Assessing OATP1B1- and OATP1B3-mediated drug-drug interaction potential of vemurafenib using R-value and physiologically based pharmacokinetic models

Abstract

Organic anion transporting polypeptides (OATP) 1B1 and OATP1B3 are important determinants of transporter-mediated drug-drug interactions (DDIs). Current studies assessed the OATP1B1 and OATP1B3-mediated DDI potential of vemurafenib, a kinase inhibitor drug with high protein binding and low aqueous solubility, using R-value and physiologically based pharmacokinetic (PBPK) models. The total half-maximal inhibitory concentration (IC_50,total) values of vemurafenib against OATP1B1 and OATP1B3 were determined in 100% human plasma in transporter-overexpressing human embryonic kidney 293 stable cell lines. The unbound fraction of vemurafenib in human plasma before (f_u,plasma) and after addition into the uptake assay plate (f_u,plasma,inc) were determined by rapid equilibrium dialysis. There was no statistically significant difference between f_u,plasma and f_u,plasma,inc. Vemurafenib IC_50,total values against OATP1B1 and OATP1B3 are 175 ± 82 and 231 ± 26 μM, respectively. The R-values [R = 1 + f_u,plasma × I_in,max/(f_u,plasma,inc × IC_50,total)] were then simplified as R=1+I_in,max/IC_50,total, and were 1.76 and 1.57 for OATP1B1 and OATP1B3, respectively. The simulated pravastatin AUC ratio was 1.28 when a single dose of pravastatin (40 mg) was co-administered with vemurafenib (960 mg, twice daily) at steady-state, compared to pravastatin alone. Both R-value and PBPK models predict that vemurafenib has the potential to cause OATP1B1- and OATP1B3-mediated DDIs.

INTRODUCTION

Organic anion transporting polypeptides (OATP) 1B1 and OATP1B3 are hepatic transport proteins that mediate the sinusoidal uptake of a variety of drugs (e.g. lipid lowering statins, anti-cancer drugs, and antibiotics). Inhibition of OATP1B1- and OATP1B3-mediated transport by co-administration with drugs that are OATP inhibitors (e.g. cyclosporine A^1,2) is often associated with an increase in the systemic exposure of OATP substrates and an elevated risk for adverse effects such as statin-related muscle toxicity or even rhabdomyolysis³. R-value and physiologically-based pharmacokinetic (PBPK) models have been utilized to assess the OATP-mediated DDI potential of new molecular entities and therapeutic drugs^4–6.

Kinase inhibitors are a fast-growing area in the drug development pipeline^7,8. Small-molecule kinase inhibitor drugs are designed to be administered orally long-term; therefore, understanding the potential for these drugs to cause DDIs with co-administered drugs that are OATP substrates, including the widely prescribed lipid lowering statins, has clinical significance³. Vemurafenib is a BRAF kinase inhibitor used for the treatment of metastatic melanoma with the BRAF V600E mutation⁹. Vemurafenib is a substrate and competitive inhibitor of P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP)^10,11. Co-administration of vemurafenib with digoxin, a P-gp substrate¹², increased digoxin systemic exposure by 1.82-fold^13,14. Currently, limited and apparently conflicting information exists regarding the inhibitory effects of vemurafenib on OATP1B1- and OATP1B3-mediated transport. Vemurafenib is documented as a non-inhibitor of OATP1B1 or OATP1B3¹¹ while the detailed information of this study is not publically available. Some reports, however, indicated that 10μM vemurafenib decreased OATP1B1-mediated [³H]-E₂17βG transport by ~75%¹⁵. Vemurafenib is a lipophilic compound with poor aqueous solubility and high protein binding (>99% protein bound in human plasma^11,16). Conventionally, the in vitro K_i values of inhibitor against OATP1B1 and OATP1B3 were determined in a protein-free assay condition as reported extensively^17–19. However, for compounds with poor aqueous solubility and high plasma protein and/or nonspecific binding, accurate determination of the unbound K_i values is often challenging. In addition, using the in vitro-determined unbound K_i values for model-based DDI prediction of such compounds also requires accurate determination of unbound fraction (f_u) in the systemic circulation. The f_u values determined by different methods (e.g. equilibrium dialysis, ultrafiltration, and ultracentrifugation) could be variable²⁰. Considering potential uncertainties in the protein binding measurements, when using the R-value model with compounds displaying high protein binding in human plasma (>99%), it is recommended in the US FDA guidance to use a fraction of unbound in the plasma (f_u,plasma) of 0.01 to avoid false negative DDI prediction⁶. However, this approach may lead to false positive or over-estimated DDI prediction²¹. Recently, a “total IC₅₀ method” was validated to determine the total half-maximal inhibitory concentration (IC_50,total) against OATP1B1, where IC_50,total was assessed in the presence of 4% bovine serum albumin (BSA) or 100% human plasma²². It is well known that plasma proteins can adsorb to the surface of many types of materials, potentially blocking non-specific binding^23,24. Hence, the presence of plasma proteins is expected to reduce the nonspecific binding of drugs to the assay apparatus during IC_{50, total} determination. Considering the inconsistent reports regarding the inhibitory effects of vemurafenib toward OATP1B in the literature and the low aqueous solubility and highly protein bound nature of vemurafenib, the inhibition potency of vemurafenib against OATP1B1 and OATP1B3 is worth re-evaluating using a “Total IC₅₀ approach”.

Human plasma is the most physiologically relevant matrix to determine the IC_50,total against OATP1B1- and OATP1B3. Hence, our current studies were designed to determine the inhibitory effects of vemurafenib on OATP1B1 and OATP1B3 in 100% human plasma. The unbound fraction of vemurafenib in human plasma before (f_u,plasma) and after adding into the uptake assay apparatus (a 24-well plate) (f_u,plasma,inc) were determined by rapid equilibrium dialysis. The f_u,plasma,inc was used to estimate the IC_50,unbound from the determined IC_50,total values against OATP1B1 and OATP1B3. OATP1B1- and OATP1B3-mediated DDI potential of vemurafenib were subsequently assessed using R-value and PBPK models.

MATERIALS AND METHODS

Materials.

