Validation of a total IC50 method which enables in vitro assessment of transporter inhibition under semi-physiological conditions.

Abstract 1. Accurate predictions of clinical transporter-mediated drug–drug interactions (DDI) from in vitro data can be challenging when compounds have poor solubility and/or high nonspecific binding. Additionally, current DDI predictions for compounds with high plasma–protein binding assume that the unbound fraction in plasma is 0.01, if the experimental value is less than 0.01 or cannot be determined. This approach may result in an overestimation of DDI risk. To overcome these challenges, it may be beneficial to conduct inhibition studies under physiologically relevant conditions. 2. Here, IC50 values, determined in the presence of 4% bovine serum albumin approximating human plasma albumin concentrations, were successfully used to predict DDI for uptake transporters, OATP1B1/1B3, OCT1/2, OAT1/3 and MATE1/2K. 3. The IC50 values of reference inhibitors with 4% bovine serum albumin, considered total IC50, were comparable to the predicted values based on nominal IC50 values determined under protein-free conditions and unbound fraction in plasma. Calculation of R-total and Cmax/IC50,total values using total plasma exposure and total IC50 values explained the clinical DDI or absence of it for these inhibitors. 4. These results suggest that IC50 determinations in the presence of 4% albumin can be used, in the context of clinical total exposure, to predict DDI involving uptake transporters.


Introduction
Drug transporters are increasingly recognized as important vehicles to govern drug absorption, distribution, and excretion. In the last decade, following extensive review of nonclinical and clinical data, including pharmacogenomics, pharmacokinetics, and transporter-mediated drug-drug interactions (DDIs), multiple transporters of clinical relevance have been identified. These transporters include P-glycoprotein, breast cancer resistance protein, organic anion transporting polypeptide (OATP) 1B1, OATP1B3, organic cation transporter (OCT) 1, OCT2, organic anion transporter (OAT) 1, OAT3, multidrug and toxin extrusion (MATE) 1 and MATE2K (Giacomini et al., 2010;Tweedie et al., 2013). To assess and mitigate the potential risk of transporter-mediated DDIs, Drug-Drug Interaction guidance documents from regulatory agencies (US Food and Drug Administration (FDA), European Medicines Agency (EMA) and Pharmaceuticals and Medical Devices Agency (PMDA)) have been updated to include broader characterization of investigational drugs as transporter substrates or inhibitors during drug development (EMA, 2012;FDA, 2012;PMDA, 2014).
Assessment of inhibition potential is recommended for all investigational drugs because such interactions may alter the pharmacokinetics and potentially pharmacodynamics of coadministered transporter substrates. Multiple in vitro tools are available to determine the transporter inhibition including cell-and vesicle-based models overexpressing the transporter of interest (Brouwer et al., 2013). Generally, these inhibition studies (e.g. IC 50 determination) are conducted in a proteinfree environment so that the result represents the unbound IC 50 value (IC 50,unbound ) of test compound under the assumption that there is no significant loss of compound during assay incubation. IC 50,unbound can be used along with clinical unbound exposure to predict the likelihood of clinical inhibition (Giacomini et al., 2010). However, when an investigational compound has poor aqueous solubility, high plasma-protein and/or high nonspecific binding, there can be uncertainty in the actual effective concentration being studied in vitro. For compounds with poor aqueous solubility or high nonspecific binding, in vitro IC 50 values are considered nominal and may not necessarily represent the intrinsic unbound IC 50 values. It is quite possible that the IC 50 determinations under the standard protein-free condition may overestimate the intrinsic unbound IC 50 values, potentially leading to an underestimation of the DDI risk. Additionally, current DDI predictions for compounds with high plasmaprotein binding assume the unbound fraction in plasma (f u,p ) is 0.01 if the experimentally derived value is less than 0.01 or if it cannot be determined (EMA, 2012;FDA, 2012). If the actual f u,p is less than 0.01, defaulting to this assumption may result in an overestimation of the DDI risk, yield a false positive prediction and lead to an unnecessary clinical DDI study.
