Ayatallah Saleh (1,2), Robin Michelet (1), Gerd Mikus (1,3), Wilhelm Huisinga (4), and Charlotte Kloft (1)
(1) Department of Clinical Pharmacy and Biochemistry, Institute of Pharmacy, Freie Universität Berlin, Germany, (2) Graduate Research Training program PharMetrX, Germany, (3) Department of Clinical Pharmacology and Pharmacoepidemiology University Hospital Heidelberg, Heidelberg, Germany, (4) Institute of Mathematics, University of Potsdam, Germany
Introduction: PBPK modelling is a powerful tool to quantitatively explore and predict PK of drugs and the magnitude of DDIs. Midazolam (MDZ), a short-acting sedative exclusively metabolised by CYP3A4 to its major active metabolite 1-hydroxy-MDZ (1-OH-MDZ) [1,2], is often used to investigate CYP3A DDI. 1-OH-MDZ’s involvement in DDI has also been suggested [3,4,5].
The utility of PBPK modelling to predict the PK of a parent drug is well-established, yet little attention has been paid for metabolites. Many kinetic and physiologic factors can determine the extent of metabolite formation and its circulation in human plasma, e.g. lysosomal trapping can occur within the tissue of formation (i.e. liver), especially for lipophilic (logP>1) and amphiphilic metabolites with ionizable amines (pKa>6) [6]. As 1-OH-MDZ is pharmacologically active, the need to adequately describe its PK arises and PBPK models form an excellent framework. The objective of this study was to evaluate and expand a whole-body PBPK model of MDZ to include 1-OH-MDZ after intravenous (i.v.) administration in healthy volunteers.
Methods: Based on an established MDZ PBPK model [7], a coupled whole-body PBPK model for MDZ and 1-OH-MDZ was built in PK-Sim®(v. 9.1.0) [8]. The model was developed using individual data from a clinical study as training dataset [9], with 1 μg MDZ i.v. (5 min), followed by a ≥72 h washout period and then 1 mg MDZ i.v. (5 min). Second dataset of 24 healthy individuals receiving 3 µg MDZ as i.v. bolus served as validation dataset.
To build 1-OH-MDZ PBPK model, an extensive literature search was performed on its (a) physicochemical properties and (b) distribution, metabolism and excretion processes. Parameters that could not be informed from literature were optimised using the training dataset and model misspecification was investigated by sensitivity analysis to generate hypotheses regarding missing processes.
The final model was evaluated by comparing predicted to observed plasma concentration-time profiles, area under the concentration-time curve (AUC) values and maximum plasma concentrations of the validation dataset.
Results: The simulated plasma concentration-time profiles for MDZ after i.v. administration of 3 different dosing regimens were in close concordance with the observed data (fold error for predicted PK parameters: 0.88 – 1.28).
For the metabolite, initial model building showed good predictions of 1-OH-MDZ exposure throughout the entire sampling interval, except for an initial steep peak in the simulation. Sensitivity analysis showed that the predicted AUC of 1-OH-MDZ, was sensitive to total hepatic clearance, lipophilicity (logP) and fraction unbound in plasma (fup). Thus, these parameters were inferred from in vivo data (logP: 2.29, fup: 11.8%, total hepatic clearance: 10.2 mL/min/kg [10]), leading to better model performance except for the first 30 minutes. An unconstrained parameter estimation for logP or the cellular permeability parameters between the hepatic interstitial and intracellular space led to an unrealistic reduction of ~98% of all three values compared to [10]. Combining this result with the fact that 1-OH-MDZ is an amphiphilic compound with a basic moiety ionisable at acidic pH led to the hypothesis of lysosomal trapping. While being able to enter the lysosome via passive diffusion, the acidic environment within the lysosome (pH 4.5–5) causes a protonation of basic groups [11]. This significantly reduces the permeability across the lipid bilayer and results in trapping of the cationic form leading to slow release from the hepatocyte to the systemic circulation. This process was included in the model by adding a hypothetical intracellular binding partner for 1-OH-MDZ within the hepatocytes. The corresponding dissociation constant Kd was fixed to 2.20 nM [12], while the concentration of that hypothetical protein was optimised to match best our training dataset resulting in 0.520 μM. Subsequent model evaluation showed close concordance with the observed data for 1-OH-MDZ (fold error for predicted PK parameters: 0.75 – 0.92).
Conclusion: The coupled whole-body PBPK model of MDZ and its main active metabolite 1-OH-MDZ, including the lysosomal trapping effect for the metabolite, was able to predict the MDZ and 1-OH-MDZ plasma profiles in healthy individuals. The hypothesis of lysosomal trapping should be further investigated in vitro or in vivo, as it can be highly relevant for other compounds as well.
