Biotech & Longevity · Friday, 4 September 2026

01 Briefing what happened

These drugs block nothing. They hold two proteins together and let the cell do the rest.

Biotech & Longevity 1 min 35 sources

The first medicine that destroys a protein instead of blocking it was approved this year, and Johnson & Johnson paid $3.05bn for a company with no approved drug at all.

$3.05bn

paid by Johnson & Johnson for Halda Therapeutics, which had no approved drug

It had one medicine in an early trial when the deal was struck in November 2025 [1]

43%

lower risk of the breast cancer worsening or the patient dying

the first approved protein-destroying drug, measured against the older drug fulvestrant [1]

20+ years

from the idea to the first approved medicine

Craig Crews proposed it at Yale; approval came on 1 May 2026 [1][2]

30+

protein-destroying drugs now being tested in people

one is approved, and the rest are still in trials [1]

The lead story — what happened

On 1 May the US drug regulator approved Veppanu, the first medicine that destroys a protein instead of blocking it. [2][1]
It was made by the biotech Arvinas with Pfizer, and it treats a form of advanced breast cancer after hormone treatment has stopped working. [2][1]
In its final-stage trial it cut the risk of the cancer worsening or the patient dying by 43%, against an older drug called fulvestrant. [1]
The drug has two ends. One end grips the faulty protein. The other grips the machinery a cell uses to shred proteins it has finished with. [1]
Holding the two together is the whole trick. The drug breaks nothing itself. The cell does the destroying. [1]
Craig Crews, a pharmacologist at Yale University in the United States, first proposed the idea more than twenty years ago. [1]
What he found was that two proteins that never evolved to touch each other can be forced to. That opened up the possibility of using induced proximity in many different ways, he says. [1]
More than 30 of these protein-destroying drugs are now being tested in people. [1]
Johnson & Johnson, one of the world's largest healthcare companies, paid $3.05bn in November 2025 for Halda Therapeutics, a Yale spin-out with no approved drug and one early trial. [1]
Halda's drug destroys nothing. It handcuffs a protein that prostate tumours carry in excess to BRD4, a protein cancer cells need to survive, and the pair kills the cell. [1]
In that early trial, 59% of men who took it for at least six weeks saw a marker of their prostate cancer fall by half or more. [1]
Thalidomide and drugs like it were later found to work this way by accident, as a glue between two proteins that would not otherwise meet. [1]
Nearly all of the molecular glues we know about were discovered by serendipity, says George Burslem, a biochemist at the University of Pennsylvania. Finding them on purpose is the open problem. [1]

Who is involved

Craig Crews

a pharmacologist at Yale University in the United States; he proposed drugs that destroy proteins more than twenty years ago and co-founded two companies built on the idea [1]
Arvinas and Pfizer

a small biotech and a large drugmaker; together they made Veppanu, the first protein-destroying drug to be approved [1][2]
Johnson & Johnson

one of the world's biggest healthcare companies; it bought Halda Therapeutics for $3.05bn before Halda had an approved drug [1]
Halda Therapeutics

a Yale spin-out; its prostate cancer drug handcuffs a tumour protein to one the cancer cell cannot live without [1]
George Burslem

a biochemist at the University of Pennsylvania; he says nearly every molecular glue known was found by accident [1]

What is pushing on this

Money moving in High↑

Johnson & Johnson paid $3.05bn for one early-stage company [1]

Proof it works in people Building↑

one approval so far, and about thirty drugs still in trials [1]

Finding the glues on purpose Steady↑

nearly all of the known ones were found by accident [1]

Reach beyond cancer Easing↑

an asthma drug holds a US fast-track label, and a Parkinson's programme is being explored [1]

How it unfolded

Over 20 years ago Craig Crews proposes that a drug could destroy a protein rather than block it [1]
Aug 2025 Halda's prostate drug gets a fast-track label from the US drug regulator [1]
Oct 2025 Halda reports that 59% of men treated for at least six weeks had their cancer marker halve [1]
Nov 2025 Johnson & Johnson buys Halda for $3.05bn [1]
1 May 2026 the US drug regulator approves Veppanu, the first protein-destroying medicine [2]

Where this points

Watch whether one of these drugs works outside cancer. An asthma drug from Kymera holds a US fast-track label, and that is the test of whether this is a cancer trick or a new way to make medicines. [1]

The rest of the day

35 more stories on this beat.

