Genetic comparison of experimental farmed strains and wild Icelandic populations of Atlantic cod (Gadus morhua L.)

Page 1

2006) 556–564www.elsevier.com/locate/aqua-online

Aquaculture 261 (

Genetic comparison of experimental farmed strains and wildIcelandic populations of Atlantic cod (Gadus morhua L.)

Christophe Pampoulie ⁎, Thóra Dögg Jörundsdóttir, Agnar Steinarsson, Gróa Pétursdóttir,Magnús Örn Stefánsson, Anna Kristín Daníelsdóttir

Marine Research Institute, Skúlagata 4, 101 Reykjavík, Iceland

Received 4 May 2006; received in revised form 28 July 2006; accepted 29 July 2006

Abstract

Microsatellite DNA loci and the Pantophysin locus (Pan I) were used to investigate levels of genetic diversity within farmedstrains of Atlantic cod Gadus morhua and to compare them with the wild source population. A total of 282 farmed samplesoriginating from a spawning ground off the south-west coast of Iceland were sampled in the years 2002 and 2003, and 258 wild codwere collected at the same spawning ground in the same years. The farmed strains exhibited a lower mean number of alleles and allelicdiversity than the wild samples at the microsatellite loci. Significant differences were observed between wild and farmed samplesboth in allele and genotype frequencies at the Pan I locus. We argue that the genetic divergence of wild and farmed samples ofAtlantic cod may be due to a small number of effective founding breeders contributing to the genetic variation of the farmed strains,inducing a reduction in allelic diversity. We discuss the potential effect of breeding practices on the genetic diversity of Atlantic cod.© 2006 Elsevier B.V. All rights reserved.

Keywords: Atlantic cod; Gadus morhua; Farming; Genetic variability; Pan I; Microsatellite; Founder effect

1. Introduction

Differences in allele frequencies between farmedstrains and the wild source populations have been shownto be mainly due to the breeding of related individuals orto the use of small numbers of parents as brood-stock,leading to a lower genetic variability of the farmedstrains (McGinnity et al., 2003). Additionally, farmedstrains of small population size are more sensitive togenetic drift and consequently have lower withinpopulation genetic diversity than wild populations(Allendorf, 1986). Therefore, it has been argued that aselective breeding program aiming at producing fast

⁎ Corresponding author. Tel.: +354 570 7216; fax: +354 570 7210.E-mail address: [email protected] (C. Pampoulie).

0044-8486/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.aquaculture.2006.07.044

growing fish with high food conversion and docilebehaviour for aquaculture should start from a basepopulation containing high genetic variability. Such aselection program will change the genetic compositionof the farmed strains compared with that of the basepopulation by reducing its genetic variability: asselection continues, ‘positive’ alleles displace ‘negative’ones (Allendorf and Ryman, 1987). Genetic variability isthe primary biological resource in the successfulartificial propagation of any species. Therefore, thepreservation of genetic variation is an importantconsideration in the maintenance of any hatchery stock(Allendorf and Phelps, 1980). Population bottlenecks ofshort duration, such as when reared strains are founded,may have little effect on heterozygosity but are expectedto reduce severely the number of alleles present. The

Page 2

557C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

number of alleles remaining is important for the longterm response to selection and survival of populationsand species (Allendorf, 1986).

The Atlantic cod, Gadus morhua, is one of the mostexploited fish in the North Atlantic. Due to overexploi-tation, most cod stocks have dramatically declined oreven collapsed (see Christensen et al., 2003 for areview). Decline in the supply of wild caught fish hascreated an opportunity for the production of cultured codand at present, farming programs are being developed inseveral countries such as Iceland, Faroe Islands, Scot-land, Norway and Canada. As research focusing on thecomparison of the genetic variability between farmedand native populations of marine fishes has been limited,we carried out a study on Atlantic cod. Samples of wildspawning cod were taken on the main spawning groundoff the south-west coast of Iceland and first generationprogeny of wild parents that came from that samelocation were sampled at the hatchery facility. Thesamples were analysed using eight microsatellite lociand the Pantophysin locus (Pan I), which encodes for anintegral membrane protein found in cytoplasmic micro-vesicles that are thought to function in a variety ofintracellular shuttling pathways. The Pan I locus hasrecently been shown to be influenced by selection(Karlsson andMork, 2003; Pogson andMesa, 2004), andtemperature has been suggested to be a selective force.As most of the microsatellite loci are expected to beselectively neutral (Schlötterer and Wiehe, 1999), it isperceived that genetic differentiation between farmedstrains and wild populations will be due to genetic driftwhen the strains are established. The main objective ofthe current study was to compare the genetic variabilityof experimental farmed strains of cod to their wild sourcepopulation and quantify any genetic differences.

