Project Topics
Get Complete Project Material File(s) Now! »
Selective breeding in fish: accessing the pedigree?
An important point, in any optimized selective breeding programme, is the capability to keep track of the pedigree. The knowledge of the pedigree has three main interests:
• allowing a better management of inbreeding, as with a known pedigree inbreeding can be calculated and constrained through optimized matings;
• permitting the estimation of heritability and genetic correlations through the evaluation of the within and between-family variance components for the trait(s) of interest, a strategy much more efficient than realized heritability, which is limited to one trait and requires a selected and a control line;
• setting up more efficient breeding programmes using family information as a means to improve the precision of the selection index.
In fish, knowledge of the pedigree is complicated by the fact that hatchlings are very small in size (from a few tens of micrograms to 150 mg) and cannot be physically tagged. There are three ways to solve the issue of using pedigrees in fish breeding programmes: (i) not use them (which is the case with mass or individual selection), (ii) use separate rearing of progenies until they reach a size where they can be tagged (usually ca. 20g mean body weight) or (iii) use genotyping of polymorphic markers to assess the parentage of individuals.
Not using the pedigree: individual selection
Using individual selection can yield interesting gains owing to the high selection intensity possible, and remains a choice of interest for selecting traits that can be measured directly on live breeding candidates. This method produced positive results selecting for body weight in channel catfish Ictalurus punctatus (Dunham and Smitherman, 1983), gilthead sea bream (Knibb et al., 1998), Nile tilapia (Basiao and Doyle, 1999), brown trout (Chevassus et al., 2004) and common carp Cyprinus carpio (Vandeputte et al., 2008). However, this was not always the case, and unsuccessful trials in common carp (Moav and Wohlfarth, 1976) and Nile tilapia (Teichert-Coddington and Smitherman, 1988; Huang and Liao, 1990) initially led some to think that selection was not operating in fish (Gjedrem, 2012). It was also the rationale to develop the “Prosper” method, an optimized individual selection method for growth that proposes to control non genetic maternal effects and competition effects (Chevassus et al., 2004) and has been the basis of the development of many breeding programmes in France (Haffray et al., 2004; Vandeputte et al., 2009a).
Other traits have also been successfully selected for by mass selection, like body shape in common carp (Ankorion et al., 1992) and muscle fat content, estimated with a Distell Fat-meter, in rainbow trout (Quillet et al., 2005).
Individual selection has the advantage that it is the easiest and cheapest to implement of selection methods, making it particularly suitable for small or medium companies. However, it also suffers from serious drawbacks. The first one is that this method is likely to generate important rates of inbreeding if not properly managed (Gjerde et al., 1996; Dupont-Nivet et al., 2006). Second, it cannot be used on traits that cannot be recorded (directly or indirectly) on the live breeding candidates. In addition, the genetic parameters of the traits selected remain generally unknown. Realized heritability for the selected trait can be estimated if a control line is maintained in parallel, but this is a rough estimate of the true heritability, and most of all the genetic parameters of other potentially interesting traits (and the genetic correlations among those) will remain unknown unless specific selection experiments for these traits are set up. Finally, selection might be more effective, especially with low heritability traits, if family information can be used (Falconer and Mackay, 1996). This will be especially true when selection applies to a combination of several traits – and it is a normal fate for breeding programmes to incorporate more traits over time.
Separate rearing of the families
Family-based selection with separate rearing of progenies is the method which has been used and developed in the first « modern » breeding programmes for fish in the 1970’s-1980’s in Norway (Gjedrem, 2010, 2012) and in North America (e.g. Hershberger et al., 1990). Typically, in such breeding programmes, each male is mated with 2-3 females in a hierarchical system, then progenies are reared separately until tagging (100-400 separate rearing units needed). After tagging3 at ca. 20g (almost 1 year in Atlantic salmon – Gjerde et al., 1994), some breeding candidates remain on the breeding site, while other tagged fish from the different families are sent to on-farm growing tests or to challenge testing for diseases (Gjedrem, 2010). This type of breeding programme was then extended to other species, with famous programmes like the GIFT (Genetically Improved Farm Tilapia) in the Philippines (Eknath and Acosta, 1998), the programme for the improvement of rohu Labeo rohita in India (Gjerde et al., 2005) or several programmes for the Pacific white shrimp Penaeus vannamei (e.g. Gitterle et al., 2005).
