New Brunswick E5B 2L9, Canada

Aquaculture 247 (2005) 211 217 www.elsevier.com/locate/aqua-online Development of an Atlantic salmon (Salmo salar) genetic improvement program: Genetic parameters of harvest body weight and carcass quality traits estimated with animal models Cheryl D. Quinton a, T, Ian McMillan a, Brian D. Glebe b a Centre for Genetic Improvement of Livestock, Department of Animal and Poultry Science, University of Guelph, Guelph, Ontario N1G 2W1, Canada b Fisheries and Oceans Canada, Aquaculture Division, St. Andrews Biological Station, 531 Brandy Cove Rd., St. Andrews, New Brunswick E5B 2L9, Canada Abstract The Atlantic Salmon Broodstock Development Program is a partnership of researchers and producers that aims to breed salmon with an optimal combination of fast growth rate, low incidence of early sexual maturation, and good carcass quality for commercial aquaculture in Atlantic Canada. Estimation of genetic parameters for these traits is an essential step in the development of this breeding program. Four year classes of Atlantic salmon were produced, each consisting of 48 93 full-sib families. Marked fish from each of these families were randomly distributed to several producers to be raised under commercial conditions. At harvest, gutted body weights and sexual maturation level were observed in 812 3471 individuals per year class at processing plants. Additionally, colour score, astaxanthin, canthaxanthin, fat and moisture contents were recorded on 472 immature individuals from one year class. Genetic parameters were estimated with single- and multiple-trait animal models. Body weight, astaxanthin, canthaxanthin, colour, fat and moisture all exhibited moderate heritabilities (0.1 0.2), indicating that these traits should respond to selection. Positive genetic correlations were found between body weight and the carotenoid pigments, colour, and fat. These results indicate that direct selection for weight may have favourable indirect responses of higher colour scores and pigmentation but also an undesirable increase in flesh fat content. A properly weighted selection index should therefore be used to select broodstock for increased harvest weight and flesh colouration, while controlling flesh fat content. D 2005 Elsevier B.V. All rights reserved. Keywords: Atlantic salmon; Salmo salar; Body weight; Carcass quality; Heritability; Correlation; Animal model; Selection 1. Introduction T Corresponding author: Tel.: +1 519 824 4120x58980; fax: +1 519 767 0573. E-mail address: cquinton@uoguelph.ca (C.D. Quinton). The Atlantic Salmon Broodstock Development Program (ASBDP) is a partnership of researchers and commercial producers based in St. Andrew s, 0044-8486/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.aquaculture.2005.02.030

212 C.D. Quinton et al. / Aquaculture 247 (2005) 211 217 New Brunswick, Canada that aims to develop a genetically-improved strain of Atlantic salmon for commercial aquaculture in Atlantic Canada. The goal of the breeding program is to develop salmon with an optimal combination of fast growth rate, good carcass quality, and low incidence of early sexual maturation. Under current New Brunswick government regulations, only Atlantic salmon descended from the wild population of the Saint John River, New Brunswick can be cultured; therefore, genetic improvement of cultured salmon must rely on selection and breeding within this strain. Genetic parameters for traits in this population, and specifically for the ASBDP stock, must therefore be estimated to predict how the ASBDP stock will respond to various selection and breeding techniques. Growth rate has generally been the trait of highest economic importance in meat livestock improvement programs and this is also the case in Atlantic salmon culture. Body weight at a fixed age has frequently been used as an indicator of growth rate in aquaculture and, additionally, body weight at harvest determines the price paid to growers by processors. Studies have generally found harvest-age (2 3.5 years) body weights in Atlantic salmon to have moderate heritability (Gunnes and Gjedrem, 1978; Gjerde and Gjedrem, 1984; Standal and Gjerde, 1987; Gjerde et al., 1994; Rye and Refstie, 1995) and are therefore suitable traits for improvement through genetic selection. As the Atlantic salmon industry has expanded, meat quality traits have become of increased interest to producers. Flesh colour is considered to be an indicator of salmon freshness and quality and processors and retailers will downgrade or even reject product with insufficient colour (Nickell and Springate, 2001). The distinctive pink red colour of salmonids