Abstract
The successful prospective incorporation of in vitro transporter kinetics in physiologically based pharmacokinetic (PBPK) models to describe drug disposition remains challenging. Although determination of scaling factors to extrapolate in vitro to in vivo transporter kinetics has been facilitated by quantitative proteomics, no robust assessment comparing membrane recoveries between different cells/tissues has been made. HEK293 cells overexpressing OCT2, MATE1, and MATE2K or human kidney cortex were homogenized and centrifuged to obtain the total membrane fractions, which were subsequently subjected to liquid-liquid extraction followed by centrifugation and precipitation to isolate plasma membrane fractions. Plasma membrane recoveries determined by quantitation of the marker Na+/K+-ATPase in lysate and plasma membrane fractions were ≤20% but within 3-fold across different cells and tissues. A separate study demonstrated that recoveries are comparable between basolateral and apical membranes of renal proximal tubules, as measured by Na+/K+-ATPase and γ-glutamyl transpeptidase 1, respectively. The plasma membrane expression of OCT2, MATE1, and MATE2K was quantified and relative expression factors (REFs) were determined as the ratio between the tissue and cell concentrations. Corrections using plasma membrane recovery had minimal impact on REF values (<2-fold). In vitro transporter kinetics of metformin were extrapolated to in vivo using the corresponding REFs in a PBPK model. The simulated metformin exposures were within 2-fold of clinical exposure. These results demonstrate that transporter REFs based on plasma membrane expression enable a prediction of transporter-mediated drug disposition. Such REFs may be estimated without the correction of plasma membrane recovery when the same procedure is applied between different matrices.
SIGNIFICANCE STATEMENT Transporter REFs based on plasma membrane expression enable in vitro-in vivo extrapolation of transporter kinetics. Plasma membrane recoveries as determined by the quantification of sodium-potassium adenosine triphosphatase were comparable between the in vitro and in vivo systems used in the present study, and therefore had minimal impact on the transporter REF values.
Introduction
PBPK models are used widely within the pharmaceutical industry and are also recommended by regulatory agencies to aid in the prediction of drug disposition and drug-drug interactions in humans (Zhao et al., 2011; EMA, 2012; FDA, 2012; Jones et al., 2015). These models have been most predictive when compounds are highly permeable and predominantly metabolized by cytochrome P450 enzymes. However, with the current industry focus on advancing metabolically stable compounds to the clinic, the drugs being developed now often have significant transporter-mediated clearance, for which building a reliable PBPK model has limitations and challenges (Jones et al., 2015; Taskar et al., 2020).
To effectively incorporate transporter kinetic data into a PBPK model, transporter abundance differences between in vitro and in vivo systems must be considered. Among a host of methodologies for transporter quantification, quantitative proteomics is considered to have the fewest limitations while offering the highest accuracy when compared with other approaches such as mRNA quantification and Western blot (Ohtsuki et al., 2011; Prasad and Unadkat, 2014; Qiu et al., 2014). Meta-analysis of quantitative proteomics data, however, has revealed interstudy variability in transporter abundances. For example, the hepatic expression of organic anion transporting polypeptide (OATP) 1B1, OATP1B3, and OATP2B1 showed statistically significant differences between studies (Badee et al., 2015). Variability may be due to differences in sample preparation methodologies, impacting recovery of target proteins (Harwood et al., 2014). However, no robust cross-system assessment has been conducted to compare the membrane protein recovery between different systems. Indeed, a recent study comparing six different methodologies of quantitative proteomics revealed nearly 100-fold variability in the measured transporter abundance starting from the same tissue samples (Wegler et al., 2017).
