Biotech & Longevity · Wednesday, 26 August 2026

01 · Briefing · what happened

A water treatment that kills bacteria left more resistance genes behind than doing nothing at all

Biotech & Longevity 1 min 33 sources

In three Chinese fish ponds, an advanced disinfection process produced water with a higher burden of antibiotic-resistance genes than untreated water. The reason is the most ordinary thing in any pond: decayed plant matter, which soaks up the killing chemistry and leaves a dose that trains rather than kills.

aquaculture ponds surveyed

in eastern China, in the field rather than a lab [1]

Higher

resistance genes after treatment

than in untreated or chlorinated water [1]

6 months

of Italian wastewater sampling

across a hospital, a university and a city sewer [2]

$2.7m

price of the week's new gene therapy

a one-time treatment for a rare metabolic disease [10]

At a glance

Researchers surveyed three aquaculture ponds in eastern China and compared water treated with an advanced oxidation process against untreated and chlorinated water. [1]
The treated water carried significantly HIGHER burdens of antibiotic-resistance genes inside its bacteria than either comparison. [1]
The culprit is humic acid - the brown residue of decayed plants, present in essentially every natural water on earth. [1]
Humic acid acts as a selective scavenger: it soaks up the radicals that were meant to kill the bacteria, shielding them from a lethal dose. [1]
What is left is a sub-lethal dose. That is the dose that stresses a bacterium rather than killing it - and a stressed bacterium starts swapping genes. [1]
The team traced it directly, using a gene-transfer sorting assay alongside metagenomics and transcriptomics, rather than inferring it from the end state. [1]
The same week, wastewater sampling in Perugia found resistant E. coli at comparable levels in a hospital, a university building and the general sewer - the resistance is already shared across the whole population, not concentrated in hospitals. [2]
Vancomycin-resistant enterococci did show a clear gradient - highest in hospital wastewater, lowest in the university building - so the two kinds of resistance are spreading by different routes. [2]
In Czech surface and waste waters, researchers mapped carbapenem resistance in Citrobacter carried on plasmids - the mobile packages that move between species. [4]
In tuberculosis, drying out the bacterium - which is exactly what happens as it forms an airborne droplet - damages its DNA and produces resistance to rifampin, a first-line drug. [3]
Thai research compared indigenous and commercial chickens and found different gut resistance profiles, which points at farming practice rather than geography. [5]
Almost none of these studies looked in a hospital. They looked at ponds, drains, rivers and birds.
A macrolide trial in healthy adults tracked what an anti-inflammatory antibiotic does to a normal gut, which is the other half of the same question. [9]

Forces in play

Sub-lethal doses High↑

treated pond water ended up with more resistance genes than untreated water, because the dose shielded rather than killed [1]

Resistance outside hospitals High↑

resistant E. coli at comparable levels in a hospital, a university and the general sewer in Perugia [2]

Detection reaching further Building↑

building-level wastewater sampling and bedside vancomycin tests are both new ways of seeing it [2][6]

New drug classes Easing→

a bacterial defence system and a nanoparticle approach both published this week, both early [7][8]

In play Humic acid — decayed plant matter, and the reason a disinfectant behaved backwards Aquaculture ponds — where antibiotics, bacteria and treatment all meet at scale Plasmids — the mobile packages that carry resistance between species Wastewater — the only place you can watch a whole population's resistance at once

How it unfolded

Feb-Jul 2025 monthly wastewater sampling runs in Perugia across three catchments [2]
This week the pond study shows treated water carrying more resistance genes than untreated [1]
This week tuberculosis is shown to gain rifampin resistance simply by drying out [3]
This week Czech river and wastewater plasmids are mapped, and Thai chicken guts compared [4][5]

Where this points

The finding to watch for is whether any of this changes a treatment standard: advanced oxidation is being deployed more widely for disinfection, and the study says its effect on resistance genes is still poorly understood.

Also today

10 more stories on this beat.

