GENOMIC SELECTION

Genomic Predictions and Genome-Wide Association
Study of Resistance Against Piscirickettsia salmonis
in Coho Salmon (Oncorhynchus kisutch) Using
ddRAD Sequencing
Agustín Barría,*,† Kris A. Christensen,‡,1 Grazyella M. Yoshida,*,§ Katharina Correa,*,** Ana Jedlicki,*
Jean P. Lhorente,** William S. Davidson,‡ and José M. Yáñez*,**,††,2

*Facultad de Ciencias Veterinarias y Pecuarias, Universidad de Chile, La Pintana, Santiago 8820808, Chile, †Doctorado en
Acuicultura, Programa Cooperativo Universidad de Chile, Universidad Católica del Norte, Pontificia Universidad Católica
de Valparaíso, 8820808 Chile, ‡Department of Molecular Biology and Biochemistry, Simon Fraser University, Burnaby, BC
V5A 1S6, Canada, §Animal Science Department, Universidade Estadual Paulista “Júlio de Mesquita Filho”, Faculdade de
Ciências Agrárias e Veterinárias, Campus Jaboticabal, Jaboticabal 14884-900, Brazil, **Aquainnovo S.A., Puerto Montt
5503032, Chile, and ††Núcleo Milenio INVASAL, Concepción 4070386, Chile

ORCID IDs: 0000-0002-2813-4559 (A.B.); 0000-0003-0021-0326 (K.A.C.); 0000-0002-6788-7369 (G.M.Y.); 0000-0003-1386-8522 (K.C.);
0000-0003-0626-1247 (A.J.); 0000-0002-9157-4231 (J.P.L.); 0000-0002-3002-3225 (W.S.D.); 0000-0002-6612-4087 (J.M.Y.)

ABSTRACT Piscirickettsia salmonis is one of the main infectious diseases affecting coho salmon (Onco-
rhynchus kisutch) farming, and current treatments have been ineffective for the control of this disease.
Genetic improvement for P. salmonis resistance has been proposed as a feasible alternative for the control
of this infectious disease in farmed fish. Genotyping by sequencing (GBS) strategies allow genotyping of
hundreds of individuals with thousands of single nucleotide polymorphisms (SNPs), which can be used to
perform genome wide association studies (GWAS) and predict genetic values using genome-wide infor-
mation. We used double-digest restriction-site associated DNA (ddRAD) sequencing to dissect the genetic
architecture of resistance against P. salmonis in a farmed coho salmon population and to identify molecular
markers associated with the trait. We also evaluated genomic selection (GS) models in order to determine
the potential to accelerate the genetic improvement of this trait by means of using genome-wide molecular
information. A total of 764 individuals from 33 full-sib families (17 highly resistant and 16 highly susceptible)
were experimentally challenged against P. salmonis and their genotypes were assayed using ddRAD sequenc-
ing. A total of 9,389 SNPs markers were identified in the population. These markers were used to test genomic
selection models and compare different GWAS methodologies for resistance measured as day of death (DD)
and binary survival (BIN). Genomic selection models showed higher accuracies than the traditional pedigree-
based best linear unbiased prediction (PBLUP) method, for both DD and BIN. The models showed an
improvement of up to 95% and 155% respectively over PBLUP. One SNP related with B-cell development
was identified as a potential functional candidate associated with resistance to P. salmonis defined as DD.

KEYWORDS

selective
breeding

genotyping by
sequencing

Oncorhynchus
kisutch

disease
resistance

GWAS
Genomic
Selection

GenPred
Shared Data
Resources

Chile is the largest producer of coho salmon (Oncorhynchus kisutch)
globally, reaching about 160,000 tons in 2014, representing more than
90% of total production (FAO 2016). However, the success and sustain-
ability of this industry is constantly threatened by infectious diseases,
including Salmon Rickettsial Syndrome (SRS). This disease is caused by
Piscirickettsia salmonis, a gram-negative and facultative intracellular
bacteria, which was isolated for the first time in Chile in coho salmon
(Cvitanich et al. 1991). Data from the Chilean National Fisheries and

Aquaculture Service (Sernapesca) indicates that during the first half of
2016, 53% of the moralities ascribed to infectious diseases in coho
salmon were associated with SRS (Sernapesca 2016). To date, control
measures and treatments for SRS are based on antibiotics and vaccines.
However, both strategies have not had the expected effectiveness under
field conditions (Rozas and Enríquez 2014).

Because of this, it is necessary to develop alternative strategies for the
control of this disease (Yáñez et al. 2014a). In this regard, breeding for

Volume 8 | April 2018 | 1183

http://orcid.org/0000-0002-2813-4559
http://orcid.org/0000-0003-0021-0326
http://orcid.org/0000-0002-6788-7369
http://orcid.org/0000-0003-1386-8522
http://orcid.org/0000-0003-0626-1247
http://orcid.org/0000-0002-9157-4231
http://orcid.org/0000-0002-3002-3225
http://orcid.org/0000-0002-6612-4087


enhanced disease resistance is a feasible and sustainable option to
improve animal health, welfare and productivity (Stear et al. 2001).
A primary requisite for including disease resistance into a breeding
program is the presence of significant additive genetic variation for
the trait (Falconer and Mackay 1996). Commonly, data to evaluate
resistance comes from experimental challenges carried out using sib-
lings of the selection candidates (Ødegård et al. 2011; Yáñez et al.
2014a). Quantitative studies have estimated significant genetic varia-
tion for resistance against different pathogens in salmonid species
(Ødegård et al. 2011; Yáñez et al. 2014a). For instance, low to mod-
erate heritabilities for resistance against P. salmonis in Atlantic
salmon (Salmo salar) (h2 = 0.11 to 0.41) (Yáñez et al. 2013; Yáñez
et al. 2014b) and coho salmon (h2 = 0.16) (Yáñez et al. 2016a) have
been estimated.

Marker assisted selection (MAS) can improve production traits in
cases where the phenotypes are difficult to measure in the selected
candidates (e.g., disease resistance traits) and the total additive ge-
netic variance explained by genetic markers is high (Hayes and
Goddard 2010). This methodology has been successfully applied
for the improvement of resistance against the Infectious Pancreatic
Necrosis Virus (IPNV) in Atlantic salmon, which is controlled by a
major quantitative trait locus (QTL) (Houston et al. 2012; Moen
et al. 2015). In the case of polygenic traits, genomic selection (GS)
(Meuwissen et al. 2001) can significantly improve selection accuracy
of breeding values compared to traditional selection, and therefore
enhance the response of selection for disease resistance in salmonid
species (Tsai et al. 2016; Vallejo et al. 2016, 2017a; Bangera et al.
2017; Correa et al. 2017; Yoshida et al. 2017).

Genotyping by sequencing (GBS) is an alternative for genotyping in
cases when SNP panels are not available. This approach reduces the
complexity of the genome, and can be used to identify thousands of
markers without prior marker discovery efforts or a reference genome.
Currently, several approaches of GBS have been developed, significantly
reducing the cost and labor (Baird et al. 2008; Elshire et al. 2011;
Peterson et al. 2012). These methodologies have been widely used in
salmonid species, to generate dense linkage maps (Brieuc et al. 2014;
Gonen et al. 2014), perform association studies to identify genomic
regions involved in the resistance against pathogens (Campbell et al.
2014; Liu et al. 2015; Palti et al. 2015b) and generate SNPs resources
(Houston et al. 2012).

Double-digest restriction-site associated DNA (ddRAD) reduces
DNA complexity by digesting DNA with two restriction enzymes
(REs) simultaneously, without random shearing (Peterson et al.
2012). This approach has been widely used in genetic studies in aqua-
culture species (reviewed in Robledo et al. 2017).

In the present study, we used ddRAD sequencing to dissect the
genetic architecture of resistance against P. salmonis in a farmed coho
salmon population and identify molecular markers associated with the
trait. Furthermore, GS models were used to evaluate the potential to
accelerate the genetic improvement of resistance against P. salmonis in
this coho salmon population by means of using genome-wide molec-
ular information.

MATERIALS AND METHODS

Coho salmon breeding population
The coho salmon population used in the present study came from a
unique 2012 year-class population. This population belongs to a
genetic improvement program that was established in 1997 and is
owned by Pesquera Antares and managed by Aquainnovo (Puerto
Montt, Chile). Further details about this breeding population, in
terms of reproductive management, rearing conditions, fish tagging
and breeding objectives are described by Yáñez et al. (2014c; 2016a)
and Dufflocq et al. (2016).

Experimental challenge
The experimental challenge against P. salmonis was performed as it is
described in detail by Correa et al. (2015a) and Yáñez et al. (2016a).
Briefly, 2,606 individuals, belonging to 107 maternal full-sib families
and 60 paternal half-sib families, were challenged against P. salmonis.
Prior to the experiment, each fish was tagged with a passive integrated
transponder (PIT-tag), placed in the abdominal cavity for genealogy
traceability during the challenge test. The P. salmonis challenge was
performed at Aquainnovo´s research station, located in Lenca River, X
Region, Chile.

For the lethal dose 50 (LD50) calculation, a random sample of 80 fish
were selected from the population. Four different dilutions from the P.
salmonis inoculum were evaluated (1/10, 1/100, 1/1000 and 1/10000).
Twenty fish were challenged at each dilution. The dilutions were in-
traperitoneally (IP) injected with a volume of 0.2ml/fish. Daily mortal-
ity was recorded. This preliminary test spanned 26 days and a dilution
of 1:680 was estimated as the LD50.