[³H]-Cholecystokinin 8 (CCK-8) (specific activity 88.0 Ci/mmol) and [³H]-estradiol 17 β-D-glucuronide (E₂17βG) (specific activity 41.4 Ci/mmol) were purchased from Perkin Elmer Life Science (Waltham, MA). Vemurafenib was purchased from LC laboratories (Woburn, MA, USA). Unlabeled CCK-8, E₂17βG, Triton X-100, dimethyl sulfoxide (DMSO), Hank’s Balanced Salt Solution (HBSS), Dulbecco’s Modified Eagle Medium (DMEM), trypsin-EDTA solution, and antibiotic antimycotic solution were purchased from Sigma-Aldrich (St. Louis, MO). Poly-L-lysine was purchased from Trevigen Inc. (Gaithersburg, MD). Geneticin® and HEPES was purchased from BD Biosciences (Bedford, MA). Fetal bovine serum (FBS) was purchased from Hyclone Laboratories (Logan, Utah). Bio-Safe II liquid scintillation mixture was purchased from Research Products International (Mt. Prospect, IL). Human plasma was purchased from BioIVT (Hicksville, NY, Lot#BRH895613).

Cell culture.

Human embryonic kidney (HEK) 293 stable cell lines over-expressing OATP1B1 (HEK293-OATP1B1) or OATP1B3 (HEK293-OATP1B3) and the HEK293-Mock cell line were generously provided by Dr. Dietrich Keppler^25,26. All HEK293 stable cell lines were cultured in DMEM containing 10% FBS and 600 μg/ml G418 as published previously^19,25,26.

Determining unbound fraction of vemurafenib in human plasma.

Vemurafenib (150 mM in DMSO) was spiked into 100% human plasma (pH 7.4) at concentrations of 5, 150, 350, and 500 μM. Rapid equilibrium dialysis (RED) (Thermo Fisher Scientific, Waltham, MA) was used to determine the unbound fraction of vemurafenib in human plasma before (f_u,plasma) and after (f_u,plasma,inc) being added to the 24-well uptake assay plate. To determine f_u,plasma, human plasma (100 μl) containing vemurafenib (5–500 μM) or vehicle control (0.1% DMSO) was added to sample chambers and phosphate buffered saline (PBS) buffer (350 μL) was added to buffer chambers of the RED device in triplicate. To determine f_u,plasma,inc of vemurafenib in the uptake assay plate, the 24-well culture plate was coated with poly-L-lysine similarly as for uptake assays¹⁹. After washing once the culture plate with 200 μL HBSS buffer, 200 μL human plasma (pH 7.4) containing vemurafenib (5–500 μM) or vehicle control was added into the wells of the culture plate and was incubated at 37°C for 3 and 30 min, respectively. Experiments were conducted in triplicate. At the end of incubation, 100 μL of human plasma containing vemurafenib or vehicle control was added to sample chambers and 350 μL PBS buffer was added to buffer chamber of the device. The RED plate was then sealed and incubated at 37°C for 4 hours on an orbital shaker (Incubator Shaker Series 25, New Brunswick Scientific, Edison, NJ) at 250 revolutions per minute following manufacture instruction. Equal volume of post-dialysis samples from the plasma and buffer chambers were used to determine vemurafenib concentration by liquid chromatography tandem-mass spectrometry (LC-MS/MS). The unbound fraction in human plasma is determined as the fraction of unbound vemurafenib concentration in the buffer chamber over the total concentration of vemurafenib in the plasma chamber as shown in Table 1.

Table 1.

Unbound fraction of vemurafenib in 100% human plasma before (f_u,plasma) and after (f_u,plasma,inc) adding into the 24-well uptake assay apparatus. The 24-well plate was with or without incubation with 100% human plasma containing vemurafenib at indicated concentrations for 30 min (triplicate each concentration). f_u,plasma and f_u,plasma,inc was determined as described in the Materials and Methods. The one-way ANOVA was used to determine the statistically significant difference of f_u,plasma values in each group (with or without incubation) among vemurafenib concentrations of 5–500 μM. At each concentration, Student’s t-test was used to determine the statistically significant difference between f_u,plasma and f_u,plasma,inc values. Data represent mean ± SD.

Vemurafenib concentrations (μM)	fu,plasma (×10−6)	fu,plasma,inc (×10−6)
5	10.2 ± 3.8	13.8 ± 4.5
150	8.7 ± 1.5	9.4 ± 2.7
350	7.9 ± 0.9	11.7 ± 2.5
500	10.2 ± 1.7	13.1 ± 4.1
Mean ± SD (of 4 cone.)	9.3 ± 1.2	12.0 ± 1.9
Average fu values from all concentrations	0.00001

LC-MS/MS analysis of vemurafenib.

The stable isotope labeled vemurafenib-d7 (Toronto Research Chemicals Inc, North York, Canada) was used as internal standard. The PBS buffer (pH 7.4) containing vemurafenib was diluted in a 1:1 ratio (v/v) with internal standard in methanol prior to LC-MS/MS analysis. The plasma samples were extracted and diluted with the internal standard (in methanol) solution. The ultra-high performance liquid chromatography (UHPLC) mobile phases A and B consist of 0.16% (v/v) heptafluorobutyric acid (HFBA) in water and acetonitrile, respectively. Vemurafenib were chromatographically separated on an ACQUITY UPLC HSS T3 1.8 μm, 100 × 2.1 mm column (Waters Corporation, Milford, MA) through a gradient elution by increasing mobile phase B from 60% to 90% in 2 minutes at a flow rate was 0.35 mL/min. Vemurafenib and vemurafenib-d7 were quantified by monitoring m/z 490>255 and 497>255 on a Thermo Scientific TSQ Vantage triple quadruple mass spectrometer equipped with Thermo Ultimate 3000 UHPLC system (Thermo Fisher Scientific, Waltham, MA). The LC-MS/MS system was operated in a positive ion mode.

Transport studies in HEK293 stable cell lines.

Cells (20–30 passages since the culture received from Dr. Keppler) were seeded at a density of 1.5 × 10⁵ cells per well in poly-L-lysine coated 24-well plates, and allowed to grow for 48 h before the transport studies. E₂17βG and CCK-8 were used as probe substrates for OATP1B1 and OATP1B3 in HEK293-OATP1B1 and HEK293-OATP1B3 cells, respectively. Accumulation of [³H]-E₂17βG (1 μM, 2 min) and [³H]-CCK-8 (1 μM, 3 min) in the protein-free condition was conducted similarly as published previously with slight modification²⁷. Cells were washed three times with HBSS buffer containing 10 mM HEPES (pH 7.4). After washing, substrate accumulation was determined in the presence of vehicle control (0.1% DMSO) or vemurafenib at highest aqueous solubility (0.78 μM)¹¹ in HBSS buffer supplemented with 10 mM HEPES (pH 7.4). For transport studies in 100% human plasma (pH 7.4), initial studies were conducted to determine linear uptake range of E₂17βG (5 μM) and CCK-8 (1 μM) in HEK293-OATP1B1 and HEK293-OATP1B3 cell lines, respectively, after subtracting accumulation in HEK293-Mock cells at each time points. Substrate incubation times fell within the linear uptake ranges determined in the current studies or published previously¹⁹.

