| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
* MTT Agrifood Research Finland, Biotechnology and Food Research, Biometrical Genetics, FI-31600 Jokioinen, Finland;
and
Finnish Game and Fisheries Research Institute, Turku Game and Fisheries Research, FI-20520, Finland; and
Finnish Game and Fisheries Research Institute Jyväskylä, Survontie 9, FI-40500 Jyväskylä, Finland
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: feed efficiency • genetic selection • genotype-environment interaction • lipid • protein
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
). Therefore, current fish breeding programs attempt to improve efficiency indirectly through direct selection on growth (Gjedrem and Thodesen, 2005). In terrestrial farm animals, body composition traits such as lipid content have potential as indicator traits for selective improvement of feed efficiency (Pym, 1990; Archer et al., 1999), but genetic correlations of composition and feed utilization traits have not been well quantified in fish.
Selection programs must also account for changing aquaculture diets. Fish meal is considered to be the superior protein source for carnivorous fishes, but feed manufacturers are increasingly replacing fish meal with plant products (Jobling et al., 2001b). With novel diets there is potential for genotype-environment interactions that may hinder selection efforts to improve feed efficiency.
In this study, we measured individual whole-body lipid and protein composition in a farmed salmonid, European whitefish, reared in a split-family design with either fish meal or soybean meal-based diets. These data were combined with individual growth and feed utilization data from an earlier study (Quinton et al., 2007) to estimate diet-specific genetic parameters and predict responses to alternative selection strategies. The study objectives were to 1) estimate genetic and phenotypic correlations of lipid and protein percentages with growth, feed intake, and feed efficiency; 2) predict whether selection for reduced lipid will improve feed efficiency; and 3) evaluate the impact of genotype-environment interactions between fish meal- and soybean meal-based diets on body composition.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
European whitefish (Coregonus lavaretus L.) is a carnivorous salmonid. This species is farmed commercially, and a selective breeding program is currently under development. The experimental diets, research population, and growth and feed intake data collection for this study were described in detail by Quinton et al. (2007) and are thus only briefly outlined here.
Diet Formulations
Two isonitrogenous and isocaloric (based on gross energy) pellet diets were formulated (Table 1; Quinton et al., 2007). In the fish meal (FM) diet, fish meal supplied 100% of the dietary protein, whereas in the soybean meal (SBM) diet, 50% of the dietary protein was replaced with defatted soybean meal- (fat content 3.5%) derived protein. Amino acid (Evira, Helsinki, Finland) and phosphorus (Raisio Ltd., Raisio, Finland) compositions of the fish and soybean meals were analyzed, and the SBM diet was supplemented with methionine, lysine, and phosphorus to bring the amounts of these nutrients equal to those in the FM diet.
|
The whitefish in the experiment originated from Kokemäki River strain broodstock housed at the FGFRI Tervo Fisheries Research and Aquaculture Station. In October 2003, 45 sires and 52 dams were mated in a partial factorial design to create 70 full- and half-sib families. Each sire was mated to an average of 1.6 dams and each dam to an average of 1.3 sires. The sires and dams belonged to a base population in which individuals were assumed to be unrelated. At the eyed-egg stage, families were transported to the FGFRI Laukaa Research Station. From hatching until the start of the experiment, families were held in separate indoor tanks. Twenty-four individually tagged fish from each family were randomly sampled for the experiment.
The diet trial was conducted from July to October 2004. To construct a split-family design, each family was first split into 2 groups to be reared with the alternative diets. Each family group was evenly distributed over 6 replicate tanks per diet. Consequently, the experiment began with a total of 1,680 fish, each replicate tank containing 140 fish (2 fish from each family). Fish were fed 6 h/d using belt feeders. Feed was supplied in excess and rations were calculated by increasing the predicted feeding rates (Koskela, 1992) by 30%.
Recording of Growth, Feed Intake, and Efficiency
Individual BW were recorded at the beginning (initial BW, BWI, n = 1,647, 
= 40.64 g, SD = 10.82 g) and end (final BW, BWF) of the trial. During the trial, fish approximately tripled in weight. Individual daily gain (DG) was calculated as the difference between BWF and BWI, divided by the number of days in the trial.