For the main challenge, fish were distributed into three tanks (7m3)
with a salt water concentration of 31 ppt. An average of eight individ-
uals (ranging from 1 to 18) from each of the 107 families were distrib-
uted into each tank. The experimental challenge was performed
through an intraperitoneal (IP) injection with 0.2ml/fish of the LD50

inoculum. The average weight of the fish at the inoculation was 279g
(SD = 138g). To ensure that these fish were free from other pathogens,
qRT-PCR was previously performed in order to control for the pres-
ence of Infectious Salmon Anemia Virus (ISAV), IPNV and Flavobac-
terium spp.

The P. salmonis challenge continued for up to 50 days post IP in-
jection. Throughout the challenge, environmental parameters (pH,
temperature, salinity and oxygen) were measured and controlled. Fish
were removed from the tanks after death, and a sample of the anterior
kidneywas taken and stored at -80� in RNALater. A necropsy assay was
performed in conjunction with qRT-PCR to confirm the cause of death
and the presence of P. salmonis. This was also done to control for the
presence of other pathogens, such as Vibrio ordalii, Renibacterium
salmoninarum and IPNV.

ddRAD library preparation and sequencing
Ten ddRAD libraries were produced by multiplexing 828 individuals
following the protocol described by Peterson et al. (2012). For this,
64 parents (males and females) and 764 offspring representing the

Copyright © 2018 Barria et al.
doi: https://doi.org/10.1534/g3.118.200053
Manuscript received September 26, 2017; accepted for publication February 1,
2018; published Early Online February 22, 2018.
This is an open-access article distributed under the terms of the Creative Commons
Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/),
which permits unrestricted use, distribution, and reproduction in any medium,
provided the original work is properly cited.
Supplemental Material is available online at www.g3journal.org/lookup/suppl/
doi:10.1534/g3.118.200053/-/DC1
1Present address: Marine Ecosystems and Aquaculture Division, Science Branch
Fisheries and Oceans Canada/Government of Canada, 4160 Marine Drive,
West Vancouver, BC, Canada V7V 1N6.

2Corresponding author: Facultad de Ciencias Agronómicas, Universidad de Chile,
Avenida Santa Rosa 11735, 8820808, La Pintana, Santiago, Chile. E-mail:
jmayanez@uchile.cl

1184 | A. Barría et al.

https://doi.org/10.1534/g3.118.200053
http://creativecommons.org/licenses/by/4.0/
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1
mailto:jmayanez@uchile.cl


17 most resistant and 16 most susceptible families were selected. An
average of 23 (ranging from 11 to 43 individuals) offspring per family
were chosen. Briefly, total DNAwas extracted using the commercial kit
Wizard SVGenomic DNA purification System (Promega) according to
the manufacturer’s protocol. Between 80 and 200 ng of DNA, from
each individual was digested with two restriction enzymes (New Eng-
land Biolabs, UK; NEB); 10 U of SbfI (specific for the CCTGCA|GG
motif)) and MseI (specific for the T|TAA motif) in a 12 ml reaction
volume, including 1ml of SbfI andMseI adapter (8.3 pM), for 90min at
37�. The ligation reaction was carried out by adding 1 ml of T4 ligase
(NEB) diluted 1:100 in T4 buffer and incubating for 150min at 37� and
subsequently at 16� overnight.

Each ligation mix was diluted with 189 ml of dilute TE buffer
(1:10). Kodak DNA Polymerase (ABM), a high-fidelity polymerase,
was used to amplify DNA fragments with the correct adapters. PCR
reactions (20 ml) were prepared containing 10 ml of PCR mix 2x,
1 ml of primer mix (10 mM each), 6 ml of diluted ligation mix and
3 ml of nuclease-free water. Each sample was PCR amplified using
the following conditions: 95� for 2 min, followed by 17 cycles of 95�
for 20s, 66� for 30s and 68� for 40s. After PCR, amplicon quality was
checked by loading 5 ml on a 2% agarose gel. Subsequently, samples
were pooled, so that the final concentration was similar among them
within each library. Each library was concentrated through an evap-
oration step for 80 min in a Centrivap Mobile Console Centrifugal
Evaporator (Labconco). This step was conducted until 300 ml of the
generated library was obtained. Final volume of each library was
loaded on a 1% agarose gel. Size of the bands selected for sequencing
ranged from 750 and 1,500 bp and between 1,800 and 2,500 bp.
DNA was purified through the QIAquick gel extraction kit (Qiagen)
following manufacturer’s instructions. Finally, libraries were se-
quenced on an Illumina Hiseq2500 platform, using 150 base
single-end.

SNP identification
Raw sequence reads obtained from Illumina sequencing were analyzed
using STACKS v. 1.41 (Catchen et al. 2011, 2013). This software was
specifically developed to analyze short-read data generated through
next generation sequencing (NGS) (Davey et al. 2013).

Sample reads were trimmed to 134 bp for all subsequent analyses,
demultiplexed and filtered using process_radtags. Rad-tags which
passed the quality filter were aligned to the Oncorhynchus kisutch
reference genome (GenBank: MPKV00000000.1) using BWA v.
0.7.12 (Li and Durbin 2009). The reference genome was indexed
(using the index function) and alignments were performed using the
mem algorithm; all parameters were set as default. Loci were then
built using pstacks with a minimum depth of coverage of three to
build a locus (-m 3).

A catalog of loci was constructed using the cstacks program using
only the parents’ loci from pstacks. To build the catalog, the maximum
number ofmismatches allowed between sample tagswas set to three (-n
3), and the matching was based on genomic location (g). After catalog
construction, the sstacks program was used in order to match rad-tags
against the catalog based again on genomic location (g), followed by the
populations software, using defaults parameters. Loci were considered
as valid if they were present in at least 75% of the individuals of the
population. As a quality control step, the following parameters were
used to filter low-confidence SNPs: Minor Allele Frequency (MAF) #
0.05, Hardy-Weinberg Equilibrium (HWE) P , 1x10-6 and genotyp-
ing call rate, 0.80. Individuals were removed from the analyses if their
genotyping call rates were below 0.70.

Trait definitions
Resistance against P. salmonis was defined as the day of death (DD)
with values ranging from 1 to 50 depending on the time of death.
Additionally, resistance was also evaluated as a binary (BIN) trait, either
dead or alive at the end of the challenge. Values for this trait were 1 in
cases where the fish died during the challenge, or 0 if the fish survived
until the end of the challenge. Initial Body Weight (IW) for each fish,
was measured prior to the IP injection.

Pedigree-based BLUP
All challenged individuals (n = 2,606) were used for the pedigree-based
approach, PBLUP, as a control for the performance evaluation of
genomic predictions. A linear univariate animal model was used to
estimate variance components and predict Estimated Breeding Values
(EBVs) for DD, while for BIN a univariate threshold animal model was
fitted (Table 1). The model used was as follows:

y ¼ Xbþ Tpþ e

In the previous equation, y is a vector of phenotypes (BIN or DD), b is
a vector of fixed effects (sex and tank as factors, and initial weight as
covariate), p is a vector of random additive polygenic genetic effects
that follows a normal distribution�N(0,As2

p), X and T are incidence
matrices, A is the additive relationship matrix, and e is the random
residual (Lynch and Walsh 1998). Both models were fitted using the
BLUPF90 set of programs (Misztal et al. 2016) by means of the AIR-
EMLF90 and THRGIBBS1F90 modules to analyze DD and BIN, re-
spectively. The MCMC Gibbs sampling scheme set for running
THRGIBBS1F90, included a total of 200,000 iterations. The first
20,000 were discarded as burn-in iterations, and then from the
remaining 180,000 samples, one from every 50 samples was saved
for analysis. This Gibbs sampling scheme collected 3,600 independent
samples for analysis.

Heritabilities for PBLUP models were computed as follows:

h2i ¼
s2
ai

s2
ai þ s2

ei

where s2
ai and s2

ei are the additive genetic and residual variances for
each trait. In the case of BIN, the residual variance was set to 1.

Genomic BLUP
The SNP based variance components and GEBVs were estimated using
genomic BLUP (GBLUP), similar to the PBLUPmodel described above,
and implemented in the BLUPF90 software. The GBLUP is a modifi-
cation of the PBLUP method, where g is a vector of random additive
genetic polygenic effects with a distribution �N(0, Gs2

g) and the nu-
merator relationship matrix A is replaced by a genomic relationship
matrix G, as described by (VanRaden 2008). Only genotyped animals,
which passed all quality controls (n = 580) were analyzed.

Single step genomic GBLUP
The single-step GBLUP (ssGBLUP) and weighted single-step GBLUP
(wssGBLUP) models were similar to the PBLUP model except for a
combined genomic and pedigree relationship. The kinship matrix used
was H (Aguilar et al. 2010), in which genotype and pedigree data are
combined. The inverse of the matrix H is:

H21 ¼ A21 þ
�
0 0
0 G21 2A21

22

�

Volume 8 April 2018 | Genomic Study of P. salmonis in Coho Salmon | 1185



where A21 is the inverse numerator relationship matrix for all ani-
mals, A21

22 is the inverse of a pedigree-based relationship matrix for
genotyped animals only, and G21 is the inverse genomic relationship
matrix. SNPs were equally weighted and given an initial value of one
in the ssGBLUP method. In the wssGBLUP method, the marker
variances were used as weights. The marker variances were estimated
from allele frequencies, and from marker effects that were calculated
in the ssGBLUP method (Wang et al. 2014). The DD trait was ana-
lyzed as a linear trait using AIREMLF90 and BLUPF90, whereas, BIN
was analyzed as a threshold trait with THRGIBBS1F90 in the
BLUPF90 family of programs (Misztal et al. 2016). TheMCMCGibbs
scheme for the estimation of the genetic parameters for BIN for
ssGBLUP and wssGBLUP, were estimated identically as described
above. The ssGBLUP and wssGBLUP models included all the geno-
typed animals which passed quality control (n = 580), and all the
phenotyped fish (n = 2,606) from 107 families.

