The Liver Partition Coefficient-Corrected Inhibitory Quotient and the Pharmacokinetic-Pharmacodynamic Relationship of Directly Acting Anti-Hepatitis C Virus Agents in Humans

Abstract

Pharmacokinetic-pharmacodynamic (PK-PD) data analyses from early hepatitis C virus (HCV) clinical trials failed to show a good correlation between the plasma inhibitory quotient (IQ) and antiviral activity of different classes of directly acting antiviral agents (DAAs). The present study explored whether use of the liver partition coefficient-corrected IQ (LCIQ) could improve the PK-PD relationship. Animal liver partition coefficients (Kp_liver) were calculated from liver to plasma exposure ratios. In vitro hepatocyte partition coefficients (Kp_hep) were determined by the ratio of cellular to medium drug concentrations. Human Kp_liver was predicted using an in vitro-in vivo proportionality method: the species-averaged animal Kp_liver multiplied by the ratio of human Kp_hep over those in animals. LCIQ was calculated using the IQ multiplied by the predicted human Kp_liver. Our results demonstrated that the in vitro-in vivo proportionality approach provided the best human Kp_liver prediction, with prediction errors of <45% for all 5 benchmark drugs evaluated (doxorubicin, verapamil, digoxin, quinidine, and imipramine). Plasma IQ values correlated poorly (r² of 0.48) with maximum viral load reduction and led to a corresponding 50% effective dose (ED₅₀) IQ of 42, with a 95% confidence interval (CI) of 0.1 to 148534. In contrast, the LCIQ-maximum VLR relationship fit into a typical sigmoidal curve with an r² value of 0.95 and an ED₅₀ LCIQ of 121, with a 95% CI of 83 to 177. The present study provides a novel human Kp_liver prediction model, and the LCIQ correlated well with the viral load reductions observed in short-term HCV monotherapy of different DAAs and provides a valuable tool to guide HCV drug discovery.

INTRODUCTION

The inhibitory quotient (IQ) is a common pharmacological predictor of antiviral activity and generally represents the ratio between the trough plasma concentration (C_min) and an in vitro potency parameter such as 50% effective concentration (EC₅₀) (3). Several modifications have been proposed to improve the prognostic value of the plasma IQ-antiviral response relationship, including the use of serum-shifted EC₅₀s, the consideration of intracellular drug concentrations, and the potential complication related to drug resistance mutations (1, 3), as well as the proposed instantaneous IQ model, which suggested that the slope of the dose-response curve is another critical fact to consider (34). Recently, several directly acting antivirals (DAAs) targeting HCV have been advanced into clinical trials, and the first DAAs have now been approved for use in combination therapy with pegylated interferon and ribavirin (2, 18, 21, 32, 35). The first DAA tested in chronic hepatitis C virus (HCV)-infected patients with potent antiviral effect was BILN 2061 (16, 20), which showed a clear dose-dependent in vivo activity. Even at the low oral twice-daily (b.i.d.) dose of 25 mg, a mean of 2-log₁₀ IU/ml viral load reduction (VLR) was observed following a brief treatment (2 days) (16). The corresponding mean plasma IQ was 6.6 at this low oral dosage. A similar plasma IQ, but significantly greater antiviral effects, was observed for telaprevir dosed at 750 mg three times a day (t.i.d.), based on published clinical pharmacokinetic (PK) data (33) and the EC₅₀s measured under the same experimental condition as used for BILN 2061 from our lab. Clinical trials for the polymerase inhibitor BILB 1941 (11) demonstrated that a mean plasma IQ of 21 was achieved with the t.i.d. oral dose of 300 mg (about 3.5-fold greater than for BILN 2061 at 25 mg b.i.d. or telaprevir at 750 mg t.i.d.). Yet this higher plasma IQ was associated with much lower antiviral effects in the clinic, with a mean viral load reduction of only 1 log₁₀ IU/ml. Correction of plasma IQ using serum-shifted EC₅₀s, or plasma protein binding-corrected IQs, failed to improve the PK-pharmacodynamics (PD) relationship (Table 1). Several factors may contribute to the inconsistent relationships between plasma IQ and clinical antiviral effects. These include mechanism-dependent differences, such as those observed for HIV protease inhibitors versus nonnucleoside reverse transcriptase inhibitors (3), or differences in drug exposure at the site of viral replication: i.e., the liver, relative to plasma. A striking feature that differentiated BILN 2061 and telaprevir from BILB 1941 was that both protease inhibitors were more highly distributed to rat liver. However, cross-species in vivo liver distributions were not consistent among rats, mice, and dogs for tested compounds, confounding any extrapolation to humans.