Daily feed intake (g) was measured by X-radiography (Talbot and Higgins, 1983; reviewed by Jobling et al., 2001a) 5 times per individual, with 2-wk intervals between measurements. Repeatability of the 5 daily feed intake records was moderate within both diets (r = 0.28). When the average of the 5 records was calculated, the repeatability of the average was 0.66, reflecting reasonable recording accuracy (Kause et al., 2006a). Individual ADFI was calculated from the repeated measurements, and individual feed efficiency (G:F) was calculated as the ratio of DG to ADFI.
Recording of Whole-Body Proximate Composition
Final whole-body composition (lipid% and protein%, based on wet weight) was measured on all fish at the end of the trial. Fish were individually minced and stored at –20°C until analysis. For each individual, 2 samples were analyzed. Each body sample was homogenized (Losmixer, Miris AB, Uppsala, Sweden) in a standard solvent (Mirasolve, Miris AB, Uppsala, Sweden), and lipid and protein were determined using midinfrared transmission spectroscopy (FMA2001 Milk Analyzer, Miris AB), as described by Elvingsson and Sjauna (1992). Standard analytical methods for lipid (Folch et al., 1957) and nitrogen (Kjeldahl, 1883) were used to construct calibration curves between single wavelength absorption of infrared light and concentration of the target substance. For each fish, the 2 sample measurements were averaged to make 1 lipid% and 1 protein% observation per individual (nFM = 752 fish, nSBM = 753). Final individual lipid and protein weights were calculated as (lipid% x BWF) and (protein% x BWF), respectively.
Statistical Analysis of Diet Effects on Trait Means
Diet effects on mean lipid% and protein% were tested with ANOVA (MIXED procedure, SAS Inst. Inc., Cary, NC), and model:
 |
where yijkl is the phenotype for lipid% and protein% for individual l; Dieti is the fixed effect of diet i (i = 1 to 2); ExpTj(Dieti) is the random effect of replicate tank j (j =1 to 6 for each diet) nested within diet i; Famk is the random effect of full-sib family k (k = 1 to 70); Diet xFamik is the random effect of diet by full-sib family interaction; and eijkl is the random residual effect for individual l. For all traits, variance due to experimental tank-family interaction was zero and thus was excluded. The F-tests, their degrees of freedom, and SEM were calculated with the Kenward and Roger option. Diet effects on composition traits independent of BWI, BWF, and total feed intake (FI = ADFI x No. of feeding days) were also analyzed by adding these traits as covariates separately to the above model. When these adjustments were done, the measurements are termed "relative" (e.g., relative lipid%).
Genetic Analysis
To examine genotype-environment interactions, observations recorded under each diet treatment were treated as separate traits. For instance, lipid percentages recorded for FM (lipid%FM) and SBM diets (lipid%SBM) were defined as different traits. Phenotypic and genetic parameters of DG, ADFI, G:F, lipid%, and protein% were estimated using a multiple-trait animal model with DMU6 software, applying restricted maximum likelihood and average information methods (Madsen and Jensen, 2006). The animal model was
 |
where i represents the traits, yijkl is the observation of trait i for individual l, ExpTij is the fixed effect of experimental tank j on trait i, Famik is the random effect of full-sib family k on trait i, Ail is the random animal genetic effect of trait i for individual l, and eijkl is the random residual error for trait i for animal l. The complete relationship matrix accounted for the animal effect, while the full-sib effect was modeled using only full-sib relationships. The individual genetic effect included additive genetic effects, and parts of potential dominance and epistatic effects. The full-sib family effect, which contained (co)variances due to common environment of full-sibs before tagging, as well as potential maternal and dominance effects, was very weak (
4% of total phenotypic variance) for lipid% and protein%. Therefore, the full-sib family effect was dropped for lipid% and protein% models, but kept for DG, ADFI, and G:F models. Residual covariances between traits measured in different diets were set to zero. Heritability (h2) was calculated as the ratio of genetic variance to total phenotypic variance. To estimate genetic parameters for relative traits independent of BWI, BWF, and FI, these covariates were added separately to the above models.
Genetic parameters were not estimated for BWF or final lipid and protein weights. In preliminary analyses, an overly large proportion of the BWF variance, and thus lipid and protein weight variances, was attributed to the full-sib family effect. This caused heritability estimates near zero and impeded genetic correlation estimation for these traits.
