Structural variants (SVs) are a major source of genetic and phenotypic variation, but remain challenging to accurately type and are hence poorly characterized in most species. We present an approach for reliable SV discovery in non-model species using whole genome sequencing and report 15,483 high-confidence SVs in 492 Atlantic salmon (Salmo salar L.) sampled from a broad phylogeographic distribution. These SVs recover population genetic structure with high resolution, include an active DNA transposon, widely affect functional features, and overlap more duplicated genes retained from an ancestral salmonid autotetraploidization event than expected. Changes in SV allele frequency between wild and farmed fish indicate polygenic selection on behavioural traits during domestication, targeting brain-expressed synaptic networks linked to neurological disorders in humans. This study offers novel insights into the role of SVs in genome evolution and the genetic architecture of domestication traits, along with resources supporting reliable SV discovery in non-model species.

Interbreeding between domesticated and wild animals occurs in several species. This gene flow has long been anticipated to induce genetic changes in life-history traits of wild populations, thereby influencing population dynamics and viability. Here, we show that individuals with high levels of introgression (domesticated ancestry) have altered age and size at maturation in 62 wild Atlantic salmon Salmo salar populations, including seven ancestral populations to breeding lines of the domesticated salmon. This study documents widespread changes to life-history traits in wild animal populations following gene flow from selectively bred, domesticated conspecifics. The continued high abundance of escaped, domesticated Atlantic salmon thus threatens wild Atlantic salmon populations by inducing genetic changes in fitness-related traits. Our results represent key evidence and a timely warning concerning the potential ecological impacts of the globally increasing use of domesticated animals.

Norway is the world’s largest producer of farmed Atlantic salmon and is home to ∼400 rivers containing wild salmon populations. Farmed escapees, a reoccurring challenge of all cage-based marine aquaculture, pose a threat to the genetic integrity, productivity, and evolutionary trajectories of wild populations. Escapees have been monitored in Norwegian rivers since 1989, and, a second-generation programme was established in 2014. The new programme includes data from summer angling, autumn angling, broodstock sampling, and snorkelling surveys in >200 rivers, and >25 000 scale samples are analysed annually. In 2014–2017, escapees were observed in two-thirds of rivers surveyed each year, and between 15 and 30 of the rivers had >10% recorded escapees annually. In the period 1989–2017, a reduction in the proportion of escapees in rivers was observed, despite a >6-fold increase in aquaculture production. This reflected improved escape prevention, and possibly changes in production methods that influence post-escape behaviour. On average, populations estimated to experience the greatest genetic introgression from farmed salmon up to 2014 also had the largest proportions of escapees in 2014–2017. Thus, populations already most affected are those at greatest risk of further impacts. These data feed into the annual risk-assessment of Norwegian aquaculture and form the basis for directing mitigation efforts.

1Structural variants (SVs) are a major source of genetic and phenotypic variation, but remain challenging to 2 accurately type and are hence poorly characterized in most species. We present an approach for reliable SV 3 discovery in non-model species using whole genome sequencing and report 15,483 high-confidence SVs in 4 492 Atlantic salmon (Salmo salar L.) sampled from a broad phylogeographic distribution. These SVs 5 recover population genetic structure with high resolution, include an active DNA transposon, widely affect 6 functional features, and overlap more duplicated genes retained from an ancestral salmonid 7 autotetraploidization event than expected. Changes in SV allele frequency between wild and farmed fish 8 indicate polygenic selection on behavioural traits during domestication, targeting brain-expressed synaptic 9 networks linked to neurological disorders in humans. This study offers novel insights into the role of SVs 10 in genome evolution and the genetic architecture of domestication traits, along with resources supporting 11 reliable SV discovery in non-model species. 12Main 13Modern genetics remains primarily focused on single nucleotide polymorphism (SNP) analyses, with a 14 growing recognition of the importance of larger structural variants (SVs) including inversions, insertions, 15 deletions and copy number variations (CNVs) (defined here as variants ≥100 bp), among others 1 . SVs 16 affect a larger proportion of bases in human genomes than SNPs 4 , are not always reliably tagged by SNPs 5 , 17 more frequently have regulatory impacts 6 , and have been shown to alter the structure, presence, number, 18 dosage, and regulation of many genes 1 . Nonetheless, SVs remain challenging to accurately type using 19 whole genome sequence data 2-3 , limiting our understanding of their biological roles and exploitation as 20 genetic markers. Consequently, there is a need for reliable SV detection approaches to fully exploit the fast-21 accumulating genome sequencing datasets in both model and non-model species, allowing for more 22 complete genetics investigations. Many tools exist for SV discovery using short-read sequencing data, but 23 all suffer from high false discovery rates (10-89%) 2,3,7 . This poses a challenge for truly de novo SV 24 detection in previously unstudied species lacking 'gold standard' reference SVs to help distinguish true 25 from false calls. Most studies rely on combining an ensemble of signals from different SV detection 26 methods, although this strategy does not reliably improve performance and can in some cases aggravate 27 false discovery 3 . Researchers therefore often apply independent experimental 8-9 or visualization methods 10 28 to validate a subset of SV calls. Overall, there remains an unsatisfactory lack of consensus on how to 29 validate the quality of de novo SV datasets in most species 3 . 31Salmonids have the highest combined economic, ecological and scientific importance among all fish 32 lineages, and have consequently been subject to hundreds of genetics stu...

Early and late life history and growth are affected by farmed salmon introgression across 105 wild populations of Atlantic salmon.