Biotech & Longevity · Thursday, 3 September 2026

01 Briefing what happened

Old mice on a weight-loss drug lived 12% longer. Mice simply fed less got most of the same.

Biotech & Longevity 1 min 40 sources

Researchers in the United States gave semaglutide to healthy mice at an age close to a 60-year-old person. The treated mice lived a median 834 days against 742. Mice fed exactly as little as the drugged ones chose to eat lived about as long -- but did worse on memory and blood sugar.

834 vs 742 days

median life of mice on the drug, against mice on nothing

treatment started at 20 months old, which is roughly a 60-year-old person [1][2]

24%

less food the mice on the drug chose to eat

a second group was fed exactly that much, to test whether the eating was the whole story [1]

11%

of Americans now taking a drug of this class

they are approved for diabetes and obesity, and this study is about neither [3]

The lead story — what happened

Researchers at the University of California, Berkeley gave the drug semaglutide to healthy 20-month-old female mice, an age close to a 60-year-old person. [1][2]
Semaglutide is the medicine sold as Ozempic and Wegovy. It copies a gut hormone that tells the brain you have eaten. [1]
Treated mice had a median lifespan of 834 days. Untreated mice had 742 days, so the gain was about 12%. [1]
The treated mice also did better on tests of coordination, muscle strength and blood sugar after three months. [2][3]
Signs of ageing eased too: less inflammation, and less of the usual loss of the body's power to repair itself. [2][3]
The mice on the drug chose to eat 24% less, which is the obvious explanation for all of it. [1][2]
So the team added a group of untreated mice fed exactly that much and nothing more. [1]
The two groups came out similar on how long they lived, so most of the extra life is the eating less. [1]
But the drugged mice beat the dieting mice on spatial memory, on exploring, and on holding blood sugar steady. [1][2]
Resting metabolic rate, which is how fast a body burns fuel while doing nothing, fell in the dieting mice and stayed roughly level in the drugged ones. [2][3]
The lead author says that points to a second route the drug takes, and that nobody can yet name it. [1][2]
The whole result is in female mice of one inbred laboratory strain, treated for three to five months. No drug is approved anywhere for ageing. [3][4][40]

Who is involved

Danica Chen

a professor of metabolic biology and nutrition at the University of California, Berkeley; she led the study and says the extra benefits point to a route that is not just eating less
Rafael de Cabo

a senior investigator at the US National Institute on Aging, the American state body that paid for the work; he wrote the commentary published beside it
Michael Corley

an ageing scientist at the University of California, San Diego, who had no part in the study; he calls it rigorous and the finding preliminary

What is pushing on this

The hunt for a pill for ageing High↑

one ageing scientist says these drugs are top of everyone's list of candidates [1]

The eating-less explanation High↓

matched-diet mice got most of the lifespan gain, but not the memory or the blood sugar [1][2]

Distance from a human answer High→

female mice of a single inbred strain, and human ageing trials are only starting [1][40]

How it unfolded

For decades cutting calories has been shown to extend life across many species [4]
20 months the mice start the drug, at roughly the human equivalent of sixty [1]
3 months in treated mice test better on muscle, coordination and blood sugar [2]
5 months in the matched-diet comparison ends, with the drugged mice ahead on memory and exploring [2]
2 September the study is published in the journal Nature [4]
Next human trials of these drugs aimed at ageing itself are starting [1]

Where this points

The next real test is a human trial that measures ageing rather than one disease at a time, and until one reports, every claim here rests on mice.

The rest of the day

32 more stories on this beat.

Each with its own sources. None of these is a link to the story above.

02
Cell-therapy rivals rush to look different

Three people died after a severe immune reaction in Novartis trials of its cell therapy rap-cel, and Bristol Myers Squibb reported inflammatory events in its own. Shares in Kyverna, Cabaletta, Allogene, CRISPR Therapeutics and Fate fell sharply before recovering. [5]

Why it matters — Analysts have settled on the faster method both firms used to grow the cells as the likely cause, and slower rivals are saying so loudly.
03
Lilly buys an immune-drug startup

Eli Lilly agreed to pay up to $2.875bn for Merida Biosciences, a Massachusetts startup whose drugs tag and destroy the specific antibodies that attack a patient's own body. It is Lilly's 13th purchase of a smaller biotech this year. [6][7]

Why it matters — Instead of switching the whole immune system down, this removes one misfiring part of it -- and it is obesity money paying for the shift.
04
Angelman syndrome drug fails

Ultragenyx said on 2 September that GTX-102 gave no benefit over a dummy treatment in a large final-stage trial in Angelman syndrome, a rare condition causing severe intellectual disability. Early trials had looked strong. [8]

Why it matters — Families had hoped it would be the first drug to improve thinking and communication in an intellectual disability, and investors had treated it as the firm's route to profit.
05
A celiac drug clears its first real test

