QTL for body weight, morphometric traits and stress response in European sea bass Dicentrarchus labrax.

Abstract

Natural mating and mass spawning in the European sea bass (Dicentrarchus labrax L., Moronidae, Teleostei) complicate genetic studies and the implementation of selective breeding schemes. We utilized a two-step experimental design for detecting QTL in mass-spawning species: 2122 offspring from natural mating between 57 parents (22 males, 34 females and one missing) phenotyped for body weight, eight morphometric traits and cortisol levels, had been previously assigned to parents based on genotypes of 31 DNA microsatellite markers. Five large full-sib families (five sires and two dams) were selected from the offspring (570 animals), which were genotyped with 67 additional markers. A new genetic map was compiled, specific to our population, but based on the previously published map. QTL mapping was performed with two methods: half-sib regression analysis (paternal and maternal) and variance component analysis accounting for all family relationships. Two significant QTL were found for body weight on linkage group 4 and 6, six significant QTL for morphometric traits on linkage groups 1B, 4, 6, 7, 15 and 23 and three suggestive QTL for stress response on linkage groups 3, 14 and 23. The QTL explained between 8% and 38% of phenotypic variance. The results are the first step towards identifying genes involved in economically important traits like body weight and stress response in European sea bass.

Full text

Page 1

doi:10.1111/j.1365-2052.2009.02010.x
QTL for body weight, morphometric traits and stress response
in European sea bass Dicentrarchus labrax
C. Massault*,†, B. Hellemans‡, B. Louro§,*, C. Batargias¶, J. K. J. Van Houdt‡, A. Canario§,
F. A. M. Volckaert‡, H. Bovenhuis†, C. Haley* and D. J. de Koning*
*Division of Genetics and Genomics, Roslin Institute and Royal (Dick) School of Veterinary Sciences, University of Edinburgh, Roslin,
Midlothian EH25 9PS, UK.†Animal Breeding and Genetics Group, Wageningen University, PO Box 338, Wageningen, NL-6700AH,
The Netherlands.‡Laboratory of Animal Diversity and Systematics, Katholieke Universiteit Leuven, Ch. Deberiotstraat 32, B-3000 Leuven,
Belgium.§Centre of Marine Sciences (CCMAR), University of Algarve, Gambelas, 8005-139 Faro, Portugal.¶Laboratory of Applied Fish
Genetics, Department of Aquaculture & Fisheries Management, Technological Educational Institute of Messolonghi, Nea Ktiria, 30200
Messolonghi, Greece
Summary
Natural mating and mass spawning in the European sea bass (Dicentrarchus labrax L.,
Moronidae, Teleostei) complicate genetic studies and the implementation of selective breeding
schemes. We utilized a two-step experimental design for detecting QTL in mass-spawning
species: 2122 offspring from natural mating between 57 parents (22 males, 34 females and
one missing) phenotyped for body weight, eight morphometric traits and cortisol levels, had
been previously assigned to parents based on genotypes of 31 DNA microsatellite markers.
Five large full-sib families (five sires and two dams) were selected from the offspring (570
animals), which were genotyped with 67 additional markers. A new genetic map was
compiled, specific to our population, but based on the previously published map. QTL
mapping was performed with two methods: half-sib regression analysis (paternal and
maternal) and variance component analysis accounting for all family relationships. Two
significant QTL were found for body weight on linkage group 4 and 6, six significant QTL for
morphometric traits on linkage groups 1B, 4, 6, 7, 15 and 23 and three suggestive QTL for
stress response on linkage groups 3, 14 and 23. The QTL explained between 8% and 38% of
phenotypic variance. The results are the first step towards identifying genes involved in
economically important traits like body weight and stress response in European sea bass.
Keywords aquaculture, body weight, cortisol, Dicentrarchus labrax, European sea bass,
morphometry, QTL mapping, selection, stress response.
Introduction
European sea bass (Dicentrarchus labrax L., Moronidae, Tele-
ostei) is a marine fish, commonly distributed along the
warm temperate coasts of the south-eastern Atlantic Ocean
and Mediterranean Sea. While in 1980 only 10 tonnes were
produced in semi-intensive aquaculture systems, production
has stabilized since 2000 and the total worldwide produc-
tion was an estimated 105 900 tonnes in 2008 (http://
www.globefish.org). The value of sea bass aquaculture
totalled 320 million Euros in 2007, at an average value of
$5.1 per kg. Greece is the main producer (44%), followed by
Turkey (26%), Italy (12%) and Spain (8%), and sea bass
represent 40% of Mediterranean aquaculture (http://
www.feap.info).
Although sea bass aquaculture started 30 years ago,
selective breeding is not widely used (Vandeputte et al.
2001). For a long time, natural mating and mass spawning
were commonly used for reproduction. Artificial reproduc-
tion and mating are now fully controlled (Moretti 1999;
Saillant et al. 2001; Dupont-Nivet et al. 2006) and have
gradually become standard operational practice. In the case
of mass selection, the highly skewed unequal contribution of
parents that is so typical of mass-spawning fish (Jones &
Hutchings 2002; Herlin et al. 2008) leads to inbreeding.
Furthermore, the unknown pedigree of the individuals
Address for correspondence
C. Massault, Division of Genetics and Genomics, Roslin Institute and
Royal (Dick) School of Veterinary Sciences, University of Edinburgh,
Roslin, Midlothian EH25 9PS, UK.
E-mail: Cecile.massault@wur.nl
Accepted for publication 8 November 2009
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
337

