Jukuri, open repository of the Natural Resources Institute Finland (Luke) 
   
 
   
All material supplied via Jukuri is protected by copyright and other intellectual property rights. Duplication 
or sale, in electronic or print form, of any part of the repository collections is prohibited. Making electronic 
or print copies of the material is permitted only for your own personal use or for educational purposes.  For 
other purposes, this article may be used in accordance with the publisher’s terms. There may be 
differences between this version and the publisher’s version. You are advised to cite the publisher’s 
version. 
 
This is an electronic reprint of the original article.  
This reprint may differ from the original in pagination and typographic detail. 
 
Author(s): Federico C.F. Calboli , Heikki Koskinen , Antti Nousianen, Clemence Fraslin , Ross D. 
Houston & Antti Kause 
Title: Conserved QTL and chromosomal inversion affect resistance to columnaris disease in 2 
rainbow trout (Oncorhyncus mykiss) populations 
Year:  2022 
Version: Publisher’s version 
Copyright:    The author(s) 2022   
Rights:  CC BY 4.0  
Rights url: https://creativecommons.org/licenses/by/4.0/  
 
Please cite the original version: 
Federico C.F. Calboli , Heikki Koskinen , Antti Nousianen, Clemence Fraslin , Ross D. Houston & Antti Kause. 
(2022). Conserved QTL and chromosomal inversion affect resistance to columnaris disease in 2 rainbow 
trout (Oncorhyncus mykiss) populations. G3 Genes, Genomes, Genetics, jkac137. 
https://doi.org/10.1093/g3journal/jkac137 
Conserved QTL and chromosomal inversion affect
resistance to columnaris disease in 2 rainbow trout
(Oncorhyncus mykiss) populations
Federico C.F. Calboli ,1 Heikki Koskinen ,2 Antti Nousianen,2 Clemence Fraslin ,3 Ross D. Houston ,3,4 and
Antti Kause 1,*
1Natural Resources Institute Finland (LUKE), FI-31600 Jokioinen, Finland
2Natural Resources Institute Finland (LUKE), FI-70210 Kuopio, Finland
3The Roslin Institute and Royal (Dick) School of Veterinary Studies, University of Edinburgh, Easter Bush EH25 9RG, UK
4Present address: Benchmark Genetics, 1 Pioneer Building, Edinburgh Technopole, Penicuik EH26 0GB, UK.
*Corresponding author: Natural Resources Institute Finland (LUKE), Myllintie 1, FI-31600 Jokioinen, Finland. Email: antti.kause@luke.fi
Abstract
We present a comparative genetic analysis of the quantitative trait loci underlying resistance to warm water columnaris disease in 2 farmed
rainbow trout (Oncorhynchus mykiss) populations. We provide evidence for the conservation of a major quantitative trait loci on Omy03,
and the putative role played by a chromosomal rearrangement on Omy05. A total of 3,962 individuals from the 2 populations experienced
a natural Flavobacterium columnare outbreak. Data for 25,823 genome-wide SNPs were generated for both cases (fatalities) and controls
(survivors). FST and pairwise additive genetic relationships suggest that, despite being currently kept as separate broodstocks, the 2 popula-
tions are closely related. Association analyses identified a major quantitative trait loci on chromosome Omy03 and a second smaller quanti-
tative trait loci on Omy05. Quantitative trait loci on Omy03 consistently explained 3–11% of genetic variation in both populations, whereas
quantitative trait loci on Omy05 showed different degree of association across populations and sexes. The quantitative trait loci on Omy05
was found within a naturally occurring, 54.84 cM long inversion which is easy to tag due to a strong linkage disequilibrium between the
375 tagging SNPs. The ancestral haplotype on Omy05 was associated with decreased mortality. Genetic correlation between mortality in
the 2 populations was estimated at 0.64, implying that the genetic basis of resistance is partly similar in the 2 populations. Our quantitative
trait loci validation identifies markers that can be potentially used to complement breeding value evaluations to increase resistance against
columnaris disease, and help to mitigate effects of climate change on aquaculture.
Keywords: rainbow trout; columnaris disease; QTL; GWAS; inversion
Introduction
Flavobacterium columnare is a well-known infectious disease agent
in freshwater fish farming (Newton et al. 1997; Waskiewicz and
Irzykowska 2014). Flavobacterium columnare is the etiological cause
of columnaris disease, which predominantly manifests itself as
ulcerations of the gills and of the skin. Mortality due to columna-
ris disease is especially high in younger fish, where the disease
produces an acute response, and depending on the virulence of
the strain, untreated fry die between 12 and 48 h after infection
(Declercq, Haesebrouck, et al. 2013, Evenhuis et al. 2014). The dis-
ease affects multiple freshwater fish species at warm water tem-
peratures, typically above 20C (Declercq, Haesebrouck, et al.
2013; Declercq, Boyen, et al. 2013), though it has been observed af-
fecting salmonid species even at colder temperatures of 12–14C
(Starliper 2011). Climate change is increasing temperatures
worldwide (Zachos et al. 2001), affecting inland water tempera-
tures (van Vliet et al. 2013), with direct repercussions on water
temperatures in fry ponds and tanks, potentially leading to more
frequent columnaris outbreaks (Karvonen et al. 2010). Currently a
live attenuated vaccine is being developed, but early testing on
channel catfish (Ictalurus punctatus) suggests negligible effects on
mortality (Malecki et al. 2021). The vaccine is not commercially
available for salmonids. Disease management therefore relies on
antibiotics (Laanto et al. 2015), the routine use and need of which
is less than ideal, due to environmental impact concerns, and
due to the risk of selection of antibiotic resistance (Declercq,
Haesebrouck, et al. 2013; Laanto et al. 2015; Zhang et al. 2017).
There is therefore a strong interest in selecting for improved
resistance in breeding programs of aquaculture species, an
approach that may adapt farmed fish also to climate change
(Sae-Lim et al. 2017).
The ability to use high density, genome-wide marker panels
for the aquaculture species of interest (either by using commer-
cially available SNP chips, or genotyping by sequencing), allows
assessment of the association between genomic regions and phe-
notype using a genome-wide association study (GWAS). Markers
showing significant association with the phenotype are poten-
tially linked to a quantitative trait locus (QTL) that is part of the
genetic architecture of the trait. Causal variants of QTLs are not
normally directly identified, and can be challenging to map
Received: March 08, 2022. Accepted: May 26, 2022
VC The Author(s) 2022. Published by Oxford University Press on behalf of Genetics Society of America.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which
permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
G3, 2022, jkac137
https://doi.org/10.1093/g3journal/jkac137
Advance Access Publication Date: 6 June 2022
Investigation
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
without sufficiently dense marker sets. Additionally, larger struc-
tural polymorphisms such as chromosome inversions may create
long chromosome regions in which markers are very tightly
linked with limited recombination. Such regions may harbor
QTLs for multiple traits and the limited recombination makes
these regions “supergenes” in which long haplotypes with pheno-
typic effects on multiple traits exist (Villoutreix et al. 2021). It has
been hypothesized that major QTLs may be more prominent in
farmed fish compared to terrestrial livestock, perhaps due to the
diversity of life-history strategies and the early phase of domesti-
cation in aquatic species (Houston et al. 2020).
There are several approaches to use QTL information in selec-
tive breeding. In genomic selection, genomic estimated breeding
values (GEBV) are calculated using genome-wide information
provided by tens of thousands of markers at once (Meuwissen
et al. 2001; Goddard and Hayes 2009). In the genomic evaluation,
phenotype–marker associations can be further emphasized by
weighting the SNPs or genomic regions by their allele substitution
effects on a phenotype (Wang et al. 2014) or by using Bayesian
methods that model non-normal distribution of the SNP effects
(Ka¨rkka¨inen and Sillanpa¨a¨ 2012). Alternatively, in marker-assis-
ted selection (MAS), a few specific markers linked to major QTLs
are used in the evaluation of individuals (Lande and Thompson
1990; You et al. 2020), and MAS can be easily combined with
breeding value evaluations that are based on relationships
among individuals (i.e. based on pedigree or realized genomic re-
lationship matrix [GRM]).
Approaches using QTL information for genomic evaluation
and selection are likely to be especially effective for those pheno-
types that are oligogenic in origin, or where QTLs are within a
chromosomal rearrangement and where recombination is sup-
pressed, conditions that allow the use of few highly informative
tagging markers. Moreover, specific QTL information is more use-
ful in smaller data sets, and when marker density is low and does
not to cover all linkage disequilibrium blocks across the genome
(Misztal et al. 2020). To effectively use QTL information in a breed-
ing program, it would be useful that the link between the markers
and the phenotypic effect would be persistent. In farmed fish,
only a few QTLs have been validated by replication in multiple
studies so far (e.g. Houston et al. 2010; Barson et al. 2015; Gonen
et al. 2015; Liu et al. 2018; Boison et al. 2019; Hillestad and
Moghadam 2019; Aslam et al. 2020; Hillestad et al. 2020), and the
understanding of the role of chromosomal inversions on deter-
mining trait variation is limited.
An important aquaculture species susceptible to columnaris
disease is the rainbow trout (Oncorhynchus mykiss). Rainbow trout
is a prominent aquaculture species worldwide due to its fast
growth and adaptability to different environments (Boudry et al.
2021). This species has been the subject of extensive genetic
analysis, which has generated a reference genome (Berthelot
et al. 2014; Gao et al. 2018), a commercial SNP chip (Palti et al.