To determine the total IC₅₀ of vemurafenib against OATP1B1 and OATP1B3, HEK293-OATP1B1 and HEK293-OATP1B3 were washed three times with HBSS buffer (pH 7.4), followed by incubation with [³H]-E₂17βG (5 μM, 3 min) or [³H]-CCH-8 (1 μM, 3 min), respectively, in the presence of vehicle control (0.1% DMSO) or vemurafenib (0.5 – 500 μM). At the end of the incubation with radiolabeled probe substrates, cells were rinsed three times with ice-cold HBSS buffer and solubilized in ice-cold PBS containing 0.5% Triton-X-100. An aliquot was subjected to liquid scintillation counting (LS6500 scintillation counter, Beckman Coulter, Brea, CA), and a BCA assay (Pierce Chemical, Rockford, IL) was performed to determine sample protein concentrations. Substrate accumulation was then normalized to the protein concentrations. Uptake studies were also conducted in poly-L-lysine-coated blank plates to correct for nonspecific binding.

To estimate the IC₅₀ values of total concentrations of vemurafenib against OATP1B1 (IC_{50,total,OATP1B1}) and OATP1B3 (IC_{50,total,OATP1B3}) in the presence of human plasma, substrate accumulation was expressed as percentage of the vehicle control and plotted against total concentrations of vemurafenib. Nonlinear regression was used to estimate the IC_50,total values by fitting a three-parameter model (Eq. 1) to the data using GraphPad Prism v.7.04 (GraphPad Software, La Jolla, CA) similar as published previously²⁷.

$$\text{E}=\text{Bottom}+\frac{\text{ Top - Bottom }}{\left(1+\frac{\text{C}}{{\text{IC}}_{50}}\right)}$$ Eq. 1

Where E is the remaining OATP1B1- or OATP1B3-mediated substrate transport at a given inhibitor concentration (C). IC₅₀ is the inhibitor concentration resulting in a response halfway between the maximal (top) response and the maximally inhibited (bottom) response. The unbound IC₅₀ values of vemurafenib (IC_50,unbound) were determined as the product of IC_50,total and f_u,plasma,inc. IC₅₀ values were expressed as mean and standard deviation (SD) from three or four independent experiments in triplicate as indicated in the legend.

Considering the reported protein binding of E₂17βG (f_u=0.39)²⁸ and CCK-8 (f_u=0.56)²⁹, the anticipated unbound substrate concentrations of E₂17βG (~ 2 μM) and CCK-8 (0.56 μM) in the uptake assay are within four- to six-fold the reported apparent K_m values of E₂17βG transport by OATP1B1 (~8.3 μM) and CCK-8 transport by OATP1B3 (~3.8 μM)^19,30,31; thus the unbound IC₅₀ values estimated in current studies would be approximately the same as the inhibition constant (K_i) values.

Prediction of OATP-mediated DDIs using the R-value model.

R-values, which represent the predicted area under the concentration-time profile (AUC) ratio of the victim drug in the presence vs. absence of an investigational drug, were calculated based on Eq. 2^6,32.

$$\text{R}=1+\left({\text{f}}_{\text{u,plasma }}\times \frac{{\text{l}}_{\text{in,max}}}{{\text{K}}_{\text{i }}}\right)$$ Eq. 2

I_in,max is the estimated total maximum plasma concentration of the inhibitor at the inlet to the liver. f_u,plasma is the unbound fraction of the vemurafenib in human plasma determined in our current study. As K_i approximates IC_50,unbound, which is the product of f_u,plasma,inc and IC_50,total, R value is expressed in Eq. 3

$$\text{R}=1+{\text{f}}_{\text{u,plasma}}\times \frac{{\text{l}}_{\text{in,max}}}{{\text{f}}_{\text{u,plasma,inc}}\times {\text{IC}}_{50,\text{total}}}$$ Eq. 3

The I_in,max is calculated as:

$${\text{I}}_{\text{in},\text{max}}={\text{I}}_{\text{max}}+\frac{{\mathrm{f}}_{\text{a}}{\mathrm{F}}_{\text{g}}\times {\mathrm{k}}_{\text{a}}\times \text{Dose}}{{\mathrm{Q}}_{\text{h}}\times {\mathrm{R}}_{\text{B}}}$$ Eq. 4

Where I_max is the maximum plasma concentration (C_max) of the vemurafenib in systemic circulation, which is 125.3 μM following the approved highest doses of 960 mg³³. The fraction absorbed (f_a) and fraction escaping gut metabolism (F_g) values are unknown, thus f_aF_g = 1 was used as a worst-case estimate. Q_h is the hepatic blood flow rate (1500 ml/min³⁴). R_B is the blood-to-plasma concentration ratio. The mean R_B for [¹⁴C]-vemurafenib was observed to be 0.58 ± 0.03¹¹. Following oral administration, the absorption rate constant (k_a) of vemurafenib was estimated to be 0.19 h⁻¹ in a population pharmacokinetic analysis^11,13.

Pharmacokinetic modeling and simulations.

A PBPK modeling approach was utilized to evaluate the OATP1B1- and OATP1B3-mediated DDI potential of vemurafenib against the OATP-substrate pravastatin using the population-based Simcyp simulator (version 17 release 1, Certara UK, Simcyp division, Sheffield, UK). The default pravastatin (Sim-Pravastatin)¹⁹ and digoxin³⁵ PBPK models within the Simcyp simulator library were used without modification as published previously. The DDI performance of the default Simcyp pravastatin model in OATP1B1- and OATP1B3- mediated DDI predictions was verified in our previous publication¹⁹. Using K_i values of rifampicin against OATP1B1 and OATP1B3 determined in our previous study¹⁹, the simulated AUC ratio (AUCR) and C_max ratio (C_maxR) of pravastatin upon co-administration of rifampicin vs. pravastatin alone are 2.49 (trial range 2.33–2.58) and 2.92 (trial range 2.69–3.30), respectively, which are comparable to the reported clinical values of 2.33 and 2.7, respectively³⁶. A PBPK model of vemurafenib was developed as a perpetrator file based on the parameters summarized in Table 2.

Table 2.