Table 1.

Initial PK-PD relationship analysis based plasma IQ and corrected plasma IQ by serum shift

Parameter	Telaprevir (VX-950)a	BILN 2061b	BILB 1941c
Daily dose (mg)	2,250	50	900
Mean log10 VLR	3.35	1.98 ± 0.86	1.04 ± 1.02
Plasma IQ (Cmin/EC50)	6.0	6.63 ± 6.04	21.0 ± 12.8
Corrected plasma IQ (by serum shift)	<0.4	0.83 ± 0.75	6.99 ± 4.27
Kpliver, rat	35.2 ± 17.3	47.8 ± 17.5	8.78 ± 1.05
Kpliver, mouse	6.04 ± 0.6	5.87 ± 0.8	5.02 ± 0.30

^aClinical data represent the median values (n = 8) from reference 33; Kp_liver values are from reference 31 (means ± SDs from 3 determinations for rats and 6 determinations for mice).

^bClinical data and EC₅₀ represent the means ± SDs (n = 7) from references 16 and 20; Kp_liver values represent the mean ± SD obtained in the current study (n = 3).

^cClinical data represent the means ± SDs (n = 7) and EC₅₀s from reference 12; Kp_liver values represent the means ± SDs obtained in the current study (n = 3).

The present study first established a proportionality method for the prediction of human liver: plasma partition coefficient (Kp_liver) from in vitro human and rodent hepatocyte partition coefficients (Kp_hep), with the incorporation of in vivo rodent Kp_liver. The predicted human Kp_liver was, in turn, used to improve our understanding of the differential PK-PD relationship observed for HCV DAAs.

MATERIALS AND METHODS

In vivo plasma pharmacokinetics (PK) and liver partition coefficient (Kp_liver).

All protocols involving animal experimentation were reviewed and approved by the Institutional Animal Care and Use Committee, and all animals received humane care according to the criteria outlined in the Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences and published by the National Institutes of Health. All chemicals used were reagent grade or better. Plasma and tissue awaiting analysis were stored frozen at −20°C. All tested drugs were stable under this storage condition within the waiting period of <1 month.

Rat PK studies were performed in male Sprague-Dawley rats (275 to 300 g; Charles River, St-Constant, Quebec, Canada), while mouse experiments were performed in female CD-1 mice weighing 24 to 30 g to avoid severe fighting observed in group-housed males. The oral dose was given at a volume of 10 ml/kg in a vehicle containing 0.3% Tween 80 and 0.5% methylcellulose. Blood samples were collected at 0, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, and 8 h postdosing, and plasma samples from 3 animals were pooled at each time point. All animal PK data were analyzed based on the standard noncompartmental model using WinNonlin (version 3.1; Scientific Consulting, Inc., Cary, NC) as described in a previous publication (10).

For comparative plasma and liver exposure studies, liver samples were collected either at multiple time points for area under the concentration-time curve (AUC) evaluations or at 8 h after the oral dose for one time point comparison for assay development and mechanistic studies as summarized in Fig. 1 and 3. The rat liver was perfused with 25 ml ice-cold saline for 1 min prior to collection of the right lateral lobe. Both plasma and liver samples were extracted and analyzed by high-performance liquid chromatography (HPLC) or liquid chromatography-tandem mass spectrometry (LC-MS-MS), as described previously (39). Kp_liver values were determined using the exposure ratios in the liver over that in the plasma (AUC or concentrations at designated time points).

Fig 1.

