individual drug-specific empirical scaling factors (ESFsd) (B, D, and F) from dog (A and B), rat (C and D), and monkey (E and F). The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, cerivastatin; 2, fexofenadine; 3, pitavastatin; 4, pravastatin; 5, repaglinide; 6, rosuvastatin; 7, telmisartan; 8, valsartan; 9, bosentan.
Species Differences in Kp Parameters.
Large interspecies differences were apparent between Kp and Kpuu (rank order of rat > dog > monkey seen for both parameters) (Fig. 7). Rat hepatocyte Kp and Kpuu values were up to 23- and 7.4-fold larger than those in dog hepatocytes, respectively (Fig. 7, A and D). The exceptions were repaglinide, telmisartan, and valsartan Kp and cerivastatin and fexofenadine Kpuu, which were comparable between the two species. In contrast, dog Kp and Kpuu were comparable or greater than the data obtained in monkey hepatocytes, with the exception of telmisartan Kpuu (Fig. 7, B and E). Differences in fexofenadine, pravastatin, and valsartan were particularly marked, as dog Kp parameters were up to 6.2-fold greater than values obtained in monkey. The overall trends seen between dog and monkey Kp parameters were also evident in the comparison of dog and human Kp and Kpuu, with the exception of valsartan. Even though Kp and Kpuu showed species-dependent values, there was a good agreement in intracellular binding parameter across species. This trend was particularly strong between dog and monkey fucell data (1.76-fold bias), whereas 2.3- and 2.6-fold differences were seen in dog fucell relative to rat and human data, respectively.
Comparison of Kp parameters among species. Kp (A–C), Kpuu (D–F), and fucell (G–I) in plated dog hepatocytes were compared with those in rats (A, D, and G) (Ménochet et al., 2012a), cynomolgus monkeys (B, E, and H) and humans (C, F, and I) (De Bruyn et al., 2018). The solid and dashed lines represent the line of unity and 2-fold difference, respectively. 1, cerivastatin; 2, fexofenadine; 3, pitavastatin; 4, pravastatin; 5, repaglinide; 6, rosuvastatin; 7, telmisartan; 8, valsartan.
Discussion
The utility of the cynomolgus monkey as a preclinical species for OATP-mediated hepatic clearance has recently been demonstrated, and a good relationship between in vitro–derived clearance in hepatocytes and in vivo clearance was observed (De Bruyn et al., 2018). As reported for metabolism-related hepatic clearance predictions from hepatocytes (and other in vitro systems), underprediction of transporter-mediated clearance also requires a bias correction to bridge the gap between extrapolated and observed values. This approach, applied retrospectively, is regarded as empirical and hence has no mechanistic basis. Recently, application of the bias correction observed in the IVIVE of cynomolgus monkey clearance improved the prediction of human hepatic clearance for the same OATP substrates from human hepatocytes. Therefore, an analogous approach was applied here to two widely used preclinical species, beagle dog and Sprague-Dawley rat. As a nonrodent preclinical species, beagle dogs are often used in pharmacokinetic studies; however, there is less information available to assess the predictive performance of hepatic transporter–mediated clearance and DDIs in this species compared with studies reported in cynomolgus monkeys (Shen et al., 2013; Chu et al., 2015; Watanabe et al., 2015; Ufuk et al., 2018). Therefore, characteristics of the hepatic uptake of a series of nine drugs were investigated to provide a data set of parameters for comparative purposes and evaluate beagle dogs as a preclinical animal model to study hepatic uptake.
In the present study, an OATP inhibitor cocktail was used to determine the CLpassive value in dog hepatocytes. Substrate-dependent inhibition has been reported for human OATP1B1, OATP1B3, and OATP2B1 (Noé et al., 2007; Izumi et al., 2015; Barnett et al., 2018) and rationalized by multiple binding sites for substrates and inhibitors of OATP transporters. Accordingly, this may also be the case for canine Oatp1b4, and the use of multiple inhibitors for OATPs is preferable to determine the involvement of OATPs in hepatic uptake. Contributions of active uptake obtained here in dog hepatocytes (Table 1) were in good agreement with estimates previously reported in human hepatocytes (Ménochet et al., 2012b), suggesting a minimal difference in active uptake contribution to cellular uptake between humans and dogs for OATP substrates. Some interspecies differences in CLpassive values were apparent, with up to 2.5-fold bias on average and an overall rank order of human ≥ monkey > dog > rat. These differences may be attributed to incomplete inhibition of active uptake in some species and/or the differences in the methodologies used to estimate this parameter [e.g., use of a single inhibitor (monkey and human), cocktail of inhibitors (dog), and mechanistic modeling (rat)]. However, for each species there was a good relationship with log D7.4, consistent with other reports based on data obtained in transfected cell lines (Li et al., 2014).
