Abstract
Suspended, plated, or sandwich-cultured human hepatocytes are routinely used for in vitro to in vivo extrapolation (IVIVE) of transporter-mediated hepatic clearance (CL) of drugs. However, these hepatocyte models have been reported to underpredict transporter-mediated in vivo hepatic uptake CL (CLuptake,in vivo) of some drugs. Therefore, we determined whether transporter-expressing cells (TECs) can accurately predict the CLuptake,in vivo of drugs. To do so, we determined the uptake CL (CLint,uptake,cells) of rosuvastatin (RSV) by TECs (organic anion transporting polypeptides/Na+-taurocholate cotransporting polypeptide) and then scaled it to that in vivo by relative expression factor (REF) (the ratio of transporter abundance in human livers and TEC) determined by liquid chromatography tandem mass spectrometry–based quantitative proteomics. Both the TEC and hepatocyte models did not meet our predefined success criteria of predicting within 2-fold the RSV CLuptake,in vivo value obtained from our positron emission tomography (PET) imaging. However, the TEC performed better than the hepatocyte models. Interestingly, using REF, TECs successfully predicted RSV CLint,uptake,hep obtained by the hepatocyte models, suggesting that the underprediction of RSV CLuptake,in vivo by TECs and hepatocytes is due to endogenous factor(s) not present in these in vitro models. Therefore, we determined whether inclusion of plasma (or albumin) in TEC uptake studies improved IVIVE of RSV CLuptake,in vivo. It did, and our predictions were close to or just fell above our lower 2-fold acceptance boundary. Despite this success, additional studies are needed to improve transporter-mediated IVIVE of hepatic uptake CL of drugs. However, using REF and TEC, we successfully predicted the magnitude of PET-imaged inhibition of RSV CLuptake,in vivo by cyclosporine A.
SIGNIFICANCE STATEMENT We showed that the in vivo transporter-mediated hepatic uptake CL of rosuvastatin, determined by PET imaging, can be predicted (within 2-fold) from in vitro studies in transporter-expressing cells (TECs) (scaled using REF), but only when plasma proteins were included in the in vitro studies. This conclusion did not hold when plasma proteins were absent in the TEC or human hepatocyte studies. Thus, additional studies are needed to improve in vitro to in vivo extrapolation of transporter-mediated drug CL.
Introduction
In drug discovery, predicting human pharmacokinetics of a drug is critical in the selection of new molecular entities with desired pharmacokinetics properties. With increased understanding of the absorption, distribution, metabolism, and excretion processes, transporters are well recognized for their significant role in the absorption, distribution, metabolism, and excretion of drugs and their metabolites. For example, transporters play an important role in the hepatobiliary clearance (CL) and therefore drug-drug interaction (DDI) and hepatic concentrations of many drugs, including drug used for hypercholesterolemia, diabetes, and hepatitis C virus infection (Endres et al., 2006; Giacomini et al., 2010; Köck and Brouwer, 2012; Shebley et al., 2017).
Although in vitro and in silico methods (e.g., physiologically based pharmacokinetic models) have been developed to successfully predict hepatic metabolic CL of drugs (Sager et al., 2015), such methods have had limited success in predicting transporter-mediated CL and DDI. Currently, hepatocyte models, such as suspended hepatocytes (SHs) or plated hepatocytes (PHs), are used for in vitro to in vivo extrapolation (IVIVE) of transporter-mediated hepatic CL of drugs. However, these hepatocyte models are widely reported to underpredict the in vivo hepatic CL of some drugs (Abe et al., 2008; Li et al., 2010; Jones et al., 2012; Zou et al., 2013). In addition, they are costly and show considerable interlot variability in transporter activity (Vildhede et al., 2014; Izumi et al., 2017). To overcome these shortcomings, we have hypothesized that the in vivo transporter-mediated CL of drugs can be predicted by first measuring the transporter-mediated CL of the drug in transporter-expressing cells (TECs). Then, this CL can be scaled by relative expression factor (REF)—that is, the relative abundance of the transporter in cells versus human tissue as determined by liquid chromatography tandem mass spectrometry (LC-MS/MS)-based quantitative targeted proteomics. As a proof of concept, we have shown that this TEC/REF approach can successfully predict the renal secretory CL of metformin in humans (Kumar et al., 2018). Moreover, we have recently shown that this proteomics-informed REF approach can successfully predict the hepatic uptake CL of rosuvastatin (RSV) and the magnitude of its inhibition by rifampin in the rat (Ishida et al., 2018a,b).
Here we report a study to determine whether TEC/REF approach can successfully predict the in vivo hepatic uptake CL of RSV in humans that we have determined using PET imaging (Billington et al., 2019). As before, we set our success criteria as being able to predict the hepatic uptake CL of RSV within 2-fold of the observed value (Kumar et al., 2017; Ishida et al., 2018a). Since human hepatocytes are widely used to predict the hepatic uptake CL of drugs, we compared the predicted hepatic uptake CL of RSV using the TEC/REF approach with that based on traditional hepatocyte models, namely SHs, plated hepatocytes (PHs), or sandwich-cultured human hepatocytes (SCHHs) (scaled using milligram of protein per gram of liver) (Fig. 1). We also determined whether RSV uptake CL in SHs, PHs, or SCHHs can be predicted from transporter-expressing cells using the REF approach. The results of this aim would be informative in case neither approach (i.e., neither TECs nor human hepatocytes) successfully predicted the in vivo RSV hepatic uptake CL. The REF approach has traditionally been based on the total transporter abundance in both the transporter-expressing cells and in human tissue. However, only transporters present in the plasma membrane actively transport drugs. Therefore, we determined whether the use of plasma membrane abundance (PMA) versus total abundance of transporters improves the predictions of the REF approach. Recent studies have shown that plasma or albumin can increase the intrinsic hepatic uptake CL of drugs mediated by organic anion transporting polypeptides (OATPs) (Kim et al., 2019; Bowman et al., 2020). Therefore, we also determined whether the inclusion of plasma or albumin in our in vitro uptake studies using TEC improves the predictions of the TEC/REF approach. Last but not least, we determined whether the magnitude of inhibition of the RSV hepatic uptake CL by cyclosporine A (CsA) observed in our PET imaging study can be predicted by using the TEC/REF approach.