2015), and detailed information about naturally occurring geno-
mic rearrangements, such as inversions (Pearse et al. 2019;
Weinstein et al. 2019; Liu et al. 2021), that are present in this spe-
cies. It is known that rainbow trout exhibits genetic-based differ-
ences in mortality caused by the bacterial cold-water disease
(Leeds et al. 2010; Evenhuis et al. 2015; Liu et al. 2018), a disease
caused by Flavobacterium psychrophilum, which is very closely re-
lated to F. columnare, and it has also been observed that rainbow
trout exhibits a positive correlation between mortality due to the
2 diseases (Evenhuis et al. 2015; Silva, Evenhuis, Vallejo, Gao, et al.
2019; Silva, Evenhuis, Vallejo, Tsuruta, et al. 2019). This observa-
tion suggests that this species could harbor genetic variation and
genetic variants that decrease mortality to columnaris disease in-
fection. If that were the case, it would be valuable to understand
the genetic basis of decreased mortality, so that this information
can be incorporated in breeding programs (Kause et al. 2005).
Here we present the results of a genome-wide association
analysis for mortality after an outbreak of F. columnare infection
in rainbow trout from 2 Finnish broodstocks. The use of fish from
2 populations allowed us to specifically address the questions of
(1) whether we would observe conservation across populations of
any effects of chromosomal regions and inversions on mortality,
(2) what the heritability of mortality to infection is in the 2 popu-
lations, and (3) whether we would observe a positive genetic
correlation between mortality in the 2 populations. We used
genomic information to calculate the realized marker-based
genomic relationships between all samples, both within and
between populations. This allowed us to obtain a direct measure
of the genetic correlation for mortality between the 2 popula-
tions. We quantify the degree of genetic differentiation between
the populations and discuss our results in the light of replicability
of association analysis results of QTLs, and other genomic struc-
tural variants.
Materials andmethods
Populations and family structure
The fish used in this study come from 2 broodstocks, one from
Savon Taimen Oy, a Finnish rainbow trout breeder and fingerling
producer based in Rautalampi, and the second from the national
breeding program maintained by the Natural Resources Institute
Finland (LUKE), based in Enonkoski research farm. The 2 brood-
stocks are located at separate farms but their offspring were
reared on the same farm. Matings were performed and families
established in May 2019 for both broodstocks. For the Savon
Taimen samples, 100 dams and 25 sires (mating 4 dam to each
sire) were used to establish 100 families in a farm located in East-
Finland on the 2nd and 3rd of May. The Savon Taimen sires were
sex reversed females, thus all the offspring were females. For the
LUKE samples, 33 dams and 48 sires were used to establish 105
families (fertilizing the eggs from each dam with multiple sires)
at the Enonkoski farm on the 15th of May. LUKE sires were natu-
ral males, and thus the offspring contained both sexes. A fin clip
was collected from all sires and dams for genotyping.
For both populations separately, 0.5 dl of eggs was sampled
from each mating and all families within a population were
thereafter pooled. The eggs were sent to the farm of Hanka-
Taimen Oy (Hankasalmi, Finland), and the eggs of the 2 popula-
tions were separately incubated and fingerlings reared in bulk.
Each broodstock produced about 35,000 (Savon Taimen) and
31,000 (LUKE) fry, which were randomly distributed into 3 tanks
for each population, 6 tanks in total. Density was hence an aver-
age of 100 fry from each family in each tank. All tanks were adja-
cent to one another, were subject to the same light conditions,
and shared the same source of water from a nearby river.
The study started on June 24, 2019 when fry were moved into
fingerling tanks. Across the whole study, all tanks were moni-
tored daily for mortality and signs of any disease, and dead indi-
vidual were collected once or twice a day. On July 4 and July 5,
2019, the first fish with signs of columnaris disease were ob-
served. Only the fry showing clear signs of columnaris disease
were randomly sampled among the fish that died in the first
5 days of the outbreak, and tail tissue was extracted and used for
DNA analysis. Some of the samples were sent for veterinary ex-
amination, and once the presence of columnaris disease was
2 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
confirmed, treatment was initiated to stop the outbreak on July
26. For both populations, 510 dead or dying fish were sampled for
tissue from each tank. Fish rearing continued in the tanks after
the end of the outbreak. On July 19, the fish were moved to new
larger tanks, all fish from 1 tank to a 1 new tank.
In September–October 2019, all the remaining fish were
counted and 510 surviving fish were sampled randomly from
each tank.
The population-wide mortality to columnaris disease was 30%
for the Savon Taimen, 10,650 dead fish over 34,951 fish in total,
and 39% for the LUKE population, 12,476 dead fish over 31,696
fish in total. The overall mortality across the 2 populations was
35%.
Genotyping
Of the samples collected as mentioned above, a total 1,450 off-
spring samples and 104 parents (1 parent was collected but acci-
dentally not sent for genotyping, and the genotyping company
genotyped 20 parents twice, omitting 20 unique parents) were
genotyped for the Savon Taimen stock, and 3,055 offspring sam-
ples and 81 parents were genotyped for the LUKE stock.
For these samples DNA was extracted and genotyped by
IdentiGen Ltd. using the commercially available 57K SNP Axiom
Trout Genotyping Array (Palti et al. 2015). After genotyping, we
used the Axiom Analysis Suite version 5.1.1 to call genotypes for
all samples, and we exported the resulting genotypes as a PLINK
format .ped and .map files for further analysis.
For all samples, the Axiom Array nominally provided data for
57,501 SNPs. After genotype calling, 36,227 SNPs were classified as
being polymorphic at high resolution by the Axiom Suite and were
retained. These SNPs were further filtered by removing all SNPs
that mapped on unplaced scaffolds and SNPs that mapped to
more than 1 position on the genome assembly, that had a missing
rate of 5% or more, that had a P-value for Hardy–-Weinberg equi-
librium smaller than 0.000001, and that had a Minor Allele
Frequency of 0.05 or lesser. After these quality control steps our
data comprised of 25,853 high-quality filtered SNPs for down-
stream analyses. We also removed all individual samples missing
10% or more of their genotypes, leaving 1,363 Savon Taimen sam-
ples (1,284 offspring and 79 parents), 2,772 LUKE samples (2,691
offspring and 81 parents), for a total of 4,135 samples after quality
control. All markers were mapped against the Omyk_1.0 genome
(Gao et al. 2018, accession number GCA_002163495.1), with each
SNP position recorded on the Axiom database.
Trait definition and sex phenotypes
Mortality was coded as a binary trait. Dead and dying offspring
fish were coded as 1 and alive coded as 0. To determine the sex of
LUKE offspring fish, we selected 7 sex-associated SNPs (Palti et al.
2015), and we used the LUKE parents, whose sex was known, to
assess the sensitivity (true positive rate) and specificity (true neg-
ative rate) (Fawcett 2005) of these markers in identifying males.
Once the specificity and sensitivity were assessed, we used the
methods described in Toli et al. (2016), which uses a Bayesian ap-
proach, to determine whether a fish is male. Starting with a 0.5
prior, each fish was tested 1 SNP at a time to calculate the poste-
rior probability that the fish is male, using the results of each test
as the new prior to the following test. After 7 consecutive tests,
we could assign a posterior probability of being male to each fish,
with the posteriors being in all cases greater than 0.995 (for
males), or smaller than 0.005 (for females). The final figures for
the LUKE population were 1,341 males and 1,337 females. For the
Savon Taimen fish, all offspring were female, so no specific sex
determination was needed.
Genomic relatedness of 2 populations
Using the SNP data, we quantified the genomic differentiation of
the 2 populations using an FST analysis, and the pairwise kinship
between samples using the additive genetic relationship coeffi-
cient, excluding all parents in both analyses. FST quantifies the
partitioning of genetic diversity in the SNPs within and across
populations, a value of zero implying panmixia and a value of 1
that all genetic variation is explained by the population structure.
We used PLINK to calculate both measures (Chang et al. 2015).
We used PLINK to provide the Weir and Cockeram FST estimate
(Weir and Cockerham 1984) by using the –fst option, and we cal-
culated a full set of Identity-By-Descent measures including the
pairwise additive genetic relationship (Weir et al. 2006) using the
–genome option. The Weir pairwise additive genetic relationship
is bound between 0 and 1 and can be directly be compared to
pedigree-based relatedness analyses.
Genome-wide association analyses
We used genome-wide complex trait analysis (GCTA) for
genome-wide association analyses (Yang et al. 2011). Analyses
were carried out on the offspring only because the parents of
both populations did not experience the F. columnare outbreak
and thus had no survival phenotype. We performed 2 kinds of
genome-wide association analyses with mortality using GCTA.
The first is a mixed linear model analysis in which GCTA calcu-
lates the GRM based on all SNPs, and then 1 SNP at a time and
the GRM are simultaneously fitted in the model, together with
covariates (-MLMA option). The random GRM factor controls for
population structure and genomic relationships (Yang et al. 2011).
The second is a mixed linear model with a leave 1 chromosome
out (MLMA-LOCO) option. With this option, each candidate SNP
is still fitted one at a time together with the covariates as de-
scribed above, but rather than using the same GRM for each SNP,
a new partial GRM is calculated every time by removing all the
other SNPs in the same chromosome as the candidate SNP, and
then calculating a new GRM on the fly (Yang et al. 2014). Yang
et al. (2014) have shown this approach to increase the power of
the analysis in those cases where the SNP under exam is not itself
associated with the phenotype but linked to a SNP that is causal,
because without this correction the causal SNP affects the analy-
sis twice, once through linkage with the SNP in analysis, and
once through the GRM, leading to a loss of power.