Parameter values used in the vemurafenib PBPK model

Vemurafenib	Values	Reference
Dose (mg)	960	9
Chemical Structure
Physicochemical properties
MW (g/mol)	489.93	11
Log PO:W	3.0	11
Compound type	Diprotic acid	11
pKa	7.9, 11.1	11
Blood-to-plasma ratio	0.58	11
Hematocrit value (%)	45	Default value
fu,p	0.00001	See Table 1 and 3
Absorption [First-order absorption model]
fa	0.69	Predicted using Caco-2 data
ka (1/h)	0.19	11
Qgut (L/h)	6.95	Predicted using Caco-2 inputs
fu.gut	1	Assumed value
Caco-2 permeability [Papp,caco-2 (7.4:7.4, passive) (10−6 cm/s)]	9.7	Corrected value (see text in methods or details)11,38
Reference [Papp,caco-2 (7.4:7.4) (10−6 cm/s)]	4	cimetidine was used as the reference compound
Peff,man (×10−4 cm/s)	1.12	Predicted
Distribution (minimal PBPK)
Vss (L/kg, CV%)	0.9 (6.67)	User input, see details in methods11,16
Elimination
CLpo (L/h, CV%)	2.67 (53.5)	45
CLR (L/h)	0.01787	See details in methods
Interaction
P-gp		0.00017 μM (see details in methods)
oatp1b1		0.0017 μM (Table 3)
OATP1B3		0.0023 μM (Table 3)

Log P_o:w :neutral species octanol:buffer partition coefficient

Q_gut: Flow rate for overall delivery of drug to the gut (drug dependent)

f_u,gut: Unbound fraction of drug in enterocytes

P_eff,man: human jejunum effective permeability

Sources for the propagated variability of system parameters are summarized in Table S5

Vemurafenib is a diprotic acid with an n-octanol:water partition coefficient (Log P_O:W) of 3 and pK_a dissociation constants of 7.9 and 11.1¹¹. Zimmerman et al. reported that vemurafenib is not a substrate of OATP1B1³⁷. The uptake of vemurafenib in OATP1B3-overexpressing cells was ~1.5 fold of that in vehicle control cells³⁷, which did not meet the recommended cut-off value of ≥2-fold recommended in the US FDA draft guidance for in vitro DDI studies for a drug to be considered as an in vitro substrate of OATP1B1 or OATP1B3⁶. Hence, a liver-plasma concentration ratio (K_p,liver) value of 1 is assumed as a conservative default value for vemurafenib since there is no reported evidence of vemurafenib hepatic uptake to the best of our knowledge. Vemurafenib passive permeability coefficient in Caco-2 cell monolayers was reported to be 2.9× 10⁻⁶ cm/s using ranitidine as the reference compound³⁸. However, to predict human jejunum effective permeability (P_eff,man) of vemurafenib, cimetidine was used as the reference compound in the current study. Therefore, the vemurafenib passive permeability in Caco-2 cells was corrected as 9.7 × 10⁻⁶ cm/s, which is the product of the observed vemurafenib value (2.9 × 10⁻⁶ cm/s) and the ratio of Simcyp library value of cimetidine (4 ×10⁻⁶ cm/s) over that of ranitidine (1.2 ×10⁻⁶ cm/s).

Vemurafenib is a substrate of P-gp and BCRP¹⁰. For the current purpose of using vemurafenib as a perpetrator drug in assessing OATP1B1- and OATP1B3-mediated DDI, a first-order absorption model was used, and therefore the clearance of vemurfenib by P-gp and BCRP was not considered. Vemurafenib is an inhibitor of P-gp and BCRP^10,11. The K_i for vemurafenib against BCRP has not been reported. A clinical DDI study reported that the AUCR and C_maxR of digoxin are 1.82 and 1.47, respectively, when digoxin was coadministered with vemurafenib compared to digoxin alone^13,14. The reported IC₅₀ value of vemurafenib against digoxin is 17 μM¹¹ from an in vitro assay where vemurafenib concentrations were upto 50 μM¹¹. No information was provided regarding if there were proteins added into the assay system¹¹. As this reported IC₅₀ value of vemurafenib (17 μM) is much greater than the aqueous solubility of vemurafenib (0.78 μM)¹¹, this value is unlikely the unbound IC₅₀ against P-gp. We assume this reported IC₅₀ against P-gp is the IC_50,total, and subsequently estimated the IC_50,unbound against P-gp as 0.00017 μM after multiplying the f_u,plasma,inc of 0.00001 (Table 1) determined in our current studies.

Vemurafenib bioavailability (F) and volume of distribution at steady state (V_ss)/F were estimated to be 0.788–1¹¹ and 1.3 L/kg (70 kg body weight)¹⁶, respectively, from population pharmacokinetic studies. The f_a and F_g values of vemurafenib from simulation output of 960 mg vemurafenib twice daily (BID) are 0.67 and 1, respectively. The product of f_a and F_g, 0.67, was assumed as the bioavailability (F) of vemurafenib, yielding a V_ss value of 0.9 L/kg for the current model. The renal clearance of vemurafenib was set to be 0.0178 L/h based upon the fact that 0.97% of the oral vemurafenib dose was recovered in urine as parent¹¹. Following multiple dosing of vemurafenib at the approved 960 mg BID, steady state was reached at around 15–21 days^11,33. Vemurafenib has extensive accumulation, AUC_{0–8 h} of which is 23.2 ± 16.5 fold higher on day 15 than on day 1³³. The DDI potential of vemurafenib against pravastatin was simulated using a trial design which comprised of a 960 mg vemurafenib dose administered BID for 21 days with a single-dose 40 mg pravastatin co-administered on day 21. The DDI simulation of vemurafenib against pravastatin was performed in 400 virtual subjects (20 trials × 20 subjects) using the default Sim-Healthy Volunteer data library. In some of the steady-state clinical trials³³, vemurafenib could be taken with or without food during the study; fasting was required only for the day of PK sample collection. Hence, all simulations involving vemurafenib were conducted under fed condition.

Given the fact that pravastatin is minimally metabolized through CYP450 enzymes³⁹, no CYP metabolism or inhibition of CYP enzymes was included in the model for either pravastatin or vemurafenib^39,40. MRP2 contributes primarily to the hepatobiliary excretion of pravastatin^41–43; however, inhibition of MRP2 by vemurafenib has not been reported. The DDI simulation of vemurafenib against pravastatin was performed assuming that vemurafenib affects pravastatin disposition only via inhibition of OATP1B1- and OATP1B3-mediated transport of pravastatin on a fit-for-purpose basis.

Data Analysis.