Determining the extent of intracellular binding of drug in hepatocytes has important implications on understanding pharmacokinetic/pharmacodynamic relationships, drug efficacy, and/or prediction of DDI risk (Zamek-Gliszczynski et al., 2013; Morse et al., 2015). Several experimental methods and in silico approaches to estimate intracellular drug concentration have been proposed (Chu et al., 2013; Guo et al., 2018), including the kinetic modeling used here. The fucell and CLpassive values obtained in plated dog hepatocytes for nine OATP substrates were strongly correlated with their log D7.4 (Supplemental Fig. S3), consistent with previous studies in both suspended and plated rat hepatocytes (Yabe et al., 2011; Ménochet et al., 2012a), as well as plated monkey and human hepatocytes (De Bruyn et al., 2018). This trend is expected, as fucell is inversely related to Kp when no active transport occurs. Estimation of hepatic fucell from the correlation to log D7.4 is well established and applicable for acidic compounds (Ménochet et al., 2012a; Chu et al., 2013) and provides a useful initial estimate of fucell in the case of limited data for implementation in the PBPK models. In addition, the current study showed a promising approach of simultaneous fitting of total uptake data ± OATP inhibitors over a longer time period (up to steady state) using a single low drug concentration which yielded fucell estimates comparable to the indirect two-step method of the estimation of this parameter via Kp and Kpuu.
Large interspecies differences were apparent in both Kp and Kpuu values, with the general rank order of rat > dog > cynomolgus monkey ≈ human for drugs investigated (Fig. 7, A–F). The Kpuu trends are not surprising, as this parameter reflects the interplay of active uptake and passive diffusion in addition to elimination processes (metabolism and biliary excretion). Interspecies differences in hepatic uptake clearance have been reported (Ménochet et al., 2012b; Watanabe et al., 2015) and correspond to some extent to differences in protein homology, but also to differences in absolute abundances of OATP/Oatp isoforms across species (Wang et al., 2015). Cynomolgus monkeys show 6- and 13-fold higher protein expressions of Oatp1b1 and Oatp1b3 compared with those in human OATP1B1 and OATP1B3, respectively, whereas comparable abundance is observed between dog Oatp1b4 and monkey Oatp1b1/Oatp1b3. In addition, the protein levels of rat Oatp1a1, Oatp1a4, and Oatp1b2 are lower than dog and monkey Oatps. In terms of the level of total protein (all OATP isoforms combined), the rank order is monkey > rat >> dog > human (Wang et al., 2015). Comparison of CLactive across species (Fig. 2, E and F) suggests that the interspecies differences of Kpuu cannot be explained solely by protein expression levels of OATP/Oatp in the liver, and that the interspecies differences in intrinsic transport affinity and capacity also need to be considered. Despite the discordance in Kp and Kpuu of drugs between hepatocytes from dog and other species, there was a good agreement in intracellular binding parameter fucell among dog, monkey, rat, and human (Fig. 7, G-I). These findings are in agreement with recent reports on correlations of fucell or fuliver between species for a diverse range of compounds (De Bruyn et al., 2018; Riccardi et al., 2018) and in relationships with log D7.4 as a species-independent parameter.