Schematic of our approach for IVIVE of RSV hepatic uptake CL using the proteomics-informed REF approach or the traditional physiologic scaling approach. HPPGL, total hepatic protein per gram of liver; WSHM, well-stirred hepatic model.
Materials and Methods
Chemicals and Reagents
Synthetic signature peptides for OATP1B1, OATP2B1, OATP1B3, and Na+-taurocholate cotransporting polypeptide (NTCP) were obtained from New England Peptides (Boston, MA). The corresponding stable isotope–labeled peptides for the above transporters, dithiothreitol, iodoacetamide, mass spectrometry–grade trypsin, Williams' E Medium (no glutamine), cryopreserved hepatocyte recovery medium, total protein quantification bicinchoninic acid assay (BCA) kit, Hank’s balanced salt solution (HBSS) with calcium and magnesium, HBSS without calcium and magnesium, human serum albumin (HSA), bovine serum albumin, and Pierce cell surface protein isolation kit were obtained from Thermo Scientific (Rockford, IL). Na+-free HBSS was prepared by replacing NaCl and NaHCO3 with choline chloride and potassium bicarbonate, respectively, from HBSS constituents. Pierce cell surface protein isolation kit contains sulfosuccinimidyl-2-(biotinamido) ethyl-1,3-dithiopropionate, quenching solution (100 mM glycine), lysis buffer, neutravidin agarose gel, wash buffer, column accessory pack, dithiothreitol, phosphate buffer, and Tris buffer. HPLC-grade acetonitrile and SDS were purchased from Fischer Scientific (Fair Lawn, NJ). Cyclosporine A and formic acid were purchased from Sigma-Aldrich (St. Louis, MO). All reagents were analytical grade. Twenty-four–well collagen-coated plates and Matrigel were purchased from Corning (Kennebunk, ME). Human hepatocyte thaw medium, INVITROGRO CP medium, INVITROGRO HI medium, and TORPEDO antibiotic mix were obtained from BioIVT (Westbury, NY). Radiolabeled [3H]RSV (>98% purity) and unlabeled RSV were purchased from American Radiolabeled Chemicals (Saint Louis, MO) and Toronto Research Chemicals (North York, ON, Canada), respectively. Human plasma was purchased from Bloodworks NW (Seattle, WA).
Procurement of Human Hepatocytes and Transporter-Expressing Cells
Human liver frozen tissues and cryopreserved human hepatocytes, which were obtained from the same donor, were obtained from BioIVT. The demographics of the human liver donors (n = 4, ADR, FEA, JEL, and YTW) were as described before (Kumar et al., 2019). OATP1B1-expressing CHO cells were generously provided by Dr. Bruno Stieger, University Hospital Zurich, Switzerland. OAPT1B1-expressing HEK293, OATP1B3-expressing HEK293, OATP2B1-expressing MDCKII, and NTCP-expressing HEK293 cells were generously provided by SOLVO Biotechnology, Hungary.
RSV Uptake Study in OATP1B1/2B1/1B3 or NTCP Transporter-Expressing Cells
[3H]RSV transport studies were conducted in OATP1B1-expressing CHO cells, OATP2B1-expressing MDCKII cells, OATP1B1-expressing HEK293 cells, OATP1B3-expressing HEK293 cells, and NTCP-expressing HEK293 cells. Transporter-expressing cells were grown in 24-well poly-D-lysine–coated plates at a density of 0.5 million cells per well with 1 ml of DMEM (CHO cells: low-glucose DMEM, HEK293 cells, and MDCKII cells: high-glucose DMEM) medium for 24 hours. After 24 hours, cells were washed twice with 1 ml/well DPBS buffer. Then, they were incubated at 37°C with Ca2+- and Mg2+-containing HBSS buffer (hereafter referred as HBSS buffer) containing 30 nM [3H]RSV and 70 nM unlabeled RSV with or without 200 µM bromsulphthalein (BSP) (OATP cells) or with HBSS or Na+-free HBSS buffer (NTCP cells) for 5–60 seconds (within linear range) (Bi et al., 2013). After incubation, the [3H]RSV solution was removed, and the cells were washed by ice-cold HBSS buffer three times (1 ml each). The uptake CL was calculated by the uptake at 5 seconds when the uptake was proportional to time. [3H]RSV uptake at 5 and 60 seconds (within linear phase) in the presence of 200 µM BSP and in the absence of sodium was used to estimate [3H]RSV passive diffusion clearance. Then, the cells were lysed at 37°C for 2 hours with 1 ml 2% SDS. Forty microliters of this lysate was used for total protein determination using the BCA Protein Assay Kit, and 700 µl was used to quantify total radioactivity by Tri-Carb Liquid Scintillation Counters (PerkinElmer, Waltham, MA). To determine the inhibitory effect of CsA, the [3H]RSV uptake study in the presence of 0.3 µM CsA was conducted in OATP1B1-, OATP2B1-, OATP1B3-, and NTCP-expressing cells as described above except that the cells were preincubated with 0.3 µM CsA for 45 minutes before conducting the uptake study. This is analogous to our in vivo PET study in which we measured the hepatic uptake CL of RSV after ∼45 minutes of CsA infusion (Billington et al., 2019).
RSV Uptake by TEC in the Presence and Absence of Albumin or Human Plasma.
100% human plasma (HP) or 5% HSA was used for these studies, and the unbound RSV concentration was maintained at ∼1 µM in the presence/absence of proteins. Hence, because of the large cost of RSV radioactivity, RSV uptake in these studies was quantified using LC-MS/MS. First, the OATP1B1-, OATP1B3-, and NTCP-expressing cells were preincubated with HBSS buffer or 100% HP or 5% HSA for 10 minutes. Then the uptake of RSV by TEC was conducted in the presence of HBSS buffer or 100% HP or 5% HSA as described above. The cell pellets were suspended in 500 μl of acetonitrile:water (80:20) containing 20 nM cerivastatin as an internal standard to precipitate the protein. After being centrifuged at 18,000g for 20 minutes, 100 μl supernatant was transferred into 96-well microplate for LC-MS/MS analysis. Extra wells in every plate were added into 1 ml 2% SDS to lyse the cells for protein quantification. The protein content in the lysate was determined using BCA Protein Assay Kit as described above.
