Report — Field Note
Same molecule, different roads
Insect-killing bacteria arrived at the same chemical weapons again and again by separate routes, and half their gene swaps changed nothing at all. The molecule does not record the road it took.
Gardeners can buy the delivery system in this story. Tiny worms called nematodes, sold to control grubs in lawns, carry bacteria into an insect's body: Steinernema carries Xenorhabdus, Heterorhabditis carries Photorhabdus. The bacteria kill the insect, and among the chemicals they make are GameXPeptides: rings of five amino acids, the small building blocks of proteins. The common one runs valine, leucine, phenylalanine, leucine, leucine, written V L F L L.
Sourced · Burnell & Stock, Nematology 2 (2000); Kaiser et al., Nature Ecology & Evolution (2026)One enzyme, GxpS, builds the ring one station at a time. (An earlier Penumbra issue drew this kind of molecular assembly line.) Each of its five stations has a pocket that picks one particular amino acid. The pocket's shape decides the bead; the order of the stations decides the ring.
Sourced · Kaiser et al. 2026Genes can trade pieces, a process called recombination. When a station's pocket is replaced by a copy from somewhere else in the genome, the station may pick a different amino acid. Swap the third pocket for one that picks leucine, and V L F L L becomes V L L L L, a different ring with a different number.
Sourced · Kaiser et al. 2026 (compounds 1 and 6)Most swaps did not do that. The team, led from the Max Planck Institute for Terrestrial Microbiology and the University of Marburg, traced the enzyme's history across 54 versions in the two bacterial groups. Of 53 swaps they counted, 27 replaced a pocket with one that picks the same amino acid. The genes changed; the ring did not. A great deal of change at the small scale left nothing to see at the large one.
Sourced · Kaiser et al. 2026: 54 GxpS versions; 27 synonymous, 26 non-synonymous recombinationsSome swaps repeated. Phenylalanine at the third position was replaced by leucine seven separate times, in different branches of the family tree; four enzymes now make the V L L L L ring as their main product. Pockets that pick valine arrived twelve separate times, producing one ring, V V L V L, by five independent origins. Biologists call this convergence.
Sourced · Kaiser et al. 2026 (compounds 6 and 4) · the tree drawn here is schematicPut the pieces together and the relation between the small scale and the large one is not a straight line. Twenty-seven swaps produced no new ring. One swap could produce a new ring outright. Separate histories ended at the same ring, so the ring alone cannot tell you which history made it. Knowing the molecule leaves the road open.
Sourced · counts from Kaiser et al. 2026 · paths drawn as schematicWhich rings can appear at all depends on where the swapped pieces come from. Similar stretches of DNA trade pieces more easily, and most swapped-in pockets came from the same gene rather than from the bacteria's other assembly-line genes. The menu of possible rings is set partly by what sits close by and looks alike, whether or not a better choice exists elsewhere.
Sourced · Kaiser et al. 2026 (donor bias, P < 10⁻⁴) · arrows drawn as schematicThe team tested eight rings on four kinds of insects, fed leaves soaked in each one. Each ring and insect pair behaved differently. Ring 36 was the most active against beet armyworm and among the weakest against diamondback moth. Ring 4 was the most active against diamondback moth and the weakest against western flower thrips. Mealworms and thrips were less affected overall. No ring came out best against all four.
Sourced · Kaiser et al. 2026, bioactivity tests (rankings as stated in the text; the paper reports no single ranking)So three things shaped the family. Selection: swaps that change the ring hit the picking pocket more often than chance predicts. Chance: many swaps drift through without effect. And bias: the swaps on offer are the ones nearby. The authors write that "strong drift will limit the degree to which all aspects of NRPS function can be perfected by selection."
Sourced · Kaiser et al. 2026 (Fisher's exact test, P < 0.05; quote from the Discussion; NRPS = the family of assembly-line enzymes)The limits are the authors' own. They traced the history of one enzyme, in two groups of bacteria. The insect tests were feeding trials on four species in a laboratory, and the authors caution that no single ring likely targets one insect alone in the wild. Side products, made in smaller amounts, are undercounted. The finding that unchanged swaps reflect chance assumes those swaps make no difference to the bacteria, which is reasonable but not proven. The link to new antibiotics, made in the press coverage, is a possibility the paper raises, not something it tested.
What remains is a picture of how a family of molecules fills out. It did not converge on one best weapon; it is the record of what was nearby, what drifted through, and what worked well enough against the insects at hand. Some of that record is gone. Recombination overwrites older history, and the authors end on that loss, writing that much of it is "probably now lost, like tears in a rain of recombination." The idea to search is evolutionary contingency.
Sourced — Kaiser, Shahmohammadi, … Bode & Hochberg, "Drift, selection and recombination bias drive the evolution of a non-ribosomal peptide", Nature Ecology & Evolution (21 Oct 2026) · Burnell & Stock, "Heterorhabditis, Steinernema and their bacterial symbionts", Nematology 2, 31–42 (2000) · Entomopathogenic nematode fact sheet, after Berry (1998). Ring sequences are shown without their mirror-image (d/l) forms. Family tree, paths and arrows are schematic; counts are the paper's. Observed — photo by the editor.