2. Materials and methods

2.1. Samples

A total of 258 mature wild fish were sampled on themain spawning ground off the south-west coast ofIceland in summer 2002 and 2003 (coded Wild S02 andWild S03, respectively) and on a feeding ground inautumn 2002 (coded Wild A02). Also, 282 maturefarmed fish were sampled at the Experimental MarineFishfarm of the Marine Research Institute located inGrindavík, Iceland during the same periods (coded FarmS02, Farm S03 and Farm A02). Additionally, an out-group from Norway was added to give an indication onhow much farmed strains and wild Icelandic samplesdiffer genetically. The farmed strain originated from the

main spawning ground south–west off Iceland and thewild source population was sampled for geneticcomparison. From 1996 to 2000 (year classes of wildcod samples collected), an experimental farming processwas carried out at the Experimental Marine Fishfarm ofthe Marine Research Institute, Iceland. Two mainbreeding practices were used: firstly, wild cod wereallowed to spawn naturally in on-shore tanks and se-condly, cod were hand-stripped on the boat at the point ofcapture. In the former, tanks contained a maximum of 20fishes at a time (2 to 4 males per female) and fertilisedeggs were collected regularly. In the latter, the eggs ofone female were stripped and fertilised with the milt oftwo to four males. Eggs and larvae obtained by the twobreeding practices were then reared together. Based onthese two breeding practices, the number of contributingbreeders was 13 to 20 females and 26 to 60 males peryear. As the experimental spawning that took place from1996 to 2000 was carried out with different wild parentson each occasion, we carried out a temporal comparisonto monitor if the genetic divergence between farmed andwild samples was temporally stable.

Gill filaments or fin clips were collected and stored in1 ml of 95% ethanol prior to DNA extraction.

2.2. DNA extraction and PCR amplification ofmicrosatellite and Pan I loci

DNA samples were extracted from gill filaments or finclips using a Chelex (Biorad, 10%) extraction protocol(Walsh et al., 1991). Samples were genotyped at eightpolymorphic microsatellite loci: Gmo2 (Brooker et al.,1994), Gmo8, Gmo19, Gmo34 (Miller et al., 2000),Tch5, Tch11, Tch14 and Tch22 (O'Reilly et al., 2002).Forward primers were labelled with VIC (Gmo8, Tch11,Tch14, Tch22), 6-FAM (Gmo2, Gmo34, Tch5) or NED(Gmo19) fluorescent dye (Applied Biosystems). Poly-merase chain reactions were performed in a 10 μl volumecontaining 2 μl of DNA product, 1 μl of 10× buffer(10mMTris-HCl, 50 mMKCl, 1.5mMMgCl2 and 0.1%Triton X-100), 1 μl of 2.5 mM DNTP, 0.2–0.4 units ofDyNAzyme™ DNA polymerase (Finnzymes) and vari-ous concentrations of primers. PCR cycles were per-formed on a multiplex basis in a GeneAmp®2700 thermalblock and preceded by an initial denaturation step of 4minat 94 °C followed by 32 cycles of: 40 s at 94 °C, 40 s atannealing temperature (Tch11, Gmo19, Gmo34 at 52 °C;Gmo2, at 50 °C; Tch14 at 56 °C; Gmo8 at 50 °C; Tch5 at50 °C; Tch22 at 49 °C), and 40 s at 72 °C. A finalelongation step of 4 min at 72 °C was performed. PCRproducts were diluted with distilled water (1:3) and wereelectrophoresed on 25 cm 6% polyacrylamide gels and

Page 3

558 C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

detected on an ABI-automatic sequencer (ABI 377,Applied Biosystems) using the software GeneScan vers.3.1.2 (Applied Biosystems, 2000). Products were scoredusing the software GeneMapper vers. 3.0 (AppliedBiosystems, 2002).

Samples were also genotyped at the Pantophysinlocus (Pan I). Polymerase chain reactions were per-formed in a 15 μl volume containing 2 μl of DNAproduct, 1.7 μl of 10× buffer (10 mM Tris-HCl, 50 mMKCl, 1.5 mM MgCl2 and 0.1% Triton X-100), 0.5 μl of50 mM MgCl2, 1.7 μl of 2.5 mM DNTP, 1 unit ofDyNAzyme™ DNA polymerase (Finnzymes) and10 mM of primers (F: TTGGTCCTCTATCTGGGCTTC; R: CGTAGCAGAAGAGTGACACAT). PCRcycles were performed in a GeneAmp®2700 thermalblock and preceded by an initial denaturation step of5 min at 95 °C followed by 30 cycles of: 1 min at 94 °C,1 min at 55 °C and 1 min at 72 °C. A final elongationstep of 7 min at 72 °C was performed. PCR productswere then cut with DraI restriction enzyme (Fermentas)using 15 μl of PCR products and 5 μl of DraI mixcontaining 18 units of enzyme and 2 μl of B+ buffer persamples. Digested PCR products were visualized in 2%agarose gel stained with ethidium bromide usingGELDOC™ 2000 reader (Biorad).

