← Longevity & Supplement Guides
No, your gut bacteria are not eating you — but fibre decides what they make
Somewhere in the middle of last week, the internet decided your gut bacteria were eating you.
The sentence spread the way the good ones do. A university write-up on 14 August, then the aggregators, then a tabloid warning that your gut bacteria could "literally start eating you" if you skimp on fibre. It is a wonderful sentence. It is also, in the way of wonderful sentences, doing a great deal of work that the research underneath it never asked for.
The underlying paper is better than the headline. It is just about something else.
The horror film was shot in 2016, and the cast were mice
The mucus-eating image comes from a genuinely important study: Desai and colleagues in Cell, 2016. They took germ-free mice, colonised them with a synthetic community of fully sequenced human gut bacteria, and withheld fermentable fibre — either continuously or in on-off cycles. The community turned to the host's own secreted mucus glycoproteins as a food source, and the colonic mucus barrier thinned. When the animals then met a mouse gut pathogen, they fared markedly worse.
Careful work, and the substrate-switching idea it introduced has held up. But look at what the system was: germ-free animals, deliberately stocked with fourteen known species, on a diet engineered to contain no fermentable fibre whatsoever. That is not a person eating too much white bread. It is an experiment built to make a mechanism visible by stripping out everything that normally obscures it.
Which is what good mechanistic biology does. It is also why the finding is ten years old and still being reported as news.
And the mechanism everybody repeats is contested
Here is the part that never makes the aggregators. In 2022, Overbeeke and colleagues put 38 conventional mice on a fibre-free diet for seven days and asked why the mucus layer thins.
It was not because mucus-degrading bacteria bloomed. Counted in absolute terms, those species barely moved. What fell was the host's own manufacturing: muc2 expression dropped, measured mucus secretion flux was lower, crypt height fell, and the intestine itself shortened inside a week. The barrier got thinner because the animal stopped building it, not because something ate it.
There is a methodological aside here worth more than the headline. Read as relative abundance, the mucus-degraders looked like they were surging — because total bacterial load was collapsing around them. Counted absolutely, they were standing still. That is the single most common way microbiome results get misread, in papers and in consumer gut reports alike: a percentage of a shrinking whole looks like growth.
The two stories point at different levers. If bacteria are chewing through the barrier, you are managing bacteria. If the host stops building the barrier when there is nothing to ferment, you are feeding a tissue. I think the second is the larger part of the story, and I hold that loosely, which is the correct grip for a seven-day mouse experiment.
What the new paper actually traced
The 2026 study behind last week's noise asked a narrower and more elegant question: where do the phenols in your bloodstream come from?
Gut bacteria build phenols and indoles out of aromatic amino acids, and those compounds end up circulating in you. Two families are of interest. One rises on plant-heavy diets — hippurate and 3-phenylpropionate. The other falls on plant-heavy diets: phenol sulfate and p-cresol sulfate, the pair the literature files under compounds you would rather have less of. That the ratio shifts with diet has been known for years. Which carbon atoms went where had not been established.
So AbuSalim, Rabinowitz and colleagues, in PNAS this August, ran isotope tracing in rodents to follow the atoms. The answer is clean. The pair that rises traces back to digestion-resistant dietary protein. The pair that falls traces back to protein the host secreted.
That is the "eating you". Not a bite — a tracer result. When your microbes run short of food that arrived with the meal, they metabolise protein your own body put into your gut, and the phenol profile tilts accordingly. Same bacteria, different substrate, different output.
Substrate, not appetite
Where the phenols in your blood come from
When the meal brings the substrate
Whole plants
Fermentable fibre and digestion-resistant plant protein, in the same bite
Microbes ferment what you sent down with it
Phenylalanine route
- Hippurate
- 3-phenylpropionate
When it does not
Little fibre, little resistant protein
The dietary substrate runs out
The same microbes turn to protein the host secreted
Tyrosine route
- Phenol sulfate
- p-cresol sulfate
The routes as traced by isotope labelling in rodents (AbuSalim et al., PNAS, August 2026). The diagram shows which substrate feeds which family of phenols — it encodes no quantities, and the tracing was done in animals, not people.
The finding they buried: the plant protein nobody counts
Now the part that should have been the headline. Fibre was not the only thing that mattered. Digestion-resistant plant protein — protein that survives your own digestion intact enough to reach the colon — did comparable work. The authors present the two as partners: together they set what the microbiome has to work with.
Which is awkward, because there is no line for it on any label. Fibre has a number, a reference value, an approved vocabulary and an entire supplements aisle. "The protein in this bean that you personally will not absorb" has none of that. It is invisible to the whole apparatus we built for telling people what to eat.
This is my favourite kind of result: not a new thing to buy, but evidence that the thing we were counting was standing in for something we weren't.
Why this is inconvenient for the fibre scoop
What follows is an argument, not a trial result, and I would rather label it than smuggle it.
Whole plants deliver fermentable fibre and digestion-resistant protein in the same mouthful. A tub of isolated fibre delivers one of the two. A bean delivers both, plus its own phenolic compounds, plus potassium and magnesium you were going to need anyway.
No human trial has put the powder against the legume on these particular metabolites, and until one does, this stays reasoning rather than evidence. But it is the direction the new tracing points, and the same direction most of the fibre literature has quietly pointed for years: whole food outperforms the isolated fraction more often than the supplement business finds convenient. If you have read our guide to fibre and the gut microbiome, read this as the amendment to it. Fibre is the shorthand. It was never the mechanism.
The gap, and its honest size
EFSA sets 25 g of fibre a day as the adult reference value. Typical intake across wealthy countries runs closer to 15 g, and the large majority of adults sit below the reference, year after year, decade after decade.
That gap is not an emergency, and it is not why you were tired on Tuesday. It is something duller and more durable: a structural shortfall that nobody experiences as a symptom. Which is precisely why the rodent work earns its place. It describes the direction of drift when the substrate thins out — and nobody is going to run the fibre-free arm in humans for a week and then biopsy the colon.
One thing to do this week
Add a legume to something you were already going to eat. Not a regimen — one addition. Lentils into the soup, chickpeas into the salad, black beans next to the eggs.
The reason is narrow and specific: of all ordinary foods, legumes are the one that carries both halves of what this work implicates — fermentable fibre and digestion-resistant protein — in a single cheap, boring package. And it is the choice that survives the caveat. Even if the phenol story turns out to matter less than it currently looks, you have added fibre, plant protein and minerals to your week, which nobody is arguing about. Compare that with a supplement bought on the strength of a mouse study, where the whole value rests on the mouse study being right.
Then notice rather than measure. There is no consumer test for your phenol profile, and I would be sceptical of anyone selling you one. What you can observe over a few weeks is unglamorous and real: bowel regularity, how long you stay full after lunch, whether your late-afternoon energy holds. None of it is diagnostic. All of it is information — the kind worth logging next to your other daily inputs, alongside what you already know about steady energy and how processed your food is.
The bottom line
Your gut bacteria are not eating you. They are opportunists with a metabolism, and when the food you sent them runs out they use what is available, which includes protein your own body put there. The new tracing work makes that traceable for the first time, and adds a genuinely fresh detail: indigestible plant protein counts alongside fibre, and nobody has been counting it.
The mucus-thinning image that carried the headline is a decade old, was demonstrated in mice, and is disputed in its details — the barrier may thin mostly because the host stops building it. All of which resolves to advice so unglamorous it barely survives a news cycle. Eat plants in roughly the form they grew in, most days. Stop counting the one component that happens to have a number on the box.
