- Email: [email protected]
Clinical Exposure Boost Predictions by Integrating Cytochrome P450 3A4–Humanized Mouse Studies With PBPK Modeling
Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
Contents lists available at ScienceDirect
Journal of Pharmaceutical Sciences journal homepage: www.jpharmsci.org
Drug DiscoveryeDevelopment Interface
Clinical Exposure Boost Predictions by Integrating Cytochrome P450 3A4eHumanized Mouse Studies With PBPK Modeling Jin Zhang 1, Tycho Heimbach 1, Nico Scheer 2, Avantika Barve 1, Wenkui Li 1, Wen Lin 1, Handan He 1, * 1 2
Drug Metabolism & Pharmacokinetics Department, Novartis Institutes for Biomedical Research, East Hanover, New Jersey 07936 Business Development Department, CEVEC Pharmaceuticals GmbH, Cologne 51105, Germany
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 December 2015 Revised 19 January 2016 Accepted 20 January 2016
NVS123 is a poorly water-soluble protease 56 inhibitor in clinical development. Data from in vitro hepatocyte studies suggested that NVS123 is mainly metabolized by CYP3A4. As a consequence of limited solubility, NVS123 therapeutic plasma exposures could not be achieved even with high doses and optimized formulations. One approach to overcome NVS123 developability issues was to increase plasma exposure by coadministrating it with an inhibitor of CYP3A4 such as ritonavir. A clinical boost effect was predicted by using physiologically based pharmacokinetic (PBPK) modeling. However, initial boost predictions lacked sufficient confidence because a key parameter, fraction of drug metabolized by CYP3A4 (fmCYP3A4), could not be estimated with accuracy on account of disconnects between in vitro and in vivo preclinical data. To accurately estimate fmCYP3A4 in human, an in vivo boost effect study was conducted using CYP3A4-humanized mouse model which showed a 33- to 56-fold exposure boost effect. Using a topdown approach, human fmCYP3A4 for NVS123 was estimated to be very high and included in the human PBPK modeling to support subsequent clinical study design. The combined use of the in vivo boost study in CYP3A4-humanized mouse model mice along with PBPK modeling accurately predicted the clinical outcome and identified a significant NVS123 exposure boost (~42-fold increase) with ritonavir. © 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
Keywords: clinical pharmacokinetics CYP enzymes drug interaction drug metabolizing enzymes elimination hepatic clearance interspecies scaling physiologically based pharmacokinetic modeling preclinical pharmacokinetics simulations
Introduction Protease inhibitors often require high-systemic exposure levels for efficacy. Exposure “boosts” can be achieved for compounds that are highly metabolized by CYP3A4.1 Coadministration with a strong CYP3A4 inhibitor such as ritonavir (RTV) can significantly increase the exposure of several CYP3A4 substrates by inhibiting their CYP3A4-mediated metabolic clearance.1,2 RTV has been shown to increase the exposure of a coadministered drug, such as Danoprevir's, thus leading to a reduction in dose, dosing frequency, and alleviating food restrictions.3 Moreover, overall high exposure variability (CV ~ 60%) can be reduced with RTV, likely by inhibiting intestinal CYP3A4.1,2
Abbreviations: AUC, area under the plasma concentration-time curve; Cmax, maximum plasma concentration; CYP/Cyp, cytochrome P450; DDI, drug-drug interaction; fmCYP3A4, fraction of drug metabolized by CYP3A4; hCYP3A4, CYP3A4 humanized mice; HLM, human liver microsomal; PBPK, physiologically based pharmacokinetic; PI, protease inhibitor; PK, pharmacokinetics; RTV, ritonavir. * Correspondence to: Handan He (Telephone: 862-778-3353; Fax: 973-781-5023). E-mail address: [email protected] (H. He).
NVS123 is an anti-HCV drug in early drug development. It had been anticipated that the NVS123 human plasma exposures could be below the efficacious concentrations due to poor solubility and dissolution along with high CYP3A4-mediated clearance. A RTVmediated exposure boost strategy to support further clinical development of NVS123 was pursued. To predict the clinical RTV boost effect on NVS123, a human physiologically based pharmacokinetic (PBPK) model was established a priori based on in vitro and in vivo preclinical pharmacokinetic (PK) data using Simcyp (Simcyp Ltd, Sheffield, UK, version 13 and Release 1), a population-based clinical trial simulator for pharmacokinetics and drug-drug interaction (DDI) predictions. The predictions provided a wide range of boost effect (2- to 40-fold increase) as a key in vitro input parameter, fraction of drug metabolized by CYP3A4 (fmCYP3A4), could not be estimated with confidence on account of the disconnect between in vitro human liver microsomal data and in vivo rat absorption, distribution, metabolism, and excretion (ADME) data. Given the uncertainty in exposure boost, using conventional in vitro and in vivo data, emerging alternative approaches were used. In vivo boost effects or DDIs can be evaluated in humanized
http://dx.doi.org/10.1016/j.xphs.2016.01.021 0022-3549/© 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
1399
Figure 1. Minimal PBPK model scheme. Dosing of NVS123 was conducted with or without RTV in humanized mice or humans. SAC representing a lump of tissues excluding the liver and portal vein; kin and kout: first-order rate constants between systemic compartment and SAC. SAC, single adjusting compartment.
mouse models, generated either by a replacement of particular mouse genes involved in drug metabolism and disposition with their corresponding human counterparts4,5 or by transplantation of human hepatocytes into immune-deficient mice to obtain chimeric liver-humanized mice.6-8 Despite numerous examples demonstrating the utility of such models to overcome species differences, data showing how these humanized mice can be used to quantitatively predict DDIs in humans are still sparse.9,10 Moreover, although previous studies focused on standard compounds with known clinical outcome, the utility of these models to prospectively predict the extent of DDIs in humans still needs to be demonstrated. Here, we describe a case example of the combined use of a genetically humanized CYP3A4 mouse model and PBPK modeling to prospectively estimate the increase in human exposure of a CYP3A4 substrate, NVS123, due to coadministration with the established strong CYP3A4 inhibitor RTV. We show that this approach accurately predicted the clinical DDI or boost effect a priori, as subsequently confirmed in a clinical study.
Key PBPK Model Input Parameters A first-order absorption model was used to estimate drug absorption in human based on Caco-2 permeability data. A minimal PBPK model described drug distribution. Enzyme kinetics and CLint for CYP3A4 were estimated using the retrograde model. Key pharmacokinetic parameters in the model have been listed in Table 2. Using a “top-down” approach, the PBPK model was verified with human PK data by comparing observed human PK profiles at 2 dose levels with simulated profiles (Figs. 2a and 2b). The simulated PK profiles could reasonably describe observed human PK profiles in healthy volunteers after a single 100-mg or 300-mg NVS123 dose. PBPK Model for Ritonavir “Boost” The default PBPK model for RTV in Simcyp, version 13.1, was slightly modified by incorporating published mechanism-based Table 2 Key Pharmacokinetic Parameters of NVS123
Materials and Methods
Parameter
PBPK Model NVS123
Absorption (model used: first-order model) Peff,man (104 cm/s) 1.52
The PBPK model for NVS123 was built by using Simcyp simulator (Simcyp Population-based Simulator, version 13.1, Certara, Princeton, NJ). Figure 1 depicts the minimal PBPK model, which consists mainly of the liver, the portal vein, the systemic compartment, and the tissue sac (single adjusting compartment). Gut absorption is captured via an absorption rate constant. Physicochemical Properties Key NVS123 physicochemical and biopharmaceutic properties, which provide a fundamental input to the PBPK model in determining drug absorption and distribution, are listed in Table 1.
Table 1 Physicochemical and Biopharmaceutic Properties of NVS123 Parameters
Measured
LogP pKa Solubility (mg/mL) Caco-2 permeability with inhibitor B/P
5.0 5.8 (acidic), 8.36 (basic) 0.09 (buffer, pH 7.4) 3.7 e6 cm/s 0.62
Initial Value
Caco-2 (106 cm/s) Qgut (L/h)
3.7 7.77
fugut
0.5
Refined Value
Estimated from Caco-2 data Measured Scaled from Caco-2 data Based on study of Yang et al.11
Distribution (model used: minimal PBPK) Vss (L/kg) 0.4 Kin (1/h)
0.511
Kout (1/h)
0.381
Vsac (L/kg)
0.23
Elimination CLpo (L/h) fm_CYP3A4
CLR (L/h)_human CLint (mL/min/pmol cyp)
From human PK data From human PK data From human PK data From human PK data
600
300
0.04-1.0
1.0
0.77 0.16-4.0
Comment
From human PK data Rat ADME, HLM, hepatocyte, human oral CL From human ADME data Retrograde model
1400
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
inhibition data12 as RTV can both inhibit and induce human CYP3A, although the inhibitory effect was expected to be predominant at a 100-mg dose.13 A possible effect of 100-mg RTV on transporters was not included in this model.
technology16 was used to analyze the plasma concentration of NVS123 and RTV in mice.