Knowledge of the multi-generational pedigree allows the use of optimal methods, like BLUP (Best Linear Unbiased Prediction) for the prediction of breeding values. Undoubtedly, such breeding programmes have generated the bulk of the genetic gain in the major genetically improved species of world aquaculture like Nile Tilapia, Atlantic salmon and Pacific white shrimp (Neira, 2010; Rye et al., 2010). While they are very convenient to include traits recorded on sibs in challenge tests (including farm ongrowing data), the initial rearing phases are done in conditions that differ a lot from industry standards, owing to the necessity to have all families reared separately in small volume tanks or hapas. A fish that starts its life at a few milligrams (or tens of milligrams in the case of salmonids) has already increased its body weight by a factor of 200 to 2000 when it reaches 20g, while the way to commercial weight only implies a further multiplication by a factor of 20 to 200.
Therefore, common environment effects (= « tank effects ») are expected to be large, and indeed they may be so when measured: 10-30% of the phenotypic variance for body weight in common carp (Ninh et al., 2011), a « substantial » amount in rohu carp (Gjerde et al., 2005), from 2 to 20% in Atlantic Salmon (Gjerde et al., 1994), from 3 to 12% in Atlantic cod Gadus morhua (Gjerde et al., 2004; Tosh et al., 2010), 14-17% in rainbow trout (Henryon et al., 2002). In some cases however, it appears that tank effects are contained to limited level (0 to 9% of phenotypic variance in rainbow trout- Elvingson and Johansson, 1993). High tank effects are problematic as they may bias the estimated family values, and then the estimated breeding value of individuals. In addition to this, the need to minimize family (and then rearing units) number tends to promote hierarchical mating designs, as for a given effective population size (needed to avoid inbreeding) they imply the production of less families than most factorial mating designs. However, hierarchical designs perform less than factorial designs both for estimation of genetic parameters (Vandeputte et al., 2001) and conservation of 3 Initially, tagging was performed by freeze branding, abaltion of fins or use of external tags (Gjedrem, 2010. Today, the majority of fish is tagged by injection of RFID glass tags (called PIT-tags) which provide reliable individual tagging at a modest cost (1-2 € per – reusable- tag) genetic variability (Dupont-Nivet et al., 2006). A last problem of using separate rearing of families is the initial cost of building the infrastructure. While the benefit to cost ratio of fish selective breeding is so high that at the industry level this initial cost should have no significant impact on the profitability of a breeding programme (Ponzoni et al., 2008), this initial investment can clearly be a constraint for a small-medium company to engage in a breeding programme.
A posteriori parentage assignment with genetic markers
he last solution to access pedigree information is the use of genetic markers. This had been thought of a long time ago in fishes (Brody et al., 1976; Moav et al., 1976; Brody et al., 1980), but at that time the available genetic markers (allozymes) did not exhibit enough variability to resolve parentage in more than a few families and involved highly invasive (even lethal) sampling. The idea had a second life when microsatellite markers became available, as those markers have a much higher genetic variability, and sampling is limited to a small piece of fin kept in ethanol at ambient temperature. Then, using either exclusion of incompatible parent pairs (Dodds et al., 1996) or maximum likelihood approaches (SanCristobal and Chevalet, 1997), a new possibility to trace family relationships arose. The first small scale trials were done in salmon and cod (Doyle et al., 1995; Doyle and Herbinger, 1995; Herbinger, 1995), and it soon became evident theoretically that large crosses with several tens of parents could be dealt with (Estoup et al., 1998; Norris et al., 2000; Villanueva et al., 2002). The first large scale trials were done in rainbow trout and Atlantic salmon (Fishback et al., 2002; Norris and Cunningham, 2004), with single assignment rates higher than 90%.
Several assignment softwares have been developed, some more focused on wildlife (CERVUS: Marshall et al., 1998; PARENTE: Cercueil et al., 2002; PAPA: Duchesne et al., 2002), on forest trees (FAMOZ: Gerber et al., 2003) or on aquaculture species (PROBMAX: Danzmann, 1997; VITASSIGN: Vandeputte et al., 2006; FAP: Taggart, 2007). Differences between assignment results can appear in complex situations, especially with likelihood-based softwares in which more hypotheses are needed than with simple exclusion (Herlin et al., 2007). The main drawback of exclusion-based softwares is their sensitivity to genotyping errors, which generates « impossible » genotypes and then unassigned offspring. This proportion of unassigned offspring can reach high levels even with modest genotyping error rates (Vandeputte et al., 2006), but this problem can be easily solved by accepting a limited number of allelic mismatches (1 to 2 in general) in the evaluation of an offspring-sire-dam triplet (Vandeputte et al., 2006; Christie, 2010). In this way, practical assignment rates higher than 90% can be obtained most of the time (Vandeputte et al., 2011).
