Selection worked; the read-out did not narrow
The 10 base populations of Drosophila simulans were built from 93 isogenic strains by round-robin crossing and each was held at about 1,000 flies, with about 500 selected every generation against 3 microlitres of octanoic acid. Mean survivability climbed from about 43 per cent at generation 0 to about 78 per cent at generation 25, flattening out near generation 15; raising the dose to 9 microlitres restarted the response, which reached about 70 per cent by generation 50, roughly fivefold the unevolved populations at that dose. Three evolved populations showed no significant drop in eggs laid, developmental viability or lifespan, so the gain came without a measured trade-off in the traits that were checked.[1]
The genomes told a messier story. Allele frequency change was judged twice on the same 5,127,840 variants, once with the Cochran-Mantel-Haenszel test and once with Bait-ER, a Bayesian method that models genetic drift explicitly. At generation 25 the two returned 439 and 401 candidate genes; at generation 50 they returned 1,238 and 174. Fewer than 30 per cent of significant genes are shared between the two time points, and the authors are plain that hitchhiking neighbours sit alongside real targets in those blocks. A denominator in the thousands is where a polygenic trait usually leaves the analyst.[1]
The second screen is a filter and a blind spot at once
The orthogonal screen was a genome-wide pooled CRISPR/Cas9 experiment in Drosophila melanogaster S2R+ cultured cells: 6 to 8 guide RNAs per gene, two replicate cultures, two acid concentrations, 26 days before sequencing. Its top 5 per cent held 87 susceptibility genes and 22 resistance genes, with no single stand-out susceptibility hit and an enrichment of membrane trafficking genes. Laid over the selection blocks, the two lists leave four susceptibility genes of broad function and two resistance genes with a plausible direct role: kraken and CG14130, the fly counterpart of mammalian Alkbh7. kraken was the screen's top resistance hit and sits in one of the most prominent selection peaks at both generations.[1]
That filter has a declared cost. Genes essential for general cell survival and genes not detectably expressed in S2R+ cells were removed before ranking, which the authors themselves name as inevitable limitations of a cell-based screen. A whole-animal resistance gene that a cultured cell line never expresses cannot reach the intersection at all, so absence from the two-gene shortlist carries less weight than presence on it. The clearest test of that reading is the 170 kilobase window on chromosome arm 3R holding 18 genes, several of them from the Osiris family, which earlier mapping work put forward and which did not come out of this analysis. The alternative reading is equally available: the earlier mapping signal may have been a wide, noisy interval that this design was right to drop.[1]
Where a repeat would have to look
The intersection selects candidates; it does not establish them, and the study does not lean on it to. Separate genetic tests carry that load: in Drosophila melanogaster, kraken mutants are more sensitive to octanoic acid while Alkbh7 overexpression raises resistance, and mutating either gene in Drosophila sechellia, the noni specialist that tolerates the toxin, lowers that tolerance. Three species, two directions of manipulation and a loss-of-function result in the specialist itself are worth more than any number of coincident peaks. The eLife assessment reaches the same place from the other side, calling the findings important and the evidence solid.[1]
One limitation decides what a repeat has to do. The 93 isogenic strains came from a single wild population in California, and the authors say the base populations might therefore miss variants found elsewhere in the species' range. If an independent evolve-and-resequence run is seeded from wild flies outside that locality and put through the same escalating dose schedule, a published result by the end of 2027 showing the kraken region rising again under selection would mean the signal belongs to the species; a run in which only Alkbh7 returns, or neither, would mean the standing variation of one Californian sample carried it. Either way the finding beats again, or it does not, and that is the difference between a strong candidate and a mechanism.[1]