3D HepaRG spheroids strengthen transcriptomic genotoxicity classification
A human-hepatic 3D model is moving genotoxicity testing closer to biology that resembles the liver rather than a flat cell monolayer on plastic. In this study, metabolically active HepaRG spheroids were paired with two transcriptomic biomarkers, GENOMARK and TGx-DDI, to determine whether gene expression readouts could classify chemicals as genotoxic or non-genotoxic with greater human relevance than older in vitro formats. The work matters because genotoxicity assessment sits at the front end of chemical safety evaluation, where a false negative can miss a DNA-damaging hazard and a false positive can unnecessarily remove a useful compound from further development.
The platform used here was deliberately more demanding than a simple 2D exposure system. HepaRG spheroids retain metabolic activity, which is essential when the chemical under study needs biotransformation before its genotoxic potential becomes visible. Several agents only become harmful after hepatic metabolism, and a screening model that cannot generate those metabolites risks undercalling risk. By extending exposure to 72 hours and testing a chemically diverse set of compounds, the investigators placed the biomarkers in a setting designed to challenge their predictive value rather than merely confirm what was already obvious.
Chemical classes tested against the liver model
Three groups of compounds were evaluated. The genotoxic set included lasiocarpine, cyclophosphamide, ethyl methanesulfonate, aflatoxin B1, and benzo[a]pyrene. The non-genotoxic non-carcinogen set included aflatoxin B2, 2-deoxy-D-glucose, D-mannitol, caffeine, and ampicillin trihydrate. The third group consisted of non-genotoxic carcinogens: perfluorooctanoic acid, diethanolamine, thioacetamide, phorbol-12-myristate-13-acetate, and di(2-ethylhexyl) phthalate.
That mix is important because genotoxicity testing cannot rely on a simple binary split between “toxic” and “safe.” Some chemicals damage DNA directly, some require metabolic activation, and others drive carcinogenesis through non-genotoxic mechanisms such as sustained cellular stress, receptor-mediated effects, or chronic tissue injury. A tool that can separate those categories in a metabolically competent human liver model has more practical value than a system that only detects overt DNA reactivity.
GENOMARK and TGx-DDI remained aligned with established hazard profiles
Gene expression data were generated with high-throughput TempO-Seq and then analyzed using the two transcriptomic biomarkers. Both biomarkers generally aligned with existing knowledge across the three chemical groups and effectively distinguished genotoxic from non-genotoxic compounds. The classification performance across the tested set supports the central premise of the study: transcriptomic signatures derived in 2D systems can retain utility when moved into a more complex 3D hepatic model.
A single sentence captures the core result. The biomarkers tracked expected genotoxicity patterns in a spheroid system that better preserves metabolic function.
The biologic rationale is straightforward. Genotoxic injury often triggers coordinated transcriptional responses tied to DNA damage recognition, cell-cycle control, stress signaling, and repair programs. If a chemical causes lesions that a hepatocyte can sense, the transcriptome should reflect that disturbance even before frank cell death appears. In a metabolically active spheroid, this becomes more meaningful because the model can convert pro-genotoxins into reactive intermediates, making the transcriptional signal closer to what might occur in human liver tissue.
Why the 3D format matters for animal-free safety testing
The search for New Approach Methodologies in genotoxicity testing is driven by two pressures at once: the push away from animal use and the need for human-relevant predictive systems. Flat cell cultures have long served as workhorses in toxicology, but they can underrepresent architecture, polarity, and metabolic competence. Three-dimensional spheroids provide a more structured microenvironment, and HepaRG cells are particularly attractive because of their hepatic characteristics and metabolic capacity.
In this setting, the combination of a 3D model and transcriptomic biomarkers does more than add technical sophistication. It addresses a central weakness of many conventional screens: a chemical may look harmless in a less metabolically active system and reveal its hazard only when the proper enzymatic machinery is present. For compounds such as cyclophosphamide, aflatoxin B1, or benzo[a]pyrene, that distinction is not academic. It determines whether the assay detects a real hazard or misses it altogether.
The fact that both GENOMARK and TGx-DDI performed consistently across genotoxic compounds, non-genotoxic non-carcinogens, and non-genotoxic carcinogens suggests that transcriptomic classification can preserve interpretability even in a more physiologically demanding model. That is the kind of behavior regulatory toxicology needs: not a black-box signal, but a pattern that maps onto known chemical behavior.
Bedside relevance is indirect, but the translational logic is strong
This is not a clinical treatment study, yet the downstream implications are tangible. Drug development and chemical safety decisions begin long before a patient ever encounters a compound. If a human-relevant assay can better discriminate DNA-reactive hazards from chemicals that cause cancer through other routes, it can refine early screening, prioritize confirmatory testing, and reduce dependence on animal models that may not capture human metabolism with equal fidelity.
The findings also fit a broader mechanistic principle in toxicology: a liver model that maintains metabolic activity is more likely to expose hidden genotoxicity. Transcriptomic biomarkers then act as molecular interpreters, reading out the cellular response to injury in a way that can be standardized across compounds. In practical terms, that could help toxicologists decide whether a candidate chemical deserves deeper genotoxic workup or whether its hazard profile is more consistent with non-genotoxic carcinogenicity.
No single assay will settle every safety question. But a metabolically active 3D hepatic system paired with transcriptomic biomarkers brings the field closer to a platform that reflects human biology rather than forcing human interpretation onto a simplified experimental artifact.
Limits of the evidence remain tied to assay scope
The study demonstrates concordance with known hazard profiles, yet its scope is still defined by a finite panel of fifteen chemicals. Broader chemical diversity will be needed before the approach can be generalized with confidence across the full range of industrial and pharmaceutical compounds. The work also centers on classification performance rather than on quantitative potency ranking, dose-response modeling across a wider chemical universe, or direct comparison against a full regulatory testing battery.
Even so, the signal is clear enough to matter. In metabolically active HepaRG spheroids, GENOMARK and TGx-DDI remained useful for separating genotoxic from non-genotoxic chemicals, including non-genotoxic carcinogens that would otherwise complicate simple hazard binaries. The next challenge is not whether the concept works in principle, but how far this transcriptomic spheroid strategy can be pushed across larger panels, mixed mechanisms, and the operational demands of routine toxicology workflows.