LC-MS/MS Analysis
The supernatant was assayed by AB Sciex Triple Quad 6500 tandem mass spectrometer coupled with Waters Acquity UPLC system (Waters, Hertfordshire, UK). The injection volume was 10 μl. The liquid chromatography separations were performed using an Acuity UPLC BEH C18 column (1.7 μm, 2.1 mm × 50 mm) with a corresponding C18 VanGuard Precolumn (1.7 μm, 2.1 mm × 5 mm) (Waters Technologies, Milford, MA). The liquid chromatography mobile phases were (A) 0.1% formic acid in water and (B) 0.1% formic acid in acetonitrile. The solvent gradient profile (min/[%B]) was 0/[10%]-0.20/[10%]-1.50/[95%]-2.20/[95%]-2.30/[10%]-3.0/[10%]. The samples were analyzed in the positive ionization mode. The transitions (Q1 > Q3) for RSV and cerivastatin (internal standard) were 482.3 > 258.2 and 460.3 > 356.3, respectively. The lower limit of RSV quantification was 0.20 nM. Declustering potential was 80 and 100 V for RSV and cerivastatin, respectively. Collision energy was 30 and 40 V for RSV and cerivastatin, respectively.
RSV Unbound Fraction in 100% Human Plasma or 5% Human Serum Albumin
Five hundred microliters of the RSV solution used in the above uptake experiment was used to determine the unbound fraction of RSV using the Centrifree Ultrafiltration Device (EMD Millipore Corporation, Billerica, MA). The HBSS solution and the plasma/albumin solution were used for nonspecific binding and protein binding, respectively. The unbound fraction of RSV in 100% HP or 5% HSA was calculated by determining the amount of RSV in the samples before filtration and in the filtrate and was corrected for nonspecific binding of RSV.
[3H]RSV Uptake Study in Human Hepatocyte Models (SHs, PHs, and SCHHs)
SHs.
Cryopreserved human hepatocytes were thawed and processed as described before (Kumar et al., 2019). Briefly, the hepatocytes were centrifuged at 1000g for 5 minutes at 4°C to remove cryopreserved hepatocyte recovery medium. Then the hepatocytes were reconstituted in Ca2+-containing or Na+-free HBSS medium, and the medium containing 0.35 × 106 viable hepatocytes (50 µl) was transferred to a 1.5-ml centrifuge tube and placed in a 37°C water bath. The transport study was started by the addition of 50 µl of Ca2+-containing or Na+-free HBSS medium containing 60 nM [3H]RSV and 140 nM RSV with or without 400 µM BSP to achieve a final concentration of 30 nM [3H]RSV, 70 nM RSV, and 200 µM BSP. At 2 and 10 minutes (within linear phase), the transport was terminated as follows using the oil-spin method. First, 150 µl of ice-cold Na+-free HBSS was added to the hepatocyte incubation. Then 200 µl of the medium containing the hepatocytes was transferred to the top of the oil layer [containing 200 µl of 3% cesium chloride with 200 µl of the mixture of silicone oil: mineral oil (5:1)] and centrifuged for 15 seconds at 12,000g. The supernatant was immediately aspirated, the tube containing the pellet was cut, and the pellet was transferred and dissolved in 1 ml of 2% SDS. The total radioactivity in SH lysate (700 µl) was measured using a liquid scintillation counter (PerkinElmer). In addition, 40 µl of SH lysate was used to measure total protein using the BCA Protein Assay Kit as described above.
PHs.
Cryopreserved human hepatocytes were thawed and plated as described before (Kumar et al., 2019). About 0.35 × 106 hepatocytes were plated per well in a 24-well collagen-coated plate for 5 hours with 0.5 ml/well TORPEDO containing INVITROGRO CP medium [1:9 (v/v)]. [3H]RSV uptake (at 2 and 10 minutes; within linear phase) by the plated human hepatocytes incubated with Ca2+-containing or Na+-free HBSS medium containing 30 nM labeled [3H]RSV and 70 nM unlabeled RSV was measured as described above in the presence or absence of Na+ or 200 µM BSP. Immediately afterward, 1 ml of 2% SDS was added to each well to lyse the cells (uptake phase) and processed as described above.
SCHHs.
Cryopreserved human hepatocytes were thawed and sandwich-cultured as described before (Kumar et al., 2019). SCHH protocol was adopted from Liu et al. (1999) and Pfeifer et al. (2013). Briefly, on day 4 after sandwich-culturing the SCHHs (0.35 × 106 hepatocytes/well) were washed twice with 1 ml of Ca2+-containing HBSS buffer and then preincubated for 10 minutes at 37°C with 500 µl of Ca2+-containing HBSS. Then, the SCHHs were incubated at 37°C with 500 µl of 30 nM [3H]RSV and 70 nM unlabeled RSV for 0.5, 2, and 5 minutes (linear phase) in Ca2+-containing HBSS with or without 200 µM BSP (to assess OATP-mediated uptake) or Na+ (to assess NTCP-mediated uptake). Uptake was terminated by washing the SCHH with ice-cold Ca2+- and Mg2+-containing HBSS. Immediately afterward, 1 ml of 2% SDS was added to each well to lyse the cells (uptake phase) and processed as described above.
Quantification of Total or Plasma Membrane Transporter Abundance in Transporter-Expressing Cell Lines; Human Liver Tissues; and Suspended, Plated, and Sandwich-Cultured Human Hepatocytes
Data on total or PMA of OATP1B1/2B1/1B3 or NTCP transporter abundance in the transporter-expressing cells, human liver tissues (n = 39), and hepatocyte homogenates were obtained from our previous publications (Wang et al., 2016; Kumar et al., 2017, 2019).
Data and Statistical Analyses
Determination of Transporter-Mediated and Passive Diffusion CLint of RSV.
OATP1B1/2B1/1B3 or NTCP transporter-expressing cells.
