Long covid may be driven from the gut, not from the lungs

Every bacterium that swims does it with a tail. That tail is a stiff helical filament, and the filament is built out of many copies of a single protein called flagellin. Your immune system has known that protein for a long time. In 2001 ², Fumitaka Hayashi and colleagues reported in Nature that one mammalian receptor, TLR5, recognises flagellin from both Gram-positive and Gram-negative bacteria, and that meeting it is enough to switch on inflammatory signalling. That is a sensible arrangement to have evolved. A bacterium with a motor is a bacterium that can travel, and travelling bacteria are the ones that reach tissue they were never meant to reach.

The question a new paper asks is what happens when that alarm gets pulled every day, quietly, from inside your own gut.

What the study did

The paper is by Jiejie Geng and colleagues, published in the journal Gut in 2026 ¹. They compared gut metagenomes across four patient groups ¹ — rheumatoid arthritis, ankylosing spondylitis, inflammatory bowel disease and long covid — and found the same thing in all of them: an expansion of flagellated bacteria. The swimmers in a human gut include members of Pseudomonadota and the flagellated clostridia inside Bacillota, so this is a shift within communities you already carry, not an infection arriving from outside. The published abstract does not name which taxa expanded, so I am not going to put a share on any of them.

They then picked long covid as the case to take apart, and their reason is worth repeating: it has a datable onset, and most patients have not yet been treated for it, so the biology is less muddied than in a long-standing arthritis.

The chain, link by link

Using single-cell RNA sequencing, functional studies in longitudinal human cohorts and co-infection mouse models, the authors lay out a chain with four links ¹:

• flagellated bacteria expand, and their flagellin reaches immune cells;
• flagellin engages TLR5 on neutrophils, which answer by casting out neutrophil extracellular traps — webs of their own DNA — and by releasing the cytokine IL-15;
• IL-15 pushes macrophages to produce arachidonic acid, the fatty acid that feeds the body's inflammatory lipid machinery;
• organs a long way from the gut inflame.

In mice co-infected to imitate the human course, the animals developed lung fibrosis and lymphoid clumps in the intestine — the multi-organ picture, reproduced in an animal. Then comes the part that matters. When the researchers genetically shut off macrophage arachidonic acid synthesis, or neutrophil IL-15, the lung damage eased. And when they cleared the gut microbiome with gentamicin, systemic inflammation fell in a way the other interventions did not match. Cut the chain and the distant organ suffers less. That is what turns a correlation into a candidate mechanism.

None of this arrives out of nowhere. Flagellins were identified as a dominant antigen in Crohn's disease in 2004 ⁴, with raised serum IgG to those commensal flagellins in Crohn's patients but not in ulcerative colitis or in controls. And mice bred without TLR5, reported in Science in 2010 ³, grow up with an altered microbiota and features of metabolic syndrome — many of those features transferable to wild-type germ-free mice by moving the microbiota alone. Flagellin has been circling this territory for a while.

What this study cannot tell you

First, and this is the big one: the cross-disease part of this paper is association, not mechanism. Finding the same bacterial signature in rheumatoid arthritis, ankylosing spondylitis, IBD and long covid does not show that one process drives all four. The mechanism was dissected in long covid and in mice. The other three diseases contributed a pattern. The authors are careful about this themselves — they call it a candidate mechanism and ask for prospective validation across disease boundaries.

Second, I cannot tell you how many people were in these cohorts. The published abstract does not give the sample sizes and the full text sits behind a subscription, so that number is not something I have verified. I would rather leave the hole visible than fill it with a figure I cannot name a source for.

Third, the rescues were genetic and pharmacological, and they were done in mice. An antibiotic quieting inflammation in a mouse is not permission to take one. Wiping out a gut community has costs this experiment was not built to measure.

What it changes today — nothing you can act on yet

There is no diet, no test and no supplement waiting at the end of this paper. What it does give is a named target somebody can go and test — flagellin, and the TLR5–IL-15–arachidonic acid line running downstream of it — plus a suggestion that arachidonic acid might serve as a marker you could measure. If that survives a prospective cohort — researchers following a group of patients forward in time instead of looking backwards at records — the interesting question becomes whether the flagellated share of a gut community can be shifted without a sledgehammer. Nobody has run that trial. When somebody does, I will report the numbers it produces, not the ones it promises.