One gene therapy approved, another halted

The FDA approved Ultragenyx's Glenglycos for glycogen storage disease type Ia, the company's first gene therapy, priced at $2.7m per patient. Days later it placed Regenxbio's Hunter syndrome therapy on clinical hold after spinal scans found a small lump or fluid-filled mass in five participants treated three to six years ago. [10][11][12]

Why it matters — The same technology, the same regulator, the same week. What separates them is follow-up long enough to see something that takes years to appear.
Where CRISPR actually goes

New profiling shows CRISPR-Cas9's off-target edits and chromosome rearrangements vary by organ and by cell, rather than being a single background rate. Separately, researchers reported reprogramming translation itself to treat inherited hearing loss. [13][14]

Why it matters — A safety figure averaged across a body is not a safety figure for the organ the therapy actually lands in.
Moderna's cancer vaccine held

Moderna's individualised cancer vaccine stopped melanoma returning in its trial, a result that restored Wall Street's confidence in the company. Researchers separately supercharged natural killer cells so they can penetrate solid tumours and keep working inside them. [15][16][17] Bristol Myers Squibb also signed an AI antibody-discovery collaboration, betting that the next immune therapy is designed rather than found. [18]

Why it matters — All three are the same idea from different ends: stop attacking the tumour directly and re-arm the immune system that was already there.
The weight-loss drugs are being reconsidered, not celebrated

Analysts expect cost and public perception to reshape the GLP-1 market; Roche bet on lean-mass preservation with a Hanmi deal, because the big weight losses come with muscle loss; and a review reappraised these drugs against bariatric surgery rather than against nothing. [19][20][21]

Why it matters — The question has moved from whether they work to what they cost you - in muscle, in money, and against the alternative nobody discusses.
Ozempic appears to work backwards in the brain

In mice, semaglutide activated hunger-linked neurons that turn out to be necessary for sustaining fat loss - close to the opposite of what was expected. A decades-old compound, TOFA, separately increased energy use rather than suppressing appetite. And a UK trial is testing semaglutide for alcohol dependency. [22][23][24]

Why it matters — A drug taken by millions is still being reverse-engineered, which is normal and worth saying out loud.
What the very old have in common

People reaching 110 and beyond carry unusually high numbers of rare CD4 cytotoxic T cells, which kill abnormal cells and become more common with age. Meanwhile nutrition researchers warned that cutting protein to extend life is a bad idea, and others found a shortcut to calorie restriction's benefits by blocking one inflammatory protein. [25][26][27]

Why it matters — Two of these are about doing less. The third is about your immune system doing more, and it is the one nobody can copy yet.
Sugar rationing became an experiment nobody designed

Britain's postwar sugar rationing created a natural experiment: people who experienced more rationing before the age of two show measurably different health seventy years later. [28]

Why it matters — You cannot run this trial. A government did it by accident, kept records, and the answer arrived in 2026.
A shark that does not go blind

Greenland sharks may live 400 years without the retinal decline that ageing normally brings, with healthy eye tissue and specially adapted proteins still working. [29]

Why it matters — Every ageing process we think is inevitable is a process something else has already found a way around.
The money is not reaching the young companies

Biotech venture funding has recovered since its 2022 bottom, but the youngest startups are missing out on it, and women's health startups are hoping a wider definition of the field can restart investment that peaked in 2024. Elsewhere researchers built contracting muscle implants that reproduce some of exercise's benefits in mice. [31][32][33]

Why it matters — A recovered sector average can sit on top of a funding drought for anyone who was not already established.
The largest Bundibugyo outbreak on record

The 2026 Bundibugyo virus outbreak across the Democratic Republic of the Congo and Uganda is the largest ever recorded for this Ebolavirus, and a vaccine-preparedness review set out what previous outbreaks did and did not teach. [30]

Why it matters — Preparedness for this specific virus was built from experience with its relatives, and the review exists because that turned out not to transfer as cleanly as hoped.

02 · Lesson · why it matters

A dose too small to kill is a dose that teaches

Half of a lethal dose is not half as effective - it is a different intervention, one that selects for whatever survived it.

How it works

A treatment is designed and tested in clean conditions
The real world contains something ordinary the lab did not
Here it is humic acid: decayed plant matter, in every pond
It absorbs the chemistry meant to kill the bacteria
So the full dose arrives as a partial dose
And a partial dose does not kill - it teaches

The twist

A dose too small to kill is not a weak version of a dose that kills. It is a different intervention entirely: a training signal. That is why the treated water ended up worse than water nobody treated at all.

Where you've seen this

Half-finished courses of antibiotics

the reason the instruction is finish the packet, not take enough to feel better

Partial pest control

a spray that reaches most of a field breeds the survivors that reach all of it

Half-enforced rules

a fine small enough to absorb teaches which rules are priced rather than prohibited

Incomplete sanctions

pressure that a state can survive is pressure it learns to route around

The catch

It does not mean disinfection is bad. Chlorinated water did fine in the same study. It means the interaction between a treatment and whatever is already in the water is not a detail - it is the result.