Table of contents :
1 General introduction
1.1 Aquaculture: a fast-growing animal production deserving optimisation
1.2 Starting from the wild: domestication and selective breeding in fish
1.3 Selective breeding in fish: accessing the pedigree?
1.3.1 Not using the pedigree: individual selection
1.3.2 Separate rearing of the families
1.3.3 A posteriori parentage assignment with genetic markers
1.4 The genetics of European sea bass
1.4.1 Population genetics of sea bass
1.4.2 Genetic variation for quantitative traits
1.5 Sex ratio in the sea bass: a difficult trait to deal with
1.6 A two-steps approach for studying genetic variation and its application to sea bass culture
2 Genetic variation as revealed by between-family variation in common garden experiments
2.1 Genetic variation for body size
2.1.1 Introduction
2.1.2 Material and methods
2.1.2.1 Animals
2.1.2.2 Data collection.
2.1.2.3 Parentage assignment
2.1.2.4 Statistical analyses
2.1.3 Results
2.1.4 Discussion
2.1.4.1 Deformities
2.1.4.2 Maternal effect
2.1.4.3 Heritability estimates
2.1.4.4 Genotype by environment interactions
2.1.4.5 Correlations between growth traits
2.1.5 Summary
2.2 Genetic variation for growth rate
2.2.1 Introduction
2.2.2 Material and methods
2.2.2.1 Animals
2.2.2.2 Data collection
2.2.2.3 Statistical analyses
2.2.3 Results and discussion
2.2.4 Summary
2.3 Additional data: genetic correlations between initial weight, slaughter weight and growth rate
2.4 Genetic variation for sex ratio
2.4.1 A polygenic hypothesis for sex determination
2.4.1.1 Introduction
2.4.1.2 Material and methods
2.4.1.3 Results
2.4.1.4 Discussion
2.4.1.5 Summary
2.4.2 Supplemental information
2.4.2.1 The threshold model for sex determination
2.4.2.2 Genetic and environmental correlations of sex tendency and body length at different ages
3 Effects of domestication and directional selection for body length
3.1 Selection response for growth
3.1.1 Introduction
3.1.2 Materials and methods
3.1.2.1 Selection of sires
3.1.2.2 Constitution of the experimental progeny
3.1.2.3 Rearing conditions and phenotyping
3.1.2.4 Parentage assignment
3.1.2.5 Statistical analyses
3.1.3 Results
3.1.3.1 Parentage assignment
3.1.3.2 Selection response in separate tanks
3.1.3.3 Selection response in mixed tanks
3.1.4 Discussion
3.1.4.1 Parentage assignment in mixed tanks
3.1.4.2 Response to selection
3.1.4.3 Effect of domestication
3.1.4.4 Possible application of the results at commercial scale
3.1.5 Summary
3.2 Selection response for sex ratio
3.2.1 A necessary baseline: sex ratios in wild sea bass populations
3.2.1.1 Introduction
3.2.1.2 Material and methods
3.2.1.3 Results
3.2.1.4 Discussion
3.2.1.5 Summary:
3.2.2 Sex ratio changes in domesticated and selected populations
3.2.2.1 Introduction
3.2.2.2 Material and methods
3.2.2.3 Results
3.2.2.4 Discussion
3.2.2.5 Summary
3.2.3 Additional data: genetic correlation of growth and sex tendency over time
3.3 Modelling the mid-term evolution of growth and sex ratio in farmed sea bass populations under selection for growth
3.3.1 Why model the evolution of sex ratio under selection for growth ?
3.3.2 A stochastic simulation model
3.3.3 Parameters tested
3.3.4 Results and discussion
4 General Discussion
4.1 Summary of the main results
4.2 A perspective for sea bass domestication and breeding programmes
4.2.1 A proof of concept of the use of marker-based parentage assignment
4.2.2 Selection for increased body weight
4.2.3 Genotype by environment interactions for growth: which consequences?
4.2.4 Practical consequences of genetic variation for sex ratio
4.2.5 How to move towards monosex female sea bass populations?
4.3 An evolutionary perspective on the genetic variation of sex ratio in sea bass, and its relations with growth
4.3.1 Polygenic sex determination: a very peculiar system ?
4.3.2 The adaptive significance of polygenic sex determination (in sea bass)
4.3.3 Sex determination and sex differentiation?
4.4 Conclusion
5 Bibliography