An apparent disconnect in the in vitro-in vivo correlation of transporter inhibition for a highly protein-bound compound has already been reported. Dolutegravir (>99% bound to plasma proteins (Bollen et al., 2015)) is an in vitro inhibitor of both OCT2 and MATE1 with apparent IC 50 values of 1.93 and 6.34 mM, respectively (GlaxoSmithKline, Reese et al., 2013). Following a 50 mg QD dose, the plasma maximum concentration of dolutegravir is 8 mM (Song et al., 2010). When the f u,p is assumed as 0.01, C u,max /IC 50 values for OCT2 and MATE1 are 0.041 and 0.013, respectively, suggesting low likelihood of clinical interactions with substrates for these transporters such as metformin. However, plasma exposures of metformin were significantly increased (1.8-2.5 fold) when co-administered with dolutegravir at 50 mg QD or BID (Zong et al., 2014). This indicates the overestimation of IC 50 value and accordingly underestimation of DDI risk when the transporter inhibition of dolutegravir is assessed in the standard protein-free condition.
The possible limitations in the currently available tools as well as an apparent in vitro-in vivo disconnect observed clinically warrant the development of a new experimental method, which enables in vitro assessment of transporter inhibition under more physiologically relevant conditions. One approach to refine DDI predictions, for challenging compounds, is to run in vitro IC 50 studies in the presence of 4% w/v albumin, which approximates human plasma albumin concentrations (Davies & Morris, 1993). IC 50 values determined under this condition can be regarded as ''total'' IC 50 values. Theoretically, these values can be compared with total, but not unbound, exposure of test compounds in DDI predictions assuming all plasma-protein binding is ascribed to albumin. The present study was aimed to validate this alternate approach in the DDI prediction for uptake transporters.

Uptake experiments
Cells were cultured as described previously (Chiou et al., 2014;Kikuchi et al., 2013). For transport studies, the cells were seeded in 24-well poly-D-lysine-coated plates and cultured for 2-3 days. Cells were washed three times with incubation buffer (Hanks Balanced Salt Solution (HBSS), supplemented with 10 mM HEPES, pH 7.4) at 37 C. For MATE1 and MATE2K studies only, the cells were instead pre-incubated with an acidification buffer (HBSS containing 10 mM HEPES and 30 mM NH 4 Cl; pH 7.4) at 37 C for 20 minutes. For the inhibition studies in the absence of proteins, uptake was initiated by adding incubation buffer containing radiolabeled probe substrate, 2 mM [ 3 H]E 2 17bG (0.5 mCi/mL) or 0. The concentrations of the other probe substrates were the same as those used in the inhibition studies in the absence of proteins. Uptake was terminated after a fixed incubation time by aspirating the incubation buffer and washing the cells three times with ice-cold HBSS. The incubation time was 5 minutes (E 2 17bG) or 2 minutes (pitavastatin) for OATP1B1 and OATP1B3, 2 minutes for OCT1, and 1 minute for OCT2, OAT1, OAT3, MATE1 and MATE2K. The uptake of probe substrates for hepatic transporters (OATP1B1, OATP1B3 and OCT1) (Supplemental Figure 1) and renal transporters (OCT2, OAT1, OAT3, MATE1 and MATE2K) (Kikuchi et al., 2013) were shown to be linear within these incubation times. Cells were then lysed in 200 mL of phosphate buffered saline containing 0.5% Triton X-100 with shaking for one hour at room temperature. Aliquots (150 mL) of cell lysate and 20 mL of incubation buffers were transferred to a 24-well micro-b counting plate containing 0.5 mL of scintillation cocktail. Radioactivity associated with the cells and incubation buffers was determined by Micro-b2 counter (PerkinElmer, Waltham, MA). The cell lysate protein concentrations were determined by the Pierce BCA Protein Assay (Thermo Scientific, Waltham, MA) according to the manufacturer's instructions. Uptake in each well was normalized by the corresponding protein concentrations and expressed as picomole per milligram protein. Transporter-specific uptake was obtained by subtracting the uptake into mock-transfected cells from that into transporter-expressing cells. IC 50 values were determined assuming competitive inhibition using the following equation: where X is log concentration of inhibitor, Y is % uptake of control, and Bottom is the lowest values of a fitted curve in GraphPad Prism (ver. 6.04) (GraphPad Software, La Jolla, CA). Mean and standard deviation (SD) of IC 50 values were calculated from three or four independent experiments each run in duplicate.

