Will you live to 120? · The Cells That Won’t Die

The most expensive lesson so far

The converse experiment seals the case. If removing senescent cells helps, does adding them harm? In 2018, a group led by Xu and Kirkland transplanted relatively small numbers of senescent cells into young, healthy mice — and the recipients developed physical dysfunction (slower walking speed, weaker grip, less endurance), persistent ill health, and, strikingly, spreading senescence in their own tissues, as the transplanted cells recruited the animals’ native cells into the same state. A modest dose of zombie cells was enough; transplanting the same number of normal, non-senescent cells did nothing. Put the two experiments side by side and the causal logic closes from both directions: clear senescent cells and aged mice improve; introduce senescent cells and young mice deteriorate. Senescent cells are not merely correlated with the damage of aging — at least in mice, they can actively cause it.

But van Deursen’s clearance experiment used a genetic switch built into the mice. To do it in a person, you need a drug that selectively kills senescent cells — a senolytic. And here the apoptosis-resistance from earlier becomes the key that unlocks the therapy. Recall that a senescent cell actively reinforces the pathways that keep it from dying. It survives only by leaning hard on a specific set of pro-survival proteins — its anti-apoptotic crutches. A normal, healthy cell isn’t leaning on those crutches; it has no pressing reason to die, so it isn’t propping itself up against death. That difference is the therapeutic window. Knock the crutch out, and the senescent cell — which depended on it — topples into the suicide it was holding off, while the healthy cell barely notices. Senolytics are drugs that knock out the crutch.

The first ones were found this way, by hypothesis rather than by accident. In 2015, James Kirkland’s group reasoned out which survival pathways senescent cells depend on and tested drugs against them, identifying dasatinib (a cancer kinase-inhibitor) and quercetin (a plant flavonoid) — dasatinib preferentially killing some senescent cell types and quercetin others, with the combination proving effective across a broader range than either drug alone. Others followed: fisetin, another plant flavonoid (though its supporting evidence remains weaker than for dasatinib-and-quercetin); navitoclax, a potent blocker of the BCL-xL survival protein. And these drugs carry a genuinely novel and appealing feature. Most medicines must be taken continuously, because they work only while the drug occupies its target. A senolytic doesn’t: once you kill a senescent cell, it is gone, and it takes time for new ones to accumulate. So senolytics could in principle be given in brief, intermittent courses — a “hit-and-run” approach, dosing for a few days, then stopping for weeks or months. That is a genuinely different therapeutic paradigm, and part of why the idea generated so much hope, and so much money.

Deeper Water — the crutch, and the cost of pulling it. The senescent cell’s survival crutches are known collectively as the senescent-cell anti-apoptotic pathways, and they include the BCL-2 family (especially BCL-xL), the PI3K/AKT survival axis, and a p53-regulating interaction involving a protein called FOXO4. Different senolytics pull different crutches — dasatinib and quercetin hit a broad set including kinase- and survival-signaling; navitoclax specifically blocks BCL-xL. And navitoclax illustrates, painfully, why “knock out the crutch” is harder than it sounds: BCL-xL is not used only by senescent cells. Platelets — the cell fragments that clot your blood — also depend on it to survive, so navitoclax causes their numbers to crash (thrombocytopenia), a dose-limiting toxicity that has hampered its use. The therapeutic window is real but it is narrow and target-dependent: the more a survival pathway is shared between zombie cells and important healthy cells, the smaller the window. This is the molecular reason senolytics cannot simply be made arbitrarily potent — and a preview of why moving from a clean mouse experiment to a safe, effective human drug is so much harder than the elegant logic suggests.

The most expensive lesson so far

Everything above is true, and the mouse data are real, and the logic is beautiful. And the first major test of that logic in human beings failed — publicly, expensively, and instructively.

The company was Unity Biotechnology, and it is important to understand that this was not some fly-by-night operation. Unity was founded by the field’s leading lights — including Judith Campisi, who reframed senescence as a driver of aging, and van Deursen and Kirkland, whose work you have just read — and it was backed by some of the most sophisticated investors in the world, with high-profile names and hundreds of millions of dollars behind it. When it went public in 2018, it carried the field’s brightest hopes. Its lead program, UBX0101, was a senolytic designed to be injected directly into an arthritic knee, to clear the senescent cells thought to be driving the inflammation and cartilage breakdown of osteoarthritis. The strategy was local (inject just the joint, sidestep whole-body toxicity), the target was concrete, and the science was first-rate.

In 2020, the Phase 2 trial results came back. UBX0101 did no better than placebo at relieving knee-osteoarthritis pain. The drug failed its primary endpoint. The company’s stock collapsed, and Unity reorganized around a different program entirely — a senolytic for retinal eye diseases, which has since produced more encouraging but still early signals.

Sit with what that failure was and was not. It was not a failure of chemistry, or of artificial intelligence, or of effort, or of expertise. The people involved understood senescence as well as anyone alive. It was, instead, the same failure this book has now shown you three times in three chapters, in three different costumes: a gorgeous, well-supported result in mice that did not transfer to humans. Maybe the senescent cells weren’t actually the dominant driver of osteoarthritis pain in people the way they appeared to be in mouse models. Maybe a single injection didn’t clear enough of them, or the right ones. Maybe the heterogeneity of senescence meant the drug killed some senescent cells but not the ones that mattered — and remember, with no clean marker, it was hard to even confirm what had been cleared. Maybe human joints and mouse joints simply differ in the way that made the experiment work. Any of these could be true; the trial cannot, by itself, tell you which. What it tells you is that the inference “clearing senescent cells fixes the disease” — so clean in the mouse — broke somewhere on the way to the human, exactly as the telomere story and the clock story broke before it.

All 10 chapters of Will you live to 120? · Foundations · All courses