Teva said its antibody TEV-408 significantly reduced gluten-caused damage to the gut lining, against a dummy, over eight weeks in a mid-stage trial. The same drug showed early promise in the skin condition vitiligo. [9]

Why it matters — Celiac disease has no approved medicine at all; the only treatment is never eating gluten again.
06
The pancreatic drug reaches lung tumours

Tumours shrank in more than 30% of people with the commonest form of lung cancer in a small trial of daraxonrasib, published on 2 September. All had a mutation in the RAS family of proteins and had already tried other treatments. [10]

Why it matters — The US regulator approved the same drug for pancreatic cancer a week ago; some tumours in this trial started growing again, and a larger randomised trial has to report first.
07
Morning infusions, longer survival

Patients given immune-system cancer drugs earlier in the day lived a median 21.4 months, against 13.1 months for those treated later, in a look back at 2,631 patients at one hospital in Manchester, England. [11]

Why it matters — The senior author says timing alone cannot be blamed -- sicker patients may simply get later slots -- and randomised trials are planned.
08
Nine more drugmakers take the price deal

Alcon, Astellas, BeOne, BridgeBio, CSL, Kyowa Kirin, Sun Pharma, Teva and UCB agreed to tie their US Medicaid prices to the lowest paid in other countries, in exchange for relief from US import tariffs. [12]

Why it matters — Medicaid already gets steep discounts by law, the terms were not published, and the deals leave out private insurance and the far bigger Medicare programme.
09
Doctors' groups write their own vaccine advice

Four US medical societies published joint flu, Covid and RSV vaccine guidance on 2 September. The US disease agency's expert advisory panel was dismissed in 2025 and its replacement is stuck in a court fight. [13][14]

Why it matters — The recommendations largely match the ones in place before, which is the point: the advice did not change, the body giving it did.
10
Cancer prices above $400,000 are normal now

Nine cancer drugs approved in the United States in 2024 launched above $400,000 a year, on a count by the Institute for Clinical and Economic Review. One cell therapy is listed at $600,000 for a single infusion. [15]

Why it matters — Twenty years ago even the prospect of a $100,000-a-year cancer drug drew fierce criticism. [15] Analysts now call the newest launch possibly the fastest in the history of cancer medicine. [16]
11
Every letter of a genome, changed

Researchers built more than 44,000 versions of the virus phi X174, altering nearly every one of its 5,386 DNA letters, and measured what each change did. One of the study's leaders says that even here they cannot explain why a quarter of the mutations kill the virus. [17]

Why it matters — The leading AI tools failed to predict them, in the most exhaustively studied small genome there is.
12
Thirty-six genes behind OCD and tics

Reading the protein-coding DNA of 3,964 people with obsessive-compulsive disorder or long-running tic disorders turned up 36 high-confidence risk genes, against four before. [18]

Why it matters — The two conditions affect one to two people in a hundred and have almost nothing aimed at their causes.
13
A stroke drug misses its mark

Adding the clot drug tirofiban to aspirin did not significantly cut worsening or new strokes in a trial of one narrow stroke type. The bad outcome hit 79 patients on the drug and 89 on the dummy. [19]

Why it matters — Patients whose feeding artery was more than 30% narrowed did appear to gain, which is the kind of hint that needs its own trial rather than a headline.
14
A lymphoma trial closed for excess deaths

Adding venetoclax to standard chemotherapy in newly diagnosed double-hit lymphoma killed more patients than it saved: six deaths on treatment out of 37, against one out of 36, and 52% alive at two years against 72%. The cohort was shut early. [20]

Why it matters — The drug is approved and effective elsewhere in blood cancer, which is how a sensible-looking combination gets built.
15
Making cell therapy at the hospital

A CD19 cell therapy grown locally in five hospitals in the Netherlands and Belgium, rather than shipped to a central factory, was tested in 24 people with lymphoma. Four died of treatment-related causes and a dose was chosen for the next stage. [21]

Why it matters — Central manufacturing is what makes these treatments slow and costly, and it is the same production question now hanging over the paused autoimmune trials.
16
Protein shells built to carry RNA

Researchers made more than 100 delivery containers from AI-designed protein assemblies, instead of copying a virus, and used them to move RNA into cells. [22]

Why it matters — Almost every gene therapy today borrows a virus's shell, and inherits the immune reactions that come with it.
17
Closing in on who gets long COVID

A Nature feature on 2 September sets out where six years of long COVID research now stands. One 2024 estimate puts the number affected worldwide at 400 million, at a yearly economic cost near $1 trillion. [23]

Why it matters — It sits alongside older illnesses that follow an infection, which medicine spent decades declining to believe in.
18
Alzheimer's blood tests hold up in Nigeria