is caused by carotenoid pigments in their diet. In current culture situations, pigments are usually added to feed in the form of astaxanthin (the most common form; Nickell and Springate, 2001) and canthaxanthin. These pigments are an expensive component of salmonid feeds so, to minimize costs, producers want animals that most efficiently absorb and retain these pigments in the flesh. Flesh carotenoid pigment content has thus also become a trait of interest for breeding programs. Family differences in flesh carotenoid pigment content have been observed in rainbow trout (Torrissen and Naevdal, 1984) and genetic parameters of carotenoid pigment contents were estimated in coho salmon fed a canthaxanthinsupplemented diet (Iwamoto et al., 1990). Flesh carotenoid pigment content genetic parameters have not been estimated in Atlantic salmon. Another carcass quality trait of increasing interest is flesh fat content because excessive fat content may have a detrimental effect on meat texture and also affect further processing; Gjedrem (1997) suggested that fat levels over 18% are too high. Genetic parameters of flesh colour and fat content have been previously estimated in Atlantic salmon (Gjerde and Gjedrem, 1984; Rye and Gjerde, 1996), coho salmon (Iwamoto et al., 1990; Neira et al., 2004), and rainbow trout (Gjerde and Schaeffer, 1989). In the present study, data were collected on harvest body weight from four ASBDP year classes. Carcass quality parameters were also measured in one year class. The objective was to estimate heritability and genetic and phenotypic correlations among harvest weight and five carcass quality traits: colour score, flesh contents of carotenoid pigments astaxanthin and canthaxanthin, fat, and moisture. These genetic parameters should help predict how the ASBDP stock will respond to various selection techniques. 2. Materials and methods 2.1. Family structures and husbandry The ASBDP began in 1998 as a continuation of the Salmon Genetic Research Program (SGRP) and four separate SGRP lines of Atlantic salmon formed the basis of the ASBDP population. Fertilisation of fouryear classes of salmon occurred in 1996, 1997, 1998 and 1999. In all year classes, random single-pair matings were done in November between four-yearold males and females to produce full-sib families. One strain within each year class was descended from separate strains produced by the SGRP (for strain histories, see Friars et al., 1995; O Flynn et al., 1992; O Flynn et al., 1999). In the 1998 and 1999 year classes, eyed eggs from single-pair matings of three commercial strains (C1, C2, and C3) were additionally raised under the same conditions. All SGRP and commercial strains were descended from wild Saint

C.D. Quinton et al. / Aquaculture 247 (2005) 211 217 213 John River, New Brunswick stock but relationships among parents were unknown. The final year class structures were as follows: 1996 year class contained 71 families from the SGRP 84JC strain; 1997 year class contained 48 families from the SGRP 89JC strain; 1998 year class contained 24 families from the SGRP 90JC strain, plus 9 families from the C1 strain and 60 families from the C2 strain; and 1999 year class contained 36 families from the SGRP 87JC strain, plus 35 families from the C2 strain and 7 families from the C3 strain. Incubation and early rearing occurred at the ASBDP hatchery in Chamcook, New Brunswick. Full-sib families were kept separate through to presmolt stage (approximately 16 months post-fertilisation), at which point individuals were marked by family with a combination of heat-branding and finclips. Smolts from each family were randomly distributed over 1 4 industry sea cages (Table 1) in the Bay of Fundy region and reared to harvest under typical commercial grow-out conditions. No grading or culling was done during the sea cage stage. Fish were harvested according to each producer s schedule, generally between 35 and 40 months post-fertilisation when animals reached an average weight of 5 kg (Table 1). Since each site within a year class was harvested at different times, the effects of age on harvest weight could not be reliably separated from environmental effects due to site. 2.2. Data collection In each year class, data were collected from fish at each site over 1 or 2 days (Table 1). At each producer s processing plant, all fish were gutted and then marked fish were identified and pulled from the processing line and given to ASBDP technicians for data collection. For each fish, technicians recorded brand and clip information, harvest body weight, and sexual maturity. Sexually mature salmon were identified by the presence