Prediction of Genetic Responses to Selection
Selection index calculations were used to predict genetic responses to phenotypic selection (Hazel, 1943) for DG, ADFI, G:F, lipid%, and protein% in FM and SBM diets. To assess the influence of direct and indirect G:F selection on genetic response in G:F, 3 selection strategies were compared for each diet: (a) selection for DG; (b) selection for maximum G:F, where increased DG and decreased ADFI were selected simultaneously and selection index weights were set to obtain maximum response in G:F; and (c) selection for maximum G:F, where increased DG and decreased lipid% were selected simultaneously and selection index weights were set to obtain maximal G:F response. Strategy (a) was comparable to most current aquaculture breeding programs. Strategy (b) was selection for maximal G:F by direct selection, and (c) maximized genetic response in G:F by indirect selection using lipid% as an indicator trait.
These selection strategies do not refer to economically optimal ones, for which index weights would be derived from economic values. We recognize that the lipid selection strategy described is only theoretical because our whole-body composition measurement required fish to be killed, and thus, individual selection is impossible in practice. However, these predictions may approximate selection based on live lipid measurement (e.g., with fat meter).
Direct and correlated genetic responses to selection were calculated by R = i(b'G)(b'Pb)–1/2 where R is the vector of responses, i is intensity of selection (set to 1), b is the vector of relative index weights, which sum to 1, G is the genetic covariance matrix, and P is the phenotypic covariance matrix. The G and P resulted from the 8-trait genetic parameter estimation model for DGFM, ADFIFM, lipid%FM, protein%FM, DGSBM, AD-FISBM, lipid%SBM, and protein%SBM described above. The parameters for G:F were not needed because genetic change in G:F was obtained through genetic changes in DG and ADFI.
To generate the alternative selection strategies, relative index weights (b) were modified. To obtain strategy (a), the index weight for DGFM was set to 1, while the weights for the 3 remaining traits were 0. To find the index weights in 2-trait selection strategies (b) and (c), a range of index weights were tested, from all weight on DG to all weight against ADFI or lipid%. Index weights found for strategy (b) were bDG = 0.50, bADFI = –0.50, and for the remaining traits were 0. Index weights for strategy (c) were bDG = 0.88, blipid% = –0.12, and for the remaining traits were 0.
Genetic responses in the traits for both diets were calculated as percentage change relative to the original trait mean in the data. To calculate genetic response in G:F, mean G:F before selection was first calculated for each diet from the real data. Then, genetic responses to selection were calculated for DG and ADFI, and the new mean G:F after selection was calculated from these.
|
RESULTS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Diet effects and genetic parameters of BWF, DG, ADFI, and G:F for the population in this trial have been previously reported (Quinton et al., 2007). For readers’ convenience, these previous results are briefly listed here. Fish fed the FM diet had significantly greater BWF (means ± SE: FM = 131.2 ± 2.38, SBM = 125.8 ± 2.24 g), faster DG (FM = 1.145 ± 0.0212, SBM = 1.091 ± 0.0201 g/d), lower ADFI (FM = 0.948 ± 0.0236, SBM = 1.027 ± 0.0285 g/d), and higher G:F (FM = 1.208 ± 0.0150, SBM = 1.078 ± 0.0240) than the SBM group. Within both diets, DG (h2 = 0.26, 0.20) and ADFI (h2 = 0.23, 0.17) were moderately heritable, G:F had low heritability (h2 = 0.06, 0.07), and DG and ADFI had high positive phenotypic and genetic correlations (rp = 0.88, 0.86; rG = 0.97, 0.93). Between-diet genetic correlations for DG (rG = 0.97) and ADFI (rG = 0.93) were very high.
Lipid Percentage
Diet Effects.
Whitefish fed the FM diet had significantly greater mean lipid% (absolute difference = 2.2%) and lipid weight (difference = 3.5 g) than those fed the SBM diet (Table 2). Differences in lipid% between diets were also independent of BWI, BWF, and FI, as shown by significant differences when these traits were used as covariates (Table 2).
|
). Heritabilities remained the same or were reduced only slightly after adjustment for BW or FI. Lipid percentage showed no major diet differences in phenotypic coefficients of variation or heritability, and confidence intervals for all heritability estimates overlapped considerably between diets (Table 3).