2.3. Statistical analysis

Allele frequencies, observed (HO) and unbiasedexpected heterozygosity (HE) were calculated inGENETIX vers. 4.03 (Belkhir et al., 1999). Hardy–Weinberg equilibrium (HWE) was calculated and testedfor significance in GENEPOP vers. 3.02 (Raymond andRousset, 1995). When appropriate, significant levelswere adjusted with a sequential Bonferroni test (Rice,1989). The quantity A (allelic diversity; Allendorf andRyman, 1987) is a measure of the proportion of geneticvariation remaining in farmed strains from the foundingpopulation based on the number of alleles (na′) retained ata polymorphic locus, where:

A ¼ ðnaV−1Þ=ðna−1Þ

and na is the original number of alleles present. Allelicdiversity ranges from 1, where all alleles are retained(wild populations), to 0 where all alleles but one are lost(Allendorf and Ryman, 1987). Average allelic diversityof microsatellite loci for farmed strains of cod was testedagainst allelic diversity for wild brood-stock (local wildpopulations). Significance was tested using Wilcoxonmatched paired sample test (Zar, 1999). Additionally,genetic diversity (allelic richness, HO and HE) was

compared between farmed strains and wild populationsof cod using FSTAT (Goudet, 1995).

Wright's single-locus F-statistics (Wright, 1969)were calculated from allele frequencies for all loci exa-mined for each population according to Weir andCockerham (1984) in GENETIX. The significance ofmulti-locus FST was assessed using the exact test inGENEPOP (Raymond and Rousset, 1995). Whenappropriate, significance levels were adjusted with asequential Bonferroni test (Rice, 1989). Pairwise geneticdistances were estimated with the Nei (1978) standardgenetic distance (microsatellites) and Roger (1972)distance (Pan I) calculated in POPULATIONS (avail-able at the following web page: http://www.pge.cnrs-gif.fr/bioinfo/wini386/) and a graphical representation of thedistance matrix was then computed with bootstrap (1000replicates) using the Neighbour joining algorithm(Saitou and Nei, 1987) available in POPULATIONS inconjunction with TREEVIEW (Page, 1996).

3. Results

3.1. Microsatellite loci

3.1.1. Within populations variationThe eight microsatellite loci studied exhibited varying

degrees of allelic diversity, ranging from eight alleles atlocus Gmo34 to 38 at locus Gmo8 (Table 1). Observedheterozygosity varied from 0.350 for Gmo34 to 0.975for Tch5 while mean allelic diversity ranged from 5.9(Gmo34) to 28.7 (Tch14) (Table 1). The mean number ofalleles was higher in wild samples than in the farmedstrains (Table 1; FSTAT test: P=0.0212). Although notsignificant (FSTAT test: HO P=0.3014; HE P=0.0870),a tendency was observed for lower observed andexpected heterozygosity in the farmed strains. Theallelic diversity assessed with A from Allendorf andRyman (1987) was significantly lower in all farmedstrains compared with wild populations (S02,Z=2.1000, P=0.0356; A02, Z=2.1974, P=0.0280;S03, Z=2.3805, P=0.0256). Almost all loci exhibitedrare alleles (frequencyb0.005) in the wild populationswhich were not present in the farmed strains (AppendixA) and these were consequently designated as diag-nostic alleles. In the farmed samples, three diagnosticalleles were found: locus Gmo2, alleles 104 and 140;locus Tch14, allele 266. All the farmed samplesexhibited a significant deficit of heterozygotes whilethe native spawning populations were in Hardy–Weinberg equilibrium (Table 1). On the other hand,heterozygote deficit was observed in the sample thatcame from the autumn feeding ground (Wild A02),

Page 4

Table 1Number of individuals scored (n), number of alleles per locus (nA), meannumber of alleles (MNA), allelic diversity (A; Allendorf and Ryman,1987) expected heterozygosity (HE), observed heterozygosity (HO) andFIS according to Weir and Cockerham (1984) for each microsatellite loci

WildS02

FarmS02

WildA02

FarmA02

WildS03

FarmS03

n 70 94 94 94 94 94Gmo2

nA 15 13 15 13 15 14HE 0.870 0.851 0.851 0.856 0.850 0.858HO 0.814 0.628 0.755 0.734 0.862 0.681FIS 0.071 0.267 0.117 0.149 −0.008 0.211