Blood markers of Alzheimer's rose step by step with disease severity in 967 older Nigerian adults, and did so on two different testing platforms. [24]

Why it matters — Nearly all the evidence for these tests comes from wealthy countries, and a test never checked elsewhere has an unknown floor.
19
Twenty women, and a hormone for brain fog

Twelve months of estriol and progesterone improved self-reported brain fog, memory and processing speed in a case series of twenty menopausal women, average age 53.5. There was no comparison group. [25]

Why it matters — About 62% of women going through menopause report brain fog, and the evidence base for treating it is close to empty.
20
Ageing is now the largest share of illness

Diseases of ageing are the biggest part of the world's disease burden and the largest lifetime burden facing a newborn, even in low-income countries, on an analysis of Global Burden of Disease data. [26]

Why it matters — The authors argue the gains compound: each reduction makes the next one worth more, which is an argument for funding ageing rather than one disease at a time.
21
Two years of antibodies, measured

Across 25,800 adults in Japan, antibody levels after a third Covid dose fell to about a tenth of their peak by two years. Infection on top of vaccination held levels higher for longer. [27]

Why it matters — Booster schedules are set from data like this, and almost all of it has stopped at one year.
22
What a pandemic costs, by strategy

Modelling of six invented respiratory pandemics puts the total loss -- health, money and lost schooling -- between 1.9% and 439% of a country's yearly output, depending on the disease and the country's income. [28]

Why it matters — Closures triggered by the outbreak generally beat long fixed school closures, and the spread of that range is the argument for preparing in advance.
23
AI put on spillover watch in Uganda

A Ugandan non-profit that has protected the mountain gorillas of Bwindi from human diseases since 2003 is moving to AI models that predict outbreaks rather than react to them. [29]

Why it matters — The park's edge is where 120 mammal species, livestock and people meet, and that meeting is where new human diseases usually begin.
24
Sixty-nine billion molecules, docked

An open platform called AdaptiveFlow screened a library of 69 billion buyable drug-like molecules on up to 5.6 million processors, and found strong blockers of two disease targets. [30]

Why it matters — Screening at this size has belonged to a handful of large companies; the code and the library are now public.
25
Mirror-image drugs, designed on a computer

A workflow called Mirror-Peptidizer designed mirror-image protein fragments that stick to three cancer and immune targets, without having to chemically build the mirrored target first. [31]

Why it matters — Mirror-image fragments survive the body's protein-chopping enzymes, which is why ordinary ones so often fail as drugs.
26
A trials company buys a first-in-human unit

Fortrea paid $45m for Worldwide Clinical Trials' early-phase business, including a 200-bed clinic and a laboratory, both in Texas. [32]

Why it matters — The buyer wants fewer handovers between the ward running a first-in-human dose and the laboratory reading its blood samples.
27
A gene therapy aimed at diabetes

Genprex hired Andelyn Biosciences to scale up manufacturing of a gene therapy delivered into the pancreas through its own duct, meant to turn one kind of pancreatic cell into an insulin-making one. [33]

Why it matters — The programme is aiming at trials rather than in them, and this deal is about being able to make enough of it to try.
28
Sixteen countries, four rulebooks

An analysis across sixteen countries found that work combining synthetic biology, AI and laboratory robots falls between biosecurity, AI, export-control and data rules, none of which were written for it. The authors say the friction multiplies rather than adds. [34]

Why it matters — A project can be legal in each country separately and impossible between them.
29
$120m for a hepatitis B therapy

AusperBio raised $120m to push AHB-137, a drug that switches off a hepatitis B protein, through late-stage testing. The trial is running in China. [35]

Why it matters — Chronic hepatitis B infects nearly 296 million people and has no cure, only suppression.
30
Proteins that switch on where it hurts

Chemists at Peking University built proteins that stay inactive until they meet nitric oxide, a gas the body releases at sites of inflammation. [36]

Why it matters — A drug that only wakes up where the disease is does not have to be aimed at it.
31
Choosing the screen before the screening

A framework called MolDockLab tested five docking programs, 15 scoring methods and three ranking strategies against about 200 compounds with known activity, then picked the best combination for that target. [37]

Why it matters — The answer a computer screen gives depends on which tools were chosen, and that choice is usually made by habit.
32
US regulators warn on foreign-only trials

Four senior US drug and device regulators wrote on 2 September that trials run with few or no American patients are harder to inspect without warning, harder to apply to Americans, and increasingly a worry for lawmakers. [38]

Why it matters — Where a trial is run is turning into a regulatory question rather than only a cost one.
33
Doctors asked about ageing they never studied

Dean Ariel, a family doctor writing in the journal Nature Aging, says patients now arrive with questions about ageing medicine that clinicians were never trained to answer, and argues the profession should prepare rather than improvise. [39]

Why it matters — Today's mouse result is exactly the kind of finding that produces those questions in a waiting room.

