Optimising genomic breeding of farmed salmon

Abstract

The relatively recent domestication of aquaculture species, such as Atlantic salmon, produces both significant challenges and substantial opportunities for the use of genomic tools in animal breeding which could improve both production and welfare. The recent development of single-nucleotide polymorphism (SNP) genotyping arrays raises the possibility of substantially increasing rates of genetic improvement using genomic selection approaches. Genomic selection utilises computational analysis to combine genome-wide SNP data with trait information to identify individuals carrying the best of the naturally occurring genetic variation for desirable characteristics and hence select the best fish to breed future generations.

The project is a collaboration with Landcatch (Hendrix Genetics), a salmon breeding company in Scotland with an extensive experience in genetic breeding programmes, which provide data from several different populations of salmon that have been recorded for traits such as body weight per year, growth and sex. This thesis aimed to investigate the efficiency of genomic prediction within and between populations, by 1) exploring the structure of farmed salmon populations and their genetic diversity, 2) simulating a similar breeding scheme and investigating the impact of various mixing strategies on the accuracy of prediction from genomic evaluation across populations under random and directional selection, 3) estimating the heritability, based on both closely and distantly related individuals, by estimating pedigree-like and SNP-based heritability simultaneously in a single model and 4) investigating the genetic basis of sexual determination and differentiation of farmed Atlantic salmon populations in order to increase the efficiency of breeding programmes.

In Atlantic salmon, the broodstock populations follow a breeding programme characterised by discrete generations with a four-year generation interval. This results in the formation of up to four separate parallel lines (‘year groups’) with only limited connectivity between them. Our results, based on Landcatch populations, show a discreet genetic differentiation between these four populations. By simulating a similar breeding scheme, under random selection with no gene flow between populations (no mixing), the lines drift apart and there is no value in combining information across populations for genomic breeding value prediction versus only within line prediction. Only a small amount of mixing between lines brings the lines closer together and facilitates the use of information across lines to improve breeding value prediction. With directional selection, the average genetic merit and genetic gain increase with the amount of mixing and the loss of genetic variation decreases. Thus optimising gene flow between year groups should be an integral part of salmon breeding programme design. Estimates of the heritability, based on both closely and distantly related individual, show that the SNP-based relationships explain the majority of the genetic variance (90%) and higher SNP density resulted in greater heritability and genetic variance estimates, potentially as higher linkage disequilibrium captures the effect of more causative variants.

SNPs in different regions of the genome that are highly associated with sex determination highlight the evolutionarily fluid state of the sex determination system in salmon and improve our understanding for a trait which has both ecological interest and important consequences for the aquaculture production. This work underlines the value of genomic information to optimise genetic progress and dissecting the genetics of commercial traits and is of relevant to salmon and potentially other species in aquaculture.