Bayes C
The Bayes C (Habier et al. 2011) analyses were performed using GS3
software. A total of 200,000 iterations were used in the Gibbs sampling,
with a burn-in period of 20,000 cycles. The results were saved every
50 cycles. The number of samples in this analysis totaled 4,000. Con-
vergence and autocorrelation were assessed by visual inspection of trace
plots of the posterior variance components. The adjusted model can be
described, in matrix notation, as follows:

y ¼ Xbþ Tpþ
Xn
i¼1

giaidi þ e

where y is the vector of phenotypic records (DD or BIN), X is an
incidence matrix of fixed effects (sex and tank as factors and IW as
covariate), b is the vector of fixed effects, T is an incidence matrix of
polygenic effects, p is a random vector of polygenic effects of all
individuals in the pedigree, gi is the vector of the genotypes for the ith

SNP for each animal, ai is the random allele substitution effect of the
ith SNP, di is an indicator variable (0, 1) sampled from a binomial
distribution with parameters determined such that 1% of the markers
were included in the model, and e is a vector of residual effects. The
following prior distributions were assumed for the genetic random
effects: Independent and identical mixture distributions for the SNP
effects; each SNP has a point mass at zero having a probabilityp and a
univariate normal distribution with a probability of 1 – p with null
mean and variance s2

a; which in turn has a scaled inverse chi-squared
prior with va ¼ 4 degrees of freedom and scale parameter s2a (or s

2
e )

(Fernando and Garrick 2013). The scale parameter was estimated as a
function of the genetic variance population, based on the mean SNP
allele frequency and number of markers assumed with nonzero effects
(Fernando et al. 2007). Only genotyped animals, which passed quality
control (n = 580) were used.

Genomic prediction accuracy
The different models were compared using a fivefold cross validation
scheme. To reduce stochastic effects of sampling, the cross-validation
analysis was replicated ten times. Briefly, all challenged individuals
(genotyped, phenotyped, or both), were randomly separated into five
validations sets. For each set, predictions were made by masking the
animals’ phenotypes and using the remaining fish as a training set to
estimate the marker effects. Thus, for each cross-validation run, the
dataset was split into a training set (80%) and a validation set (20%).
Accuracy was used to assess the performance of each model and was
estimated as follows:

rEBV ;BV ¼ rEBV ;y
h

where rEBV;y is the correlation between the EBV of a given model
(predicted for the validation set using information from the training
set) and the actual phenotype, while h is the square root of the ped-
igree-based estimate of heritability (Correa et al. 2017; Legarra 2008;
Palaiokostas et al. 2016; Tsai et al. 2016). Finally, accuracies were
calculated for each model and compared to those obtained with the
PBLUP model.

Genome-wide association study
In order to identify associations between genetic markers and P.
salmonis resistance, as DD or BIN, four genome-wide association
methodologies were performed using the BLUPF90 set of programs
(Misztal et al. 2016). GBLUP, ssGBLUP, wssGBLUP and
w3ssGBLUP models were used to analyze the DD and BIN traits,
using a linear and a threshold model respectively (as described
above in model 1). For the GBLUP model, the pedigree-based re-
lationship matrix (A) was replaced by a genomic matrix (G). For the
ssGBLUP, wssGBLUP and w3ssGBLUP GWAS models the H ma-
trix was used as described above. SNPs were weighted equally (and
given a weight of one), for the ssGBLUP model. For the wssGBLUP
and w3ssGBLUP methods, weights were determined based on indi-
vidual marker variances that were estimated using both the marker
effects, which were calculated previously for the ssGBLUP model,
and marker allele frequencies (Wang et al. 2014). For the GBLUP
GWAS model, only genotyped animals, which passed quality con-
trol (n = 580), were analyzed. For the ssGBLUP, wssGBLUP and
w3ssGBLUP GWAS models, all the genotyped animals passing
quality control (n = 580), and all the phenotyped fish (n = 2,606)
were used. Due to practical reasons, the molecular markers an-
chored to the different scaffolds not placed in chromosomes, were
assigned as chromosome 31. The parents of the challenged individ-
uals were not included in the GWAS analysis because they did not
have associated phenotype information, as they were not submitted
to the challenge experiment.

Data availability
Table S1 contains genotypic data (available at the public dryad digital
repository https://doi.org/10.5061/dryad.b273q6p), Table S2 contains
phenotypic data, and Table S3 contains the pedigree information. Table
S4 contains the full list of genes located within the top ten 1-Mbp
windows proximate to each SNP associated with P. salmonis resistance
for DD and BIN identified through wssGBLUP. Table S5 contains in-
formation of the top ten markers which explain the highest percentage
of the genetic variance for each method and trait. Table S6 contains
results from the 10 replicated CV.

n Table 1 Estimated genetic parameters and accuracy of breeding
values (EBV) estimation for resistance against P. salmonis using a
pedigree-based model

Phenotypea s2
a s2

e h2(SE) Accuracy (R)

DD 12.55 77.60 0.14(0.034) 0.271
BIN 0.38 1.00 0.27 (0.043) 0.316
a
The BLUP analysis included the phenotype of all the progeny of 107 families
challenged against Piscirickettsia salmonis (n = 2,606)

1186 | A. Barría et al.

https://doi.org/10.5061/dryad.b273q6p
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS2.txt
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS3.txt
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS4.docx
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS4.docx
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS5.xlsx
http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS6.xlsx


RESULTS

Challenge test
Mortality began on the 10th day after the P. salmonis challenge, with
evident symptoms of SRS and pathological lesions typical of SRS. These
signs include swollen kidney, splenomegaly and yellowish liver tone
and coloration (Rozas and Enríquez 2014). Challenged families showed
considerable phenotypic variation for P. salmonis resistance. Average
mortality of all the 107 families reached 38.53% during the 50-day
challenge. The average cumulative mortality rate among the 17 best
and 16 worst families, selected for genotyping, reached 19% and 63%,
respectively (Figure 1).

ddRAD sequencing
Prior to quality control (QC), per base quality (Phred score) was
evaluated. The average quality score ranged from 36 to 38 among
libraries, indicating high quality of data. Illumina sequencing, including
parents, yielded an average of 156,058,078 (6 16 million) of raw se-
quences per library. After initial QC, which included the removal of low
quality sequences and reads with either missing or ambiguous barco-
des, an average of 31,660,024 of the reads were removed. In parallel
with QC, reads were trimmed to 134 bp. Thus, 79% of the raw reads
were retained for subsequent analysis. To create a set of all possible
alleles in the population, data sets of the parental samples were used to
create a STACKS catalog. This catalog consisted of 106,309 unique
ddRAD loci from which 20,068 markers from 757 individuals were
identified. Quality filtering reduced this to 9,389 putative bi allelic SNPs
(see Table S1) with an average sequencing depth of 38x ranging from
11x to 501x. These markers were identified segregating along the ge-
nome of 580 individuals (see Table S1).

Genome-wide association analysis
Fourgenome-wideassociationmethodologieswereperformedeither for
DD and BIN. These approximations include GBLUP, ssGBLUP,
wssGBLUP and w3ssGBLUP. For both traits, all the models showed
a similar association pattern. In the case of the ssGBLUPmethodology,
the GWAS plots become less noisy as the iterations progress, and the
peaks associated with the traits become more distinct (Figure 2 and
Figure 3). For DD, a marker potentially associated with P. salmonis
resistance was located on chromosome 11 (Figure 2). This marker was
identified in all of the four models and was within the top ten markers
explaining most of the percentage of the genetic variance (Table 2).

The availability of a high quality coho salmon reference genome
(GenBank accession number MPVK00000000.1), made it possible to
identify genes near this marker. Within �55 Kbp of this marker is the
phosphoinositide-3-kinase adaptor protein 1 (pik3ap1) gene; a gene re-
lated with innate host defense through B-cells development (Aiba et al.
2008; Herzog et al. 2009).

In the case of BIN, two molecular markers, explaining most of the
genetic variance, were identified in all of the fourmodels (Table 2). One
of these markers is located on chromosome 29, while the second
marker was identified only at a scaffold level (Scaffold04124) (Figure
3). For BIN, host immune response related genes were not found
proximate to any of the suggestive molecular markers.

However, some genes within a 1-Mb window of these markers have
been suggested to be involved with P. salmonis infection. The 14 KDa
Phosphohistidine phosphatase-like (PHPT1), gelsolin-like (GSN), and
glutamine synthase-like (GS) genes are located on chromosome
29 and near associated markers. Claudin-10 was found on scaf-
fold04124. These genes have been previously identified as being
up-regulated in Salmo salar individuals with low susceptibility to P.
salmonis (Pulgar et al. 2015). Moreover, the retinoic acid receptor RXR-
alpha-A-like (RXRA) gene, located on chromosome 29, has previously
been identified as a molecular biomarker for P. salmonis infections, and
has been found to be down-regulated in macrophages during infection
(Rise et al. 2004).

A full list of genes that are locatedwithin a 1-Mbpwindowproximate
to the suggestive markers associated with P. salmonis resistance for
Okis11, Okis29 and at a scaffold level, identified through wssGBLUP
model, is shown in Table S4. In the case of the marker located on the
scaffold, the surrounding sequence was blasted against the Salmo salar
reference genome (NC_027300.1).