Correlation of Kp_hep between freshly prepared and cryopreserved hepatocytes (A) and the effects of 50% plasma protein on the relationship between Kp_hep and Kp_liver in rats (B and C). In panel A, open circles and squares represent mean values from rat and mouse hepatocytes (n = 3), and the inset presents data plotted in log₁₀ scales. In panels B and C, Kp_liver values of warfarin (1), phenytoin (2), diazepam (3), verapamil (4), quinidine (5), and imipramine (6) are from previous publications (7, 14, 15, 27, 36, 38), while the Kp_liver values of HCV DAAs represent the means from 3 determinations. All Kp_hep values in panel B represent the means obtained in the absence of rat plasma (n = 6), while the values in panel C were obtained in the presence of 50% rat plasma (n = 3).

Fig 3.

Effect of rifampin on Kp_hep and in vivo PK of BILB 1941 and inhibitor A in the rat. In vitro effects of rifampin were evaluated with BILB 1941 and compound A at an initial concentration of 2 μM (A and B) in rat hepatocytes. All the Kp_hep values represent the means and SDs from 6 to 8 determinations. Open bars represent Kp_hep obtained on ice (1.2°C), compared with data obtained at 37°C (solid bars). In vivo effects of rifampin at 20 mg/kg intravenously (C and D) were evaluated with BILB 1941 and compound A at the oral dose of 5 mg/kg of body weight. All the data points represent the means and SDs (n = 3 to 4). The numbers above the histogram in the bottom panels indicate the mean fold changes with coadministration of rifampin.

Kp_hep.

In vitro hepatocyte Kp (Kp_hep) evaluations were performed using either freshly isolated or cryopreserved hepatocytes suspended in an incubation medium containing the following ingredients (in g/liter): Krebs powder mix with d-glucose (9.6), CaCl₂ (0.15), HEPES (4.8), and NaHCO₃ (2.4); the pH was adjusted to 7.4 with NaOH. Fresh hepatocytes were isolated from rats and mice as used in PK experiments, using methods as described in the literature (4, 28, 40). Cryopreserved rat, mouse, and human hepatocytes were obtained from CellzDirect, Pittsboro, NC, and prepared according to the manufacturer's instructions. In experiments including plasma protein in the hepatocyte incubation, rat plasma was obtained in-house by cardiac puncture on fasted male Sprague-Dawley rats. Human and mouse plasma from CD-1 mice was obtained from Bioreclamation Inc.

The hepatocyte suspension was incubated at 37°C under 95% O₂ and 5% CO₂ for 10 min and then mixed with 3 volumes of incubation medium to achieve the targeted final concentration of 2 million cells/ml. After initial experiments which compared the impact of plasma protein on in vivo Kp_liver predictability, all in vitro Kp_hep studies were performed in the presence of 50% plasma in the incubation media, and the initial time course evaluation led to the choice of the incubation time of 16 min. All incubation samples were placed on ice and centrifuged at 11,000 × g for 15 s. A 400-μl aliquot of the supernatant was kept for extraction. The remaining supernatant was aspirated. The cell pellet was resuspended and washed twice with washing medium that contained the same components as the incubation medium, with the elimination of d-glucose. All tested compounds were prepared in a stock solution in dimethyl sulfoxide (DMSO) and diluted into the incubation medium to achieve a final concentration of 2 μM, a median value corresponding to the highly effective plasma exposure of HCV DAAs in the clinic. Our initial experiments also demonstrated that concentration changes in the clinically relevant range (submicromolar to 15 μM) did not lead to concentration-dependent Kp_hep changes.

Compounds from the final pellet and incubation medium were extracted by acetonitrile containing 2% (vol/vol) formic acid. Hepatocyte and incubation medium concentrations of tested compounds were determined against a corresponding standard curve by HPLC or LC-MS-MS as described in detail in a previous publication (10). Selection of bioanalysis methods was based on compound properties, and the best experimental condition was chosen to obtain analysis precision and accuracy (coefficient of variation [CV] < 12%). The in vitro hepatocyte partition coefficient (Kp_hep) was determined as follows: (nmol of compound in pellet/10⁶ cells) × 1.20 × 10⁸ hepatocytes/g liver/(medium, nmol/ml).

Human Kp_liver model.

