Modeling the Pathogenesis of Drug-Induced Liver Injury with Organoids and Microfluidic Devices

Abstract

Drug-induced liver injury (DILI) is a critical concern in modern medicine, encompassing a broad spectrum of liver-related abnormalities caused by various pharmaceutical compounds. DILI is classified as either intrinsic, occurring in a dose-dependent and predictable manner, or idiosyncratic, occurring seemingly spontaneously. Ranging from mild liver enzyme elevations to acute liver failure, DILI is an unpredictable and multifaceted problem that poses significant risks and challenges in both drug discovery and patient experiences. DILI is currently a leading cause of both clinical trial failures and withdrawals post-FDA approval. This is due to a lack of human models that robustly predict hepatotoxicity. Although the current standard, primary human hepatocytes assays are costly, subject to batch variations, and quickly lose hepatocyte function. Drug responses in animal models are not predictive for humans. This results in clinical trial failures where patients suffer severe hepatotoxicity. Even post-approval, market withdrawals and black box warnings for drug safety are often caused by idiosyncratic DILI, where a hepatotoxicity event may only occur in 1:10,000 or less patients given the prescription drug and therefore not adequately sampled through the course of an entire clinical trial. There have been a multitude of advancements in human-derived liver models in recent years mostly in terms of 3D spheroid/organoid culture. These have been developed either from culturing a mixture of cell types in a confined space to force cell contact or differentiated from induced-pluripotent stem cells (iPSCs). For this dissertation, I adapted a previous protocol of generating iPSC-derived human liver organoids (HLOs) to a high-throughput screening platform and microfluidic chips. The iPSC-liver model was chosen as it potentiates development of a genetically diverse patient-derived biobank or to study rare idiosyncratic DILI. Human liver organoids (HLOs) contain not only the parenchymal hepatocyte-like cells but also non-parenchymal stellate-like, Kupffer-like, and cholangiocyte-like cells. HLOs dispersed into 384-well plates retained liver function as measured by marker expression, albumin production, and CYP450 activity and showed viability for high-throughput pre-clinical screening for drug safety. Morphological profiling of treated inform on DILI mechanism. Dispersed HLOs transferred onto microfluidic chips with media flow show remarkable improvement for modelling liver with notable increases in albumin expression and CYP activity. These liver chips reliably model DILI based on cytotoxicity and morphological perturbations (lipid accumulation and mitochondrial impairment) at physiologic drug concentrations and captured a recent case of synergistic DILI not discovered until clinical trials. Single-cell transcriptomics of liver chips predicted an additional case of synergistic DILI that was confirmed in a high-throughput assay. A 16-well microfluidic liver chip was also developed to balance throughput and physiological relevance. This platform shared all the improvements seen with the original microfluidic chip but with an added benefit of culture longevity. 16-well liver chips were shown to maintain cell viability for up to 28-days and serve as a long-term liver model. Finally, we isolated PBMCs from DILI patient whole blood to reprogram into iPSCs. Reprogrammed iPSCs were differentiated into HLOs. 4 patient-derived HLOs were used to screen a panel of 64 hepatotoxins. While neither HLO line responded to all 64 compounds, screening through all 4 lines captured hepatotoxicity of the drug panel with high accuracy, emphasizing the need for genetic diversity and redundant screening. Future research will attempt to capture the original patient’s DILI events through added model complexity and inclusion of other cell types.