|
). Similarly, lipid% had moderate to high positive genetic correlations with DG and ADFI (0.64 to 0.88). Correlations of lipid% with relative DG and relative ADFI (adjusted for BWI) were slightly lower compared with the correlations with absolute DG and ADFI. Phenotypic and genetic correlations of lipid% with DG and ADFI were of similar magnitude in both diets; phenotypic correlations were slightly higher in the SBM diet, but genetic correlations were slightly higher in the FM diet and confidence intervals overlapped. Phenotypic correlations of lipid% and G:F were close to zero (0.03 and –0.06, Table 4). Genetic correlations of lipid% and G:F differed in the 2 diets: a weak positive correlation in FM, and a weak negative correlation in SBM. Both of these estimates, however, had confidence intervals that overlapped.
|
Protein Percentage
Diet Effects.
Protein weights did not differ significantly between diets (Table 2
). Similarly, diet differences in absolute and relative mean protein% were small, and the statistical tests reached only marginally significant levels.
Genetic Variation.
Absolute and relative protein% heritabilities were low (0.05 to 0.07), and phenotypic coefficients of variation and heritabilities were similar in both diets (Table 3).
Correlations.
Phenotypic correlations between protein% and absolute or relative DG and ADFI were close to zero (0.03 to 0.09, Table 4). Similarly, protein% showed weak genetic correlations with absolute or relative DG and ADFI (–0.04 to 0.33). The phenotypic correlation between protein% and G:F was near zero for both diets (0.00 and 0.03). The respective genetic correlation was negative in the FM diet, but positive in the SBM diet; however, these estimates had very large standard errors. Protein percentage and lipid% had weak positive phenotypic (0.2 and 0.3) and genetic (0.27 and 0.16) correlations.
Between-diet genetic correlations for absolute and relative protein% were high and positive, though all had large standard errors (absolute: 0.77 ± 0.500; relative to BWF: 0.89 ± 0.532; relative to FI: 0.89 ± 0.530).
Genetic Responses to Selection
Direct selection for G:F (simultaneous selection on DG and ADFI) was the best strategy to improve G:F, followed by indirect selection using lipid% as an indicator (simultaneous selection on DG and lipid%). Selection on DG alone was the least effective strategy for increasing G:F.
Selection solely for high DG caused correlated genetic increases in both G:F (absolute responses 0.49 and 0.46%) and lipid% (2.7 and 1.7%) in both diets (Table 5). As in our previous study (Quinton et al., 2007), the high positive genetic correlation between DG and ADFI caused correlated increases in ADFI as well, but because ADFI had lower heritability than DG, ADFI increased proportionally less than DG, thus increasing G:F.
|
). Increased G:F was associated with slightly increased lipid% in the FM diet and slightly decreased lipid% in the SBM diet, in accordance with the estimated genetic correlations between G:F and lipid% (Table 4). Because ADFI had a positive genetic correlation with lipid%, simultaneous selection on DG and ADFI controlled lipid%. Compared with the DG strategy, simultaneous selection on DG and ADFI caused only one-third as much increase in lipid% in the FM diet and resulted in a very small decrease in lipid% in the SBM diet.
Simultaneous selection for DG and against lipid% for maximum G:F increased the genetic response in G:F by factors of 1.49 and 1.55 in the FM and SBM diets, respectively, compared with the DG strategy (Table 5). The positive genetic correlation between lipid% and ADFI in this case reduced gains in ADFI, thus increasing G:F. Nevertheless, genetic response in G:F remained low (e.g., in the FM diet, DG and DG-lipid% strategies resulted in G:F genetic response of 0.49 and 0.73%, respectively). In the FM diet, selection on DG and lipid% resulted in a small increase in lipid%, whereas in the SBM diet this strategy decreased lipid%. Therefore, although lipid% was unfavorably or weakly associated with G:F in the FM and SBM diets, respectively, simultaneous selection for rapid growth and against body lipid content did increase G:F.