LC-MS/MS Analysis
Humanized Mouse Model and In Vivo Studies Liver/GuteCYP3A4-humanized mouse model (hCYP3A4; 9049M, Taconic, Hudson, NY) mice described previously14 were used. Animals were allowed to acclimatize for 7 days before the experimental procedure. Mice were kept in accord with local laws and regulations and in temperature-controlled environments with a 12-h light cycle and given standard diets and water ad libitum. All animal procedures were approved by the Novartis Animal Care and Use Committee of Novartis Pharmaceuticals Corporation. Each animal was orally dosed with NVS123 at 50 mg/kg alone or coadministrated with RTV at 35 mg/kg single or daily as described in Table 3. NVS123 at 50 mg/kg in mouse is equivalent to human dose at 300 mg based on body surface area, whereas RTV dose at 35 mg/kg is equivalent to a 200-mg human dose.15 Blood samples were collected by serial bleeding from tail vein at 0.25, 0.5, 1, 3, 7, 24 h after dose of NVS123. Dried Blood Spot
A 1/800 (~3mm) dried blood spot punch of blank plasma (for control samples and zero time samples), calibration standards, QCs, and unknown samples were added into the appropriate well of a 96-well assay plate. A 200-mL aliquot of the ISTD working solution (200 ng/mL of [Mþ6]NVS123 and 200 ng/mL of [Mþ6]RTV with 20% water (containing 0.2% ZnSO4) in methanol, v/v) was added into all wells except the wells for the control blanks, to which a 200-mL aliquot of 20% water (containing 0.2% ZnSO4) in methanol (v/v) was added. The plate was vortex-mixed for approximately 5 min on a pulse-vortex mixer at a speed setting of about 50. The plate was sonicated for about 10 min. Using a TomTec system, the resulting supernatant was transferred into the corresponding well of a clean 1-mL 96-well plate and evaporated at approximately 45 C under a stream of nitrogen, followed by the addition of a 150-mL aliquot of 50% aqueous methanol with 0.1% formic acid (v/v). The plate was vortex-mixed well at a speed setting of about 50. A 20-mL aliquot of the extract was injected onto the LC-MS/MS system.16 The lower limit of quantification was 50.0 ng/mL, and the upper limit of quantification was 5000 ng/mL.
Figure 2. Comparison of Simcyp PBPK model predicted and observed plasma PK profiles of NVS123 in humans at different dose levels. (a) Single NVS123 dose of 100 mg, (b) Single NVS123 dose of 300 mg. The solid line with open circles shows the observed clinical PK profile (mean ± SD); the dashed line shows the predicted PK profiles.
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
1401
Table 3 Study Design for In Vivo DDI Study in hCYP3A4 Mice Group
Compound
Number of Mice
Dose
Strain of Mouse
1 2
NVS123 only NVS123 RTV NVS123 RTV
4 4
PO 50-mg/kg single dose PO 50-mg/kg single dose PO 35-mg/kg single dose PO 50-mg/kg single dose on day 8 PO 35 mg/kg QD 8 days
Male, hCYP3A4 Male, hCYP3A4
3
4
Male, hCYP3A4
Clinical Study A randomized, double-blind, placebo-controlled oral dose study was conducted in healthy volunteers to assess the boost effect of the CYP3A4 inhibitor RTV coadministered with NVS123. In this study, 4 subjects were dosed with 50-mg NVS123 alone (period 1) followed by a washout of approximately 5 days. In period 2, the same subjects were dosed by 100-mg bid RTV from the evening of day 1 through day 3, along with a single dose of 50-mg NVS123 on the morning of day 2. The plasma samples were collected at designated time points for analysis of NVS123 and RTV by an LC-MS/MS system. Results Estimation of Human fmCYP3A4 From Rat ADME and Human Liver Microsome Data Metabolic profile in human liver microsomes suggested that CYP3A4 is primarily responsible for NVS123 metabolism. However, rat in vivo ADME data and metabolic profile in hepatocytes showed the possibility that other pathways could also contribute to NVS123 clearance in vivo. As a result, the fmCYP3A4 for NVS123 estimated from rat ADME, in vitro human liver microsomal (HLM) data, hepatocyte data, and human oral CL ranged widely from 0.04 to 1 (4%-100%) as summarized in Table 4. When these fm values were incorporated into the Simcyp PBPK model, the differences in fmCYP3A4 values resulted in a wide range of predicted NVS123 exposure increases. Increases in NVS123 area under the plasma concentration-time curve (AUC) were predicted to be between 1- to 45-fold. NVS123 maximum plasma concentrations (Cmax) were predicted to increase between 1- to 18-fold (Fig. 3). These wide prediction ranges left the clinical team with a significant uncertainty to select an optimal starting dose for a clinical RTV boost study.
Figure 3. Predicted change of plasma Cmax and AUC of NVS123 in humans when codosed with ritonavir. Predictions were made using Simcyp PBPK modeling for single dose of 100-mg NVS123 and 100-mg bid ritonavir. The fmCYP3A4 estimates covered the range of from values obtained from HLM data or rat ADME studies.
a single dose 35-mg/kg RTV dose or dosed after pretreatment with 35-mg/kg RTV once a day for 8 days. The NVS123 Cmax and AUC were increased by 8.3- and 32.7-fold, respectively, when codosed simultaneously with a single dose of 35-mg/kg RTV. After multiple dose RTV pretreatment, the Cmax and AUC were increased by 10.1- and 56.3-fold, respectively, suggesting stronger inhibition of CYP3A4 likely due to higher RTV concentration (Table 5 and Fig. 4). Extrapolating the data from the in vivoehumanized mouse DDI study to humans suggested fmCYP3A4 for NVS123 to be near 1.0, that is, NVS123 is expected to be mainly metabolized by CYP3A4 in man. Refined PBPK Model Using a “Bottom-Up Approach” for fmCYP3A4 Obtained From hCYP3A4 Mice The fmCYP3A4 value obtained from the in vivo study in hCYP3A4 mice described previously was used to estimate the boost effect in human exposure of NVS123 following codosing with RTV by Simcyp PBPK modeling. The observed human oral clearance, CLpo, was 300 L/h (from a human multiple ascending NVS123 only dose study). The predicted profiles of NVS123 in plasma with or without coadministration of RTV are shown in Figure 5. With pretreatment of 100-mg bid RTV, a significant boost effect was predicted a priori with an average of ~34.6-fold increase in AUC and ~12.8-fold increase in Cmax of NVS123 compared with those with NVS123 alone (Table 6).
In Vivo Boost Study Using CYP3A4-Humanized Mice Clinical DDI Study With NVS123 and RTV An in vivo DDI study in which NVS123 was dosed alone at 50 mg/kg (equivalent to human dose of 300 mg based on body surface area, FDA Guidance for Industry and Reviewers, 2005), or codosed with the CYP3A4 inhibitor RTV at 35 mg/kg (equivalent to human dose of 200 mg based on body surface area, FDA Guidance for Industry and Reviewers, 2005), was conducted in hCYP3A4 mice (n ¼ 4 mice per dose group). In the codosed groups, 50-mg/kg NVS123 was either coadministrated simultaneously with
A clinical DDI study with NVS123 at 50-mg single dose and a 100-mg bid RTV dose was subsequently conducted in healthy Table 5 Plasma Exposure of NVS123 in Humanized Mouse With or Without Coadministration of RTV PK parameters NVS123 Only NVS123 SD þ RTV SD
NVS123 SD þ RTV MD
Fold Increased
Table 4 Estimation of fmCYP3A4 for NVS123 fmCYP3A4
Data
1.0 0.7 0.2 0.04
Metabolic profile in HLM Human oral CL assuming F ¼ 10% Metabolic profile hepatocyte Metabolic profile in rat excreta
Cmax (ng/mL) AUC (ng$h/mL) T1/2 (h)
495
4090
899
29,400
1.2
3.1
SD, single dose; MD, multiple dose.