Page 2

complicates selective breeding. However, the use of infor-
mation from QTL is a first step towards selective breeding.
Regions of the genome that are linked to a quantitative trait
of interest are detected using QTL mapping.
The steady increase in genetic and molecular biological
studies of European sea bass (for a review, see Volckaert
et al. 2008) provides a platform to implement breeding
programmes that do not depend solely on the measurement
of phenotypic traits, but rely also on information from
genetic markers, i.e. marker assisted selection. Large scale
profiling of paternity with microsatellite markers has facili-
tated the breeding of families on an experimental and
commercial scale in a reliable and affordable fashion (Garcı ´a
de Le ´on et al. 1995; Chatziplis et al. 2007). From such
communally bred families, quantitative genetic parameters
have been reported for heritabilities and phenotypic and
genotypic correlations for some traits, such as sex (Vande-
putte et al. 2007), body weight and length (Vandeputte &
Launey 2004; Saillant et al. 2006; Chatziplis et al. 2007)
and carcass traits (Saillant et al. 2009). Heritabilities for
growth are relatively high and vary from 0.29 to 0.60 in
European sea bass (Saillant et al. 2006; Dupont-Nivet et al.
2008). These high heritabilities have allowed a doubling in
growth rate in just four generations (B. Chatain, personal
communication). A medium density genetic map is avail-
able and has been updated (Chistiakov et al. 2005, 2008),
based on more than 200 microsatellites and more than 200
amplified fragment length polymorphisms (AFLPs). The
ESTs of several tissue-specific cDNA banks have been
sequenced, analysed (A. Canario, pers. comm.), and used as
a source of microsatellite (Chistiakov et al. 2008) and SNP
markers (Souche 2009). At the same time, a QTL analysis
for body weight has been performed in European sea bass
(Chatziplis et al. 2007) and has identified a QTL for growth
on linkage group 1.
An important economical trait is stress, either linked to
pathogen infection or behaviour (such as confinement and
handling). Cortisol, a widely accepted proxy for stress in
fish, affects the immune system response (Engelsma et al.
2002). Therefore, if genomic regions are found that influ-
ence the stress response, this could have a beneficial effect
on the management of fitness. All these features represent
multiple steps towards the use of genetic and molecular
tools for improvement of production.
Massault et al. (2008) suggested a two-step experimental
design for QTL mapping in mass-spawning species with
natural mating. In the first step, individuals are assigned to
parents using a limited number of markers, and in a second
step a subset of large families is selected for a whole genome
scan. So far, such a design has not been applied in practice
to fish. The strategy has been implemented in a single study
on European sea bass. Volckaert et al. (submitted) described
how the families for QTL analysis were established by
assigning parentage among the progeny of a 1-day batch
spawning. Heritabilities for growth, morphometry and
stress response were calculated on these data and are
reported by Volckaert et al. (submitted). The aim of this
study was to perform the QTL analysis on stress response,
body weight and morphological traits.
Materials and methods
Experiment
We followed a two-step procedure for the stress experiment
that takes into account the specificity of the natural mating
population, as suggested by Massault et al. (2008): (i) take a
random sample of 2122 offspring, originating from a single
day natural mating by mass spawning and assign parents
using a limited number of genetic markers, (ii) select large
families for further genotyping and QTL mapping.
Fifty-seven parents originating from commercial and wild
lines made up the base population. Offspring were chosen at
random and genotyped for 31 loci at 50 days of age. Based
on these genotypes, offspring were assigned to parents.
Confinement stress was induced on all 2122 fish at
approximately 8 months old, by reducing the volume of
water to 0.2 m3/kg of fish for 4 h (for details see Volckaert
et al. submitted).
Phenotyping
Following the stress experiment, individuals were pheno-
typed for ten traits: body weight, eight various morpho-
metric traits, and cortisol level (for full details see Volckaert
et al. submitted; see Fig. 1 and Table 1). About half of the
offspring were stunned in the morning using icy water
whereas the other half were stunned in the afternoon. After
being stunned, each fish was bled, weighted and digitally
photographed. The plasma, after separation from the red
blood cells, was conserved at )20 ?C. When photographed,
fish were placed as laterally as possible to avoid shape
variation, and landmarks were placed at various points on
the body after determining the common scale using a ruler
Figure 1 Morphometric measurement of European sea bass. Numbers
are related to the traits described in Table 1.
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
Massault et al.
338

