Prediction of Multidrug Resistance Protein 3 (MDR3) Inhibition-mediated Cholestatic Drug-induced Liver Injury (DILI) Using Quantitative Systems Toxicology (QST) Modeling

Conference: SOT
Software: DILIsym®
Division: DILIsym Services

Background & Purpose

  • DILI is a primary cause of acute liver failure and reason for the termination of drug development programs
  • To successfully predict and prevent DILI events, it is critical to understand the various types of underlying DILI mechanisms
  • Inhibition of hepatic efflux transporters is a well-recognized mechanism that can lead to DILI (e.g., bile salt export pump (BSEP) inhibition-mediated accumulation of toxic bile acids (BAs) in hepatocytes)
  • MDR3 inhibition is a key mechanism that can manifest into cholestatic DILI, clinically defined by alkaline phosphatase (ALP) >2x upper limit of normal (ULN) in combination with a major elevation of γ-glutamyl-transferase (GGT) and alanine aminotransferase (ALT)/ALP (fold ULN) <2, and characterized by cholangiocellular injury
  • MDR3 is a phospholipid (PL) floppase that translocates PLs to the apical side of the canalicular membrane where PLs can form mixed micelles with biliary BAs, thereby reducing BA monomer-induced injury to cholangiocytes
  • This important hepatic function can be compromised by compounds that inhibit MDR3 activity, and could result in the development of clinically defined cholestatic liver injury
  • A computational QST model for this phenomenon in humans has recently been developed
  • In the current work, MDR3 inhibitors with and without cholestatic DILI liability were used to validate this novel QST model of cholestatic DILI


  • DILIsym® (version 8A), a commercially available QST model of DILI, was extended to mechanistically represent MDR3 inhibition-mediated cholestatic DILI
  • This model consists of previously developed representations of BA homeostasis, mitochondrial function, oxidative stress, innate immunity, among other submodels important to liver health and injury, that are solved computationally in the DILIsym software
  • To predict MDR3 inhibition-mediated cholestatic DILI, new relevant features were mathematically represented in DILIsym (Fig. 1)
  • A variety of publicly available clinical data with and without drug effects was used to calibrate and validate the updated model and to construct a new virtual population (SimPops®) of healthy volunteers (n=285) representing variability in both BA toxicity and cholestasis mechanisms
  • Physiologically based pharmacokinetic (PBPK) models of four selected MDR3 inhibitors were developed in GastroPlus® (version 9.8.2) to inform the hepatocellular exposure of these drugs (Fig. 2)
  • Dosing protocol-specific exposure predictions along with in vitro MDR3 and BSEP inhibition potential data (e.g., half-maximal inhibitory concentration, IC50) were implemented in the extended DILIsym model to evaluate cholestatic DILI predictions for each of the MDR3 inhibitors (Fig. 3)


  • The novel cholestatic DILI representation in DILIsym predicted hepatotoxicity for the two DILI-associated MDR3 inhibitors itraconazole and verapamil, while no hepatotoxicity signals were predicted for the two clean MDR3 inhibitors loratadine and chlorpheniramine
  • Hepatic exposure, MDR3 inhibition potential and MDR3 mode of inhibition were important drivers of the predicted cholestatic liver injury
  • This work shows that QST modeling is a promising approach to reasonably predict clinically defined cholestatic DILI liability in humans

By James J. Beaudoin, Jeffry Adiwidjaja, Kyunghee Yang, and Jeffrey L. Woodhead

2024 SOT Annual Meeting and ToxExpo, March 10–14, 2024, Salt Lake City, Utah