For more than a decade, a large part of what aging research thought it knew about “zombie cells” was measured with the wrong tool. Not a subtly miscalibrated tool. The wrong protein entirely.

This year a researcher named Sholto David did something nobody in the field had bothered to do: he checked the catalog numbers. He pulled together hundreds of published studies that claimed to measure p16, the most-cited marker of cellular senescence, and looked at which antibody each one had actually bought. Of the 334 papers he could read in full, only 17 had used an antibody that detects p16. The rest, without knowing it, had been staining for a completely different protein. The mistake sits in Nature, Nature Medicine, Cancer Cell, and a long list of other journals that are supposed to be the careful ones.

If you have read how science actually works, you know I think the interesting failures are not the frauds but the honest, systemic ones. This is a textbook example, and it is worth walking through slowly, because the lesson is bigger than one reagent.

Two proteins, one name

There are two unrelated proteins that both got tagged with “p16.”

The first is p16INK4a, the product of the CDKN2A gene. It is a brake on the cell cycle and one of the central markers of senescence, the state where a cell stops dividing but stays alive and leaks inflammatory signals into the tissue around it. This is the p16 that aging research cares about.

The second is p16-Arc, also called ARPC5, a small subunit of the Arp2/3 complex that helps build the actin cytoskeleton. It has nothing to do with aging. It is a housekeeping structural protein.

The names collide by pure coincidence, and the antibody catalogs made the collision easy to fall into. Several widely sold antibodies listed under “p16” (clones such as EP1551Y, and catalog codes ab51243, ab151303, and sc-166760) detect p16-Arc, not p16INK4a. A researcher searching a supplier’s site for “p16” could add the wrong vial to the cart in seconds and never notice.

And here is why the error survived so long: a band on a western blot looks like a result no matter which protein made it. p16-Arc is expressed almost everywhere and roughly tracks the amount of cells you loaded, so the signal even looked plausible. The tool always produced a number. The number was just measuring the cytoskeleton.

What survives, and what does not

It would be easy to read this as “senescence was a myth.” That is the wrong conclusion, and getting the calibration right matters.

The causal backbone of the field does not rest on the antibody. The strongest evidence that senescent cells actively drive aging comes from genetics, not staining. In Jan van Deursen’s lab at the Mayo Clinic, mice were engineered with a construct called INK-ATTAC, which makes any cell that switches on the p16 gene killable on command. Flip the switch and the p16-positive cells die. Do that and the animals age better: clearing those cells delayed age-related disease (Nature, 2011), and in naturally aged mice it extended median lifespan (Nature, 2016). None of this used the faulty antibody. It used the gene’s own promoter and a genetic kill switch.

So the idea is intact. Senescent cells are real, they accumulate with age, and removing them helps mice. What is shaky is the decade of measurement stacked on top of that idea, the thousands of experiments that claimed to count senescent cells or watch p16 rise and fall by staining for it.

What it means for longevity research

Senescence is one of the load-bearing pillars of modern aging biology, and p16 is its single most-used marker. If most p16 staining was quietly reading an actin protein, then a large slice of the literature is now uncertain: findings that “senescent-cell burden rose in this aging tissue,” or “fell after this candidate drug,” or “correlated with this disease in human samples,” cannot be taken at face value until someone checks which antibody produced them.

Not everything falls. Studies that measured CDKN2A messenger RNA, or used genetic reporters, or leaned on other senescence markers, are untouched. The contamination is wide, not total. But the antibody was the workhorse, which is exactly why it does so much damage.

And the damage compounds. Papers cite papers. Grant applications, biotech pipelines, and clinical rationales were all built partly on a measurement that was frequently wrong, and each new result inherited the error from the ones it stood on. A full decade passed, in the most selective journals in biology, before anyone thought to read the label on the bottle. The work ahead is unglamorous: an audit of which conclusions leaned on the bad reagent, and a re-run of the ones that did.

The lesson for anyone reading longevity claims is the same one that runs under every good scientific instinct. One method, however ubiquitous, is a single point of failure. Real signal is convergence: the same conclusion reached by orthogonal methods that fail in different ways. That is precisely why a handful of genetic mouse experiments now outweigh a thousand western blots. The blots all shared one blind spot. The genetics did not.

The casualty: fisetin

Nowhere did the hype outrun the evidence faster than with fisetin, a plant flavonoid sold as a senolytic, a compound that supposedly clears senescent cells. The supplement industry built a category on early mouse studies and a compelling mechanism story.

Then came the rigorous test. The National Institute on Aging runs the Interventions Testing Program, which deliberately re-runs lifespan claims across multiple independent labs to weed out results that only work in one pair of hands. Fisetin went through it and did not extend life in mice. In humans the senolytic story is thinner still: the trials are small and early, such as a phase I/II randomized study of fisetin in knee osteoarthritis, and none has yet delivered the convincing efficacy result the marketing quietly implies. The bottles shipped years before the proof, and the proof has not arrived.

The boring lever

The biology is real. The supplement aisle is not the biology. Those are two different sentences, and the space between them is where a lot of money changes hands.

If you want to act on cellular senescence today with something that has actually earned its evidence, the answer is dull enough to be disappointing: exercise. Its effects on senescence and inflammation show up across many methods that do not depend on one contested antibody, and it clears every other bar that longevity supplements keep tripping over. The boring foundation beats the exciting bet, again, the way it almost always does.

That is the quiet moral of the p16 story too. The flashy layer, the marker everyone stained for, the molecule everyone sold, turned out to be the fragile part. The unglamorous work underneath, the genetics and the sweat, is what held.

References

1. Science (AAAS). Protein name confusion created antibody mix-up affecting hundreds of papers. https://www.science.org/content/article/protein-name-confusion-created-antibody-mix-affecting-hundreds-papers
2. For Better Science. Mind over Antibody (June 2026). https://forbetterscience.com/2026/06/02/mind-over-antibody/
3. Baker, D.J., et al. (2011). Clearance of p16Ink4a-positive senescent cells delays ageing-associated disorders. Nature. https://www.nature.com/articles/nature10600
4. Baker, D.J., et al. (2016). Naturally occurring p16Ink4a-positive cells shorten healthy lifespan. Nature, 530, 184-189. https://www.nature.com/articles/nature16932
5. Fight Aging! (2023). The NIA Interventions Testing Program shows that fisetin does not extend life in mice. https://www.fightaging.org/archives/2023/12/the-nia-interventions-testing-program-shows-that-fisetin-does-not-extend-life-in-mice/