Sixty Years On, Senescence Is Better Understood Than Ever, and Senolytics Still Have Not Delivered in Humans

News
#1

Senolytics
Senolytics2754×1536 452 KB

This is a historical review from Marco Demaria’s group in Groningen. It traces how cellular senescence went from a cell-culture curiosity in 1961 to a drug target. The main message is that senescence is not one thing. It suppresses tumors, helps wounds heal and shapes embryos, yet when senescent cells persist they drive inflammation and age-related disease. The review concedes that clinical translation has been poor: two company-led senolytic programs failed in Phase 2, and the field lacks validated biomarkers to show a drug has hit its target.

In 1961, Leonard Hayflick and Paul Moorhead showed that normal human cells in a dish divide a limited number of times and then stop for good. That observation overturned fifty years of belief that cultured cells were immortal. A new review in The EMBO Journal follows what happened to the idea over the next six decades, and the story is one of repeated redefinition.

First came a mechanism. In 1990, shortening telomeres were proposed as the counting device behind the division limit. Then came a purpose. In 1997, researchers found that switching on a cancer gene pushed normal cells into the same arrested state, which recast senescence as a defense against tumors. Radiation, chemotherapy and oxidative stress turned out to do the same thing.

The third shift mattered most for aging research. Senescent cells do not sit quietly. They release a mix of inflammatory signals, growth factors and tissue-digesting enzymes known as the senescence-associated secretory phenotype, or SASP. This secretion explains how a small number of arrested cells can disturb a whole tissue, and why the same process can stop a tumor in one setting and feed one in another.

The causal evidence arrived in 2011 and 2016, when engineered mice allowed researchers to kill cells expressing the senescence marker p16. Clearing those cells delayed age-related disease and, in one study, lengthened life. Drugs followed in 2015 with the combination of dasatinib and quercetin, and later with navitoclax and others.

The review gives equal weight to the awkward findings. Senescent cells help close wounds. They limit liver scarring. They appear on schedule in the embryo and are cleared by immune cells. One mouse study found that removing p16-expressing cells damaged the liver. The working model is now one of balance: senescence is helpful when brief and harmful when the cells linger, either because damage keeps producing them or because an aging immune system stops removing them. Senescent cells can also hide from immune attack using a surface molecule called HLA-E.

On treatments, the authors describe two routes. Senolytics kill senescent cells. Senomorphics, which include rapamycin, metformin and JAK inhibitors, quiet the secretions without killing the cells.

The human record so far is weak, and the review says so. Early trials of dasatinib and quercetin showed the approach was feasible. But the osteoarthritis senolytic UBX0101 failed in Phase 2, and UBX1325 for diabetic eye disease missed its main goal in Phase 2b. The authors blame the absence of reliable biomarkers. Without them, nobody can confirm that a drug cleared senescent cells in a person.

Readers should know that the senior author holds equity or board positions in three senescence-targeting companies.

Actionable Insights

This review offers no evidence that would justify self-treating with senolytics, and it reports no effect sizes for any human benefit.

  1. No senolytic has shown clinical efficacy in a controlled human trial. The two programs that reached Phase 2 efficacy testing failed.
  2. The mouse results are real but modest in translation value. From the original studies (not this review, and from my recollection), genetic clearance of p16-positive cells raised median lifespan by roughly 24 to 27 percent, and late-life dasatinib plus quercetin raised remaining lifespan by about 36 percent. A gain in remaining lifespan from old age is a much smaller share of total lifespan.
  3. Senescent cells do useful work. Removing them slowed wound healing in mice. Senolytic dosing around surgery, injury or active wound healing carries a plausible risk.
  4. You cannot currently measure your own senescent-cell burden in a meaningful way. Commercial “senescence” panels rely on markers the field itself calls insufficiently specific.
  5. Rapamycin and metformin reduce senescent-cell secretions in the laboratory. Human trials of these drugs did not measure senescent cells, so that mechanism is unconfirmed in people.

Context/Source

  • Open Access Paper: Cellular senescence: six decades of discovery and reinvention
  • Institution: European Research Institute for the Biology of Ageing (ERIBA), University of Groningen and University Medical Center Groningen
  • Country: Netherlands
  • Journal: The EMBO Journal, published 2 October 2026
  • Competing interests: Demaria has board membership and equity in Rubedo Life Sciences, a board and advisory role at Oisin Biotechnologies, and equity in Cleara Biotech
    Impact evaluation: The impact score of this journal is 8.3, evaluated against a typical high-end range of 0–60+ for top general science, therefore this is a High impact journal.
1 Like
#2

HTwdnPSaUAATPbl
HTwdnPSaUAATPbl1312×1199 376 KB

Source: This post on x.

#3

There was a paper in 2021

https://www.nature.com/articles/s43587-021-00105-8

Which highlighted to me the role of citrate efflux from the mitochondria and I think explained what is happening with senescence.

I think a large proportion of senescent cells are cells which came from stem cells, but stuck on the way towards turning into somatic cells. (In the same way as the osteoblasts in the cited paper got stuck).

#4

You can then identify which diseases of ageing are as a result of an increase in senescence.

agebaseddiseases
agebaseddiseases1500×1400 334 KB

1 Like
#5

Clocks Built to Detect Senescence Fail to Budge After Senolytics

Yale researchers asked why epigenetic clocks do not fall after senolytic treatment, and tested whether clocks built specifically from senescence-related DNA methylation sites would do better. They found that only 9,363 of 396,333 methylation sites (2.4 percent) change in the same direction with cell-culture senescence, with age and with mortality risk. From those sites they trained three new “senescence-enriched” clocks. None declined after senolytic treatment in cultured cells, in mouse muscle, or in the blood of 19 people given dasatinib plus quercetin, with or without fisetin, for six months. One score rose at three months. The senolytic datasets were small and had no placebo arm, and none measured whether senescent cells were actually cleared.