Protein percentage had small genetic responses to all selection strategies; the largest response was <0.4% (Table 5).
|
DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
; Kolstad et al., 2004; Kause et al., 2006b). Body composition traits such as lipid and protein contents have traditionally been of interest in fish breeding programs because of their association with product yield and quality. However, composition traits also have potential use as indicators for selective improvement of feed efficiency (reviewed by Pym, 1990; Archer et al., 1999). Until recently, quantification of the genetic relationships of feed utilization and body composition traits in fish has been challenging because of the difficulty and expense of measuring individual feed intake in large numbers of animals (Kause et al., 2006a). These difficulties were overcome in the current study by measuring individual daily feed intake with X-radiography. In the current study, we estimated genetic correlations of whole-body lipid% and protein% with feed utilization traits to examine potential physiological mechanisms causing variation in G:F and to assess these measurements as indicator traits for selective improvement of G:F in European whitefish. The first major finding of this study was that simultaneous selection for rapid growth and reduced body lipid% led to greater genetic response in G:F compared with selection for rapid growth alone.
In fish breeding programs, selective improvement of feed utilization must also account for changing aquaculture diets. Fish meal is considered to be a superior protein source in many livestock diets and is currently the major component of diets for carnivorous fish species such as salmonids (Jobling et al., 2001b; New and Wijkström, 2002; Tacon, 2004). However, there are considerable ecological and economic motivators to reduce fish meal use (e.g., Naylor et al., 2000; FAO Fisheries Department, 2004). Research into replacing fish meal with plant products such as soybean has been carried out since the 1970s (e.g., Cho et al., 1974; Fowler, 1980). Presently, commercial fish feed manufacturers are increasing the use of plant products (reviewed by Storebakken et al., 2000; Jobling et al., 2001b), and near-future diets are expected to be vegetarian (Powell, 2003). In the current study, we also examined the potential for related genotype-environment interactions on feed utilization and body composition traits. The second major result of this study was that there were only weak genotype-environment interactions for body composition, and thus, changing from fish meal-based to soybean meal-based diets was predicted to have only a minor influence on fish breeding programs.
Lipid Percentage
Genetic Variation. Whole-body lipid% showed moderate heritability within both diet environments. Furthermore, heritabilities for lipid% scaled to a common FI remained moderate. Therefore, genetic family differences in body lipid content persisted after adjusting to a common total amount of feed consumed, indicating genetic variation in lipid digestion or metabolism.
The estimates in this study (0.12 to 0.22) were smaller than whole-body lipid% heritabilities found in larger rainbow trout (0.32 to 0.59, Tobin et al., 2006) and broad-sense heritability estimated in juvenile rainbow trout (0.47, Kinghorn, 1983). Most previous estimates for lipid percentage have been calculated from abdominal, fillet, or muscle lipid measurements in adult or harvest-size animals because such measures are economically important traits. Lipid deposits at different body locations are genetically different traits (Gjerde and Schaeffer, 1989; Kause et al., 2002; Tobin et al., 2006), and therefore, direct comparison between composition heritabilities for whole-body and specific tissues is not straightforward. However, the estimates in the current study were within typical heritability ranges for abdominal or visceral lipid content (0.25 to 0.30, Gjerde and Schaeffer, 1989; 0.22, Rye and Gjerde, 1996; 0.03 to 0.28, Kause et al., 2002; 0.10 to 0.41, Neira et al., 2004) and flesh lipid percentage (0.47, Gjerde and Schaeffer, 1989; 0.18, 0.19, Iwamoto et al., 1990; 0.30, Rye and Gjerde, 1996; 0.16, Kause et al., 2002; 0.17, 0.26, Neira et al., 2004; 0.19, Quinton et al., 2005).
Correlations with Growth and Feed Utilization.
Daily gain had moderate positive genetic and phenotypic correlations with whole-body lipid%. Therefore, in selection index calculations, selection for increased DG also increased lipid%. The correlation estimates in this study were comparable with that found in juvenile rainbow trout (0.47, Kinghorn, 1983). Other published correlations of growth traits and lipid percentage are mostly positive but vary widely depending on the measurement made, age, and species studied (Gjerde and Schaeffer, 1989; Iwamoto et al., 1990; Rye and Gjerde, 1996; Kause et al., 2002; Neira et al., 2004; Quinton et al., 2005; also reviewed by Gjedrem, 1997). Correlated increase in lipid in response to selection on growth is likely to be detrimental to product quality and yield.
Average daily feed intake had moderate to high positive genetic and phenotypic correlations with body lipid%. Therefore, selection against lipid% indirectly reduced ADFI. Likewise, selection index calculations showed that simultaneous selection on DG and ADFI had a correlated response of restraining increases in lipid% by one-third compared with selection for high DG alone.