of external secondary sexual characteristics (dark skin colour, presence of hooked jaw); this is the identification method typically used in commercial processing situations. Since fish were gutted prior to data collection, sex identification was unreliable and thus not recorded. In addition to the above data, tissue samples were taken from 472 immature individuals in the 1996 year class for carcass quality trait analysis. For each individual sampled, tissue weighing approximately 200 g was obtained from the area behind the right-side gill cover. Samples were bagged and identified to family (no record was taken to match weights with tissue samples). Tissue analysis was done at Moore- Table 1 Descriptive statistics for harvest body weight (HWT), flesh colour (COL), astaxanthin (AX), canthaxanthin (CX), fat, and moisture (MOI) data collected at harvest from salmon raised at sea cage sites within each of four year classes, including harvest dates and numbers of immature and mature animals observed Year class Site Harvest dates Trait n Immature n Mature Mean SD Min Max 1996 A 2000-07-05 HWT (kg) 776 36 6.70 1.50 1.77 11.25 COL (score) 472 0 26.20 0.79 21.63 28.78 AX (ppm) 472 0 6.62 0.68 3.50 8.66 CX (ppm) 472 0 6.50 0.88 2.29 9.04 Fat (%) 472 0 15.88 1.97 6.13 21.07 MOI (%) 472 0 64.88 1.83 59.82 74.07 1997 B 2000-11-13, 15 HWT (kg) 811 179 4.71 1.08 1.65 8.14 C 2001-03-25 HWT (kg) 1017 27 5.70 1.43 1.78 11.04 D 2001-04-01 HWT (kg) 730 2 5.25 1.07 0.80 7.80 E 2001-05-22 HWT (kg) 636 69 6.57 1.71 0.96 10.96 1998 F 2001-09-26, 27 HWT (kg) 697 302 4.05 1.32 0.64 7.28 G 2001-10-11 HWT (kg) 676 127 5.45 0.96 1.54 8.75 H 2002-03-15 HWT (kg) 626 126 4.74 0.78 1.35 7.37 1999 I 2002-10-16, 23 HWT (kg) 771 190 4.89 0.92 1.20 7.30 J 2003-02-26, 27 HWT (kg) 1922 210 4.56 1.10 0.80 8.00

214 C.D. Quinton et al. / Aquaculture 247 (2005) 211 217 Clark in St. Andrews, New Brunswick. Each sample was divided into two equally sized replicates that were analysed separately. Each replicate was visually scored for colour by 2 technicians according to the Roche Colour Fan and the two scores were averaged to give one record per replicate. Near infrared spectroscopy analysis was done to determine tissue replicate concentrations of astaxanthin, canthaxanthin, and wet weight percentages of fat and moisture. 2.3. Data analysis Descriptive statistics for data collected at each site are summarised in Table 1. Variance components were estimated with Average Information Restricted Maximum Likelihood (AI-REML) in the software package DMU Version 6 (Madsen and Jensen, 2002). The four unconnected year classes were analysed separately. Harvest body weight was analysed with the singletrait animal models: y ijkl ¼ l þ P i þ S j þ M k þ A l þ e ijkl ð1þ where y ijkl is the observation of harvest gutted body weight in animal l, l is the population mean, P i is the fixed effect of production site i, S j is the fixed effect of strain j, M k is the fixed effect of maturation class k (immature or mature), A l is the random animal effect of animal l, and e ijkl is the random residual error for animal l. In the 1996 and 1997 year classes there was only one strain represented, so the strain effect was removed from the model for those year classes. Harvest body weight and carcass traits in immature fish from the 1996 year class were also analysed with the following multiple-trait animal model: y ij ¼ l i þ A ij þ e ij ð2þ where i represents the traits harvest body weight, astaxanthin content, canthaxanthin content, colour score, fat content, and moisture content, y ij is the observation of trait i for animal j, l i is the population mean for trait i, A ij is the random animal effect of trait i for animal j, and e ij is the random residual error for trait i for animal j. The AI-information matrix for Eq. (2) was not positive definite (probably because harvest weight and carcass quality traits were measured on different animals), so standard errors of genetic parameters could not be calculated. Eq. (2) was also used to analyse only the carcass quality traits and this did allow estimation of standard errors. 