The OATP or NTCP transporter-mediated intrinsic uptake CL (CLint,uptake,cells,1B1, CLint,uptake,cells,1B3, CLint,uptake,cells,2B1, CLint,uptake,cells,NTCP, hereafter referred to as CLint,uptake,cells) of [3H]RSV was determined first by taking the difference in [3H]RSV uptake into the cells in the presence and absence of BSP or Na+, respectively. Then, the OATP- or NTCP-mediated uptake CLint of [3H]RSV into these cells was calculated as the ratio of the rate of transporter-mediated uptake of [3H]RSV and [3H]RSV concentration in the medium (30 nM). The passive diffusion CLint (CLint,passive,cells) of [3H]RSV into the cells was calculated as above except that the rate of uptake of [3H]RSV uptake into the cells in the presence of 200 µM BSP or in the absence of sodium was used. The CLint,passive,cells used was the average of passive diffusion CL obtained in CHO, MDCKII, and HEK293 cells.
SHs, PHs, and SCHHs.
The total transporter-mediated CLint of [3H]RSV in hepatocyte models (CLint,uptake,hep) was calculated in the presence of 200 µM BSP (an OATP and NTCP inhibitor) as above. Of this, the OATP-mediated CLint of [3H]RSV was calculated as above, but the passive diffusion CLint (CLint,passive,hep) was determined by using only the data on uptake of [3H]RSV in Na+-free HBSS buffer and in the presence of BSP. NTCP-mediated CLint of [3H]RSV was calculated as above but by using only the data on uptake of [3H]RSV in Ca2+-containing and Na+-free HBSS buffer:(1)in which CLint,OATP,hep, CLint,NTCP,hep, and CLint,passive,hep are OATP-mediated, NTCP-mediated, and passive diffusion CLint of RSV in the hepatocyte models, respectively.
IVIVE of Intrinsic Hepatic Uptake CL of RSV from Transporter-Expressing Cells in the Presence or Absence of Plasma Proteins Using REF.
The CLint,in vivo,pred,cells from the transporter-expressing cells was calculated as follows as the sum of the contribution of uptake CL of [3H]RSV by each transporter (using plasma membrane abundance, Table 1) and by passive diffusion:(2)Where i represents the value calculated using transporter abundance in the ith human liver sample [n = 39; previously determined in our laboratory (Wang et al., 2016)]. CLint,in vivo,pred,passive was scaled as in eq. 4 and was assumed to be constant.
OATP1B1/2B1/1B3- and NTCP-mediated CLint,uptake,cells of [3H]RSV based on their total or PMA in the transporter-expressing cells
In Eq. 2, for each transporter, the prediction of RSV CLint,in vivo,pred,cells was calculated using eq. 3 as described below for OATP1B1 transporter:(3)in which [OATP1B1]i represents average total OATP1B1 transporter abundance in the ith human liver samples (n = 39). Since the cell-surface biotinylation methodology to estimate PMA of transporters cannot be used with intact tissue (Kumar et al., 2019), all the transporters quantified in liver tissue were assumed to be present in the plasma membrane. The RSV uptake experiments in the presence of HPs or HSA were conducted only with OATP1B1-, OATP1B3-, and NTCP-expressing cells since these three transporters account for more than 90% active uptake of RSV. Notably, since our data showed that RSV OATP2B1 and passive diffusion CL in TEC in absence of proteins accounted for only 7.6% and 0.6% of total RSV uptake, the sensitivity of our instrument was insufficient to estimate these CL in the presence of proteins because of the extensive protein binding of RSV. Therefore, we assumed OATP2B1-mediated and passive diffusion CL were identical with or without plasma proteins.
IVIVE of Intrinsic Hepatic Uptake CL of RSV from Hepatocyte Models.
CLint,uptake,hep of RSV obtained from hepatocyte models was scaled to human liver as follows:(4)in which CLint,in vivo,pred,hep is the predicted in vivo intrinsic hepatic uptake CL from data in each hepatocyte model. Liver weight is 1500 g, and the total protein per unit liver weight is 88 mg/g of liver (Karlgren et al., 2012).
Prediction of In Vivo Total Hepatic Uptake CL of RSV Based on Transporter-Expressing Cells and Hepatocyte Models.
RSV in vivo hepatic uptake CL was computed from its hepatic intrinsic uptake CL based on TEC/REF or hepatocytes as follows:(5)in which (
) is the liver blood flow, and
is the RSV fraction unbound in blood (0.17) estimated using eq. 6. The CLin vivo,pred,cell reported in the Results is the average of the 39 predicted values (each based on transporter abundance in the 39 individual livers).
(6)fu,p is the RSV fraction unbound in plasma (0.12) (Martin et al., 2003), B:P is the RSV blood to plasma ratio [0.69 (Martin et al., 2003)].
Prediction of
of [3H]RSV in Hepatocyte Models from TECs.
The CLint,uptake,hep of [3H]RSV in each human hepatocyte model was predicted from as follows:
(7)in which, [Transporter]hep represents total or PMA of the transporter (picomole per milligram protein) in the hepatocyte model (Kumar et al., 2019).
IVIVE of Inhibition of RSV CLin vivo,uptake by CsA Based on TEC/REF.
(8)
(9)Where [Transporter(i)]ave represents the average abundance of the ith transporter (i.e., OATP1B1, OATP1B3, OATP2B1 or NTCP) in the 39 human liver samples.
The predicted % inhibitory effect of 0.3 μM CsA on total RSV hepatic uptake CL was calculated as follows:(10)
Statistical Analysis.
Wilcoxon matched-pair signed rank statistical test (using Prism 7, version 7.03) was used to compare the observed CLint,uptake,total, CLint,OATP,hep, CLint,NTCP,hep, and CLint,passive,hep of [3H]RSV across the various hepatocyte models (Fig. 2). For analyses of the rest of the data shown in Figs. 2 and 3 and Supplemental Fig. 1, the Tukey’s multiple comparison test was used.
[3H]RSV CLint,uptake,hep in SHs, PHs, and SCHHs. [3H]RSV CLint,uptake,hep was determined in four lots of SHs (A), PHs (B), or SCHHs (C). The average contribution to [3H]RSV CLint,uptake,hep followed the order OATPs > NTCP > passive diffusion (D). [3H]RSV CLint,uptake,hep, CLint,OATP,hep, CLint,NTCP,hep, and CLint,passive,hep were not significantly different between the hepatocyte models (Wilcoxon matched-pair signed rank test). The average data (Avg) from (A–C) are shown in (D), and they are mean ± S.D. of four lots of hepatocytes, each conducted in triplicate.