Results
Uptake of probe substrates in the presence of 100% human plasma or 4% BSA In order to determine and compare IC 50 values of test compounds under the same unbound concentration of probe substrates, the total concentration of E 2 17bG (OATP1B1/ OATP1B3) and ES (OAT3) was increased from 2 to 5 mM and from 0.1 to 6 mM, respectively. Probe substrate concentrations were determined by considering their plasma unbound fractions, 0.39 and 0.016, respectively (Giorgi & Crosignani, 1969;Rosenthal et al., 1972), and assuming similar binding in 4% BSA. No adjustment for probe substrate concentrations was made for PAH (OAT1) and MPP + (OCT1, OCT2, MATE1 and MATE2K) because the protein binding of these compounds was expected to be minimal. The fold difference values of the uptake of probe substrates between transporter-expressing cells and mocktransfected cells in the absence of inhibitors are summarized in Supplemental Table 1; these are at least 3.7-fold greater than that in mock-transfected cells, giving sufficient window for the subsequent IC 50 studies.
Comparison of OATP1B1 total IC 50 values between 4% BSA and 100% human plasma Total IC 50 values of reference inhibitors for OATP1B1 (rifampicin, rifamycin SV, indinavir and saqunavir) were determined in the presence of 4% BSA or 100% human plasma ( Table 1). The observed total IC 50 values were compared (within three-fold) whether using 4% BSA or 100% human plasma ( Figure 1). Representative IC 50 curves are shown in Figure 2 and Supplemental Figure 2. Due to the greater uptake window in the presence of 4% BSA than human plasma (Supplemental Table 1), 4% BSA was used in the subsequent total IC 50 studies for the other transporters.  Table 1). The IC 50 values determined in the presence of 4% BSA were higher than those determined in the absence of proteins. The unbound IC 50 values were extrapolated to total IC 50 values using the following equation (Table 1): IC 50, total, predicted ¼ IC 50, unbound =f u, p Predicted total IC 50 values (IC 50,total,predicted ) were compared with the observed total IC 50 values (IC 50,total ) ( Figure 3). The observed total IC 50 values of reference inhibitors in 4% BSA were within three-fold of the predicted values based on their unbound IC 50 and f u,p values.

OATP1B1 IC 50 value of rifampicin using a clinically relevant probe substrate pitavastatin
Additional IC 50 studies were conducted in order to validate the translatability of the total IC 50 method in predicting a clinical DDI. The IC 50 values of rifampicin, a clinically relevant inhibitor, on the OATP1B1-mediated uptake of pitavastatin were 0.56 mM and 4.1 mM in the absence and presence of 4% BSA, respectively ( Figure 4 and Table 1).

DDI prediction using total IC 50 values
Static DDI predictions for reference inhibitors on hepatic transporters (OATP1B1, OATP1B3 and OCT1) and renal transporters (OCT2, OAT1, OAT3, MATE1 and MATE2K) were conducted using either unbound or total IC 50 values. For hepatic uptake transporters, R-free and R-total were calculated using the following equations: where I in,max is the estimated maximum inhibitor concentration at the inlet to the liver and is equal to: C max + (k a Â Dose Â F a F g )/Q h . C max is the maximum systemic plasma concentration of inhibitor; Dose is the inhibitor dose; F a F g is the fraction of the dose of inhibitor which is absorbed; k a is the absorption rate constant of the inhibitor, and Q h is the estimated hepatic blood flow (1500 mL/min) (FDA, 2012). For renal transporters, unbound C max (C u,max )/IC 50,unbound and C max /IC 50,total values were calculated. A summary of the predicted inhibition of hepatic transporters and renal transporters by reference inhibitors (tested under both assay conditions) is listed in Tables 2 and 3, respectively. For hepatic uptake transporters, R-total values, which were derived from total plasma exposure and total IC 50 values, were comparable to the R-free values which are calculated using unbound exposure and unbound IC 50 values (Table 2). Likewise, C u,max /IC 50,unbound and C max /IC 50,total values were comparable for renal transporters (Table 3).