Page 3

and coordinated axes (x, y and z). From those landmarks,
morphometric traits were observed.
The cortisol level was determined using radio immuno-
assay technology (RIA) based on antiserum raised in rabbit
against cortisol-3-CMO-BSA and tritriated cortisol (Fitzger-
ald Industries International). RIA cross-reacted with only
one plasma component, co-migrating with cortisol in layer
chromatography. Plasma samples were diluted in phosphate
buffer containing 0.5 g/l gelatine (pH 7.6) and heated at
80 ?C. We extracted the cortisol according to the protocol of
Scott et al. (1982), as detailed in Volckaert et al. (submit-
ted).
Genotyping and genetic map
Volckaert et al. (submitted), using a triple multiplex-PCR of
31 microsatellite markers, identified 11 large full-sib fami-
lies (n = 922) from the 2122 offspring for heritability and
correlation analysis (see above). Five of the eleven families
(n = 570 fish) and their seven parents were fully genotyped
for 98 microsatellite markers (with eight markers not
fulfilling minimal quality requirements). The population
structure for the QTL mapping includes the offspring from
five males and two females; the number of offspring per
family varies between 93 and 143 (Table 2). We generated
a genetic map specific to our population using the soft-
ware CRI-MAP V.2.4. (http://compgen.rutgers.edu/multimap/
crimap).
QTL mapping
We used two methods to detect QTL: half-sib regression
analysis as described by Knott et al. (1996), available on the
web (http://www.gridqtl.org.uk, Hernandez & Knott 2009),
and variance component analysis (VCA) as described by
George et al. (2000) (http://qtl.cap.ed.ac.uk/puccinoserv
lets/hkloaderLoki). The model used for half-sib analyses is as
follows:
yij¼ aiþ bixijþ eij;
where yijis the offspring phenotype, aithe mean for family i,
bithe regression coefficient for family i, xijthe probability of
inheriting parental allele 1 conditional on the marker
information and eijthe residual. Information content (IC)
was calculated as described by Knott et al. (1996). The QTL
effect is estimated within families as the allele substitution
effect. Knott et al. (1996) provide a method to calculate the
QTL effect across families in terms of variance because of the
QTL by looking at the difference between the residuals of full
and reduced models:
QTL effect ¼ 4 ? ð1 ? ðRES full=RES reducedÞÞ;
where RES full and RES reduced are the residuals of full and
reduced models respectively.
Regression was performed at each centiMorgan and the
test statistic for presence of a QTL was calculated as de-
scribed by Knott et al. (1996). We used permutation tests
with 500 iterations and bootstrap analysis with 2000 iter-
ations to evaluate the significance and confidence intervals
of detected QTL respectively. A QTL is considered significant
if it exceeds the 5% genome-wide threshold and a QTL is
considered suggestive if it exceeds the 5% chromosome-wide
threshold.
For VCA, we used the genotypes and the linkage map to
first calculate the identity-by-descent (IBD) matrix at each
Table 1 Number of observations, mean, standard deviation, coefficient of variation and heritabilities (standard error) for 10 traits in 570 European
sea bass.
Trait Abbreviation*No. individuals MeanSD CV (%)h2** (±SE)
Body weight
Cortisol
1. Standard length
2. Head length
3. Body length
4. Pre anal length
5. Abdominal length
6. Post-anal length
7. Head depth
8. Body depth
BW ())
CORT ())
SL (SL)
HL (SNOP)
BL (OPCA)
PrAnl (SNAN)
AL (OPAN)
PsAnl (ANCA)
HD ())
BD (DOPV)
566
444
540
540
540
540
540
540
540
540
40.58
319.5
13.26
3.77
9.50
9.20
5.57
4.46
2.46
3.38
14.21
136.40
1.61
0.46
1.17
1.16
0.72
0.54
0.33
0.41
35
43
12
12
12
13
13
12
13
12
0.54 (±0.2)
0.08 (±0.06)
0.65 (±0.22)
0.61 (±0.21)
0.64 (±0.22)
0.68 (±0.23)
0.66 (±0.23)
0.52 (±0.20)
0.64 (±0.22)
0.56 (±0.21)
*Abbreviations according to Chatziplis et al. 2007.
**Heritabilities calculated from 922 fishes (including QTL population) in Volckaert et al. (submitted).
Table 2 The structure of the QTL population of European sea bass. Five
males were mated with two females, providing full-sib families and
paternal and maternal half-sib families.
Half-sib IDFull-sib IDDam IDSire ID No. offspring
1
1
2
2
2
1
2
3
4
5
2
2
6
6
6
1
3
4
5
7
98
93
92
143
142
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
QTL in European sea bass
339