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

02
A Chinese drug beats the biggest cancer medicine

A new cut of trial data showed that ivonescimab, made by Akeso in China and licensed to Summit, helped people with lung cancer live longer than Merck's Keytruda. One analyst expects the full figures to show a 24% lower risk of death. Summit shares rose as much as 14%. [3]

Why it matters — Drugs like Keytruda transformed cancer care over the past decade, and drugmakers have largely failed to top them despite much trying. [3]
03
The same drug's other trial could not show it

In a 438-patient final-stage trial published on 1 September, ivonescimab plus chemotherapy held lung cancer back for 6.8 months against 4.4 on chemotherapy alone. People lived a median 16.8 months against 14.0, a gap the trial could not separate from chance. [4]

Why it matters — Two results on the same drug in one week, pointing different ways on the only question that finally matters.
04
GSK pays up to $1.3bn for a two-target cancer drug

GSK agreed on 3 September to pay Hutchmed $110m now and up to $1.19bn later for most rights to HMPL-A830. It is an antibody that finds the protein EGFR and carries a small drug that blocks another, KRAS. [5]

Why it matters — Both proteins drive common cancers, and a great many KRAS-driven tumours still have no safe, lasting treatment. [5]
05
Alzheimer's leaves a mark years before the scan

Researchers at the University of Oslo scanned healthy older adults for nearly twenty years, then looked back. Brain structure had already changed at least seven years before amyloid plaques showed up on the scan that is used to spot the disease earliest. [6]

Why it matters — Either something starts the plaques before they can be seen, or something else damages the brain first. The second would mean the drug hunt has been aimed too narrowly. [6]
06
The brain's attackers get their orders in the neck

Washington University researchers found that the immune cells which worsen tau damage in the brain are given their instructions by cells in lymph nodes outside it. Mice that lacked those instructing cells were protected from the damage. [7][8]

Why it matters — The harm is in the brain and the decision is made elsewhere, which is a place nobody had been treating. [8]
07
Block one receptor, and old mice stay young

Stanford researchers found that ageing macrophages, the body's clean-up cells, stop clearing the roughly 100 billion worn-out white blood cells made each day. Blocking one receptor on them, called EP2, kept the brain, heart, liver, kidney and muscle of old mice more youthful. [9]

Why it matters — It points at chronic inflammation as a cause of ageing rather than a symptom, though this was mice and human cells, not people. [9]
08
A lupus drug missed everything, and goes on

Alumis said on 1 September that envudeucitinib failed both its main and its backup measures in a 408-person mid-stage lupus trial. The company is taking it into final-stage testing anyway, on the strength of one group of patients picked out in advance. Its shares fell 54%. [10]

Why it matters — About 60% of the trial's patients were in that group, against the 80% seen in other trials, and the company plans to enrol mainly those people next time. [10]
09
Huntington's gene therapy is filed at last

uniQure asked the US drug regulator on 2 September to approve its gene therapy for Huntington's disease, an inherited and fatal brain disorder with no treatment for its cause. [11]

Why it matters — The regulator had earlier refused; the filing follows its reversal. [11]
10
Overdose deaths fell, and nobody can name one reason

US overdose deaths dropped 27% in 2024 and fell again in 2025. Scotland was down 13% and Canada 9%. A Commonwealth Fund report published on 3 September asked officials in more than two dozen places, and every one said no single thing did it. [12]

Why it matters — The things named repeatedly were unglamorous: wider naloxone supply, treatment in and around prisons, and better data. [12]
11
One gene edit, cholesterol still down a year on

Fifteen people with stubbornly high blood fats got a single infusion of CTX310, which uses CRISPR to switch off a liver gene called ANGPTL3. A year later the effect held: at the top dose, triglycerides down 48% and LDL cholesterol down 53%. [13][14]

Why it matters — It is a safety study of fifteen people, and switching a gene off permanently is not something you can undo if it turns out to be a mistake. [13]
12
A clot drug proved safe and prevented nothing

Milvexian, from Bristol Myers Squibb and Johnson & Johnson, was meant to stop dangerous clots without the bleeding that clot drugs cause. In about 14,000 patients treated within a week of a sudden blockage in a heart artery, it did not increase serious bleeding, and it did not cut heart attacks, strokes or deaths either: 5.4% against 5.1%. [13]

Why it matters — The trial was stopped in November 2025 once it was clear it would not work; the full data explains why. Two more trials of the same drug are still running. [13]
13
A heart gene therapy holds at twelve months