IVIVE of RSV hepatic CLuptake,in vivo based on transporter-expressing cells or the hepatocyte models, SHs, PHs, or SCHHs in HBSS buffer (A), or adjusted for the protein-mediated uptake of RSV in transporter-expressing cells (B). The transporter-expressing cells predicted the hepatic CLuptake,in vivo within 3-fold of the observed value (1205.6 ml/min). In contrast, the hepatocyte models underpredicted RSV CLuptake,in vivo by 5- to 10-fold. OATPs were the major contributors to the total RSV CLint,in vivo,pred,cells and CLint,in vivo,pred,hep followed by NTCP, with smallest contributor being passive diffusion CL. The solid and dashed lines show the 95% confidence interval (CI) of the observed hepatic CLuptake,in vivo and 2-fold lower limit of the mean observed hepatic CLuptake,in vivo (603 ml/min). RSV CLint,in vivo,pred,cells was significantly higher (Tukey’s multiple comparison test) than CLint,in vivo,pred,hep obtained from any of the hepatocyte models. Data shown for SH, PH and SCHH are mean ± S.D. of four lots of hepatocytes, each conducted in triplicate. Data shown for transporter-expressing cells are mean ± SD of the predicted values based on transporter abundance in 39 human liver samples. (A) When the transporter-mediated CLint,uptake,cells was adjusted for the increase in the RSV uptake in the presence of 100% HP or 5% HSA, the transporter-expressing cells better predicted hepatic CLuptake,in vivo (595.3 ml/min with 100% HP; 621.4 ml/min in 5% HSA) and came close to or fell just above the lower 2-fold boundary of the observed value (B).
Results
OATP1B1-, OATP1B3-, OATP2B1-, and NTCP-Mediated RSV CLint,uptake,cells Based on Their Total or PMA in Transporter-Expressing Cells.
When expressed with respect to milligram of cellular protein, except for OATP2B1 the transporter-mediated [3H]RSV CLint,uptake,cells was similar across all the transporters (Table 1). However, when [3H]RSV CLint,uptake,cells was expressed per picomole of total or PMA of each transporter, the picture changed. In this case, the NTCP-mediated [3H]RSV was ∼2-fold greater than that by OATPs irrespective of whether CLint,uptake,cells was obtained using total or the PMA of the transporter (Table 1). Except for OATP2B1, the majority (>60%) of each transporter was found to be expressed in the plasma membrane (PM). The estimated CLint,passive,cells of [3H]RSV in CHO, MDCKII, and HEK293 cells was similar across the cells (mean: 0.19 ± 0.04 µl/min per milligram cellular protein, range: 0.15–0.24 µl/min per milligram cellular protein) and statistically not different from the corresponding passive diffusion CL in hepatocyte models (CLint,passive,hep) (SHs, PHs, and SCHHs: 0.58 ± 0.24, 0.30 ± 0.20, and 0.67 ± 0.75 µl/min per milligram protein, respectively) (Supplemental Fig. 1).
A Comparison of [3H]RSV
in SHs, PHs, and SCHHs.
The total uptake of [3H]RSV in all lots of hepatocytes was about the same except for in JEL, in which it was consistently higher than the other three lots because of greater OATP-mediated uptake (Fig. 2). Overall, CLint,uptake,hep, CLint,OATP,hep, CLint,NTCP,hep, and CLint,passive,hep of RSV were not significantly different between the hepatocyte models (Fig. 2D). In addition, the average contribution of OATPs was higher than NTCP (56% vs. 40%) to total [3H]RSV CLint,uptake,hep (Fig. 2D). Notably, the average CLint,passive,hep of [3H]RSV in SHs, PHs, and SCHHs was only 5.6% ± 3.2%, 2.2% ± 1.0%, and 7.4% ± 8.0% of the total [3H]RSV CLint,uptake,hep, respectively (Fig. 2).
Prediction of RSV CLuptake,in vivo from Transporter-Expressing Cells (CLin vivo,pred,cells) and Hepatocyte Models (CLin vivo,pred,hep) (IVIVE).
CLuptake,in vivo predicted using REF and transporter-expressing cells (OATP1B1/1B3/2B1, NTCP, and passive diffusion) fell outside our predefined acceptance criteria of being within 2-fold of the observed hepatic uptake blood CL (CLuptake,in vivo). The predicted RSV hepatic CLin vivo,pred,cells was 37.0% (446.4 ± 130.9 ml/min) of the mean observed hepatic CLuptake,in vivo (mean 1205.6 ml/min, 95% confidence interval 802–1609 ml/min) (Billington et al., 2019). Of the predicted total RSV CLuptake,in vivo by transporter-expressing cells, ∼99% was via transporters, and ∼1% was via passive diffusion (Fig. 3). Of the total predicted RSV CLint,in vivo,pred,cells (computed based on the average transporter abundance in 39 human liver samples), the contribution of OATP1B1, OATP1B3, OATP2B1, NTCP, and passive diffusion clearance was 37.9%, 15.5%, 7.6%, 38.3%, and 0.6%, respectively (Fig. 3). The CLin vivo,pred,hep based on SHs, PHs, and SCHHs was about 18% (220.4 ± 79.4 ml/min), 23% (273.0 ± 100.4 ml/min), and 16% (195.0 ± 47.2 ml/min) of the observed value, respectively (Fig. 3). The CLint,in vivo,pred,cells of RSV based on transporter-expressing cells was 3934.2 ± 1634.2 ml/min (OATP1B1: 1489.8 ± 726.8 ml/min, OATP1B3: 610.8 ± 306.6 ml/min, OATP2B1: 300.9 ± 112.9 ml/min, and NTCP: 1507.6 ± 667.5 ml/min, passive diffusion: 25.1 ml/min). The average CLint,in vivo,pred,hep of RSV based on SHs, PHs, and SCHHs (1551 ± 672.6, 2022.6 ± 965.6, and 1328.7 ± 376.2 ml/min, respectively) was ∼50%–70% lower than that based on transporter-expressing cells.