3. Results and discussion 3.1. Trait heritability Harvest body weights showed low to moderate heritability estimates ranging from 0.1 to 0.2 over all year classes (Table 2). These estimates were somewhat lower than heritability estimates found in other studies, which generally average 0.2 0.3 for Atlantic salmon 2 3 years of age (Gunnes and Gjedrem, 1978; Gjerde and Gjedrem, 1984; Standal and Gjerde, 1987; Gjerde et al., 1994; Rye and Refstie, 1995). Colour score had slightly lower heritability than body weight, at about 0.14 (Table 2). This result follows trends of other studies in which Atlantic salmon flesh colour has lower heritability than growth traits. Gjerde and Gjedrem (1984) estimated heritability of colour score to be quite low at 0.03. Rye and Gjerde (1996) reported fairly low heritabilities (0.09 and 0.16) for flesh colour when measured on an observed scale, but when values were transformed to a Table 2 Variance component estimates for harvest body weights in four year classes as estimated by single-trait animal models and harvest body weight and flesh astaxanthin, canthaxanthin, colour, fat and moisture in one year class as estimated by multiple-trait animal models Year class Trait 2 r A 2 r E 2 r P h 2 FSE 1996 a Harvest weight 0.31 1.75 2.06 0.15F0.05 1996 b Harvest weight 0.33 1.76 2.09 0.16 Colour 0.09 0.54 0.63 0.14 Astaxanthin 0.05 0.41 0.46 0.10 Canthaxanthin 0.09 0.69 0.78 0.12 Fat 0.73 3.16 3.89 0.19 Moisture 0.65 2.70 3.35 0.19 1996 b Colour 0.08 0.55 0.63 0.13F0.07 Astaxanthin 0.04 0.42 0.46 0.09F0.06 Canthaxanthin 0.08 0.69 0.78 0.11F0.06 Fat 0.74 3.15 3.89 0.19F0.08 Moisture 0.65 2.70 3.35 0.19F0.08 1997 a Harvest weight 0.34 1.31 1.65 0.21F0.04 1998 a Harvest weight 0.12 1.01 1.12 0.10F0.02 1999 a Harvest weight 0.15 0.88 1.03 0.15F0.03 a Single-trait animal model. b Multiple-trait animal model.

C.D. Quinton et al. / Aquaculture 247 (2005) 211 217 215 liability scale, heritabilities were slightly higher (0.12 and 0.18). The carotenoid pigments showed the lowest heritabilities of all traits; astaxanthin and canthaxanthin heritabilities were approximately 0.10 and 0.12, respectively (Table 2). These estimates are considerably lower than carotenoid level heritabilities of 0.3 and 0.5 estimated by Iwamoto et al. (1990) in coho salmon. In the study by Iwamoto et al. (1990), the diets were supplemented with canthaxanthin, so those results may not be comparable to the ones in the present study. The feed at the grow-out sites in the present study was supplemented with astaxanthin, which is the carotenoid pigment most used in current Atlantic salmon production (Nickell and Springate, 2001). Flesh fat and moisture contents had the highest heritability estimates of the carcass quality traits, both at approximately 0.19 (Table 2). The fat content heritability was lower than fat content heritabilities of 0.30 and 0.38 reported by Rye and Gjerde (1996) in Atlantic salmon but similar to fat content heritability estimates of 0.17 and 0.26 reported by Neira et al. (2004) in coho salmon. Overall, the heritability estimates indicate that traits that the ASBDP may be interested in improving harvest body weight, flesh colour, carotenoid pigment content, and fat content should show reasonable response to direct selection. It is important to note, however, that since the ASBDP year classes contained full-sib families only, dominance and maternal effects could not be separated from additive genetic effects. Therefore, the heritability estimates reported here represent upper limits (Falconer and Mackay, 1996). The magnitudes of the calculated values tend to be smaller than those found in some other studies, but the reasons for this are unknown. Multiple grow-out sites with differing environmental conditions, leading to additional environmental variance, may have caused lower than expected heritability estimates, but such site differences should have been largely accounted for by site fixed effects in the animal models. However, site environmental effects were confounded with age at harvest, so there may be additional age effects acting on phenotype. The estimates may also have been affected by factors that could not be recorded at harvest, such as sex effects, or errors in family assignment (due to mis-reading brands and fin clips) and stage of sexual maturation. 3.2. Trait correlations Genetic correlations between harvest body weight and carcass quality traits were mostly positive, with the exception of the negative genetic correlation Table 3 Trait correlations (100) among harvest weight and carcass quality traits estimated from multiple-trait animal models containing all traits (first row) and carcass quality traits only (second row). Genetic correlations (FSE where calculated) are above the diagonal and phenotypic correlations are below the diagonal Harvest weight Colour Astaxanthin Canthaxanthin Fat Moisture Harvest weight 49 73 23 31 32 Colour 7 67 11 59 60 65F29 18F46 61F31 61F30 Astaxanthin 9 50 17 67 65 50 8F48 73F30 70F32 Canthaxanthin 3 54 54 7 12 54 54 7F37 12F36 Fat 5 6 31 12 100 6 31 12 100F0 Moisture 6 11 27 8 99 11 27 8 99

