In 1678, Antonie van Leeuwenhoek pointed his microscope at human semen and observed crystalline structures that nobody could explain. The crystals were spermine phosphate, though it took 250 years for Otto Rosenheim to work out the chemical structure. In the intervening centuries, spermine accumulated known functions the way old buildings accumulate plaques: it stabilizes DNA, regulates cell growth, modulates ion channels. Each role was catalogued, studied, and filed away as understood.
What nobody tested, for 340 years, was whether spermine binds iron.
The answer turns out to be physically obvious. Spermine carries a 4+ charge at physiological pH. Four protonated amino groups spaced along a carbon backbone. Ferrous iron, Fe2+, is a divalent cation that sits comfortably in the electrostatic cage those amino groups create. The chelation is not a biological innovation. It is a physical inevitability of spermine's molecular structure. It was happening in Leeuwenhoek's crystallization dishes. It was happening in every cell in every organism that produces polyamines. The function was always there. The question was always available. Nobody asked it.
Li and colleagues at Sun Yat-sen University asked it. Their paper, published in Nature in June 2026, demonstrates that spermine is an endogenous iron chelator that inhibits ferroptosis. Ferroptosis is cell death driven by iron. The mechanism is blunt: free ferrous iron catalyzes the Fenton reaction, generating hydroxyl radicals that rip through membrane lipids. The polyunsaturated fatty acids in cell membranes are particularly vulnerable. Once lipid peroxidation cascades past the cell's repair capacity, the membrane fails and the cell dies.
Every cell that handles iron faces this problem. You need iron for oxygen transport, electron transfer, DNA synthesis. But the same reactivity that makes iron useful makes it lethal when it's loose in the cytoplasm. The cell's standard defense is GPX4, an enzyme that reduces lipid peroxides before they can propagate. Deplete GPX4 or overwhelm it with iron, and ferroptosis follows.
Spermine, it turns out, works upstream of all this. By chelating Fe2+ directly, it prevents the Fenton reaction from starting. It physically sequesters the iron so it never touches the membrane lipids. The research team tracked the mechanism using stable isotope tracing and metabolomics, demonstrating that labeled glutamine flowed through a specific biosynthetic route to produce spermine that then bound intracellular iron.
The protective function makes sense. Cells produce spermine as part of normal polyamine metabolism. The iron chelation is a bonus, a physical freebie that has been supplementing the cell's defenses against accidental lipid damage for as long as polyamine-producing organisms have existed.
Here is where the story turns.
Hepatocellular carcinoma cells are glutamine-addicted. They consume enormous quantities of glutamine to fuel proliferation, synthesize nucleotides, and maintain antioxidant defenses. Li's team discovered that these cancer cells upregulate an enzyme called ALDH18A1, also known as P5CS, which opens an alternative pathway: glutamine to glutamate to pyrroline-5-carboxylate to ornithine. Ornithine then feeds into the standard polyamine cascade: ornithine to putrescine to spermidine to spermine.
This is a precursor bypass. The conventional route to ornithine runs through the urea cycle. The ALDH18A1 pathway taps the cancer cell's existing glutamine addiction and reroutes it to produce more spermine than a healthy cell would ever need. The extra spermine chelates extra iron. The extra chelation suppresses ferroptosis. The cancer cell survives conditions that would kill it otherwise.
The same physical property that protects healthy cells from accidental lipid damage now protects cancer cells from programmed death. The molecule hasn't changed. The chemistry hasn't changed. What changed is who benefits.
Li's team tested this from multiple angles. Genetic ablation of ALDH18A1 triggered ferroptosis in hepatocellular carcinoma cells. AAV-delivered short hairpin RNA against ALDH18A1 did the same. A small molecule inhibitor, YG1702, impaired both spontaneous and chemically induced liver tumorigenesis in mouse models. Block the precursor bypass, and the cancer cell's extra spermine supply disappears. Block the spermine supply, and iron goes back to doing what free iron does. The membranes fail. The cell dies.
The flip side: exogenous spermine, administered directly, protected healthy organs from ferroptosis-driven ischemia-reperfusion injury in liver, intestines, and kidneys. Same molecule. Opposite clinical context. Shield in one setting, weapon in the other.
The specific cancer biology matters, but the structural insight is what I keep returning to.
Spermine's iron chelation is what you might call a found mechanism. It was not evolved for the purpose of binding iron. It does not require any protein machinery to perform the chelation. The binding is an intrinsic consequence of the molecule's charge distribution. It is a physical property sitting latently inside a metabolite that cells already produce for other reasons.
Found mechanisms are everywhere in biology. Metabolites with latent physical properties. Side reactions of enzymes that happen to produce useful byproducts. Structural features of proteins that serve functions unrelated to their evolved role. Most of the time, these are invisible. They work in the background, providing a diffuse benefit that never gets attributed to any specific selective pressure because it was never selected for in the first place.
The problem becomes visible when selection levels conflict. A healthy organism benefits from spermine's iron chelation because it prevents accidental cell death. A tumor benefits from the same chelation because it prevents deliberate cell death. The molecule doesn't know the difference. It is a polycation in the presence of a divalent cation. It chelates. The physics doesn't have an opinion about who lives and who dies.
Cancer's trick, in this case, is not to invent new chemistry. It is to build new logistics. ALDH18A1 upregulation is a metabolic reroute, not a novel enzyme. Glutamine addiction is a pre-existing condition that the pathway exploits. The cancer cell identifies a physical property already present in the cellular commons and amplifies its supply line. Everything downstream is just physics following its instructions.
This reframes how we think about metabolic resistance in cancer. The standard model assumes cancer evolves novel mechanisms to resist therapy. Some of it does. But some of it may be amplification of physical properties that were always there, always doing something, just never doing enough to matter until a somatic lineage figured out how to turn up the supply. The search for novel resistance mechanisms might be looking in the wrong place. The resistance was sitting in the metabolite pool the whole time, below the resolution of anyone who thought spermine's known functions were its only functions.
Three hundred and forty years of studying a molecule. Its most consequential function, hiding in plain sight, explained by first-year electrostatics. The shield was always there. The question is who holds it.