As a result, there have been both successes and challenges for the in vitro-in vivo extrapolation (IVIVE) of transporter kinetics using quantitative proteomics. For example, the intrinsic hepatic clearance of rosuvastatin in vivo was accurately predicted by extrapolating the uptake clearance measured in human embryonic kidney (HEK) 293 cells stably expressing OATP1B1, OATP1B3, or OATP2B1 (major transporters involved in rosuvastatin hepatic uptake) to whole liver using a PBPK model after accounting for the difference in transporter abundance on plasma membrane (Bosgra et al., 2014). Similarly, the correction with protein expression on plasma membrane in vitro and consideration of membrane potential difference between in vitro and in vivo enabled a successful IVIVE of metformin uptake clearance via organic cation transporter (OCT) 2 (Kumar et al., 2018). Conversely, an average empirical scaling factor of 58 was required to explain the clinical exposure of seven OATP substrates when the uptake clearance determined in sandwich culture human hepatocytes was used in a PBPK model (Jones et al., 2012). This empirical scalar is not fully explained by human liver to sandwich culture hepatocyte abundance ratios based on quantification of total membrane transporter proteins; 1.7-fold (OATP1B1), 7.5-fold (OATP1B3), and 3.6-fold (OATP2B1) (Kimoto et al., 2012). These observations are consistent with the hypothesis that transporters expressed on plasma membrane, not total membranes, would determine functional activity, and may provide a better scalar for IVIVE.
Recently, an optimized biotinylation method was developed and applied to quantify the plasma membrane expression of transporters in vitro (Kumar et al., 2017). This methodology demonstrated that the fraction of transporters expressed on the plasma membrane can vary between different transporters and different host cells, further emphasizing the importance of considering transporter abundance on plasma membrane in IVIVE. However, this biotinylation method can only be applied to in vitro cells and does not allow the quantification of transporter expression on plasma membranes from whole tissue samples.
The aim of this study was 1) to establish a novel methodology to quantify the transporter protein expression on plasma membranes, and 2) to investigate if there are differences in plasma membrane recovery between in vitro and in vivo systems and whether corrections are warranted in transporter IVIVE using REFs. Sodium-potassium adenosine triphosphatase (Na+/K+-ATPase) is a well characterized plasma membrane protein marker (Rajasekaran et al., 2010; Imai et al., 2013), and in this study, its abundance was used to assess plasma membrane recovery. Plasma membrane recovery was determined in HEK293 cells overexpressing the renal transporters OCT2, multidrug and toxin exclusion protein (MATE) 1 and MATE2K, and human kidney cortex. A potential difference in the recovery between basolateral and apical membranes was also explored. The plasma membrane expression of these transporters was quantified using proteomics, and REFs determined with or without the correction for plasma membrane recovery.
Materials and Methods
Materials.
HEK293 cell lines stably expressing OCT2, MATE1, or MATE2K (HEK-OCT2, HEK-MATE1, or HEK-MATE2K), and the corresponding mock-transfected cell line (HEK-EV) were obtained from Prof. Kathleen M. Giacomini (Department of Biopharmaceutical sciences, School of Pharmacy, University of California San Francisco). These cells were cultured as described previously (Kikuchi et al., 2013). Snap-frozen human kidney cortexes from 9 individuals were obtained from Amsbio (Cambridge, MA). The demographics of these kidney tissues are summarized in Supplemental Table 1. All other materials were purchased from a commercial source unless stated otherwise.
Plasma Membrane Purification.
Plasma membrane fractions were prepared from transporter-expressing HEK293 cells and human kidney cortex using a plasma membrane protein extraction kit (ab65400; Abcam, Cambridge, MA) according to the manufacturer’s instructions. Briefly, cells or tissue were resuspended in the homogenization buffer containing a protease inhibitor cocktail included in the kit and homogenized using either Dounce homogenizer (cells) or OMNI-YHO electric homogenizer (tissues). Centrifugation steps were then performed (10,000g for 30 minutes at 4°C) to separate the cytosolic fraction from the total membrane fraction, which consists of plasma membranes and cellular organelle membranes. To isolate plasma membrane proteins, the total membrane fraction was resuspended in upper and lower phase solution from the kit (liquid-liquid extraction). This mixture was then centrifuged (1000g for 10 minutes), with the upper phase collected and extracted twice to maximize the yield of plasma membrane proteins. The pooled upper phase was combined, and 5 volumes of cold water was added, and then it was stored at 4°C for overnight. The plasma membrane fraction was then pelleted by centrifugation (17,000g for 30 minutes at 4°C).