In the presence of 100% HP, the unbound RSV CLint,uptake,cells into OATP1B1-, OATP1B3-, and NTCP-expressing cells increased by 1.77 ± 0.19–, 1.67 ± 0.06–, and 1.49 ± 0.20–fold, respectively. In the presence of 5% HSA, unbound RSV CLint,uptake,cells into OATP1B1-, OATP1B3-, and NTCP-expressing cells increased by 2.07 ± 0.75–, 1.47 ± 0.26–, 1.60 ± 0.05–fold, respectively. Notably, the nonspecific binding of RSV to the apparatus (measured in HBSS) was 0.08 ± 0.06. After correcting for this, the unbound fraction of RSV in 100% HP and 5% HSA was 0.15 ± 0.01 and 0.13 ± 0.01, respectively.
Prediction of RSV CLint,uptake,hep from Transporter-Expressing Cells.
The total transporter-mediated CLint,uptake,hep in hepatocyte models predicted from cell lines (i.e., the sum of CLint predicted by the individual transporter-expressing cells, OATP1B1/1B3/2B1, and NTCP as well as passive diffusion CL) was not significantly different from the observed values irrespective of whether the predictions were based on the total or PMA of the transporters in the cells (Fig. 4). The average contribution of OATP1B1, OATP1B3, OATP2B1, NTCP, and passive diffusion to the [3H]RSV uptake in hepatocyte models was about 42% ± 8%, 9% ± 1%, 9% ± 1%, 33% ± 0.3%, and 7% ± 7%, respectively. The [3H]RSV CLint,passive,hep and CLint,passive,cells were not significantly different (Supplemental Fig. 1).
IVCIVH of [3H]RSV CLint,uptake,hep. The predicted [3H]RSV CLint,uptake,hep from transporter-expressing cells was not significantly different from the observed [3H]RSV CLint,uptake,hep, irrespective of whether the predictions were made based on the total or PMA of the uptake transporters. Data shown are mean ± S.D. of four lots of hepatocytes, each conducted in triplicate. NS, not significantly different based on Wilcoxon matched-pair signed rank statistical test.
IVIVE of the Magnitude of Inhibition of RSV Hepatic Uptake CL by 0.3 µM CsA Based on Transporter-Expressing Cells.
The %inhibition of [3H]RSV uptake into OATP1B1, OATP1B3, OATP2B1, and NTCP cells by CsA (0.3 µM) was 56.5% ± 6.1%, 47.5% ± 12.5%, 56.6% ± 32.0%, and 21.9% ± 6.7%, respectively. This translated to CsA (0.3 µM) inhibiting the total CLuptake,in vivo of RSV by 32.9%. In our PET imaging study, the in vivo % inhibitory effect of 3 µM CsA (=0.1) on [11C]RSV hepatic uptake in three subjects was 40.3% ± 6.8%, with the 95% confidence interval being 23.4%–57.2% (Billington et al., 2019).
Transporter Abundance-Activity Correlation in Human Hepatocyte Models.
The observed total CLint,uptake,hep of [3H]RSV in the human hepatocyte models was well correlated with total transporter abundance of OATP1B1 (R2 = 0.80; Fig. 5A) but not with total abundance of OATP1B3, OATP2B1, or NTCP (Fig. 5, B–D). This conclusion did not change if the PMA of the transporters was used as the independent variable (unpublished data).
Correlation of [3H]RSV CLint,uptake,hep and total transporter protein abundance of OATP1B1 (A), OATP1B3 (B), OATP2B1 (C) and NTCP (D) in hepatocyte models. The [3H]RSV CLint,uptake,hep was highly correlated with the total abundance of OATP1B1. However, the correlation with OATP1B3, OATP2B1, and NTCP transporter abundance was poor. Similar results were obtained if the x-axis was PMA of the respective transporter. Data shown are mean of triplicates in four lots of hepatocytes models.
Discussion
We designed our study to incorporate several unique features hitherto not considered when conducting IVIVE of transporter-based CL. First, we did not assume as others have (Li et al., 2010; Kim et al., 2019) that in vivo CL of OATP substrates, such as RSV, is rate-determined by its uptake into the liver. This assumption can only be made if the sinusoidal efflux CL of the drug is much less than its metabolic plus biliary CL (Patilea-Vrana and Unadkat, 2016, 2018). Our PET imaging data have confirmed that this assumption is not correct for RSV (Billington et al., 2019) and, as discussed below, will result in biased evaluation of the CLin vivo,pred. Here, we correctly compared the in vivo RSV hepatic uptake CL determined by PET imaging with that predicted based on TEC/REF and hepatocytes without any assumptions of the rate-determining step. Second, we studied RSV uptake in multiple hepatocyte models (using cells isolated from the same donor) in which we previously quantified the total and PMA abundance of OATPs and NTCP (Kumar et al., 2018). Hence, we were able to test which hepatocyte model (SHs, PHs, or SCHHs) better predicted in vivo hepatic uptake CL of RSV and compared these predictions with that obtained from transporter-expressing cells. Third, quantifying RSV uptake CL into cells and hepatocytes allowed us to determine whether there was an in vitro cells to in vivo hepatocytes extrapolation (IVCIVH) agreement in the predicted transporter-based hepatic uptake CL of RSV. This comparison is important (as described below) because both in vitro models underestimated the in vivo hepatic uptake CL of RSV, albeit to a different extent. Fourth, because there is limited data on IVIVE of transporter-mediated DDI, we determined whether the magnitude of in vivo inhibition of hepatic uptake CL of RSV by CsA observed in our PET imaging study (Billington et al., 2019) could be predicted from in vitro uptake studies in TEC.
The predicted RSV CLin vivo,pred,cells by transporter-expressing cells was 446.4 ± 130.9 ml/min when plasma/albumin was not included in the uptake studies, a value much less than (∼37.0%) of the observed value (1205.6 ml/min; calculated from Billington et al. (2019)). Thus, IVIVE of RSV CLin vivo,pred,cells based on TEC/REF did not meet our predefined success criteria of being within 2-fold of the observed value, a criterion widely accepted for wide therapeutic index drugs. The goal of IVIVE of CL is not to accurately predict CL of a drug but to provide an estimate of the eventual CL of the drug likely to be expected in a phase 1 trial of the drug. If the % PMA of transporters in liver tissue is similar to that in the hepatocytes (and not 100% as assumed), our predicted CLin vivo,pred,cells would not change substantially (although it would be slightly lower) because the major contributors to the uptake of RSV (OATP1B1 and NTCP) were predominately present in the plasma membrane of the hepatocytes and the transporter-expressing cells.