216 C.D. Quinton et al. / Aquaculture 247 (2005) 211 217 between harvest weight and flesh moisture content (Table 3). Body weight was most highly correlated with flesh astaxanthin level (0.73), followed by colour (0.49), moisture content ( 0.32) and fat contents (0.31). These results are similar to those of Rye and Gjerde (1996) who estimated genetic correlation of gutted body weight with flesh colour to be 0.31 and with fat content to be 0.42. Neira et al. (2004) found a higher genetic correlation (0.73) between body weight and fat content in coho salmon. Colour, astaxanthin, and fat are therefore likely to show some indirect response to direct selection on harvest body weight. Canthaxanthin content had the lowest genetic correlation with harvest weight (0.23), so this trait would probably have less indirect response to selection on harvest body weight. A higher genetic correlation (0.6) was estimated between weight and carotenoid levels in coho salmon (Iwamoto et al., 1990) but, as discussed above, that study may not be comparable to the present one. An important limitation of the correlations estimated between weight and carcass quality traits in this study is that there were no animals with both weight and quality records. Therefore, all environmental covariances between weight and quality traits were set to zero and the corresponding phenotypic correlations shown were also close to zero (Table 3). In a preliminary study with this data, phenotypic correlations among the traits were examined with multivariate analysis of variance and only the phenotypic correlation between weight and astaxanthin level was significantly different from zero. The limitation of these results indicates the importance of measuring both weight and carcass quality traits in some individuals when conducting this type of study. Genetic correlations among the carcass quality traits were similar in both multiple-trait analyses (Table 3). Flesh colour, astaxanthin content, and fat content had moderate positive genetic correlations ranging from 0.59 to 0.67. Rye and Gjerde (1996) found a very different genetic correlation of 0.82 between fat content and colour, although the corresponding phenotypic correlation (0.03) was similar to the one estimated in this study. Moisture content had moderate negative genetic correlations with colour and astaxanthin and a very high negative genetic correlation with fat content, which was expected due to the inverse relationship between fat and moisture. There was concern that the almost perfect correlations between moisture and fat (Table 3) could have affected the other covariances calculated but, when the same model was run without moisture data, very similar results were found (not presented) so this did not appear to be a problem. Flesh canthaxanthin content had very little to no genetic correlation with the other quality traits. The correlations found in this study are mostly favourable for the ASBDP breeding program because they indicate that direct selection for harvest body weight will have indirect responses to improve flesh colour and carotenoid pigment content. There is, however, an unfavourable positive genetic correlation between harvest body weight and flesh fat content. This result has also been found in other studies (Rye and Gjerde, 1996; Neira et al., 2004) and predicts that selection for high harvest weight could lead to higher fat content, an undesirable characteristic. These findings support a warning by Alderson (2001) that a selection program that concentrates on growth rate but neglects flesh quality traits could lead to undesirable product quality. To counteract this correlated response, a selection index should be developed that includes negative economic weighting on flesh fat content. This approach has been taken in other Atlantic salmon selection programs (Gjedrem, 2000). Since the genetic correlation between harvest body weight and fat content is fairly low, the population should contain broodstock with good estimated breeding values for all traits of interest. 4. Conclusions Many Atlantic salmon traits are economically important to aquaculture and the ASBDP aims to create a strain with an optimal combination of fast growth rate, good flesh quality, and low incidence of early sexual maturation. The current study indicates that direct selection for weight in the ASBDP stock will have favourable indirect responses of higher colour scores and pigmentation but also an undesirable increase in flesh fat content. Future studies will examine the economic values of these traits, in order to develop a broodstock selection index that simultaneously improves harvest weight and flesh colour, while controlling flesh fat content.