Gmo8nA 28 19 18 20 31 19HE 0.922 0.899 0.913 0.914 0.926 0.899HO 0.942 0.830 0.957 0.894 0.851 0.862FIS −0.015 0.082 −0.044 0.028 0.087 0.047

Gmo19nA 23 16 23 18 27 17HE 0.922 0.811 0.904 0.854 0.914 0.861HO 0.900 0.830 0.904 0.809 0.904 0.840FIS 0.031 −0.018 0.005 0.059 0.017 −0.024

Gmo34nA 7 5 7 4 8 5HE 0.420 0.525 0.492 0.503 0.531 0.561HO 0.442 0.521 0.404 0.500 0.457 0.510FIS −0.047 0.013 0.185 0.013 0.143 0.094

Tch5nA 20 15 19 18 19 17HE 0.925 0.897 0.925 0.894 0.926 0.901HO 0.929 0.904 0.809 0.819 0.968 0.883FIS 0.003 −0.003 0.131 0.089 −0.045 0.025

Tch11nA 20 20 21 20 21 20HE 0.922 0.911 0.928 0.914 0.930 0.903HO 0.861 0.926 0.957 0.926 0.947 0.957FIS 0.078 −0.011 −0.027 −0.008 −0.013 −0.055

Tch14nA 25 26 36 28 28 33HE 0.940 0.935 0.952 0.947 0.946 0.924HO 0.900 0.766 0.957 0.660 0.915 0.883FIS 0.049 0.186 0.000 0.308 0.038 0.049

Tch22nA 7 6 6 5 7 5HE 0.657 0.585 0.600 0.632 0.621 0.629HO 0.614 0.510 0.617 0.649 0.638 0.713FIS 0.072 0.132 −0.022 −0.021 −0.022 −0.129

TotalMNA 18.1 15.0 18.1 15.8 20.1 15.0A 1 0.811 ⁎ 1 0.839 ⁎ 1 0.784 ⁎

HE 0.822 0.802 0.821 0.815 0.830 0.811HO 0.800 0.739 0.795 0.749 0.834 0.791FIS 0.034 0.083 0.036 0.086 −0.004 0.030

Bold values: FIS values deviating significantly from HWE assessed inGENEPOP with exact test (Raymond and Rousset, 1995), after multipletest correction.⁎ Indicates significant differences with Wilcoxon matched pairedtest (Zar, 1999).

559C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

probably due to Wahlund effect as mixture of differentspawning populations feed on the same basin.

3.1.2. Among samples variationThe partitioning of genetic variance as estimated by

F-statistics among and within the six Icelandic samplesshowed overall highly significant FST and FIS values of0.013 (confidence interval 0.011–0.015) and 0.045(0.029–0.082) respectively. Out of 15 FST pairwisecomparisons, 11 were significantly different from zero(Pb0.001). The comparisons which did not show anysignificant differences were the three comparisonsbetween wild samples, and between the spring samplesof the farmed strains (Farm S02 and Farm S03). The levelof genetic differentiation between farmed and wildsamples varied from 0.6 to 1.8%.We calculated pairwiseestimates of Nei (1978) genetic distances and con-structed a dendrogram using the Neighbour joininganalysis in the software POPULATIONS and assessedsupport for nodes by bootstrapping over loci (1000replicates) (Fig. 1). The dendrogram topology was con-cordant with the level of differentiation observed withthe FST values, showing strong support for separateclusters of farmed and wild populations (Icelandicsamples and Norwegian out-group). The pairwise FST

values which were lower between Icelandic wild samplesand theNorwegian out-group (0.003bFST valuesb0.008)than between the farmed strains and the Norwegian out-

Fig. 1. Neighbour joining tree with Nei (1978) genetic distance forseven samples (three Icelandic farmed strains and wild samples, and aNorwegian out-group) using eight microsatellite loci. The scalerepresents the distance between branches. Bootstrap values whichrepresent the percentage of all replicates consistent with each branchare shown beside each node.

Page 5

Fig. 2. Neighbour joining tree with Roger (1972) genetic distance forseven samples (three Icelandic farmed strains and wild samples, and aNorwegian out-group) using the Pan I locus. The scale represents thedistance between branches. Bootstrap values which represent thepercentage of all replicates consistent with each branch are shownbeside each node.

560 C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

group (0.015bFST valuesb0.021) strongly supportedobserved clustering.