References:
[1] M. Balk, H. Hentschke, U. Rudolph, B. Antkowiak, B. Drexler. Differential depression of neuronal network activity by midazolam and its main metabolite 1-hydroxymidazolam in cultured neocortical slices. Sci. Rep. 7: 3503 (2017).
[2] K.C. Patki, L.L. Von Moltke, D.J. Greenblatt. In vitro metabolism of midazolam, triazolam, nifedipine, and testosterone by human liver microsomes and recombinant cytochromes p450: role of cyp3a4 and cyp3a5. Drug Metab. Dispos. 31: 938–44 (2003).
[3] S.T. Wiebe, A.D. Meid, G. Mikus. Composite midazolam and 1’-OH midazolam population pharmacokinetic model for constitutive, inhibited and induced CYP3A activity. J. Pharmacokinet. Pharmacodyn. 47: 527–542 (2020).
[4] X. Li, L. Junge, M. Taubert, A. von Georg, D. Dahlinger, C. Starke, S. Frechen, C. Stelzer, M. Kinzig, F. Sörgel, U. Jaehde, U. Töx, T. Goeser, U. Fuhr. A Novel Study Design Using Continuous Intravenous and Intraduodenal Infusions of Midazolam and Voriconazole for Mechanistic Quantitative Assessment of Hepatic and Intestinal CYP3A Inhibition. J. Clin. Pharmacol. 60: 1237–1253 (2020).
[5] X. Li, S. Frechen, D. Moj, T. Lehr, M. Taubert, C.-H. Hsin, G. Mikus, P.J. Neuvonen, K.T. Olkkola, T.I. Saari, U. Fuhr. A Physiologically Based Pharmacokinetic Model of Voriconazole Integrating Time-Dependent Inhibition of CYP3A4, Genetic Polymorphisms of CYP2C19 and Predictions of Drug-Drug Interactions. Clin. Pharmacokinet. 59: 781–808 (2020).
[6] M. V. Schmitt, P. Lienau, G. Fricker, A. Reichel. Quantitation of Lysosomal Trapping of Basic Lipophilic Compounds Using In Vitro Assays and In Silico Predictions Based on the Determination of the Full pH Profile of the Endo-/Lysosomal System in Rat Hepatocytes. Drug Metab. Dispos. 47: 49–57 (2019).
[7] N. Hanke, S. Frechen, D. Moj, H. Britz, T. Eissing, T. Wendl, T. Lehr. PBPK Models for CYP3A4 and P-gp DDI Prediction: A Modeling Network of Rifampicin, Itraconazole, Clarithromycin, Midazolam, Alfentanil, and Digoxin. CPT pharmacometrics Syst. Pharmacol. 7: 647–659 (2018).
[8] T. Eissing, L. Kuepfer, C. Becker, M. Block, K. Coboeken, T. Gaub, L. Goerlitz, J. Jaeger, R. Loosen, B. Ludewig, M. Meyer, C. Niederalt, M. Sevestre, H.-U. Siegmund, J. Solodenko, K. Thelen, U. Telle, W. Weiss, T. Wendl, S. Willmann, J. Lippert. A computational systems biology software platform for multiscale modeling and simulation: integrating whole-body physiology, disease biology, and molecular reaction networks. Front. Physiol. 2: 4 (2011).
[9] N. Hohmann, F. Kocheise, A. Carls, J. Burhenne, W.E. Haefeli, G. Mikus. Midazolam microdose to determine systemic and pre-systemic metabolic CYP3A activity in humans. Br. J. Clin. Pharmacol. 79: 278–285 (2015).
[10] H.Q. Nguyen, E. Kimoto, E. Callegari, R.S. Obach. Mechanistic Modeling to Predict Midazolam Metabolite Exposure from In Vitro Data. Drug Metab. Dispos. 44: 781–91 (2016).
[11] M. V. Schmitt, P. Lienau, G. Fricker, A. Reichel. Quantitation of lysosomal trapping of basic lipophilic compounds using in vitro assays and in silico predictions based on the determination of the full pH profile of the endo-/lysosomal system in rat hepatocytes. Drug Metab. Dispos. 47: 49–57 (2019).
[12] A. Buhr, R. Baur, E. Sigel. Subtle changes in residue 77 of the gamma subunit of alpha1beta2gamma2 GABAA receptors drastically alter the affinity for ligands of the benzodiazepine binding site. J. Biol. Chem. 272: 11799–804 (1997).
Reference: PAGE 29 (2021) Abstr 9842 [www.page-meeting.org/?abstract=9842]
Poster: Drug/Disease Modelling - Absorption & PBPK