For the statistical analysis in Fig. 3, linear mixed effects models were used to estimate fold changes and associated standard errors (SE) of substrate accumulation in treatment group(s) vs. control for triplicate data with a fixed group effect and a random effect (experiment date), adjusting for group-specific variances similar to that published previously^19,44. One-way analysis of variance (ANOVA) and Student’s t-test was used to determine the statistically significant difference in f_u,plasma as detailed in legend of Table 1. A two-sided p-value of <0.05 defines statistical significance. Data analyses were conducted using the SAS software (version 9.3, Cary, NC).

Fig. 3. Effects of vemurafenib on OATP1B1- and OATP1B3-mediated transport in HBSS buffer and in 100% human plasma.

(A) [³H]-E₂17βG accumulation in HEK293-OATP1B1 cells. In HBSS buffer containing 10 mM HEPES (pH 7.4), [³H]-E₂17βG accumulation (2.5 μM, 2 min) was determined in the presence of vehicle control or 0.78 μM vemurafenib. In 100% human plasma (pH 7.4), [³H]-E₂17βG accumulation (5 μM, 3 min) was determined in the presence of vehicle control or 500 μM vemurafenib. (B) [³H]-CCK-8 accumulation in HEK293-OATP1B3 cells. [³H]-CCK-8 accumulation (1 μM, 3 min) was determined in HBSS buffer containing 10 mM HEPES (pH 7.4) in the presence of vehicle control or 0.78 μM vemurafenib or in 100% human plasma (pH 7.4) in the presence of vehicle control or 500 μM vemurafenib. Linear mixed effects models were fit to the data (n=3 in triplicate). * indicates a statistically significant difference (p<0.05 vs. CTL).

RESULTS

Time-dependent uptake of [³H]-E₂17βG and [³H]-CCK-8 in HEK293-OATP1B1 and −1B3 stable cell lines.

In 100% human plasma, OATP1B1-mediated accumulation of [³H]-E₂17βG (5 μM) was linear up to at least 3 min, and OATP1B3-mediated accumulation of [³H]-CCK-8 (1 μM) was linear up to at least 5 min (Fig. 1). Subsequent accumulation of [³H]-E₂17βG (5 μM) and [³H]-CCK-8 (1 μM) were determined at 3 minutes in HEK293-OATP1B1 and HEK293-OATP1B3 cells, respectively.

Fig. 1. Time-dependent accumulation of OATP1B1- and OATP1B3-mediated substrate transport in the presence of human plasma.

Time-dependent accumulation of [³H-]E₂17βG (5 μM) (A) and [³H]-CCK-8 (1 μM) in the presence of 100% human plasma (pH 7.4). OATP1B1-mediated [³H]-E₂17βG accumulation and OATP1B3-mediated [³H]-CCK-8 accumulation were determined by subtracting the accumulation values obtained in HEK293-Mock cells from those obtained in HEK293-OATP1B1 (A) and HEK293-OATP1B3 (B) cells, respectively, under the same conditions. Data represent the mean ± SD in triplicate from a single experiment.

Inhibitory effects of vemurafenib on OATP1B1 and OATP1B3 in 100% human plasma

Co-incubation with vemurafenib at total concentrations of 0.5–500 μM in 100% human plasma (pH 7.4) yielded an IC_50,total value of 174.6 ± 82.0 (mean ± SD, n=3) and 231.3 ± 26 μM (mean ± SD, n=4) against OATP1B1 and OATP1B3, respectively (Fig. 2, Table 3). Rifampicin has been reported to inhibit OATP1B1-mediated E₂17βG transport at a IC_50,total of 2.4 ± 0.2 μM²². Rifampicin was used as a positive control in the current study for inhibition of OATP1B1 in human plasma. The current study used HEK293-OATP1B1 cells from the same source²⁶, and used human plasma from the same vendor and with the same specifics (information obtained by personal communication) as the previous study²². As shown in supplemental Fig. S1, inhibition of OATP1B1 by rifampicin yielded a IC_50,total of 2.1 μM, which is comparable to the previously reported value of 2.4 μM²².

Fig. 2. Inhibitory effects of vemurafenib on OATP1B1 and OATP1B3 in 100% human plasma.

(A) Accumulation of [³H]-E₂17βG (5 μM, 3 min) in HEK293-OATP1B1 cells. (B) Accumulation of [³H]-CCK-8 (1 μM, 3 min) in HEK293-OATP1B3 cells. Substrate accumulation was determined in the presence of vehicle control or vemurafenib (0.5 – 500 μM) in 100% human plasma (pH 7.4), and is expressed as percentage of vehicle control. Lines represent the fitted lines of IC₅₀ by nonlinear regression analysis. Data represent mean ± SD from n=3 experiments for OATP1B1 and n=4 experiments for OATP1B3.

Table 3.

IC₅₀ values of vemurafenib against OATP1B1 and OATP1B3 and predicted AUC or C_max ratios of OATP1B1- and OATP1B3-mediated DDIs using R-value and PBPK models. IC_50,total values are mean±SD from n=3 and 4 experiments for OATP1B1 and OATP1B3, respectively. For the PBPK model, the AUCR and C_max ratio were predicted from the DDI simulation described in Fig. 6 using pravastatin as substrate.

Transporter	Iin,max (μM)	Observed IC50,total (μM)	fu,plasma = fu,plasma,inc	Estimated IC50,unbound (μM)	R-values	PBPK model
						AUCR	Cmax ratio
0ATP1B1	132.5	174.6 ± 82	0.00001	0.0017	1.76	1.28	1.26
0ATP1B3		231.6 ± 26		0.0023	1.57

Determining f_u,plasma,inc and f_u,plasma

As shown in Table 1, the unbound fractions of vemurafenib in human plasma before (f_u,plasma) and after (f_u,plasma,inc) adding into the 24-well uptake assay apparatus ranged 7.9 ± 0.9 – 10.2 ± 1.7 and 9.4 ± 2.7–13.8 ± 4.5 ×10⁻⁶, with mean values of 9.3 ± 1.2 and 12.0 ± 1.9 ×10⁻⁶, respectively. There is no significant difference in unbound fractions either among concentrations within each group or between two groups at each concentration. Hence, an average unbound fraction value of 0.00001 from all concentrations with and without incubation in the 24-well uptake apparatus was used for both the f_u,plasma and f_u,plasma,inc in the current study. The IC_50,unbound values against OATP1B1 and OATP1B3 were estimated by multiplying the IC_50,total values against OATP1B1 (174.6 μM) and OATP1B3 (231.3 μM) with the f_u,plasma,inc, yielding 0.0017 and 0.0023 μM for OATP1B1 and OATP1B3, respectively. We also tested that there is no time-dependent changes in the incubation time between 3 min (duration of substrate transport in uptake assay) and 30 min at 350 μM total vemurafenib concentration in human plasma as shown in Fig. S2.

