Establishment of In Vitro P-Glycoprotein inhibition assay and its exclusion criteria to assess the risk of drug–drug interaction at the drug discovery stage

Materials

Digoxin, verapamil, and lucifer yellow were purchased from Sigma–Aldrich (St. Louis, Missouri). Elacridar and cyclosporin A were from Toronto Research Chemicals, Inc. (North York, Ontario, Canada). Itraconazole was from MP Biomedicals, LLC (Tokyo, Japan). Ketoconazole and quinidine were from Wako Pure Chemicals Industries (Osaka, Japan). All other chemicals were of the highest reagent grade available from commercial sources.

HTS Transwell® 96-well permeable support plates with a polyethylene terephthalate membrane, 0.4-μm pore size and 0.143-cm² surface area, were purchased from Corning Life Sciences (Lowell, Massachusetts).

Cell Culture

LLC-PK1 cells expressing human MDR1 were cultured with slight modification as previously described.17 Cells were collected by adding 0.05% trypsin/ethylenediaminetetraacetic acid solution; next, cells were suspended in cell culture media (Dulbecco's modified eagle medium with 20% fetal bovine serum, 1% penicillin, streptomycin, and 150 ng/mL colchicine) and seeded onto 96-well insert plates (1.125×10⁵ cells/well). These plates were incubated in an atmosphere of 5% CO₂–95% air at 37°C. Fresh medium was replaced on the third or fourth day and the transcellular transport study was conducted on the seventh day after seeding.

In Vitro P-gp Inhibition Assay for Digoxin Transport

The inhibitory effect on digoxin transport across LLC-PK1 cells expressing human MDR1 was determined in the presence or absence of potential P-gp inhibitors. Inhibitors were added to the M199 medium, which included 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid at pH 7.4, on both sides of the transwell chamber before adding digoxin. After preincubation for 30 min, the transcellular transport study of digoxin (10 μM) was conducted in an atmosphere of 5% CO₂–95% air at 37°C for 2 h. The Michaelis constant (K_m) value for digoxin transport was found to be more than 500 μM (unpublished data). This result is also supported by Keogh and Kunta9 that no saturable transport was observed over the range of 0.05–10 μM. Therefore, it should not be a problem to use 10 μM of digoxin in order to determine the concentration of samples from in vitro P-gp inhibition assay without using radiolabeled compounds. Lucifer yellow (200 μM) was added in every transwell along with digoxin and inhibitors to monitor cell monolayer integrity. Dimethyl sulfoxide (DMSO) was set up to 0.9%, a level at which cellular damage by DMSO was negligible. Samples from the receiver side were taken and the concentrations of aliquots were determined using tandem mass spectrometry (MS/MS) analysis (API4000; AB SCIEX, Foster City, California) equipped with an ultra-fast liquid chromatography (UFLC) apparatus (Shimadzu, Kyoto, Japan). The P_app values of digoxin in the receiver wells were determined using Eq. 1:

(1)

where amount, area, C₀, and time represent the transported amount of digoxin per well of the cell monolayer, the surface area of the cell monolayer, the initial concentration of digoxin in the donor side, and the incubation time, respectively. The magnitude of inhibition was calculated using Eqs. 2 and 3:

(2)

(3)

where ε_i and ε_i,app are the degrees of inhibition of digoxin ER and P_app,BtoA, respectively. ER_max and ER_min represent the ER in the absence or presence of 4 μM of elacridar, which is an adequate concentration to achieve complete inhibition of P-gp function. The P_{app,BtoA max} and P_{app,BtoA min} also represent P_app,BtoA in the absence or presence of elacridar. The IC₅₀ values of potential P-gp inhibitors were calculated in GraphPad Prism version 5.00 for Windows, GraphPad Software (San Diego, California).

Quinidine was used as a marker of IC₅₀ variability to ensure the reliability of the generated data. The mean (SD) of the IC₅₀ values based on ER and P_app,BtoA were 1.52 (0.35) and 8.53 (2.41) μM, respectively, indicating that the results of inhibition studies can be independently replicated.

