In ¢sh, sexual maturation is often considered to be a problem because it perturbs growth and product quality. Therefore, it is common to select against early-maturing males in commercial breeding programmes (Gjedrem 2000). Genetic determinism of age at sexual maturation has been extensively studied in many species, and even some QTLs have been found (e.g. Gjerde 1986; Hankin, Nicholas & Downey 1993; Longalong, Eknath & Bentsen 1999; Kause, Ritola, Paananen, Mantysaari & Eskelinen 2003; Haidle, Janssen, Gharbi, Moghadam, Ferguson & Danzmann 2008). However, the eiect of sexual maturation on heritability estimates for other traits is not documented. In this paper, we present data collected during the beginningof sexualmaturationof a group of rainbow trout and showhow it biases heritability estimations. The ¢sh studied were issued froma full-factorial mating between two dams and 45 sires. Fish were all reared in the same tank since the eyed stage under a natural photoperiod and pedigrees were redrawn using10 microsatellites. Fishwere harvested in April at 17 months of age. They were killed on ice and several growth and quality traits were determined. Body weight was the main trait studied in this paper. The sex of the ¢sh was recorded by visual inspection of the gonads. For females, none of the individuals showed signs of maturation. For males, observers attempted to diierentiate non-maturing and maturing males but it turned out to be di⁄cult because we were at the very beginning of the maturation (this rainbow trout strain usually spawns in November^ December). However, we recorded gonads weight to enable calculation of the gonado-somatic index (GSI5100 gonads weight/body weight), which was used to determine which males were maturing or not.We studied several GSI thresholds: 0.1, 0.2, 0.3, 0.4 and 0.5. For each threshold, males with GSI above the threshold were considered to be maturing males. We studied sex eiect and heritability estimates for each threshold. Heritability was estimated using VCE5 (Groeneveld & Kovac 1990) with a sire model and sex and dam as ¢xed eiects. Dam was set as a ¢xed eiect because there were only two dams in our mating design, and therefore between-dams variance is of no interest. Two dams were used instead of one in order to avoid confusion of dominance eiects with additive genetic eiects in the between-sires variance. Moreover, this design was demonstrated to be e⁄cient for estimating heritabilities when the total number of oispring analysed is ¢xed (Dupont-Nivet, Vandeputte & Chevassus 2002). First, we analysed all the datasets with a sex eiect comprising three levels (female, male and maturing male). Second, maturing males were removed from the dataset and the sex effect was set to two levels (male and female). Using ‘FAP’ software (Taggart 2007), 87% of ¢sh could be attributed to their parents. Oispring from two sires with less than three oispring were removed from the data set. Four hundred and thirty-three animals were ¢nally analysed. The sex ratio was well equilibrated with 217 females, i.e., 50.1%. In Table 1, Aquaculture Research, 2010, 41, e878^e880 doi:10.1111/j.1365-2109.2009.02448.x
Read full abstract