Effects of vemurafenib on OATP1B1- and OATP1B3-mediated transport in HBSS buffer.

Effects of vemurafenib at maximum aqueous solubility (0.78 μM) on OATP1B1-mediated transport of [³H]-E₂17βG (2.5 μM, 2 min) and OATP1B3-mediated transport of [³H]-CCK-8 (1 μM, 3 min) were determined in HBSS buffer (pH 7.4) containing 10 mM HEPES in HEK293-OATP1B1 and HEK293-OATP1B3, respectively. As shown in Fig. 3, co-incubation with vemurafenib (0.78 μM) does not affect OATP1B1- or OATP1B3-mediated transport compared to vehicle control. As a comparison, in the same experiment, effects of vemurafenib (500 μM) on OATP1B1- and OATP1B3-mediated transport were determined in human plasma. Similarly as shown in Fig. 2, in 100% human plasma, co-incubation with vemurafenib (500 μM) significantly decreased OATP1B1- and OATP1B3-mediated transport of [³H]-E₂17βG (5 μM, 3 min) and [³H]-CCK-8 (1 μM, 3 min) to 0.38 ± 0.1 and 0.13 ± 0.04 fold of control, respectively (p<0.05).

Assessing OATP1B1- and OATP1B3-mediated DDI potential of vemurafenib using R-value and PBPK models.

Following Eq. 4, I_in,max of vemurafenib is estimated to be 132 μM (Table 3). Because there is no statistically significant difference between f_u,plasma and f_u,plasma,inc determined in the current studies (Table 1), Eq. 3 was simplified as R=1+I_in,max/IC_50,total. The R-values for vemurafenib against OATP1B1 and OATP1B3 are 1.76 and 1.57, respectively (Table 3).

The vemurafenib PBPK model well described the plasma concentration-time profile of vemurafenib administered following a single oral 960 mg in patients (Fig. 4 A), which is the clinical data used to develop the model⁴⁵. The PBPK model was then verified using all other vemurafenib clinic data available to the best of our knowledge (Fig. 4 B–D, Fig. S3–S7 and Supplemental Table S1 and S2). It has been documented that the AUC and C_max of vemurafenib are dose-proportional over the dose range of 240, 480, 720, and 960 mg BID oral dose on day 15 steady-state¹¹. The current vemurafenib PBPK model well describes all available steady-state concentration-time profiles of vemurafenib (240–960 mg, BID). The model also decently described the majority of the concentration-time profiles following the first doses of vemurafenib (240–960 mg).

Fig. 4. Simulated and observed plasma concentration-time profiles of vemurafenib following oral administration.

(A) Vemurafenib 960 mg single dose and (B) vemurafenib 960 mg twice daily for 15 days. The grey lines represent the simulated individual trials (20) of 15 subjects using a population of 300 virtual subjects (40% female, 34–65 years) (A) and 20 trials of 16 subjects using a population of 320 virtual subjects (50% female, 39–62 years) (B). The black thin lines represent the upper (95^th) and lower (5^th) percentiles and the black thick line represents the simulated mean of the healthy volunteer population. The closed circles denote the observed values from the clinical study by Ribas et al., 2014 (A) and Grippo et al., 2014 (B). C and D are the same simulation results and clinical data as in B but displayed on different scales of x-axis.

The vemurafenib model was then verified by an independent clinical DDI study of vemurafenib against digoxin¹⁴. To the best of our knowledge, this is the only available clinical DDI study involving vemurafenib that evaluated transporter-mediated interactions. The digoxin PBPK model decently described the reported plasma-concentration-time profile of digoxin following a 0.25 mg single oral dose with and without vemurafenib co-administration (Fig. 5 A and Fig. S8). The simulated AUC0–168h ratio and C_maxR of a single dose of digoxin when coadministered with 960 mg BID vemurafenib at steady-state, compared with digoxin alone, were 1.46 (trial range 1.30–1.68) and 1.45 (trial range 1.33–1.55), respectively (Fig. 5 A), compared to the reported AUC_lastR (up to 168 h after digoxin administration) and C_maxR of 1.82 and 1.47, respectively^13,14. The simulated and observed AUC0–24h ratios of digoxin, when co-administered with vemurafenib compared to digoxin alone, were 1.37 (trial range 1.26–1.50) and 1.38¹⁴, respectively. The vemurafenib PBPK model modestly predicts the clinical DDI of vemurafenib against digoxin with slight underestimation of the AUC_last ratio, which is due to the fact that the mean terminal half-life after a single dose of digoxin was increased from 35.0 to 56.4 hours by co-administration with vemurafenib¹⁴. The overestimation of the terminal clearance, and therefore the underestimation of AUCR, of digoxin could be due to a missing interaction of vemurafenib with a renal basolateral uptake transporter involved, besides the predominant glomerular filtration, in digoxin renal elimination. Unfortunately, such transporter(s) are yet to be identified. A transporter that fulfills this requirements could be easily added to the model, but with the currently available data is not verifiable. Sensitivity analyses of the in vivo K_i values of vemurafenib against P-gp was tested. As shown in Fig. 5 B, the AUCR and C_maxR of a single dose of digoxin was sensitive to the in vivo K_i values of vemurafenib against P-gp, consistent to the report that vemurafenib is a P-gp inhibitor¹¹. AUCR, C_maxR and C_min ratio (C_minR) of digoxin at steady-state were similar to that of single dose, all of which are also sensitive to vemurafenib K_i,P-gp (Fig. S 9). These data suggest that the plasma concentration of the vemurafenib model can be decently verified by an independent clinical DDI study¹⁴. For OATP1B1 and OATP1B3-mediated DDI prediction, the interacting concentration from the inhibitor vemurafenib is the corrected portal vein (corrected for the uptake of 1 and f_u/B:P) concentration, since a minimal PBPK model was used. In our first-order absorption model for vemurafenib, the interacting concentration is also derived from the portal vein concentration. The modest prediction of vemurafenib against digoxin DDI support that the interacting concentrations of vemurafenib is decently verified by an independent clinical DDI study¹⁴. It has been documented that the median accumulation ratio of vemurafenib was 7.36 and the steady state was reached within 22 days for most patients¹¹. Therefore, in the current DDI trial design, the victim drug pravastatin was administered after 21 days of administration of vemurafenib at the US FDA-approved doses of 960 mg BID. Using the IC_{50,unbound,OATP1B1} (0.0017 μM) and IC_{50,unbound,OATP1B3} (0.0023 μM) values, the AUCR and C_maxR of pravastatin predicted following the 21 days of 960 mg BID doses of vemurafenib and a 40 mg daily dose of pravastatin on day 21 were 1.29 and 1.27, respectively (Fig. 6). Using half of the IC_50,unbound values of vemurafenib against OATP1B1 and OATP1B3 as a worst-case scenario, the predicted pravastatin AUC and C_max ratios were 1.50 and 1.46, respectively (Fig. 6). Parameter sensitivity analyses indicated that the AUCR and C_maxR of pravastatin are sensitive to the K_i values of vemurafenib against OATP1B1 and OATP1B3 (Table S 3).