Western Blot.
Total and plasma membrane fractions were subject to Western blot analysis to confirm the enrichment of plasma membrane using Na+/K+-ATPase and calreticulin as markers for plasma and endoplasmic reticulum membranes, respectively. Proteins (10 μg) from each fraction were electrophoresed on 4%–12% SDS-polyacrylamide gel and transferred to a polyvinylidene difluoride membrane. Anti-Na+/K+-ATPase monoclonal (1:1000; Abcam) and anti-calreticulin polyclonal antibodies (1:1000; Rockland Immu, Limerick, PA) were used as the primary antibodies, with anti-mouse IgG AlexaFluor 680 and anti-rabbit IgG IRDye 800 used as secondary antibodies, respectively. The signals were detected using an LI-COR ODYSSEY system.
Trypsin Digestion.
Plasma membrane fractions were suspended in 0.2% of RapiGest (Waters, Milford, MA) in 50 mM ammonium bicarbonate, and the protein concentrations were determined by the Pierce BCA protein assay (Thermo Scientific, Waltham, MA) with bovine serum albumin as a standard according to the manufacturer’s instructions. Protein fractions were digested using an In-Solution Tryptic Digestion Kit (Thermo, Rockford, IL) according to the manufacturer’s suggestions. Briefly, the lysate or plasma membrane proteins (5–24 µg from transporter-expressing HEK293 cells and 20–500 µg from human kidney cortex) were diluted in 50 mM ammonium bicarbonate and heat-denatured (5 mM dithiothreitol; 5 minutes at 95°C) followed by alkylation (10 mM iodoacetamide; 20 minutes in the dark at room temperature). The incubation was then digested with trypsin (37°C for 3 hours), followed by an additional trypsinization at the same trypsin/protein ratio at 30°C overnight. In a preliminary experiment, various ratios of trypsin/protein were evaluated, with the condition yielding the highest overall peptide amounts resulting from a trypsin/protein ratio of 1:20. The digestion was terminated by the addition of trifluoroacetic acid (final concentration of 0.1%), and the samples were analyzed by liquid chromatography/tandem mass spectrometry (LC-MS/MS) as described below. In a separate experiment conducted to compare the plasma membrane recovery between basolateral and apical membranes as measured by Na+/K+-ATPase or γ-glutamyl transpeptidase 1 (GGT1), respectively, trypsin digestion was performed on acetone-precipitated protein pellets using the iST kit (PREOMICS GMBH).
Assumptions in Quantitative Proteomics.
The following key assumptions were made in the quantitative proteomics and data analysis in this study:
Na+/K+-ATPase and GGT1 are predominantly expressed on the basolateral and apical membranes of renal proximal tubules, respectively, as supported by immunohistochemistry (Katz, 1982; Hanigan and Frierson, 1996; Wetzel and Sweadner, 2001; Peti-Peterdi et al., 2002; Hanigan et al., 2015).
The recovery of membrane proteins is the same whether looking at the transporter of interest or plasma membrane markers.
Only transporters expressed on the plasma membrane are involved in drug transport into or out of the cells.
OCT2, MATE1, and MATE2K are equally distributed through proximal tubules in human kidney cortex.
The value for the number of proximal tubule cells in human kidney cortex reported by McLaren et al., 1995 (16 × 106 PTC/g of kidney cortex) was used in this study to convert the transporter expression in human kidney cortex to the level at proximal tubule cells.
Quantitative Proteomics for Plasma Membrane Markers and Transporters.