Page 4

position of the genome with Loki (Heath 1997). In the
analysis, only linkage information on the transmission of
alleles from parent to offspring was considered and IBD
relationships between parental alleles were not considered
(i.e. no linkage disequilibrium). The resulting IBD matrix
is used to model a QTL in a linear mixed model using
ASREML:
y ¼ Xb þ Zu þ Zv þ e;
where y is the phenotype, X is the incidence matrix relating
phenotypes to systematic environmental effects, b is the
vector with solutions for systematic environmental effects, Z
the incidence matrix relating animals to phenotype, u vector
of additive polygenic effect, v vector of additive QTL effect
and e vector of environmental effect (varðuÞ ¼ Ar2u and
ðvÞ ¼ Gr2v; with A the additive relationship matrix and G
the IBD matrix). For CORT, the environmental effect of
sampling time was included. The sampling time corresponds
to a half-day, resulting in three classes: first day morning,
first day afternoon and second day morning. For the other
traits the only other systematic effect was the mean. The
QTL effect is given by the VCA as the proportion of phe-
notypic variance because of the QTL.
The null hypothesis model is:
y ¼ Xb þ Zu þ e
with the same variable definitions as for the QTL model.
The presence of a QTL, treated as a random effect, was
tested at every centiMorgan using the likelihood ratio test:
LRT ¼ 2 ? ðlf ? lrÞ;
where LRT is the log likelihood ratio test and lf is the log
likelihood of the full model i.e. the model including the QTL
and lr the log likelihood from the reduced model i.e. the
model without a QTL. Under the null hypothesis, the like-
lihood ratio test follows a chi-square distribution with one
degree of freedom. For VCA, we computed 1% chromosome-
wide thresholds according to the method described by
Piepho (2001). The thresholds vary as a function of the
chromosome length and the trait studied.
Results
Summary statistics
Heritabilities for each trait (listed in Table 1 and including
abbreviations) were calculated for 922 animals, including
the QTL population (Volckaert et al. submitted), while all
other statistics were derived from the actual QTL population
(n = 570). CORT heritability is 0.08, while heritabilities of
the others traits ranged from 0.52 to 0.68. We note that all
morphometric traits (SL, HL, BL, PRAL, AL, POAL, HD and
BD) have a coefficient of variation (CV) within the same
range (12–13%), while BW and CORT are more variable
(CV of 35% and 43% respectively). Morphometric traits are
highly positively correlated among themselves and with BW
(see also Volckaert et al. submitted).
Genetic map
Based on the marker genotypes collected from 570 offspring
and their parents, we built a genetic map containing 20
linkage groups and covering 639 cM (Fig. 2). Of the 90
markers, 87 were located in linkage groups consisting of
two or more markers. Three markers were unlinked to any
of the other markers (linkage group LG9, LG18 and LG25).
The average marker spacing is 7.70 cM. The LG1 was cut
into two parts (1a and 1b) because of a gap >100 cM.
LG16, LG21 and LG22 are not represented in this genetic
map as no markers were located on them.
Information content
Table 3 summarizes the average IC of each linkage group
for both paternal and maternal half-sib (PHS and MHS)
analysis, as well as their number of markers, average
number of alleles and their length. The average IC of the
genome amounts to 0.77 for PHS and 0.81 for MHS: it
varied from 0.42 to 0.96 for PHS and 0.64 to 0.97 for MHS.
We observed an average IC lower than 0.5 on LG2 and LG4
for both PHS and MHS. Another region on linkage group
LG20 had low IC for PHS only.
QTL detected
Variance component analysis and regression analysis were
performed for each trait for the whole genome (on the 20
linkage groups). Regression analysis was divided into PHS
and MHS analysis (Table 4 and Fig. 3). Significant QTL for
regression analysis were found when the F-statistics were
above the 5% genome-wide threshold and suggestive when
above the 5% chromosome-wide thresholds, given by
GridQTL and significant when log likelihood (LR) is above
the 1% chromosome-wide threshold.
There were two BW QTL on LG4 and on LG6. The one on
LG4 was present and highly significant in all analyses,
although it was placed at different positions on the chro-
mosome (F = 5.65 for PHS with significance thresholds of
5% CW of 3.10 and 5% GW of 4.48, F = 7.09 for MHS with
significance thresholds of 5% CW of 3.76 and 5% GW of
7.02 and LR = 5.78 for VCA with a significance threshold
of 1%CW of 13.5). It was not possible to estimate the var-
iation because of the BW QTL on LG4 detected with VCA, as
the likelihood converged at the boundary of the parameter
space, i.e. at a heritability of that QTL of 0.99, but the other
analyses indicated a large QTL effect on this trait. Another
large BW QTL was detected on LG6 with the MHS method
(F = 8.59 with 5% CW of 3.82 and 5% GW of 7.02).
There was a large number of highly significant mor-
phology QTL, in particular for the PHS analysis. A common
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
Massault et al.
340

Page 5

QTL was found in LG4 by all analyses, but again at different
positions (F = 4.76 for PHS with 5% CW of 2.90 and 5%
GW of 4.18, F = 10.31 with 5% CW of 4.37 and 5% GW of
6.98 and LR = 15.46 for VCA with 1%CW of 13.03). VCA
showed a large proportion of phenotypic variation explained
by the QTL (38%), while this proportion was less for PHS
and MHS (13% and 14% respectively). Another morphology
QTL was detected with VCA on LG6 with a QTL effect of
about 9.4% (LR = 22.79 with 1% CW of 12.07). The other
morphology QTL found were located on four different
chromosomes and discovered with PHS analysis (F between
4.36 and 4.96 with 5% CW of 2.41–2.69 and 5% GW of
4.18). The QTL effects explained between 12% and 16% of
phenotypic variation.
Three QTL for CORT were detected with PHS (LG3 and
LG14, F = 2.69 with 5% CW of 2.70 and F = 3.01 with 5%
Figure 2 Genetic map of QTL-mapping sea bass population. The linkage groups LG16, LG21 and LG22 from Chistiakov et al. (2005) are missing as
no markers from those groups are present.
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
QTL in European sea bass
341