8.3 32.7 2.6
Fold Increased 4980
10.1
50,600
56.3
2.2
1.8
1402
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404 Table 6 Simulated Boost Effect of 100-mg Ritonavir Pretreatment on 50-mg NVS123 in Human Using Refined PBPK Model
Figure 4. Pharmacokinetics of NVS123 in hCYP3A4 mice (n ¼ 4) with and without coadministration of RTV. The dashed line with open circle shows the PK profile of NVS123 alone at 50-mg/kg PO in hCYP3A4 mice; the solid line with open diamond shows the PK profile of NVS123 at 50 mg/kg after codosing a 35-mg/kg single RTV dose; the solid line with open triangles shows the PK profile of NVS123 at 50 mg/kg with a pretreatment of RTV at 35 mg/kg for 8 days. Data are presented as mean ± SD.
volunteers (n ¼ 4). The observed increase in plasma AUC of NVS123 when codosed with RTV in this study was high and was 42.5-fold (Figs. 6a and 6b). C12 h, Cmin, and Cmax were also increased by 156- and 9.3-fold, respectively (Table 7). The observed human exposure data were in good agreement with the prediction from the PBPK model using the high fmCYP3A4 value (0.99) estimated from the humanized mouse DDI boost study (Fig. 7). The boost effect (fold change in AUC and Cmax) observed in human was also in good agreement with that from the DDI study in hCYP3A4 mice.
Discussion The boosting of protease inhibitors (PIs) is a therapeutic strategy wherein a low RTV dose is given concurrently with another PI to enhance the PI's systemic exposure through the inhibition of cytochrome P450 enzymes, particularly CYP3A4. For many of the PIs, RTV boosting can result in improved drug levels that can increase efficacy, decrease pill burden, add flexibility to the dosing schedule, and remove fasting restrictions. Accurate estimation of
Figure 5. Predicted change of plasma PK profiles of 50-mg NVS123 in human with or without pretreatment with ritonavir at 100-mg bid starting at 12 h (12 h before dosing NVS123) using Simcyp PBPK modeling. The fmCYP3A4 value estimated from this in vivo DDI study in hCYP3A4 mice was high (0.99). The solid line shows the PK profile of NVS123 with RTV pretreatment, and the dashed line shows the PK profile of NVS123 without RTV pretreatment. 0 h is the time of starting RTV pretreatment.
Treatment
Cmax (ng/mL)
AUC (ng$h/mL)
NVS123 alone NVS123 þ RTV Fold increase
54.4 694 12.8
161 5593 34.6
the boost effect usually depends on preclinical data, which can now be facilitated by modeling and simulation approaches. The PBPK model is a mathematical approach for predicting the ADME of compounds in humans or animal species, by transcribing anatomic, physiological, physical, and chemical properties into mathematical algorithms.17 In addition to other applications, PBPK modeling can be used to anticipate the quantitative extent of PK-based DDIs,18 and there is an increasing demand from regulatory agencies to include such predictions in submission documents.19 This requirement is often satisfied by using commercially available PBPK-modeling softwares, such as Simcyp or GastroPlus™.20,21 Although these tools are continuously improved and often provide reliable outcomes, the quality of their predictions relies on an accurate appraisal of key parameters used for PBPK modeling. Accordingly, such predictions can be challenging if key parameters cannot be estimated with sufficient confidence. An important parameter for the estimation of the extent of such a DDI is the fraction of drug metabolized by CYP3A4 (fmCYP3A4) in humans, which is usually determined by various in vitro approaches such as HLM or hepatocyte data22,23 and preclinical in vivo studies, for example, rat ADME data.24 However, for NVS123, HLM, hepatocyte, and rat ADME data were not in good agreement and generated a wide range of fmCYP3A4 from 0.04 to 1. This range of fmCYP3A4 values resulted in a predicted increase in NVS123 AUC by coadministrated RTV of 1- to 45-fold, leading to a lack of confidence in this PBPK model to accurately predict the DDI or boost effect of RTV on NVS123 in human. Accordingly, an alternative approach had to be considered. We conducted an in vivo DDI study using a CYP3A4-humanized mouse model expressing human CYP3A4 in the liver and intestine to predict the clinical outcome and estimate human fmCYP3A4 for NVS123 and to use these data for further PBPK model refinement. The hCYP3A4 mice used in our study were generated by combining 2 transgenes with liver- and gut-specific CYP3A4 expression, respectively, on a mouse Cyp3a gene cluster knockout background.14 The impact of blocking CYP3A4 activity in the intestine on the plasma levels of a probe substrate was previously shown by a significant increase in docetaxel exposure in transgenic mice with gut-specific expression of CYP3A4 after treatment with RTV.14 A profound increase in triazolam exposure was reported in the humanized mice expressing CYP3A4 in both liver and intestine in response to the prototypical CYP3A4 inhibitor ketoconazole.25 Furthermore, ter Heine et al.26 used the humanized mice with liver- or gut-specific CYP3A4 expression to dissect the role of hepatic and intestinal CYP3A4 on lopinavir disposition by using RTV as an inhibitor. In spite of these promising results, data showing the utility of the hCYP3A4 mice to predict clinical DDIs are still very limited, and the prospective prediction of the extent of DDI in humans using these models has not yet been reported. To accurately predict the clinical DDI outcome, the human equivalent dose for both NVS123 and RTV based on body surface area was used in the in vivo DDI study with hCYP3A4 mice. Pretreatment with RTV at single dose and multiple doses for 8 days was assessed in terms of changes in NVS123 AUC and Cmax. The data showed that a single dose of RTV at 35 mg/kg can inhibit CYP3A4, reducing the metabolism of NVS123 in the liver and
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
1403
Figure 6. Observed pharmacokinetics of NVS123 in human healthy volunteers (n ¼ 4) with and without pretreatment of RTV. (a) Linear scale and (b) logarithmic scale. The solid line with open triangle shows the NVS123 PK at 50 mg without RTV pretreatment and the solid line with open circle shows the NVS123 PK at 50 mg with RTV pretreatment at 100-mg bid starting 12 h before NVS123 dosing (12 h). 0 h is the time of dosing NVS123. Data are presented as mean ± SD.
increasing the Cmax and AUC significantly by 8.3- and 32.7-fold, respectively. Multiple doses of RTV increased the Cmax and AUC of NVS123 to 10.1- and 56.3-fold, suggesting a stronger inhibition of CYP3A4. We speculate that the stronger inhibition of CYP3A4 with multiple doses of RTV is a likely consequence of higher RTV exposure, but we did not measure the RTV plasma profile in our study. RTV not only increased the plasma exposure of NVS123 but also prolonged the T1/2. Considering that NVS123 has minimal first-pass metabolism based on animal data, we believe the mechanism of the boost effect of RTV on NVS123 is predominantly related to the inhibition of CYP3A4 in the liver rather than the first-pass metabolism. Because plasma exposure of NVS123 could be boosted by 30- to 50-fold with low doses of RTV, we rationalized that a lower starting dose of NVS123 should be considered in the clinical trial. Another important outcome from the in vivo DDI study in the hCYP3A4 mice was the estimation of fmCYP3A4 for NVS123 in humans to be 1.0, meaning that NVS123 is metabolized only by CYP3A4. When this refined fmCYP3A4 value was used for PBPK modeling, the clinical boosting effect of RTV on NVS123 at different dosing combinations was accurately predicted (Fig. 7). The clinical team acknowledged the data from the mouse in vivo DDI study using hCYP3A4 as well as the updated PBPK model prediction and reduced the NVS123 starting dose in the clinical DDI study from 200 to 50 mg. This was considered a safe dose, as a single oral 1000-mg NVS123 dose had been tested in the clinic without issues (not shown). The clinical study at 50 mg resulted in a 9.3-fold increase in Cmax and 42.5-fold increase in AUC when NVS123 was given to healthy volunteers pretreated with RTV at 100-mg bid and thus confirmed the outcome from the humanized
Table 7 Human Plasma Exposure Parameters of NVS123 at 50-mg Single Dose With or Without Pretreatment With RTV at 100-mg Bid PK parameters
NVS123 Only
NVS123 SD þ RTV SD
Fold Increased
C12h (ng/mL) Cmax (ng/mL) AUC (ng$h/mL) T1/2 (h)
0.828 47.8 128 1
129 444 5438 6
156 9.3 42.5 6.0
mouse DDI study. In addition, C12 h of NVS123 was also increased 156 fold when pretreated with RTV at 100-mg bid. It has been reported that the trough drug concentration for HCV PIs may be important in minimizing or preventing the emergence of resistance, and trough drug concentration had the strongest relationship with maximum change in HCV RNA.27,28 Because RTV coadministration increases the minimum plasma concentration (Cmin) more pronounced than the Cmax and AUC of NVS123, we speculate that lower dose of NVS123 may achieve efficacious trough concentration while overall NVS123 exposure can be reduced. In summary, we demonstrated that the combination of an in vivo DDI study with NVS123 and the potent CYP3A4 inhibitor RTV in hCYP3A4 mice along with PBPK modeling allowed for a prospective prediction of the extent of DDI and boost effect in humans, which was subsequently confirmed in a clinical study.