Surrogate peptides for each protein were selected as described previously (Prasad and Unadkat, 2014). The sequences and multiple reaction monitoring transitions of surrogate peptides, native and stable isotope label, are shown in Supplemental Table 2. Na+/K+-ATPase was quantified as the average of two different surrogate peptides (<25% difference between peptides, indicating similar digestion efficiency), except in the relative quantification in human kidney cortex (subject #7–9) where a single peptide was used, whereas GGT1, OCT2, MATE1, and MATE2K were quantified using the surrogate peptide which yielded the highest signal (representing best digestion efficiency, extraction, and/or MS ionization) among multiple peptides evaluated in the preliminary assessment. All analyses were performed using a Sciex QTRAP-6500, whereas the liquid chromatography conditions used three methods (Supplemental Table 3). OCT2, MATE1, and MATE2K peptides were quantified in three separate preparations of plasma membrane fractions from transporter-expressing HEK293 cells and plasma membrane fractions of human kidney cortex from six subjects (#1–6). Na+/K+-ATPase and GGT1 were quantified in kidney cortex from additional three subjects (#7–9) to compare the recovery of basolateral and apical membranes.
Determination of Plasma Membrane Recovery.
Abundances of the basolateral membrane marker Na+/K+-ATPase in transporter-expressing HEK293 cells and human kidney cortex, as well as the apical membrane marker GGT1 in human kidney cortex, were determined by LC-MS/MS in lysate and plasma membrane fractions prepared from each cell and tissue as described above. The recovery of plasma membranes (PM recovery, %) relative to cellular lysate during the sample preparation was calculated using the equation below, assuming that Na+/K+-ATPase and GGT1 are exclusively expressed on basolateral and apical plasma membranes, respectively: where the numerator and denominator represent the abundance of Na+/K+-ATPase or GGT1 in the plasma membrane and lysate fractions, respectively.
Determination of Transporter Relative Expression Factors.
The expression of OCT2, MATE1, and MATE2K was corrected for the plasma membrane recovery as follows: where [Transporter]PM,apparent and [Transporter]PM,normalized represent the plasma membrane expression of transporters before and after the correction for plasma membrane recovery, respectively. The transporter abundance in human kidney cortex was further converted to the proximal tubule cells (PTC) level using the number of proximal tubule cells per gram of kidney cortex (16 × 106 PTC/g of kidney cortex) (McLaren et al., 1995). REFs were calculated as the ratio of transporter abundance between proximal tubule cells and in vitro cell lines prior to or after the correction by plasma membrane recovery.
Uptake Assays.
Uptake studies using HEK-OCT2, HEK-MATE1, HEK-MATE2K, or HEK-EV were conducted as described previously (Kikuchi et al., 2013). The uptake of metformin by HEK-OCT2 was linear up to 4 minutes (data not shown). The uptake linearity by HEK-MATE1 and HEK-MATE2K couldn’t be accurately determined due to the known overshoot of uptake under intracellular acidification condition (increase followed by decrease due to shortage of driving force) (Kikuchi et al., 2013); the uptake of metformin reached its maximum at 1.5–3 minutes (data not shown). The uptake of metformin for 2 minutes (OCT2) or 1 minute (MATE1 and MATE2K) was determined at various concentrations (13.7–30,000 μM). Transporter-specific uptake was obtained by subtracting the uptake into mock-transfected cells from that into transporter-expressing cells. Km and Vmax values were obtained using the following equation: where v is the initial uptake velocity of the substrate and S is the substrate concentration in the medium. Fitting was performed by the nonlinear least-squares method using GraphPad Prism. Vmax values were normalized to account for number of cells (nmol/min/106 cells) using a cellular protein concentration (mg protein/106 cells) determined in a separate experiment, where the protein abundance in one million cells was determined by Pierce BCA protein assay after being precipitated in acetonitrile: water (1:1, v/v) then dissolved in 0.1 N NaOH/0.5% SDS.
In Vitro to In Vivo Extrapolation of Metformin Transport Kinetics in the Simcyp PBPK Model.