A model for the prediction of human Kp_liver was based on in vitro-in vivo proportionality averaging the data from all preclinical species tested with the following equation:

$$\text{human}{\text{K}}_{\text{Pliver}}=\frac{1}{n}{\displaystyle \sum _{i=1}^n\left[\left({\text{K}}_{\text{Phep}},\text{human}/{\text{K}}_{\text{phep}},{\text{animal}}_{\text{i}}\right)\times {\text{K}}_{\text{Pliver}},{\text{animal}}_{\text{i}}\right]}$$

In the absence of human Kp_liver values for HCV DAAs, the published human Kp_liver values of 5 marketed drugs for the treatment of other diseases were used as a benchmark to validate the human Kp_liver model (doxorubicin, verapamil, digoxin, quinidine, and imipramine). Prediction errors were calculated by the percent differences of the predicted human Kp_liver values over those published values [(predicted − observed)/observed × 100].

Clinical PK and pharmacodynamics.

To minimize variability of other factors such as the emergence of resistance, the PK-PD relationship analysis for all four in-house DAAs tested in the clinic were based on the mean plasma C_min and the mean maximum viral load reduction obtained following a short monotherapy, 2 to 5 days (11, 16, 20, 22, 26). All research using human subjects cited in these references used study protocols which conformed to the ethical guidelines of the 1975 Declaration of Helsinki, with informed consent in writing obtained from each patient. C_min represents mean values of each dose group obtained on the day closest to the maximum VLR. Specifically, the data for BILB 1941 and BI 207127 were obtained following 5 days of t.i.d. treatment at the oral doses of 10 to 450 mg and 100 to 800 mg, respectively. BILN 2061 was dosed b.i.d. orally for 2 days at 25 to 500 mg per dose. Faldaprevir (BI 201335) was dosed once daily for 2 days at 20 to 120 mg per dose. Maximum VLRs from all patients within the first 5 days of drug treatment were collected, and the mean from each dosing group was calculated as the PD parameter. The relationships between plasma IQ or LCIQ and mean maximum VLR were analyzed based on a four-parameter logistic equation (also called sigmoidal dose-response curve equation with variable slope) (GraphPad Prism software, version 4.0; GraphPad Software, CA):

$$\text{VLR}={\text{VLR}}_{\min }+\frac{{\text{VLR}}_{\max }-{\text{VLR}}_{\min }}{1+{\left(\frac{{10}^{\log E{D}_{50}}}{{10}^{\log X}}\right)}^{\text{Hill}\text{slope}}}$$

where VLR refers to the mean maximum viral load reduction in log scale, X refers to the values of IQ or LCIQ, and 50% effective dose (ED₅₀) here refers to the IQ or LCIQ that corresponds to the 50% maximum VLR on the log scale. The reported median plasma C_min and short-term viral load reduction of telaprevir (33) were also included in the PK-PD relationship plot.

In vitro-in vivo evaluation of potential role of active uptake transporters.

To obtain a mechanistic understanding of the differences in liver uptake observed for different inhibitors, the potential role of active uptake transporters was examined by the comparative in vitro and in vivo uptake of two HCV polymerase antivirals, BILB 1941 and BI compound A. Inhibition of uptake by the well-known organic anion-transporting polypeptide (OATP) inhibitor rifampin (12) was tested over the concentration range of 10 to 1,000 μM. Experiments were usually carried out at 37°C, but in some cases, they were performed under an ice-cold condition (1.2°C) to evaluate the role of the active uptake mechanism, as described in Results.

RESULTS

We first explored whether a simple model could be easily set up to predict in vivo Kp_liver. As shown in Fig. 1A, in vitro Kp_hep derived from freshly prepared hepatocytes correlated well the Kp_hep from cryopreserved hepatocytes, which enabled the use of cryopreserved hepatocytes from pooled donors to minimize donor-related variability. Initial experiments using a set of marketed drugs (warfarin, phenytoin, diazepam, verapamil, quinidine, and imipramine) showed a reasonable correlation between in vivo Kp_liver and Kp_hep, in the hepatocyte suspension without plasma protein, but the correlation was very poor for the HCV protease and polymerase inhibitors that we tested (Fig. 1B). Inclusion of 50% plasma protein in the hepatocyte distribution assay improved the correlation, particularly for the HCV DAA compounds (Fig. 1C). Nevertheless, the correlation was very poor if the data for marketed drugs and HCV DAAs were pooled.