Page 6

CW of 2.74 respectively) and MHS (LG23, F = 6.59 with
5% CW of 3.25) and were 5% significant chromosome-wide.
These were therefore considered as suggestive QTL. None
were detected with VCA. The proportion of phenotypic
variation explained by the QTL varied between 8% and
10%.
Morphology and BW QTL on LG4 were detected in the
PHS as well as in the MHS analysis. The most likely
location of the QTL was different in the two analyses:
55 cM in the HS analysis compared with 0 cM in the
MHS analysis. However, the confidence intervals for QTL
position were very large and overlapped, indicating that
this might represent the same QTL. We can observe an
interesting phenomenon between BW and morphological
trait QTL; there were three BW-morphology QTL, one for
PHS on LG4 and two for MHS on LG4 and LG6. For each
pair of QTL, we noticed that both BW and morphology
QTL were located at the same position in the linkage
group. Body weight and all morphometric traits in this
study were highly correlated,
the observed link between both QTL (Volckaert et al.
submitted).
whichmight explain
Discussion
Linkage map and information content
The genetic map that we built resembles closely the previ-
ous genetic maps published by Chistiakov et al. (2005,
2008), with the difference that 23 new markers were added
to our map. Although it is specific to our QTL population,
the similarities between both of them show the consistency
and relevance of our results. The map is of medium density
with an average gap of about 8 cM, which is sufficient for
QTL mapping according to Chistiakov et al. (2005).
The low IC detected in some regions on the linkage
groups may impact the results, in particular for the QTL
detected on LG4, which are numerous. Results in some
regions become less reliable as the extremity of the
chromosome is not very informative.
Identification of QTL
We found a total of 15 different QTL using the three analyses
summarized in Table 3. The results found with VCA are
similar tothose found withMHS regression analysis. BW and
morphology QTL were discovered at the same linkage groups
and at similar positions; those traits were found to be highly
positivelycorrelated.MorphologyQTLonLG4explain38%of
phenotypic variation, but some of the effects might be
attributabletotheactualBWQTLonthesamelinkagegroup.
PHS detected more QTL, especially for the morphological
trait, than either MHS or VCA. PHS and MHS differ in the
number of QTL detected and in their position on the genome.
MoreQTLwerediscoveredforPHS.Thereareonlytwofemale
parents included in the analysis, and this might have an
impact on QTL discovery in MHS analysis. Potential QTL for
MHS might not have been detected, because the only two
females in the parental population did not segregate for this
trait. The 95% confidence intervals are very large for nearly
all QTL, covering the entire length of chromosomes. Fine
mapping, which is the next logical step in this process, must
use a more dense genetic map to be able to pinpoint inter-
estinggenes,butourresultsarepreciseenoughforafirstQTL
analysis.Forfine-mapping,theuseofmoremarkersmayhelp
toreducethe confidenceintervals andnarrowtheinteresting
regions that we found. At the moment, only microsatellites
are available, but development of AFLPs (Chistiakov et al.
2008) and soon SNPs may accelerate the process of fine-
mapping. Although the increase in the number of markers
influences the precision of the results, population structure
has an impact on the power to detect QTL. In this study, we
have a mixed population of half-sib and full-sib. To have the
best results, we selected the largest families. However,
because of the unequal contribution of parents, the largest
families will represent only very few parents.
The number of QTL found with the VCA is lower than
with half-sib regression. This might be attributed to the
Table 3 Number of markers, average number of alleles per marker,
length of the linkage group, and average information content for
paternal and maternal half-sib regression analysis (PHS and MHS) of
European sea bass.
LG
No.
markers
Average no.
allelesLength
Avg IC
PHS
Avg IC
MHS
LG1a
LG1b
LG2
LG3
LG4
LG5
LG6
LG7
LG8
LG10
LG11
LG12
LG13
LG14
LG15
LG17
LG19
LG20
LG23
LG24
5
6
7
4
6
6
5
4
7
4
2
4
3
6
4
3
3
3
2
2
5.2
4.3
3.9
3.5
3.7
4
3.8
6
4.3
4.3
3.5
5.5
5
3.8
3
4.8
4
3
4.5
7
530.80
0.96
0.701
0.83
0.592
0.88
0.81
0.89
0.85
0.83
0.86
0.92
0.92
0.79
0.82
0.89
0.82
0.423
0.88
0.93
0.76
0.97
0.684
0.81
0.645
0.91
0.80
0.84
0.88
0.80
0.67
0.90
0.87
0.77
0.81
0.92
0.67
0.62
0.92
0.97
7
73
13
106
45
24
34
42
36
5
14
15
59
11
17
34
36
9
6
1Average of 0.49 between 50 and 46 cM.
2Average of 0.27 between 0 and 34 cM.
3Average of 0.32 between 12 and 36 cM.
4Average of 0.46 between 44 and 61 cM.
5Average of 0.37 between 0 and 33 cM.
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
Massault et al.
342