Figure 7. Comparison of predicted and observed human NVS123 PK profiles at 50-mg single dose with or without boost by RTV at 100-mg bid from 12 h (12 h before NVS123 was administered). The open triangle shows the observed NVS123 PK alone at 50-mg single dose (data are presented as mean ± SD), and the open circle shows the observed NVS123 PK at 50 mg with RTV pretreatment at 100 mg from 12 h (both data are presented as mean ± SD). The dashed line represents the simulated NVS123 PK alone at 50-mg single dose, and the solid line represents the simulated NVS123 PK at 50 mg with RTV pretreatment at 100 mg from 12 h.
1404
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
Acknowledgments The authors thank Dr. Binfeng Xia for helpful discussions and for sharing his expertise in PBPK modeling. Tycho Heimbach, Handan He, Avantika Barve, Wen Lin, and Jin Zhang participated in research design. Jin Zhang conducted experiments. Tycho Heimbach, Wenkui Li, and Jin Zhang performed data analysis. Jin Zhang, Nico Scheer, Tycho Heimbach, Handan He, and Avantika Barve wrote or contributed to the writing of the article.
References 1. Kempf DJ, Klein C, Chen HJ, et al. Pharmacokinetic enhancement of the hepatitis C virus protease inhibitors VX-950 and SCH 503034 by co-dosing with ritonavir. Antivir Chem Chemother. 2007;18(3):163-167. 2. King JR, Wynn H, Brundage R, Acosta EP. Pharmacokinetic enhancement of protease inhibitor therapy. Clin Pharm. 2004;43(5):291-310. 3. Reddy MB, Chen Y, Haznedar JO, et al. Impact of low-dose ritonavir on danoprevir pharmacokinetics: results of computer-based simulations and a clinical drug-drug interaction study. Clin Pharm. 2012;51(7):457-465. 4. Scheer N, Wolf CR. Genetically humanized mouse models of drug metabolizing enzymes and transporters and their applications. Xenobiotica. 2014;44(2):96-108. 5. Shen HW, Jiang XL, Gonzalez FJ, Yu AM. Humanized transgenic mouse models for drug metabolism and pharmacokinetic research. Curr Drug Metab. 2011;12(10):997-1006. 6. Foster JR, Lund G, Sapelnikova S, Tyrrell DL, Kneteman NM. Chimeric rodents with humanized liver: bridging the preclinical/clinical trial gap in ADME/ toxicity studies. Xenobiotica. 2014;44(2):109-122. 7. Grompe M, Strom S. Mice with human livers. Gastroenterology. 2013;145(6): 1209-1214. 8. Peltz G. Can 'humanized' mice improve drug development in the 21st century? Trends Pharmacol Sci. 2013;34(5):255-260. 9. Hasegawa M, Kapelyukh Y, Tahara H, et al. Quantitative prediction of human pregnane X receptor and cytochrome P450 3A4 mediated drug-drug interaction in a novel multiple humanized mouse line. Mol Pharmacol. 2011;80(3): 518-528. 10. Hasegawa M, Tahara H, Inoue R, Kakuni M, Tateno C, Ushiki J. Investigation of drug-drug interactions caused by human pregnane X receptor-mediated induction of CYP3A4 and CYP2C subfamilies in chimeric mice with a humanized liver. Drug Metab Dispos. 2012;40(3):474-480. 11. Yang J, Jamei M, Yeo KR, et al. Prediction of intestinal first-pass drug metabolism. Current Drug Metabolism. 2007;8(7):676-684. 12. Mao J, Mohutsky MA, Harrelson JP, Wrighton SA, Hall SD. Prediction of CYP3A4-mediated drug-drug interactions using human hepatocytes suspended in human plasma. Drug Metab Dispos. 2011;39(4):591-602.
13. Foisy MM, Yakiwchuk EM, Hughes CA. Induction effects of ritonavir: implications of drug interactions. Ann Pharmacother. 2008;42(7):1048-1059. 14. van Herwaarden AE, Wagenaar E, van der Kruijssen CM, et al. Knockout of cytochrome P450 3A yields new mouse models for understanding xenobiotic metabolism. J Clin Invest. 2007;117(11):3583-3592. 15. Federal Drug Administration. Guidance for Industry and Reviewers: Estimating the safe starting dose in clinical trials for therapeutics in adult healthy volunteers; 2005. Available at: http://www.fda.gov/OHRMS/DOCKETS/98fr/02d-0492gdl0001-vol1.pdf. Accessed January 17, 2016. 16. Li W, Tse FL. Dried blood spot sampling in combination with LC-MS/MS for quantitative analysis of small molecules. Biomed Chromatogr. 2010;24(1): 49-65. 17. Espie P, Tytgat D, Sargentini-Maier ML, Poggesi I, Watelet JB. Physiologically based pharmacokinetics (PBPK). Drug Metab Rev. 2009;41(3):391-407. 18. Jones HM, Chen Y, Gibson C, et al. Physiological based pharmacokinetic modeling in drug discovery and development: a pharmaceutical industry perspective. Clin Pharm & Ther. 2015;97(3):247-262. 19. Prueksaritanont T, Chu X, Gibson C, et al. Drug-drug interaction studies: regulatory guidance and an industry perspective. AAPS J. 2013;15(3):629-645. 20. Heimbach T, Lakshminarayana SB, Hu W, He H. Practical anticipation of human efficacious doses and pharmacokinetics using in vitro and preclinical in vivo data. AAPS J. 2009;11(3):602-614. 21. Jamei M, Marciniak S, Feng K, Barnett A, Tucker G, Rostami-Hodjegan A. The Simcyp population-based ADME simulator. Expert Opin Drug Metab Toxicol. 2009;5(2):211-223. 22. Desbans C, Hilgendorf C, Lutz M, et al. Prediction of fraction metabolized via CYP3A in humans utilizing cryopreserved human hepatocytes from a set of 12 single donors. Xenobiotica. 2014;44(1):17-27. 23. Youdim KA, Zayed A, Dickins M, et al. Application of CYP3A4 in vitro data to predict clinical drug-drug interactions; predictions of compounds as objects of interaction. Br J Clin Pharmacol. 2008;65(5):680-692. 24. Chen Y, Jin JY, Mukadam S, Malhi V, Kenny JR. Application of IVIVE and PBPK modeling in prospective prediction of clinical pharmacokinetics: strategy and approach during the drug discovery phase with four case studies. Biopharm Drug Dispos. 2012;33(2):85-98. 25. van Waterschoot RA, Rooswinkel RW, Sparidans RW, van Herwaarden AE, Beijnen JH, Schinkel AH. Inhibition and stimulation of intestinal and hepatic CYP3A activity: studies in humanized CYP3A4 transgenic mice using triazolam. Drug Metab Dispos. 2009;37(12):2305-2313. 26. ter Heine R, Van Waterschoot RA, Keizer RJ, Beijnen JH, Schinkel AH, Huitema AD. An integrated pharmacokinetic model for the influence of CYP3A4 expression on the in vivo disposition of lopinavir and its modulation by ritonavir. J Pharm Sci. 2011;100(6):2508-2515. 27. Reesink HW, Zeuzem S, Weegink CJ, et al. Rapid decline of viral RNA in hepatitis C patients treated with VX-950: a phase Ib, placebo-controlled, randomized study. Gastroenterology. 2006;131(4):997-1002. 28. Rubino C, Bradford WZ, Forrest A, et al. Pharmacokinetic-pharmacodynamic (PK-PD) relationships for ITMN-191 in a phase 1 multiple ascending dose trial in patients with genotype 1 chronic hepatitis C (CHC) infection. Hepatology. 2008;48(Suppl.2):1140A-1141A.