A 33-fold range was observed in CLH values in dog following a single intravenous infusion of nine drugs investigated; the CLtotal or CLH values obtained in the present study were consistent with previous reports (pitavastatin CLH, 6.8 ml/min/kg; pravastatin CLH, 9.0 ml/min/kg; telmisartan CLtotal, 6.75 ml/min/kg; valsartan CLH, 8.3 ml/min/kg) (Deguchi et al., 2011; Wilby et al., 2011). In vitro prediction in dogs resulted in a good agreement with in vivo values with a 2.73-fold bias, and the predicted CLint,H values of 5/9 drugs investigated within 2-fold of in vivo CLint,H (Fig. 3). This drug set average scaling factor obtained in dog improved human IVIVE performance by reducing the prediction bias (to 2.11-fold) and increasing precision. Use of ESFav from other species resulted in a similar prediction success (2-fold bias in human IVIVE when using monkey and rat ESFav). In contrast, use of ESFsd resulted in mixed success. Application of monkey ESFsd for human prediction resulted in a prediction bias of <3-fold, with 67% of drugs predicted within 2-fold of the in vivo CLint,H values. In contrast, both the prediction accuracy and the proportion of drugs within the same error threshold were lower using dog and rat drug-specific scalars (Figs. 4 and 5; Table 4). Although relatively smaller differences in drug-specific scaling factors (ratio of observed to predicted CLint,H) were observed between the species for a number of drugs (e.g., within 2.5-fold for repaglinide, telmisartan, and pitavastatin), the differences in ESFsd were more pronounced for pravastatin and cerivastatin (up to 27- and 4.6-fold larger in monkey and rat relative to the value obtained in dog, respectively). The present findings are based on a relatively small data set of well established OATP substrates; further studies with a larger number of transporter substrates, in particular with more challenging low-clearance OATP drugs and/or candidates with complex metabolism-transporter interplay, are required to establish wider application of the ESFsd method and confirm distinct trends between the preclinical species noted here. When using designated lots of human hepatocytes for assessment of new molecular entities, it would be prudent to evaluate uptake of a range of OATP substrates (low-medium-high clearance, different percentage of active vs. passive contribution to the overall uptake) to assess the need for ESF application from preclinical species.
In conclusion, the present study represents the most comprehensive assessment of OATP-mediated hepatic clearance performed to date in beagle dogs. In vitro uptake parameters obtained in this preclinical species showed low interindividual variability, with the exception of fexofenadine and pravastatin. Consistency in the extent of hepatocellular binding in three preclinical species and humans is encouraging and suggests that detailed mechanistic studies in preclinical species may be valuable to inform modeling of human hepatocyte data and subsequent PBPK model development. IVIVE of dog hepatocyte data resulted in a good agreement with the observed CLint,H (2.73-fold bias). The use of this value as an average empirical scaling factor improved human clearance IVIVE, suggesting utility of beagle dog as a preclinical model for the assessment of hepatic uptake mediated by OATPs. Use of dog ESFav resulted in success comparable to monkey in improving human IVIVE, in contrast to use of drug-specific scaling factors, rationalized by species differences in protein abundance and/or relative contribution of particular transporters to the overall hepatic uptake. Further investigations and expansion of the data set to include cationic drugs and assessment of DDIs associated with transporter-mediated uptake (in isolation or in conjunction with metabolism and biliary excretion) would be informative to strengthen the use of dog as a preclinical model for evaluation of transporter-mediated hepatic disposition.
Acknowledgments
We thank Dr. David Hallifax and Susan Murby for developing the methods of LC-MS/MS for the in vitro hepatocyte experiments.
Authorship Contributions
Participated in research design: Matsunaga, Morse, Hillgren, Hall, Houston, Galetin.
Conducted experiments: Matsunaga, Morse, Mohutsky, Bedwell, Bao.
Performed data analysis: Matsunaga, Ufuk, Morse.
Wrote or contributed to the writing of the manuscript: Matsunaga, Ufuk, Morse, Bedwell, Bao, Mohutsky, Hillgren, Hall, Houston, Galetin.
Footnotes
- Received August 24, 2018.
- Accepted December 17, 2018.
This work was supported by a consortium of pharmaceutical companies (GlaxoSmithKline, Lilly, and Pfizer) within the Centre for Applied Pharmacokinetic Research at the University of Manchester.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- ABT
- 1-aminobenzotriazole
- BSP
- sulfobromophthalein, CLactive, active uptake clearance
- CLint,H
- intrinsic hepatic clearance
- CLint,met
- intrinsic metabolic clearance
- CLpassive
- passive diffusion clearance
- CLuptake
- total uptake clearance
- DDI
- drug-drug interaction
- DPBS
- Dulbecco’s phosphate-buffered saline
- ESF
- empirical scaling factor
- ESFav
- average empirical scaling factor
- ESFsd
- drug-specific empirical scaling factor
- fucell
- fraction unbound in cells
- gmfe
- geometric mean fold error
- IVIVE
- in vitro–in vivo extrapolation
- Kp
- total cell-to-medium concentration ratio
- Kpuu
- cell-to-medium concentration ratio for unbound drug
- LC-MS/MS
- liquid chromatography with tandem mass spectrometry
- OATP/Oatp
- organic anion transporting polypeptide
- P450
- cytochrome P450
- PBPK
- physiologically based pharmacokinetic
- Copyright © 2019 by The American Society for Pharmacology and Experimental Therapeutics
