Most of this batch used the strains as equipment

This sweep queried all 508 active strains over a window reaching back to June and brought in 271 papers, every one published in 2026. After off-context pruning removed 526 link rows, 181 links across 37 strains remain, and 84 of those are scored primary or secondary across 25 strains. Three strains hold 53 of the 84: Lacticaseibacillus rhamnosus GG with 18, Akkermansia muciniphila with 18, and Escherichia coli Nissle 1917 with 17. What those papers are doing is more interesting than how many there are: in the great majority the strain is the instrument rather than the question.

The strain that has become a chassis

E. coli Nissle 1917 gained 17 scored links, and 11 of them are engineering papers: the organism modified to express an enzyme, carry a payload, or respond to an external trigger, then evaluated in a disease model. Work of that kind is evidence about a construct, not about a probiotic anyone can buy, and filing it under the strain is a limitation of how this index is organised rather than a claim about the strain. We record it because the engineering literature is where most of the money and attention now sit, and a reader comparing strain pages should know that a tall paper count on this one is largely counting a different activity.

Six of the 17 do study the organism as it is, and two of those are worth more than their citation counts will suggest. A perfused three-dimensional gut-on-chip found that Nissle produced transient changes in epithelial organisation followed by recovery, while a pathogen damaged static two-dimensional cultures at lower doses than the chip, which held its tissue integrity. A co-culture biofilm assay reported that Nissle formed significantly more biofilm than E. coli K12 at 30 degrees and that neither strain formed measurable biofilm at 37, and its authors conclude their own assay is unsuited to the question they built it for. A paper that reports its method did not work is doing something the 11 engineering papers mostly cannot.

For LGG the question being asked is survival, not effect

The largest group of the 18 scored LGG links, six of them, is about getting the organism somewhere intact and keeping it alive. Alginate microcapsules with quince and flaxseed mucilage kept 88.55% of cells viable through simulated gastrointestinal exposure against 43.04% of free cells. Capsules built with grape pomace flour held 8.0 log CFU per millilitre or more after 28 days at 4 degrees while free cells fell to roughly 3.3 to 4.0 log, and also protected cells through pasteurisation and storage in juice. A third group used the strain's own CRISPR-Cas system to restore lactose use with a marker-free 20 base pair deletion, producing an edited strain that acidified milk and survived yogurt storage better than the wild type. A fourth found that digested soybean protein and peptides increased adhesion of LGG to intestinal cells along with higher mucin secretion.

Those are formulation and manufacturing results, and they are the honest answer to a question this catalogue often gets asked, which is whether the organism in a product is still alive when it arrives. None of them measures an outcome in a person. Both facts belong next to each other.

Akkermansia gained a standards proposal and a way to remove it

A review proposing a five-module framework for A. muciniphila research argues for strain-specific evaluation and for standards that make live, pasteurised and postbiotic forms comparable. That the field needs such a proposal in 2026, for its most-studied next-generation candidate, is itself the finding.

The sharper paper runs the other way. Researchers isolated 27 lytic phages that kill A. muciniphila without affecting the other gut bacteria they tested, and an optimised cocktail suppressed it both in stool-derived communities and in colonised mice. A tool that subtracts a single organism is how necessity gets tested rather than assumed, and it is also a caution: an organism worth adding on purpose is an organism whose abundance can be pushed in either direction, and neither direction is automatically good.

One exploratory result, labelled as such

A reanalysis of four public metagenomic datasets found Faecalibacterium prausnitzii enriched after resistant starch exposure, and reported that baseline co-occurrence with Ruminococcus bromii was linked to larger increases in some acetate and lactate pathways. The authors label that second part exploratory, and the label is the point: a pattern found by reanalysing data collected to answer other questions is a reason to design a study, not a result. The batch's other null is smaller and cleaner. Fed to worms, LGG did not extend lifespan in either wild-type or daf-16 C. elegans compared with standard E. coli feeding. A negative result in a nematode says little about a person, which is exactly why it is worth recording in a catalogue where positive animal results are the most common kind of evidence on a page.

The contaminant links, and the marks that did not move

Three links in this batch carry a contaminant tag, and no strain's contaminant mark changed as a result. All three evidence columns for all 544 strains were compared against production at the end of the local pipeline and came back identical. It is worth saying plainly, because a tagged paper and a changed mark are different events and the first is far more common than the second.

The closest to this catalogue's own subject is the weakest as evidence: a consortium of PETase-expressing Nissle with a biofilm-forming Pseudomonas formed about 1.5 times more biofilm on PET film than the control, and roughening the film increased it further. That is engineered plastic-surface colonisation on a bulk film in a flask, filed as plastic interaction with an unclear setting because the paper states neither a host nor an environmental one. It is scored secondary, and contaminant categories count primary evidence only, so it cannot move a mark by design.

The second is a heavy metal study with the strongest design of the three: in cadmium-exposed mice, A. muciniphila supplementation reversed changes in hippocampal m6A methylation and alternative splicing across more than a thousand genes, largely without changing total messenger RNA levels, and the authors propose that RNA regulation as the mechanism of protection. It is primary and host-context, and it still moved nothing, because Akkermansia already carries host-context evidence in all three contaminant categories. The paper deepens a mark rather than creating one.

The third is a null that reports its own confounding. An in vitro test of the plasticizer DEHP on E. coli BL21 found limited growth inhibition in MIC assays but could not separate the compound's effect from poor aeration caused by phase separation, and its growth curves conflicted. It is filed as chemical interaction rather than plastic interaction, because a leached additive is not a particle, and that distinction is the reason the categories were split apart.