02 Lesson why it matters

What a remedy asks for decides who it reaches

Eating less has extended animal lives for decades, and almost nobody can keep it up, so the search moved to a version that asks for nothing.

The twist

How much good a remedy does in the world is its effect multiplied by how many people can keep doing it, and the second number is nearly always the one that decides.

How it works

Eating about a quarter less extends life in animals
But it has to be done every day, for years
So almost nobody sustains it, and the benefit stays in the laboratory
The search shifts from a bigger effect to a smaller demand
A weekly injection asks for almost nothing, so it can reach people the evidence never did

Where you've seen this

Tooth decay

fluoride in the water reached the people that telling everyone to brush never did

Car safety

an airbag that fires by itself protects the driver who forgot the belt

Income tax

money taken from the payslip is collected from people who would never file a form

Computer security

an update that installs itself reaches the machines whose owner never clicks it

The catch

Removing the demand does not delete it. It moves onto a factory, a price and a list of side effects -- and here it swaps ninety years of evidence for five months of mice.

And the whole of it

The reader is already inside several of these: the water, the payslip, the update that installed itself last night. Each one works on people who were never going to do it themselves, which is most people most of the time, and none of them feels like a decision anyone made.

03 Truth what's really going on

What is really going on

Cutting calories has extended life in animals for decades and hardly anyone can sustain it, so ageing research has largely become a hunt for a drug that copies the result without the discipline -- and in mice, one nearly does.

Why it works on us — A lifespan number is the rarest thing in ageing research and the easiest to carry away, so 12% longer travels and 20-month-old female mice of one inbred strain does not.

Who gains

The sellers of Ozempic and Wegovy — Every new effect found for this class widens the market for whoever sells it, and an estimated 11% of Americans already take one of these drugs. [1][3]
Eli Lilly — Its obesity sales paid for a 13th biotech purchase this year, at up to $2.875bn, moving its pipeline away from the drugs that funded it. [6][7]
Kyverna, Cabaletta and Autolus — Their rivals' trial halts let them argue in public that their slower way of growing cells is the safer one. [5]
The nine drugmakers in the US pricing deal — They were spared import tariffs in exchange for discounts on Medicaid, which is a small share of most of their sales. [12]
Revolution Medicines — Its cancer drug now has lung-cancer results a week after its pancreatic approval, and analysts are calling the launch possibly the fastest in cancer medicine. [10][16]

Who pays

Families living with Angelman syndrome — A drug that had looked strong in early trials showed no benefit in the large one, and nothing else is close to the clinic. [8]
People in the paused autoimmune cell-therapy trials — Three died after a severe immune reaction, and enrolment in lupus, myasthenia gravis and multiple sclerosis studies is on hold. [5]
Patients in the double-hit lymphoma trial — Adding venetoclax to standard chemotherapy killed six of 37 on treatment against one of 36, and the cohort was closed early. [20]
Americans with cancer, and whoever insures them — Nine drugs approved in 2024 launched above $400,000 a year, and the bill lands on US state health programmes, employers and patients. [15]
People in trials run entirely outside the United States — US regulators say they cannot inspect foreign sites the same way, including without warning, and the host country may not have the capacity to do it either. [38]

What nobody knows yet

Open questions from across today’s stories — ours included.

01

What the drugged mice got that the dieting mice did not.

Both groups ate the same amount and lived about as long, yet the drugged ones did better on memory, exploring and blood sugar, and the lead author says nobody can name the cause. [1][2]
02

Whether any of this happens in a person.

Every result is in female mice of a single inbred laboratory strain, and human trials aimed at ageing itself are only starting. [1][40]
03

What happens in male mice.

The paper is titled for female mice and reports no male group, so nothing here says what the drug does to a male. [4][40]
04

What killed the three people in the Novartis cell-therapy trials.

The company has said only that they died after a severe immune reaction. Analysts suspect the faster method used to grow the cells, and no finding has been published. [5]
05

How much the nine new US price deals actually save.

Medicaid already gets large discounts required by a 1990 law, and neither the White House nor the companies published the terms. [12]
06

Whether the time of day really changes cancer survival.

The Manchester study found 21.4 months against 13.1, but it looked backwards at existing records, and its senior author said timing alone cannot be held responsible. [11]
07

Why a quarter of the lethal mutations in the most-studied small genome are lethal.

Researchers changed nearly every letter of phi X174 and still cannot explain them, and the AI tools built to predict exactly this could not either. [17]
08

What Ultragenyx does now with its Angelman drug.

The company said only that the final-stage trial showed no benefit against a sham, and has not said whether the programme continues. [8]

04 Hope carry this

In 1973 a person born anywhere on Earth could expect to live about 58 years. Fifty years later it is more than 73, and it went up in every country.