IC 50 determination of troglitazone in the absence or presence of 4% BSA Troglitazone is extensively bound in human plasma: f u,p equals 0.0011 (Shibukawa et al., 1995). The IC 50 value of troglitazone for OATP1B1 was determined in the absence and presence of 4% BSA ( Figure 5). Troglitazone inhibited OATP1B1 in the absence and presence of 4% BSA with IC 50 values of 0.32 ± 0.20 mM and 40 ± 16 mM (mean ± SD), respectively.

Discussion
The in vitro characterization of transporter interactions can be challenged by poor aqueous solubility and/or high nonspecific binding of a compound because there is an uncertainty in the effective concentration of test compound during the assay incubation, i.e. it can be lower than the nominal concentrations due to a potential loss of compounds by precipitation or nonspecific binding in the assay system (e.g. plastic wells). There are multiple approaches to improve transporter DDI characterization for these classes of compounds. For example, an IC 50 value may be estimated based on experimentally determined concentrations of solubilized compound in the assay, rather than nominal concentrations. Another approach may be to conduct an IC 50 study in the presence of proteins (e.g. albumin) or plasma and normalize the resulting IC 50 value (i.e. total IC 50 value) by the in vitro unbound fractions which are determined in a separate experiment. These approaches would be considered ideal because they may provide the intrinsic unbound IC 50 value even for a compound with poor aqueous solubility and/or high nonspecific binding. However, an accurate determination of aqueous solubility and (plasma) protein binding, which are required to provide a reliable DDI prediction in the current methodology, can be challenging for such compounds (Riccardi et al., 2015). In the present study, a new in vitro method which enables the in vitro assessment of transporter inhibition under semiphysiological conditions was developed and validated using reference inhibitors for uptake transporters. IC 50 values determined for rifampicin, rifamycin SV, indinavir and saqunavir against OATP1B1 in 4% BSA or 100% human plasma were comparable (Figure 1), suggesting that 4% BSA can be used as a surrogate for plasma. The IC 50 values of reference inhibitors for hepatic and renal transporters in the presence of 4% BSA were comparable to the predicted values based on their nominal IC 50 values determined under proteinfree conditions (i.e. unbound IC 50 ) corrected for unbound fraction in plasma (Table 1, Figure 3). Accordingly, the Rtotal and C max /IC 50,total values, which are considered DDI predictions using total plasma exposure of test compounds  instead of unbound exposure, were comparable to the R-free and C u,max /IC 50,unbound values, respectively (Tables 2 and 3).
Rifampicin, rifamycin SV and indinavir have been associated with clinical inhibition of OATP1B1 and/or OATP1B3. The area under plasma concentration-time curve (AUC) of OATP1B1 substrates such as atorvastatin, pravastatin, pitavastatin and rosuvastatin increased following co-administration of a single oral dose of 600 mg rifampicin (Prueksaritanont et al., 2014;Yoshida et al., 2012).