Two of the most prominent ideas in aging science do not seem to be on speaking terms. Epigenetic clocks, which estimate biological age from chemical tags on DNA, predict who will get sick and die. Senescent cells, the damaged cells that stop dividing and linger, are thought to drive that same decline. Yet earlier studies found that clocks respond erratically to senescence, and do not fall when people take senolytics, the drugs designed to kill those cells.

A team at Yale proposed a simple explanation: perhaps existing clocks are built from the wrong sites. Clocks are trained to track age or mortality without regard to the underlying biology, so only a sliver of their signal might relate to senescence. A clock built only from senescence-linked sites might succeed where others failed.

To find those sites, the researchers pooled public data on human cells driven into senescence in three ways: by repeated division, by radiation or chemotherapy damage, and by an activated cancer gene. They removed sites that simply track cell division. About 37,800 sites changed consistently with senescence.

Then came the first surprise. When the team checked those sites in human blood, more than half moved in the opposite direction with age. Only 9,363 sites shifted the same way with senescence, with aging and with higher death risk. That is 2.4 percent of the sites examined.

From this shortlist the group trained three clocks, one to recognize senescent cells in culture, one to predict age and one to predict mortality. All three tracked age and predicted death in an independent sample. The mortality version performed best: each standard step up in its score went with roughly double the rate of death.

The second surprise was the test that mattered. The team applied the clocks to cells treated with senolytic compounds, to muscle from old mice given an experimental senolytic, and to blood from two small clinic-based trials in which 19 adults took dasatinib and quercetin, or that pair plus fisetin, three days a month for six months. None of the clocks went down. If anything they drifted up, and the culture-trained score rose significantly three months into dasatinib and quercetin.

The authors offer several explanations. The drugs may harm healthy cells while killing senescent ones. Surviving cells may divide to fill the gap, which itself shifts methylation. Or the methylation signal may come from healthy cells reacting to senescent neighbors, and may not reset quickly.

There is a plainer possibility that the paper cannot exclude. No dataset confirmed that senescent cells were removed in the people treated. The human studies had no placebo group. The senescence signature came from skin and bone marrow cells in a dish, then was read in blood. And the culture-trained clock recognized only radiation-induced senescence in independent data.

The authors conclude that methylation clocks may be poor endpoints for senolytic trials. That is a fair reading. The data say much less about whether senolytics work.

Actionable Insights

This study tested measurement tools, not a health outcome. It offers no evidence for or against taking senolytics.

  • Epigenetic age tests cannot currently tell you whether a senolytic “worked”. Neither standard clocks nor these purpose-built ones fell after treatment. A flat or higher score after dasatinib and quercetin is not proof of failure or harm, and a lower score would not be proof of benefit.
  • Size of the senolytic effect on the clocks. Zero or slightly upward. Of 18 comparisons, one reached significance: a rise of roughly half a score unit in the culture-trained clock at three months, which had partly faded by six months. With that many comparisons and no correction, one such result is expected by chance.
  • What the clocks do predict. In about 10 years of Framingham follow-up, each one-standard-deviation rise in the scores went with a 45 to 117 percent higher death rate (hazard ratios 1.45, 1.52 and 2.17). That describes risk prediction, not the effect of any treatment.
  • No health outcomes. The human data include no measures of function, disease or senescent cell burden.
  • Dasatinib is a prescription leukemia drug with real toxicity. The trials here were uncontrolled, with 19 participants each.
  • Bottom line for self-experimenters. Paying for before-and-after clock tests around a senolytic course is unlikely to be informative. [Confidence: Medium to High]

Context and Source

  • Full title: DNA Methylation Signatures of Cellular Senescence Are Not Reversed by Senolytic Treatment
  • Institution: Yale University and Yale School of Medicine, with the Institute for Hormonal Balance (Orlando) and TruDiagnostic (Lexington, Kentucky)
  • Country: United States
  • Journal: Aging Cell, 2026, volume 25, article e70430. DOI 10.1111/acel.70430
  • Funding: National Institute on Aging (1R01AG065403) and an NIH training grant
  • Competing interests: The senior author and one co-author report consulting fees from TruDiagnostic and are named inventors on an epigenetic clock patent, described as unrelated. Two co-authors are affiliated with TruDiagnostic, which sells epigenetic age tests and supplied the human senolytic data. The finding runs against the commercial interest in clocks as intervention readouts, which counts in its favor.
  • Impact evaluation: The impact score of this journal is 7.7 (Journal Impact Factor, 2025 citation data), evaluated against a typical high-end range of 0 to 60+ for top general science, therefore this is a Medium impact journal.

Senescence versus aging. More than half of senescence-linked sites move opposite to aging in blood. This supports the view that in vitro senescence and organismal epigenetic aging are largely distinct processes. [Confidence: Medium] It may equally reflect a tissue mismatch between cultured fibroblasts and blood. [Confidence: Medium]

Why no reversal: the authors’ hypotheses, all untested

  • Some senescent cells resist the specific drugs.
  • Senolytics damage non-senescent cells, offsetting any gain.
  • Surviving cells proliferate to compensate, which shifts methylation.
  • The signal comes from non-senescent cells responding to senescent neighbors and does not reset within months.
1 Like