Polyamines are among the most studied molecules in aging research. Spermidine supplementation extends lifespan in yeast, worms, flies, and mice, primarily by promoting autophagy — the cellular recycling process that clears damaged organelles. The mechanism runs through a protein called eIF5A, which requires a polyamine-derived modification (hypusination) to function. Activate eIF5A, enhance autophagy, slow aging. The pipeline is clean.
Sato and colleagues at Tokyo University of Science, publishing in the Journal of Biological Chemistry in 2026, identified the complication. Humans have two eIF5A genes: eIF5A1 and eIF5A2. They share 84% amino acid sequence identity. Both require hypusination by the same enzyme. eIF5A1 drives autophagy and cellular repair. eIF5A2 drives ribosomal biogenesis, cellular proliferation, and is associated with cancer progression. Polyamines activate both.
The molecule cannot choose. The binding site is too similar. Spermidine promotes the repair protein and the growth protein simultaneously, because the structural features that the hypusination enzyme recognizes are shared between the two targets. The specificity that the supplement would need to be purely therapeutic — distinguishing eIF5A1 from eIF5A2 — exceeds what the chemistry can provide.
The structural lesson is about what gene duplication does to specificity. Before the duplication, there was one eIF5A gene and one activator. The specificity was perfect — not because the activator was precise, but because there was only one target. The activator didn't need to discriminate. When the gene duplicated and the two copies diverged in function — one toward repair, one toward proliferation — the activator's binding affinity didn't change. It still recognizes the conserved features. But now those features appear on two proteins with opposing biological consequences.
The duplication created the ambiguity retroactively. The original system was specific by default: one key, one lock, no discrimination required. The duplication produced a second lock similar enough to accept the same key. The key didn't become less precise. The world became more ambiguous around it. The supplement that extends lifespan through eIF5A1 also feeds the machinery of cancer through eIF5A2, and the 16% sequence divergence between them is not enough for a small molecule to tell the difference.