GWAS was performed within populations, and in all combina-
tions of populations and sexes. The model used in the GWAS
was:
y ¼ aþ bxþ Gþ tank þ sex þ e ðmodel 1Þ
with y the phenotype, a the mean value, b the fixed regression
slope (beta) corresponding to the allele substitution effect of the
SNP tested, x the SNP genotype (with levels 0/1/2 corresponding
to the number of minor alleles in the genotype of an individual),
G is the random genetic effect captured by the GRM, tank the
fixed tank effect (with 3 levels within a stock, and 6 levels across
the stocks), sex is the fixed effect for the sex of the fish (with 2
levels), and e is the random error. The GRM was calculated using
either the MLMA or LOCO approach. When analyzing data on one
sex only, the gender effect was excluded. Because the families
were split across tanks, we did not investigate maternal effects.
F. C. F. Calboli et al. | 3
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
The proportion of genetic variance explained by a SNP (%SNP
r2SNP) was calculated as (Lynch and Walsh 1997):
%r2SNP ¼ pð1 pÞb2=r2grm  100
in which b is as determined before, p is the frequency of the com-
mon allele, r2grm is the total genetic variance explained by the ge-
nomic relationships, estimated in GCTA with the null model:
y ¼ aþ Gþ tank þ sex þ e ðmodel 2Þ
To correct for multiple statistical tests, we employed a
genome-wide significance threshold of 1.9  106 (the Bonferroni
corrected threshold of 0.05/25,823 SNPs), and a chromosome-
wide significance threshold of 5.6  105 (the chromosome aver-
aged Bonferroni correction 0.05/(25,823 SNPs/29 chromosomes)).
These thresholds were applied to all analyses because in all anal-
yses the number of SNPs were the same, and because the
Bonferroni correction is dependent on the number of SNPs.
Tagging SNPs for Omy5 inversion
We tested for the presence of a previously identified large inver-
sion on chromosome Omy05 (Pearse et al. 2019), using tagging
SNPs identified in the wild rainbow trout populations in the
United States (Gao et al. 2021). Of the 475 SNPs listed by Gao et al.
(2021) as fully tagging the inversion, 375 were common with our
data, allowing us to identify unequivocally the genotype of each
fish for the presence of the Ancestral (A) or Rearranged (R) haplo-
types based on the identification of Pearse et al. (2019).
Estimation of heritability and genetic correlation
To estimate the overall genetic variance in mortality within the
populations and the genetic correlation between the populations,
BLUPF90 was used (Misztal et al. 2002). Phenotypes of the 2 popu-
lations were coded to be 2 different traits. Thus, it was possible to
calculate the genetic correlation between these 2 traits, and the
h2 for mortality in each population.
For the estimation of the genetic parameters, we used 2 mod-
els. The first is the linear mixed model that estimates the herita-
bility on the observed scale, as implemented in the AIREMLF90
module (which uses a maximum likelihood approach). Using the
formula of Dempster and Lerner (1950), the observed, 0 and 1,
scale estimate of heritability was corrected for the prevalence of
mortality in each population to obtain the heritability at the un-
derlying liability scale.
The second approach uses a threshold model that estimates
h2 directly on an underlying normally distributed liability scale,
as implemented by the THRGIBBS3F90 module (which uses Gibbs
sampling). Posterior distribution of the parameters was obtained
by Gibbs sampling using the THRGIBBS3F90 module run with a
burn in of 250,000 iterations, a sampling run of 25,000 iterations
that saved every 50th iteration.
A bivariate model was used in which the Model 2 was used for
both population-specific traits, and the residual covariance be-
tween the traits was assumed to be zero. Heritability was com-
puted as:
h2i ¼ r2gi=ðr2gi þ r2e Þ
with h2i the heritability in population i, r
2
gi the genetic variance for
mortality for population i, and r2e the residual variance, and the
genomic correlation was
corij ¼ covij=
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðr2gir2gjÞ
q
with corij the genetic correlation between mortality in popula-
tions i and j, and covij the genetic covariance between popula-
tions i and j.
Results
FST and additive genetic relationship within and
between populations
The average FST between Savon Taimen and LUKE offspring was
0.03, indicating that on average the population difference
explained 3% of the variation in allele frequencies across all
markers. Based on this figure, we consider that the 2 populations
were closely related.
Pairwise additive genetic relationship between the offspring of
Savon Taimen population ranged between 0 and 0.70, with a me-
dian of 0.04 and a mean of 0.06. For the offspring of the LUKE
population, the pairwise additive genetic relationship ranged be-
tween 0 and 0.76, with a median of 0.02 and a mean of 0.05. For
the Savon Taimen fish, 62% of the pairwise comparisons were
greater than 0, with a median value of 0.075. For the LUKE popu-
lation, 56% of the pairwise comparisons were greater than 0, with
a median value of 0.063. These figures indicate that, in both pop-
ulations, the median genetic relationship between samples that
are related broadly matches the value of half first cousins
(0.0625).
When comparing the 2 populations, the additive genetic rela-
tionship between Savon Taimen offspring and LUKE offspring
ranged between 0 and 0.34, with median of 0.00 and mean of
0.01. Comparing the additive genetic relationship between the 2
populations, we also noted that 20% of the pairwise comparisons
were greater than 0, with a median value of 0.05, slightly below
the value of half first cousins, based on the chain counting rule
used in pedigrees (Lynch and Walsh 1997). Very importantly, ev-
ery single fish in both populations has at least one nonzero addi-
tive genetic relationship value with a fish in the other population.
MLMA results for Omy03 QTL
The MLMA analysis identified a significant (above genome-wide
Bonferroni correction) peak in association of mortality with SNPs
on chromosome Omy03 when all samples were pooled together
(Table 1a), similarly to what was observed by Fraslin et al. (2022)
for the LUKE samples alone. When the 2 populations and the 2
sexes are analyzed separately, we observe the same peak on
chromosome Omy03 in both the LUKE female and LUKE male
samples at the level of chromosome-wide significance. The
Savon Taimen samples show a peak in the same location, that
does not reach significance, either genome, or chromosome-wide
(Table 1a).
For the main SNP on Omy3 (Table 1a), the beta value for the
SNP shows a positive value, indicating that the minor allele is as-
sociated with increased mortality, and in all cases the beta is on
average 0.10 (range 0.09–0.11) irrespective of the P-value. The
most significant SNP on Omy03 explains on average 10% (range
9–13%) of the additive genetic variance observed in mortality.
MLMA-LOCO results for Omy03 QTL
We observed that LUKE females (Figs. 1a and 2a), LUKE males
(Figs. 1b and 2b), and Savon Taimen fish (Figs. 1c and 2c) all show
the presence of QTL on chromosome Omy03, with P-values
reaching at least chromosome-wide level. When all samples are
4 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
combined (Fig. 1, d and e) and when Savon Taimen fish and LUKE
males are combined (Fig. 2, d and e), the QTL reaches genome-
wide significance. Thus, this QTL is present in both sexes and
populations (Table 1b).
In all analyses, the betas of the most significant SNPs indicate
an average 9% increase in mortality for each rare allele substitu-
tion (beta values range 0.08–0.11) (Table 1b). The proportion of ge-
netic variance explained by the top SNPs is 0.09–0.13.
MLMA-LOCO identifies a large Omy05 QTL
We observed that LUKE females (Figs. 1a and 3a) have no QTL on
chromosome Omy05, LUKE males show a long flat QTL (55 Mb in
length) that does not reach significance (Figs. 1b and 3b), and
Savon Taimen fish show a strong long QTL containing SNPs with
P-values reaching chromosome-wide and genome-wide signifi-
cance (Figs. 1c and 3c). In the analysis combining all samples, the
long QTL contains multiple SNPs with P-values reaching
chromosome-wide significance (Figs. 1d and 3d). In the dataset
combing Savon Taimen and LUKE males, the long QTL reaches
chromosome-wide significance and contains multiple SNPs with
P-values reaching genome-wide significance (Figs. 1e and 3e). The
results show that the QTL is present in Savon Taimen fish, and
hint that it may exist also in LUKE males. The QTL corresponds
to the genetic region affected by the Omy05 inversion (in Fig. 3,
the SNPs that tag the inversion are shown in red). In both popula-
tions, and in the LUKE males and LUKE females separately, the
linkage between SNPs is very high (median value for all fish com-
bined r2 ¼ 0.75; in both LUKE males and LUKE females, and all
LUKE samples the median value r2 ¼ 0.77; in Savon Taimen
median value r2 ¼ 0.76).
In those cases where SNPs showed P-values above
chromosome-wide or genome-wide level, the betas indicate that
the effect on mortality is on average 10% (beta values range
0.07–0.15) (Table 1c), but in LUKE females close to zero
(0.00–0.02). The proportion of genetic variance explained by the
top SNPs is 0.09–0.16 for Savon Taimen fish, 0.03–0.06 for LUKE
males, and 0.000–0.004 for LUKE females.
We identified that in our data the R haplotype has a frequency
of 16%, and the A haplotype has a frequency of 84% (Table 2).
When looking at the raw data, even a simple chi-square test
shows an association with mortality for the R haplotype
(v-squared¼ 19.49, df¼ 2, P-value¼ 5.85e-05). The A haplotype
shows consistent association with reduced mortality in the
Savon Taimen data and in the LUKE males data (Table 2).
