G3, 2023, 00(0),

https://doi.org/10.1093/g3journal/jkad138
Advance Access Publication Date: 19 June 2023
Genome Report

The generation of the first chromosome-level de novo

genome assembly and the development and validation

Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023

of a 50K SNP array for the St. John River aquaculture
strain of North American Atlantic salmon
Guangtu Gao,1,* Geoffrey C. Waldbieser,2 Ramey C. Youngblood,3 Dongyan Zhao,4 Michael R. Pietrak,5 Melissa S. Allen,6
Jason A. Stannard,6 John T. Buchanan,6 Roseanna L. Long,1 Melissa Milligan,5 Gary Burr,5 Katherine Mejía-Guerra,4
Moira J. Sheehan,4 Brian E. Scheffler,7 Caird E. Rexroad III,8 Brian C. Peterson,5 Yniv Palti 1,
*
1
USDA-ARS National Center for Cool and Cold Water Aquaculture, 11861 Leetown Road, Kearneysville, WV 25430, USA
2
USDA-ARS Warmwater Aquaculture Research Unit, 141 Experimental Station Road, Stoneville, MS 38776, USA
3
Institute for Genomics, Biocomputing and Biotechnology, Mississippi State University, Mississippi State, MS 39762, USA
4
Breeding Insight, 119 CALS Surge Facility, Cornell University, 525 Tower Road, Ithaca, NY 14853, USA
5
USDA-ARS National Cold Water Marine Aquaculture Center, 25 Salmon Farm Road, Franklin, ME 04634, USA
6
Center for Aquaculture Technologies, 8395 Camino Santa Fe, San Diego, CA 92121, USA
7
USDA-ARS Genomics and Bioinformatics Research Unit, 141 Experimental Station Road, Stoneville, MS 38776, USA
8
USDA-ARS Office of National Programs, George Washington Carver Center Room 4-2106, 5601 Sunnyside Avenue, Beltsville, MD 20705, USA

*Corresponding author: Email: yniv.palti@usda.gov; *Corresponding author: Email: Guangtu.gao@usda.gov.

Abstract
Atlantic salmon (Salmo salar) in Northeastern US and Eastern Canada has high economic value for the sport fishing and aquaculture in-
dustries. Large differences exist between the genomes of Atlantic salmon of European origin and North American (N.A.) origin. Given the
genetic and genomic differences between the 2 lineages, it is crucial to develop unique genomic resources for N.A. Atlantic salmon.
Here, we describe the resources that we recently developed for genomic and genetic research in N.A. Atlantic salmon aquaculture.
Firstly, a new single nucleotide polymorphism (SNP) database for N.A. Atlantic salmon consisting of 3.1 million putative SNPs was gen-
erated using data from whole-genome resequencing of 80 N.A. Atlantic salmon individuals. Secondly, a high-density 50K SNP array en-
riched for the genic regions of the genome and containing 3 sex determination and 61 putative continent of origin markers was
developed and validated. Thirdly, a genetic map composed of 27 linkage groups with 36K SNP markers was generated from 2,512 in-
dividuals in 141 full-sib families. Finally, a chromosome-level de novo genome assembly from a male N.A. Atlantic salmon from the
St. John River aquaculture strain was generated using PacBio long reads. Information from Hi-C proximity ligation sequences and
Bionano optical mapping was used to concatenate the contigs into scaffolds. The assembly contains 1,755 scaffolds and only 1,253
gaps, with a total length of 2.83 Gb and N50 of 17.2 Mb. A BUSCO analysis detected 96.2% of the conserved Actinopterygii genes
in the assembly, and the genetic linkage information was used to guide the formation of 27 chromosome sequences. Comparative ana-
lysis with the reference genome assembly of the European Atlantic salmon confirmed that the karyotype differences between the 2
lineages are caused by a fission in chromosome Ssa01 and 3 chromosome fusions including the p arm of chromosome Ssa01 with
Ssa23, Ssa08 with Ssa29, and Ssa26 with Ssa28. The genomic resources we have generated for Atlantic salmon provide a crucial boost
for genetic research and for management of farmed and wild populations in this highly valued species.

Keywords: genome map, SNP array, Atlantic salmon, North American, lineage, karyotype

Introduction
Atlantic salmon (Salmo salar) in Northeastern US and Eastern
Canada has high economic value for the sport fishing and aqua-
culture industries and is of significant ecological and social im-
portance (Council 2004). Atlantic salmon of North American
(N.A). origin is also being farmed in Tasmania, Australia (Kijas
et al. 2017), and Chile (Yáñez et al. 2016). The historic declines of
wild Atlantic salmon populations in the state of Maine, USA, led
to the listing of wild or natural Atlantic salmon from 8 rivers as en-
dangered under the Federal Endangered Species Act (Council
2004). Evidence from microsatellite DNA markers has clearly de-
monstrated that Atlantic salmon of N.A. origin is genetically dis-
tinct from European Atlantic salmon (King et al. 2001; Council
2002; King et al. 2005). This led to a judicial ruling that strictly regu-
lated farming of Atlantic salmon in Maine and Eastern Canada,
banned the farming of European and genetically modified salmon
strains in these areas, and required routine genetic certification of
the N.A. aquaculture stocks (Council 2004).
Large differences exist between the genomes of the 2 lineages
or subspecies. The haploid chromosome number of the N.A. and
European Atlantic salmon is N = 27 and N = 29, respectively