Genetic parameters and predictions
Significantadditivegeneticvariationwasestimated forbothDDandBIN
when using all the data from challenged individuals from the 107 ma-
ternal, full-sib families (Table 1). Using the pedigree-basedmodel with-
out genomic data, estimates of the narrow sense heritability for DD and
BIN were equal to 0.14 (6 0.034) and 0.27 (6 0.043), respectively.

Based on a fivefold cross validation, the accuracy of the PBLUP
model was slightly lower forDD (0.271) than for BIN (0.316) (Figure 4).
When genomic data were included, accuracies for DD and BIN were
higher than those achieved using only phenotypic data. However, there
is considerable variation between models and trait definitions. The
accuracies for the different models ranged from 0.299 (ssGBLUP) to
0.529 (GBLUP) for the DD trait, and from 0.314 (ssGBLUP) to 0.807
(GBLUP) for the BIN. For DD, all the models with genomic data out-
performed the pedigree-based model. The relative increase in accuracy
ranged from 10 (ssGBLUP) to 95% (GBLUP) (Figure 4). In the case of
BIN, the relative increase in accuracy ranged from 20 (wssGBLUP) to
155% (GBLUP). However, one of the genomic models (ssGBLUP) had
a similar accuracy to the PBLUP model, with a relative accuracy 1%
lower than the pedigree-based model.

Interestingly, the accuracies obtainedwithGBLUPwere higher than
ssGBLUP andwssGBLUP for both traits. These accuracy values reached
0.529, 0.299 and 0.417 for DD and 0.807, 0.314 and 0.3797 for BIN. For
both traits, the models with better performance were, GBLUP. Bayes
C . wssGBLUP . ssGBLUP.

DISCUSSION
Significant genetic variation for P. salmonis resistance was detected in
the present study.Moderate heritabilities were estimated using different

Figure 1 Kaplan-Meier curves for Piscirickettsia salmonis experimental
challenge in coho salmon. Average mortality curves for the 107 full-sib
families, and the 17 best and 16 worst families.

Volume 8 April 2018 | Genomic Study of P. salmonis in Coho Salmon | 1187

http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS4.docx


Figure 2 Genomic association analyses for resistance against Pis-
cirickettsia salmonis in a coho salmon population (defined as day
of death) for four different models; GBLUP (A), ssGBLUP (B),
wssGBLUP (C) and w3ssGBLUP (D). The gray area highlights the
SNPs (in green), which are among the top ten markers explaining a
high percentage of the genetic variance in the four models.

1188 | A. Barría et al.



Figure 3 Genomic association analyses for resistance against
Piscirickettsia salmonis in a coho salmon population (for survival
as a binary trait) for four different models; GBLUP (A), ssGBLUP
(B), wssGBLUP (C) and w3ssGBLUP (D). The gray area highlights
SNPs (in green) that were among the top ten markers explaining
a high percentage of the genetic variance in the four models.

Volume 8 April 2018 | Genomic Study of P. salmonis in Coho Salmon | 1189



trait definitions, either for DD or BIN. Estimated heritabilities were
higher for resistance as a binary trait when compared to DD.

Previously, Yáñez et al. (2016a), estimated a heritability of 0.16 for
resistance against P. salmonis, defined as day of death (from the same
coho salmon population) through a bivariate linear model. Our study
estimated a similar heritability value (0.14) in the same population.
Differences in the estimations are likely due to the univariate model
we used instead of a bivariate model. We also estimated heritability for
P. salmonis resistance using a threshold model for the binary trait,
which was higher (0.27) than the value reported for DD. These results
are consistent with previous findings for resistance against P. salmonis
in Atlantic salmon using pedigree information (Yáñez et al. 2013).

When resistance, defined as day of death, was analyzed using a linear
model, heritabilitywas estimated as 0.18 (0.03).When using a threshold
model to analyze resistance as a binary trait, a heritability of 0.24 (0.04)
was calculated (Yáñez et al. 2013, 2014b). The genetic variation and
heritability values for P. salmonis resistance are in accord with different
studies that have also found significant genetic resistance to other bac-
terial diseases in salmonid species (Gjøen et al. 1997; Ødegård et al.
2006; Vallejo et al. 2016).

Bacterial disease resistance has been suggested to be a polygenic trait
in aquaculture species. For example, Palaiokostas et al. (2016) suggested
that resistance against Photobacterium damselae subsp. has a polygenic
architecture in Gilthead Sea Bream (Sparus aurata). Using a 50K SNP
genotyping array, it was possible to elucidate a moderately polygenic
architecture of P. salmonis resistance in Atlantic salmon (Correa et al.
2015b). In the current study, and using 9K SNPs, a similar genetic
architecture for resistance against P. salmonis in coho salmon popula-
tion was found. However, it is likely that the moderate number of

individuals could limit the power to detect QTL of larger effect con-
trolling P. salmonis resistance in coho salmon.

Among the top ten genetic markers for DD and BIN, one and two
markers were identified among all four models, respectively. The
availability of an annotated, coho salmon genome made it possible to
identify the phosphoinositide-3-kinase adaptor protein 1 (pik3ap1) (also
known as the adaptor protein B-cell PI3K adaptor (BCAP)), a gene that
is related with B-cell development (Herzog et al. 2009), proximate to
the genetic marker found to be associated with resistance. The role of
B-cells, through the humoral response, has been widely investigated
and elucidated (reviewed in Janeway et al. 2001). B-cells may also have
phagocytic activity in both rainbow trout (Li et al. 2006) and Atlantic
salmon (Øverland et al. 2010). These cells are capable of ingesting large
particles and bacteria; killing them through phagolysosome fusion (Li
et al. 2006). Studies in vitro, showed that from the total phagocytic
leukocytes isolated from Atlantic salmon head kidney, 37% were
B-cells. Additionally, 77% were B-cells when leukocytes were isolated
from peripheral blood (PB). The phagocytic ability of B-cells was three
times higher than those observed in neutrophils in head kidney, while
in PB no differences were observed (Øverland et al. 2010).

We hypothesize that B-cell development could help in the immune
response against P. salmonis in coho salmon population through its
phagocytic activity. Further studies are needed in order to have a better
understanding of the role of these cells in the resistance against P.
salmonis, and its ability to digest and kill bacteria. For resistance defined
as the BIN trait, no genes related with immune response were identified
near genetic markers associated with this trait. A possible explanation
of this observation, is that the region near the genetic marker acts as a
regulatory sequence. Also, it could be due to the relative small sample

n Table 2 Top ten markers associated with Piscirickettsia salmonis resistance defined as DD and BIN in coho salmon, using wssGBLUP
method

Ranking Name Chrc Pos (BP)d PEVe Genesf

DD
1 24987_127 11a 30525921 5.302 PIK3AP1, TIAL1, PCBD1
2 6135_83 3 37136738 3.225 NOXA1, UBAC1, KLHL20
3 7914_47 4 18281619 3.056 LRP5, NTRK3, KLHL25
4 25096_120 11 33047788 2.744 VA, INPP5A, CSAD
5 34697_43 15 32261231 2.081 CCDC153, KMT2A, LXN
6 52922_94 25 656773 1.802 NHLRC2, NRG3, LRRC4
7 41979_18 18 61818285 1.750 ROBO2, KCNJ1, LCE
8 22393_114 10 24996755 1.451 SLC34A2, SH3RF1, FYB
9 24553_70 11 19605173 1.294 VOS41, CHMP5, FAM49B

10 58185_41 29 22363292 1.211 TSC1, GFI1B, STOM
BIN

1 58185_41 29a 22363292 4.270 PHPT1, GSN, GS
2 66451_65 31ab 12406 2.076 PSMD14, CLDN10, CTSM
3 68326_79 31b 211676 2.073 ROBO1, PIK3CB, KCNJ1
4 45949_127 21 17289066 2.073 FLVCR1, VTA1, HIVEP2
5 36367_15 16 15005555 1.807 SEC24D, KACNIP4, MYOZ2
6 6135_83 3 37136738 1.291 NOXA1, UBAC1, KLHL20
7 37641_86 17 19982735 1.184 NR0B2, HIVEP3, EDN2
8 23665_61 10 55479897 1.170 HDAC5, CADM1, ICAM1
9 47149_112 22 17215562 1.133 PPARA, CDKN1B, KCNQ1

10 18750_95 8 26751149 1.094 GRID2, SMARCAD1, TSPAN3,
a
Markers in common within the top ten along the four models.

b
Salmo salar used as reference specie.

c
Chromosome.

d
Position in coho salmon reference genome.

e
Percentage of Phenotypic variance.

f
Summary of the genes located within 1-Mb window are in supporting information Table S4.

1190 | A. Barría et al.

http://www.g3journal.org/lookup/suppl/doi:10.1534/g3.118.200053/-/DC1/TableS4.docx


size for the GWAS, which ideally should be over 1000 animals (limiting
the resolution of the GWAS).

The underlying genetic basis for this trait may have an important
impact on the accuracy of genomic predictions. Thus, performance
comparisonsof different algorithms is required for eitherGSandGWAS
when a complex trait is studied for the first time within a population
(Vallejo et al. 2017b). The current study is the first to evaluate the
genetic architecture of resistance againstPiscirickettsia salmonis in coho
salmon through different algorithms (to our knowledge). We used a
genomicmodel that assumes that genetic variances are controlled by an
infinite number of markers with minimum effects on the trait (i.e.,
GBLUP). The GBLUP method calculates genomic relationship (G ma-
trix) using all the genotyped markers for this reason. An extension of
this method, which combines genomic (G) and pedigree-based (A)
relationship information into the H relationship matrix (Aguilar
et al. 2010; Legarra et al. 2014) was also evaluated. This method, called
single-step GBLUP (ssGBLUP), mimics a Bayesian selection model in
that it only fits SNPs that explain moderate to large genetic variance of
the trait (Wang et al. 2012). Results of these different genomic-wide
association analyses suggest that resistance against Piscirickettsia sal-
monis has a polygenic architecture, with no major QTL (i.e.,
explaining.= 10% of the genetic variation) segregating in the current
population (Figure 2 and Figure 3).

