PAGE. Abstracts of the Annual Meeting of the Population Approach Group in Europe.
PAGE 28 (2019) Abstr 9050 [www.page-meeting.org/?abstract=9050]
Poster: Drug/Disease modelling - Absorption & PBPK
Vanessa Baier (1,2), Lars M. Blank (1), Florian Caiment (3), Olivia Clayton (4), Henrik Cordes (1), David A. Fluri (5), Hans Gmuender (6), Jens M. Kelm (5), Ramona Nudischer (4), Adrian Roth (4), Ralph Schlapbach (6), Nathalie Selevsek (6), Christoph Thiel (1), José V. Castell (7), Jos Kleinjans (3), Lars Kuepfer (1), and members of the HeCaToS Consortium
(1) Institute of Applied Microbiology - iAMB, Aachen Biology and Biotechnology – ABBt, RWTH Aachen University, Germany (2) esqLABS GmbH, Saterland, Germany (3) Department of Toxicogenomics, Maastricht University, Maastricht, Netherlands (4) Roche Pharmaceutical Research and Early Development, Roche Innovation Center Basel, Basel, Switzerland (5) InSphero AG, Schlieren, Switzerland (6) Functional Genomics Center Zurich, ETH Zurich and University of Zurich, Zurich, Switzerland (7) Instituto de Investigación Sanitaria. Hospital Universitario La Fe, Valencia, Spain
Objectives: Cholestasis is a major clinical manifestation of drug induced liver injuries (DILI) characterized by an impaired bile flow. In consequence, bile acids (BAs) accumulate in the liver and other tissues ultimately leading to clinically diagnosable symptoms. BA metabolism is a systemic process that involves the interplay of multiple tissues along enterohepatic circulation including the gastrointestinal tract and the liver. Due to the many processes involved in BA distribution a mechanistic understanding of the events underlying drug-induced cholestasis is challenging to achieve . Computational modelling bears the chance to significantly contribute to a comprehensive description of BA metabolism and its interplay with drugs. For this purpose, the ability of these models to compile existing knowledge in a mathematical representation, to simulate tissue concentration profiles which are experimentally inaccessible, and to formulate hypotheses to be specifically addressed in further experimental studies are of great use. We here present the application of a previously developed physiology-based bile acid (PBBA) model  for the quantification of the cholestatic risk of different known hepatotoxicants.
Methods: The PBBA model was based on well-established concepts from physiology-based pharmacokinetic (PBPK) modelling and built with the Open Systems Pharmacology Suite . It was initially informed by clinical plasma BA measurements and subsequently used for the prediction of changes in body BA levels after drug administration in healthy individuals as well as in patients with a high-risk genotype . In the presented approach, adaptation of the liver in response to repeated drug administration was investigated in a PBPK-informed in vitro assay setup with liver spheroids for 10 known hepatotoxicants . For each compound, a simulated drug PK profile was used to determine in vivo drug exposures at liver tissue. These drug profiles were then transferred to in vitro assay concentrations in an assay with human liver spheroids to reproduce physiologically relevant drug levels for up to two weeks. It was thus possible to track changes in the expression of drug-ADME genes and to include this in the PBBA model.
Results: Transcriptomics were generated from liver spheroids reflecting the adaptation of liver tissue in the face of up to two weeks of drug exposure. The measured mRNA fold changes of CYP7A1, bile salt export pump (BSEP), and NTCP were dynamically integrated into the PBBA model to simulate the changes of BA levels in various tissues after drug administration. The drugs were then ranked according to the resulting average BA levels in different body compartments. In a complementary approach, plasma BA levels were measured from patients hospitalised after a DILI event and ranked according to their cholestatic risk. Using these clinical data for the overlapping set of drugs as a benchmark we identified the venous blood compartment in the PBBA model as the one showing the highest agreement with patient measurements. This correlation was higher than for the case of BSEP fold changes, which is frequently used as an experimental marker for cholestasis. We thus obtained a full ranking of all 10 investigated drugs with respect to their cholestatic potential. In agreement with a previous study , the results suggest that plasma BA levels might not be representative for tissue BA levels where changes can be more severe than in the plasma.
Conclusions: Our results show that the contextualisation of specifically-generated in vitro data in the model provides mechanistic insights which would otherwise not have been accessible. PBPK modelling is particularly suited for such detailed analyses since it allows the integration of heterogenous in vitro and other preclinical data. The prospective design of an in vitro assay with physiologically relevant tissue concentrations simulated with drug-specific PBPK models significantly supported these analyses. It could be shown that the PBBA model with the contextualised assay data reveals better insight than for example simple in vitro transport assays of BSEP. Also, our model based-analyses enhance the mechanistic understanding of the occurrence of drug-induced cholestasis. Our approach could therefore in the future provide a platform for advanced preclinical testing in pharmaceutical development programs with lower incidence rates of cholestatic DILI as such enhancing patient safety and in consequence success rates in pharmaceutical development.