Fat or lean? The quantitative genetic basis for selection strategies of muscle and body composition traits in breeding schemes of rainbow trout (Oncorhynchus mykiss)

Quantitative genetic analyses of fish composition have been strongly biased towards lipid deposition, rather than protein deposition. This is partly at odds with desires of the modern aquaculture industry, to improve the efficiency of lean growth. Using a quantitative genetic approach, we examined the selection potential in both protein and lipid components of wet weight growth in rainbow trout over a two-year growth period. Two diet treatments were applied to test the hypothesis that an experimental, high protein, low lipid diet (HP) would enhance selection potential compared to the current modern, normal protein, high lipid diet (NP). We found that lipid traits (lipid body weight, percent muscle and body lipid; h=0.40) were more heritable than corresponding protein traits (protein body weight, percent muscle and body protein; h=0.18), indicating a higher selection potential for lipid traits. The results revealed further that breeding for improved lipid composition over the whole growth period is easier than for protein composition. This was shown by the high favourable genetic correlations between differently aged fish for lipid traits. In contrast, the respective correlations for protein traits were low or even negative. Similarly, the genetic correlations between muscle and body composition were higher for lipid than for protein, enhancing selection efforts to change lipid traits. Heritabilities increased with age, implying that selection practiced on old (N800 g) rather than young (b60 g) fish should bemore effective in achieving a compositional response. Although the diet had a significant effect on the composition traits, there was no general trend for diet differences in heritabilities of either protein or lipid traits. Thus, the hypothesis of increased selection potential on HP diet was not supported. In conclusion, lipid traits are both more variable and exhibit more favourable genetic architecture for selection compared to protein traits. © 2006 Elsevier B.V. All rights reserved.