When resistance was defined as DD, the accuracy for the PBLUP
model was slightly lower than resistance measured as BIN (0.271 and
0.316, respectively). Thesevalues are consistentwith the results obtained
for resistance against the bacterial disease pasteurellosis in S. aurata
defined as day of death, authors reached an accuracy up to 0.30 through
a pedigree-based model (Palaiokostas et al. 2016). However, both ac-
curacy values are slightly lower when compared with values obtained
for resistance against sea lice in Atlantic salmon. In this regard, Correa
et al. (2017) reached an accuracy of 0.41 for resistance against Caligus
rogercresseyi, while Tsai et al. (2016) reached a prediction accuracy
�0.5 for resistance against Lepeophtheirus salmonis using PBLUP.

The results from our genomic predictions are in agreement with
previous studies that showed higher estimated accuracies usingGS than
withPBLUPin the samehalf/full-sib familystructure insalmonbreeding
programs (Nielsen et al. 2009; Lillehammer et al. 2013; Bangera et al.
2017).

TheGSmodel that showedthebestperformance in termsofaccuracy
of predictions for DD was GBLUP. This was followed by Bayes C,

wssGBLUPandssGBLUP.Theadditionofgenomic informationallowed
an improvement up to 95% in accuracy for this trait. The current study
showed a relative accuracy prediction improvement of 54% when
comparing PBLUP to wssGBLUP for this trait. This improvement is
different to that obtained by Vallejo et al. (2016) for resistance against
Bacterial Cold Water Disease (BCWD), defined as day of death, in
rainbow trout (Oncorhynchusmykiss). The authors reported a reduction
in the predictive ability (PA) of 20 and 26% using wssGBLUP, either
with a chip array (40K) or through RAD sequencing (10K) respectively.
The authors attributed this pronounced reduction to stochastic fluctu-
ations due to a small training group of their study (n = 583). However,
when the number of individuals was increased, this model outper-
formed the pedigree-based model, reaching a relative increase in accu-
racy up to 108%, for the same trait definition (Vallejo et al. 2017a).

In the case of BIN, the use of the ssGBLUP model, did not show an
improvement in the accuracy prediction (with a reduction of 1% in
comparison to the accuracyobtained though thepedigree-basedmodel).
However, there was an estimated relative increase in accuracy ranging
from20to155%whencomparingPBLUPto theothers genomicmodels.
In this regard, Bayes C and GBLUP showed an increase in accuracy of
140% and 155%, respectively. These high values are similar to the
improvements seen with resistance against BCWD defined as a binary
trait, which reached a relative increase up to 97% (Vallejo et al. 2017a).

The relative improvements ranged in accuracy from -1 to 155% for
BIN and from 10 to 95% for DD are greater to those obtained for other
diseases resistance studies in Atlantic salmon; even with lower marker
numbers. Using an identical random selection design as in the current
study, sea lice resistance showed a relative improvement in accuracy of
22% relative to PBLUP when using 37K SNPs (Correa et al. 2017). For
reliability, an improvement up to 52% with 220K SNPs was reached
(Ødegård et al. 2014). Tsai et al. (2016) reported that when a non-full
siblings design was used, an improvement in accuracy of 250% and
500% was reached in two different populations compared to PBLUP.
However, when the methodology was changed to a random selection
scheme this improvement only reached up to 27% using 33K SNPs.

In case of Piscirickettsia salmonis resistance, and using the same
cross-validation scheme in our study, the relative reliability was in-
creased by 25% and 30% for resistance, defined as day of death or as
a binary trait, respectively with 50K SNPs (Bangera et al. 2017).

In the current study we evaluated a wide range of differentmodels of
GS for their potential implementation in aquaculture. The performance
of each implemented model varied according to the underlying genetic
architecture of the trait (Meuwissen et al. 2001; Daetwyler et al. 2010).
Thus, it is valuable to perform these comparisons to identify the best
performing method using real data. Similar accuracies among PBLUP
and ssGLBUP models could be due to the predicted GEBV as both
models are overrepresented by polygenic EBV (Bangera et al. 2017).
GBLUP estimated genetic relationships using genotype and pedigree
data rather than just average relationship as PBLUP (Habier et al.
2007). This allows a more accurate genetic relationship matrix and
provides an increase in performance, as seen in the current data, due
to the close family relationship. Moreover, GBLUP had significantly
better performance when resistance was defined as a linear or binary
trait compared to the other evaluated genomic models. We suggest that
this could be an effect of only genotyping families from the opposite
sides of the mortality distribution (i.e., most resistant and most suscep-
tible), and not all the challenged individuals. Bangera et al. (2017),
reported similar accuracies among GBLUP and Bayesian methods be-
tween GBLUP and Bayes C using 10K SNPs in Atlantic salmon. Similar
results that have been seen in dairy cattle for most traits (de Roos et al.
2009; Goddard 2009)

Figure 4 Comparison of predicted accuracies (R) for Piscirickettsia
salmonis resistance in a coho salmon population comparing between
PBLUP and models with genomic data for DD (red bars) and BIN (blue
bars).

Volume 8 April 2018 | Genomic Study of P. salmonis in Coho Salmon | 1191



The greater improvement in accuracy for the binary trait BIN,
compared with the linear trait DD, could be due a better fit of the
thresholdmodel for BIN than the fit of the linearmodel forDD. It could
also be due to the higher estimated heritability.

We hypothesize that the large improvement values seen in the
current study are likely due to an increased level of linkage disequilib-
rium (LD) found within this farmed coho salmon population. Addi-
tional studies are needed to elucidate the minimum number of markers
necessary for GS.

Our results, using ddRAD sequencing, are in agreement with other
genomic studies, whichutilizeGBS techniqueswith aquaculture species.
Using RAD sequencing, some authors have previously performed
genomic-wide association studies in rainbow trout looking for associ-
ations with disease resistance. From a total of 4K identified SNPs,
31markerswere significantlyassociatedwitheitherBCWDorInfectious
Hematopoietic Necrosis Virus (IHNV) resistance as a binary trait
(Campbell et al. 2014). These authors also showed the potential of these
RAD markers to predict an animal’s phenotype. In the case of BCWD
resistance, defined as a linear and binary trait, Palti et al. (2015b)
identified suggestive and significant SNPs in two different families
and candidate genes associated with this trait using �5K markers per
family. Similar numbers of SNPs were used by Liu et al. (2015) to
significantly associate 18 SNPs.

Genomic selection predictions are in accordance with studies aimed
to evaluate GS using others GBS techniques, in both, relative increase in
accuracy and number of discovered SNPs. Dou et al. (2016) predicted
higher accuracies for shell height and shell width using GBLUP and
Bayes methods in Yesso scallop (Patinopecten yessoensis) using 2K
SNPs identified by 2b-RAD. Identical methodology allowed
Palaiokostas et al. (2016) to reach a relative increase in accuracy up
to 53%with 12K SNPs using Bayesianmethods compared to PBLUP. A
study in rainbow trout using RAD sequencing identified 10K SNPs.
Even then, the accuracies were similar with GS models compared to
PBLUP, and the authors predicted that increasing the number of indi-
viduals could lead to a relative increase in accuracy up to 69% (Vallejo
et al. 2016).

The genotyping strategy was aimed at the i) evaluation of genomic
selection methods; and ii) allowing the identification of molecular
markers associated with the trait by means of GWAS. The aim was
maximizing the phenotypic variance within the sample while keeping a
balancedrepresentationoffishper family.Thus, genotyping strategywas
not totally random,but specifically focusedon themostextreme families;
17 resistant and 16 susceptible families. We aimed at genotyping all the
fishbelonging toeach selected family. Thus, each familywas represented
within the sample with an average of 23 (ranging from 11 to 43) fish/
family.

The availability of dense SNP arrays for coho salmon, as it is already
the case for Atlantic salmon (Houston et al. 2014; Yáñez et al. 2016b)
and rainbow trout (Palti et al. 2015a), may increase the accuracy for
predicting genomic breeding values and the power for the determina-
tion of the genetic factors involved in economically-important traits. It
is also expected that in the near future, further functional studies for a
better understanding of P. salmonis resistance and other complex traits
in salmonids will be facilitated by the international initiative on the
Functional Annotation of All Salmonid Genomes, FAASG (Macqueen
et al. 2017).

We have evaluated different GS models, and demonstrated that the
use of genomic prediction is a feasible strategy for the improvement of
breeding value prediction. This information could be used for the
implementation of genomic information in genetic programs for Pis-
cirickettsia salmonis resistance in farmed coho salmon populations.

Conclusions
Moderate significant genetic variation was estimated for resistance
against Piscirickettsia salmonis in coho salmon, using either pedigree
or genomic information. These results highlight the feasibility of in-
cluding this trait into genetic improvement programs. Our study shows
that genomic prediction methods, using ddRAD genotypes (including
9K SNPs), has a substantial advantage in terms of accuracy when
compared to pedigree-basedmodel for eitherDDor BIN. The improve-
mentwas up to 95 and 155% respectively in the current population. The
association analyses were used to identify a gene related with B-cell
development, which could also be involved in resistance against P.
salmonis. To our knowledge, this is the first study aimed at dissecting
the genetic architecture of P. salmonis resistance in a coho salmon
population.