Contents lists available at ScienceDirect
Journal of Pharmaceutical Sciences journal homepage: www.jpharmsci.org
Drug DiscoveryeDevelopment Interface
Clinical Exposure Boost Predictions by Integrating Cytochrome P450 3A4eHumanized Mouse Studies With PBPK Modeling Jin Zhang 1, Tycho Heimbach 1, Nico Scheer 2, Avantika Barve 1, Wenkui Li 1, Wen Lin 1, Handan He 1, * 1 2
Drug Metabolism & Pharmacokinetics Department, Novartis Institutes for Biomedical Research, East Hanover, New Jersey 07936 Business Development Department, CEVEC Pharmaceuticals GmbH, Cologne 51105, Germany
a r t i c l e i n f o
a b s t r a c t
Article history: Received 10 December 2015 Revised 19 January 2016 Accepted 20 January 2016
NVS123 is a poorly water-soluble protease 56 inhibitor in clinical development. Data from in vitro hepatocyte studies suggested that NVS123 is mainly metabolized by CYP3A4. As a consequence of limited solubility, NVS123 therapeutic plasma exposures could not be achieved even with high doses and optimized formulations. One approach to overcome NVS123 developability issues was to increase plasma exposure by coadministrating it with an inhibitor of CYP3A4 such as ritonavir. A clinical boost effect was predicted by using physiologically based pharmacokinetic (PBPK) modeling. However, initial boost predictions lacked sufficient confidence because a key parameter, fraction of drug metabolized by CYP3A4 (fmCYP3A4), could not be estimated with accuracy on account of disconnects between in vitro and in vivo preclinical data. To accurately estimate fmCYP3A4 in human, an in vivo boost effect study was conducted using CYP3A4-humanized mouse model which showed a 33- to 56-fold exposure boost effect. Using a topdown approach, human fmCYP3A4 for NVS123 was estimated to be very high and included in the human PBPK modeling to support subsequent clinical study design. The combined use of the in vivo boost study in CYP3A4-humanized mouse model mice along with PBPK modeling accurately predicted the clinical outcome and identified a significant NVS123 exposure boost (~42-fold increase) with ritonavir. © 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
Keywords: clinical pharmacokinetics CYP enzymes drug interaction drug metabolizing enzymes elimination hepatic clearance interspecies scaling physiologically based pharmacokinetic modeling preclinical pharmacokinetics simulations
Introduction Protease inhibitors often require high-systemic exposure levels for efficacy. Exposure “boosts” can be achieved for compounds that are highly metabolized by CYP3A4.1 Coadministration with a strong CYP3A4 inhibitor such as ritonavir (RTV) can significantly increase the exposure of several CYP3A4 substrates by inhibiting their CYP3A4-mediated metabolic clearance.1,2 RTV has been shown to increase the exposure of a coadministered drug, such as Danoprevir's, thus leading to a reduction in dose, dosing frequency, and alleviating food restrictions.3 Moreover, overall high exposure variability (CV ~ 60%) can be reduced with RTV, likely by inhibiting intestinal CYP3A4.1,2
Abbreviations: AUC, area under the plasma concentration-time curve; Cmax, maximum plasma concentration; CYP/Cyp, cytochrome P450; DDI, drug-drug interaction; fmCYP3A4, fraction of drug metabolized by CYP3A4; hCYP3A4, CYP3A4 humanized mice; HLM, human liver microsomal; PBPK, physiologically based pharmacokinetic; PI, protease inhibitor; PK, pharmacokinetics; RTV, ritonavir. * Correspondence to: Handan He (Telephone: 862-778-3353; Fax: 973-781-5023). E-mail address: [email protected] (H. He).
NVS123 is an anti-HCV drug in early drug development. It had been anticipated that the NVS123 human plasma exposures could be below the efficacious concentrations due to poor solubility and dissolution along with high CYP3A4-mediated clearance. A RTVmediated exposure boost strategy to support further clinical development of NVS123 was pursued. To predict the clinical RTV boost effect on NVS123, a human physiologically based pharmacokinetic (PBPK) model was established a priori based on in vitro and in vivo preclinical pharmacokinetic (PK) data using Simcyp (Simcyp Ltd, Sheffield, UK, version 13 and Release 1), a population-based clinical trial simulator for pharmacokinetics and drug-drug interaction (DDI) predictions. The predictions provided a wide range of boost effect (2- to 40-fold increase) as a key in vitro input parameter, fraction of drug metabolized by CYP3A4 (fmCYP3A4), could not be estimated with confidence on account of the disconnect between in vitro human liver microsomal data and in vivo rat absorption, distribution, metabolism, and excretion (ADME) data. Given the uncertainty in exposure boost, using conventional in vitro and in vivo data, emerging alternative approaches were used. In vivo boost effects or DDIs can be evaluated in humanized
http://dx.doi.org/10.1016/j.xphs.2016.01.021 0022-3549/© 2016 American Pharmacists Association®. Published by Elsevier Inc. All rights reserved.
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
1399
Figure 1. Minimal PBPK model scheme. Dosing of NVS123 was conducted with or without RTV in humanized mice or humans. SAC representing a lump of tissues excluding the liver and portal vein; kin and kout: first-order rate constants between systemic compartment and SAC. SAC, single adjusting compartment.
mouse models, generated either by a replacement of particular mouse genes involved in drug metabolism and disposition with their corresponding human counterparts4,5 or by transplantation of human hepatocytes into immune-deficient mice to obtain chimeric liver-humanized mice.6-8 Despite numerous examples demonstrating the utility of such models to overcome species differences, data showing how these humanized mice can be used to quantitatively predict DDIs in humans are still sparse.9,10 Moreover, although previous studies focused on standard compounds with known clinical outcome, the utility of these models to prospectively predict the extent of DDIs in humans still needs to be demonstrated. Here, we describe a case example of the combined use of a genetically humanized CYP3A4 mouse model and PBPK modeling to prospectively estimate the increase in human exposure of a CYP3A4 substrate, NVS123, due to coadministration with the established strong CYP3A4 inhibitor RTV. We show that this approach accurately predicted the clinical DDI or boost effect a priori, as subsequently confirmed in a clinical study.
Key PBPK Model Input Parameters A first-order absorption model was used to estimate drug absorption in human based on Caco-2 permeability data. A minimal PBPK model described drug distribution. Enzyme kinetics and CLint for CYP3A4 were estimated using the retrograde model. Key pharmacokinetic parameters in the model have been listed in Table 2. Using a “top-down” approach, the PBPK model was verified with human PK data by comparing observed human PK profiles at 2 dose levels with simulated profiles (Figs. 2a and 2b). The simulated PK profiles could reasonably describe observed human PK profiles in healthy volunteers after a single 100-mg or 300-mg NVS123 dose. PBPK Model for Ritonavir “Boost” The default PBPK model for RTV in Simcyp, version 13.1, was slightly modified by incorporating published mechanism-based Table 2 Key Pharmacokinetic Parameters of NVS123
Materials and Methods
Parameter
PBPK Model NVS123
Absorption (model used: first-order model) Peff,man (104 cm/s) 1.52
The PBPK model for NVS123 was built by using Simcyp simulator (Simcyp Population-based Simulator, version 13.1, Certara, Princeton, NJ). Figure 1 depicts the minimal PBPK model, which consists mainly of the liver, the portal vein, the systemic compartment, and the tissue sac (single adjusting compartment). Gut absorption is captured via an absorption rate constant. Physicochemical Properties Key NVS123 physicochemical and biopharmaceutic properties, which provide a fundamental input to the PBPK model in determining drug absorption and distribution, are listed in Table 1.
Table 1 Physicochemical and Biopharmaceutic Properties of NVS123 Parameters
Measured
LogP pKa Solubility (mg/mL) Caco-2 permeability with inhibitor B/P
5.0 5.8 (acidic), 8.36 (basic) 0.09 (buffer, pH 7.4) 3.7 e6 cm/s 0.62
Initial Value
Caco-2 (106 cm/s) Qgut (L/h)
3.7 7.77
fugut
0.5
Refined Value
Estimated from Caco-2 data Measured Scaled from Caco-2 data Based on study of Yang et al.11
Distribution (model used: minimal PBPK) Vss (L/kg) 0.4 Kin (1/h)
0.511
Kout (1/h)
0.381
Vsac (L/kg)
0.23
Elimination CLpo (L/h) fm_CYP3A4
CLR (L/h)_human CLint (mL/min/pmol cyp)
From human PK data From human PK data From human PK data From human PK data
600
300
0.04-1.0
1.0
0.77 0.16-4.0
Comment
From human PK data Rat ADME, HLM, hepatocyte, human oral CL From human ADME data Retrograde model
1400
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
inhibition data12 as RTV can both inhibit and induce human CYP3A, although the inhibitory effect was expected to be predominant at a 100-mg dose.13 A possible effect of 100-mg RTV on transporters was not included in this model.
technology16 was used to analyze the plasma concentration of NVS123 and RTV in mice.