In vitro transport kinetics of metformin by OCT2, MATE1, and MATE2K were extrapolated to in vivo using experimentally determined Km and Vmax values and the corresponding REFs (following plasma membrane recovery correction) in the Simcyp metformin PBPK model (mechanistic kidney model; version 16; Simcyp Ltd, Sheffield, UK). Except for the in vitro transport kinetics and REF values, the default input parameters were used in the simulation. The electrochemical gradient model was activated for OCT2 as described previously (Burt et al., 2016). A sensitivity analysis for REFs was conducted in the metformin PBPK model to confirm the model’s utility in evaluating the experimentally determined REFs for the basolateral uptake (OCT2) and apical efflux transporters (MATE1 and MATE2K; MATE1 was selected only for the purpose of sensitivity analysis). The metformin exposure-time profile was simulated at two different doses (250 and 500 mg) across 70 (10 trials × 7 subjects) and 60 (10 trials × 6 subjects) healthy volunteers for the 250 mg and 500 mg dose, respectively, using the experimentally determined in vitro transporter kinetics and REF values for OCT2, MATE1, and MATE2K. Model performance was assessed by comparing the simulated profiles to the clinically observed exposures (Somogyi et al., 1987; Wang et al., 2008).
Results
Enrichment of Plasma Membranes and Recovery.
The total and plasma membrane fractions obtained from HEK293 cells transfected with either an empty vector (EV) or OCT2 were analyzed by Western blot to verify the enrichment of plasma membranes (Fig. 1A). In both cell lines, the plasma membrane marker Na+/K+-ATPase was significantly enriched in the plasma membrane fraction compared with the total membrane fraction. The levels of calreticulin, an intracellular membrane marker associated with endoplasmic reticulum, were substantially lower in plasma membrane fractions than total membrane fractions. These results demonstrate the effectiveness in decreasing intracellular membrane proteins while retaining plasma membrane proteins in the plasma membrane fractions. The recovery of plasma membranes from transporter-expressing HEK293 cells and human kidney cortex is shown in Fig. 1B and Supplemental Table 4. Although higher variability was seen in HEK-OCT2 cells, average recoveries were comparable (<3-fold) between different cells and tissues; 20%, 13%, 7.4%, and 12% in HEK-OCT2, -MATE1, -MATE2K cells, and human kidney cortex, respectively.
Comparison of Basolateral and Apical Membrane Recovery.
The basolateral membrane marker Na+/K+-ATPase and apical membrane marker GGT1 were quantified in the lysate and plasma membrane fractions of kidney cortex from three additional subjects to compare the plasma membrane recoveries between the two sides of renal proximal tubule epithelial cells (Supplemental Table 5). The recovery values of Na+/K+-ATPase were comparable between 1) when the absolute quantification values (reference standard of known concentration; fmol/106 cells) are used in calculation and 2) when the relative quantification values (reference standard of approximate concentration; arbitrary unit/106 cells) are used, indicating that the relative quantification approach is sufficient to determine the membrane recovery. The recovery of GGT1 determined by relative quantification was comparable (<1.5-fold) to that of Na+/K+-ATPase. Thus, the plasma membrane recovery determined by the quantification of Na+/K+-ATPase in individual kidney cortex samples was used to normalize the transporter expression in subsequent analysis regardless of the localization of transporters.
Quantification of Transporters and Determination of Relative Expression Factors.
The mean abundance of OCT2, MATE1, and MATE2K on the plasma membrane of overexpressing cells was 771, 471, and 391 fmol/106 cells, respectively, before plasma membrane recovery correction (Table 1). Values were corrected using the plasma membrane recovery from the individual experiments, resulting in 4064, 4276, and 5351 fmol/106 cells, respectively (Table 1). The mean abundance in human kidney cortex before and after plasma membrane recovery (determined by Na+/K+-ATPase) correction was 4233, 6774, and 789 fmol/106 PTC, and 37355, 58982, and 6947 fmol/106 PTC for OCT2, MATE1, and MATE2K, respectively (Table 2). Accordingly, the relative expression factors for OCT2, MATE1, and MATE2K were 5.5, 14, and 2.0 prior to correction and 9.2, 14, and 1.3 after plasma membrane recovery correction, respectively (Table 3).
