The diabetes drug metformin blocked a cancer-killing process in lab cells — the opposite of what earlier studies found

Metformin, one of the most widely prescribed drugs on Earth, has spent a decade building a reputation as a potential anticancer agent — study after study suggested the cheap diabetes drug could push cancer cells toward ferroptosis, an iron-driven form of cell death. A new study in Pharmacology Research & Perspectives (2026) delivers a genuine plot twist: under cysteine starvation, metformin and its cousin phenformin did the exact opposite, shielding cancer cells from ferroptotic death.

What the researchers did

The team at the Korea Institute of Radiological and Medical Sciences cultured H1299 human lung cancer cells and HEYA8 ovarian cancer cells in custom cysteine-free medium, then added metformin or phenformin. Cysteine starvation normally collapses glutathione synthesis and leaves membranes exposed to runaway lipid peroxidation — ferroptosis’s signature. Both drugs dose-dependently rescued viability, and cell-death staining confirmed the dead-cell fraction dropped sharply. Control experiments confirmed the death program was genuinely ferroptotic: the ferroptosis inhibitors ferrostatin-1 and liproxstatin-1 and the iron chelator deferoxamine all restored viability, while the apoptosis blocker z-VAD-fmk did nothing. Intriguingly, a necroptosis inhibitor also rescued the cells, suggesting a hybrid death program with lipid peroxidation as the common thread.

The protection runs through mitochondria — not antioxidants

Here’s the mechanistic surprise: metformin did not restore glutathione, did not fix the oxidized-to-reduced glutathione ratio, and did not rescue the GPX4 enzyme. Its protection was entirely independent of the antioxidant axis ferroptosis researchers usually target. Instead, the trail led to mitochondria. Within two hours of cysteine deprivation — before major cell loss — mitochondria became hyperpolarized and oxygen consumption spiked, changes previously shown to be required for this form of ferroptosis. Metformin and phenformin, both established Complex I inhibitors, suppressed the hyperpolarization, curbed the oxygen spike, and blunted the pileup of ferrous iron inside mitochondria. Rotenone (a direct Complex I blocker) did the same, pinning the effect on mitochondrial bioenergetics. A bonus clue: cyclosporine A, which blocks the mitochondrial permeability transition pore, also protected cells — hinting that transient pore opening drives ferroptosis in ways standard measurements miss.

Why this doesn’t change diabetes care — but could change drug design

Three caveats keep this in the lab where it belongs. First, the protective doses were millimolar — far above the roughly 10–40 micromolar plasma levels in treated patients, where the authors saw no protection. Second, the effect is cell-type specific: other lung cancer lines (A549, H460, H23, H820) were far less sensitive to cysteine deprivation in the first place. Third, prior studies reported metformin inducing ferroptosis in lung, breast, and liver cancer models via NRF2/HO-1 and SLC7A11 pathways — the direction genuinely flips with context.

The practical warning is for drug developers, not patients: as pharma chases therapies that starve tumors of cysteine or block the SLC7A11/xCT cystine importer, co-administered biguanides — or other mitochondrial modulators — could quietly undermine them. If you take metformin for diabetes, nothing here is a reason to change anything; talk to your doctor, not a headline.

Where the evidence stands

• Preliminary: Dose-dependent rescue from cysteine-deprivation ferroptosis, with mechanistic probes (Complex I inhibition, mitochondrial iron and ROS measurements), in two human cancer cell lines.
• Contested: The direction of metformin’s effect flips with context — earlier studies found it induces ferroptosis via NRF2/HO-1 and SLC7A11 pathways.
• Absent: Clinical relevance at therapeutic doses (millimolar in the dish vs. micromolar in patients); animal or human data; generalizability across cancer types.