Ten people with heart failure got SRD-002, a one-off gene therapy that adds a copy of the gene for SERCA2a, a calcium pump that lets heart muscle relax after each beat. At a year, eight of ten met a pre-set target for normal filling pressure. [13]

Why it matters — The same target and delivery method failed a 250-patient trial in a different kind of heart failure, so ten patients with no comparison group settles nothing. [13]
14
A pregnancy hormone, rebuilt as a pill

AstraZeneca's AZD5462 copies relaxin, a hormone that widens blood vessels during pregnancy. In 375 heart failure patients it significantly lowered the resistance the heart pumps against, in those whose pumping was only mildly reduced. Another measure narrowly missed. [13]

Why it matters — An injected version of the same hormone failed a final-stage trial years ago; this is the same biology in a tablet. [13]
15
$60m for a pill for rare inherited obesity

Superluminal Medicines raised $60m to develop a daily tablet aimed at MC4R, a receptor that controls appetite and how much energy the body burns. Human testing should start by the end of the year. Nvidia and Eli Lilly joined the round. [15]

Why it matters — The existing drug for these rare genetic conditions is an injection, and several rivals aimed at the same problem have failed or been paused. [15]
16
Brain tumour cells turned into alarm cells

Researchers used CRISPR to switch on genes that reprogram glioblastoma cells into cells resembling dendritic cells, the immune system's messengers, so the body attacks the tumour. Glioblastoma is the deadliest primary brain tumour and survival after it returns is very poor. [16]

Why it matters — It changes the tumour into something that calls for help, rather than sending a drug in after it.
17
China writes one rulebook for new biology

China's State Council Order No. 818, a single framework for new biomedical technologies, came into force on 1 May. A Nature comment on 1 September set out how it works, noting the case of a six-year-old girl who died last year after an experimental gene-editing treatment aimed at her brain, in Shanghai. [17]

Why it matters — The death happened before the rules existed, and is the reason the authors say the rules are needed. [17]
18
Stiff tissue stops the brain making new cells

Peking University researchers measured the stiffness of the part of the mouse hippocampus that makes new brain cells. It stiffened early in Alzheimer's mice, and stiffness alone was enough to stop new cells forming. An enzyme that softened it again brought them back. [18]

Why it matters — The same stiffening was found in brain tissue from people who had died with Alzheimer's, which makes it more than a mouse finding. [18]
19
Tau in one tiny spot made mice anxious first

Researchers put damaged tau into the locus coeruleus, a small brain area that supplies the chemical noradrenaline, in mice. Three months later the mice were anxious and repetitive, and had fewer sleep spindles, the short bursts of brain activity that punctuate sleep. No brain cells had died. [19]

Why it matters — People with Alzheimer's often become anxious or agitated years before they start forgetting, and this is where that may begin. [19]
20
A brain scan reader built to show its working

A framework published on 3 September tries to make automated reading of Alzheimer's brain scans more trustworthy, by tying its answer to named brain regions rather than reading the image as a whole. [20]

Why it matters — The stages of the disease look alike on a scan, which is why these tools carry little confidence and travel badly to new data. [20]
21
A US spending bill blocks control of grants

The US House of Representatives voted 370 to 48 on 1 September for a stopgap bill funding the US federal government until 11 December. It temporarily blocks a plan by the Trump administration to give political appointees more say over federal research grants. [21]

Why it matters — The block expires on the same day the money does. [21]
22
A malaria project may have found a meat-allergy fix

Researchers at the US National Institutes of Health looking for a way to prevent malaria may instead have found a way to treat alpha-gal syndrome, the allergy to red meat caused by tick bites. The paper appeared in the Journal of Clinical Investigation. [21]

Why it matters — Nobody was looking for it, which is how a fair number of treatments arrive.
23
Making a blind eye light-sensitive again

A review published on 2 September set out where optogenetic sight restoration stands. Viruses carry light-sensitive proteins into the retinal nerve cells that survive after the light-detecting cells have died, and the clinical pipeline is growing. [22]

Why it matters — It does not care which gene caused the blindness, which matters once there are no light-detecting cells left to fix. [22]
24
Nine in ten people carry this virus for life

A reference review published on 3 September sets out what is known about varicella zoster virus, which infects more than 90% of people worldwide. It causes chickenpox, then hides in nerve cells for life, and can wake up decades later as shingles. [23]

Why it matters — Waking up is triggered by the immune system weakening with age or illness, which is why shingles is a disease of later life. [23]
25
A fifth of dementia is a blood-vessel disease