MLMA-LOCO results for Omy06, Omy09, and
Omy10 QTLs
In other chromosomes excluding Omy03 and Omy05, we do not
observe any other QTL reaching genome-wide significance
(Fig. 1). SNPs that are found to be chromosome-wide significant
in more than one analysis are SNPs on chromosomes Omy06 and
Omy10, with chromosome-wide results found in LUKE males,
and in the combined Savon Taimen and LUKE males data. In
LUKE males, 2 SNPs on Omy06 are at chromosome-wide signifi-
cance with positive betas (0.10 and 0.13). In the combined data of
Savon Taimen and LUKE males, these 2 SNPs have a positive but
lower beta (range 0.07–0.09), and 1 extra SNP with negative beta
(0.07). The distances between these 3 SNPs are 3 and 4 Mb, sug-
gesting they would not be part of the same QTL, irrespective of
the directions of their effects: the linkage between these SNPs
was low, ranging between r2 ¼ 0.002 and r2¼ 0.17, indicating they
are not part of one linkage block. The genetic variance explainedT
ab
le
1.
T
h
e
m
os
t
si
gn
ifi
ca
n
t
SN
Ps
w
it
h
as
so
ci
at
io
n
to
m
or
ta
li
ty
in
ge
n
om
e-
w
id
e
as
so
ci
at
io
n
an
al
ys
es
.
A
ll
el
es
LU
K
E
fe
m
al
es
LU
K
E
m
al
es
S
av
on
T
ai
m
en
A
ll
fi
sh
to
ge
th
er
S
av
on
T
ai
m
en
an
d
LU
K
E
m
al
es
S
N
P
ch
r
b
p
M
aj
or
M
in
or
b
et
a
6
S
E
V
ar
ia
n
ce
ex
p
la
in
ed
P
b
et
a
6
S
E
V
ar
ia
n
ce
ex
p
la
in
ed
P
b
et
a
6
S
E
V
ar
ia
n
ce
ex
p
la
in
ed
P
b
et
a
6
S
E
V
ar
ia
n
ce
ex
p
la
in
ed
P
b
et
a
6
S
E
V
ar
ia
n
ce
ex
p
la
in
ed
P
(a
)O
m
y0
3,
M
LM
A
A
X
-8
99
74
38
5
3
64
,0
17
,0
36
T
C
0.
09
6
0.
02
0.
10
1.
98
E-
04
0.
1
6
0.
02
0.
13
9.
94
E-
06
0.
10
6
0.
03
0.
09
1.
95
E-
04
0.
10
6
0.
02
0.
10
2.
56
E-
10
0.
11
6
0.
02
0.
11
1.
16
E-
08
(b
)O
m
y0
3,
M
LM
A
-L
O
C
O
A
X
-8
99
74
38
5
3
64
0,
17
,0
36
T
C
0.
09
6
0.
02
0.
10
4.
70
E-
05
0.
10
6
0.
02
0.
12
2.
25
E-
06
0.
11
6
0.
03
0.
09
4.
58
E-
05
0.
11
6
0.
01
0.
11
7.
19
E-
14
0.
11
6
0.
02
0.
11
1.
22
E-
10
A
X
-8
99
53
54
4
3
64
,0
32
,4
77
A
G
0.
08
6
0.
02
0.
08
2.
07
E-
04
0.
09
6
0.
02
0.
09
5.
87
E-
05
0.
11
6
0.
04
0.
06
1.
22
E-
03
0.
10
6
0.
02
0.
08
1.
14
E-
10
0.
10
6
0.
02
0.
08
1.
50
E-
07
A
X
-8
99
76
03
2
3
65
,7
33
,7
33
G
A
0.
08
6
0.
02
0.
07
1.
82
E-
03
0.
08
6
0.
02
0.
08
1.
23
E-
04
0.
08
6
0.
03
0.
04
3.
48
E-
03
0.
08
6
0.
02
0.
07
1.
47
E-
08
0.
09
6
0.
02
0.
06
7.
72
E-
07
(c
)O
m
y0
5,
M
LM
A
-L
O
C
O
A
X
-8
99
41
92
0
5
33
,3
28
,6
65
T
C
0.
01
6
0.
03
0.
00
1
6.
17
E-
01
0.
04
6
0.
02
0.
03
1.
78
E-
02
0.
15
6
0.
03
0.
16
1.
03
E-
06
0.
08
6
0.
02
0.
04
9.
25
E-
06
0.
11
6
0.
02
0.
07
5.
76
E-
07
A
X
-8
99
42
56
2
5
50
,3
26
,0
41
T
G
0.
02
6
0.
03
0.
00
4
4.
48
E-
01
0.
05
6
0.
02
0.
06
2.
10
E-
03
0.
10
6
0.
03
0.
09
1.
29
E-
04
0.
07
6
0.
02
0.
04
5.
23
E-
06
0.
09
6
0.
02
0.
07
8.
52
E-
07
A
X
-8
99
57
53
9
5
54
,7
95
,3
22
A
G
0.
00
6
0.
03
0.
00
0
8.
85
E-
01
0.
04
6
0.
02
0.
05
2.
80
E-
03
0.
13
6
0.
03
0.
14
5.
69
E-
06
0.
07
6
0.
02
0.
04
1.
37
E-
05
0.
10
6
0.
02
0.
08
2.
67
E-
07
A
X
-8
99
31
73
4
5
55
,5
55
,4
33
A
G
0.
01
6
0.
03
0.
00
2
6.
31
E-
01
0.
05
6
0.
02
0.
03
1.
08
E-
02
0.
14
6
0.
03
0.
13
1.
28
E-
05
0.
07
6
0.
02
0.
03
2.
63
E-
05
0.
10
6
0.
02
0.
06
1.
78
E-
06
T
h
e
re
su
lt
s
ar
e
gr
ou
p
ed
by
se
x
an
d
p
op
u
la
ti
on
,a
n
d
b
y
2
co
m
bi
n
ed
d
at
as
et
s
(a
ll
sa
m
p
le
s
to
ge
th
er
,a
n
d
Sa
vo
n
T
ai
m
en
an
d
LU
K
E
m
al
es
to
ge
th
er
).
(a
)M
LM
A
an
al
ys
is
,c
h
ro
m
os
om
e
O
m
y0
3.
(b
)M
LM
A
-L
O
C
O
an
al
ys
is
,
ch
ro
m
os
om
e
O
m
y0
3.
(c
)M
LM
A
-L
O
C
O
an
al
ys
is
,c
h
ro
m
os
om
e
O
m
y0
5.
F. C. F. Calboli et al. | 5
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
ranges from 5% to 9% for the top SNPs on Omy06 and between
4% and 11% for the top SNPs on Omy10.
One SNP on chromosome Omy09 was also found to be
chromosome-wide significant in all the fish combined, and the
combined Savon Taimen and LUKE males, datasets. In both
cases, the beta is negative (all fish combined beta¼0.08,
Savon Taimen and LUKE males beta¼0.09), explaining 4%
and 5% of the genetic variance in the 2 datasets.
Heritability and genetic correlation
We observed that heritability estimates are moderate and of the
same magnitude in both populations (Table 3). The linear model esti-
mates of h2 on the liability scale are, for LUKE h2 ¼ 0.32, SD¼ 0.05,
and for Savon Taimen h2 ¼ 0.35, SD¼ 0.08. The heritability estimates
from the threshold model were similar to the ones from the linear
model. From threshold model, the posterior estimates of h2 are, for
LUKE h2 ¼ 0.32, SD¼ 0.04, and for Savon Taimen h2 ¼ 0.36, SD¼ 0.06.
Fig. 1. Genome-wide association analysis across all chromosomes from the MLMA-LOCO (leave-one-chromosome out) approach. For all panels, in blue
the chromosome significance P-value, in red the genome-wide significance line, both based on Bonferroni correction. a) Manhattan plot of LUKE female
fish. b) Manhattan plot of LUKE male fish. c) Manhattan plot of Savon Taimen female fish. d) Manhattan plot of all fish combined. e) Manhattan plot of
Savon Taimen females and LUKE males combined.
6 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
Genetic correlation between mortality in the 2 populations is
positive and moderate. The linear model estimates a genetic cor-
relation of 0.62 (SD¼ 0.25), and the threshold model estimates a
genetic correlation of 0.64 (SD¼ 0.23) (Table 3).
Discussion
Our results provide an assessment of the conservation of the ge-
netic architecture, and the role of an extensive chromosomal
inversion, underlying a disease resistance phenotype in 2 popula-
tions whose kinship has been estimated at the genome-wide
level.
Population differentiation
Selective breeding for resistance would strongly benefit from
knowing whether the same loci are associated with survival in all
(or most) breeding stocks, and what is the range of possible
responses that can be attributed to a QTL. The quantification of
Fig. 2. Manhattan plots for chromosome Omy3 from MLMA-LOCO analysis. For all panels, in blue the chromosome significance P-value, in red the
genome-wide significance line, both based on Bonferroni correction. a) Manhattan plot of LUKE female fish. b) Manhattan plot of LUKE male fish. c)
Manhattan plot of Savon Taimen fish. d) Manhattan plot of all fish combined. e) Manhattan plot of Savon Taimen and LUKE males combined.
F. C. F. Calboli et al. | 7
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
the genetic distance between populations allows to create test-
able expectations on the permanence and effect of QTLs based
on the genetic distance between stocks. To give a frame of refer-
ence to the FST figures we obtained, we directly compared them to
the FST results from the analysis of 7 French broodstocks geno-
typed using the very same SNP chip we used (D’Ambrosio et al.
2019). Based on the FST comparisons reported by D’Ambrosio, the
FST value we observe is of the same order of magnitude with the
FST observed between closely related French INRA broodstock,
where the populations are known to be separated by about 5–10
generations. This result is also consistent with the observation
that every fish in 1 population has at least 1 nonzero additive ge-
netic relationship with the other population, with a value com-
patible with the generational distance we assess by FST analysis.