Page 7

approximate 1% chromosome-wide thresholds calculated
with Piepho?s method (Piepho 2001). Those thresholds are
adjusted for the marker density of the genetic map.
Chatziplis et al. (2007) mention a QTL associated with
morphometric traits located at the beginning of LG1; we
also detected a morphology QTL on LG1. The two mapping
Table 4 Body weight, morphological and cortisol QTL for paternal half-sib (PHS), maternal half-sib (MHS) and variance component analysis (VCA) of
European sea bass.
TraitMethod
Position
in cM
No. segregating
parents
Linkage
groupF
5% CW
(HS)
5% GW
(HS)LR
1% CW
(VCA)
QTL
effect (%)1
Confidence
Interval (cM)
BW
BW
BW
BW
MORPH
MORPH
MORPH
MORPH
MORPH
MORPH
MORPH
MORPH
CORT
CORT
CORT
PHS
MHS
VCA
MHS
PHS
MHS
VCA
VCA
PHS
PHS
PHS
PHS
PHS
PHS
MHS
543
2
–
2
3
2
–
–
2
3
1
2
2
2
1
4
4
4
6
4
4
4
6
5.65
7.09
–
8.59
4.76
10.31
–
–
4.36
5.71
4.21
4.93
2.69
3.01
6.59
3.10
3.76
–
3.82
2.90
4.37
–
–
2.41
2.60
2.56
2.50
2.70
2.74
3.25
4.48
7.02
–
7.02
4.18
6.98
–
–
4.18
4.18
4.18
4.18
4.59
4.59
7.61
–
–
5.78
–
–
–
15.46
22.79
–
–
–
–
–
–
–
–
–
2.23
–
–
–
13.03
12.07
–
–
–
–
–
–
–
16 1–88
0–76
NA2
5–22
1–105
0–4
NA2
NA2
1–7
0–34
0–11
0–6
0–13
0–59
0–9
0
3
8
–
12
55
10
13
14
38
2
2
19 9.4
14
16
12
13
8
9
10
51B
7
15
24
3
14
23
30
0
0
3
1
3
BW, body weight; CORT, cortisol level; MORPH, combination of all morphometric traits (HD, HD, BD, BL, AL, PRAL, POAL and SL); PHS, paternal
half-sib analysis; MHS, maternal half-sib analysis; VCA, variance component analysis; cM, centiMorgan.
1QTL effect: proportion of the phenotypic variation explained by the QTL.
2Not applicable.
Figure 3 QTL in bold are 5% genome-wide,
whereas those underlined are 5% chromo-
some-wide. MHS stands for maternal half-sib
regression, PHS for paternal half-sib regression
and VCA for variance component analysis.
The traits are MORPH for morphology and BW
for body weight.
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
QTL in European sea bass
343