Genetic relationships of body composition with feed intake have rarely been studied in fish. To our knowledge, the only comparable estimate is by Kinghorn (1983), who found food consumption had a negative genetic correlation (–0.40, based on family means) with lipid% in rainbow trout. Possibly, our results differed due to the feed intake measurement methodology; Kinghorn (1983) estimated feed intake from oxygen consumption. However, our results were consistent with estimates in terrestrial livestock species. In broiler chickens, genetic correlations of abdominal fat percentage and feed consumption traits have also been positive (0.24, Chambers et al., 1984; 0.55, Wang et al., 1991). Backfat and feed intake also have similar positive genetic correlations in pigs (average = 0.37, Clutter and Brascamp, 1998; 0.64, Johnson et al., 1999) and beef cattle (0.24, Schenkel et al., 2004), although Schenkel et al. (2004) found no significant genetic correlation of intramuscular fat with feed intake in beef cattle.
Feed efficiency and lipid% had a weak positive genetic correlation in the FM diet, a weak negative genetic correlation in the SBM diet, and phenotypic correlations were close to zero. These results were unexpected, and especially in the FM diet, appeared to disagree with previous studies. Terrestrial livestock literature suggests that fat traits have negative genetic correlations with G:F, and furthermore, lean animals may be more efficient because on a wet weight basis, fat tissue growth is 3 times more energetically costly than muscle growth (reviewed by Pym, 1990; Archer et al., 1999). Gain-to-feed ratio has shown negative genetic correlations with lipid% in rainbow trout (family mean phenotypic correlation = –0.15, Kinghorn, 1983) and with abdominal fat percentage in broilers (–0.68, Chambers et al., 1984; –0.70, Wang et al., 1991). Correspondingly, the inverse feed-to-gain ratio has shown positive genetic correlations with abdominal fat percentage in broilers (average = 0.46, Leenstra and Pit, 1988), and with backfat in pigs (average = 0.30, Clutter and Brascamp, 1998; 0.40, Johnson et al., 1999). Schenkel et al. (2004), however, found no significant genetic correlation of feed-to-gain ratio with backfat or intramuscular fat in beef cattle. The probable reason for the weak relationships between lipid% and G:F observed in the current study was the very low heritability of G:F as a ratio trait. We previously found that DG and ADFI in whitefish were each moderately heritable but very strongly correlated, which caused low variability in the G:F ratio and, thus, low heritability (Quinton et al., 2007). Therefore, in the current study, despite high correlations of lipid% with both DG and ADFI, lipid% may have had a weak relationship with the relatively invariable G:F ratio.
However, the selection index calculations revealed that multitrait selection for high DG and reduced lipid% considerably increased the correlated genetic response in G:F compared with selection for DG alone. In this strategy, the high positive correlation of lipid% with ADFI caused direct selection for reduced lipid% to indirectly reduce ADFI (the G:F denominator), thus increasing G:F. These index calculation results did not contradict the presented phenotypic and genetic correlations between lipid% and G:F. By including several traits in the selection index calculations, we revealed relationships among multiple traits that were not directly evident from pairwise correlations. This was the motivation for using index selection calculations in the current study.
Inclusion of lipid% as an indicator trait in a breeding program that already selects for rapid growth should therefore enhance genetic improvement of G:F. Selection for reduced fat as a means to improve G:F has already been proposed for fish breeding programs by Gjerde et al. (2002) and Gjedrem and Thodesen (2005). The selection index predicted genetic responses in G:F were evidently small, but would likely have substantial economic impact on a production scale (Kolstad et al., 2004).
To improve product quality and avoid increased lipid deposition in response to growth selection (Kause et al., 2007), fish breeding programs need to select on lipid stores measured from different body parts; these may be only weakly genetically correlated (Tobin et al., 2006). Lipid traits are already under selection in many breeding programs, and increased G:F is a likely by-product of growth and product quality improvement efforts. To compare the relative advantages of recording of body lipid or feed intake for G:F improvement, a cost-benefit analysis should be conducted. Currently, X-radiography is a research tool not automated for routine use. Other feed intake recording methods may be more practical, such as the use of marker ingredients added to feed that can be used to record long-term feed intake (Kause et al., 2006a).
Protein Percentage
Genetic Variation.
