Stem Cell Platform Predicts Chemotherapy Damage to Healthy Tissue Better Than Animal Tests

One of the cruel ironies of cancer chemotherapy is that the drugs designed to kill tumors often devastate the patient’s own healthy tissue first. Hair follicles collapse, bone marrow fails, the gut lining erodes, and fertility can vanish — all because rapidly dividing stem and progenitor cells in these tissues are nearly as vulnerable to DNA-damaging agents as the malignant cells the treatment targets. A new study from Hungarian researchers, published in Pathology and Oncology Research, offers a strikingly direct way to model this collateral damage in a dish: a platform built from human induced pluripotent stem cells (hiPSCs) that can predict which cell types will be most sensitive to cytostatic drugs, potentially reducing reliance on animal testing in early drug safety screening.

The research team, led by Ágota Apáti and Balázs Sarkadi at the HUN-REN Research Center for Natural Sciences in Budapest, with colleagues at Semmelweis University, Salus Ltd., Innocell Ltd., and the University of Szeged, began with an unmet need that regulators themselves have begun to acknowledge. More than 90 percent of compounds that appear safe and effective in animal studies ultimately fail in human clinical trials, frequently because of unexpected toxicity. In response, frameworks such as the US FDA Modernization Act 2.0, the FDA’s 2025 Strategic Roadmap, and revised European Medicines Agency guidelines now formally encourage human cell-based approaches — so-called new approach methodologies (NAMs) — to establish drug safety. Yet despite this regulatory shift, healthy human stem and progenitor cells remain rare as preclinical toxicity controls.

The Budapest team’s solution was elegantly systematic. They reprogrammed umbilical cord blood cells from two donors — one female and one male — into induced pluripotent stem cell lines, named KUT-2 and KUT-6, using non-integrating Sendai virus vectors that leave no genetic trace. Cord blood was a deliberate choice: these newborn cells carry high telomerase activity, fewer accumulated somatic mutations, and a naïve epigenetic state that facilitates clean reprogramming, while avoiding the age-related damage and chronic inflammation that can persist in lines derived from adult donors. The lines were registered at hpscreg.eu, karyotyped, verified by short tandem repeat profiling, confirmed free of mycoplasma, and rigorously validated for pluripotency through marker staining and spontaneous differentiation into all three germ layers.

From these two isogenic starting points, the researchers differentiated the cells along distinct developmental lineages. Neural stem cells (NSCs) and hippocampal neural progenitor cells (hNPCs) represented ectoderm — a model of both the developing brain and adult neurogenesis in the dentate gyrus. Mesenchymal stem cell-like cells (MSCs) represented mesoderm, modeling the stromal cells that support tissue regeneration. Because all the derivatives shared the same genetic background within each line, any difference in drug response could be attributed to cell type rather than donor variability — a confound that has long plagued studies comparing unrelated primary cell lines. The neurons generated from the neural stem cells were even shown to be functionally active, firing spontaneous electrical activity recorded by multielectrode arrays and responding to glutamate in calcium imaging assays.

To probe different cellular stress pathways, the team assembled a panel of eight pharmacologically active compounds. Four were classic DNA-damaging anticancer agents: doxorubicin, which intercalates into DNA and poisons topoisomerase II; mitoxantrone, another topoisomerase II inhibitor; SN-38, the active metabolite that stabilizes topoisomerase I–DNA complexes; and laromustine, an alkylating agent that crosslinks DNA while blocking its own repair. The other four acted through non-genotoxic mechanisms: deferiprone, an iron chelator that disrupts mitochondrial function; celecoxib, a COX-2 inhibitor with pro-apoptotic effects; phenothiazine, which interferes with multiple signaling pathways; and quinacrine, an antimalarial with lysosome-disrupting and antitumor properties. Dose–response curves were generated in 384-well plates, and half-maximal inhibitory concentrations (IC50 values) were computed from viability measurements using a resazurin-based fluorescent assay.

The results revealed a striking biological gradient. When challenged with the DNA-damaging agents, the pluripotent stem cells themselves were by far the most sensitive, showing the lowest IC50 values, followed closely by the neural progenitors. The mesenchymal derivatives were dramatically more resistant — with IC50 values one to two orders of magnitude higher than the other cell types. Crucially, this pattern held across all four genotoxic compounds and both donor lines, indicating that sensitivity was dictated by cell type, not by the specific chemistry of the drug or by the donor’s sex. When the same drugs were tested against conventional cancer cell lines — a uterine sarcoma line and an ovarian adenocarcinoma line — their IC50 values were closer to those of the resistant MSCs than to the hypersensitive stem cells.

For the non-genotoxic compounds, a different pattern emerged. Mesenchymal cells, pluripotent cells, and neural stem cells showed comparable sensitivities, while the hippocampal neural progenitors were consistently and significantly the most vulnerable — a finding the authors link to the known dependence of neural progenitors on lysosomal homeostasis and autophagic flux, pathways disrupted by quinacrine and deferiprone. The hNPCs’ heightened susceptibility suggests this platform could serve as a model for chemotherapy-related cognitive impairment — the notorious

Subject of Research: A human iPSC-derived stem and progenitor cell platform for predicting chemotherapy cytotoxicity in healthy tissues

Article Title: Beyond animal models: using human pluripotent stem cell-derived stem and progenitor cells to predict cytotoxicity in normal tissues

Article References: Vincze, K., Varga, N., Tóth, S., Kis, A. K., Béres, R., Réthelyi, J. M., Veréb, Z. J., Erdei, Z., Sarkadi, B., & Apáti, Á. (2026). Beyond animal models: using human pluripotent stem cell-derived stem and progenitor cells to predict cytotoxicity in normal tissues. Pathology & Oncology Research, 32, Article 1612490. https://doi.org/10.3389/pore.2026.1612490

Image Credits: AI Generated

DOI: 10.3389/pore.2026.1612490

Keywords: induced pluripotent stem cells, cytotoxicity, chemotherapy, drug safety screening, neural progenitor cells, mesenchymal stem cells, DNA damage, new approach methodologies, umbilical cord blood, animal model alternatives, IC50, toxicology