Quantification of Samples from In Vitro P-gp Inhibition Assay

Samples from the in vitro P-gp inhibition assay for digoxin transport were mixed with an equal volume of acetonitrile, including 100 nM of digoxin-d3 (internal standard), and centrifuged at 4°C and 5084 × g for 5 min (Sorvall Legend RT; Thermo Fisher Scientific, Yokohama, Japan). The supernatant was diluted with an equal volume of 0.1% formic acid solution. The sample was mixed, centrifuged at 4°C and 5084 × g for 5 min, and the supernatant was analyzed using MS/MS analysis (API4000, AB SCIEX) equipped with an UFLC apparatus (Shimadzu). The analytical column used was an Atlantis T3, C18 column (20 mm × 2.1 mm, 3 μm) (Waters, Milford, Massachusetts). The total run time, flow rate, and column temperature were 1 min, 1 mL/min, and 50°C, respectively. Mobile phases A and B consisted of 0.2% formic acid in distilled water and 0.2% formic acid in methanol, respectively. The initial concentration of mobile phase B was 5% and this condition was maintained for 0.02 min followed by linear increase of B to 95% over the next 0.03 min, and held for 0.45 min. The condition was then returned to initial concentrations and held for 0.5 min for reequilibration. The lowest limit of quantification was 4.6 nM and the standard curve was linear (r² > 0.99) over the digoxin concentration range of 4.6–10,000 nM. Detailed MS conditions are shown in Table 1. Accuracy of standards and intraday and interday variability was within ±15%.

Determination of Plasma Protein Binding

Unbound fractions in human plasma for itraconazole, cyclosporin A, ketoconazole, verapamil, quinidine, and talinolol were determined using the 96-well equilibrium dialysis apparatus (HTDialysis, LLC, Gales Ferry, Connecticut). Pooled human plasma was purchased from KAC (Kyoto, Japan). Tested compounds were added into the control plasma, (final concentration = 1 μM) and 150-μL aliquots were added into the dialysis well, which was separated by the dialysis membrane (singlestrip HTD96a/b; HTDialysis, LLC). The receiver chamber was loaded with 150 μL of 50 mM phosphate buffered saline (pH 7.4). The apparatus was placed on a shaker (87 rpm) (NR-1; TAITEC, Tokyo, Japan) for 20 h at the room temperature. Previous in-house experiments revealed that equilibrium would be adequately achieved within an incubation period of 20 h (data not shown). After incubation, aliquots (50 μL) from donor and receiver sides were mixed with 50-μL aliquots of either the dialyzed control buffer or human plasma to yield an identical matrix so that all samples could be quantified using the same matrix component. Additionally, 150 μL of acetonitrile containing 100 ng/mL alprenolol as an internal standard was added to every sample, vigorously mixed, and centrifuged at 4°C (5084 x g) for 5 min. The supernatant was diluted with 0.2% (v/v) formic acid in 10 mM ammonium formate (pH 3) and the concentration of each aliquot was determined using MS/MS analysis equipped with an UFLC apparatus. The unbound fraction in plasma was calculated by the ratio of concentrations in receiver and donor sides of the dialysate. The analytical column used was a Shim-pack XR-ODS, C18 column (20 mm × 2.0 mm, 5 μm) from Shimadzu. Total run time, flow rate, and column temperature were 2.6 min, 0.5 mL/min, and 50°C, respectively. Mobile phases A and B consisted of 0.2% (v/v) formic acid in 10 mM ammonium formate and acetonitrile, respectively. The initial concentration of mobile phase B was 10%, and this condition was maintained for 0.2 min followed by a linear increase of B to 95% over the next 0.2 min, and held for 1 min. The condition was then returned to the initial concentration and held for 1.2 min for reequilibration. Detailed MS conditions are shown in Table 1. Accuracy of standards and interassay variability were within ±15%.

Risk Assessment of P-gp-Mediated Digoxin DDI

Literature research on DDI studies with digoxin was conducted to assess the effects of potential P-gp inhibitors on digoxin exposure. In this report, a 25% increase of digoxin exposure [area under the curve (AUC) and C_max] in the presence of P-gp inhibitors was categorized as a clinically significant change because the digitalis-induced toxicity was observed.16 The mean unbound steady-state C_max following administration of the highest clinical dose is defined as [I₁], and the theoretically maximal gastrointestinal concentration (dose/250 mL) is defined as [I₂] according to a previous report.14 The unbound fractions in human plasma were ascertained from the literature research18, 19 and in-house experimental data.

The result of risk assessment of P-gp-mediated digoxin DDI can be categorized into four groups. True positive and true negative mean that in vitro P-gp inhibition data are in accordance with the clinical digoxin DDI. False negative means in vitro data are not in agreement with a clinical digoxin DDI, where the increase of AUC is greater than or equal to 1.25-fold; on the contrary, false positive means in vitro data are not in agreement with a clinical digoxin DDI, where the increase of AUC is less than 1.25-fold.

