Genotype-by-environment interaction in production traits of yellowtail kingfish (Seriola lalandi) reared in RAS and flow through systems

Seriola lalandi (Yellowtail Kingfish or Haku in te reo Māori) is emerging as an ideal species for aquaculture in New Zealand. The majority of S. lalandi production occurs in sea cages, however there is increased interest in using Recirculating Aquaculture Systems (RAS) that provide increased environmental control in the face of climate change, improved biosecurity, reduced water usage, and better waste capture. It is currently unknown whether a breeding program designed to generate genetic improvement in RAS would produce similar genetic gain in sea cages or if genotype-by-environment (GxE) interactions would limit the transfer of genetic improvement between growing systems. In this study we investigate the GxE interaction for both harvest length and harvest weight between S. lalandi reared in land-based RAS and flow through (FT) systems. We used genotyping-by-sequencing (GBS) SNP genotypes to estimate the genomic relationship matrix (GRM) for all individuals and used the GRM to estimate the genetic correlation of the same trait between the two systems. We found genetic correlations of 0.57 ± 0.1 and 0.54 ± 0.1 for harvest length and harvest weight respectively between the two systems. These lower-to-moderate genetic correlations indicate moderate-to-strong GxE because genotypes rank differently across the two systems for the same trait. We hypothesize that this GxE is primarily due to the large temperature and stocking density differences between the RAS and flow through systems. We also aimed to assess the relationship between the level of GBS SNP coverage and the standard errors of our genetic correlation estimates to determine possibilities for developing low or moderate density genotyping methods to reduce genotyping costs for future GxE experiments. Reducing the number of SNPs by 98.3 % from 167 K to 2.8 K only increased the standard error of the genetic correlations between environments from 0.10 to 0.13 in harvest weight and from 0.10 to 0.12 in harvest length. Therefore, it should be possible to reduce per-sample sequencing costs for future GxE experiments by using low or moderate density genotyping methods. • Harvest weight and length show moderate GxE between flow through and RAS systems. • More re-ranking occurs between individuals with moderate compared to extreme EBVs. • Standard errors of genetic correlations only slightly increased when using fewer SNPs.