Since human hepatocytes are routinely used to predict transporter-mediated CL in vivo, we next examined whether the various human hepatocytes models (SHs, PHs, or SCHHs) could better predict RSV CLuptake,in vivo than the TEC/REF approach. They did not; all the models considerably underpredicted (by about 5–10-fold) RSV CLuptake,in vivo, among which the PHs showed the best performance, whereas the SCHHs showed the worst. This finding is consistent with previous literature reports that hepatocytes underpredict (sometimes drastically) transporter-mediated hepatic CL of drugs (Jones et al., 2012). Based on the above discrepancy, we asked whether the RSV CLint,uptake,cells by transporter-expressing cells could predict CLint,uptake,hep by hepatocyte models using the proteomics-informed approach (IVCIVH). Indeed, it could, irrespective of whether the total or PMA of the transporters was used as REF (Fig. 4). In addition, the RSV CLint,passive,cells predicted by the TECs was not significantly different from that observed in hepatocytes (CLint,passive,hep) (Supplemental Fig. 1). This result immediately poses the question: why was the RSV CLin vivo,pred,cells significantly greater than that predicted from the hepatocytes (CLin vivo,pred,hep) (Fig. 3)? This was because we used PMA-based REF to predict CLin vivo,pred,cells, whereas CLin vivo,pred,hep did not take into consideration PMA of transporters. Instead, it used the total protein in human liver as a scaling factor and assumed 100% PMA of the transporters in the hepatocyte models even though it was only 60%–80% (OATPs and NTCP) (Kumar et al., 2019). Indeed, when a hybrid (REF and total protein) approach was used to scale CLint,hep,predicted (Fig. 4; Table 1), the CLin vivo,pred,hep was not significantly different from CLin vivo,pred,cells (unpublished data).
The above IVCIVH comparison was informative to address the question as to why both approaches (TEC/REF and hepatocytes) underestimated RSV CLuptake,in vivo, with the hepatocytes doing so more than the cells. A possible explanation is that the in vitro models are not adequately capturing the in vivo transporter (OATP/NTCP)-mediated or the passive hepatic uptake clearance of RSV or both. This could be due to differences in the in vivo RSV affinity (Km) or kcat (catalytic turnover) of the drug by the transporters from those in vitro. Mechanistically, this could be due to a difference in post-translational modifications of the transporter protein(s) between in vitro and in vivo resulting in either a changed kcat, or the affinity (Km) for the drug. In vitro to in vivo differences in kcat or Km could also be caused by endogenous factors present in in vivo but absent in in vitro models, such as plasma proteins (Bowman et al., 2019). However, when we adjusted the transporter-mediated CLint,uptake,cells of RSV for the HP or HSA-mediated increase in RSV CLint,uptake,cells, the predicted hepatic CLint,in vivo,pred,cells improved and was close to (HP) or fell just above the lower boundary (HSA) of our success criteria (Fig. 3B). With 100% HP the 95% CI range of the predicted value (547.4∼643.2 ml/min) overlapped with the lower 2-fold boundary of our acceptance criteria. Based on our acceptance criteria, we can conclude that the TEC/REF approach was successful in predicting RSV CLuptake,in vivo.Nevertheless, the TEC/REF approach under-estimated the observed CLuptake,in vivo by about 50%. Therefore, other explanations need to be invoked to explain the remaining IVIVE discrepancy in prediction of RSV CLuptake,in vivo. One possibility is that our PET imaging study overestimated the true CLuptake,in vivo because PET imaging does not allow us to distinguish between RSV present in the hepatocytes versus that in the canaliculi. Interestingly, the above explanations for underestimation of in vivo RSV hepatic uptake CL by in vitro systems must be species-dependent because we have previously shown that TEC/REF approach successfully and accurately predicted the in vivo transporter-mediated RSV hepatic uptake CL in rats obtained by PET imaging (Ishida et al., 2018a).
To illustrate why comparing the predicted RSV CLin vivo,pred,cells in TECs to the observed in vivo systemic RSV blood CL is erroneous, let us consider the following. Our predicted (446.4 ± 130.9 ml/min without HP/HSA; 595.3 ± 152.6 ml/min with HP; 621.4 ± 155.0 ml/min with HSA) value is well within the 2-fold of the observed in vivo systemic RSV blood CL (850.5 ml/min; Martin et al., 2003) (Supplemental Fig. 2). This is because when hepatic uptake CL is not the rate-determining step, the in vivo systemic CL of a drug will always be less than the hepatic uptake CL of the drug—hence comparing the hepatic uptake CL predicted based on TEC or hepatocyte studies, even though an underestimation of the true hepatic uptake CL may fortuitously fall within the 2-fold boundary of the observed in vivo systemic CL of the drug (Supplemental Fig. 2). Thus, previous reports of success with such comparisons should be viewed with great caution (e.g., Kim et al., 2019).
Although NTCP-mediated RSV CLint,uptake,cells,NTCP was about twice that by OATP1B1 (Table 1) due to the greater hepatic/hepatocyte abundance of OATP1B1, the latter was the dominant contributor (33%–48%) to the total RSV CLint,uptake,hep. The contribution of OATP1B1 (33%–48%) and NTCP (33%) to RSV CLint,uptake,hep was higher than that of OATP2B1 (8%–10%) or OATP1B3 (8%–9%). Thus, it is not surprising that OATP1B1 total abundance (or PMA) showed good correlation (R2 ≥ 0.80) with total RSV CLint,uptake,hep into SHs, PHs, and SCHHs (Fig. 5). The poorer correlation of total RSV CLint,uptake,hep with NTCP abundance is surprising and needs further exploration.
REF based on PMA of a transporter is expected to better predict transporter-mediated clearance than REF based on total cellular transporter abundance. This was not the case here because the major contributors to RSV CLint,uptake,hep OATP1B1 and NTCP were found to be predominately in the plasma membrane of cells and hepatocytes (Kumar et al., 2018). Nevertheless, as was shown previously for OCT2 (Kumar et al., 2018), it is important that PMA of transporters along with the in vivo transport mechanism be considered in IVIVE using TEC/REF approach.