Fig. 5. Simulated and observed plasma concentration-time profiles of digoxin in the presence and absence of vemurafenib following oral administration.

(A) Simulated plasma concentrations of digoxin following a single oral dose (0.25 mg) on day 1 alone (black lines) or on day 22 with vemurafenib (red lines). In the trials with vemurafenib coadministration, vemurafenib was administered on day 1–28 at 960 mg BID, and a single oral dose of digoxin (0.25 mg) was administered on day 22. X-axis indicates times after digoxin dose. The simulation was conducted with 20 trials of 26 subjects using a population of 520 virtual subjects (55% female, 21–65 years) same as in Zhang et al., 2018¹⁴. Dashed lines are 5^th and 95^th percentiles of simulated digoxin concentrations with (red) and without (black) vemurafenib. Closed circles denote observed clinical data with (red) and without (black) vemurafenib. The same simulation results during 0–24 h are plotted on a different scale in y-axis in Fig. S9. (B) Sensitivity analysis of vemurafenib K_i against P-gp on digoxin AUC and C_maxR. The same trial condition was used as described above in A and in previous publication (Zhang et al., 2018)¹⁴. Closed circles and squares denote AUCR and C_maxR, respectively. Red symbols are simulated AUCR (1.46) and C_maxR (1.45) using K_i of 0.00017 μM against P-gp as described in the methods. The dashed line indicates AUCR and C_maxR of 1.

Fig. 6. Simulated and observed plasma concentration-time profiles of pravastatin in the presence and absence of vemurafenib following oral administration.

(A) Simulated versus observed plasma concentrations of pravastatin following a single dose of pravastatin (40 mg) co-administered with placebo (solid line), or vemurafenib (960 mg twice daily steady state) (dashed lines). Vemurafenib was administered on day 1–22 at 960 mg BID. On day 21, a single 40 mg pravastatin was administered orally. Closed circles denotes the observed values of pravastatin⁵⁴. Lines represent the means of simulated virtual populations from 20 trials of 20 subjects (50% female, 20–50 years). Dashed lines are simulated using IC_{50,unbound,OATP1B1} (0.0017 μM) and IC_{50,unbound,OATP1B3} (0.0023 μM) determined in current studies (black) and half of these IC₅₀ values as the worst-case scenario (red), respectively.

The rifampicin K_{i,OATP1B1,total} determined in the current study in 100% human plasma is 2.1 μM (Fig. S1). The rifampicin IC_{50,unbound,OATP1B1} is estimated to be 0.24 μM after multiplying the rifampicin f_u,plasma of 0.116^46,47. Interestingly, this rifampicin Ki,OATP1B1 value (0.24 μM) is similar to the in vivo Ki,OATP1B1 value of rifampicin we and others reported previously^19,27, yielding a pravastatin AUCR of 2.49 (trial range 2.33–2.58) in PBPK model simulation comparable to the reported pravastatin AUCR of 2.33³⁶ with rifampicin as the perpetrator.

DISCUSSION

Evaluation of the DDI potential of kinase inhibitor drugs with widely prescribed statins is of clinical significance. The current study, for the first time, estimated the inhibition potency of vemurafenib against OATP1B1 and OATP1B3 in human plasma and assessed the OATP1B1 and OATP1B3-mediated DDI potential.

The estimated maximum total plasma concentration of vemurafenib at the inlet to the liver (I_in,max) is 132 μM (Table 3). In human plasma, vemurafenib at 150 μM, a concentration comparable to a clinically relevant concentration, significantly inhibited OATP1B1- and OATP1B3-mediated transport to 60.2 and 72.1% of control, respectively (Fig. 2), suggesting that inhibition of OATP1B1 and OATP1B3 is likely to occur in vivo under the physiologically relevant condition. Vemurafenib is a highly protein bound drug (protein binding>99%) with low aqueous solubility (~0.78 μM)¹¹. Based on the f_u,plasma,inc value of 0.00001 determined in our current study, even at the highest, 500 μM total concentration of vemurafenib used in the current study in human plasma, the unbound concentration vemurafenib (0.005 μM) is within the soluble range in aqueous solution. Therefore, inhibition data in our current study is not anticipated to be confounded by solubility of vemurafenib. It has been documented in the FDA documentation that vemurafenib is not an inhibitor of OATP1B1 or OATP1B3¹¹. A previous report describing inhibition of vemurafenib on OATP1B1 by 75% fold was conducted at a 10-μM concentration in the aqueous buffer, which is greater than the soluble concentration in aqueous solution¹⁵. Due to the low solubility of vemurafenib, the inhibition data of vemurafenib against OATP1B1 and OATP1B3 from this previous study needs to be interpreted with caution given the f_u,inc of vemurafenib in the uptake assay buffer in this study¹⁵ is unknown.

For compounds with high protein binding and/or non-specific binding, it is a challenge to accurately determine the in vitro IC₅₀ values against transporters in traditional protein-free buffer as the actual unbound drug concentration in the assay apparatus could be lower than the expected values. For example, troglitazone is a highly-bound compound with a f_u,plasma value of 0.0011⁴⁸. The in vitro IC₅₀ values of troglitazone against OATP1B1 determined in studies from two different laboratories using similar OATP1B1-overexpressing cell lines varied ~8-fold (0.32 μM²² vs. 2.5 μM⁴⁹). Differential binding of such compounds to the uptake assay system used in different laboratories may contribute to the observed large inter-laboratory variability.