In Vitro P-gp Inhibition Assay for Digoxin Transport

To estimate the inhibitory potency of typical P-gp inhibitors (elacridar, itraconazole, cyclosporin A, ketoconazole, verapamil, and quinidine) on digoxin transport (10 μM), in vitro P-gp inhibition assays using LLC-PK1 cells expressing human MDR1 were performed. The sigmoid inhibition curve indicates that ER and P_app,BtoA of digoxin decreased in a concentration dependent manner (Fig. 1). The inhibitory potency of elacridar on digoxin transport was over 10 times greater than other P-gp inhibitors. These results are comparable to those of a previous report in which the inhibitory effect of compounds was determined according to P_app,BtoA of [³H]digoxin (30 nM) using MDCKII–MDR1 cells.9

Comparison of IC₅₀ Values Generated from ER and P_app,BtoA

The IC₅₀ values of 48 compounds were ascertained to assess the potential of P-gp inhibitory activity (Table 2). The IC₅₀ values of compounds that did not achieve more than 50% inhibition at the highest concentration were greater than 60 μM. The IC₅₀ values were calculated using inhibition of ER and P_app,BtoA by using Eqs. 2 and 3, respectively. Elacridar showed the strongest P-gp inhibitory effect with IC₅₀ values of 0.022 [IC₅₀(ER)] and 0.074 [IC₅₀(P_app,BtoA)] μM, followed by carvedilol, itraconazole, verapamil, and cyclosporin A. The IC₅₀ values generated from P_app,BtoA were larger than those from ER for all compounds tested, and there was a positive correlation (r² = 0.61) between IC₅₀(ER) and IC₅₀(P_app,BtoA) (Fig. 2).

Theoretsical Consideration of the Difference Between IC₅₀ Values Generated from ER and P_app,BtoA

The method to calculate IC₅₀ values vary among laboratories.10,13,15, 16,20 This may lead to misunderstanding of the rank order of the P-gp inhibitory potency. To choose the most definitive parameter (ER or P_app,BtoA) for the estimation of in vitro P-gp inhibition, kinetic analysis was performed. An appropriate kinetic model has been reported to describe in vitro and in vivo P-gp function.11,21 On the contrary, a simple two-compartment model cannot explain the difference between IC₅₀(ER) and IC₅₀(P_app,BtoA) (Fig. 2). Therefore, we used three-compartment model analysis (Fig. 3). The mass balance of digoxin in the apical (dX_A/dt), basolateral (dX_B/dt), and intracellular (dX_C/dt) areas are described by differential Eqs. 4, 5, and 6, respectively:

(4)

(5)

(6)

where C_A, C_B, and C_C indicate substrate concentration in each compartment. The PS_P-gp represents the intrinsic permeability–surface area (PS) product mediated by P-gp, and PS_A and PS_B represent the PS product mediated by passive permeability. Assuming that the back diffusion of digoxin on the receiver side to the intracellular area is negligible (sink condition), P_app,AtoB and P_app,BtoA fluxes are described in Eqs. 7 and 8. Because donor concentrations in the apical and basolateral chambers are equivalent, ER can be described using Eq. 9. Equation 9 indicates that ER is proportional to PS_P-gp:

(7)

(8)

(9)

When the function of P-gp is completely inhibited, the maximal change in P_app,BtoA is described by Eq. 10, where P_{app,BtoA max} and P_{app,BtoA min} are efflux activities in the absence or presence of elacridar, respectively. The maximal change of P_app,BtoA in the presence of P-gp inhibitor is twofold and is not reflected by PS_P-gp. On the contrary, the theoretically maximal increase of ER is described by Eq. 11. (average value was approximately 20 in this study), where ER_max and ER_min are ERs in the absence or presence of elacridar, respectively. The maximal increase of ER is directly depending on the change of intrinsic P-gp activity (PS_P-gp):

(10)

(11)

The inhibitory rate (ε_i and ε_i,app) is calculated using Eqs. 2 and 3. The expression formulae of ε_i and ε_i,app can be developed by substitution of the intrinsic P-gp function (PS_P-gp) and P_app,BtoA for Eqs. 12 and 13, respectively, where PS_P-gp,max is the maximal efflux activity in the absence of P-gp inhibitors. It is assumed that PS_A = PS_B:

(12)

(13)

Equations 14 and 15 indicate intrinsic P-gp function (PS_P-gp) using ε_i and ε_i,app in the presence of P-gp inhibitors. The PS_P-gp with a use of ER directly reflects the changes of the intrinsic P-gp activity, although PS_P-gp with a use of P_app,BtoA may be affected by the passive permeability:

(14)

(15)