In Vitro Transporter Kinetics of Metformin by OCT2, MATE1, and MATE2K.
Metformin uptake was saturable with mean Km and Vmax values of 1872 μM and 7.29 nmol/min/106 cells (OCT2), 454 μM and 2.93 nmol/min/106 cells (MATE1) and 1840 μM and 3.40 nmol/min/106 cells (MATE2K) (Table 4). A representative data and individual values are shown in Fig. 2 and Supplemental Table 6, respectively.
IVIVE of Metformin Transport Kinetics.
A sensitivity analysis for REFs was conducted in the metformin PBPK model to confirm the model’s utility in evaluating the experimentally determined REFs for the basolateral uptake (OCT2) and apical efflux transporters (MATE1 and MATE2K). Although both MATE transporters exist on the apical membrane of proximal tubule cells, the sensitivity analysis here focused on MATE1, as it likely plays a more predominant role in metformin transport; ∼8.5-fold higher plasma membrane expression assuming the same relative digestion, extraction, and ionization efficiencies between the two surrogate peptides (Table 2) and 3.5-fold higher metformin uptake clearance (Vmax/Km) (Table 4) than MATE2K. The metformin renal clearance and plasma area under the curve (AUC) were sensitive to both OCT2 and MATE1 REFs; a corresponding increase in renal clearance and decrease in AUC were seen as REF values for OCT2 and MATE1 were increased (Fig. 3). In the subsequent PBPK simulation of metformin exposure, all three metformin transporters (OCT2, MATE1, and MATE2K) were included using the corresponding in vitro transport kinetics and REF values. The simulated exposure-time profiles of metformin at two doses (250 and 500 mg) based on the experimental parameters were within 2-fold of the clinically observed exposure (Fig. 4; Table 5).
Discussion
The IVIVE of transporter kinetics coupled with transporter quantification is an emerging area of interest that has shown promise in the prediction of drug disposition for transporter substrates, but with varying degrees of success. One complicating factor is that transporter proteins are not only found on the plasma membrane but also in the intracellular compartment, and only transporters on the plasma membrane surface can contribute to compound uptake to or efflux from the cell (Bow et al., 2008; Hayashi and Sugiyama, 2013). This indicates that the extrapolation of in vitro kinetics data to in vivo outcomes should consider differences in transporter abundance on plasma membranes, rather than on whole cell lysate or total membranes. Furthermore, a correction for plasma membrane recovery may be required to determine the absolute abundance of transporters due to the different matrices involved. In the present study, a simple and reliable method using quantitative proteomics was developed to accurately determine the plasma membrane expression of target proteins.
Enrichment of the plasma membrane fraction is advantageous when quantifying low abundance proteins such as transporters, which are often difficult to detect in the lysate or total membrane fractions from cells or tissues. These enrichment methods can increase the concentration of plasma membrane proteins while reducing background protein levels. This results in a large increase in overall LC-MS/MS signal from equivalent amounts of starting total protein. In this study, plasma membrane fractions were successfully enriched from both transporter-overexpressing cells and whole tissue samples (Fig. 1; Supplemental Table 4). In addition to enabling the successful protein quantification of OCT2 and MATE1, the quantitation of MATE2K was also possible, even though it had a much lower abundance in the plasma membrane fractions from human kidney cortex (assuming the same digestion, extraction, and ionization efficiencies between different surrogate peptides) (Table 2). This is in contrast to observations by Prasad et al. (2016), who were able to detect, but not quantify, MATE2K (below the lower limit of quantification) in the total membrane fraction from human kidney cortex. In the same study, OCT2 and MATE1 were successfully quantified, which is consistent with the higher abundance of these transporters compared with MATE2K on the plasma membrane fraction from human kidney cortex determined in the present study (Table 2).
