Research Digest
Abundance moved, and four papers checked whether it mattered
Akkermansia muciniphila took 23 of this batch's 64 scored links. Five of them stop at reporting that its abundance shifted. Four went further and tested whether the organism was carrying the effect rather than just accompanying it.
This sweep queried all 508 active strains over a window reaching back to June and brought in 234 papers, every one published in 2026. It created 823 link rows; off-context pruning removed 689 of them and an alias-pollution row accounted for one more, leaving 133 links across 26 strains, of which 64 are scored primary or secondary. Those 64 links sit on 59 distinct papers: 43 in vitro or animal, 16 human, and only three carrying randomised evidence. Akkermansia muciniphila alone took 23 of the 64, and the interesting split is not how many but what kind.
Most papers that report a bacterium rising after some intervention stop there. This one did not. In mice given a plant-derived compound, A. muciniphila became enriched, and the authors report the enrichment was necessary for the compound's effect rather than incidental to it, and trace the route through an indole metabolite acting on a host receptor. Whether the organism is doing the work is a separate question from whether its abundance moved, and it takes a separate experiment. Three other papers in this batch supplied the organism directly: a cannabinoid study in which antibiotic depletion weakened the effect and giving the organism restored it, a mouse knockout study in which the organism or its extracellular vesicles attenuated the outcome, and a short-chain fatty acid study in which a propionate-producing A. muciniphila was used to test the metabolite route. Four of fourteen secondary links, which is both better than this literature's reputation and a long way from a majority.
The five that stopped at the correlation
Against those four, five of the fourteen report an abundance difference and leave it there. Three are human observational studies that measured its abundance alongside something else: 64 patients starting immunotherapy, where baseline gut composition differed between those who went on to respond and those who did not; 28 women with primary ovarian insufficiency against 23 controls, where the two groups differed in the relative abundance of several taxa including this one, with no difference in alpha diversity; and 79 patients followed for six months after bariatric surgery, whose authors call the link between the organism's change and the hormone absorption measure they tracked modest and exploratory. That last label is the correct one for all three. A difference in abundance between groups is a reason to run an experiment, not a result about what the organism does.
One of this batch's own Akkermansia reviews says so directly. Surveying medicinal plants reported to raise its abundance, the authors note that human evidence is limited and mixed and caution that abundance changes do not prove causality, efficacy or safety. When the field's own review literature is making that argument, a catalogue organised by strain should be making it too.
A quieter result in the same batch shows why the distinction has teeth. In ovariectomized rats given pasteurized A. muciniphila, faecal butyrate rose without any detectable enrichment of Akkermansia and without the altered microbiota being restored. Whatever the preparation did, it did not require the organism to establish. Abundance and effect came apart in opposite directions in the same genus, in the same batch.
The strongest design in the batch rates itself low
Three of the 59 scored papers carry randomised evidence, and two of them involve the same strain. Limosilactobacillus reuteri DSM 17938 gained a pooled analysis of six randomised trials in 565 children, a separate randomised trial in 50 infants that gave it together with BG-R46, a second strain derived from it, and an animal study in newborn rats reporting changes in lung tissue measures and antioxidant expression. The third randomised paper tested a multi-strain product, linked to Lacticaseibacillus rhamnosus GG as one ingredient among several. A combination trial cannot isolate any one strain in it, which leaves the pooled analysis as the only result in the batch that speaks to a single strain at the strength of a trial. We are describing designs rather than outcomes here on purpose, and the pooled analysis supplies the reason: it rates the certainty of its own evidence very low to low. That rating is the finding a reader should carry away. It is the strongest design in the batch and it still does not settle much, which is the ordinary condition of strain-level probiotic evidence rather than a defect of this particular set of trials.
Six links and not one trial of the organism
Faecalibacterium prausnitzii took six scored links, and five are human observational studies in which it appears among the taxa measured while something else is being tested or described. The sixth is a review of its role in neurological conditions whose own summary states the position plainly: the human evidence is observational, the interventional evidence is mostly from rodents, and there are no human trials of the live organism at all. Both halves of that sentence are worth keeping next to each other on a strain page.
The five also show how wide the net has become, and they are not one study design. One sampled the uterine microbiome of 34 healthy women, found every sample positive for bacterial genetic material, and sorted them into three community types, with no comparison group at all. That is a legitimate descriptive study of a site this organism is not usually discussed in, and it adds a link to the strain page without adding anything a reader could act on. Counting links is a poor proxy for counting evidence, which is why these pages report the design mix.
Nissle keeps being used as a chassis, and this time it moved a meter
Escherichia coli Nissle 1917 took seven scored links, and six study it modified, instrumented or killed rather than as a probiotic anyone can buy: engineered to express an enzyme or a payload, chemically decorated, processed into empty envelopes, or radiolabelled for imaging. Only one, a growth test on mushroom processing by-products, studies the unmodified organism. The 2026-10-12 batch split 11 engineered against 6 unmodified on the same strain, so this is the second batch running the same way, and it means a tall paper count on this strain page is largely counting a different activity.
One of the six is the only paper in the entire batch to carry a contaminant tag, and it is the one closest to this catalogue's own subject. An engineered Nissle expressing a PET hydrolase, coated to prolong gut retention, lowered uptake of polyethylene terephthalate nanoplastics by Caco-2 cells, and was then carried into a mouse model exposed to the same nanoplastics. It is scored primary with a host-context setting, and it is the first literature in a while to lift a contaminant confidence meter on this index: Nissle's plastic-interaction rollup moved from none to weak, the only confidence change among 5,984 rollup rows. No strain's contaminant mark changed, because Nissle already carried host-context evidence in all three contaminant categories.
What the paper is not is evidence about a product. The organism in it was given a gene it does not have, then coated in a synthetic polymer layer, and the work ran in cell culture and mice. A reader comparing strain pages should read it as a demonstration that the chemistry can be made to work inside a living organism, which is genuinely interesting, and not as information about anything on a shelf. We file it under the strain because that is how this index is built, and the limitation belongs in the open.
The most useful of Nissle's other papers is the one that reports its own method failing. Seven pigs were given radiolabelled Nissle capsules so the organism could be tracked by imaging; the label stayed in the gut with negligible systemic absorption, which was the question, but the labelling and freeze-thaw encapsulation also cut bacterial viability and the radiolabel came loose to a considerable degree. A paper that publishes the limitations of its own tracer is more useful to the next group than a cleaner result would have been. A second engineered strain outcompeted a pathogen for attachment to cultured cells by more than a hundredfold and lowered its intestinal carriage in three kinds of poultry by more than tenfold, which is a livestock result and is recorded as one.