C.D. Quinton et al. / Aquaculture 247 (2005) 211 217 217 Acknowledgements The authors would like to acknowledge the work and assistance of staff from the ASBDP and Skretting Canada (formerly Moore-Clark), Dr. L. R. McKay, Dr. L. R. Schaeffer, and the following companies: Aqua Fish Farms Ltd., Cooke Aquaculture Ltd., Harbour de Loutre Products Ltd., Heritage Salmon Ltd., Jail Island Aquaculture Ltd., and Stolt Sea Farms Inc. References Alderson, R., 2001. The potential impact of genetic change on harvest and eating quality of Atlantic salmon. In: Kestin, S.C., Warriss, P.D. (Eds.), Farmed Fish Quality. Fishing News Books, Malden, MA, pp. 137 144. Falconer, D.S., Mackay, T.F.C., 1996. Introduction to Quantitative Genetics. 4th ed. Prentice Hall, Toronto, ON. 464 pp. Friars, G.W., Bailey, J.K., O Flynn, F.M., 1995. Applications of selection for multiple traits in cage-reared Atlantic salmon (Salmo salar). Aquaculture 137, 213 217. Gjedrem, T., 1997. Flesh quality improvement in fish through breeding. Aquaculture International 5, 197 206. Gjedrem, T., 2000. Genetic improvement of cold-water fish species. Aquaculture Research 31, 25 33. Gjerde, B., Gjedrem, T., 1984. Estimates of phenotypic and genetic parameters for carcass traits in Atlantic salmon and rainbow trout. Aquaculture 36, 97 110. Gjerde, B., Schaeffer, L.R., 1989. Body traits in rainbow trout: II. Estimates of heritabilities and of phenotypic and genetic correlations. Aquaculture 80, 25 44. Gjerde, B., Simianer, H., Refstie, T., 1994. Estimates of genetic and phenotypic parameters for body weight, growth rate, and sexual maturity in Atlantic salmon. Livestock Production Science 38, 133 143. Gunnes, K., Gjedrem, T., 1978. Selection experiments with salmon: IV. Growth of Atlantic salmon during two years in the sea. Aquaculture 15, 19 33. Iwamoto, R.N., Myers, J.M., Hershberger, W.K., 1990. Heritability and genetic correlations for flesh coloration in pen-reared coho salmon. Aquaculture 86, 181 190. Madsen, P., Jensen, J., 2002. A package for analysing multivariate mixed models. Danish Institute of Agricultural Sciences (DIAS), Denmark. Neira, R., Lhorente, J.P., Araneda, C., Diaz, N., Bustos, E., Alert, A., 2004. Studies on carcass quality traits in two populations of Coho salmon (Oncorhynchus kisutch): phenotypic and genetic parameters. Aquaculture 241, 117 131. Nickell, D.C., Springate, J.R.C., 2001. Pigmentation of farmed salmonids. In: Kestin, S.C., Warriss, P.D. (Eds.), Farmed Fish Quality. Fishing News Books, Malden, MA, pp. 58 75. O Flynn, F.M., Friars, G.W., Bailey, J.K., Terhune, J.M., 1992. Development of a selection index to improve market value of cultured Atlantic salmon (Salmo salar). Genome 35, 304 310. O Flynn, F.M., Bailey, J.K., Friars, G.W., 1999. Responses to two generations of index selection in Atlantic salmon (Salmo salar). Aquaculture 173, 143 147. Rye, M., Refstie, T., 1995. Phenotypic and genetic parameters of body size traits in Atlantic salmon Salmon salar L. Aquaculture Research 26, 875 885. Rye, M., Gjerde, B., 1996. Phenotypic and genetic parameters of body composition traits and flesh colour in Atlantic salmon Salmo salar L. Aquaculture Research 27, 121 133. Standal, M., Gjerde, B., 1987. Genetic variation in survival of Atlantic salmon during the sea-ranching period. Aquaculture 66, 197 207. Torrissen, O.J., Naevdal, G., 1984. Pigmentation of salmonids genetical variation in carotenoid deposition in rainbow trout. Aquaculture 38, 59 66.