LC-MS/MS Analysis
Humanized Mouse Model and In Vivo Studies Liver/GuteCYP3A4-humanized mouse model (hCYP3A4; 9049M, Taconic, Hudson, NY) mice described previously14 were used. Animals were allowed to acclimatize for 7 days before the experimental procedure. Mice were kept in accord with local laws and regulations and in temperature-controlled environments with a 12-h light cycle and given standard diets and water ad libitum. All animal procedures were approved by the Novartis Animal Care and Use Committee of Novartis Pharmaceuticals Corporation. Each animal was orally dosed with NVS123 at 50 mg/kg alone or coadministrated with RTV at 35 mg/kg single or daily as described in Table 3. NVS123 at 50 mg/kg in mouse is equivalent to human dose at 300 mg based on body surface area, whereas RTV dose at 35 mg/kg is equivalent to a 200-mg human dose.15 Blood samples were collected by serial bleeding from tail vein at 0.25, 0.5, 1, 3, 7, 24 h after dose of NVS123. Dried Blood Spot
A 1/800 (~3mm) dried blood spot punch of blank plasma (for control samples and zero time samples), calibration standards, QCs, and unknown samples were added into the appropriate well of a 96-well assay plate. A 200-mL aliquot of the ISTD working solution (200 ng/mL of [Mþ6]NVS123 and 200 ng/mL of [Mþ6]RTV with 20% water (containing 0.2% ZnSO4) in methanol, v/v) was added into all wells except the wells for the control blanks, to which a 200-mL aliquot of 20% water (containing 0.2% ZnSO4) in methanol (v/v) was added. The plate was vortex-mixed for approximately 5 min on a pulse-vortex mixer at a speed setting of about 50. The plate was sonicated for about 10 min. Using a TomTec system, the resulting supernatant was transferred into the corresponding well of a clean 1-mL 96-well plate and evaporated at approximately 45 C under a stream of nitrogen, followed by the addition of a 150-mL aliquot of 50% aqueous methanol with 0.1% formic acid (v/v). The plate was vortex-mixed well at a speed setting of about 50. A 20-mL aliquot of the extract was injected onto the LC-MS/MS system.16 The lower limit of quantification was 50.0 ng/mL, and the upper limit of quantification was 5000 ng/mL.
Figure 2. Comparison of Simcyp PBPK model predicted and observed plasma PK profiles of NVS123 in humans at different dose levels. (a) Single NVS123 dose of 100 mg, (b) Single NVS123 dose of 300 mg. The solid line with open circles shows the observed clinical PK profile (mean ± SD); the dashed line shows the predicted PK profiles.
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
1401
Table 3 Study Design for In Vivo DDI Study in hCYP3A4 Mice Group
Compound
Number of Mice
Dose
Strain of Mouse
1 2
NVS123 only NVS123 RTV NVS123 RTV
4 4
PO 50-mg/kg single dose PO 50-mg/kg single dose PO 35-mg/kg single dose PO 50-mg/kg single dose on day 8 PO 35 mg/kg QD 8 days
Male, hCYP3A4 Male, hCYP3A4
3
4
Male, hCYP3A4
Clinical Study A randomized, double-blind, placebo-controlled oral dose study was conducted in healthy volunteers to assess the boost effect of the CYP3A4 inhibitor RTV coadministered with NVS123. In this study, 4 subjects were dosed with 50-mg NVS123 alone (period 1) followed by a washout of approximately 5 days. In period 2, the same subjects were dosed by 100-mg bid RTV from the evening of day 1 through day 3, along with a single dose of 50-mg NVS123 on the morning of day 2. The plasma samples were collected at designated time points for analysis of NVS123 and RTV by an LC-MS/MS system. Results Estimation of Human fmCYP3A4 From Rat ADME and Human Liver Microsome Data Metabolic profile in human liver microsomes suggested that CYP3A4 is primarily responsible for NVS123 metabolism. However, rat in vivo ADME data and metabolic profile in hepatocytes showed the possibility that other pathways could also contribute to NVS123 clearance in vivo. As a result, the fmCYP3A4 for NVS123 estimated from rat ADME, in vitro human liver microsomal (HLM) data, hepatocyte data, and human oral CL ranged widely from 0.04 to 1 (4%-100%) as summarized in Table 4. When these fm values were incorporated into the Simcyp PBPK model, the differences in fmCYP3A4 values resulted in a wide range of predicted NVS123 exposure increases. Increases in NVS123 area under the plasma concentration-time curve (AUC) were predicted to be between 1- to 45-fold. NVS123 maximum plasma concentrations (Cmax) were predicted to increase between 1- to 18-fold (Fig. 3). These wide prediction ranges left the clinical team with a significant uncertainty to select an optimal starting dose for a clinical RTV boost study.
Figure 3. Predicted change of plasma Cmax and AUC of NVS123 in humans when codosed with ritonavir. Predictions were made using Simcyp PBPK modeling for single dose of 100-mg NVS123 and 100-mg bid ritonavir. The fmCYP3A4 estimates covered the range of from values obtained from HLM data or rat ADME studies.
a single dose 35-mg/kg RTV dose or dosed after pretreatment with 35-mg/kg RTV once a day for 8 days. The NVS123 Cmax and AUC were increased by 8.3- and 32.7-fold, respectively, when codosed simultaneously with a single dose of 35-mg/kg RTV. After multiple dose RTV pretreatment, the Cmax and AUC were increased by 10.1- and 56.3-fold, respectively, suggesting stronger inhibition of CYP3A4 likely due to higher RTV concentration (Table 5 and Fig. 4). Extrapolating the data from the in vivoehumanized mouse DDI study to humans suggested fmCYP3A4 for NVS123 to be near 1.0, that is, NVS123 is expected to be mainly metabolized by CYP3A4 in man. Refined PBPK Model Using a “Bottom-Up Approach” for fmCYP3A4 Obtained From hCYP3A4 Mice The fmCYP3A4 value obtained from the in vivo study in hCYP3A4 mice described previously was used to estimate the boost effect in human exposure of NVS123 following codosing with RTV by Simcyp PBPK modeling. The observed human oral clearance, CLpo, was 300 L/h (from a human multiple ascending NVS123 only dose study). The predicted profiles of NVS123 in plasma with or without coadministration of RTV are shown in Figure 5. With pretreatment of 100-mg bid RTV, a significant boost effect was predicted a priori with an average of ~34.6-fold increase in AUC and ~12.8-fold increase in Cmax of NVS123 compared with those with NVS123 alone (Table 6).
In Vivo Boost Study Using CYP3A4-Humanized Mice Clinical DDI Study With NVS123 and RTV An in vivo DDI study in which NVS123 was dosed alone at 50 mg/kg (equivalent to human dose of 300 mg based on body surface area, FDA Guidance for Industry and Reviewers, 2005), or codosed with the CYP3A4 inhibitor RTV at 35 mg/kg (equivalent to human dose of 200 mg based on body surface area, FDA Guidance for Industry and Reviewers, 2005), was conducted in hCYP3A4 mice (n ¼ 4 mice per dose group). In the codosed groups, 50-mg/kg NVS123 was either coadministrated simultaneously with
A clinical DDI study with NVS123 at 50-mg single dose and a 100-mg bid RTV dose was subsequently conducted in healthy Table 5 Plasma Exposure of NVS123 in Humanized Mouse With or Without Coadministration of RTV PK parameters NVS123 Only NVS123 SD þ RTV SD
NVS123 SD þ RTV MD
Fold Increased
Table 4 Estimation of fmCYP3A4 for NVS123 fmCYP3A4
Data
1.0 0.7 0.2 0.04
Metabolic profile in HLM Human oral CL assuming F ¼ 10% Metabolic profile hepatocyte Metabolic profile in rat excreta
Cmax (ng/mL) AUC (ng$h/mL) T1/2 (h)
495
4090
899
29,400
1.2
3.1
SD, single dose; MD, multiple dose.
