Each year, thousands of patients experience severe liver damage from drugs that passed extensive preclinical testing and were considered safe. These adverse reactions are rare and unpredictable, affecting only certain individuals without warning. When liver failure occurs, the implicated drug is withdrawn, leaving the medical community searching for explanations that conventional laboratory tests are not equipped to deliver.
This type of injury, known as idiosyncratic drug-induced liver injury (iDILI), is a leading cause of drug withdrawals and regulatory warnings globally. It represents a substantial proportion of acute liver failure cases unrelated to overdose. Despite extensive research over several decades, IDILI remains a persistent and unresolved challenge in drug development.
The primary difficulty in studying and predicting iDILI arises from its underlying mechanism: it is not the drug itself that directly damages liver cells, but rather the patient’s immune system. Specifically, CD8+ T cells attack the liver following a drug-induced, allergy-like immune response. This response is determined by the individual’s genetic makeup, particularly genes involved in immune recognition. Conventional laboratory models of liver toxicity lack immune cells and are therefore incapable of detecting this form of injury.
Researchers at Cincinnati Children’s Hospital Medical Center, in collaboration with Roche and Genentech, have developed a system that fundamentally addresses this limitation.
Mini-Livers Meet the Immune System
Published in Advanced Science in 2025, the new platform combines two cutting-edge technologies into a single, integrated system:
Human liver organoids (HLOs) are miniature, three-dimensional liver structures grown from a patient’s own induced pluripotent stem cells (iPSCs). These are not simple cell cultures; they contain functional hepatocytes (the main liver cells that process drugs), hepatic stellate cells, cholangiocytes, and vascular cells, essentially a microscale liver that produces albumin, responds to drugs, and behaves remarkably like real liver tissue.
Autologous CD8+ T cells are immune cells derived from the exact same donor as the liver organoids, primed and trained to respond to the drug being tested.
A critical aspect of this approach is the use of autologous cells, meaning both the liver and immune cells originate from the same individual. This is significant because the immune system’s recognition of a drug as a threat depends on specific genetic markers known as Human Leukocyte Antigens (HLA), which are cell-surface proteins that present molecular fragments to T cells. The high degree of individual variability in these markers explains why iDILI affects only certain people.
By pairing genetically matched liver organoids and T cells from the same donor within a system that allows direct interaction, this platform can, for the first time, replicate the complete sequence of events leading to immune-mediated liver damage in vitro.
The Engineering Challenge: Getting Rid of Matrigel
One of the less visible innovations in this platform is the exclusion of Matrigel. Culture has long depended on Matrigel, a gel derived from mouse tumor tissue that acts as a biological scaffold for organoid growth. It works well for growing organoids, but it creates a serious problem when you want to add immune cells: Matrigel physically blocks T cells from contacting the hepatocytes they need to attack, and it introduces mouse-derived signals that interfere with immune responses.
The Cincinnati team replaced Matrigel with a micropatterned hydrogel system, Gri3D, consisting of plates with 500-micrometer wells that accommodate approximately 250 cells, forming uniform organoids. Each well generates organoids of consistent size and structure without animal-derived scaffolds. This matrix-free environment enables immune cells to move freely and interact naturally with liver cells, which is essential for accurately modeling immune-mediated injury.
This approach produced organoids with significantly greater uniformity than those formed with Matrigel, resulting in highly consistent diameters and numbers across donors. Such consistency enables standardized and reproducible experiments at scale.
Testing the System: Can It Tell Good Drugs From Bad?
Prior to utilizing the platform for immune-mediated injury modeling, the researchers verified its ability to accurately assess conventional drug toxicity and to avoid generating false-positive alarms.
They tested two compounds with very different toxicity profiles:
Chlorpromazine is a psychiatric drug known to directly damage liver cells in a dose-dependent way, which scientists call intrinsic hepatotoxicity. As expected, the organoid platform showed clear, dose-dependent liver damage: cells died, ATP levels declined, and the organoids ceased secreting albumin, a marker of liver function. The results matched those reported by other 3D liver models.
Flucloxacillin is an antibiotic that causes liver injury only in certain patients via immune-mediated mechanisms, rather than through direct toxicity. When the platform was exposed to flucloxacillin in the absence of immune cells, no cell death or functional impairment was observed, even at concentrations exceeding those typically found in patients’ bloodstreams by more than 30-fold. This outcome demonstrates that the platform does not produce the false positives observed in other models.