Received: April 04, 2023. Accepted: May 31, 2023

Published by Oxford University Press on behalf of The Genetics Society of America 2023.
This work is written by (a) US Government employee(s) and is in the public domain in the US.
2 | G3, 2023, Vol. 00, No. 0
(Lubieniecki et al. 2010; Brenna-Hansen et al. 2012). Recent studies
that used whole-genome scans with high-density single nucleo-
tide polymorphism (SNP) assays have provided further evidence
for the genomic differentiation between the N.A. and European
lineages (Lehnert et al. 2020). At the time we started this project,
there was only a reference genome for the European lineage
(Lien et al. 2016). A reference genome is required for discovering
DNA sequence variants among individuals and between popula-
tions and for using those variants to develop high-density geno-
typing assays. The Atlantic salmon genome is large and complex
with a genome size estimate of 3.0 Gbp (Hardie and Hebert
2003). Although it is similar in size to that of most mammals, its
architecture and organization are more complex. In addition to
the ancestral 3R whole-genome duplication shared by all ray-
finned fish, the salmonids have undergone a more recent (4R)
whole-genome duplication event (Allendorf and Thorgaard 1984;
Lien et al. 2016). As a result of the salmonid ancestral 4R whole-
genome duplication, 9 pairs of chromosome arms in the Atlantic
salmon genome have retained more than 90% of sequence simi-
larity to each other with patterns of tetrasomic inheritance. In to-
tal, 94.4% of the genome is composed of 98 pairs of collinear blocks
known as homeologous chromosome regions (Lien et al. 2016). In
addition, nearly 60% of the genome contains repetitive sequences
(Lien et al. 2016). To that end, we have recently demonstrated that
long-read DNA sequencing and associated assembly algorithms
can help to overcome the technical difficulties associated with
generating de novo assemblies for the complex and large genomes
of salmonid species (Gao et al. 2022).
In addition to a reference genome, SNP arrays are also needed
for genetic studies in N.A. Atlantic salmon. High-density SNP ar-
rays are used for collecting large amounts of genome-wide geno-
type data. This information is useful for dissecting the genetic
basis of quantitative traits in agriculture and for implementing
models of genomic selection, which has revolutionized the field
of selective breeding over the past decade (Meuwissen et al.
2016). High-density SNP arrays are publicly available for Atlantic
salmon of European origin (Houston et al. 2014). However, less
than half of the SNPs from the high-density SNP arrays designed
for the European salmon are informative for N.A. salmon (John
Buchanan, unpublished data) (Yáñez et al. 2016). The 2 N.A. popu-
lations that were sampled by Yanez et al. to quantify polymorph-
ism of markers from the array they developed were different from
the aquaculture strains used by the USDA breeding program in
Maine. In addition, Yanez et al. reported that only ∼5% of the
SNPs from the array they developed for Chilean aquaculture
strains of European origin overlapped with SNP arrays that were
developed in Europe at the time. Other researchers and commer-
cial breeding companies have developed a 50K SNP chip for N.A.
Atlantic salmon using information from the European salmon ar-
rays, but those are currently not available for the public due to in-
tellectual property constrains and competitive commercial
interests (John Buchanan, unpublished data) (Kijas et al. 2018).
Currently, resources such as high-density SNP arrays are lacking
for N.A. Atlantic salmon and there is a need for generating a refer-
ence genome and an associated high-density genotyping assay.
Recently, we initiated efforts to generate genomic resources for
the breeding program at the USDA-ARS National Cold Water
Marine Aquaculture Center (NCWMAC) in Franklin, Maine. We
used high-coverage whole-genome Illumina resequencing for
SNP discovery in 80 fish from 3 distinctive aquaculture stocks of
N.A. Atlantic salmon that are propagated in the NCWMAC (Gao
et al. 2020). The Gaspe strain is thought to have originated from
Atlantic salmon from the Grand Cascapedia river, Gaspé Bay,
Quebec (Nash and Waknitz 2003), and propagated for
Aquaculture on the west US coast before being added as an aqua-
culture stock to the USDA breeding program in Maine. The
Penobscot River strain is thought to have originated from a wild
population of N.A. Atlantic salmon in Maine that was spawned
at the Craig brook National Fish Hatchery. The St. John River
(SJR) aquaculture strain is widely used for Atlantic salmon farm-
ing in North America. Wild fish from tributaries above the
Mactaquac Dam (New Brunswick province, Canada) were used
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
to establish the SJR aquaculture strain (Liu et al. 2017). Overall,
we discovered 6.6 million SNP markers, including more than 1.5
million markers with a minor allele frequency (MAF) ≥ 0.25. In
addition, we identified 5,822 candidate markers that can poten-
tially be used to distinguish between N.A. and European Atlantic
salmon by comparing genotypes of the 80 N.A. Atlantic salmon
with publicly available whole-genome sequence information
from 31 Atlantic salmon representing a diversity of European po-
pulations. Although this SNP database provided a foundational re-
source for designing new SNP genotyping arrays specifically
tailored for genomic and genetic analyses in N.A. Atlantic salmon
aquaculture, we also found a high percentage of sequence mis-
match from the alignment of N.A. salmon sequencing reads to
the European salmon reference genome assembly. This indicated
that a reference genome from N.A. salmon can improve the qual-
ity and genome coverage of SNP discovery in these fish.
To enable better resource for SNP discovery and genotyping
and for other comparative genomics and genetic analyses, we
generated and report here the first de novo genome assembly of
nearly complete chromosome sequences for the N.A. Atlantic sal-
mon subspecies. The source DNA for the genome assembly was
from a male fish from the St. John River aquaculture strain.
Advances in long-read DNA sequencing technologies, together
with the development of new assembly algorithms and pipelines,
have recently enabled improved genome assemblies for complex
salmonid genomes like the rainbow trout (Gao et al. 2021). The
new genome assembly for Atlantic salmon was generated using
PacBio long-read sequencing technology followed by scaffolding
with Hi-C contact maps and Bionano optical mapping. The result
was a high-quality, contiguous chromosome sequence. The con-
tigs of the new assembly were used as reference to generate a
new SNP discovery dataset for N.A. Atlantic salmon from which
polymorphic SNP markers that are distributed on all chromo-
somes were selected for a high-density 50K SNP array. The SNP ar-
ray also includes sequence probes from the sdY gene as markers
for sex determination and a limited set of markers that were
used in a proof of concept for preliminary assessment of their use-
fulness for continent of origin (COO) classification (Europe or
North America). The new SNP array was validated by genotyping
pedigreed fish from the USDA breeding program and generating
a linkage map that facilitated anchoring and ordering of the se-
quence scaffolds in 27 chromosome sequences. The reference
genome and high-density SNP array provide very important re-
sources for more precise breeding in aquaculture, and for com-
parative genomics research on the ecology and evolution of
Atlantic salmon.
Materials and methods
Testing with the King 7 microsatellites
The grandparents and parents of all the Atlantic salmon fish used
for SNP discovery, genome sequencing, and SNP array genotyping
in this study were tested by the US Fish and Wildlife Service using