Page 8

populations differ, although they belonged to the same
company. We split this linkage group into LG1a and LG1b
because of a large region in the middle of LG1 devoid of
markers. Our QTL is located at the end of the linkage group,
in the LG1b section. The high confidence intervals suggest
that the location of the QTL could be anywhere on the
chromosome, and therefore this result potentially confirms
the results of Chatziplis et al. (2007). These authors also
mention a possible BW QTL on LG1, but we could not
support this finding. No other QTL were reported in Euro-
pean sea bass, suggesting that we found novel significant
and suggestive QTL for this species.
We found three suggestive QTL for stress response.
Although they are not highly significant, it is an important
discovery, because QTL for stress response have not been
reported before. A candidate gene for this trait could be the
glucocorticoid receptor, which has been sequenced and whose
position remains unknown (Terova et al. 2005). Several
genes related to carbohydrate metabolism, genes involved
in Na+/K+transport, polyamine biosynthesis and iron
homeostasis (Sarropoulou et al. 2005) could be also taken
as candidate genes, as they are involved in the mechanism
of stress response. Our identified regions could be of high
interest for candidate gene approach studies.
Methodological comparison
We compared half-sib regression and VCA, and each
method had its advantages and drawbacks. We performed
separately a PHS regression analysis and MHS regression
analysis. Therefore, in the regression analysis, the full-sib
relationships are not taken into account. We additionally
performed a VCA that deals with all family relations in our
experimental population. Half-sib regression does not
require large computing resources and time, and therefore it
allows a permutation test to be performed and confidence
intervals to be calculated using bootstrapping.
Variance component analysis has been implemented to
simultaneously account for all relations in our data. The
QTL is modelled as a random effect in the model; therefore,
the QTL may have an infinite number of alleles. However,
VCA is computationally demanding and time consuming,
especially for the calculation of IBD and thresholds for 5%
chromosome-wide significance when computed according
to an approximation given by Piepho (2001).
Power of the experimental design
This study has implemented the two-step procedure devel-
oped by Massault et al. (2008). When applying simulated
parameters from our specific population (rounded to 600
individuals), we found that, for an 80% power, we were
theoretically able to detect QTL explaining at least 14.5% of
phenotypic variation using half-sib regression analysis and
at least 6.5% of phenotypic variation using VCA. Mor-
phology QTL on LG4 explain a large proportion of pheno-
typic variation (38%) and will have been detected with a
less powerful experimental design. In the case of PHS
regression, this QTL could be associated with marker
DLA0175, whereas for MHS and VCA, it will be more linked
to a marker at the beginning of the linkage group DLA0166.
In contrast, morphology QTL on LG6 might not have been
detected, as the QTL effect was 9.4% of the phenotypic
variation (against 14.5% expected theoretically). Our design
was powerful enough to detect a QTL of medium size, which
is proof that the design is perfectly adapted for a natural
mating population. The proportion of phenotypic variation
explained by QTL detected with half-sib regression was of
medium size effect (8–16%). This shows that our method is
able to detect not only QTL with large effects, but also
medium effects for sea bass, and could be generalized to
mass-spawning species.
The fact that there were only two female parents involved
in the experiment may also have an impact on the success
of detecting QTL. The heterozygosity for each allele is
therefore none, half or one. As explained earlier, existing
QTL might be ignored because none of the females are
segregating for them. The power therefore might be lower,
as in theory we assumed a heterozygosity of 0.5, which in
reality is impossible to predict.
The theory was based on a sparse genetic map, with
marker spacing of 20 cM. Our genetic map had average
marker intervals close to 8 cM. A denser map could have
improved the power, and therefore our experiment might
not have detected some QTL.
The power of the experiment depended on the herita-
bility of the trait. We found three suggestive CORT QTL
for stress response. The heritability of this trait shown in
Table 1 is 0.08. Massault et al. (2008) have set a high
heritability of 0.5 for the two-step experimental design.
Trait heritability is known to play a role in QTL discovery
(Kolbehdari et al. 2005), and in the case of low herita-
bility such as cortisol, there is less chance of detecting
important QTL.
Our study shows that QTL mapping is possible for natural
mating mass-spawning species, using a specific experimen-
tal design and tools such as parentage assignment. The
results of this experiment are preliminary, but they are
promising. QTL were detected and could be located at spe-
cific chromosomes, so that complementary fine-mapping
can be undertaken.
Acknowledgements
This study has been funded by the European Commission
(project AQUAFIRST,contract
2004–513692). DJK and CSH acknowledge financial sup-
port from the BBSRC. E. Couto, S. Darivianaku, J. Fuentes,
P. Guerreiro, M. Kampakli, G. Kotoulas and D. Power helped
with phenotyping on-site.
numberFP6-STREP-
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
Massault et al.
344