3.2. The Pan I locus

3.2.1. Within samples variationObserved heterozygosity varied from 0.16 in a farmed

sample to 0.49 in a wild one (Table 2). A tendency wasobserved for lower observed and expected heterozygos-ity in the farmed strains but significance was only de-tected for the expected heterozygosity (FSTAT test:HO P=0.5944; HE P=0.0382). Genotype distributionswere in HWE (Table 2) except for the wild samples ofspring 2003. The frequency of B alleles was significantlyhigher in wild samples than in the farmed ones.

3.2.2. Among samples variationThe partitioning of genetic variance among and with-

in the six Icelandic samples as estimated by F-statisticsshowed an overall highly significant FST and FIS valuesof 0.074 and 0.035, respectively. Out of 15 FST pairwisecomparisons, 10 were significant (Pb0.001). The com-parisons which did not show any significance were thecomparisons between wild samples, the farmed samplesS02 and S03, and the farmed samples A02 and S03. Thelevel of genetic differentiation between farmed and wildsamples varied from −0.3 to 20.2%. Very few B alleleswere present in the farmed samples due to the highfrequency of AA genotype. Genotype frequencies weresignificantly different between farmed and wild samples(v2[1]=18.30, Pb0.001; Table 2). The BB genotype waspresent in a relatively high proportion of the wildsamples but not in the farmed samples. No significantdifference was found between male and female allelefrequencies in our samples (v2[6]=4.31, PN0.05). Wecalculated pairwise estimates of Roger (1972) genetic

Table 2Number of individuals scored (n), number of fish in each sample typedwith AA, AB or BB genotypes, expected heterozygosity (HE),observed heterozygosity (HO) and FIS according to Weir andCockerham (1984) for the Pan I locus

WildS02

FarmS02

WildA02

FarmA02

WildS03

FarmS03

n 70 94 94 93 94 94Genotype AA 25 66 33 44 47 76Genotype AB 33 28 44 46 30 15Genotype BB 12 0 17 4 17 3HE 0.48 0.26 0.49 0.41 0.45 0.20HO 0.49 0.31 0.46 0.48 0.33 0.16FIS −0.005 −0.177 0.067 −0.182 0.276 0.201

Bold values: FIS values deviating significantly from HWE assessed inGENETIX (Belkhir et al., 1999) with 1000 permutations (P=0.01).

distances and constructed a dendrogram using theNeighbour joining analysis in the software POPULA-TIONS and assessed support for nodes by bootstrappingover loci (1000 replicates) (Fig. 2). The dendrogramtopology was concordant with the level of differentiationobserved with the FST values, showing strong supportfor separate clusters of farmed strains and wild po-pulations. Because of the low genotypic diversity of thePan I locus in Norwegian cod, the Norwegian out-groupclustered with the farmed strains. This was supported bythe pairwise FST values which were higher betweenIcelandic wild samples and the Norwegian out-group(0.212bFST valuesb0.283) than between the farmedstrains and the Norwegian out-group (0.0158bFST

valuesb0.157).

4. Discussion

Although exceptions have been documented (Palmet al., 2003), a reduction in genetic diversity measured

Page 6

561C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

as mean number of alleles at neutral markers such asmicrosatellites is generally observed in farmed strainscompared with wild populations (Norris et al., 1999).This also applies to both heterozygosity and allelic di-versity (Mjolnerold et al., 1997; Clifford et al., 1998;Kim et al., 2004). Our study showed that the micro-satellite loci exhibited a significantly lower meannumber of alleles and allelic diversity (A of Allendorfand Ryman, 1987) in the farmed strains compared withthe wild populations. There was also a tendency towardsa higher frequency of diagnostic alleles in the wildsamples when compared with the farmed ones (seeAppendix A). However, the presence of three diagnosticalleles (that were also rare alleles) in the farmed samplessuggest either high selection of those alleles in thefarmed strains or sampling error. The latter could havehappened at any of the sampling occasions: when theoriginal brood-stocks were selected, when the farmedprogenies were sampled or when the wild samples thatwere used for comparison were taken. Loss of rarealleles in farmed strains is generally viewed as a moremeaningful measure of genetic variability than hetero-zygosity mainly because the latter is insensitive tosubstantial genetic change occurring in the firstgeneration of cultured fish (Hedgecock and Sly, 1990;Evans et al., 2004). Loss of rare alleles has consequentlybeen considered to be more harmful than a loss ofheterozygosity (Evans et al., 2004). The reduced geneticvariability we observed in the farmed strains is probablydue to a low number of successful breeders during thefoundation of those strains and is similar to a recentbottleneck effect in terms of impact on geneticvariability (Allendorf, 1986). Loss of genetic variabilityat selectively neutral loci such as microsatellites mightin itself not be considered harmful for aquaculturebreeding programs. However, microsatellite loci can beconsidered representative of changes that happened inthe whole genome (Schlötterer and Wiehe, 1999) andthey could be indicative of the loss of genetic variabilityat coding regions. This can be of crucial importance foraquaculture management, as loss of variation at codingregions can reflect erosion of the underlying geneticvariation of commercially important traits such asgrowth and disease resistance (Vuorinen, 1984; Evanset al., 2004). Moreover, as Evans et al. (2004) stated,when variation is lost during the first generation ofbreeding practices, it will be lost for all subsequentgenerations within a closed breeding program, and willreduce the opportunities for genetic improvementavailable within that culture stock. Because of theobserved erosion of genetic variation in the farmedstrains, we found a significant genetic differentiation

