Reliability of in vitro data for the mechanistic prediction of brain extracellular fluid pharmacokinetics of P-glycoprotein substrates in vivo; are we scaling correctly?

D.W. van Valkengoed [1], V. Rottschäfer [2,3], E.C.M de Lange [1]

[1]: Division of Systems Pharmacology and Pharmacy, Leiden Academic Center for Drug Research, Leiden University, Leiden, The Netherlands [2]: Mathematical Institute, Leiden University, Leiden, The Netherlands [3]: Korteweg-de Vries Institute for Mathematics, University of Amsterdam, P.O. Box 94248, 1090 GE Amsterdam, The Netherlands

Introduction: Drug pharmacokinetic (PK) profiles observed in the blood stream are not always informative about the PK within the central nervous system (CNS) due to the presence of active efflux transporters like P-glycoprotein (P-gp/MDR1/mdr1a) at the blood-brain-barrier (BBB) [1]. However, knowing CNS PK is crucial for the development and optimization of CNS-based therapies. Physiologically based pharmacokinetic (PBPK) models are mechanistically informed mathematical models that predict drug disposition, and as such, these models are perfect candidates for improving our understanding of CNS PK. Ideally, CNS PBPK models would make use of in vitro derived data to predict CNS PK, thereby removing the need for animal experimentation during drug development. In vitro transwell transport assays on P-gp expressing cell monolayers allow the calculation of apparent permeability (P­_app) and corrected efflux ratios (ER_c) of P-gp substrates. These parameters contain mechanistic information on P-gp mediated efflux clearance (CL_pgp), and therefore show promise for use in PBPK models for the prediction of brain extracellular fluid (ECF) PK of P-gp substrates.

Objectives: This study investigated whether P_app values derived from literature could be repurposed as input in a CNS-PBPK model to confidently predict rat brainECF PK of the four P-gp substrates morphine, quinidine, risperidone and paliperidone. An important aspect to in vitro data is its variable nature [2,3], which might impact its use for mechanistic modelling. Therefore, we also aimed to establish how variability in transport data for the same drug from different sources would influence prediction outcomes and model robustness.

Methods: Transport data and P-gp protein expression in cell lines and rats were extracted from literature to calculate P-gp mediated efflux CL (CL_pgp). CL­_pgp was scaled from in vitro to in vivo through a relative expression factor (REF), which accounts for the difference in P-gp expression between in vivo and in vitro. CL_pgp was determined using the method outlined by Kalvass et al. [4], and introduced into the existing LeiCNS CNS-PBPK model [5].  Passive diffusion across the BBB was described with P_app data retrieved from assays without functional P-gp in the cell lines. Predictions were validated using brainECF microdialysis data after constant infusions (steady state, SS) and short (non-SS) infusions of the P-gp substrates. Predictions were deemed accurate when the median prediction error (PE) fell within 2-fold of the observed data.

Results: 26 transport values were extracted from literature from 4 cell lines of non-BBB origin, namely Caco-2, LLC-PK1-mdr1a/MDR1, and MDCKII-MDR1. P-gp protein expression in rats was consistent in literature (19.4 fmol/µg protein), whereas in vitro expression varied and was split into three categories (high, average, low) for each cell line. This allowed prediction intervals to be generated, which were used to investigate the impact of uncertainty in in vitro P-gp expression on the prediction accuracy. Brain distribution of three passively diffusing drugs was predicted well using in vitro data. Moreover, the model was able to accurately predict the P-gp substrates’ brainECF PK for all SS dosing regimens and the morphine and risperidone non-SS dosing schemes. However, certain combinations of P_app and in vitro P-gp expression showed superior predictive performance compared to others. Additionally, the variability in reported in vitro P-gp expression impacted prediction accuracy and model robustness when studies did not provide P_app values together with in vitro P-gp expression values. However, even when P_app and in vitro P-gp expression were reported in conjunction, this did not guarantee a more accurate prediction, even giving worse predictions compared to other in vitro expression values. Subsequent removal of the REF improved predictions for some, but not all drugs.

Discussion and conclusions: The developed CNS-PBPK model shows promise in predicting brainECF PK using in vitro data of P_app, ER_c and P-gp expression as input, but the in vitro to in vivo translation approach is not yet robust, as it seems that not only P-gp expression and P_app play a role. It can be concluded that more needs to be learned about context and drug dependency of in vitro study data for confident and robust predictions of brainECF PK.

References:

1. Schinkel, A.H., P-Glycoprotein, a gatekeeper in the blood-brain barrier. Adv Drug Deliv Rev, 1999. 36(2-3): p. 179-194.
2. Hirsch, C. and S. Schildknecht, In Vitro Research Reproducibility: Keeping Up High Standards. Front Pharmacol, 2019. 10: p. 1484.
3. Volpe, D.A., Variability in Caco-2 and MDCK cell-based intestinal permeability assays. J Pharm Sci, 2008. 97(2): p. 712-25.
4. Kalvass, J.C. and G.M. Pollack, Kinetic considerations for the quantitative assessment of efflux activity and inhibition: implications for understanding and predicting the effects of efflux inhibition. Pharm Res, 2007. 24(2): p. 265-76.
5. Saleh, M.A.A., et al., Lumbar cerebrospinal fluid-to-brain extracellular fluid surrogacy is context-specific: insights from LeiCNS-PK3.0 simulations. J Pharmacokinet Pharmacodyn, 2021. 48(5): p. 725-741.

Reference: PAGE 32 (2024) Abstr 10901 [www.page-meeting.org/?abstract=10901]

Poster: Drug/Disease Modelling - Absorption & PBPK

PDF poster / presentation (click to open)