Page 9

References
ChatziplisD.,BartagiasC.,TsigenopoulosC.S.,MagoulasA.,KolliasS.,
Kotoulas G., Volckaert F.A.M. & Haley C.S. (2007) Mapping
quantitative trait loci in European sea bass (Dicentrarchus labrax):
theBASSMAPpilotstudy.Aquaculture 272,S172–82.
Chistiakov D.A., Hellemans B., Haley C.S., Law A.S., Tsigenopoulos
C.S., Kotoulas G., Bertotto D., Libertini A. & Volckaert F.A.M.
(2005) A microsatellite linkage map of the European sea bass
Dicentrarchus labrax L. Genetics 170, 1821–6.
Chistiakov D.A., Tsigenopoulos C.S., Lagnel J., Yuanmei G., Helle-
mans B., Haley C.S., Volckaert F.A.M. & Kotoulas G. (2008) A
combined AFLP and microsatellite linkage map and pilot com-
parative genomic analysis of European sea bass Dicentrarchus
labrax L. Animal Genetics 39, 623–34.
Dupont-Nivet M., Vandeputte M., Haffray P. & Chevassus B. (2006)
Effect of different mating designs on inbreeding, genetic variance
and response to selection when applying individual selection in
fish breeding programs. Aquaculture 252, 161–70.
Dupont-Nivet M., Vandeputte M., Vergnet A., Merdy O., Haffray P.,
Chavanne H. & Chatain B. (2008) Heritabilities and GXE inter-
actions for growth in European sea bass (Dicentrarchus labrax L.)
using a marker-based pedigree. Aquaculture 275, 81–7.
Engelsma M.Y., Huising M.O., van Muiswinkel W.B., Flik G., Kwang
J., Savelkoul H.F.J. & Verburg-van Kemenade B.M.L. (2002)
Neuroendocrine-immune interactions in fish: a role for interleu-
kin-1. Veterinary Immunology and Immunopathology 87, 467–79.
Garcı ´a de Leo ´n F.J., Dallas D.J., Chatain B., Canonne M., Versini J.J.
& Bonhomme F. (1995) Development and use of microsatellite
markers in seabass, Dicentrarchus labrax (Linnaeus, 1758)
(Perciformes: Serranidae). Molecular Marine Biology and Biotech-
nology 4, 62–8.
George A., Visscher P. & Haley C.S. (2000) Mapping quantitative
trait loci for complex pedigree: a two-step variance component
approach. Genetics 156, 2081–92.
Heath S.C. (1997) Markov chain Monte Carlo methods for radiation
hybrid mapping. Journal of Computational Biology 4, 505–15.
Herlin M., Delghandi M., Wesmajervi M., Taggart J.B., McAndrew
B.J. & Penman D.J. (2008) Analysis of the parental contribution
to a group of fry from a single day of spawning from a com-
mercial Atlantic cod (Gadus morhua) breeding tank. Aquaculture
274, 218–24.
Hernandez J. & Knott S. (2009) Haplotyping via minimum re-
combinant paradigm. BMC Proceedings 3(Suppl. 1), S7.
Jones M.W. & Hutchings J.A. (2002) Individual variation in
Atlantic salmon fertilization success: implications for effective
population size. Ecological applications 12, 184–93.
Knott S., Elsen J. & Haley C.S. (1996) Methods for multiple-marker
mapping of quantitative trait loci in half-sib populations. Theo-
retical and Applied Genetics 93, 71–80.
Kolbehdari D., Jansen G.B., Schaeffer L.R. & Allen B.O. (2005)
Power of QTL detection by either fixed or random models in half-
sib designs. Genetics Selection Evolution 37, 601–14.
Massault C., Bovenhuis H., Haley C.S. & de Koning D.J. (2008) QTL
mapping designs for aquaculture. Aquaculture 285, 23–9.
Moretti A. (1999) Manual on Hatchery Production of Seabass and
Gilthead Seabream. Food and Agricultural Organisation, Rome.
Piepho H.P. (2001) A quick method for computing approximate
thresholds for quantitative trait loci detection. Genetics 157, 425–
32.
Saillant E., Chatain B., Forstier A., Przybyla C. & Fauvel C. (2001)
Parental influence on early development in the European sea
bass. Journal of Fish Biology 58, 1585–600.
Saillant E., Dupont-Nivet M., Haffray P. & Chatain B. (2006) Esti-
mates of heritability and genotype-environment interactions for
body weight in sea bass (Dicentrarchus labrax L.) raised under
communal rearing conditions. Aquaculture 254, 139–47.
Saillant E., Dupont-Nivet M., Sabourault M., Haffray P., Laureau S.,
Vidal M.O. & Chatain B. (2009) Genetic variation for carcass
quality traits in cultured sea bass (Dicentrarchus labrax). Aquatic
Living Resources 22, 105–12.
Sarropoulou E., Kotoulas G., Power D.M. & Geisler R. (2005) Gene
expression profiling of gilthead sea bream during early devel-
opment and detection of stress-related genes by application of
cDNA microarray technology. Physiological Genomics 23, 182–
91.
Scott A., Sheldrick E. & Flint A. (1982) Measurement of 17[alpha],
20[beta]-dihydroxy-4-pregen-3-one in plasma in trout (Salmo
gairdneri Richardson): seasonal changes and response to salmon
pituitary extract. General and Comparative Endocrinology 46, 444–
51.
Souche E. (2009) Genomic Variation in European Sea Bass Dicen-
trarchus labrax: From SNP Discovery Within ESTs to Genome Scan.
PhD Thesis, Katholieke Universiteit Leuven, Leuven, Belgium.
Terova G., Gornati R., Rimoldi S., Bernardini G. & Saroglia M.
(2005) Quantification of a glucocorticoid receptor in sea bass
(Dicentrarchus labrax, L.) reared at high stocking density. Gene
357, 144–51.
Vandeputte M. & Launey S. (2004) The genetic management of fish
domestication. Productions Animales 17, 237–42.
Vandeputte M., Dupont-Nivet M., Chatain B. & Chevassus B. (2001)
Setting up a strain-testing design for the seabass, Dicentrarchus
labrax: a simulation study. Aquaculture 202, 329–42.
Vandeputte M., Dupont-Nivet M., Chavanne H. & Chatain B.A.
(2007) Polygenic hypothesis for sex determination in the Euro-
pean sea bass Dicentrarchus labrax. Genetics 176, 1049–57.
Volckaert F.A.M., Batargias C., Canario A., Chatziplis D., Chistiakov
D., Haley C., Libertini A. & Tsigenopoulos C. (2008) European sea
bass. In: Genome Mapping and Genomics in Animals (Ed. by T.D.
Kocher & C. Kole), Volume 2: 117–33. Springer-Heidelberg,
Berlin.
Volckaert F.A.M., Hellemans B., Batargias C., Louro B., Massault C.,
Van Houdt J.K.J., Haley C., de Koning D.J. & Canario A.V.M.
(submitted) Family assignment and heritability of body weight,
morphometric traits and stress response in aquacultured Euro-
pean sea bass Dicentrarchus labrax.
? 2009 The Authors, Journal compilation ? 2009 Stichting International Foundation for Animal Genetics, Animal Genetics, 41, 337–345
QTL in European sea bass
345