The immune cell that multiplies after 100, in a week spent at the far edge of life

The cells that arrive late

Almost everything we know about the ageing immune system is a story of decline. This week a team in Japan reported the opposite, in the oldest people alive.

Researchers led by Kosuke Hashimoto at the University of Osaka looked at killer T cells [1]. These are immune cells that do not just raise the alarm. They destroy infected and cancerous cells themselves [1]. The work was published in Cell Reports on Wednesday [1]. In most people these particular cells, called CD4 cytotoxic T lymphocytes, make up under 5% of all T cells [1].

In the study’s youngest group, aged 70 to 90, they were 4% [1]. In centenarians, people aged 100 to 109, they were around 10%. In supercentenarians - people past 110 - they were around 18% [1][2]. The cells were not just present. They had been copying themselves in response to specific threats - what an immune system does when it is still adapting, not merely fading [1].

“Most of immune ageing research has focused on decline,” Hashimoto said. “Our study suggests that even at extreme old age, the immune system may still selectively adapt” [1].

The caveat is the size. The whole study was 28 people - eight aged 70 to 90, ten centenarians, ten supercentenarians - and about 40,000 of their immune cells [1][2]. Nobody has shown these cells cause the long life rather than accompany it, and the authors say they are not yet sure what the cells actually do [1].

Only about 150 people in Japan are old enough to ask

Hashimoto’s own explanation for the small numbers is blunt: supercentenarians are hard to study because so few exist. In all of Japan there are roughly 150 people aged 110 or over [1].

That thinness runs through the whole question. The record for a human life belongs to Jeanne Calment, who died in 1997 at 122 years and 164 days [3]. Even that is disputed. Some researchers argue the woman who died was her daughter [3]. The second-longest verified life was Kane Tanaka’s, 119 years and 107 days. The oldest person alive has just turned 117 [3].

So the estimates of how long a person could live disagree wildly. One demographic analysis put the ceiling near 115 while not ruling out 125 [3]. Another model tracked short-term swings in the blood counts of people over 85. It put the collapse of the body’s ability to recover at about 120, and a hard limit near 150 [3]. A Moscow team ran a thought experiment - cure every cause of ageing except the DNA damage that accumulates in cells - and got 194 [3]. The gap between those answers is not mostly a disagreement about biology. It is what happens when you estimate from a handful of people.

The ageing clocks go on trial

The field’s workaround is a clock - an estimate of biological age read from chemical tags on your DNA. It lets a trial finish in a couple of years instead of forty. On Friday Nature Medicine published the first serious audit of whether those clocks actually move when you try to slow ageing [4].

The team assembled a database of 51 studies that tracked people through an intervention, then ran 16 well-known ageing clocks plus 94 other DNA-tag measures across all of them [4]. Clocks trained to predict death or the pace of ageing responded most strongly, and agreed with one another [4]. Drug and lifestyle changes moved the markers most [4]. And how long a study ran, and who was in it, largely decided whether anything showed up at all [4].

Alongside it, Nature Medicine ran a commentary by Steve Horvath of UCLA, who built two of the best-known clocks. He discloses patents on both, a role at the foundation that licenses them, and equity in the anti-ageing company Altos Labs [5].

”A beautiful mouse story, sold as human advice”

A separate longevity claim got a sharp public correction. A review from the University of Wisconsin-Madison, drawing on 350 earlier papers, argued that cutting dietary protein promotes metabolic health and a longer life [6].

Stuart Phillips, a muscle researcher at McMaster University, called it “a beautiful mouse story, sold as human advice” [6]. In people over 65, he said, low protein is tied to worse outcomes, not better: “More frailty. More falls. Higher mortality” [6]. Much of the evidence the review leaned on came from animals, and its own authors acknowledge that the effect of protein restriction on human lifespan is unknown [6].

A brain that grew for five years in a dish

Ageing also showed up somewhere new. In Nature, Paola Arlotta’s group at Harvard reported growing brain organoids for more than five years [8]. An organoid is a pea-sized clump of brain tissue grown from stem cells. These are the longest-lived lab-grown organs so far [8].

They grew 34 of them and sampled every three to six months, then yearly [8]. At 15 days to two months the gene activity matched a first-trimester fetal brain; at three to six months, the second trimester; by 12 months, a newborn’s [8]. Ageing clocks trained on real people said the tissue’s molecular age matched its actual age, meaning it developed at a human pace despite sitting in a dish [10]. Neurons wired up and kept firing for at least two years [10].

The oddest finding: cells remembered how old they were. Old cells taken from year-old organoids and mixed with 15-day-old cells did not revert to making early neurons - they carried on producing the mature types [8]. These are still clumps of tissue, not brains. But the NIH, which funded the work, says the length is the point [10]. Conditions like autism unfold over years, and until now the models died in months [10][9].

In mice, meanwhile, a Nature Aging paper looked at worn-out cells - cells that stop dividing but refuse to die, and leak inflammation. Blocking a pair of proteins in them cut DNA damage and inflammation in old mouse livers. Palbociclib, an approved breast-cancer drug, reduced frailty and improved physical performance in aged mice [7]. Mice, not people.

At the other end of medicine, the same thin numbers

The week’s approvals came from the rarest corner of the field, where patient counts are tiny for a different reason.

On Wednesday the FDA, the US drug regulator, approved Regeneron’s Pasatru for fibrodysplasia ossificans progressiva, a disease that grows bone where bone should not be [11]. People with it typically use wheelchairs by 25, and only some live into their 50s. It capped roughly three decades of work. “You’re almost completely stopping the new bone forming,” said Richard Keen of London’s Royal National Orthopaedic Hospital, who ran the pivotal trial [11].

A day later Ultragenyx won the first approval for a treatment for glycogen storage disease type Ia [12]. It is an inherited fault that leaves people unable to hold their blood sugar steady between meals [12]. It is a one-time gene therapy priced at $2.7 million per patient, and it came with a priority review voucher - a regulatory fast pass companies can sell. Three sold this year for $180 million, $195 million and $200 million [12].

Not everything cleared. Capricor’s cell therapy for Duchenne muscular dystrophy is heading for rejection [14]. The clean result sat in the smaller group of boys who already had signs of heart damage [13]. Across the full trial, which also included boys whose hearts were still fine, the data turned noisy [13]. A parent of a trial participant, writing in STAT, described designing a rare-disease study as “hundreds of tiny decisions” made without enough prior data to know which is right [13].

Two deals followed the same map. BioMarin paid $275 million for a drug for hypophosphatasia, a rare bone-softening disorder [15]. Leo Pharma committed up to $435 million for Tanabe’s pill for two rare conditions that make sunlight painful [16]. And STAT asked gene-editing specialists to review video of a presentation Huidagene gave three months before a boy died in its trial last year [17]. They split sharply on both the science and the judgement behind it [17].