A human liver organoid screening platform for DILI risk prediction

Abstract

Background & aims: Drug-induced liver injury (DILI), both intrinsic and idiosyncratic, causes frequent morbidity, mortality, clinical trial failures and post-approval withdrawal. This suggests an unmet need for improved in vitro models for DILI risk prediction that can account for diverse host genetics and other clinical factors. In this study, we evaluated the utility of human liver organoids (HLOs) for high-throughput DILI risk prediction and in an organ-on-chip system.

Methods: HLOs were derived from three separate iPSC lines and benchmarked on two platforms for their ability to model in vitro liver function and identify hepatotoxic compounds using biochemical assays for albumin, ALT, AST, microscopy-based morphological profiling, and single-cell transcriptomics: i) HLOs dispersed in 384-well-formatted plates and exposed to a library of compounds; ii) HLOs adapted to a liver-on-chip system.

Results: Dispersed HLOs derived from the three iPSC lines had similar DILI predictive capacity as intact HLOs in a high-throughput screening format, allowing for measurable IC₅₀ values of compound cytotoxicity. Distinct morphological differences were observed in cells treated with drugs exerting differing mechanisms of toxicity. On-chip HLOs significantly increased albumin production, CYP450 expression, and ALT/AST release when treated with known hepatoxic drugs compared to dispersed HLOs and primary human hepatocytes. On-chip HLOs were able to predict the synergistic hepatotoxicity of tenofovir-inarigivir and displayed steatosis and mitochondrial perturbation, via phenotypic and transcriptomic analysis, on exposure to fialuridine and acetaminophen, respectively.

Conclusions: The high-throughput and liver-on-chip systems exhibit enhanced in vivo-like functions and demonstrate the potential utility of these platforms for DILI risk assessment. Tenofovir-inarigivr-associated hepatotoxicity was observed and correlates with the clinical manifestation of DILI observed in patients.

Impact and implications: Idiosyncratic (spontaneous, patient-specific) drug-induced liver injury (DILI) is difficult to study due to the lack of liver models that function as human liver tissue and are adaptable for large-scale drug screening. Human liver organoids grown from patient stem cells respond to known DILI-causing drugs in both a high-throughput and on a physiological "chip" culture system. These platforms show promise for researchers in their use as predictive models for novel drugs before entering clinical trials and as a potential in vitro diagnostic tool. Our findings support further development of patient-derived liver organoid lines and their use in the context of DILI research.

Keywords: drug development; hepatotoxicity; high-content imaging; microfluidic devices.

Fig. 1.. 384-well adaptation of HLOs.

HLOs grown from iPSC lines 72.3, 2E, and CC3 are dispersed into 384-well plates and treated with a 10-point dose response of 12 common DILI causing compounds. After 120 hrs incubation, cells are fixed and stained with Hoechst 33342, MitoView Green, HCS CellMaskOrange, and LipidTox DeepRed and imaged with an automated confocal microscope. (A) IC₅₀ values of these compounds through cell viability counts are calculated (n=4 per concentration, per cell line). (B) CellProfiler was used to extract features at each compound’s respective IC₅₀ values for 72.3-derived HLOs and embedded into UMAP. Plot points represent individual cells. Color intensity dictates the percentage of max measurement for each feature.

Fig. 2.. Development of a HLO-based Liver Chip.

HLOs developed from iPSC lines 72.3, 2E, and CC3 are disrupted into single-cell suspension and cultured into patient-derived liver organoids on chip (PaDLOCs) and compared against intact organoids on 12-well plates. (A) Albumin released in PaDLOCs is identical to that of plate HLOs at day 0 but increases over 7 days (day 21–28 of differentiation). (B) PaDLOCs turnover CYP1A, 2B, and 3A family substrates acetaminophen, cyclophosphamide, and darunavir at increased rate compared to plate HLOs. (C) Cells are treated with DMSO control and known hepatotoxins APAP (100 μM) and FIAU (1 μM). PaDLOCs demonstrated both ALT (D) and AST release and (E) albumin production diminishment across 7 days. Bars and plot points represent mean ± SD (n=3 PaDLOC chips and n=3 plate HLO wells). Statistical significance was calculated using ANOVA with multiple comparison Dunnett’s test. *, **, ***, and **** denote P values of less than 0.05, 0.01, 0.001, and 0.0001 respectively. (F) Confocal images of PaDLOCs at day 7 of treatment stained with CellMask Orange (magenta) and LipidTOX Deep Red (cyan). Images shown are scaled to identical intensity ranges. (G) UMAP clustering of 72.3-derived PaDLOCs highlighting a selection of liver specific genes. Each point represents one cell. Gray values represent no detected expression. (H) Volcano plot comparing gene differential expression between 72.3-derived PaDLOC and HLOs with genes most upregulated in PaDLOC highlighted (>0 designates higher expression in PaDLOCs).

Fig. 3.. Assessment of known DILI-causing drug combination: tenofovi-inarigivir.

(A) ALT, (B) AST, (C) and albumin released by 72,3, 2E, and CC3 PaDLOCs over 7 days of treatment with tenofovir (500 nM), inarigivir soproxil (500 nM), and tenofovir-xf inarigivir combination (250 + 250 nM) (n=3 chips per condition). Plot points represent mean ± SD. Statistical significance was calculated using ANOVA with multiple comparison Dunnett’s test. *, **, ***, and **** denote P values of less than 0.05, 0.01, 0.001, and 0.0001 respectively. (D) PaDLOCs treated with DMSO control, individual agents, combinations, APAP, and FIAU were stained with Hoechst 33342, CellMask Orange, and LipidTOX Deep Red. Images shown are scaled to identical intensity ranges. (E) CellProfiler extracted cell-level features were embedded into UMAP demonstrating morphological clustering. Plot points represent individual cells.

Fig. 4.. Single Cell Transcriptomics of treated PaDLOCs.

(A) Hepatocytes across treatments are identified and subset through marker expression and embedded into a UMAP to visualize similarities between treatments. Plot points represent individual cells. (B) Relative expression of DGAT1, PLIN4, FABP4, NDUFA4, PRDX4, and GSTP1 in vehicle control, fialuridine, tenofovir, and tenofovir-inarigivir treated PaDLOCs. (C) Volcano plots highlighting significant differential expression between control and drug treatments (>0 designates higher expression in treatment). (D) In 384-well cultures, 2-dimensional dose response assays show inarigivir and both tenofovir and fialuridine are synergistic with Bliss scores of 17.624 and 22.964, respectively, as calculated by SynergyFinder 2.0.