8.3 32.7 2.6
Fold Increased 4980
10.1
50,600
56.3
2.2
1.8
1402
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404 Table 6 Simulated Boost Effect of 100-mg Ritonavir Pretreatment on 50-mg NVS123 in Human Using Refined PBPK Model
Figure 4. Pharmacokinetics of NVS123 in hCYP3A4 mice (n ¼ 4) with and without coadministration of RTV. The dashed line with open circle shows the PK profile of NVS123 alone at 50-mg/kg PO in hCYP3A4 mice; the solid line with open diamond shows the PK profile of NVS123 at 50 mg/kg after codosing a 35-mg/kg single RTV dose; the solid line with open triangles shows the PK profile of NVS123 at 50 mg/kg with a pretreatment of RTV at 35 mg/kg for 8 days. Data are presented as mean ± SD.
volunteers (n ¼ 4). The observed increase in plasma AUC of NVS123 when codosed with RTV in this study was high and was 42.5-fold (Figs. 6a and 6b). C12 h, Cmin, and Cmax were also increased by 156- and 9.3-fold, respectively (Table 7). The observed human exposure data were in good agreement with the prediction from the PBPK model using the high fmCYP3A4 value (0.99) estimated from the humanized mouse DDI boost study (Fig. 7). The boost effect (fold change in AUC and Cmax) observed in human was also in good agreement with that from the DDI study in hCYP3A4 mice.
Discussion The boosting of protease inhibitors (PIs) is a therapeutic strategy wherein a low RTV dose is given concurrently with another PI to enhance the PI's systemic exposure through the inhibition of cytochrome P450 enzymes, particularly CYP3A4. For many of the PIs, RTV boosting can result in improved drug levels that can increase efficacy, decrease pill burden, add flexibility to the dosing schedule, and remove fasting restrictions. Accurate estimation of
Figure 5. Predicted change of plasma PK profiles of 50-mg NVS123 in human with or without pretreatment with ritonavir at 100-mg bid starting at 12 h (12 h before dosing NVS123) using Simcyp PBPK modeling. The fmCYP3A4 value estimated from this in vivo DDI study in hCYP3A4 mice was high (0.99). The solid line shows the PK profile of NVS123 with RTV pretreatment, and the dashed line shows the PK profile of NVS123 without RTV pretreatment. 0 h is the time of starting RTV pretreatment.
Treatment
Cmax (ng/mL)
AUC (ng$h/mL)
NVS123 alone NVS123 þ RTV Fold increase
54.4 694 12.8
161 5593 34.6
the boost effect usually depends on preclinical data, which can now be facilitated by modeling and simulation approaches. The PBPK model is a mathematical approach for predicting the ADME of compounds in humans or animal species, by transcribing anatomic, physiological, physical, and chemical properties into mathematical algorithms.17 In addition to other applications, PBPK modeling can be used to anticipate the quantitative extent of PK-based DDIs,18 and there is an increasing demand from regulatory agencies to include such predictions in submission documents.19 This requirement is often satisfied by using commercially available PBPK-modeling softwares, such as Simcyp or GastroPlus™.20,21 Although these tools are continuously improved and often provide reliable outcomes, the quality of their predictions relies on an accurate appraisal of key parameters used for PBPK modeling. Accordingly, such predictions can be challenging if key parameters cannot be estimated with sufficient confidence. An important parameter for the estimation of the extent of such a DDI is the fraction of drug metabolized by CYP3A4 (fmCYP3A4) in humans, which is usually determined by various in vitro approaches such as HLM or hepatocyte data22,23 and preclinical in vivo studies, for example, rat ADME data.24 However, for NVS123, HLM, hepatocyte, and rat ADME data were not in good agreement and generated a wide range of fmCYP3A4 from 0.04 to 1. This range of fmCYP3A4 values resulted in a predicted increase in NVS123 AUC by coadministrated RTV of 1- to 45-fold, leading to a lack of confidence in this PBPK model to accurately predict the DDI or boost effect of RTV on NVS123 in human. Accordingly, an alternative approach had to be considered. We conducted an in vivo DDI study using a CYP3A4-humanized mouse model expressing human CYP3A4 in the liver and intestine to predict the clinical outcome and estimate human fmCYP3A4 for NVS123 and to use these data for further PBPK model refinement. The hCYP3A4 mice used in our study were generated by combining 2 transgenes with liver- and gut-specific CYP3A4 expression, respectively, on a mouse Cyp3a gene cluster knockout background.14 The impact of blocking CYP3A4 activity in the intestine on the plasma levels of a probe substrate was previously shown by a significant increase in docetaxel exposure in transgenic mice with gut-specific expression of CYP3A4 after treatment with RTV.14 A profound increase in triazolam exposure was reported in the humanized mice expressing CYP3A4 in both liver and intestine in response to the prototypical CYP3A4 inhibitor ketoconazole.25 Furthermore, ter Heine et al.26 used the humanized mice with liver- or gut-specific CYP3A4 expression to dissect the role of hepatic and intestinal CYP3A4 on lopinavir disposition by using RTV as an inhibitor. In spite of these promising results, data showing the utility of the hCYP3A4 mice to predict clinical DDIs are still very limited, and the prospective prediction of the extent of DDI in humans using these models has not yet been reported. To accurately predict the clinical DDI outcome, the human equivalent dose for both NVS123 and RTV based on body surface area was used in the in vivo DDI study with hCYP3A4 mice. Pretreatment with RTV at single dose and multiple doses for 8 days was assessed in terms of changes in NVS123 AUC and Cmax. The data showed that a single dose of RTV at 35 mg/kg can inhibit CYP3A4, reducing the metabolism of NVS123 in the liver and
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
1403
Figure 6. Observed pharmacokinetics of NVS123 in human healthy volunteers (n ¼ 4) with and without pretreatment of RTV. (a) Linear scale and (b) logarithmic scale. The solid line with open triangle shows the NVS123 PK at 50 mg without RTV pretreatment and the solid line with open circle shows the NVS123 PK at 50 mg with RTV pretreatment at 100-mg bid starting 12 h before NVS123 dosing (12 h). 0 h is the time of dosing NVS123. Data are presented as mean ± SD.
increasing the Cmax and AUC significantly by 8.3- and 32.7-fold, respectively. Multiple doses of RTV increased the Cmax and AUC of NVS123 to 10.1- and 56.3-fold, suggesting a stronger inhibition of CYP3A4. We speculate that the stronger inhibition of CYP3A4 with multiple doses of RTV is a likely consequence of higher RTV exposure, but we did not measure the RTV plasma profile in our study. RTV not only increased the plasma exposure of NVS123 but also prolonged the T1/2. Considering that NVS123 has minimal first-pass metabolism based on animal data, we believe the mechanism of the boost effect of RTV on NVS123 is predominantly related to the inhibition of CYP3A4 in the liver rather than the first-pass metabolism. Because plasma exposure of NVS123 could be boosted by 30- to 50-fold with low doses of RTV, we rationalized that a lower starting dose of NVS123 should be considered in the clinical trial. Another important outcome from the in vivo DDI study in the hCYP3A4 mice was the estimation of fmCYP3A4 for NVS123 in humans to be 1.0, meaning that NVS123 is metabolized only by CYP3A4. When this refined fmCYP3A4 value was used for PBPK modeling, the clinical boosting effect of RTV on NVS123 at different dosing combinations was accurately predicted (Fig. 7). The clinical team acknowledged the data from the mouse in vivo DDI study using hCYP3A4 as well as the updated PBPK model prediction and reduced the NVS123 starting dose in the clinical DDI study from 200 to 50 mg. This was considered a safe dose, as a single oral 1000-mg NVS123 dose had been tested in the clinic without issues (not shown). The clinical study at 50 mg resulted in a 9.3-fold increase in Cmax and 42.5-fold increase in AUC when NVS123 was given to healthy volunteers pretreated with RTV at 100-mg bid and thus confirmed the outcome from the humanized
Table 7 Human Plasma Exposure Parameters of NVS123 at 50-mg Single Dose With or Without Pretreatment With RTV at 100-mg Bid PK parameters
NVS123 Only
NVS123 SD þ RTV SD
Fold Increased
C12h (ng/mL) Cmax (ng/mL) AUC (ng$h/mL) T1/2 (h)
0.828 47.8 128 1
129 444 5438 6
156 9.3 42.5 6.0
mouse DDI study. In addition, C12 h of NVS123 was also increased 156 fold when pretreated with RTV at 100-mg bid. It has been reported that the trough drug concentration for HCV PIs may be important in minimizing or preventing the emergence of resistance, and trough drug concentration had the strongest relationship with maximum change in HCV RNA.27,28 Because RTV coadministration increases the minimum plasma concentration (Cmin) more pronounced than the Cmax and AUC of NVS123, we speculate that lower dose of NVS123 may achieve efficacious trough concentration while overall NVS123 exposure can be reduced. In summary, we demonstrated that the combination of an in vivo DDI study with NVS123 and the potent CYP3A4 inhibitor RTV in hCYP3A4 mice along with PBPK modeling allowed for a prospective prediction of the extent of DDI and boost effect in humans, which was subsequently confirmed in a clinical study.