level of almost 2% between the wild samples andfarmed strains. This value corresponds approximately tothe differentiation level observed between very distinctwild stocks of cod in the North Atlantic such as the Isleof Man and Norway (Hutchinson et al., 2001). Highmortality can be observed in production of cod juvenilesand overall final survival after weaning is normally low(Herbinger et al., 1995; Olsen et al., 1999; Gitterle et al.,1995). This has also been observed during our study(Steinarsson A., pers. comm.). High mortality in thehatchery could influence how different families arerepresented in the surviving juvenile population. Somefamilies can be over-represented with others under-re-presented or not represented at all. Therefore, thenumber of breeders (Nb) could be considerably higherthan the effective number of breeders (Ne) represented inthe juvenile population if mortalities in the hatchery arehigh. To maintain levels of genetic variation withinfarmed strains that are comparable to the wild sourcepopulation, good management practices are required.This is especially important in terms of the use of ade-quate numbers of effective parents (high Ne) andthereby minimising the effects of genetic drift (Allen-dorf, 1986). In this way, the maximum amount of bothheterozygosity and allelic diversity will be preserved.

The analysis of a genetic marker coding for Panto-physin (Pan I locus), shows that a loss of heterozy-gosity (both HE and HO) in the coding region of theDNA can occur. Indeed, the analysis revealed the quasiabsence of the allele B and Pan IBB genotypes in thefarmed strains. The observed genetic differencesbetween farmed strains and wild samples couldentirely be due to the selection of brood-stock duringthe foundation of that strain (see above). However, ithas recently been suggested that the Pan I locus isunder Darwinian selection (Karlsson and Mork, 2003;Pogson and Mesa, 2004), and temperature and depthhave been suggested as the selective forces. In thenatural environment, cod is exposed to environmentalchanges such as fluctuations in temperature, salinityand food resources availability. On the other hand, therearing environmental conditions were relatively stableand they were kept closer to the optimum for growthperformance compared with the natural environmentalconditions (temperature was maintained at 9–10 °Cand food was supplied twice a day in the farm;Steinarsson A., pers. comm.). The environmentalconditions maintained at the farm might have exerteda selective pressure that was different to the oneoccurring in the natural environment of the AtlanticOcean. This may have led to the predominance of theallele A in the farmed strains.

Page 7

562 C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

Declining wild Atlantic cod stocks have increased thedemand for farming in Europe and as a consequence,farming programs have been developed in various coun-tries. Already, a fully developed cod farming programwith traceability of families and brood-stock managementis being carried out at the Marine Research Institute ofIceland. Genetic comparison of those new farmed strainswith local wild populations will give a better insight in theefficiency of such programs to conserve the geneticvariability of native cod populations within the farmedstrains. Introgression of farmed and wild population ofexploited fishes has been documented in several species(Clifford et al., 1998; McGinnity et al., 2003; Alarcón etal., 2004), and is known to be more frequent in small thanlarge wild populations (freshwater vs. marine species).Therefore, future fish-farming projects should take thisinformation into account and avoid sea-farming in areaswhere local spawning occurs.

5. Conclusion

We are among the first to investigate the geneticvariability in farmed strains of Atlantic codGadusmorhua,using both neutral markers such as microsatellite loci and amarker under selection such as the Pan I locus. Our studyclearly shows that nuclear markers such as microsatelliteloci are useful to detect genetic variation between a F1-generation of farmed strains and the native populations theywere derived from. Additionally, the use of a marker underselection can be useful to detect whether farmed and nativepopulations are under similar selective pressures. Thefarmed samples were obtained during the experimental set-up of a hatchery facility with the sole aim of collecting eggsfor initial hatching and rearing experiments. Consequently,no care was taken in brood-stock management whichcurrent findings show is vital if breeding programs are to bedesigned to retain genetic variability. Current results shouldbe of particular interest for future farming programs ofmarine exploited fishes which are declining because ofhigh fishing pressure.