Fig. 3. Manhattan plots for chromosome Omy5 from MLMA-LOCO analysis. For all panels, in blue the chromosome significance P-value, in red the
genome-wide significance line, both based on Bonferroni correction. The red dots correspond to the P-values of SNPs marked as tagging SNPs for the
large inversion complex on Omy5. a) Manhattan plot of LUKE female fish. b) Manhattan plot of LUKE male fish. c) Manhattan plot of Savon Taimen fish.
d) Manhattan plot of all fish combined. e) Manhattan plot of Savon Taimen and LUKE males combined.
8 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
Our assessment of the genetic distance between 2 stocks used
in the present QTL mapping analysis informs our understanding
of the degree to which the stocks share common genetic determi-
nation.
Effect of the inversion on Omy05 on mortality in
the 2 populations
The most novel result we observed is the presence of a QTL asso-
ciated with mortality on the large double inversion present on
chromosome Omy05. The presence of an inversion on Omy05 has
been documented in the wild (Pearse et al. 2019) and in commer-
cial broodstocks of rainbow trout (Gao et al. 2021). Our analysis
revealed the presence of at least 1 QTL on the area on Omy05 af-
fected by the inversion. This inversion is composed by 2 adjacent
inversions, covering approximately 60% of the chromosome, of
21.99 and 32.83 Mb in length, respectively, the first inversion of
which is pericentric. Despite the presence of 2 inversions, in prac-
tice in our data they behave as one single, large inversion, and
the same behavior is also reported by Pearse et al. (2019).
The inversion is naturally occurring in wild populations in
North America (Pearse et al. 2019), and it is estimated that it has
been maintained in wild populations for about 1.5 million years.
Different analytical approaches also suggest that the inversion is
also in farmed populations in the United States (Vallejo et al.
2018; Gao et al. 2021) and France (D’Ambrosio et al. 2019). Both
haplotypes reported in the literature, A (ancestral) and R (rear-
ranged) are found in our data, matching the tagging SNPs
reported by Gao et al. (2021), and we could assign every sample to
the 3 possible AA/AR/RR genotypes.
In our data, the most common haplotype is the A haplotype,
and it is associated with a reduced mortality to columnaris dis-
ease especially in the Savon Taimen population. Pearse et al.
(2019) highlight how this inversion acts a “super gene” which
contains multiple loci. The inversion contains a large number of
genes analogous to genes whose effect has been characterized in
other model organisms, and thus have a putative function in
O. mykiss, including genes associated with circadian rhythm, pho-
tosensory, adiposity, and age at maturity. Inversions may play an
important role in evolution by reducing recombination between
favorable combinations of alleles. In natural populations, the A
haplotype is associated with the steelhead anadromous morph in
females, and R haplotype is associated with a sedentary freshwa-
ter morph in males. The A haplotype is also associated with
slightly slower growth. The A haplotype, that in our data
increases resistance against warm water columnaris disease,
exhibits a clinal distribution in wild populations of O. mykiss
ranging from Southern Alaska to Southern California, with the
highest prevalence in southern rivers (Pearse et al. 2019).
Capture-recapture data of wild O. mykiss in the Big Creek river in
California show that the frequency of individuals carrying the A
haplotype increases with warming water temperature (Pearse
et al. 2019). The inversion was tagged in a wild population by 2
SNP markers by Pearse et al. (2019), which were further expanded
in a farmed population to 475 SNPs by Gao et al. (2021) in a
farmed stock, providing clear evidence of the conservation of this
chromosomal rearrangement between wild O. mykiss and aqua-
culture stock. In our populations, these tagging SNPs match the
known location and the size of the inversion, and showed associ-
ation with mortality, indicating the presence of at least 1 QTL in
the inversion, the exact location which will have to be further
elucidated.
Our analysis of the Omy05 inversion QTL showed, however,
that the magnitude of the allele substitution effect depended on
sex and population. Without dividing the data in its constituent
populations, and without dividing the LUKE population in males
and females, the impact of Omy5 QTL would have been substan-
tially less clear. We detected a clear effect on the Savon Taimen
females (9–16% of the genetic variance explained, beta range
0.10–0.15), and nonexisting effect in the LUKE females (0.01–0.2%
of the genetic variance explained, beta range 0.00–0.02). The
LUKE males showed a nonsignificant effect, but the direction of
the effect was similar to Savon Taimen fish (3–6% of the genetic
variance explained, beta range 0.04–0.05). Moreover, by combin-
ing the data of Savon Taimen and LUKE males, the P-values on a
log10 scale of the SNPs on Omy05 were strongly elevated, imply-
ing there may be the same underlying QTL in both groups (Figs. 1
and 3 and Table 2). However, there is no immediate explanation
of the discrepancy we observe between sexes and populations.
There is evidence that the Omy05 inversion contains loci that are
associated with sex determination and sex-specific expression of
traits (Pearse et al. 2019), and we know that the sires of the Savon
Taimen samples are sex reversed females. However, the cause of
Table 2. Distribution of the genotyped samples by population, sex, and survival outcome, subdivided by the Omy05 inversion genotypes
(AA for homozygote ancestral, AR for heterozygote, and RR for homozygote rearranged genotypes).
Savon Taimen LUKE
Females Males Females
Omy5 genotype A/A A/R R/R Total A/A A/R R/R Total A/A A/R R/R Total
Alive (0) 635 208 7 850 501 171 8 680 486 174 23 683
Dead (1) 268 148 18 434 448 194 19 661 472 161 21 654
Mortality 0.30 0.42 0.72 0.47 0.53 0.70 0.49 0.48 0.48
The last line provides the mortality for each population and sex, according to each Omy05 inversion genotype.
Table 3. Genetic correlation between Savon Taimen and LUKE
population for mortality, and heritability (h2) of mortality,
estimated with a linear model on both observed and liability
scales, and the posterior estimates obtained with a threshold
model.
Sample mean SD
Linear model
Genetic correlation 0.62 0.25
Savon Taimen: h2 observed scale 0.21 0.05
Savon Taimen: h2 liability scale 0.35 0.08
LUKE: h2 observed scale 0.20 0.03
LUKE: h2 liability scale 0.32 0.05
Posterior mean Posterior sd
Posterior estimate of threshold model
Genetic correlation 0.64 0.23
Savon Taimen: h2 0.36 0.06
LUKE: h2 0.32 0.04
F. C. F. Calboli et al. | 9
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
the discrepancy of effects in the 2 populations and sexes still
remains unexplained.
The QTL on Omy03 is conserved in the 2
populations
Our analysis successfully identified a QTL on chromosome 3 as-
sociated with mortality. The broad picture we can take from this
analysis is that the QTL on chromosome Omy03 is strongly con-
served between these 2 populations, both in terms of location,
and the proportion of genetic variance explained (LUKE females
7–10%; LUKE males 8–12%; Savon Taimen 4–9%; all fish together
7–11%; LUKE males and Savon Taimen 6–11%), and with what
seems to be a negligible, or limited, effect of sex on the QTL.
With the full data analyzed together, the QTL on chromosome
Omy03 shows the highest P-value peak (on a log10 scale) com-
pared to any other locus (Figs. 1 and 2). The significance of the P-
value for association with mortality diminished depending on
what subset of the data was included in analysis, though in all
cases the QTL on Omy03 reached chromosome-wide or genome-
wide significance. The loss of power in resolving the QTL is obvi-
ously not at all surprising, given that fewer samples lead to less
statistical power. In the LUKE samples, both males and females
showed a P-value for association between the QTL and mortality
that was broadly of the same order of magnitude, though it is
less significant than when all samples are considered together.
The Savon Taimen data showed a slightly lower significance
compared to either sex in the LUKE data. These consistent pat-
terns differed from the results of the Omy05.
The only previous effort to map QTL for mortality to columna-
ris disease by GWAS does not report association loci matching
our results (Silva, Evenhuis, Vallejo, Gao, et al. 2019). Silva,
Evenhuis, Vallejo, Gao, et al. (2019) used 2 breeding lines of North
American rainbow trout exposed to a virulent strain of F. colum-
nare, analyzed independently with the same SNP chip used here,
but with an exposure/challenge experimental setup, rather than
a naturally occurring outbreaks (as it is for our data). The results
of that experiment identified QTLs that were mostly not overlap-
ping in the 2 strains, and none of them were either on chromo-
somes Omy03 or Omy05. On the other hand, a positive
correlation has been reported between resistance to columnaris
disease and BCWD (Evenhuis et al. 2015; Silva, Evenhuis, Vallejo,
Tsuruta, et al. 2019). Three QTLs recently reported for BCWD re-
sistance (Vallejo, Liu, et al. 2017; Liu et al. 2018; Fraslin et al. 2018;
Fraslin et al. 2019) have been identified on chromosome Omy03, a
result that is comparable with our results. Nevertheless, the
genome-wide significant SNPs we identified on Omy03 span from
63,527,178 to 70,261,448 bp, placing this QTL between the 2 of the
reported for BCWD (the first spanning from 61,621,949 to
62,558,467 bp, the second spanning from 77,108,538 to78,076,592
bp). Our results raise the question of the degree to which the cor-
relation between CD and BCWD is due to the fact that the same
QTLs control both traits, or whether different QTLs are located in
the same genomic region in linkage disequilibrium.
Heritability and genetic correlation
We estimated heritability (on both observed and liability scales),
and genetic correlation using the realized GRM generated from
the genome-wide SNP data. This approach captures the full ge-
netic contribution to the mortality to columnaris disease, rather
than just the effects of the main QTLs, and is parallel to the ap-
proach that can be used in genomic evaluation to estimate geno-
mic breeding values. The liability-scale heritability values
indicate a moderate genetic effect on the phenotype, which
broadly matches those previously reported for columnaris dis-
ease (0.23 and 0.34, Vallejo, Liu, et al. 2019), using the same SNP
chip.