Assuming that the Michaelis–Menten kinetics is applicable for P-gp inhibition, ε_i and ε_i,app equal [I]/{[I] + IC₅₀(ER)} and [I]/{[I] + IC₅₀(P_app,BtoA)}, respectively, where [I] represents inhibitor concentration. According to Eqs. 14 and 15, the ratio of IC₅₀ values [IC₅₀(P_app,BtoA)/IC₅₀(ER)] can be described using Eq. 16, indicating that IC₅₀ values generated from P_app,BtoA [IC₅₀(P_app,BtoA)] were larger than those from ER [IC₅₀(ER)] by 1 + PS_P-gp,max/2PS_A. The difference between IC₅₀(ER) and IC₅₀(P_app,BtoA) in the transcellular transport study (Fig. 2 and Table 2) was comparable to the results based on a compartment model. Therefore, the compartment model (Fig. 3) was appropriate for the theoretical consideration of the difference between IC₅₀ values generated from ER and P_app,BtoA:

(16)

Consequently, the difference of IC₅₀ values generated from two parameters was kinetically described using the compartment model analysis. The difference between the IC₅₀ values generated from ER and P_app,BtoA may be attributed to the fact that the inhibitory rate based on P_app,BtoA (ε_i,app) cannot neglect the influence of passive permeability. Additionally, our experimental data and model analysis indicate that ER is a highly sensitive parameter for estimating the degree of P-gp inhibition as compared to P_app,BtoA. Thus, ER is an appropriate parameter for estimating the IC₅₀ values of P-gp inhibitors and was used for the subsequent risk assessment of clinical P-gp-mediated DDI studies.

Risk Assessment of P-gp-Mediated Digoxin DDI

To evaluate whether the investigational drug leads to clinically relevant DDI with digoxin, risk assessment of P-gp-mediated DDI was performed considering IC₅₀(ER) and the increase of digoxin plasma concentration in clinical DDI studies. Clinical drug interaction studies of digoxin and concomitant drugs are summarized in Table 3. In this report, a 25% increase in digoxin exposure (AUC and C_max) in the presence of P-gp inhibitors was determined to be clinically significant. Literature research indicated that 15 out of 36 DDI studies with digoxin were clinically relevant. Ratios of [I₁]/IC₅₀ and [I₂]/IC₅₀ were calculated to determine the cutoff limit to predict a clinically relevant digoxin DDI. The highest values of [I₁]/IC₅₀ and [I₂]/IC₅₀ (0.79 and 7182) were observed in the clinical digoxin DDI concomitantly administered with quinidine. The AUC ratio of digoxin was also the highest value (2.65) in this study (Table 3).

The consideration of [I₁] and [I₂] is one of the key factors to predict the clinical digoxin DDI. The definition of [I₁] used in the White paper¹ is the steady-state unbound C_max following administration of the highest clinical dose. Indeed, the unbound drug is thought to be pharmacologically active and accessible to transporters as a substrate and/or an inhibitor. As shown in Table 3, [I₁]/IC₅₀ values were higher than 0.1, which is the recommended cutoff limit by the White paper, for cyclosporin A and quinidine that significantly increase the digoxin concentration in the clinical DDI study. However, in most cases, [I₁]/IC₅₀ values were less than 0.1, even though clinically relevant DDIs were observed. The [I₁] may not be suitable to predict clinically relevant DDIs at the intestinal absorption. The [I₂] is defined as the clinical dose divided by 250 mL, which is the volume of a glass of water. Although the potential of P-gp inhibition may be overestimated for poorly soluble drugs, it is important to take a safety risk assessment when it comes to selection of candidate compounds. To improve the predictability of actual intestinal concentration of P-gp inhibitors, consideration of the intestinal fluid volume and solubility of compounds is necessary. Tachibana et al reported that the estimated intestinal fluid volume was 1.92 L. In addition, the solubility of compounds in the estimated intestinal fluid was evaluated for the prediction of P-gp-mediated DDI.22 Further investigations may be required to validate the estimated volume in either fasted or fed condition.

In Vitro Cutoff Values ([I₂]/IC₅₀) to Predict the Clinical Digoxin Interaction Studies

