Single cell transcriptomic landscapes of human liver organoids stratify models of non-alcoholic fatty liver disease

Non-alcoholic fatty liver disease (NAFLD) is a rapidly growing cause of morbidity with few treatment options available. Thus, accurate in vitro systems to test new therapies are indispensable. Recently, human liver organoid (HLO) NAFLD models have emerged. However, a systematic evaluation of their translational potential is currently missing. Here, we develop a structured approach to evaluate NAFLD-HLO models, testing oleic acid (OA) and palmitic acid (PA) in comparison to TGF-{beta}1 for disease induction. Through analysis of [~]100K single-cell transcriptomes of the HLO injury landscape, we find all three models induce inflammatory signatures. However, only TGF-{beta}1 promotes collagen production, fibrosis, and hepatic stellate cell (HSC) expansion. In striking contrast, OA ameliorates fibrotic signatures and reduces the HSC population. Integrating data from each model with that of NAFLD patients across disease progression further demonstrates PA and TGF-{beta}1 more robustly model inflammation and fibrosis. Our findings highlight the importance to stratify NAFLD-HLO models by clinical disease progression, provide a single-cell reference to benchmark future organoid injury models, and allow us to study evolving steatohepatitis, fibrosis, and HSC susceptibility to injury in a dynamic, multi-lineage human in vitro system.