ACKNOWLEDGMENTS
AB and KC acknowledge the National Commission of Scientific and
Technologic Research (CONICYT) for the funding through the
National PhD funding program. AB acknowledges the Government
of Canada for the funding through the Canada-Chile Leadership
Exchange Scholarship. This project was funded by the U-Inicia grant,
from the Vicerrectoria de Investigación y Desarrollo, Universidad de
Chile. This work has been conceived on the frame of the grant FON-
DEF NEWTON-PICARTE (IT14I10100), funded by CONICYT
(Government of Chile) and the Newton Fund - The British Council
(Government of United Kingdom). This work has been partially sup-
ported by Núcleo Milenio INVASAL from Iniciativa Científica Mile-
nio (Ministerio de Economía, Fomento y Turismo, Gobierno de
Chile). This research was carried out in conjunction with EPIC4
(Enhanced Production in Coho: Culture, Community, Catch), a proj-
ect supported by the government of Canada through Genome Canada,
Genome British Columbia, and Genome Quebec.

Authors’ contributions: AB performed DNA extraction, library con-
struction, ddRAD analysis, GWAS analysis, and wrote the initial ver-
sion of the manuscript. KrC performed library construction and
contributed to the data analysis. AB, GY and KC performed genomic
prediction analysis. AJ performed DNA extraction. JPL contributed
with study design. WD contributed with analysis and discussion. JMY
conceived and designed the study, supervised work of AB and con-
tributed to the analysis, discussion and writing. All authors have
reviewed and approved the manuscript.
Animal ethics approval: Challenge and sampling procedures were
approved by the Comité de Bioética Animal from the Facultad de
Ciencias Veterinarias y Pecuarias, Universidad de Chile (Certificate
N08-2015).

LITERATURE CITED
Aguilar, I., I. Misztal, D. L. Johnson, A. Legarra, S. Tsuruta et al., 2010 Hot

topic: A unified approach to utilize phenotypic, full pedigree, and geno-

mic information for genetic evaluation of Holstein final score. J. Dairy Sci.

93(2): 743–752. https://doi.org/10.3168/jds.2009-2730
Aiba, Y., M. Kameyama, T. Yamazaki, T. F. Tedder, and T. Kurosaki,

2008 Regulation of B-cell development by BCAP and CD19 through

their binding to phosphoinositide 3-kinase. Blood 111(3): 1497–1503.

https://doi.org/10.1182/blood-2007-08-109769
Baird, N. A., P. D. Etter, T. S. Atwood, M. C. Currey, A. L. Shiver et al.,

2008 Rapid SNP discovery and genetic mapping using sequenced RAD

markers. PLoS One 3(10): e3376. https://doi.org/10.1371/journal.

pone.0003376

1192 | A. Barría et al.

https://doi.org/10.3168/jds.2009-2730
https://doi.org/10.1182/blood-2007-08-109769
https://doi.org/10.1371/journal.pone.0003376
https://doi.org/10.1371/journal.pone.0003376


Bangera, R., K. Correa, J. P. Lhorente, R. Figueroa, and J. M. Yáñez,
2017 Genomic predictions can accelerate selection for resistance against
Piscirickettsia salmonis in Atlantic salmon (Salmo salar). BMC Genomics
18(1): 121. https://doi.org/10.1186/s12864-017-3487-y

Brieuc, M. S. O., C. D. Waters, and J. E. Seeb, and K. A. Naish, 2014 A
dense linkage map for Chinook salmon (Oncorhynchus tshawytscha) re-
veals variable chromosomal divergence after an ancestral whole genome
duplication event. G3 Genes Genomes Genet. 4: 447–460. https://doi.org/
10.1534/g3.113.009316

Campbell, N. R., S. E. LaPatra, K. Overturf, R. Towner, and S. R. Narum,
2014 Association mapping of disease resistance traits in rainbow trout
using restriction site associated DNA sequencing. G3 Genes Genomes
Genet. 4: 2473–81. https://doi.org/10.1534/g3.114.014621

Catchen, J. M., A. Amores, P. Hohenlohe, W. Cresko, and J. H. Postlethwait,
2011 Stacks: building and genotyping loci de novo from short-read
sequences. G3 Genes Genomes Genet. 1: 171–182. https://doi.org/
10.1534/g3.111.000240

Catchen, J., P. A. Hohenlohe, S. Bassham, A. Amores, and W. A. Cresko,
2013 Stacks: An analysis tool set for population genomics. Mol. Ecol. 22
(11): 3124–3140. https://doi.org/10.1111/mec.12354

Correa, K., R. Bangera, R. Figueroa, J. P. Lhorente, and J. M. Yáñez, 2017 The
use of genomic information increases the accuracy of breeding value pre-
dictions for sea louse (Caligus rogercresseyi) resistance in Atlantic salmon
(Salmo salar). Genet. Sel. Evol. 49(1): 15. https://doi.org/10.1186/s12711-
017-0291-8

Correa, K., M. Filp, D. Cisterna, M. E. Cabrejos, C. Gallardo-Escárate et al.,
2015a Effect of triploidy in the expression of immune-related genes in
coho salmon Oncorhynchus kisutch (Walbaum) infected with Piscirick-
ettsia salmonis. Aquacult. Res. 46: 59–63. https://doi.org/10.1111/
are.12584

Correa, K., J. Lhorente, M. Lopez, L. Bassini, S. Naswa et al.,
2015b Genome-wide association analysis reveals loci associated with
resistance against Piscirickettsia salmonis in two Atlantic salmon (Salmo
salar L.) chromosomes. BMC Genomics 16(1): 854. https://doi.org/
10.1186/s12864-015-2038-7

Cvitanich, J., O. Garate, and C. E. Smith, 1991 The isolation of a rickettsia‐
like organism causing disease and mortality in Chilean salmonids and its
confirmation by Koch’s postulate. J. Fish Dis. 14(2): 121–145. https://doi.
org/10.1111/j.1365-2761.1991.tb00584.x

Daetwyler, H. D., R. Pong-Wong, B. Villanueva, and J. A. Woolliams,
2010 The impact of genetic architecture on genome-wide evaluation
methods. Genetics 185(3): 1021–1031. https://doi.org/10.1534/genet-
ics.110.116855

Davey, J. W., T. Cezard, P. Fuentes-Utrilla, C. Eland, K. Gharbi et al.,
2013 Special features of RAD Sequencing data: Implications for geno-
typing. Mol. Ecol. 22(11): 3151–3164. https://doi.org/10.1111/mec.12084

Dou, J., X. Li, Q. Fu, W. Jiao, Y. Li et al., 2016 Evaluation of the 2b-RAD
method for genomic selection in scallop breeding. Sci. Rep. 6(1): 19244.
https://doi.org/10.1038/srep19244

Dufflocq, P., J. P. Lhorente, R. Bangera, R. Neira, S. Newman et al.,
2016 Correlated response of flesh color to selection for harvest weight
in coho salmon (Oncorhynchus kisutch). Aquaculture 472: 6–11.

Elshire, R. J., J. C. Glaubitz, Q. Sun, J. A. Poland, K. Kawamoto et al.,
2011 A robust, simple genotyping-by-sequencing (GBS) approach for
high diversity species. PLoS One 6(5): e19379. https://doi.org/10.1371/
journal.pone.0019379

Falconer, D.S., and Mackay T.F.C., 1996. Introduction to Quantitative Ge-
netics. fourth ed. Longman Group Limited, Harlow, Essex, U.K.

FAO, 2016. Fisheries and Aquaculture Department [Online]. (Rome. Updated 31
January 2016. http://www.fao.org/fishery/statistics/global-aquaculture-production/
query/en

Fernando, R.L., and D. Garrick, 2013 Bayesian Methods Applied to GWAS.
In: Gondro C., van der Werf J., Hayes B. (eds) Genome-Wide Association
Studies and Genomic Prediction. Methods in Molecular Biology (Meth-
ods and Protocols), vol 1019. Humana Press, Totowa, NJ.

Fernando, R. L., D. Habier, C. Stricker, J. C. M. Dekkers, and L. R. Totir,
2007 Genomic selection. Acta Agric. Scand. Sect. Anim. Sci. 57:
192–195. https://doi.org/10.1080/09064700801959395

Gjøen, H. M., T. Refstie, O. Ulla, and B. Gjerde, 1997 Genetic correlations
between survival of Atlantic salmon in challenge and field tests. Aquaculture
158(3-4): 277–288. https://doi.org/10.1016/S0044-8486(97)00203-2

Goddard, M., 2009 Genomic selection: Prediction of accuracy and max-
imisation of long term response. Genetica 136(2): 245–257. https://doi.
org/10.1007/s10709-008-9308-0

Gonen, S., N. R. Lowe, T. Cezard, K. Gharbi, S. C. Bishop et al., 2014 Linkage
maps of the Atlantic salmon (Salmo salar) genome derived from RAD
sequencing. BMC Genomics 15(1): 166. https://doi.org/10.1186/1471-2164-
15-166

Habier, D., R. L. Fernando, and J. C. M. Dekkers, 2007 The impact of
genetic relationship information on genome-assisted breeding values.
Genetics 177: 2389–2397. https://doi.org/10.1534/genetics.107.081190

Habier, D., R. L. Fernando, K. Kizilkaya, D. J. Garrick, T. Meuwissen et al.,
2011 Extension of the bayesian alphabet for genomic selection. BMC
Bioinformatics 12(1): 186. https://doi.org/10.1186/1471-2105-12-186

Hayes, B., and M. Goddard, 2010 Genome-wide association and genomic
selection in animal breeding. Genome 53(11): 876–883. https://doi.org/
10.1139/G10-076

Herzog, S., M. Reth, and H. Jumaa, 2009 Regulation of B-cell proliferation
and differentiation by pre-B-cell receptor signalling. Nat. Rev. Immunol.
9(3): 195–205. https://doi.org/10.1038/nri2491

Houston, R. D., J. W. Davey, S. C. Bishop, N. R. Lowe, J. C. Mota-Velasco
et al., 2012 Characterisation of QTL-linked and genome-wide restric-
tion site-associated DNA (RAD) markers in farmed Atlantic salmon.
BMC Genomics 13(1): 244. https://doi.org/10.1186/1471-2164-13-244

Houston, R. D., J. B. Taggart, T. Cézard, M. Bekaert, N. R. Lowe et al.,
2014 Development and validation of a high density SNP genotyping
array for Atlantic salmon (Salmo salar). BMC Genomics 15(1): 90.
https://doi.org/10.1186/1471-2164-15-90

Janeway, C., P. Travers, M. Walport, and M. Shlomchik, 2001 B-cell acti-
vation by armed helper T cells., pp. 343–360 in Immunobiology 5, The
Immune System in Health and Disease, Ed. 5. Churchill Livingstone, New
York.