An advantage of our approach of using a genome-wide SNP
dataset in the 2 populations is the ability to have a measure of
the realized genomic relationship between individuals of both
populations. This allows to directly assess the genetic correlation
between populations for mortality, even for distinct populations
and when the traditional sire-dam-offspring pedigree does not
exist (Table 3). In our data, the genetic correlation was positive
(rG ¼ 0.620.64) indicating that the genetic basis of resistance is
indeed partly similar in the 2 populations. Yet, these figures are
likely an underestimate of the true genomic correlation between
populations, given that the causal variants were likely not di-
rectly genotyped, there is recombination between SNPs and the
causal variant, and the allele frequencies at the causal variant
differ in the 2 populations (Wientjes et al. 2018). To summarize,
tn line with our QTL analysis these results together reflect a mod-
erate degree of consistency of QTL effects and of other minor
polygenetic genetic effects in determining the phenotype.
QTL validation
The conservation of QTL locations is typically ascertained by test-
ing for the same phenotype in multiple populations (Vallejo et al.
2014; Evenhuis et al. 2015; Vallejo, Liu, et al. 2017; Silva, Evenhuis,
Vallejo, Gao, et al. 2019) or generations (Liu et al. 2018) for the
presence and location of the QTLs, as we present here with the
Savon Taimen and LUKE populations. Our results indicate that
mortality to columnaris disease is either oligogenic or polygenic
in nature, a fact that makes QTL validation more challenging
compared to monogenic traits. Nevertheless, our data does pro-
vide evidence that some QTLs associated with decreased mortal-
ity to columnaris disease are conserved in the 2 separate
populations of rainbow trout, supporting the hypothesis that this
trait is indeed controlled by few replicable QTLs (that explain
around 10% of genetic variation) and the remaining polygenetic
effects. There are 2 mutually nonexclusive reason for finding
similar QTLs in the 2 populations. First, the most significant SNPs
may be strongly linked to the causal variant, with minimal chan-
ces of recombination breaking the linkage between SNPs and
causal variant. Second, the close kinship between the 2 popula-
tions in our analysis suggests that the genetic architecture for a
phenotype will overlap.
The large inversion on Omy05 provides one special case of a
genomic region with limited or no recombination. The effect of
this inversion is to create a long haplotype with very strong or
complete linkage disequilibrium between all markers. Any QTL
located on this inversion can be effectively tagged by any and all
the SNPs known to tag the inversion, making allele detection for
the inversion extremely easy and reliable.
Marker-assisted selection and genomic selection
Our results show that there is potential to improve resistance
against F. columnare by both genomic selection and MAS. To date,
the most successful implementations of MAS in aquaculture are
the well-publicized infectious pancreatic necrosis (IPN) case
(Houston et al. 2010; Moen et al. 2009), resistance for lymphocystis
disease in Japanese flounder (Fuji et al. 2007), and, potentially,
rainbow trout selected for resistance to BCWD using 2 QTLs
(these are commercially marketed by Aquagen, but no results
have been published in the literature). The VGLL3 locus (Ayllon
et al. 2015; Barson et al. 2015) could also be used for selection of
maturity and growth traits in European stocks of Atlantic salmon
10 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
(Debes et al. 2021), but most likely not in North American ones,
because of the different genetic architecture of traits in North
American stocks of Atlantic salmon (Boulding et al. 2019). In our
data, we observe that the main QTL explains 7–11% of the genetic
variance with the extreme genotypes having a 8–11% difference
in mortality, implying they would be also a valid choice for MAS
for decreased mortality. The inversion on Omy05 would be an ef-
fective target of MAS, because it can be extremely reliably tagged,
and because tagging markers are unlikely to decrease their tag-
ging potential even over a long time period because the inversion
precludes recombination. The presence of QTLs for decreased
mortality, and the opportunity to use MAS in a warming climate,
where outbreaks of columnaris disease are more likely, can have
substantial practical implications.
GS that uses genome-wide marker sets is a second viable ap-
proach, and it allows to estimate a genomic breeding value for
each sample, the GEBV (Meuwissen et al. 2001; Goddard and
Hayes 2007). GS is especially beneficial for polygenic traits, be-
cause the effect of each locus is small, and thus MAS would not
be very effective (Meuwissen et al. 2001). GS provides improve-
ment over pedigree-based evaluations, for instance in the LUKE
population, GS shows an 13.6% increase in the accuracy of GEBVs
for resistance to F. columnare infection compared to the use of
pedigree data alone (Fraslin et al. 2022). By specifically modeling
the effects of SNPs on phenotype in GS, selection accuracies can
be potentially increased further when QTLs exists (Fraslin et al.
2022). Effectiveness of GS depends on the presence of a dense set
of genome-wide markers capturing all haplotype blocks across
the genome and thus all possible QTLs affecting the trait
(Meuwissen et al. 2001; Goddard and Hayes 2007), and on a suffi-
ciently large reference population (VanRaden et al. 2009). Yet, be-
cause of the potential costs of genotyping many thousands of
markers in thousands of samples, current attempts of using a
much smaller subsets of SNPs than those present in a SNP chip
have the potential of major savings. It is theoretically possible to
impute a genome-wide marker set based on the genotyped
markers, obtaining the benefits of a whole genome scan at a
lower cost (Dufflocq et al. 2019; Tsairidou et al. 2020). It is also pos-
sible to use these marker sets to reconstruct pedigrees, for a sin-
gle step GBLUP approach. Nevertheless, a much smaller subset of
SNPs might prove problematic in reconstructing the GRM
(Misztal et al. 2020) and might not capture all loci involved in the
phenotype, and thus it might decrease the effectiveness of selec-
tion approaches. Despite these concerns, empirical evidence
shows that small marker sets with markers chosen to tag known
QTLs can provide better power and more accurate genomic pre-
dictions than pedigree based approaches, at least for disease re-
sistance (Vallejo, Leeds, et al. 2017; Vallejo et al. 2018).
Aquaculture species have been only recently domesticated and
subject to selection, which might also explain the presence of
QTLs explaining a large amount of variation for many commer-
cially relevant traits (Houston et al. 2020), due to the fact selection
for the most important QTLs has not brought them to fixation.
To conclude, our results bring an important insight into this
discussion. We show that resistance to columnaris disease is oli-
gogenic in the 2 populations we analyzed, with QTL conservation
between them. Additionally, our results show the potential im-
portance of genomic rearrangements in the genetic architecture
of a trait. Genomic rearrangements reduce or ablate recombina-
tion and can be successfully tagged by one or 2 SNPs, with fewer
chances that recombination would weaken the linkage between
QTL and tagging SNPs over generations. Our results also show
that, while inversions can play an important role in determining
a phenotype and can have potential to be used to augment selec-
tive breeding, their actual effect needs to be characterized on a
case by case.
Ethic statement
The establishment of progeny families at LUKE’s facilities fol-
lowed the protocols approved by the LUKE’s Animal Care
Committee, Helsinki, Finland. Savon Taimen Oy and Hanka-
Taimen Oy, 2 fish farming companies, have authorization for fish
management, rearing and experiments, and all parties comply
with the EU Directive 2010/63/EU for animal experiments.
Data availability
Data are made available on figshare at the address: https://doi.
org/10.6084/m9.figshare.19323506.
Conflicts of interest
None declared.
Literature cited
Aslam ML, Robledo D, Krasnov A, Moghadam HK, Hillestad B,
Houston RD, Baranski M, Boison S, Robinson NA. Quantitative
trait loci and genes associated with salmonid alphavirus load in
Atlantic salmon: implications for pancreas disease resistance
and tolerance. Sci Rep. 2020;10(1):10393. doi:10.1038/s41598-02
0–67405-8.
Ayllon F, Kjærner-Semb E, Furmanek T, Wennevik V, Solberg MF,
Dahle G, Taranger GL, Glover KA, Almen MS, Rubin CJ, et al. The
vgll3 locus controls age at maturity in wild and domesticated
Atlantic salmon (Salmo salar L.) males. PLoS Genet. 2015;11(11):
e1005628. doi:10.1371/journal.pgen.1005628.
Barson NJ, Aykanat T, Hindar K, Baranski M, Bolstad GH, Fiske P, Jacq
C, Jensen AJ, Johnston SE, Karlsson S, et al. Sex-dependent domi-
nance at a single locus maintains variation in age at maturity in
salmon. Nature. 2015;528(7582):405–408. doi:10.1038/na-
ture16062.
Berthelot C, Brunet F, Chalopin D, Juanchich A, Bernard M, Noe¨l B,
Bento P, Da Silva C, Labadie K, Alberti A, et al. The rainbow trout
genome provides novel insights into evolution after whole-
genome duplication in vertebrates. Nat Commun. 2014;5(1):1–10.
doi:10.1038/ncomms4657.
Boison S, Ding J, Leder E, Gjerde B, Bergtun PH, Norris A, Baranski M,
Robinson N. QTLs associated with resistance to cardiomyopathy
syndrome in Atlantic salmon. J Hered. 2019;110(6):727–737. doi:
10.1093/jhered/esz042.
Boudry P, Allal F, Aslam ML, Bargelloni L, Bean TP, Brard-Fudulea S,
Brieuc MS, Calboli FC, Gilbey J, Haffray P, et al. Current status and
potential of genomic selection to improve selective breeding in
the main aquaculture species of International Council for the
Exploration of the Sea (ICES) member countries. Aquac Rep.