Acknowledgement

This research has been carried out under the COD-TRACE EU-project (Q5RS-2000-01697). We would liketo thank the Experimental Marine Fishfarm staff of theMarine Research Institute Iceland who reared the farmedstrains, and A. Ragnarsdóttir, B. Þorgilsson, V. Chossonand H. Pétursdóttir for their help during the sampling.Special thanks are addressed to two anonymous refereesfor their useful comments on the manuscript and toL. Taylor for improving the English.

Appendix A. Presence (+) of diagnostic alleles foreach microsatellite locus and each sample set

WildS02

FarmS02

WildA02

FarmA02

WildS03

FarmS03

n

70 94 94 94 94 94 Gmo2 104 + + + 124 + 126 + 130 + 134 + + 140 + + 142 + + 144 + + 146 +

Gmo8

192 + 196 + 200 + 212 + 232 + 236 + + 256 + 260 + + 264 + + 276 + 280 + 284 + 292 + + 300 + 312 +

Gmo19

124 + + 132 + + 136 + + + 200 + + + 204 + + + 212 + + + 220 + + 224 + 228 + 232 +

Gmo34

112 + + + 116 + + +

Tch5

188 + + 192 + + + 260 + + 264 + 268 + 276 + +

Page 8

Appendix A (continued )

563C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

WildS02

FarmS02

WildA02

FarmA02

WildS03

FarmS03

Tch11

118 + 122 + 134 + 138 +

Tch14

102 + 110 + + + 242 + + 250 + 254 + 266 +

Tch22

107 + 115 +

References

Alarcón, J.A., Magoulas, A., Georgakopoulos, T., Zouros, E., Alvarez,M.C., 2004. Genetic comparison of wild and cultivated Europeanpopulations of the Gilthead Sea bream (Sparus aurata). Aquacul-ture 230, 65–80.

Allendorf, F.W., 1986. Genetic drift and the loss of alleles versusheterozygosity. Zoo Biol. 5, 181–190.

Allendorf, F.W., Phelps, S.R., 1980. Loss of genetic variation in ahatchery stock of cutthroat trout. Trans. Am. Fish. Soc. 109,537–543.

Allendorf, F.W., Ryman, N., 1987. Genetic management of hatcherystock. In: Ryman, N., Utter, F. (Eds.), Population Genetics andFishery Management. University of Washington Press, London,pp. 141–160.

Belkhir, K., Borsa, P., Goudet, J., Bonhomme, F., 1999. GENETIX:logiciel sous Windows pour la génétique des populations Version3.0. Laboratoire Génome and Population, CNRS-UPR. Universitéde Montpellier II, Montpellier (France).

Brooker, A.L., Cook, D., Bentzen, P., Wright, J.M., Doyle, R.W.,1994. Organization of microsatellites differs between mammalsand cold-water teleost fishes. Can. J. Fish. Aquat. Sci. 51,1959–1966.

Christensen, V., Guénette, S., Heymans, J.J., Walters, C.J., Watson, R.,Zeller, D., Pauly, D., 2003. Hundred-year decline of North Atlanticpredatory fishes. Fish Fish. 4, 1–24.

Clifford, S.L., McGinnity, P., Ferguson, A., 1998. Genetic changes inan Atlantic salmon populations resulting from escaped juvenilefarm salmon. J. Fish Biol. 52, 118–127.

Evans, B., Bartlett, J., Sweijd, N., Cook, P., Elliott, N.G., 2004. Loss ofgenetic variation at microsatellite loci in hatchery producedabalone in Australia (Haliotis rubra) and South Africa (Haliotismidae). Aquaculture 233, 109–127.

Gitterle, T., Rye, M., Salte, R., Cock, J., Johansen, H., Lozano, C.,Suárez, J.A., Gjerde, B., 1995. Genetic (co)variation in harvestbody weight and survival in Penaues (Litopenaeus) vannameiunder standard commercial condition. Aquaculture 243, 83–92.

Goudet, J., 1995. FSTAT (Version 1.2): a computer program tocalculate F-statistics. J. Heredity 86, 485–486.

Hedgecock, D., Sly, F., 1990. Genetic drift and effective sizes ofhatchery-propagated stocks of the Pacific oyster Crassostreagigas. Aquaculture 88, 21–38.

Herbinger, C.M., Doyle, R.W., Pitman, E.R., Paquet, D., Mesa, K.A.,Morris, D.B., Wright, J., Cook, D., 1995. DNA fingerprint basedanalysis of paternal and maternal effects on offspring growth andsurvival in communally reared rainbow trout. Aquaculture 137,245–256.

Hutchinson, W.F., Carvalho, G.R., Rogers, S.I., 2001. Marked geneticstructuring in localised spawning populations of cod Gadusmorhua in the North Sea and adjoining waters, as revealed bymicrosatellites. Mar. Ecol. Prog. Ser. 223, 251–260.