the King 7 microsatellites (King et al. 2001; King et al. 2005) and
were certified to have genotypes consistent of the N.A. COO.
SNP discovery
The same DNA sequence data reported in our previous SNP dis-
covery effort (Gao et al. 2020) were used in a new SNP discovery
analysis to generate the source dataset for the design of a 50K
SNP array. The main difference was that here, we used the N.A.
Atlantic salmon reference genome assembly from a fish from
the SJR strain as the reference for the SNP discovery. All raw se-
quence data used in the SNP discovery can be found in the NCBI
SRA repository (BioProject accession PRJNA559280). Cleaned reads
were aligned to an earlier draft of the Canu contig assembly of the
N.A. Atlantic salmon genome assembly (final version in GenBank
Accession GCA_021399835.1, reported here for the first time)
using BWA-MEM with the default parameters (Li 2013).
Duplicated alignments were marked using PicardTools v2.19.2
(http://broadinstitute.github.io/picard). Reads in regions of de-
tected indels were realigned to remove alignment artifacts, and
base quality scores were recalibrated based on the realignments
to generate analysis-ready reads in bam format using GATK
v3.8.1 (Poplin et al. 2018). Raw variants, including SNPs and indels,
were generated using HaplotypeCaller and GenotypeGVCF tools of
GATK v3.8.1 (Poplin et al. 2018). SNPs were retained if they met the
following filtering criteria: (1) located more than 5 bp from an in-
del; (2) QUAL > 30; (3) minimum and maximum read depths of
100 and 3000, respectively; (4) for each sample, at least 1 read sup-
porting the reference allele and 2 reads supporting the alternative
allele; (5) no missing genotype per SNP position; and (6) MAF > 0.3.
Selection of SNPs for the 50K array
The annotation of the European Atlantic salmon reference gen-
ome was used as a proxy to locate genic regions in the N.A. salmon
genome. Whole-genome alignments between sequence contigs
from the N.A. and European Atlantic salmon genome were gener-
ated using minimap2 (Li 2018). Alignments passing the following
requirements were retained for downstream analyses: (1) align-
ments to the 29 chromosomes of the European salmon genome
(Lien et al. 2016) and contigs of the N.A. salmon genome assembly
that are longer than 5,000 bp; (2) minimum alignment length of
5,000 bp; (3) minimum mapping quality of 40; (4) a N.A. salmon
contig has to have the “assigned” European salmon chromosome
(using the sum of base pairs aligned to the European genome as
a voting system, discard duplicated, or multiple hits with similar
alignment results); and (5) only 1 N.A. salmon contig with the
best alignment score was kept if multiple contigs aligned to the
same region of the European genome. To select 100K SNPs for
technical ranking by the SNP array manufacturer, ∼35 SNPs
were selected in each 1-Mb window, with a minimum distance be-
tween adjacent SNPs of 1.5 kb and prioritized SNPs in genic re-
gions based on the alignments to the European genome
annotation. The dataset of 100K SNPs included 8,654 from 1 of
the European-based high-density SNP arrays (Houston et al.
2014) that were found to be of high quality and polymorphic in
the N.A. stock of the Mowi ASA aquaculture company. The SNP in-
formation for those markers was provided courtesy of Serap
Gonen and Matthew Baranski, and the original Affymetrix SNP
IDs for those markers were retained in the new 50K SNP array.
The ∼100K SNPs were then evaluated by the array manufactur-
er (ThermoFisher, USA) for best compatibility with the SNP array
platform, from which 49,804 were selected as the core SNPs of the
50K Axiom SNP array. The total number of markers on the array is
49,880 which also included 72 SNPs that can potentially be used
G. Gao et al. | 3
for COO classification, including 44 nuclear markers and 20 mito-
chondrial markers identified previously (Gao et al. 2020) and 8 nu-
clear markers that were previously found to be informative for
COO classification by the Center for Aquaculture Technologies
(unpublished data). Four additional sequence probes from the
sdY gene were also added to the array for sex determination ana-
lysis. Those probes were previously included and tested in other
Axiom SNP arrays for Atlantic salmon (Kijas et al. 2018). The pres-
ence of the sequences compatible with the sdY probes would indi-
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
cate that the fish is a male and the absence that it is a female.
Genotyping and validation of the SNP array
and putative COO markers
Fin tissues were collected, and genomic DNA was extracted from
2,512 pedigreed fish from the SJR strain of the USDA breeding pro-
gram. Additional DNA samples (Supplementary Table 1) were
used to evaluate the COO markers in the Axiom array. Genomic
DNA samples were fluorescently labeled, hybridized to the SNP ar-
ray, and scanned for initial genotype calls (Neogen, Lincoln, NE)
according to the Axiom genotyping procedures described by
