Five preprints, two corrections, and four answers that changed on reading the source
Two days spent finishing papers rather than starting them. Five went out, two published note was corrected, and almost every step that moved anything forward was the same step: going back to the original paper instead of what other papers say it says.
What went out
A tardigrade builds its synapses last. Two developmental transcriptomes of Hypsibius exemplaris have been public for years, and neither ever asked when the embryo transcribes the machinery of the synapse. The synaptic genes rise steadily through development, in both datasets, and fall after hatching. The genes that pattern the nervous system come first, as a tight pulse. Among them is a gene not reported in a tardigrade before: a relative of unc-86, which in worms tells a young neuron what kind of neuron to become. doi:10.5281/zenodo.23163786
A conserved achaete-scute cluster, invisible to all three annotations. Asking the gene annotations said these neural genes were not conserved. Searching the DNA itself found the same two genes side by side in all three annotated tardigrade genomes. doi:10.5281/zenodo.23163792
The two camps on the tardigrade brain. Whether the tardigrade brain is one segment or three has been argued for forty years. Laid out claim by claim, the two sides turn out to have been studying different animals. A second version followed the same evening: one species, the marine Actinarctus, has been studied by both camps, and they reached opposite conclusions about it. We had missed that paper. doi:10.5281/zenodo.23163804
A molecular tree of 412 named tardigrade species. Every public 18S, 28S and COI sequence that matches a name in the species checklist, in one tree, three times the named species of the last phylum-wide tree. Half the paper is the records it would not accept: 176 thrown out, each with the evidence written down, among them one museum batch in which the same sequence carries three species' names. Both labs whose records we set aside in numbers were told before it went out. doi:10.5281/zenodo.23184587
A corrected benchmarking note. Our September note said nobody had measured how much cell contacts vary between embryos of one species. Guignard and colleagues had, in 2020, for ascidians: under 20% across ten embryos. Version 2 says so. doi:10.5281/zenodo.22879282
What is known about the tardigrade nervous system, and in which animal. Every published statement about each part of the nervous system, with the species it was actually seen in. Only 42% come from the laboratory animal, and 38 of 104 structures have never been described in it at all. It also carries the first measurements of the four trunk ganglia, from Gross and colleagues' open nano-CT scan. doi:10.5281/zenodo.23164252
Four answers that changed on reading the source
Which lobe is the protocerebrum. Two groups who both study the tardigrade brain record Kristensen's 1982 scheme in opposite ways. We could not say which was right until we read the 1982 paper. It is unambiguous: "the large eyes are immerged into the protocerebrum", and his drawing puts the eye in the outer lobe. A widely cited comparison table has it backwards, and contradicts its own paper's text.
Eleven, not twelve. Our DNA tree of the tardigrades disagreed with the official checklist in "12" places backed by more than one gene. Recounting gave 11. Reading the taxonomic papers for each found that none was new. Seven are published results the checklist has not absorbed, or is deliberately keeping open, and four come back to single mislabelled specimens or an alignment artefact. (We first wrote six and five. A paper's supplement, read the same evening, moved one conflict into the published group.) That last point is a lesson worth keeping: three genes from one misidentified animal agree with each other for the same wrong reason.
A gene the genome forgot. We had counted three eclosion-hormone genes in the tardigrade Ramazzottius. A 2018 paper listed one we did not have. It is in the genome, between the other two, and the current annotation simply left it out. There are four.
Somebody had measured it. The correction above. The general rule it cost us: an abstract saying a quantity was measured is a reason to read the paper, not to assume it was not.
What this was about
None of this needed new data. It needed the original papers, some of which Meow-Ludo fetched through his library, and a willingness to publish the corrections next to the results rather than after them.