OCT2 is localized on the basolateral membrane of renal proximal tubules, whereas MATE1 and MATE2K are on the apical membrane (Hillgren et al., 2013). A separate experiment demonstrated that the recovery is comparable between basolateral and apical membranes from human kidney cortex using the current methodology (Supplemental Table 5), thus supporting the use of a unified plasma membrane recovery number, as measured by Na+/K+-ATPase, to normalize the transporter expression regardless of cellular localizations. Mean plasma membrane recovery was comparable between different cells and tissues, and the transporter REF values did not change substantially (<2-fold change) when recovery was considered (Table 3). Although there was no significant impact in this study, correction may normalize variability when different matrices or methods are used. In addition, it is important to note that plasma membrane recovery was generally low (ranged from 6.5%–20%, except one replicate from HEK-OCT2, 41%; Supplemental Table 4), and correction would be necessary if the purpose is to determine absolute transporter abundance. Higher data variability is also likely if different methods are used to enrich the membrane proteins and recovery is not considered. These factors may contribute to the notable cross-laboratory differences (up to nearly 100-fold) observed for transporter quantitation (Harwood et al., 2014, 2016; Wegler et al., 2017).
In vitro kinetic parameters of metformin for OCT2, MATE1, and MATE2K were determined (Fig. 2; Table 4) and Km values in the present study are similar to those previously reported (present study versus literature range, μM); OCT2 (1872 versus 1465–3356), MATE1 (454 versus 228–780) and MATE2K (1840 versus 819–1980) (Tanihara et al., 2007; Song et al., 2008; Zolk et al., 2009; Shen et al., 2016). The Vmax values (nmol/min/mg protein) obtained do highlight experimental system-dependent differences, particularly for MATE1 and MATE2K; OCT2 (37.0 versus 12–153), MATE1 (15.6 versus 2.2–4.5), and MATE2K (16.9 versus 0.85–7.1). These observations emphasize the importance of determining REFs in the same experimental systems used for transport characterization. The differences in Vmax values are likely attributed to the difference in transporter abundance on the plasma membrane. As a result, caution should be exercised when using REF values reported in the literature for IVIVE.
In vitro kinetics of metformin for OCT2, MATE1, and MATE2K were translated to in vivo clearance using the corresponding REFs (corrected with plasma membrane recovery) in a PBPK model. The simulated exposure-time profiles of metformin at two doses (250 and 500 mg) correctly captured the elimination phase of metformin pharmacokinetics (Fig. 4). Furthermore, the predicted exposures were within 2-fold of clinically observed exposures (Table 5) (Somogyi et al., 1987; Wang et al., 2008). These results demonstrate that in vitro transporter kinetics can be accurately extrapolated to in vivo by using REFs derived from the plasma membrane expression of transporters.
Metformin is predominantly cleared from the systemic circulation by urinary excretion of unchanged drug via glomerular filtration and tubular secretion (Graham et al., 2011). Considering the lack of binding to plasma proteins (Tucker et al., 1981) and low lipid solubility, which likely leads to negligible passive reabsorption (Graham et al., 2011), the highest achievable renal clearance of metformin when the activities of secretory transporters are very significant can be approximated by the sum of glomerular filtration rate (7.2 l/h) and renal blood flow (60 l/h) (Neuhoff et al., 2013). It is worthwhile to note that the simulated renal clearance of metformin is close to the theoretically highest value (67.2 l/h) when the REFs for both OCT2 and MATE1 are greater than 3 (Fig. 3). The experimentally determined REFs for these transporters (5.5 and 14 before and 9.2 and 14 after the correction for plasma membrane recovery, respectively; Table 3) are in this range, and therefore, a 3- to 4-fold change in REF values is not expected to change the predicted renal clearance and resulting exposure of metformin.