In addition, transient benign unconjugated hyperbilirubinemia has been observed clinically with these compounds, which has been ascribed to the inhibition of OATP1B1 and/ or OATP1B3 (Chiou et al., 2014). In contrast, while saquinavir is an OATP1B1 and OATP1B3 inhibitor in vitro, clinical inhibition has not been reported. Saquinavir did not increase the AUC of pravastatin after a 400 mg dose with co-administration of 400 mg ritonavir (Fichtenbaum et al., 2002) and no clinical hyperbilirubinemia has been reported (Chiou et al., 2014), suggesting that saquinavir does not reach the concentration sufficient to inhibit OATP1B1 and/or OATP1B3 clinically. The draft DDI guidance from FDA recommends that compounds should be investigated clinically as potential OATP1B1 or OATP1B3 inhibitors if the R-free value is greater than 1.25 (FDA, 2012). Indeed, Rfree values calculated in the present study successfully differentiated the clinical inhibition potential of inhibitors of OATP1B1 and/or OATP1B3 for rifampicin, rifamycin SV, indinavir and saquinavir using the cutoff value of 1.25 (Table  2). Importantly, R-total values can also explain a clinical inhibition or its absence for these inhibitors by applying the same cutoff value. R-total values were greater than 1.25 for rifampicin, rifamycin SV and indinavir, while it was borderline (1.3) for saquinavir (similar to the R-free value of 1.2) ( Table 2). Pyrimethamine is predicted to be a clinical inhibitor of OCT1 (R-free and R-total values greater than 1.25), however, this has not been examined clinically. For DDI predictions on renal transporters, the International Transporter Consortium recommends the cutoff value of 0.1 for C u,max /IC 50,unbound (Tweedie et al., 2013). In the present study, both C u,max /IC 50,unbound and C max /IC 50,total values were greater than 0.1 for probenecid on OAT1 and OAT3 and pyrimethamine on MATE1 and MATE2K (Table  3). This is consistent with known clinical DDIs between probenecid and OAT1 or OAT3 substrates such as famotidine, and between pyrimethamine and MATE1 or MATE2K substrate (metformin) (Morrissey et al., 2013). The translatability of total IC 50 method in predicting a clinical DDI was further explored using a clinically relevant combination of victim and perpetrator drugs. Rifampicin inhibited the OATP1B1-mediated uptake of pitavastatin in the presence of 4% BSA (Figure 4). The resulting R-total value (4.2) was greater than the cutoff value (1.25), consistent with the clinical DDI observed between these two compounds ( Table   Table 2. Static DDI prediction for hepatic transporters using unbound or total IC 50 values.

Transporter
Compound Mean unbound and total (4% BSA) IC 50 and f u,p values summarized in Table 1 were used in calculations.
2) (Prueksaritanont et al., 2014). Overall, either the unbound IC 50 values combined with unbound plasma exposure or total IC 50 values combined with total plasma exposure could explain the clinical transporter interactions for these reference inhibitors. These results suggest that IC 50 determinations in the presence of 4% albumin can be used in the prediction of transporter DDI by using clinical total plasma exposure in the calculation of R-total and C max /IC 50,total values. The in vitro method developed in the present study, the ''total IC 50 method'', may offer an advantage to the conventional protein-free method in the transporter characterization for compounds with poor solubility and/or high plasma protein-and nonspecific binding because: (1) the presence of albumin is expected to minimize nonspecific binding and uncertainty in the effective concentration in the test system, and (2) an accurate f u,p value is no longer required in DDI predictions as in vitro IC 50 values are determined in the presence of a physiological albumin concentration and are directly compared to total plasma concentrations. The total IC 50 method was applied for a highly-bound compound, troglitazone, of which f u,p equals 0.0011 (Shibukawa et al., 1995) (Figure 5). Troglitazone was confirmed as an OATP1B1 inhibitor with a nominal IC 50 value of 0.32 mM in the conventional protein-free assay condition. This value is 8-fold lower than the previously reported value (2.5 mM) (van de Steeg et al., 2013) even though the studies were conducted in almost identical experimental conditions; inhibition of [ 3 H]E 2 17bG uptake in HEK cells stably expressing OATP1B1 was examined. This may imply an uncertainty in the nominal IC 50 values potentially due to the nonspecific adsorption of troglitazone to assay apparatus, which could vary across different laboratories. Albumin plays a major role in the protein binding of troglitazone in plasma (Shibukawa et al., 1995). With the experimentally determined f u,p of 0.0011, the total IC 50 value of troglitazone for OATP1B1 was expected to be 291 mM. However, in contrast to the reference inhibitors used to validate the method, the observed total IC 50 value of troglitazone (40 mM) was significantly lower than the predicted value. This indicates a significant loss of compound in the absence of added protein, resulting in an overestimation of intrinsic unbound IC 50 value. The total IC 50 method may improve the DDI characterization for compounds with poor solubility and/or high plasma protein-and nonspecific binding like troglitazone.
In conclusion, the present study describes a new in vitro method, namely the total IC 50 method, which enables the in vitro assessment of transporter inhibition under semiphysiological conditions. IC 50 determinations in the presence of 4% bovine serum albumin can be used, in context of clinical total exposure, as an alternate to traditional proteinfree methods to predict clinical DDI involving uptake transporters.