2021;20:100700. doi:10.1016/j.aqrep.2021.100700.
Boulding EG, Ang KP, Elliott JAK, Powell F, Schaeffer LR. Differences in
genetic architecture between continents at a major locus previously
associated with sea age at sexual maturity in European Atlantic
salmon. Aquaculture. 2019;500:670–678. doi:10.1016/j.aquacul-
ture.2018.09.025.
F. C. F. Calboli et al. | 11
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, Lee JJ. Second-
generation PLINK: rising to the challenge of larger and richer data-
sets. Gigascience. 2015;4:7. doi:10.1186/s13742-015-0047-8.
Debes PV, Piavchenko N, Ruokolainen A, Ovaskainen O, Moustakas-
Verho JE, Parre N, Aykanat T, Erkinaro J, Primmer CR. Polygenic
and major-locus contributions to sexual maturation timing in
Atlantic salmon. Mol Ecol. 2021;30(18):4505–4519. doi:
10.1111/mec.16062.
Declercq AM, Boyen F, Van den Broeck W, Bossier P, Karsi A, Haesebrouck
F, Decostere A. Antimicrobial susceptibility pattern of Flavobacterium
columnare isolates collected worldwide from 17 fish species. J Fish Dis.
2013;36(1):45–55. doi:10.1111/j.1365–2761.2012.01410.x.
Declercq AM, Haesebrouck F, Van Den Broeck W, Bossier P,
Decostere A. Columnaris disease in fish: a review with emphasis
on bacterium-host interactions. Vet Res. 2013;44:27. doi:
10.1186/1297–9716-44-27.
Dempster ER, Lerner IM. Heritability of threshold characters.
Genetics. 1950;35(2):212–236.
Dufflocq P, Perez-Enciso M, Lhorente JP, Ya´~nez JM. Accuracy of geno-
mic predictions using different imputation error rates in aqua-
culture breeding programs: a simulation study. Aquaculture.
2019;503:225–230. doi:10.1016/j.aquaculture.2018.12.061.
D’Ambrosio J, Phocas F, Haffray P, Bestin A, Brard-Fudulea S, Poncet
C, Quillet E, Dechamp N, Fraslin C, Charles M, et al. Genome-wide
estimates of genetic diversity, inbreeding and effective size of ex-
perimental and commercial rainbow trout lines undergoing se-
lective breeding. Genet Sel Evol. 2019;51(1):26. doi:
10.1186/s12711-019–0468-4.
Evenhuis JP, Lapatra SE, Marancik D. Early life stage rainbow trout
(Oncorhynchus mykiss) mortalities due to Flavobacterium columnare
in Idaho, USA. Aquaculture. 2014;418–419:126–131. doi:
10.1016/j.aquaculture.2013.09.044.
Evenhuis JP, Leeds TD, Marancik DP, Lapatra SE, Wiens GD. Rainbow
trout (Oncorhynchus mykiss) resistance to columnaris disease is
heritable and favorably correlated with bacterial cold water dis-
ease resistance. J Anim Sci. 2015;93(4):1546–1554.
Fraslin C, Brard-Fudulea S, D’Ambrosio J, Bestin A, Charles M,
Haffray P, Quillet E, Phocas F. Rainbow trout resistance to bacte-
rial cold water disease: two new quantitative trait loci identified
after a natural disease outbreak on a French farm. Anim Genet.
2019;50(3):293–297. doi:10.1111/age.12777.
Fraslin C, Dechamp N, Bernard M, Krieg F, Hervet C, Guyomard R,
Esquerre D, Barbieri J, Kuchly C, Duchaud E, et al. Quantitative trait
loci for resistance to Flavobacterium psychrophilum in rainbow trout:
effect of the mode of infection and evidence of epistatic interac-
tions. Genet Sel Evol. 2018;50(1). doi:1186/s12711-018–0431-9.
Fraslin C, Koskinen H, Nousianen A, Houston RD, Kause A. Genome-wide
association and genomic prediction of resistance to Flavobacterium col-
umnare in a farmed rainbow trout population. Aquaculture. 2022;557:
738332. doi:10.1016/j.aquaculture.2022.738332.
Fuji K, Hasegawa O, Honda K, Kumasaka K, Sakamoto T, Okamoto N.
Marker-assisted breeding of a lymphocystis disease-resistant
Japanese flounder (Paralichthys olivaceus). Aquaculture. 2007;
272(1–4):291–295. doi:10.1016/j.aquaculture.2007.07.210.
Gao G, Magadan S, Waldbieser GC, Youngblood RC, Wheeler PA,
Scheffler BE, Thorgaard GH, Palti Y. A long reads-based de-novo as-
sembly of the genome of the Arlee homozygous line reveals chro-
mosomal rearrangements in rainbow trout. G3 (Bethesda). 2021;11:
jka052. doi:1093/g3journal/jkab052.
Gao G, Nome T, Pearse DE, Moen T, Naish KA, Thorgaard GH, Lien S,
Palti Y. A new single nucleotide polymorphism database for rain-
bow trout generated through whole genome resequencing. Front
Genet. 2018;9:147. doi:10.3389/fgene.2018.00147.
Goddard ME, Hayes BJ. Genomic selection. J Anim Breed Genet. 2007;
124(6):323–330. doi:10.1534/genetics.107.077784.
Goddard ME, Hayes BJ. Mapping genes for complex traits in domestic
animals and their use in breeding programmes. Nat Rev Genet.
2009;10(6):381–391. doi:10.1038/nrg2575.
Gonen S, Baranski M, Thorland I, Norris A, Grove H, Arnesen P,
Bakke H, Lien S, Bishop SC, Houston RD, et al. Mapping and vali-
dation of a major QTL affecting resistance to pancreas disease
(salmonid alphavirus) in Atlantic salmon (Salmo salar). Heredity.
2015;115(5):405–414. doi:10.1038/hdy.2015.37.
Hayes BJ, Bowman PJ, Chamberlain AJ, Goddard ME. Invited review:
genomic selection in dairy cattle: progress and challenges. J Dairy
Sci. 2009;92(2):433–443. doi:10.3168/jds.2008-1646.
Hillestad B, Makvandi-Nejad S, Krasnov A, Moghadam HK.
Identification of genetic loci associated with higher resistance to
pancreas disease (PD) in Atlantic salmon (Salmo salar L.). BMC
Genomics. 2020;21(1):388. doi:10.1186/s12864-020–06788-4.
Hillestad B, Moghadam HK. Genome-wide association study of pi-
scine myocarditis virus (PMCV) Resistance in Atlantic Salmon
(Salmo salar). J Hered. 2019;110(6):720–726. doi:10.1093/jhere-
d/esz040.
Houston RD, Bean TP, Macqueen DJ, Gundappa MK, Jin YH, Jenkins TL,
Selly SLC, Martin SAM, Stevens JR, Santos EM, et al. Harnessing ge-
nomics to fast-track genetic improvement in aquaculture. Nat Rev
Genet. 2020;21(7):389–409. doi:10.1038/s41576-020–0227-y.
Houston RD, Haley CS, Hamilton A, Guy DR, Mota-Velasco JC,
Gheyas AA, Tinch AE, Taggart JB, Bron JE, Starkey WG, et al. The
susceptibility of Atlantic salmon fry to freshwater infectious pan-
creatic necrosis is largely explained by a major QTL. Heredity.
2010;105(3):318–327. doi:10.1038/hdy.2009.171.
Ka¨rkka¨inen HP, Sillanpa¨a¨ MJ. Back to basics for Bayesian model
building in genomic selection. Genetics. 2012;191(3):969–987. doi:
10.1534/genetics.112.139014.
Karvonen A, Rintama¨ki P, Jokela J, Valtonen ET. Increasing water
temperature and disease risks in aquatic systems: climate
change increases the risk of some, but not all, diseases. Int J
Parasitol. 2010;40(13):1483–1488. doi:10.1016/j.ijpara.2010.04.015.
Kause A, Ritola O, Paananen T, Wahlroos H, Ma¨ntysaari EA. Genetic
trends in growth, sexual maturity and skeletal deformations, and
rate of inbreeding in a breeding programme for rainbow trout
(Oncorhynchus mykiss). Aquaculture. 2005;247(1–4):177–187. doi:
10.1016/j.aquaculture.2005.02.023.
Laanto E, Bamford JK, Ravantti JJ, Sundberg LR. The use of phage
FCL-2 as an alternative to chemotherapy against columnaris dis-
ease in aquaculture. Front Microbiol. 2015;6:829. doi:
10.3389/fmicb.2015.00829.
Lande R, Thompson R. Efficiency of marker-assisted selection in the
improvement of quantitative traits. Genetics. 1990;124(3):
743–756.
Leeds TD, Silverstein JT, Weber GM, Vallejo RL, Palti Y, Rexroad CE,
Evenhuis J, Hadidi S, Welch TJ, Wiens GD, et al. Response to selec-
tion for bacterial cold water disease resistance in rainbow trout. J
Anim Sci. 2010;88(6):1936–1946. doi:10.2527/jas.2009-2538.
Liu S, Gao G, Layer RM, Thorgaard GH, Wiens GD, Leeds TD, Martin
KE, Palti Y. Identification of high-confidence structural variants
in domesticated rainbow trout using whole-genome sequencing.
Front Genet. 2021;12:639355. doi:10.3389/fgene.2021.639355.
Liu S, Vallejo RL, Evenhuis JP, Martin KE, Hamilton A, Gao G, Leeds
TD, Wiens GD, Palti Y. Retrospective evaluation of marker-
assisted selection for resistance to bacterial cold water disease in
three generations of a commercial rainbow trout breeding popu-
lation. Front Genet. 2018;9:286. doi:10.3389/fgene.2018.00286.