Dementia affects more than 55 million people and is expected to triple by 2050. About 20% of cases are vascular dementia, caused by injured blood vessels, a leaking barrier around the brain and inflammation together. [24]

Why it matters — It is the second most common cause after Alzheimer's, and it is caused by things that damage blood vessels anywhere. [24]
26
Midlife weight carries the highest brain risk

A review published on 1 September finds obesity across life is linked to cognitive decline, with obesity in middle age carrying a particularly high risk. It acts through the hypothalamus and through inflammation and blood-vessel damage in the hippocampus. [25]

Why it matters — The measured risk is not evenly spread across a life; it is concentrated in the middle of it. [25]
27
Sleep and stress named as changeable drivers

A review on 2 September argues sleep and chronic stress feed Alzheimer's directly. Sleep runs the clearing systems that remove waste protein from the brain, and those systems are impaired in the disease; long-term stress adds inflammation. [26]

Why it matters — Both are things that can change, unlike the genes that make up most of the known risk. [26]
28
The people who most need vaccines get fewest

A review on 3 September finds people with autoimmune joint diseases catch more preventable infections, because of the disease, the drugs that damp their immune systems, and age. Their coverage for routine flu and pneumonia vaccines remains low. [27]

Why it matters — The same drugs that raise their risk also make a vaccine work less well, which is why the timing has to be planned. [27]
29
Growing the immune system's own brakes

A review on 1 September collects the ways researchers try to make regulatory T cells, the immune cells that hold the rest back and stop the body attacking itself. The methods run from feeding them signals to engineering them. [28]

Why it matters — Most autoimmune treatment blunts the whole immune system; this would strengthen the part that already says stop. [28]
30
Cancer's off-switches turn up in diabetes

A review on 2 September finds that immune checkpoint proteins, the off-switches cancer drugs release to let the immune system attack tumours, also appear to shape obesity, diabetes and thyroid disease. [29]

Why it matters — Existing drugs already move these switches, which is why anyone is looking. [29]
31
The brake chemical inside immune cells

A review published on 4 September sets out what is known about itaconate, a chemical immune cells make that acts as a brake on inflammation and on damage from oxygen. It shows up across infections, autoimmune disease and cancer. [30]

Why it matters — It is made by the cell's own energy machinery, which ties how a cell burns fuel to how hard it inflames. [30]
32
An Ebola-family outbreak was spotted late

A paper in a Lancet journal on 3 September reports that recognition of the 2026 Bundibugyo virus outbreak was delayed by political circumstances and by problems getting people and equipment there. It describes qualifying diagnostic tests while the outbreak was still running. [31]

Why it matters — The tests were still being qualified while the outbreak was running. [31]
33
Fauci answers the too-many-vaccines claim

In an interview published on 3 September, Anthony Fauci, who ran the US government's infectious disease institute for decades, addressed the claim that many vaccines overwhelm a child's immune system. That is just scientifically not the case, he said. [32]

Why it matters — The claim is central to current US vaccine policy arguments, and he is the person it is usually aimed at.
34
A frog experiment names a heart-defect gene

Researchers testing rare human gene variants in Xenopus frogs identified ARHGAP10 as a gene behind congenital heart disease and ciliopathies, disorders of the tiny hairs cells use to sense and move fluid. [33]

Why it matters — It shows a route from a rare variant in one family to a mechanism, which is the step that usually stalls. [33]
35
What would count as reversing ageing

A Cell Metabolism piece on 1 September sets out what an age-reversal claim should have to show. It notes that several ordinary things reverse multiple aspects of ageing, including rapamycin in laboratory mice and regular exercise. [34]

Why it matters — The bar matters commercially: without one, every intervention that improves anything can be sold as reversal. [34]
36
Alzheimer's drugs in brains with two diseases

A viewpoint in a JAMA journal asks how doctors should use the new amyloid-clearing Alzheimer's drugs in patients who are likely to have a second brain disease alongside it. [35]

Why it matters — Most older brains carry more than one problem, and the trials were run in people chosen for having one. [35]

02 Lesson why it matters

The drug that does none of the work

A molecule far too weak to break a protein can still destroy it, if it holds the target against machinery that was already running.

The twist

A molecule far too weak to break a protein can still destroy it, because it never had to do the breaking. It only had to hold the target where the breaking already happens.