Affymetrix. For final genotyping calls and quality control ana-
lyses, we used the Affymetrix Power Tools (APT) and SNPolisher
software packages following the procedures and practices sug-
gested by the Affymetrix Axiom array genotyping manual and in
the same way we previously described for the rainbow trout 57K
Axiom array (Palti et al. 2015).
For preliminary evaluation of putative markers that can be
used to differentiate COO, we genotyped 60 samples from
European origin populations and 14 from N.A. origin. These
same samples were part of the panel that was originally used to
test the King 7 microsatellites (King et al. 2001; King et al. 2005).
Principal components analysis (PCA) was conducted with the pro-
gram PLINK version 1.9 (Purcell et al. 2007) using genotypes ob-
tained with 61 validated COO SNP markers. The input .ped file
was generated from the genotype data for each fish treated as un-
related individuals and the .map file was formed using all the COO
markers in the analysis.
Linkage map
The fish used in the linkage analysis were from the St. John River
stock of N.A. Atlantic salmon (NCWMAC-USDA). In total, we gen-
otyped 2,512 fish from 141 full-sib families, including 230 parents
and 2,282 offspring. The program Lep-map3 (Rastas 2017) was
used for generating the linkage map with the same iterative
step-wise approach used for the rainbow trout linkage map
(Gonzalez-Pena et al. 2016). Briefly, the markers were separated
into 27 linkage groups by running the SeparateChromosomes2
module at LOD score 110. More markers were added to the linkage
groups by running the JoinSingles2All module with a more relaxed
LOD threshold. Finally, markers were ordered in each linkage
group by running the OrderMarkers2 module. Linkage groups
were aligned with the karyotype chromosomes by comapping
the flanking sequences of SNP markers from this linkage map
and the map of Brenna-Hansen et al. (Brenna-Hansen et al. 2012)
to the 27 primary sequence scaffolds of the new N.A. Atlantic sal-
mon genome assembly. Two programs were used to map the
flanking sequences of the SNP markers to the genome assembly
sequence scaffolds. The program NovoAlign (http://www.
novocraft.com/products/novoalign/) with the parameter -t 60
was used for the SNP markers of this linkage map, and Blastn
(Zhang et al. 2000) with the parameters of -evalue 1.0e−5,
-perc_identity 80, and -qcov_hsp_perc 80 was used for the mar-
kers of Brenna-Hansen et al. (2012).
4 | G3, 2023, Vol. 00, No. 0
Genome assembly: DNA extraction and
sequencing
DNA from a single male (sample NCWMAC SJR YC14-15 Fam3)
from the USDA St. John River stock of N.A. Atlantic salmon was ex-
tracted using the Qiagen DNeasy kit from a fresh blood sample
collected in tri-potassium EDTA. The blood was stored on ice until
DNA extraction within 24 hours to enhance the yield of high-
molecular weight DNA for long-read sequencing. The DNA sam-
ple was sequenced at the core facility of the University of
Delaware (Newark, Delaware, USA) using 29 SMRT cells of the
PacBio RS-II system in a continuous long-read (CLR) mode, which
generated approximately 312 Gb of sequence data (104× cover-
age). The PacBio read length N50 was ∼33 kb and the average
read length was ∼19 kb. For polishing the genome assembly,
∼2.1 billion paired-end reads (2× 150 bp) of Illumina short reads
were generated by Admera Health (South Plainfield, NJ, USA)
from the same DNA sample. The short-read sequences were pro-
cessed with the program cutadapt to remove the Illumina sequen-
cing adaptors producing a total of ∼316 Gb (∼105× genome
coverage). In general, high genome coverage above 30× is desired
for polishing with Illumina reads.
Construction of the assembly contigs
The initial assembly of sequence contigs was generated from the
PacBio long reads using the Canu (Version 1.8) assembler (Koren
et al. 2017) with the options of correctedErrorRate = 0.085,
corMhapSensitivity = normal, and ovlMerDistinct = 0.975. After
the “correction” and “trimming” steps, which involve alignments
of reads, error correction, trimming of poor-sequence quality seg-
ments, and removal of overlapping reads after those corrections,
Canu retained 2.64 million high-quality reads with a 103.5 Gb
(∼34.5×) total length from the raw PacBio reads. The trimmed
reads were then assembled into 14,845 contigs.
Next, the programs purge_haplotigs (Roach et al. 2018) and pur-
ge_dups (Guan et al. 2020) were used to correct the Canu assembly
errors and identify regional duplications due to high heterogen-
eity in some of the genomic regions. First, we ran purge_haplotigs
to identify artefactual contigs caused by very low or very high
coverage detected by mapping of the original PacBio long read to
the Canu contigs. As a result, 1,394 Canu contigs were character-
ized as artefacts and thus removed from further analysis. Second,
the retained sequences were processed with purge_dups to iden-
tify highly overlapping haplotigs. With this step, 2,263 and