The number of proximal tubule cells in human kidney cortex has not been well established (Scotcher et al., 2016). In the present study, 16 × 106 PTC/g of kidney cortex (average of four kidney preparations) (McLaren et al., 1995) was used to convert the transporter expression in human kidney cortex to the level at proximal tubule cells and calculate the REF values comparing to the expression in corresponding in vitro cell lines. Another literature reports that approximately 70 × 106 kidney cells primarily of proximal tubule origin were obtained from 1 g of cortical tissue (Cummings and Lash, 2000). Using this higher value in the analysis results in the REF values of 2.1, 3.2, and 0.30 for OCT2, MATE1, and MATE2K after the correction for plasma membrane recovery, respectively, while those were originally 9.2 (OCT2), 14 (MATE1), and 1.3 (MATE2K) (Table 3). The predicted AUCs of metformin using these REF values are also comparable to the clinical AUCs (data not shown), which is in line with the sensitivity analysis where metformin AUC is not sensitive to REFs in that range because of its already high renal clearance (Fig. 3). A compound with lower renal clearance mediated by these transporters would be required for additional validation of the experimental REFs. The current work does show that the quantification of renal drug transporters allows for the prediction of metformin disposition. Additional work will be required to determine whether this approach will be successful for other renally cleared drugs.
In conclusion, a novel approach was established to reproducibly quantify the transporter protein expression on plasma membranes in both cells and tissue. The plasma membrane recoveries determined by the quantification of Na+/K+-ATPase as the marker protein were comparable between HEK293 cells overexpressing OCT2, MATE1, and MATE2K and human kidney cortex in this study, suggesting that transporter REFs may be able to be estimated without the correction of plasma membrane recovery as long as the same procedure is applied between different matrices. Further studies are warranted to expand this observation to other transporters and in vitro/in vivo systems. In vitro transporter kinetics of metformin was successfully extrapolated to in vivo using the REFs for OCT2, MATE1, and MATE2K based on the plasma membrane transporter expression. These results demonstrate that the careful quantification of transporter proteins on plasma membrane enables prediction of human drug disposition using PBPK modeling.
Authorship Contributions
Participated in research design: Kikuchi, Chiou, Durbin, Savaryn, de Morais, Jenkins, Bow.
Conducted experiments: Chiou, Savaryn, Ma, Emami Riedmaier.
Performed data analysis: Kikuchi, Chiou, Durbin, Savaryn, Ma, Emami Riedmaier.
Wrote or contributed to the writing of the manuscript: Kikuchi, Chiou, Durbin, Savaryn, Ma, Emami Riedmaier, de Morais, Jenkins, Bow.
Footnotes
- Received March 29, 2021.
- Accepted July 6, 2021.
↵1 Current affiliation: Proteinaceous. Kenneth R. Durbin (Proteinaceous): Evanston, Illinois
↵2 Current affiliation: Bristol Myers Squibb; Drug Development and Preclinical Studies, DMPK; Research and Early Development. Arian Emami Riedmaier (BMS): Princeton, New Jersey
AbbVie sponsored and funded the study; contributed to the design; participated in collection, analysis, and interpretation of data; and in writing, reviewing, and approval of the final version.
All authors are employees, former employees or retirees of AbbVie and may own AbbVie stock.
↵This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- AUC
- area under the curve
- EV
- empty vector
- GGT1
- γ-glutamyl transpeptidase 1
- HEK
- human embryonic kidney
- IVIVE
- in vitro-in vivo extrapolation
- LC-MS/MS
- liquid chromatography/tandem mass spectrometry
- MATE
- multidrug and toxin exclusion protein
- Na+/K+-ATPase
- sodium-potassium adenosine triphosphatase
- OATP
- organic anion transporting polypeptide
- OCT
- organic cation transporter
- PBPK
- physiologically based pharmacokinetic
- PM
- plasma membrane
- PTC
- proximal tubule cells
- REF
- relative expression factor
- Copyright © 2021 by The American Society for Pharmacology and Experimental Therapeutics