How it works

A protein inside a cell is doing harm
The old way is to grip it and hold it shut
The new drug grips it with one end only
Its other end grips machinery the cell already runs
Held there, the machinery does what it always does
The protein is destroyed, and the drug destroyed nothing

Where you've seen this

Handcuffs

two people who could each walk away are stopped by a thing not strong enough to stop either one

An introduction

whoever puts two people in a room does none of the work and changes what happens

A coin in the gears

it never has to out-muscle the engine, only to sit where two parts meet

The catch

It only works where two proteins can actually be brought together, and where the cell's machinery is present and working. It took more than twenty years to get from the idea to one approved drug, and nearly every glue found so far was found by accident.

And the whole of it

The cell was already taking proteins apart by the billion long before any of this was invented; these drugs only change which ones end up in front of the shredder. Nobody involved built that machinery, and the people who worked out how to borrow it spent twenty years finding out what it would accept.

03 Truth what's really going on

What is really going on

A twenty-year-old university idea got its first approved medicine and its first three-billion-dollar price tag in the same year, and the buying is running well ahead of what has been shown in people.

Why it works on us — One approval makes a whole field feel settled. A biochemist quoted in the reporting says people got nervous about these drugs until one was cleared, and nothing about the other thirty trials changed on the day it was. [1]

Who gains

Craig Crews and the companies built on his idea — He co-founded both Halda, which Johnson & Johnson bought for $3.05bn, and Arvinas, which made the first approved drug of the type. [1]
Akeso and Summit — Their shares moved on a data cut whose full figures have not been shown yet; Summit's rose as much as 14% in a morning. [3]
GSK — It bought most rights to a first-of-its-kind cancer drug for $110m up front, with the other $1.19bn payable only as the drug hits milestones or earns royalties. [5]

Who pays

Families affected by lupus — The drug that failed is going to a final-stage trial that will mostly enrol one subgroup, so the answer for everyone else moves years further away. [10]
Alumis shareholders — The stock fell 54% in pre-market trading on the morning the results were released. [10]
Patients whose lung cancer has stopped responding to a targeted drug — The trial they were in delayed the cancer clearly and could not show it extended life, so the choice they face is unresolved for now. [4]
People in the 2026 Bundibugyo outbreak — Recognition of the outbreak was delayed by political circumstances and by the difficulty of getting people and equipment there, and the diagnostic tests were being qualified while it ran. [31]

What nobody knows yet

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

01

Whether the new lung cancer drug actually helps people live longer.

Two results landed in one week and point different ways. The published final-stage trial gives a median 16.8 months against 14.0, close enough to chance that the trial could not call it. A different trial's new data cut is reported as showing a survival gain, with one analyst expecting a 24% lower risk of death when the full figures are shown. [4][3]
02

What starts the brain changes that show up seven years before plaques.

The Oslo team offers two explanations and cannot choose between them: something invisible is already driving the plaques, or something else is damaging the brain first. Which one is right decides whether the amyloid drug hunt is aimed correctly. [6]
03

Why overdose deaths fell across several countries at once.

The report's author asked officials in more than two dozen places and every one of them said no single measure explains it. State health authorities who have fought the crisis for decades say they were surprised. [12]
04

Whether Alumis's lupus drug works in anybody.

It missed its main and backup measures across 408 patients. The company is relying on one group picked out in advance, which turned out to be about 60% of enrolment instead of the 80% expected, and it has not yet agreed the next trial with the regulator. [10]
05

What Halda's handcuff drug does in a real comparison.

The 59% figure comes from an early trial with no comparison group, in men who had already had other treatments. Johnson & Johnson paid $3.05bn on that evidence. [1]
06

How much of the mouse ageing result reaches people.

Blocking the EP2 receptor kept organs across the bodies of old mice more youthful, and the human work was in cells. Nobody has given it to a person. [9]
07

What killed the six-year-old girl in Shanghai.

She died last year after an experimental gene-editing treatment aimed at her brain. The case was reported by an investigation in July, and it pre-dates the framework China brought in on 1 May. [17]
08

Whether one permanent gene edit for cholesterol is safe over a life.

Fifteen people were dosed and followed for a year. Switching off ANGPTL3 in the liver cannot be undone, and a year is not a lifetime. [13]

04 Hope carry this

Craig Crews wrote down the idea that a drug could destroy a protein instead of blocking it more than twenty years ago, when nothing of the kind existed. On the first of May a medicine built that way was approved for a form of advanced breast cancer.