11,977 sequences were identified as the primary contigs and hap-
lotigs, respectively. This information was then used in construc-
tion of the genome assembly scaffolds.
Construction of the assembly scaffolds and
chromosome sequences
To improve the contiguity of the assembly, we first used Bionano
optical mapping with the Saphyr platform (Bionano Genomics,
San Diego, CA, USA) to scaffold the assembled contigs.
High-molecular weight DNA was extracted in agarose gel plugs
from fresh blood of the same Atlantic salmon fish by Amplicon
Express (Pullman, Washington), and the extracted DNA molecules
were labeled with the Direct Label and Stain (DLS) DNA Labeling
kit utilizing the Direct Label Enzyme (DLE-1) (Bionano Genomics,
San Diego, CA, USA). A total of 8.5 million DNA molecules were
generated with a total length of 1.2 Tb (average length 141 kb).
The raw mapping data were processed with Bionano Solve
software (Solve3.3_10252018) (https://bionanogenomics.com/
support-page/bionano-access-software/) to generate a de novo
Bionano assembly guided by the contigs of the PacBio sequence
assembly. About 1.9 million molecules with N50 of ∼305 kb
(∼150× genome coverage) and average label density of 10 sites
per 100 kb were selected from the raw data after the filtering
step, and 1,881 Bionano genome maps were generated with a total
length of ∼3.09 Gb and N50 of ∼5.1 Mb. The Bionano genome maps
and the 14,240 primary contigs and haplotigs were then scaffolded
using the Bionano hybrid scaffold pipeline. This resulted in the
joining of 2,174 contigs (∼2.61 Gb) into 817 scaffolds (∼2.65 Gb)
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
and breaking of 140 contigs that were considered by the Bionano
pipeline to be chimeric joins. The Bionano software utilized
13-bp gaps to indicate the presence of small gaps of undetermined
size or undetermined flanking sequence overlaps (Gao et al. 2021),
and the sequences flanking these gaps were manually corrected.
This resulted in the removal of 21 sequence overlaps, and the total
length of the scaffolds was reduced to ∼2.63 Gb with N50 of
∼14.2 Mb. Finally, 1,195 primary contigs (∼0.2 Gb) that were not
joined with other sequences and were longer than 3 kb were
added to the Bionano scaffolds for further construction of the as-
sembly. Altogether, the Bionano assembly was composed of 2,012
sequence scaffolds with a total length of ∼2.84 Gb and N50 of
∼11.94 Mb.
To improve the accuracy of the Bionano scaffolds, we first used
gcpp (http://github.com/PacificBiosciences/GenomicConsensus)
to polish the sequences. The PacBio raw reads were first mapped
to the Bionano scaffolds using minimap2 (Version 2.15-r915) (Li
2018), and the scaffolds were polished using the “arrow” algorithm
in gcpp based on the alignment results. To further polish the scaf-
folds, the Illumina short-read sequences were aligned to the scaf-
folds using BWA-MEM (version 0.7.17-r188) (Li 2013) with the
default parameters and the genomic variants were called using
the freebayes (version 1.3.1) program (Garrison and Marth 2012)
with the default parameters. Variants with a quality score > 1
and the alternative allele in the Illumina sequence reads were
identified, and the scaffolds were corrected based on the selected
variants using bcftools (Li et al. 2009). This polishing procedure
was performed twice to increase the accuracy of the genome
assembly.
The Hi-C proximity ligation sequence data were used to gener-
ate chromosome-scale scaffolds from the polished Bionano scaf-
fold assembly. Fresh blood (100 µl) was collected in ethanol
(750 µl) and immediately placed in dry ice and shipped. The Hi-C
library was prepared with the Arima HiC Kit (April 2018) using re-
striction enzymes that cut at GATC and GANTC recognition sites
and sequenced to produce 833 million Illumina paired-end reads
(2 × 150 bp) (Arima Genomics, San Diego, CA, USA). The reads
were aligned to the Bionano scaffolds, and the alignments were fil-
tered following the Arima-HiC mapping pipeline (http://github.
com/ArimaGenomics/mapping_pipeline). Briefly, the forward
and reverse sequence reads of the Illumina pair-ends were
mapped separately to the Bionano hybrid scaffolds using the
BWA-MEM aligner (Li 2013). The chimeric 3′ side of the aligned
reads that crossed the ligation junctions was removed, and the
reads that were mapped from both ends with the mapping quality
score higher than 10 were retained. PCR duplicates in the Hi-C se-
quences were identified and removed using the Picard Toolkit
(http://broadinstitute.github.io/picard/). After these filtering
steps, ∼300 million paired read alignments were retained, includ-
ing ∼79 million interscaffold alignments. The paired-read align-
ment data were then processed with SALSA (Ghurye et al. 2019)
which broke 251 chimeric scaffolds and merged the polished
Bionano assembly scaffolds into 1,663 Hi-C scaffolds with a total
length of ∼2.83 Gb and N50 of ∼37.9 Mb.