12 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
Lynch M, Walsh B. Genetics and the Analysis of Quantitative Traits.
Sunderland (MA): Sinauer; 1997.
Malecki JK, Roy LA, Arias CR, Nhat Truong T, Hanson TR, Lange MD,
Shoemaker CA, Beck BH. Bioeconomic analysis of Flavobacterium
columnare vaccine pond trials with channel catfish. North Am J
Aquaculture. 2021;83(3):207–217. doi:10.1002/naaq.10191.
Meuwissen TH, Hayes BJ, Goddard ME. Prediction of total genetic
value using genome-wide dense marker maps. Genetics. 2001;
157(4):1819–1829.
Misztal I, Lourenco D, Legarra A. Current status of genomic evalua-
tion. J Anim Sci. 2020;98:1–14. doi:10.1093/jas/skaa101.
Misztal I, Tsuruta S, Strabel T, Auvray B, Druet T, et al. BLUPF90 and
related programs (BGF90). Proceedings of the 7th World Congress
on Genetics Applied to Livestock Production. Montpellier
(France). 28; 2002.
Moen T, Baranski M, Sonesson AK, Kjøglum S. Confirmation and
fine-mapping of a major QTL for resistance to infectious pancre-
atic necrosis in Atlantic salmon (Salmo salar): population-level
associations between markers and trait. BMC Genomics. 2009;10:
368. doi:10.1186/1471–2164-10-368.
Newton JC, Wood TM, Hartley MM. Isolation and partial characteri-
zation of extracellular proteases produced by isolates of
Flavobacterium columnare derived from channel catfish. J Aquat
Anim Health. 1997;9(2):75–85.
Palti Y, Gao G, Liu S, Kent MP, Lien S, Miller MR, Rexroad CE, Moen T.
The development and characterization of a 57K single nucleotide
polymorphism array for rainbow trout. Mol Ecol Resour. 2015;
15(3):662–672. doi:10.1111/1755-0998.12337.
Pearse DE, Barson NJ, Nome T, Gao G, Campbell MA, Abadıa-Cardoso A,
Anderson EC, Rundio DE, Williams TH, Naish KA, et al. Sex-depen-
dent dominance maintains migration supergene in rainbow trout.
Nat Ecol Evol. 2019;3(12):1731–1742. doi:10.1038/s41559-019–1044-6.
Sae-Lim P, Kause A, Mulder HA, Olesen I. Breeding and genetics sym-
posium: climate change and selective breeding in aquaculture. J
Anim Sci. 2017;95(4):1801–1812. doi:10.2527/jas2016.1066.
Silva RMO, Evenhuis JP, Vallejo RL, Gao G, Martin KE, Leeds TD, Palti
Y, Lourenco DAL. Whole-genome mapping of quantitative trait
loci and accuracy of genomic predictions for resistance to colum-
naris disease in two rainbow trout breeding populations. Genet
Sel Evol. 2019;51(1):42. doi:10.1186/s12711-019–0484-4.
Silva RMO, Evenhuis JP, Vallejo RL, Tsuruta S, Wiens GD, Martin KE,
Parsons JE, Palti Y, Lourenco DAL, Leeds TD, et al. Variance and co-
variance estimates for resistance to bacterial cold water disease
and columnaris disease in two rainbow trout breeding popula-
tions1. J Anim Sci. 2019;97(3):1124–1132. doi:10.1093/jas/sky478.
Starliper CE. Bacterial coldwater disease of fishes caused by
Flavobacterium psychrophilum. J Adv Res. 2011;2(2):97–108. doi:
10.1016/j.jare.2010.04.001.
Toli EA, Calboli FCF, Shikano T, Merila¨ J. A universal and reliable as-
say for molecular sex identification of three-spined sticklebacks
(Gasterosteus aculeatus). Mol Ecol Resources. 2016;16(6):1389–1400.
doi:10.1111/1755-0998.12543.
Tsairidou S, Hamilton A, Robledo D, Bron JE, Houston RD. Optimizing
low-cost genotyping and imputation strategies for genomic selec-
tion in Atlantic salmon. G3 (Bethesda). 2020;10(2):581–590. doi:
10.1534/g3.119.400800.
Vallejo RL, Leeds TD, Gao G, Parsons JE, Martin KE, Evenhuis JP,
Fragomeni BO, Wiens GD, Palti Y. Genomic selection models dou-
ble 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. 2017;
49(1):17. doi:10.1186/s12711-017-0293-6.
Vallejo RL, Liu S, Gao G, Fragomeni BO, Hernandez AG, Leeds TD,
Parsons JE, Martin KE, Evenhuis JP, Welch TJ, et al. Similar genetic
architecture with shared and unique quantitative trait loci for
bacterial cold water disease resistance in two rainbow trout
breeding populations. Front Genet. 2017;8:156. doi:10.3389/fgene.
2017.00156.
Vallejo RL, Palti Y, Liu S, Marancik DP, Wiens GD. Validation of linked
QTL for bacterial cold water disease resistance and spleen size on
rainbow trout chromosome Omy19. Aquaculture. 2014;432:
139–143. doi:10.1016/j.aquaculture.2014.05.003.
Vallejo RL, Silva RMO, Evenhuis JP, Gao G, Liu S, Parsons JE, Martin
KE, Wiens GD, Lourenco DAL, Leeds TD, et al. Accurate genomic
predictions for BCWD resistance in rainbow trout are achieved
using low-density SNP panels: evidence that long-range LD is a
major contributing factor. J Anim Breed Genet. 2018;135(4):
263–274. doi:10.1111/jbg.12335.
VanRaden PM, Van Tassell CP, Wiggans GR, Sonstegard TS, Schnabel
RD, Taylor JF, Schenkel FS. Invited review: reliability of genomic
predictions for North American Holstein bulls. J Dairy Sci. 2009;
92(1):16–24. doi:10.3168/jds.2008-1514.
van Vliet MT, Franssen WH, Yearsley JR, Ludwig F, Haddeland I,
Lettenmaier DP, Kabat P. Global river discharge and water tem-
perature under climate change. Global Environ Change. 2013;
23(2):450–464. doi:10.1016/j.gloenvcha.2012.11.002.
Villoutreix R, Ayala D, Joron M, Gompert Z, Feder JL, Nosil P.
Inversion breakpoints and the evolution of supergenes. Mol Ecol.
2021;30(12):2738–2755. doi:10.1111/mec.15907.
Waskiewicz A, Irzykowska L. 2014. Flavobacterium
spp.—Characteristics, occurrence, and toxicity. In: Batt CA,
Tortorello ML, editors. Encyclopedia of Food Microbiology.
Cambridge, Massachusetts: Academic Press. p. 938–942.
Wang H, Misztal I, Aguilar I, Legarra A, Fernando RL, Vitezica Z,
Okimoto R, Wing T, Hawken R, Muir WM. Genome-wide associa-
tion mapping including phenotypes from relatives without
genotypes in a single-step (ssGWAS) for 6-week body weight in
broiler chickens. Front Genet. 2014;5:134. doi:10.3389/fgene.
2014.00134
Weinstein SY, Thrower FP, Nichols KM, Hale MC. A large-scale chro-
mosomal inversion is not associated with life history develop-
ment in rainbow trout from Southeast Alaska. PLoS One. 2019;
14(9):e0223018. doi:10.1371/journal.pone.0223018.
Weir BS, Anderson AD, Hepler AB. Genetic relatedness analysis:
modern data and new challenges. Nat Rev Genet. 2006;7(10):
771–780. doi:10.1038/nrg1960.
Weir BS, Cockerham CC. Estimating F-statistics for the analysis of
population structure. Evolution. 1984;38(6):1358–1370.
Wientjes YCJ, Calus MPL, Duenk P, Bijma P. Required properties for
markers used to calculate unbiased estimates of the genetic cor-
relation between populations. Genet Sel Evol. 2018;50(1):65. doi:
10.1186/s12711-018–0434-6.
Yang J, Lee SH, Goddard ME, Visscher PM. GCTA: a tool for genome-
wide complex trait analysis. Am J Hum Genet. 2011;88(1):76–82.
doi:10.1016/j.ajhg.2010.11.011.
Yang J, Zaitlen NA, Goddard ME, Visscher PM, Price AL. Advantages and
pitfalls in the application of mixed-model association methods. Nat
Genet. 2014;46(2):100–106. doi:10.1038/ng.2876.
F. C. F. Calboli et al. | 13
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022
You X, Shan X, Shi Q. Research advances in the genomics and applica-
tions for molecular breeding of aquaculture animals. Aquaculture.
2020;526:735357. doi:10.1016/j.aquaculture.2020.735357.
Zachos J, Pagani M, Sloan L, Thomas E, Billups K. Trends, rhythms,
and aberrations in global climate 65 Ma to present. Science. 2001;
292(5517):686–693. doi:10.1126/science.1059412.
Zhang Y, Zhao L, Chen W, Huang Y, Yang L, Sarathbabu V, Wu Z, Li J,
Nie P, Lin L, et al. Complete genome sequence analysis of the fish
pathogen Flavobacterium columnare provides insights into antibi-
otic resistance and pathogenicity related genes. Microb Pathog.
2017;111:203–211. doi:10.1016/j.micpath.2017.08.035.
Communicating editor: D.J. de Koning
14 | G3, 2022, Vol. 00, No. 0
D
ow
nloaded from
 https://academ
ic.oup.com
/g3journal/advance-article/doi/10.1093/g3journal/jkac137/6603111 by guest on 30 June 2022