An expert from CN Bio explores considerations when choosing an in vitro liver model
As the major drug metabolising organ, the liver is central to safety and DMPK assessments, however, a diverse set of in vitro liver models are available from simple suspension cultures to advanced microphysiological systems (MPS). No model is universally best. Each provides different strengths, limitations and practical considerations. When selecting the right tool for the job, it is imperative to match the model to the specific question you need to answer.
Across our industry the same challenges hinder decision making, unclear or conflicting animal data, late stage failure risk and more mechanistic clarity. Addressing these issues requires aligning model capabilities to what you need to predict, detect, or understand. This article explores three common scenarios where liver models diverge in suitability, highlighting where MPS provides clarity to proceed with greater confidence.
Scenario 1. Predicting drug clearance or metabolite-induced DILI
When the question is ‘what metabolism affects my molecules and how quickly are they cleared?’ your model needs to demonstrate high human metabolic fidelity, broad enzyme coverage and stability over time.
Kratochwil et al., (2017)1 compared the metabolic activities of multiple liver cultures across 11 enzyme markers. HepaRG, HepatoPac, and HµREL were broadly comparable to primary human hepatocytes (PHH) suspensions, with some differences. Contrastingly, iPSC derived hepatocytes and HepG2 activities were tenfold lower and therefore less suited. HepaRGs offer cost and scalability advantages but can underpredict due to slower metabolic turnover, while PHH suspension assays are short-lived. To extend study windows, researchers previously turned to HµREL, or HepatoPac formats.
However, certain questions require longer experimental windows, including clearance studies beyond a week, or studying enzyme induction – which typically requires multiple dosing cycles. Additionally, more sophisticated systems may be required to assess the combined effects of metabolism by gut and liver to improve oral bioavailability predictions. These previously un-addressable questions require a different approach using complementary perfused Liver MPS (Liver-on-a-chip) offering the highest possible metabolic relevance. Their dynamic flow conditions extend performance over weeks and enable human gut and liver models to be interconnected2.
Where your question relates to predicting metabolite-induced DILI, is a simple yes/no triage sufficient, or are deeper mechanistic insights required? Spheroid models represent a practical screening tool to improve human relevance, but sensitivity is restricted, as demonstrated by Rubiano et al., (2020), who showed that PHH CYP3A4 activity was significantly higher and more stable in Liver MPS cultures3. The enhanced capabilities of MPS also benefit mechanistic pathway interrogation and latent hepatotoxicity identification. This is particularly relevant for idiosyncratic DILI which can develop slowly as metabolites accumulate. Therefore, combining spheroid screening with more translational MPS insights delivers a comprehensive workflow for reducing late-stage risk.
Scenario 2. Detecting the mechanism of cholestatic injury
When the question becomes “Do my lead candidates pose cholestasis risks?” requirements shift to high bileacid transport fidelity, stable long-term function and detecting a panel of cholestasis-relevant endpoints.
MPS systems are the primary option, but sensitivity varies between platforms. In a comparative study, Nitsche et al., (2025) highlighted that only CN Bio’s PhysioMimix Liver MPS detected consistent bileacid reductions and early mechanistic changes across cholestatic toxicants. Both perfused HepaRG and PHH cultures reported changes in total bile acid (late-stage cholestatic biomarker). Glycine conjugated bile acids, consistent with human liver, were most abundant in PHH cultures, whereas HepaRG mainly produced taurine conjugates4. This suggests an optimal PhysioMimix Liver MPS workflow of HepaRG for initial cost efficient risk identification, then PHHs for mechanistic confirmation with clinically aligned biomarkers. Bile acids are measured in media (via kits or LC-MS), whilst bile acid synthesis/metabolism/ transporter expression is via OMICs analysis of recovered microtissue.
Scenario 3. Understanding if preclinical in vivo toxicity will translate to humans
Another challenge is determining ‘which animal species predicts the human response’, when two animal species give opposing or inconsistent results. This requires consistent cross-species culture conditions, stable function >10 days and the detection of interspecies divergence.
In 2025, Negi et al., compared primary hepatocyte function and drug effects in human, monkey, dog and rat using a PhysioMimix Liver MPS and 2D assays5. MPS assays maintained longer-term function and captured more species specific toxicity responses than 2D. Although more expensive and lower throughput, the number of drugs to be screened in this context is nominal. Therefore, the translational clarity offered by MPS is justified to minimise drug misclassification risk, support confident clinical development progression and prevent safe asset loss from pipelines.
Summary:
Each liver model provides different strengths in metabolic and mechanistic competence, practicality and cost. Rather than asking which model is ‘best’, understanding and strategically combining these tools to answer specific questions will derisk programmes earlier.
As global regulatory momentum shifts towards NAMs approaches, MPS have become a risk reduction necessity for guiding decision making and building holistic, mechanistically informed regulatory packages that accelerate the path to market.
1. https://doi.org/10.1208/s12248-016-0019-7
2. https://doi.org/10.1016/j.dmd.2025.100130
3. https://doi.org/10.1111/cts.12969
4. https://doi.org/10.1007/s00204-025-04263-1
5. https://doi.org/10.1021/acsptsci.5c00554
For more information visit: www.cn-bio.com