In vitro-in vivo liver distribution data (Table 2) clearly demonstrated that only a few drugs showed a simple linear relationship with a cross-species slope close to unity between the in vitro Kp_hep and the observed in vivo Kp_liver (e.g., doxorubicin and verapamil [Table 2]). In comparison, a proportional relationship could be observed across multiple species for all drugs examined. Using this in vitro-in vivo proportionality method, based on the combined data from two rodent species plus the in vitro human Kp_hep, the human Kp_liver of all tested literature drugs were well modeled, with a prediction error less than 45% of the published values (Table 2).

Table 2.

Cross-species proportionality of Kp_hep and Kp_liver and validation of predictability of human liver Kp^a

Drug	Kphep			Kpliver			Predicted human Kpliver
	Human	Rat	Mouse	Human	Rat	Mouse	Value	Error (%)
Doxorubicin	30.5 ± 4.7	154 ± 22	228 ± 11	38.1	136	207	28 ± 5.4	−28
Verapamil	7.7 ± 1.4	8.9 ± 1.2	6.7 ± 2.3	6.9	6.7	10.3	9.2 ± 1.1	33
Digoxin	1.5 ± 0.1	1.8 ± 0.4	1.1 ± 0.8	11.0	14.2	2.2	7.9 ± 1.6	−28
Quinidine	4.8 ± 0.5	11.4 ± 3.2	10.5 ± 1.4	13.3	25.4	9.2	7.7 ± 0.8	−42
Imipramine	5.4 ± 0.9	21.4 ± 4.5	14.8 ± 1.4	8.5	32.5	15.0	6.9 ± 1.4	−18

^aIn vivo mean Kp_liver values were taken from previous publications (n = 2 to 8 for rats, n = 11 to 18 for humans, n = 3 to 6 for mice), as cited from references 5, 7, 8, 9, 13, 14, 15, 19, 24, 27, 36, and 38). In vitro Kp_hep (obtained from cryopreserved hepatocytes in the presence of 50% plasma protein) and predicted human Kp_liver values represent the means ± SDs from 3 determinations. Prediction errors were calculated by the percent differences of the predicted human Kp_liver values over those published values [(predicted − observed)/observed × 100].

The significantly improved exposure-response relationship with the incorporation of LCIQ is summarized in Fig. 2. As shown in Fig. 2A, the retrospective PK-PD data analysis based on short-term monotherapy data from four Boehringer Ingelheim HCV antivirals, two NS3 proteases (BILN 2061 and faldaprevir), and two nonnucleoside NS5B polymerase inhibitors (BILB 1941 and BI 207127) (11, 16, 20, 22, 26) led to a poor correlation (r²) of 0.48 for the IQ-maximum viral load reduction (VLR) relationship analysis and a corresponding 50% effective dose (ED₅₀) IQ of 42, with a large 95% confidence interval (CI), 0.1 to 148534. Exploration of the IQ-VLR relationship analysis using both maximum concentration of drug in plasma (C_max) and AUC exposures failed to improve that relationship compared to the analysis using C_min. In comparison, incorporation of the modeled human liver Kp as a correction factor permitted LCIQ-antiviral response to be fit into a typical sigmoidal exposure-response curve with an improved r² of 0.95 (Fig. 2B), giving a maximum predicted VLR plateau of 3.4 log₁₀ IU/ml, with a 95% CI of 3.1 to 3.7. The corresponding ED₅₀ was LCIQ 121, with a 95% CI ranging from 83 to 177 (Fig. 2B). PK-PD data reported for telaprevir also fit reasonably well into the LCIQ-VLR relationship curve (Fig. 2B).

Fig 2.

The relationship between plasma IQ (A) or LCIQ (B) and the mean maximum viral load reductions observed following monotherapy with four BI anti-HCV DAAs over 2 to 5 days. Each point represents the group mean from each dose (n = 6 to 12). The data for all DAAs were fit into a sigmoidal dose-response curve equation with variable slope as described in Materials and Methods. The LCIQ-VLR relationship of telaprevir (2,250 mg/day) was also plotted for comparison.