And the whole of it

Nobody in this chain did anything careless. The engineer chose a process with good laboratory evidence; the pond owner treated the water as advised; the chemistry did precisely what chemistry does in the presence of decayed leaves. The failure lives in the gap between a controlled test and a real pond, and almost every failure worth studying lives in a gap of that shape.

03 · Lab · your turn

The Dose That Teaches

Treat a pond three times and watch a light dose breed the resistance a full dose does not, then discover that murky water turns your full dose into a light one.

04 · Truth · what's really going on

Stripped of the framing

A disinfection process that performs well in a laboratory performed backwards in three real ponds, because real water contains decayed plant matter and laboratory water does not. The finding is not about that chemical - it is about how much of what we know about treating water was learned in water nobody would recognise.

Why it lands — 'Advanced oxidation process' sounds like an upgrade on chlorine, and 'humic acid' sounds like a contaminant. In fact humic acid is just rotted leaves, it is in every pond on earth, and the older, plainer treatment was the one that behaved.

Claimed

What people said. Not yet a fact.

The study authors

Treated waters carried significantly higher resistance-gene burdens than untreated or chlorinated ones, and humic acid is the mechanism. [1]

They went after the mechanism directly with a gene-transfer sorting assay rather than inferring it from the end state, which is the difference between a correlation and a cause.
The paper itself

The impact of these processes on resistance-gene spread remains poorly understood. [1]

An unusual thing for a paper to volunteer, and it is the sentence that matters most: the process is being deployed more widely while this is still true.
Regenxbio

It does not expect to resubmit its Hunter syndrome application in the near term. [11][12]

The scans found lumps or fluid-filled masses in five participants dosed three to six years ago. The finding is old; the detection is new.

Verified

What we could actually stand behind.

The effect was measured in three separate aquaculture ponds, in the field. [1]

How we checked — Field surveys rather than a bench experiment, with metagenomics and transcriptomics alongside the transfer assay - three methods pointing the same way.
Resistant E. coli was found at comparable levels in hospital, university and general-population wastewater in Perugia. [2]

How we checked — Six months of monthly sampling across three catchments, using both culture-based and molecular methods.
Ultragenyx's approved gene therapy is priced at $2.7m per patient. [10]

How we checked — Stated by company officials on an investor call, which is the most reliable kind of source for a price - it is a number they are bound by.

Nobody knows

Open questions — ours included.

Whether this happens at the scale of a municipal water plant.

The study is three aquaculture ponds. Ponds are warm, shallow, full of organic matter and often dosed with antibiotics. A drinking-water plant is none of those things, and nothing here tests one. [1]
How much resistance is actually moving from water into people.

The Perugia work shows the same resistance across a whole population's wastewater. It cannot show the direction of travel. [2]
What caused the masses in the Regenxbio patients.

Five people, spinal scans, three to six years after dosing. The company has not said, and a hold is what a regulator does when nobody knows yet. [11][12]
Whether the tuberculosis finding changes how transmission is understood.

Drying out damages the bacterium's DNA and produces rifampin resistance, and drying out is what happens in an airborne droplet - but the study is in the laboratory, on a filter, not in a person. [3]

Who gains

Chlorine — The older, cheaper, less impressive treatment is the one that came out of this study without a problem, having been steadily displaced by processes marketed as advanced. [1]
Holders of priority review vouchers — Ultragenyx's approval came with one it plans to sell; three sold this year for $180m, $195m and more, which can be worth more than the therapy that earned them. [10]
Established biotech firms — Venture funding has recovered on average while the youngest companies miss out, which concentrates the recovery in companies that already existed. [30]

Who pays

People downstream of treated water — Aquaculture ponds drain into rivers, and the resistance genes measured here are carried on plasmids that move between species. [1][4]
The five Regenxbio participants — They were dosed three to six years ago with an experimental therapy, and the scans that found the masses were done now. [11][12]
Patients priced out — A one-time therapy at $2.7m is a cure and a rationing decision in the same sentence, and nothing in the reporting says who gets it. [10]

05 · Hope · carry this

The team that found this went looking for a mechanism rather than reporting the odd result and moving on. Chlorine, the cheap old treatment nobody writes papers about, came through the same study fine.