Whole-body protein% showed low heritability in both diets. This result was consistent with those of Gjerde and Schaeffer (1989) and Kause et al. (2002), who estimated flesh protein percentage heritabilities close to zero in adult rainbow trout. Protein levels typically show low variability among individuals and families (Shearer, 1994; Tobin et al., 2006). Therefore, body protein% in juvenile whitefish is not expected to respond strongly to direct selection. It is likely, however, that in larger fish heritability for protein% will be larger because in maturing fish, fillet weight percentage of wet BW shows moderate heritability (Kause et al., 2002; Neira et al., 2004; Rutten et al., 2005). Because the fillet is a major location for protein storage, genetic variation in fillet percentage tends to result in genetic variation in whole-body protein%, even when fillet protein percentage has zero heritability. Accordingly, Tobin et al. (2006) observed that heritability for body protein% in large (>2 kg) rainbow trout was as large as 0.39, although heritability for muscle protein percentage was simultaneously low.
Correlations With Growth and Feed Utilization.
Body protein% was less related to growth and feed utilization traits compared with body lipid%. Phenotypic and genetic correlations between protein% and DG were weakly positive under both diets. These results were similar to weak correlations observed in rainbow trout (0.10 to 0.12, Kause et al., 2002) and agreed with observations from diet comparisons that protein mass increases at a slower rate than lipid mass in response to increased wet weight mass (Shearer, 1994).
In the current study, ADFI phenotypic correlations with protein% were close to zero, as was the genetic correlation in the SBM diet, although the genetic correlation was positive in the FM diet. Feed efficiency phenotypic correlations with protein were near zero, and genetic correlations were either negative (FM diet) or positive (SBM diet), but with large standard errors. Due to low heritability and weak genetic correlations with other traits, protein% in whitefish did not respond substantially to selection strategies in the selection index calculations. No comparable estimates of genetic correlations of body protein traits with feed intake or feed efficiency were found in the scientific literature.
Protein percentage seems to be under strong internal control and maintained in stable homeostasis in fish (Shearer, 1994). This stability control may prevent the occurrence of strong correlations with growth and feed intake traits. Our results indicated that changing the lipid component of wet weight growth by selection would be easier than changing protein.
Whole-body lipid and protein were not competing growth strategies in whitefish, as evidenced by weakly positive, not negative, genetic and phenotypic correlations between lipid% and protein%. These may be characteristics of small fish, because Tobin et al. (2006) found that phenotypic and genetic correlations between whole-body protein% and lipid% were near zero or positive in younger rainbow trout (up to 19 mo postfertilization; BW 800 g), but became negative as fish reached heavier weights (31 mo; 2,200 g). This indicates changes in growth strategies of wet weight protein and lipid components as fish grow. Correlation estimates between protein and lipid components also vary with measurement type, as shown by the range of estimates in other studies (rP = –0.46, Gjerde and Schaeffer, 1989; rG = –0.58 to 0.34, Kause et al., 2002).
Novel Diet Impacts
Body Composition Trait Means.
In a previous study, we found that European whitefish fed the novel SBM diet had significantly slower growth than those fed the FM diet (Quinton et al., 2007). The current study revealed that this growth difference was largely caused by lower body lipid content, not lower protein content, with the SBM diet. Lower body lipid% in the SBM group was contrary to our expectation. We previously observed that whitefish had greater ADFI with the SBM diet and, therefore, greater energy intake (Quinton et al., 2007). Within a diet, fish with high feed intake also had high body lipid%. When comparing diet means, we therefore expected surplus nutrients from the SBM feed to be stored as fat (Pym, 1990; Shearer, 1994). However, this did not occur; rather, comparison of diet means revealed the opposite result of the within-population phenotypic and genetic correlations. The difference in lipid% was also found independent of FI. Therefore, the whitefish appeared to digest, synthesize, or deposit lipids less effectively when nutrients were derived from the SBM source. The SBM-derived nutrients may have consumed more energy in metabolism, or were absorbed less well and thus excreted. Lower body lipid% resulting from less well utilized nutrients in the SBM diet may also explain the greater ADFI found in the SBM group in the previous study (Quinton et al., 2007). Although the 2 diets in this study were balanced for energy and composition on a gross basis, the SBM diet likely contained antinutritional factors (reviewed by Francis et al., 2001) that caused reduced energy on a digestible basis compared with the FM diet (J. Koskela, unpublished data) and may have limited the availability of some amino or fatty acids. Restricted or poor quality dietary nutrients, as well as reduced lipid deposits, are well documented to increase feed intake in salmonids (e.g., Jobling and Koskela, 1996; Jobling and Johansen, 1999; Silverstein et al., 1999; Johansen et al., 2002). In this case, whitefish on the SBM diet may have increased intake to meet a biologically determined nutritional demand. Our results differed from other studies that found no effect of dietary soy on body composition (Smith et al., 1988; Kaushik et al., 1995), but due to processing differences, soy digestibility likely differed in these studies.