Linkage information from the new genetic maps was used to
generate chromosome sequences from the Hi-C assembly scaf-
folds. The program NovoAlign (http://www.novocraft.com/
products/novoalign/) was used to map the flanking sequences of
each SNP marker to the scaffolds, and the genetic linkage infor-
mation was used to anchor the scaffolds to chromosomes and
then to order, orient, and concatenate the scaffolds into 27 chro-
mosomes. With the guidance of the linkage information, 56 Hi-C
scaffolds were broken into 150 smaller scaffolds at 94 Hi-C join
sites due to incorrect ordering or wrong orientation of the joined
scaffolds, or due to the insertion of a smaller scaffold at the join
site of the larger scaffold.
Alignment with the assembly of the European
salmon genome
To align the chromosome sequences from the new assembly
(GCA_021399835.1) to the chromosome sequences of the
European Atlantic salmon genome assembly, repeat masked
chromosome sequences of assembly Ssal_v3.1 (GCA_
905237065.2) were used as the reference, and Blastn was used as
the aligner with the parameters -evalue 1.0e−100, -word_size 51.
All alignments longer than 2,000 bp were selected and plotted
with the plot function of the program R.
Results and discussion
New SNP database for N.A. Atlantic salmon
A new SNP database of ∼3.1 million SNPs mapped to contigs from
an earlier draft of the new N.A. Atlantic salmon genome assembly
(GCA_021399835.1) was generated using GATK v3.8.1 (Poplin et al.
2018) as described in the methods. Of those SNPs, ∼2.3M can be
uniquely mapped to the chromosome sequences in the current as-
sembly. The new database of ∼2.3M SNPs uniquely mapped to a
single chromosome position is available for download including
the chromosome position and alternate alleles for each SNP
(Supplementary File 1). To generate this database, we mapped
the SNPs in retrospect to the final version of the chromosome se-
quences. We used the Novoalign program (www.Novocraft.com)
to align a 30-bp flanking sequence from each side of the SNP to
find exact matches to the genome assembly chromosome se-
quences. Of the 0.8M SNPs that were not uniquely mapped, 0.2M
were mapped to unplaced scaffolds, 0.4M were mapped to mul-
tiple locations in the chromosomes (MSVs), and 0.2M were dupli-
cated (PSVs) as each allele mapped uniquely to a different position
in the chromosome sequences.
Genotyping and validation with the 50K SNP array
The SNPs were validated with the pedigreed population from the
SJR aquaculture strain that was used to produce the linkage
map (N = 2,512). The marker categories were based on the
Affymetrix Axiom array cluster properties. The total number of
putative SNPs placed on the chip was 49,880. A total of 36,206
SNPs (73%) were categorized as of high quality and polymorphic
and an additional 1,825 SNPs (4%) as of high quality but mono-
morphic, using the default quality filtering of the Affymetrix
SNPolisher software. The number of SNPs passing the quality fil-
tering steps is summarized in Table 1. Among the polymorphic
high-resolution markers, the average and median MAF were 0.32
and 0.34, respectively. The marker distribution by MAF is shown
in Fig. 1. The average and median marker heterozygosity were
40 and 45%, respectively. A high rate of polymorphism is very im-
portant for genome scans to map QTL and for the utilization of
marker information in genomic selection (Vallejo and Leeds
G. Gao et al. | 5
Table 1. Validation of the SNPs with DNA samples from N.A.
Atlantic salmon (N = 2,512). The marker categories are based on
the Affymetrix Axiom array cluster properties. Number of SNP
markers assigned to each category are listed next to the
percentage of total number of markers in parenthese.
Category SNP markers
Total SNPs on the array 49,880 (100%)
Call rate < 97% 5,421 (10.9%)
No minor homozygotes 797 (1.6%)
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
Off-target variant 25 (0.05%)
Other 5,586 (11.2%)
Monomorphic high resolution 1,825 (3.7%)
Hemizygous (mitochondrial) 20 (0.04%)
Polymorphic high resolution 36,206 (72.6%)
et al. 2017; Vallejo and Liu et al. 2017). The high polymorphism of
the validated SNP arrays as measured by average MAF and hetero-
zygosity in the target aquaculture population is substantially bet-
ter than what was reported for similar high-density SNP arrays in
salmonids (Houston et al. 2014; Palti et al. 2015; Yáñez et al. 2016).
Of the 36,206 polymorphic SNPs, 34,302 (95%) had MAF greater
than 0.05 in the SJR aquaculture population that was sampled in
this study, suggesting very efficient utilization of the genotype
data from this array for genomic and genetic analyses in the target
population. In comparison, in the array developed for Chilean
aquaculture populations, the fraction of validated SNPs with
MAF greater than 0.05 in 2 populations of N.A. origin was between
27 and 43% (Yáñez et al. 2016).
COO markers
Forty-one SNPs from the nuclear chromosomes and 20 mitochon-
drial SNPs were shown to be potentially useful for COO differenti-
ation using 60 samples from European origin populations and 14
from N.A. origin. These same samples were part of the panel
that was originally used to test the King 7 microsatellites (King
et al. 2001; King et al. 2005). The 2,512 SJR aquaculture strain fish
of N.A. origin that were used for general validation of the SNP ar-
ray quality and for generating the linkage map were also used to
validate the COO markers. The populations sampled are listed
in Supplementary Table 1, and the markers validated for COO
identification including the allele associated with each COO and
their genome positions are listed in Supplementary Table 2.
The nuclear COO markers were not randomly distributed in chro-
mosomes. The flanking sequences of 39 of the 41 COO nuclear
markers were mapped to 13 of the 27 chromosomes with 7 and
6 markers mapped to chromosomes 12 and 14, respectively
(Supplementary Table 2). This nonrandom distribution may