One goal of IVIVE of hepatobiliary CL of drugs is to predict the hepatic concentrations of the drug when the site of toxicity or efficacy is within the liver. Obviously, predicting the in vivo hepatic CL of a drug is insufficient to predict the hepatic concentration of a drug. This is because all in vivo hepatobiliary clearances [sinusoidal uptake CL (CLs,uptake), metabolic CL (CLmetabolism), biliary efflux CL (CLb,efflux), sinusoidal efflux CL (CLs,efflux)] need to be estimated or predicted in vitro. Although the SCHHs have been suggested as a tool to estimate these clearances, as previously shown by us (Kumar et al., 2020) and the data presented here on the shortfall of the SCHH model, alternative methods are needed to estimate all these hepatobiliary clearances. One alternative method is to use proteomics-based REF for IVIVE of CLs,efflux and CLb,efflux. CLmetabolism can readily be obtained by using human liver microsomes/cytosol or S9 fraction. As to whether this proteomics-informed REF approach can successfully predict these clearances and tissue concentration of drugs in humans is yet to be determined. However, initial success in IVIVE of hepatic 11C-RSV concentrations in our rat study demonstrates the potential of this approach (Ishida et al., 2018a).
In our PET imaging study, we found that 0.3 µM CsA reduced RSV hepatic CLuptake,in vivo in three of the four subjects (Billington et al., 2019). In those in which it did, the in vivo %inhibition by CsA was 40.3% ± 6.8%. Our TEC successfully predicted the %inhibition (32.9%) of CLuptake,in vivo because it fell within the 95% confidence interval (23.4%–57.2%) of the % inhibition observed in vivo. This observation adds to the growing evidence that in vivo transporter-mediated DDI can be predicted from in vitro studies (Ishida et al., 2018b).
In summary, we compared the ability of two in vitro systems, TEC and human hepatocyte models, to predict in vivo RSV hepatic CLuptake,in vivo in humans. We show that the TEC/REF approach resulted in a better prediction of the RSV hepatic CLuptake,in vivo compared with the hepatocyte models. However, both in vitro models failed to accurately predict (within 2-fold) the observed RSV CLuptake,in vivo. The reason for this in vitro to in vivo discrepancy remains unclear and needs further investigation. Clearly, inclusion of albumin or plasma in the in vitro uptake TEC studies could bridge this discrepancy (but just barely). That is, the predicted RSV CLuptake,in vivo, based on TEC/REF approach, was about 50% of the observed value. Despite this success, further research is needed to determine how predictions by the TEC/REF and hepatocyte models can be improved. Interestingly, this result is in contrast to our previous success using TEC/REF to predict the in vivo hepatic CL of RSV in rats (Ishida et al., 2018a) and renal secretory CL of metformin in humans (Kumar et al., 2018). Collectively, our data suggest that the TEC/REF approach appears to be superior to hepatocytes in IVIVE of transporter-mediated hepatic CL of drugs. In addition, as we have previously shown (Ishida et al., 2018a), the REF approach can successfully predict the magnitude of in vivo DDI.
Acknowledgments
The authors thank Tot Bui Nguyen (University of Washington) for her support in human hepatocytes, transporter-expressing cell line culture, and LC-MS/MS proteomics. The authors also thank Bhagwat Prasad (University of Washington) for his support in LC-MS/MS proteomics and Bruno Stieger (University of Zurich) for providing OATP1B1-expressing CHO cells.
Authorship Contributions
Participated in research design: Kumar, Yin, Ishida, Salphati, Hop, Rowbottom, Xiao, Lai, Mathias, Chu, Humphreys, Liao, Heyward, Unadkat.
Conducted experiments: Kumar, Yin.
Contributed new reagents or analytic tools: Nerada, Szilvásy, Heyward.
Performed data analysis: Kumar, Yin, Ishida, Unadkat.
Wrote or contributed to the writing of the manuscript: Kumar, Yin, Ishida, Salphati, Hop, Rowbottom, Xiao, Lai, Mathias, Chu, Humphreys, Liao, Nerada, Szilvásy, Heyward, Unadkat.
Note Added in Proof: An error in Equation 6 of the Method section (propagated to Fig. 3 and Supplemental Fig. 2 of results) was found in the Fast Forward version published Oct 13, 2020. This error, as well as Fig. 3 and Supplemental Fig. 2 of the Results, and its impact on the rest of the manuscript (including Abstract, Results, Significance Statement and Discussion) have been corrected.
Footnotes
- Received August 4, 2020.
- Accepted September 24, 2020.
↵1 Current affiliation: Gilead Sciences, Inc., Foster City, California.
↵2 Current affiliation: Sunovion Pharmaceuticals, Inc., Marlborough, Massachusetts.
↵3 Current affiliation: Aranmore Pharma Consultant, Trenton, New Jersey.
↵4 Current affiliation: Clovis Oncology, San Francisco, California.
V.K. and M.Y. were supported in part by the Simcyp Grant and Partnership Scheme and University of Washington Research Affiliate Program on Transporters (UWRAPT) funded by Genentech, Biogen, Gilead, Merck, Bristol-Myers Squibb, Pfizer, and Takeda.
↵
This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- BCA
- bicinchoninic acid assay
- BSP
- bromsulphthalein
- CL
- clearance
- CsA
- cyclosporine A
- DDI
- drug-drug interaction
- DMEM
- Dulbecco’s modified Eagle’s medium
- HBSS
- Hank’s balanced salt solution
- HEK293
- human embryonic kidney 293
- HP
- human plasma
- HSA
- human serum albumin
- IVCIVH
- in vitro cells to in vivo hepatocytes extrapolation
- IVIVE
- in vitro to in vivo extrapolation
- LC-MS/MS
- liquid chromatography tandem mass spectrometry
- NTCP
- Na+-taurocholate cotransporting polypeptide
- OATP
- organic anion transporting polypeptide
- PET
- positron emission tomography
- PH
- plated hepatocyte
- PMA
- plasma membrane abundance
- REF
- relative expression factor
- RSV
- rosuvastatin
- SCHH
- sandwich-cultured human hepatocyte
- SH
- suspended hepatocyte
- TEC
- transporter-expressing cell
- Copyright © 2021 by The American Society for Pharmacology and Experimental Therapeutics