Whole-body protein% and protein weight did not differ between the diets. Thus, protein% differences did not contribute greatly to diet differences in wet weight growth. This finding was also in agreement with Shearer (1994), who concluded that body protein percentage is constrained within narrow limits for fish within a given weight, life-cycle stage, and species. In this case the greater intake in the SBM group (Quinton et al., 2007) may have enabled this group to deposit sufficient protein for their body size (but not additional lipid), thus bringing protein% equal to those of the FM group.
Genotype-Diet Interactions for Body Composition. Genetic parameters showed that body lipid% and protein% were not greatly affected by genotype-diet interactions. Coefficients of phenotypic variation and heritabilities of lipid% and protein% did not differ substantially between diets. Between-diet genetic correlations of lipid% measured in FM and SBM diets were close to unity. Protein percentage cross-diet correlations were lower, but still highly positive with very large standard errors. These findings are encouraging for fish breeders for 2 reasons. First, a novel SBM-based diet is unlikely to substantially impact selection potential for body composition. Second, selection applied within currently used FM diets will have a correlated genetic response of improved body composition within SBM-based diets as well.
Possible evidence for genotype-environment interaction was seen by the within-diet differences in trait genetic correlations. Lipid percentage genetic correlations with DG and ADFI tended to be weaker (average absolute difference = 0.19), and thus more favorable, under the SBM diet. A possible explanation is that the more suitable protein and lipid quality of the FM diet increased lipid deposition in fish and, thus, also amplified the relationship between rapid growth and increased lipid deposition. A similar phenomenon has been observed in rainbow trout fed high energy diets (Kause et al., 2007). Also, protein% and ADFI exhibited a positive genetic correlation in the FM diet, but zero genetic correlation in the SBM diet. Genetic correlations of G:F and lipid% (positive in FM, negative in SBM) and protein% (negative in FM, positive in SBM) were more favorable in the SBM diet as well. These correlations together suggest the SBM diet may provide a more favorable environment for selective breeding. However, these small differences are speculative and should be confirmed with further studies because confidence intervals of the compared genetic estimates overlapped and phenotypic correlations did not show large differences between diets.
Generally, the observed negligible genotype-environment interaction agreed with other farmed fish studies, where composition traits have not displayed diet interactions with family (Austreng et al., 1977; Austreng and Refstie, 1979; Tobin et al., 2006) or strain (Smith et al., 1988; Li et al., 2006). Similarly, farmed fish studies on growth and feed efficiency have found little or no genotype interactions with diet nutrient levels (Austreng et al., 1977; Austreng and Refstie, 1979; Smith et al., 1988; Romana-Eguia and Doyle, 1992; Blanc, 2002; Glover et al., 2004; Kause et al., 2006b; Li et al., 2006; Palti et al., 2006; Quinton et al., 2007). Therefore, selection strategies for improving growth, feed efficiency, and body composition within current fish meal-based diets should also improve populations for future soybean meal-based diets.
In conclusion, results of this study demonstrated that simultaneous selection for rapid growth and reduced body lipid percentage was an effective strategy to improve feed efficiency in European whitefish and achieved greater genetic gains in efficiency compared with selection on growth alone. Furthermore, increasing use of soybean meal in aquaculture feeds should not greatly impact salmonid selective breeding programs. Current selection that occurs within fish meal-based diets should also improve populations for future plant-based diets.
|
Footnotes |
|---|
2 Corresponding author: cheryl.quinton{at}mtt.fi
Received for publication December 4, 2006. Accepted for publication August 15, 2007.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
P2), coefficients of phenotypic variation (CVP), heritabilities (h2), and regression coefficients (Reg) for absolute and relative whole-body wet weight lipid and protein percentages measured under fish meal and soybean meal diets