Figure 7. Comparison of predicted and observed human NVS123 PK profiles at 50-mg single dose with or without boost by RTV at 100-mg bid from 12 h (12 h before NVS123 was administered). The open triangle shows the observed NVS123 PK alone at 50-mg single dose (data are presented as mean ± SD), and the open circle shows the observed NVS123 PK at 50 mg with RTV pretreatment at 100 mg from 12 h (both data are presented as mean ± SD). The dashed line represents the simulated NVS123 PK alone at 50-mg single dose, and the solid line represents the simulated NVS123 PK at 50 mg with RTV pretreatment at 100 mg from 12 h.
1404
J. Zhang et al. / Journal of Pharmaceutical Sciences 105 (2016) 1398e1404
Acknowledgments The authors thank Dr. Binfeng Xia for helpful discussions and for sharing his expertise in PBPK modeling. Tycho Heimbach, Handan He, Avantika Barve, Wen Lin, and Jin Zhang participated in research design. Jin Zhang conducted experiments. Tycho Heimbach, Wenkui Li, and Jin Zhang performed data analysis. Jin Zhang, Nico Scheer, Tycho Heimbach, Handan He, and Avantika Barve wrote or contributed to the writing of the article.
References 1. Kempf DJ, Klein C, Chen HJ, et al. Pharmacokinetic enhancement of the hepatitis C virus protease inhibitors VX-950 and SCH 503034 by co-dosing with ritonavir. Antivir Chem Chemother. 2007;18(3):163-167. 2. King JR, Wynn H, Brundage R, Acosta EP. Pharmacokinetic enhancement of protease inhibitor therapy. Clin Pharm. 2004;43(5):291-310. 3. Reddy MB, Chen Y, Haznedar JO, et al. Impact of low-dose ritonavir on danoprevir pharmacokinetics: results of computer-based simulations and a clinical drug-drug interaction study. Clin Pharm. 2012;51(7):457-465. 4. Scheer N, Wolf CR. Genetically humanized mouse models of drug metabolizing enzymes and transporters and their applications. Xenobiotica. 2014;44(2):96-108. 5. Shen HW, Jiang XL, Gonzalez FJ, Yu AM. Humanized transgenic mouse models for drug metabolism and pharmacokinetic research. Curr Drug Metab. 2011;12(10):997-1006. 6. Foster JR, Lund G, Sapelnikova S, Tyrrell DL, Kneteman NM. Chimeric rodents with humanized liver: bridging the preclinical/clinical trial gap in ADME/ toxicity studies. Xenobiotica. 2014;44(2):109-122. 7. Grompe M, Strom S. Mice with human livers. Gastroenterology. 2013;145(6): 1209-1214. 8. Peltz G. Can 'humanized' mice improve drug development in the 21st century? Trends Pharmacol Sci. 2013;34(5):255-260. 9. Hasegawa M, Kapelyukh Y, Tahara H, et al. Quantitative prediction of human pregnane X receptor and cytochrome P450 3A4 mediated drug-drug interaction in a novel multiple humanized mouse line. Mol Pharmacol. 2011;80(3): 518-528. 10. Hasegawa M, Tahara H, Inoue R, Kakuni M, Tateno C, Ushiki J. Investigation of drug-drug interactions caused by human pregnane X receptor-mediated induction of CYP3A4 and CYP2C subfamilies in chimeric mice with a humanized liver. Drug Metab Dispos. 2012;40(3):474-480. 11. Yang J, Jamei M, Yeo KR, et al. Prediction of intestinal first-pass drug metabolism. Current Drug Metabolism. 2007;8(7):676-684. 12. Mao J, Mohutsky MA, Harrelson JP, Wrighton SA, Hall SD. Prediction of CYP3A4-mediated drug-drug interactions using human hepatocytes suspended in human plasma. Drug Metab Dispos. 2011;39(4):591-602.
13. Foisy MM, Yakiwchuk EM, Hughes CA. Induction effects of ritonavir: implications of drug interactions. Ann Pharmacother. 2008;42(7):1048-1059. 14. van Herwaarden AE, Wagenaar E, van der Kruijssen CM, et al. Knockout of cytochrome P450 3A yields new mouse models for understanding xenobiotic metabolism. J Clin Invest. 2007;117(11):3583-3592. 15. Federal Drug Administration. Guidance for Industry and Reviewers: Estimating the safe starting dose in clinical trials for therapeutics in adult healthy volunteers; 2005. Available at: http://www.fda.gov/OHRMS/DOCKETS/98fr/02d-0492gdl0001-vol1.pdf. Accessed January 17, 2016. 16. Li W, Tse FL. Dried blood spot sampling in combination with LC-MS/MS for quantitative analysis of small molecules. Biomed Chromatogr. 2010;24(1): 49-65. 17. Espie P, Tytgat D, Sargentini-Maier ML, Poggesi I, Watelet JB. Physiologically based pharmacokinetics (PBPK). Drug Metab Rev. 2009;41(3):391-407. 18. Jones HM, Chen Y, Gibson C, et al. Physiological based pharmacokinetic modeling in drug discovery and development: a pharmaceutical industry perspective. Clin Pharm & Ther. 2015;97(3):247-262. 19. Prueksaritanont T, Chu X, Gibson C, et al. Drug-drug interaction studies: regulatory guidance and an industry perspective. AAPS J. 2013;15(3):629-645. 20. Heimbach T, Lakshminarayana SB, Hu W, He H. Practical anticipation of human efficacious doses and pharmacokinetics using in vitro and preclinical in vivo data. AAPS J. 2009;11(3):602-614. 21. Jamei M, Marciniak S, Feng K, Barnett A, Tucker G, Rostami-Hodjegan A. The Simcyp population-based ADME simulator. Expert Opin Drug Metab Toxicol. 2009;5(2):211-223. 22. Desbans C, Hilgendorf C, Lutz M, et al. Prediction of fraction metabolized via CYP3A in humans utilizing cryopreserved human hepatocytes from a set of 12 single donors. Xenobiotica. 2014;44(1):17-27. 23. Youdim KA, Zayed A, Dickins M, et al. Application of CYP3A4 in vitro data to predict clinical drug-drug interactions; predictions of compounds as objects of interaction. Br J Clin Pharmacol. 2008;65(5):680-692. 24. Chen Y, Jin JY, Mukadam S, Malhi V, Kenny JR. Application of IVIVE and PBPK modeling in prospective prediction of clinical pharmacokinetics: strategy and approach during the drug discovery phase with four case studies. Biopharm Drug Dispos. 2012;33(2):85-98. 25. van Waterschoot RA, Rooswinkel RW, Sparidans RW, van Herwaarden AE, Beijnen JH, Schinkel AH. Inhibition and stimulation of intestinal and hepatic CYP3A activity: studies in humanized CYP3A4 transgenic mice using triazolam. Drug Metab Dispos. 2009;37(12):2305-2313. 26. ter Heine R, Van Waterschoot RA, Keizer RJ, Beijnen JH, Schinkel AH, Huitema AD. An integrated pharmacokinetic model for the influence of CYP3A4 expression on the in vivo disposition of lopinavir and its modulation by ritonavir. J Pharm Sci. 2011;100(6):2508-2515. 27. Reesink HW, Zeuzem S, Weegink CJ, et al. Rapid decline of viral RNA in hepatitis C patients treated with VX-950: a phase Ib, placebo-controlled, randomized study. Gastroenterology. 2006;131(4):997-1002. 28. Rubino C, Bradford WZ, Forrest A, et al. Pharmacokinetic-pharmacodynamic (PK-PD) relationships for ITMN-191 in a phase 1 multiple ascending dose trial in patients with genotype 1 chronic hepatitis C (CHC) infection. Hepatology. 2008;48(Suppl.2):1140A-1141A.