Karlsson, S., Mork, J., 2003. Selection-induced variation at thepantophysin locus (Pan I) in a Norwegian fjord population of cod(Gadus morhua L.). Mol. Ecol. 12, 3265–3274.

Kim, J.E., Withler, R.E., Ritland, C., Cheng, K.M., 2004. Geneticvariation within and between domesticated chinook salmon, On-corhynchus tshawytscha, strains and their progenitor populations.Environ. Biol. Fishes 69, 371–387.

McGinnity, P., Prodöhl, P.A., Ferguson, A., Hynes, Ó., Maoiléidigh, R.N., Baker, N., Cotter, D., O'Hea, B., Cooke, D., Rogan, G.,Taggart, J., Cross, T., 2003. Fitness reduction and potentialextinction of wild populations of Atlantic salmon Salmo salar as aresult of interactions with escaped farm salmon. Proc. R. Soc.Lond. B 270, 2443–2450.

Miller, K.M., Le, K.D., Beacham, T.D., 2000. Development of tri- andtetranucleotide repeat microsatellite loci in Atlantic cod (Gadusmorhua). Mol. Ecol. 9, 238–239.

Mjolnerold, I.B, Refseth, U.H., Karlsen, E., Balstad, T., Jakobsen, K.S.,Hindar, K., 1997. Genetic differences between two wild and onefarmed population of Atlantic salmon (Salmo salar) revealed bythree classes of genetic markers. Hereditas 127, 239–248.

Nei,M., 1978. Estimation of average heterozygosity and genetic distancefrom a small number of individuals. Genetics 89, 583–590.

Norris, A.T., Bradley, D.G., Cunningham, E.P., 1999. Microsatellitegenetic variation between and within/farmed and wild Atlanticsalmon Salmo salar populations. Aquaculture 180, 247–264.

O'Reilly, P.T., Canino, M.F., Bailey, K.M., Bentzen, P., 2002. Isolationof twenty low stutter di- and tetranucleotide microsatellites forpopulation analyses of walleye pollock and other gadoids. J. FishBiol. 56, 1074–1086.

Olsen, Y., Evjemo, J.O., Olsen, A., 1999. Status of the cultivationtechnology for production of Atlantic halibut (Hippoglossushippoglossus) juveniles in Norway/Europe. Aquaculture 176, 3–13.

Page, R.D.M., 1996. TREEVIEW: an application to display phylogenetictrees on personal computers. Comput. Appl. Biosci. 12, 357–358.

Palm, S., Dannewitz, J., Jarvi, T., Petersson, E., Prestegaard, P., Ryman,N., 2003. Lack ofmolecular genetic divergence between sea-ranchedand wild sea trout (Salmo trutta). Mol. Ecol. 12, 2057–2071.

Pogson, G.H., Mesa, K.A., 2004. Positive Darwinian selection at thepantophysin (Pan I) locus in Marine Gadid fishes. Mol. Biol. Evol.21, 65–75.

Raymond, R., Rousset, F., 1995. GENEPOP (Version 1.2): populationgenetic software for exact tests and ecumenicism. J. Heredity 86,248–249.

Rice, W.R., 1989. Analyzing table of statistical tests. Evolution 43,223–225.

Roger, J.S., 1972. Measures of genetic similarity and genetic distance.Stud. Genet. 7213, 145–153.

Saitou, N., Nei, M., 1987. The neighbor joining method, a newmethods for reconstructing phylogenetic trees. Mol. Biol. Evol. 4,406–425.

Page 9

564 C. Pampoulie et al. / Aquaculture 261 (2006) 556–564

Schlötterer, C., Wiehe, T., 1999. Microsatellites: a neutral marker toinfer selective sweeps. In: Goldstein, D.B., Schlötterer, C. (Eds.),Microsatellites: Evolution and Applications. Oxford Univ. Press,Oxford, UK, pp. 238–247.

Vuorinen, J., 1984. Reduction of genetic variability in a hatchery stockof brown trout, Salmo trutta. J. Fish Biol. 24, 339–348.

Walsh, P.S., Metzger, D.A., Higuchi, R., 1991. Chelex 100 as amedium for simple extraction of DNA for PCR-based typing fromforensic material. Biotechniques 10, 506–513.

Weir, B.S., Cockerham, C.C., 1984. Estimating F-statistics for theanalysis of population structure. Evolution 38, 1358–1370.

Wright, S., 1969. Evolution and the genetics of populations. TheTheory of Gene Frequencies, vol. 2. University of Chicago Press,Chicago.

Zar, J.H., 1999. Biostatistical Analysis, 4th ed. Prentice Hall, NewJersey.