Legarra, A., O. F. Christensen, I. Aguilar, and I. Misztal, 2014 Single Step, a
general approach for genomic selection. Livest. Sci. 166: 54–65. https://
doi.org/10.1016/j.livsci.2014.04.029

Legarra, A., C. Robert-Granié, E. Manfredi, and J.-M. Elsen, 2008 Performance
of genomic selection in mice. Genetics 180: 611–618. https://doi:10.1534/ge-
netics.108.088575

Li, J., D. R. Barreda, Y.-A. Zhang, H. Boshra, A. E. Gelman et al., 2006 B
lymphocytes from early vertebrates have potent phagocytic and micro-
bicidal abilities. Nat. Immunol. 7(10): 1116–1124. https://doi.org/
10.1038/ni1389

Li, H., and R. Durbin, 2009 Fast and accurate short read alignment with
Burrows-Wheeler transform. Bioinformatics 25(14): 1754–1760. https://
doi.org/10.1093/bioinformatics/btp324

Lillehammer, M., T. H. E. Meuwissen, and A. K. Sonesson, 2013 A low-
marker density implementation of genomic selection in aquaculture using
within-family genomic breeding values. Genet. Sel. Evol. 45(1): 39.
https://doi.org/10.1186/1297-9686-45-39

Liu, S., R. L. Vallejo, Y. Palti, G. Gao, D. P. Marancik et al., 2015 Identification
of single nucleotide polymorphism markers associated with bacterial cold
water disease resistance and spleen size in rainbow trout. Front. Genet. 6:
1–10. https://doi.org/10.3389/fgene.2015.00298

Lynch, M., and B. Walsh, 1998 Genetics and analysis of quantiative traits,
Sinauer Associates, Sunderland.

Macqueen, D. J., C. R. Primmer, R. D. Houston, B. F. Nowak, L. Bernatchez
et al., 2017 Functional Annotation of All Salmonid Genomes (FAASG):
an international initiative supporting future salmonid research, conser-
vation and aquaculture. BMC Genomics 18: 484. https://doi.org/10.1186/
s12864-017-3862-8

Volume 8 April 2018 | Genomic Study of P. salmonis in Coho Salmon | 1193

https://doi.org/10.1186/s12864-017-3487-y
https://doi.org/10.1534/g3.113.009316
https://doi.org/10.1534/g3.113.009316
https://doi.org/10.1534/g3.114.014621
https://doi.org/10.1534/g3.111.000240
https://doi.org/10.1534/g3.111.000240
https://doi.org/10.1111/mec.12354
https://doi.org/10.1186/s12711-017-0291-8
https://doi.org/10.1186/s12711-017-0291-8
https://doi.org/10.1111/are.12584
https://doi.org/10.1111/are.12584
https://doi.org/10.1186/s12864-015-2038-7
https://doi.org/10.1186/s12864-015-2038-7
https://doi.org/10.1111/j.1365-2761.1991.tb00584.x
https://doi.org/10.1111/j.1365-2761.1991.tb00584.x
https://doi.org/10.1534/genetics.110.116855
https://doi.org/10.1534/genetics.110.116855
https://doi.org/10.1111/mec.12084
https://doi.org/10.1038/srep19244
https://doi.org/10.1371/journal.pone.0019379
https://doi.org/10.1371/journal.pone.0019379
http://www.fao.org/fishery/statistics/global-aquaculture-production/query/en
http://www.fao.org/fishery/statistics/global-aquaculture-production/query/en
https://doi.org/10.1080/09064700801959395
https://doi.org/10.1016/S0044-8486(97)00203-2
https://doi.org/10.1007/s10709-008-9308-0
https://doi.org/10.1007/s10709-008-9308-0
https://doi.org/10.1186/1471-2164-15-166
https://doi.org/10.1186/1471-2164-15-166
https://doi.org/10.1534/genetics.107.081190
https://doi.org/10.1186/1471-2105-12-186
https://doi.org/10.1139/G10-076
https://doi.org/10.1139/G10-076
https://doi.org/10.1038/nri2491
https://doi.org/10.1186/1471-2164-13-244
https://doi.org/10.1186/1471-2164-15-90
https://doi.org/10.1016/j.livsci.2014.04.029
https://doi.org/10.1016/j.livsci.2014.04.029
https://doi:10.1534/genetics.108.088575
https://doi:10.1534/genetics.108.088575
https://doi.org/10.1038/ni1389
https://doi.org/10.1038/ni1389
https://doi.org/10.1093/bioinformatics/btp324
https://doi.org/10.1093/bioinformatics/btp324
https://doi.org/10.1186/1297-9686-45-39
https://doi.org/10.3389/fgene.2015.00298
https://doi.org/10.1186/s12864-017-3862-8
https://doi.org/10.1186/s12864-017-3862-8


Meuwissen, T. H., B. J. Hayes, and M. E. Goddard, 2001 Prediction of total
genetic value using genome-wide dense marker maps. Genetics 157:
1819–1829.

Misztal, I., S. Tsuruta, D. Lourenco, Y. Masuda, I. Aguilar et al., 2016 Manual
for BLUPF90 Family of Programs. University of Georgia, Athens, GA.

Moen, T., J. Torgersen, N. Santi, W. S. Davidson, M. Baranski et al., 2015 Epithelial
cadherin determines resistance to infectious pancreatic necrosis virus in Atlantic
salmon. Genetics 200: 1313–1326. https://doi: 10.1534/genetics.115.175406

Nielsen, H.M., A. K. Sonesson, H. Yazdi, and T. H. E. Meuwissen, 2009 Comparison
of accuracy of genome-wide and BLUP breeding value estimates in sib based
aquaculture breeding schemes. Aquaculture 289(3-4): 259–264. https://doi.org/
10.1016/j.aquaculture.2009.01.027

Ødegård, J., M. Baranski, B. Gjerde, and T. Gjedrem, 2011 Methodology for
genetic evaluation of disease resistance in aquaculture species: Challenges
and future prospects. Aquacult. Res. 42: 103–114. https://doi.org/10.1111/
j.1365-2109.2010.02669.x

Ødegård, J., T. Moen, N. Santi, S. A. Korsvoll, S. Kjøglum et al.,
2014 Genomic prediction in an admixed population of Atlantic salmon
(Salmo salar). Front. Genet. 5: 1–8. https://doi: 10.3389/fgene.2014.00402

Ødegård, J., I. Olesen, B. Gjerde, and G. Klemetsdal, 2006 Evaluation of
statistical models for genetic analysis of challenge test data on furunculosis
resistance in Atlantic salmon (Salmo salar): Prediction of field survival.
Aquaculture 259(1-4): 116–123. https://doi.org/10.1016/j.aquacul-
ture.2006.05.034

Øverland, H. S., E. F. Pettersen, A. Rønneseth, and H. I. Wergeland,
2010 Phagocytosis by B-cells and neutrophils in Atlantic salmon (Salmo
salar L.) and Atlantic cod (Gadus morhua L.). Fish Shellfish Immunol. 28
(1): 193–204. https://doi.org/10.1016/j.fsi.2009.10.021

Palaiokostas, C., S. Ferarreso, R. Franch, R. D. Houston, and L. Bargelloni,
2016 Genomic prediction of resistance to pasteurellosis in gilthead sea
bream (Sparus aurata) using 2b-RAD sequencing. G3 Genes Genomes
Genet. Available at: https://doi:10.1534/g3.116.035220

Palti, Y., G. Gao, S. Liu, M. P. Kent, S. Lien et al., 2015a The development
and characterization of a 57K single nucleotide polymorphism array for
rainbow trout. Mol. Ecol. Resour. 15(3): 662–672. https://doi.org/10.1111/
1755-0998.12337

Palti, Y., R. L. Vallejo, G. Gao, S. Liu, A. G. Hernandez et al.,
2015b Detection and validation of QTL affecting bacterial cold water
disease resistance in rainbow trout using restriction-site associated dna
sequencing. PLoS One 10(9): e0138435. https://doi.org/10.1371/journal.
pone.0138435

Peterson, B. K., J. N. Weber, E. H. Kay, H. S. Fisher, and H. E. Hoekstra,
2012 Double digest RADseq: An inexpensive method for de novo SNP
discovery and genotyping in model and non-model species. PLoS One 7:
e37135. https://doi:10.1371/journal.pone.0037135