Our observations indicated that liver Kp has a significant impact on antiviral efficacy of HCV DAAs, and we sought a mechanistic explanation for the differences in liver uptake observed for different inhibitors. The potential role of active uptake transporters was examined by the comparative in vitro and in vivo results for two HCV polymerase inhibitors, BILB 1941 and BI compound A (Fig. 3). The higher Kp_hep of BI compound A is clearly more temperature dependent (Fig. 3A and B) and more affected by the well-known OATP inhibitor rifampin over the concentration range of 10 to 1,000 μM. Since the transporter-mediated active uptake is known to be mostly inhibited at ice-cold temperature (29), the differences between Kp_hep at 37°C and the ice-cold condition reflect hepatocyte uptake mostly attributable to the active process, and thus more affected by the inhibitor rifampin. In vivo, rifampin increased the plasma exposure in rats and decreased the liver Kp of BI compound A more than that of BILB 1941, consistent with OATP-mediated liver uptake of compound A (Fig. 3C and D).

DISCUSSION

Allometry has been used as one of the common approaches for human PK predictions (25). However, based on data from the present study and those published from others, cross-species liver distribution clearly does not follow the trend of allometry. Part of the reason for the lack of human liver Kp allometry is the cross-species variability of active uptake transporters (17). Indeed, several members of the OATP transporter family that have been suggested to mediate active liver uptake are not evolutionarily well conserved, and human OATP orthologues may not exist in rodents (12). In addition, profound differences in substrate affinity between rat and mouse OATP2 have been reported (17, 37). In the present study, the effects of OATP uptake inhibition by rifampin further supported the contribution of active uptake transporters in the observed high Kp_liver and highlight the importance of in vitro Kp_hep in the correct cross-species liver Kp prediction. Therefore, in vitro hepatocyte data measured as Kp_hep may provide an important value to index the cross-species differences in liver active uptake. On the other hand, in vitro Kp_hep alone was not able to predict human Kp_liver correctly for all the drugs, due to the contribution of other factors. These seem to be captured reasonably well by the incorporation of in vivo liver Kp from two rodent species. Similar to the improved human in vivo clearance prediction with the in vitro-in vivo extrapolation method (23), many in vivo factors may be important for the prediction of in vivo Kp_liver, such as plasma and tissue binding and passive permeability. Some of these factors may be consistent across species (23, 25) and captured well by the incorporation of in vivo liver Kp from more than one preclinical species.

The importance of plasma protein binding was evidenced by the significant improvement in the correlation between the observed rat Kp_liver and the Kp_hep measured in the presence of 50% plasma protein for these highly protein-bound HCV DAAs (>99%). In contrast, the effect of plasma protein binding was not very obvious for the literature drugs, for which protein binding is below 95%. In addition, the relationship between in vitro Kp_hep and in vivo rat Kp_liver seems to follow two distinct correlation relationships, further supporting the idea that other factors contribute to the in vivo liver distribution.

Liver is the predominant organ for HCV viral replication (6, 30), and the data from the present study strongly support explicit consideration of liver exposure for PK-PD data analysis of HCV DAAs. One of the potential reasons for the poor IQ-VLR relationship might be related to the interpatient variability, as well as limited dose groups. Therefore, IQ determinations from all individual patients, with a broader dose range in the clinical studies, might lead to improved PK-PD analysis. However, this additional interpatient variability would not account for the differential distribution to the liver and would not fully explain drug versus drug differences in a PK-PD relationship. Nonetheless, incorporation of interpatient variability into the LCIQ-PD relationship analysis may further improve the model.

The present study provides a novel in vitro-in vivo proportionality method to model human Kp_liver. The human Kp_liver-corrected plasma IQ, or LCIQ, correlated well with the observed viral load reduction in the setting of short-term monotherapy with several different anti-HCV agents. The plasma exposure required to achieve a targeted LCIQ can be coupled with allometry to guide clinical dosage estimation and thus safety window assessment, as well as candidate selection of drugs targeting the liver, as illustrated for agents against HCV.