Five-Cell Liver Organoids Reveal an Indirect Imipramine Injury Pathway

A Five-Lineage Model Adds Intercellular Injury Routes

Drug-induced liver injury can begin in one cell population and damage another through intercellular signals. Researchers led by Tsinghua University constructed multi-lineage hepatic organoids to investigate this indirect toxicity. Their Nature Communications study combines a drug panel with detailed experiments tracing how imipramine-treated hepatic stellate cells affect neighboring hepatocytes.

All five cell populations were differentiated from the established H1 human embryonic stem-cell line: hepatocytes, cholangiocytes, endothelial cells, Kupffer cells and hepatic stellate cells. The starting mixture contained 60% hepatocytes and 10% of each other population, totaling 30,000 cells per organoid. The aggregates retained structural integrity over 35 days, with albumin secretion and selected metabolic activities maintained during culture.

Hepatocyte-only monolayers and spheroids provided controls for cell composition and three-dimensional organization. The multi-lineage organoids more closely resembled human-liver gene expression and showed higher measured CYP3A4, CYP1A2 and CYP2C9 activities. These functional comparisons established a context for the toxicity experiments; the organoids remained stem-cell-derived constructs with only partially characterized vascular and biliary functions.

The Drug Panel Finds Both Additional Hits and Missed Toxicities

The screen included 34 drugs associated with clinical liver injury and 24 negative comparators. The Results describe 14-day continuous exposure at the respective human clinical peak plasma concentrations, with medium replacement every three days. An ATP-based viability signal below 80% of vehicle control, together with statistical significance, defined a positive result.

The organoids identified 28 of 34 positive drugs, giving 82.4% sensitivity. Six of 24 negative drugs were incorrectly flagged, giving 75% specificity. Seven compounds reduced viability in the multi-lineage organoids without significantly affecting hepatocyte-only spheroids. These differential responses identified candidates for cell-interaction-dependent toxicity, rather than establishing the mechanism of every hit.

The larger cellular mixture was not uniformly more sensitive. The authors report that methotrexate and dantrolene injured hepatocyte-only spheroids but not the organoids; four other clinically liver-injury-associated drugs caused no cytotoxicity in either model. Adaptive immune mechanisms and systemic context were incompletely represented. These findings place the favorable screening performance alongside its missed and model-dependent responses.

Imipramine produced the most pronounced differential response and was selected for mechanistic investigation. In separate 24-hour experiments at 0.87 μM, corresponding to ten times the cited clinical peak concentration, it did not significantly increase IL-1β, IL-6 or TNF-α secretion. Positive-control inflammatory stimulation elicited those cytokines, supporting investigation of a different injury pathway under the imipramine condition.

Changed Vesicle Cargo Connects Stellate Cells to Hepatocyte Apoptosis

Hepatic stellate cells strongly expressed TRKB, the receptor encoded by NTRK2. Reducing TRKB expression attenuated their apoptotic response to imipramine. Conditioned medium from treated stellate cells subsequently caused hepatocyte apoptosis, with a stronger effect than medium from treated endothelial cells. Direct imipramine exposure did not produce the same hepatocyte response under the compared conditions.

The investigators then isolated extracellular vesicles described as exosomes from stellate cells exposed to 0.87 μM imipramine for three days. Vesicle number and average diameter did not increase significantly across the tested concentration range. Instead, their molecular cargo changed. Hepatocytes took up vesicles from treated cells and underwent more apoptosis; CD9-antibody treatment reduced uptake and partially protected the recipient cells.

The altered cargo included miR-34, identified in the study as miR-34a-3p. In recipient hepatocytes, it reduced XIAP, an inhibitor of apoptosis, alongside increased caspase-3 activation. Reporter experiments supported direct regulation of the XIAP transcript. Increasing or reducing p53 in stellate cells changed vesicular miR-34 and the subsequent hepatocyte injury, while exosome-production inhibition partly rescued organoid viability. The evidence therefore points to cargo remodeling rather than a general increase in vesicle production.

Separate 35-day mouse experiments supported this sequence: stellate-cell changes preceded hepatocyte injury, and interventions against TRKB, vesicle production or miR-34 reduced liver-injury measurements. The apoptosis comparisons used six mice per group and serum-enzyme measurements five per group. These animal experiments examined the proposed mechanism; the organoid screen and the mouse follow-up answer different validation questions.

The model’s ATP endpoint can miss transient or pathway-specific effects, and its static culture lacks physiological flow and adaptive immunity. Nevertheless, the imipramine experiments show how a multicellular assay can reveal an injury signal absent from a hepatocyte-only comparison, then connect that signal to a defined intercellular mechanism. The six false negatives and six false positives in the drug panel remain part of that experimental result.