The appropriate in vitro cutoff value to predict P-gp-mediated DDIs was investigated using [I₂]/IC₅₀ because [I₂] indicates the theoretical maximal gastrointestinal concentration. Results of risk assessment are categorized into the true positive, true negative, false positive, and false negative. A true positive indicates an in vitro prediction corresponding to the observation that digoxin AUC and C_max increased by 25% in the presence of P-gp inhibitors. On the contrary, a true negative indicates an in vitro prediction corresponding to the observation that digoxin AUC and C_max did not significantly change. The accuracy of DDI risk assessment is shown in Figure 4, indicating that the risk criteria of [I₂]/IC₅₀ = 30 was the optimal cutoff value to predict a clinically relevant DDI [the average of true positive and negative percentage (85% accuracy)]. The reliability of the criteria was supported by a previous report using a literature search.22 The IC₅₀ values for compounds that showed less than 50% inhibition of ER at the highest concentration (60 μM) were assumed to be 60 μM in this investigation. The ITC recommended that investigational drugs with an [I₂]/IC₅₀ value greater than or equal to 10 should be evaluated using the clinical P-gp-mediated DDI potential.1 However, using these criteria, the false negative rate increased to 52%, which may result in the overestimation of P-gp inhibition potential for drug candidate. At the drug discovery stage, it is primarily important to raise an alert for compounds with severe potential of DDIs in advance to the following clinical study. To set the criteria for compounds selection, the weighting between false positives and false negatives should be well-balanced. From a safety perspective, a false negative error has greater consequences than a false positive because it may result in unexpected side effect caused by P-gp-mediated DDI in clinic. However, from the view of accuracy of risk assessment, the exclusion criteria should not be too conservative; otherwise promising candidate compounds may be discarded with a false positive data. In Figure 4, only captopril was categorized into false negative at [I₂]/IC₅₀ values of 10 and 30. And the false negative result with captopril is not due to P-gp inhibition, but is attributed to the reduction of renal clearance of digoxin in the presence of captopril.23 Accordingly, we propose that the cutoff value of [I₂]/IC₅₀ = 30 generated from ER would be the appropriate parameter to evaluate the potential of P-gp-mediated DDI in intestinal absorption.

Risk Assessment of P-gp-Mediated Digoxin DDI from In Vitro IC₅₀ Values

At the drug discovery stage, it is challenging to set the efficacious doses unless a precise estimation of pharmacodynamics using pharmacokinetic–pharmacodynamic modeling and analysis in humans is possible.24 It is also important to provide the appropriate exclusion criteria of P-gp inhibition for the candidate selection; otherwise a false positive prediction may lead to exclude a number of candidate compounds that may not have the potential DDI in clinic. Therefore, we investigated whether in vitro IC₅₀ value itself is applicable for determining the risk assessment of P-gp-mediated DDIs. A retrospective analysis of clinical digoxin DDI studies revealed that compounds with IC₅₀ values less than 2 μM significantly increased digoxin exposure (9 out of 10 compounds). A false positive result was observed for nicardipine, with an IC₅₀(ER) value of 1.1 μM and [I₂]/IC₅₀ value of 232.5. This discrepancy is potentially due to the overestimation of actual intestinal concentration of nicardipine in humans, leading to a false positive in DDI prediction. Indeed, nicardipine is a representative CYP3A substrate in humans.25 Additionally, the intrinsic clearance value of nicardipine determined on the basis of its metabolic stability with pooled human intestinal microsomes was larger than most CYP3A substrates. The fraction of the dose not metabolized by intestinal metabolic enzymes in humans (F_g in human) was 0.64.26 Therefore, [I₂] may overestimate the actual intestinal concentration of nicardipine. On the contrary, a false negative result was observed for captopril in DDI prediction (Table 4). The 39% increase of digoxin AUC in the presence of captopril27 was observed because captopril decreases renal clearance of digoxin.23 As a result, a retrospective analysis of clinical digoxin DDI studies suggested that compounds with IC₅₀ values less than 2 μM have high risk for P-gp-mediated DDI.

Extensive research on P-gp-mediated DDIs at the intestinal absorption process with P-gp-specific substrates (i.e., digoxin and talinolol) has been performed in the pharmaceutical industry and academia.10,13,15, 16 At the same time, P-gp-mediated DDIs with CYP3A4/5 and P-gp dual substrates in the intestine have been of great interest. Cyclosporin A and tacrolimus are immunosuppressant drugs and metabolic substrates for CYP3A4 and CYP3A5. They are also actively transported out of cells by P-gp.28 The increase in plasma concentrations of cyclosporin A29 and tacrolimus30 in the presence of ketoconazole, which is a dual inhibitor for CYP3A4/5 and P-gp, were reported. The inhibition of CYP3A4/5 may increase F_g, whereas the inhibition of P-gp may increase the fraction absorbed (F_a) in intestine. In the case of DDI study of paclitaxel (F_a × F_g = 0.045), the increase in systemic exposure of paclitaxel in combination with elacridar is same in magnitude (sevenfold increase) as in combination with cyclosporin A, which is a dual inhibitor for CYP3A4/5 and P-gp.31 Thus, it is important to pay attention to the impact of P-gp inhibition as well as CYP3A4/5 inhibition. Further quantitative investigation will be needed to predict the risk of DDIs involving both CYP3A4/5 and P-gp.