The unbound fraction of vemurafenib in human plasma was reported to be 0.0014, and the plasma protein binding was independent of vemurafenib concentrations ranging 0.5–100 μM¹⁶; however, the detailed experimental and analytical methods used to determine the f_u,plasma is not publically available. Our finding of lack of concentration-dependent binding of vemurafenib in human plasma in the concentration range of 5–500 μM (Table 1) is similar to previously documented results¹⁶. Different experiment methods used to determine protein binding (e.g. equilibrium dialysis, ultrafiltration, and ultracentrifugation) may yield different f_u values as reported previously for valproic acid as an example⁵⁰. In our current study, we determined the f_u,plasma and f_u,plasma,inc values side-by-side from the same experiment using RED and LC-MS/MS. We further used f_u,plasma,inc to estimate the IC_50,unbound and input the f_u,plasma into R-value and PBPK models for subsequent DDI prediction. Using this approach, the R-values of vemurafenib against OATP1B1- and OATP1B3-mediated pravastatin transport (Table 3) are greater than the US-FDA recommended cut-off values of 1.1⁶. At the highest recommended dose of 960 mg, using measured vemurafenib K_{i,OATP1B1/1B3} and half of the K_i values as the worst-case scenario, PBPK model predicts AUCR of 1.29 and 1.50 and C_maxR of 1.27 and 1.46, respectively, for vemurafenib against pravastatin. These data suggest that vemurafenib may have potential to cause clinically significant DDIs against OATP1B1 and OATP1B3 substrates, e.g. statins, however, with a modest magnitude in AUCR and C_maxR (< 2-fold). In addition to the 960 mg dose, two reduced doses of 720 and 480 mg are recommended in under certain adverse reactions⁹. We further evaluated the OATP1B1 and OATP1B3-mediated DDI potential of vemurafenib at lower doses against pravastatin using PBPK modelling. As shown in Table S4, at the lowest clinical dose of 480 mg, vemurafenib is not anticipated to cause clinically relevant DDI using the measured Ki,OATP1B1/3. With increased dose of vemurafenib and with reduced K_i,OATP1B1/3 values under worst-case scenarios, an OATP1B1 and OATP1B3-mediated DDI is more likely to occur. To date, a clinical DDI of vemurafenib against statins, such as pravastatin, has not been reported; however, based on the results stated above, co-medication of vemurafenib and OATP1B1- and OATP1B3-transported compounds should be monitored in the clinic, specifically when high doses of vemurafenib are used.

Based on the estimated IC_{50,unbound,OATP1B1} (0.0017 μM) and IC_{50,unbound,OATP1B3} (0.0023 μM) values and Eq. 1, at an unbound concentration of 0.78 μM in HBSS buffer (Fig. 3 A and B), vemurafenib would be anticipated to yield an ~96% and ~99% inhibition against OATP1B1- and OATP1B3-mediated transport, respectively. However, in our current study (Fig. 3), 0.78 μM vemurafenib did not inhibit OATP1B1 and OATP1B3-mediated transport in HBSS buffer without protein. This discrepancy may be explained by the high non-specific binding of vemurafenib to the uptake assay apparatus in the absence of proteins. Human plasma proteins can block nonspecific binding of drugs to the surface of various materials^23,24. As a matter of fact, our data suggest that there is no significant difference between f_u,plasma,inc and f_u,plasma, suggesting that in the presence of human plasma, the non-specific binding of vemurafenib to the uptake assay apparatus is minimal. It is plausible to conclude that including human plasma in the uptake assay system may help to reduce non-specific binding to the uptake apparatus in the case of vemurafenib.

Many drugs on the market and in the drug development pipeline are highly protein bound^51,52. Due to historical limitation of methodologies for measuring f_u,plasma of highly bound drugs in plasma, the US FDA recommended using 0.01 as f_u,plasma for compound with high bound >99%^6,32. With the recent improvements in the analytic methods, using measured f_u,plasma values is proposed by scientists in the field, instead of using the conservative f_u,plasma of 0.01, in order to minimize the false positive DDI prediction using the R-value model^21,53. Nevertheless, the accurate determination of in vitro IC_50,unbound values of highly bound drugs remained challenging. The current study presents a novel strategy of using IC_50,unbound values estimated from the IC_50,total taking into consideration f_u,plasma,inc in the R-value and PBPK models when predicting OATP1B1- and OATP1B3-mediated DDIs. The predictive performance of this approach is worthy to be further evaluated using more highly bound drugs and to compare the predicted DDI with the available clinical DDI data. A benefit of using the current IC_50,total method to predict OATP1B1- and OATP1B3-mediated DDI is that the R-value and AUCR ratio are related to the ratio of f_u,plasma vs. f_{u,plasma,incubation,}, other than the absolute values of f_u,plasma. f_u values determined using different methods can be variable^20,21,50. f_u values can even be variable with slightly different experiment condition (e.g. incubation time) using the same method²¹. A benefit of using current IC₅₀ total method to predict OATP1B1- and OATP1B3-mediated DDI is that the R-value and AUCR ratio are related to the ratio of f_u,plasma vs. f_{u,plasma,incubation,}, other than the absolute values of f_u,plasma. We believe it is critical to use f_u,plasma and f_u,plasma,inc values from the same study, as we did in the current study, to avoid the potential uncertainty in f_u determined in different studies due to different experimental methods and/or conditions.

E₂17βG is a sensitive substrate of OATP1B1 when determining K_i values in vitro; it yielded similar Ki,OATP1B1 as pravastatin when using CsA and rifampicin as inhibitors^17,18. We used K_i values generated using E₂17βG and CCK-8 as probe substrates of OATP1B1 and OATP1B3, respectively, to assess OATP1B1- and OATP1B3-mediated DDI potential in our previous publications^19,27. The sensitivity of various OATP1B1 and OATP1B3 substrates, including pravastatin, in vitro in human plasma for determining IC_50,total worth to be systemically evaluated in future studies, similarly as published in conventional protein-free buffer^17,18.

In conclusion, the current studies report that vemurafenib can inhibit OATP1B1 and OATP1B3 in vitro at clinically relevant plasma concentrations and predict that vemurafenib has potential to cause clinically relevant DDIs against OATP substrates such as pravastatin. The current study is a concept proof to apply the total IC₅₀ method into R-value and PBPK models to predict OATP1B1- and OATP1B3-mediated DDIs. The prediction performance using the IC_50,total methods warrants further validation with clinical DDI studies. The prediction performance using the IC_50,total methods warrants further validation with clinical DDI studies. Assessing the IC_50,unbound from the IC_50,total for compounds that are highly protein bound and/or with low aqueous solubility may be a beneficial approach to help mitigate false negative DDI prediction.

Abbreviations:

DMEM

Dulbecco’s Modified Eagle’s Medium

DMSO

dimethyl sulfoxide

DDI

Drug-Drug Interaction

HBSS

Hanks’ Balanced Salt Solution

OATP

Organic Anion Transporting Polypeptide

P-gp

P-glycoprotein

PBPK

physiologically-based pharmacokinetic

CCK-8

Cholecystokinin 8

E₂17βG

estradiol 17 β-D-glucuronide

HEK

Human embryonic kidney