The platform successfully passed this critical validation by accurately identifying direct toxicants while remaining unresponsive to immune-mediated agents, thereby confirming the essential and non-redundant role of the immune component.
The Genetic Key: HLA-B*57:01 and Flucloxacillin
Flucloxacillin is one of the best-characterized examples of a genetically restricted drug reaction in medicine. People who carry a genetic variant called HLA-B*57:01 are at significantly higher risk of developing liver injury when taking this antibiotic. The reason: their immune system’s recognition machinery is subtly shaped to identify flucloxacillin as a threat in a way that non-carriers’ immune system. However, possession of the HLA-B*57:01 allele does not guarantee liver injury; most carriers tolerate the drug without adverse effects. The occurrence of injury depends on the individual’s immune response, suggesting that genetic screening alone is insufficient to predict safety.
The research team collected blood from four HLA-B*57:01 carriers and four non-carriers. They isolated naïve CD8+ T cells and monocytes from each donor, converted the monocytes into dendritic cells (the immune system’s antigen-presenting specialists), and used those dendritic cells to train the T cells against flucloxacillin over 12 days.
All four HLA-B*57:01 carriers showed some degree of immune response. Two of them, designated as “strong responders,” showed robust T cell activation, as confirmed by a more than twofold increase in T cell proliferation and the appearance of multiple activation markers on the T cell surface. The two non-carriers showed no comparable response, confirming the HLA-restricted nature of the reaction.
The Final Test: Organ Meets Immune Cell
With primed T cells in hand, the team co-cultured them with liver organoids grown from iPSCs generated from each donor’s own blood cells, creating a fully personalized model of the immune–liver interaction.
The result was striking. Flucloxacillin-primed T cells from the two strong-responding HLA-B*57:01 carriers caused a fourfold increase in liver cell death in their matched organoids. The injured organoids released cytokeratin-18 (CK-18), a highly specific biomarker of hepatocyte death used clinically to detect liver injury, along with elevated levels of TNF-alpha and Granzyme B, two key molecular weapons used by CD8+ T cells to kill their targets.
T cells from the two weaker responders among HLA-B*57:01 carriers, as well as from all four non-carriers, did not induce comparable injury. The platform thus captured not only HLA restriction but also the individual variability that, in practice, determines whether a genetically susceptible person develops liver injury.
This represents a remarkable level of resolution, closely reflecting clinical observations in which two patients with the same genetic risk allele may have divergent outcomes: one develops liver failure while the other tolerates the drug without incident. The platform can distinguish between these cases.
What This Means for Drug Development and Patients
The implications of this platform extend well beyond flucloxacillin. The modular design means it can be adapted to study:
Immune checkpoint inhibitor toxicity is a growing clinical problem as cancer immunotherapy drugs increasingly cause liver inflammation in some patients, sometimes severely enough to require stopping treatment.
Autoimmune liver diseases, including autoimmune hepatitis, where the body’s immune system turns on its own liver without any drug involvement.
Biologic drug immunogenicity is the phenomenon where the immune system mounts a response against a biologic medicine, sometimes causing organ damage.
The platform introduces a capability previously lacking in drug safety testing: the evaluation of inter-individual immune variability prior to clinical trials. This approach enables pharmaceutical developers to identify genetic risk groups during drug development, rather than discovering adverse effects only after widespread patient exposure.
The researchers plan to scale the platform through automation, increase donor diversity for population-level analyses, incorporate single-cell profiling to identify predictive immune signatures, and ultimately integrate iPSC-derived T cells. This would enable a fully stem cell–derived, standardized immune–liver system that does not require donor blood samples.
The Bottom Line
Immune-mediated drug-induced liver injury has long represented a significant challenge in medicine, remaining undetectable by standard tests, unpredictable by genetics alone, and often resulting in severe outcomes. The platform described in Advanced Science by researchers at Cincinnati Children’s Hospital Medical Center, Roche, and Genentech constitutes the first genuinely functional solution to this problem.
By combining patient-derived liver organoids with genetically matched immune cells in a matrix-free, scalable microarray system, this platform recapitulates the entire sequence of events leading to immune-mediated liver damage, from T cell priming to hepatocyte death, and captures the individual variability that determines risk. New kind of safety test. For patients, it is the beginning of a future in which a drug’s danger to your liver can be assessed using your own cells, before you ever take a dose.