Pulgar, R., C. Hödar, D. Travisany, A. Zuñiga, C. Domínguez et al.,
2015 Transcriptional response of Atlantic salmon families to Piscir-
ickettsia salmonis infection highlights the relevance of the iron-depriva-
tion defence system. BMC Genomics 16(1): 495. https://doi.org/10.1186/
s12864-015-1716-9

Rise, M. L., S. R. M. Jones, G. D. Brown, K. R. von Schalburg, W. S. Davidson
et al., 2004 Microarray analyses identify molecular biomarkers of At-
lantic salmon macrophage and hematopoietic kidney response to Piscir-
ickettsia salmonis infection. Physiol. Genomics 20(1): 21–35. https://doi.
org/10.1152/physiolgenomics.00036.2004

Robledo, D., C. Palaiokostas, L. Bargelloni, P. Martínez, and R. Houston,
2017 Applications of genotyping by sequencing in aquaculture breeding
and genetics. Rev. Aquac. 1–13. https://doi.org/10.1111/raq.12193

de Roos, A. P. W., B. J. Hayes, and M. E. Goddard, 2009 Reliability of
genomic predictions across multiple populations. Genetics 183(4):
1545–1553. https://doi.org/10.1534/genetics.109.104935

Rozas, M., and R. Enríquez, 2014 Piscirickettsiosis and Piscirickettsia sal-
monis in fish: a review. J. Fish Dis. 37(3): 163–188. https://doi.org/
10.1111/jfd.12211

Sernapesca, 2016 Informe Sanitario de Salmonicultura en Centros Marinos 2016.
([Online]. Valparaiso, Chile. Updated 14 August 2017. http://www.sernapesca.
cl/index.php?option=com_remository&Itemid=246&func=fileinfo&id=27560).

Stear, M. J., S. C. Bishop, B. A. Mallard, and H. Raadsma, 2001 The sus-
tainability, feasibility and desirability of breeding livestock for disease
resistance. Res. Vet. Sci. 71(1): 1–7. https://doi.org/10.1053/
rvsc.2001.0496

Tsai, H.-Y., A. Hamilton, A. E. Tinch, D. R. Guy, J. E. Bron et al., 2016 Genomic
prediction of host resistance to sea lice in farmed Atlantic salmon populations.
Genet. Sel. Evol. 48(1): 47. https://doi.org/10.1186/s12711-016-0226-9

Vallejo, R. L., T. D. Leeds, B. O. Fragomeni, G. Gao, A. G. Hernandez et al.,
2016 Evaluation of genome-enabled selection for bacterial cold water
disease resistance using progeny performance data in rainbow trout: in-
sights on genotyping methods and genomic prediction models. Front.
Genet. 7: 1–13. https://doi.org/10.3389/fgene.2016.00096

Vallejo, R. L., T. D. Leeds, G. Gao, J. E. Parsons, K. E. Martin et al.,
2017a Genomic selection models double the accuracy of predicted
breeding values for bacterial cold water disease resistance compared to a
traditional pedigree-based model in rainbow trout aquaculture. Genet.
Sel. Evol. 49(1): 17. https://doi.org/10.1186/s12711-017-0293-6

Vallejo, R. L., S. Liu, G. Gao, B. O. Fragomeni, A. G. Hernandez et al.,
2017b Similar genetic architecture with shared and unique quantitative
trait loci for bacterial cold water disease resistance in two rainbow trout
breeding populations. Front. Genet. 8: 156. https://doi.org/10.3389/
fgene.2017.00156

VanRaden, P. M., 2008 Efficient methods to compute genomic predictions.
J. Dairy Sci. 91(11): 4414–4423. https://doi.org/10.3168/jds.2007-0980

Wang, H., I. Misztal, I. Aguilar, A. Legarra, R. L. Fernando et al.,
2014 Genome-wide association mapping including phenotypes from
relatives without genotypes in a single-step (ssGWAS) for 6-week body
weight in broiler chickens. Front. Genet. 5: 1–10. https://doi.org/10.3389/
fgene.2014.00134

Wang, H., I. Misztal, I. Aguilar, A. Legarra, and W. M. Muir,
2012 Genome-wide association mapping including phenotypes from
relatives without genotypes. Genet. Res. 94(02): 73–83. https://doi.org/
10.1017/S0016672312000274

Yáñez, J. M., R. Bangera, J. P. Lhorente, A. Barria, M. Oyarzun et al.,
2016a Negative genetic correlation between resistance against Piscir-
ickettsia salmonis and harvest weight in coho salmon (Oncorhynchus
kisutch). Aquaculture 459: 8–13. https://doi.org/10.1016/j.aquacul-
ture.2016.03.020

Yáñez, J. M., S. Naswa, M. E. Lopez, L. Bassini, K. Correa et al., 2016b Genomewide
single nucleotide polymorphism discovery in Atlantic salmon (Salmo salar):
validation in wild and farmed American and European populations. Mol. Ecol.
Resour. 16(4): 1002–1011. https://doi.org/10.1111/1755-0998.12503

Yáñez, J. M., R. Bangera, J. P. Lhorente, M. Oyarzún, and R. Neira,
2013 Quantitative genetic variation of resistance against Piscirickettsia
salmonis in Atlantic salmon (Salmo salar). Aquaculture 414–415:
155–159. https://doi.org/10.1016/j.aquaculture.2013.08.009

Yáñez, J. M., L. N. Bassini, M. Filp, J. P. Lhorente, R. W. Ponzoni et al.,
2014a Inbreeding and effective population size in a coho salmon
(Oncorhynchus kisutch) breeding nucleus in Chile. Aquaculture 420–421:
S15–S19. https://doi.org/10.1016/j.aquaculture.2013.05.028

Yáñez, J. M., R. D. Houston, and S. Newman, 2014b Genetics and genomics
of disease resistance in salmonid species. Front. Genet. 5: 1–13. https://
doi:10.3389/fgene.2014.00415

Yáñez, J. M., J. P. Lhorente, L. N. Bassini, M. Oyarzún, R. Neira et al.,
2014c Genetic co-variation between resistance against both Caligus
rogercresseyi and Piscirickettsia salmonis, and body weight in Atlantic
salmon (Salmo salar). Aquaculture 433: 295–298. https://doi.org/10.1016/
j.aquaculture.2014.06.026

Yoshida, G.M., R. Bangera, R. Carvalheiro, K. Correa, R. Figueroa, J. P.
Lhorente, and J. M. Yáñez, 2017 Genomic prediction accuracy for
resistance against Piscirickettsia salmonis in farmed rainbow trout. G3
Genes Genomes Genet. 8:719–726. https://doi.org/10.1534/g3.117.300499

Communicating editor: R. Houston

1194 | A. Barría et al.

https://doi:%2010.1534/genetics.115.175406
https://doi.org/10.1016/j.aquaculture.2009.01.027
https://doi.org/10.1016/j.aquaculture.2009.01.027
https://doi.org/10.1111/j.1365-2109.2010.02669.x
https://doi.org/10.1111/j.1365-2109.2010.02669.x
https://doi:%2010.3389/fgene.2014.00402
https://doi.org/10.1016/j.aquaculture.2006.05.034
https://doi.org/10.1016/j.aquaculture.2006.05.034
https://doi.org/10.1016/j.fsi.2009.10.021
https://doi:10.1534/g3.116.035220
https://doi.org/10.1111/1755-0998.12337
https://doi.org/10.1111/1755-0998.12337
https://doi.org/10.1371/journal.pone.0138435
https://doi.org/10.1371/journal.pone.0138435
https://doi:10.1371/journal.pone.0037135
https://doi.org/10.1186/s12864-015-1716-9
https://doi.org/10.1186/s12864-015-1716-9
https://doi.org/10.1152/physiolgenomics.00036.2004
https://doi.org/10.1152/physiolgenomics.00036.2004
https://doi.org/10.1111/raq.12193
https://doi.org/10.1534/genetics.109.104935
https://doi.org/10.1111/jfd.12211
https://doi.org/10.1111/jfd.12211
http://www.sernapesca.cl/index.php?option=com_remository&Itemid=246&func=fileinfo&id=27560
http://www.sernapesca.cl/index.php?option=com_remository&Itemid=246&func=fileinfo&id=27560
https://doi.org/10.1053/rvsc.2001.0496
https://doi.org/10.1053/rvsc.2001.0496
https://doi.org/10.1186/s12711-016-0226-9
https://doi.org/10.3389/fgene.2016.00096
https://doi.org/10.1186/s12711-017-0293-6
https://doi.org/10.3389/fgene.2017.00156
https://doi.org/10.3389/fgene.2017.00156
https://doi.org/10.3168/jds.2007-0980
https://doi.org/10.3389/fgene.2014.00134
https://doi.org/10.3389/fgene.2014.00134
https://doi.org/10.1017/S0016672312000274
https://doi.org/10.1017/S0016672312000274
https://doi.org/10.1016/j.aquaculture.2016.03.020
https://doi.org/10.1016/j.aquaculture.2016.03.020
https://doi.org/10.1111/1755-0998.12503
https://doi.org/10.1016/j.aquaculture.2013.08.009
https://doi.org/10.1016/j.aquaculture.2013.05.028
https://doi:10.3389/fgene.2014.00415
https://doi:10.3389/fgene.2014.00415
https://doi.org/10.1016/j.aquaculture.2014.06.026
https://doi.org/10.1016/j.aquaculture.2014.06.026
https://doi.org/10.1534/g3.117.300499