indicate association of genes on those 2 chromosomes with the
genetic differentiation between the 2 lineages of Atlantic salmon.
However, in contrast to our findings of enriched presence of
COO-associated SNP markers on chromosomes 12 and 14,
Lehnert et al. (Lehnert et al. 2020) identified genomic regions
associated with high differentiation between the 2 lineages on
chromosomes 06, 13, 16, and 19. Therefore, a much more compre-
hensive evaluation of candidate COO SNPs that are distributed on
all chromosomes with a larger panel of samples representing mul-
tiple wild and aquaculture populations from Europe and North
America is needed towards developing a reliable set of COO
SNPs for genotyping assay. However, this type of comprehensive
evaluation of candidate COO SNPs is beyond the proof of concept
we conducted in this study.
6 | G3, 2023, Vol. 00, No. 0
Fig. 1. Polymorphism of SNPs from the Axiom 50K array. MAF distribution among the ∼36K of polymorphic high-resolution SNPs.
One of the N.A. origin fish from Gander—Jonathan’s Brook
(sample NF2-015) had European mitochondria genotypes and
N.A. nuclear genotypes, indicating possible introgression of the
European mitochondrial genome due to ancestral hybridization.
Evidence for introgression of European genotypes into N.A. popu-
lations was previously reported for Atlantic salmon in
Newfoundland (Bradbury et al. 2015). Also, European mtDNA al-
leles have been confirmed in several Newfoundland populations
of Atlantic salmon (Ian Bradbury, personal communication).
The results of the principal component analysis we conducted
using genotypes from the 61 validated COO markers are shown in
Fig. 2. The reference samples were clearly separated into 2 clus-
ters in concordance with their COO (Fig. 2a), except for the single
sample from Gander—Jonathan’s Brook described above due to its
nucleus markers being with N.A. genotypes and mitochondria
markers with European genotypes. As expected, the 2,512 sam-
ples from the aquaculture SJR strain also clustered with the refer-
ence samples from N.A. origin (Fig. 2b). Among the reference
samples, the primary component that differentiated the
European origin salmon from the N.A. origin accounted for 94%
of the total variance, and after adding the 2,512 SJR fish to the ana-
lysis, the primary component still explained ∼83% of the variance.
Sex determination markers
Three markers derived from Atlantic salmon sdY sequences (Kijas
et al. 2018) were placed on the array in duplicates (i.e. 2 Probset IDs
for each SNP ID). The positions of the flanking sequences in the
sdY gene sequence, Probset IDs, and the flanking sequences are
listed in Supplementary Table 3. The alternative alleles for SNP
ID Affx-158802265, Affx-158802258, and Affx-158802275 were C/
T, C/T, and G/T, respectively. Homozygous T alleles correlated
with the absence of the sdY sequence (female), and homozygous
alternative alleles or heterozygous genotypes correlated with
sdY presence (male). To evaluate the accuracy of the sex deter-
mination markers, we genotyped 135 female and 95 male parents
of the families used for the linkage analysis and found 100% con-
cordance of the sex phenotype with the sdY marker genotypes.
Linkage map
A total of 36,158 polymorphic high-resolution markers were
mapped to 27 linkage groups with an average of 1,339 markers
per group (Supplementary File 2). The number of markers per
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
linkage group ranged from 712 to 2,224. The average genetic
length per chromosome in cM was 63 and 91 with a minimum
length of 36 and 60 and maximum length of 129 and 127 for the
male and female maps, respectively. The distribution of markers
in linkage groups labeled to match the nomenclature of the
European Atlantic salmon karyotype is shown in Fig. 3. Similar
to previous reports for Atlantic salmon and rainbow trout (Lien
et al. 2016; Pearse et al. 2019), the pattern of genetic distance in
the paternal map was composed of high levels of recombination
in the telomeric regions and high interference in the centromeric
region and throughout most of the chromosome. Conversely, the
typical maternal pattern of recombination in acrocentric chromo-
somes demonstrated higher rates of recombination near the
centromere and then a steady level of recombination throughout
the rest of the chromosome that gradually decreased to a plateau
in the telomere region. In the metacentric chromosome, the ma-
ternal pattern demonstrated strong recombination interference
in the centromeric region and then a similar pattern to the acro-
centric chromosome in the direction of either telomere whereas
the paternal chromosome was demonstrated to have high levels
of interference except near both telomeres (Fig. 4).
Genome assembly
The genome assembly presented in the present research is based
on DNA from a single male from the USDA St. John River stock of
N.A. Atlantic salmon that was sequenced using PacBio long-read
sequencing technology. The PacBio sequences were assembled
into contigs with Canu, and those contigs were joined into scaf-
folds using Bionano optical mapping and Hi-C proximity ligation
sequence data. Using genetic linkage information, the scaffolds
from the Hi-C assembly were then anchored to and ordered on
chromosomes to generate the chromosome sequences of the
USDA_NASsal_1.1 assembly (GCA_021399835.1).
The Canu sequence assembly was composed of 14,845 contigs
with a total length of ∼3.97 Gb, N50 of 1.45 Mb, and BUSCO score of
95.8% (using actinopterygii_odb9 as the lineage dataset). While
the estimated genome size of Atlantic salmon is ∼3.0 Gb (Lien
et al. 2016), it was clear that heterozygosity in this individual
genome caused duplication of haplotypes for some of the hetero-
zygous genomic regions. After the purging steps removed dupli-
cated haplotypes resulting from heterozygous sequences, the
BUSCO score was improved slightly to 96.0% but the total length
 G. Gao et al. | 7

Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023

Fig. 2. Principal components analysis with the 61 validated COO SNP markers. a) Reference samples of known origin [NA or Europe (EU)]. b) The same
reference samples with the samples from the aquaculture SJR strain fish that were used for the SNP array validation and for generating a genetic linkage
map. The dot in the middle at the bottom of both panels A and B represents 1 of the N.A. origin fish from Gander—Jonathan’s Brook (sample NF2-015),
which had European mitochondria genotypes and N.A. nuclear genotypes, indicating possible introgression of the European mitochondrial genome due
to ancestral hybridization.

of the assembly contigs at ∼3.89 Gb was still greater than the ex-
pected genome size. More detailed statistics for each step in the
assembly pipeline are shown in Table 2. The Bionano hybrid scaf-
folding pipeline used the optical map to join neighboring or over-
lapping Canu contigs and to break chimeric sequence contigs that
were misassembled by Canu. The final Bionano scaffolded assem-
bly included 2,012 scaffolds of joined primary contigs and haplo-
tigs and primary contigs that could not be joined into scaffolds.
After this step, the total length of the assembly was finally re-
duced to within the expected genome size at ∼2.83 Gb and the
N50 increased to 11.9 Mb (Table 2). The Hi-C ligation proximity
mapping further reduced the number of scaffolds to 1,633 and in-
creased the N50 to 37.9 Mb and the BUSCO score to 96.2%, with
negligible impact on the total length of the assembly (Table 2).
Using information from the new linkage map to anchor the
scaffolds to karyotype chromosomes broke some of the largest
scaffolds due to their alignment with different linkage groups.
Overall, 633 scaffolds were anchored to 27 linkage groups with
914 spanned gaps and 606 unspanned gaps to generate 27
chromosome sequences in a total length of ∼2.53 Gb, or 89% of
the total genome assembly length. The distribution of chromo-
somes by physical length in base pairs from this assembly is
shown in comparison with the chromosomes of the European
Atlantic salmon reference genome assembly in Fig. 5. The total
length of the 1,122 unplaced scaffolds that are not anchored to a
chromosome is ∼300 Mb. The total number of scaffolds in the
chromosome-level assembly for the N.A. Atlantic salmon is
therefore 1,755, total length is ∼2.83 Gb, total ungapped sequence
length is ∼2.79 Gb, and scaffold N50 and L50 are 16.2 Mb and 42,
respectively.
A comparison of the assembly statistics between this assembly
and the 2 annotated reference genome assemblies for the
European Atlantic salmon lineage is shown in Supplementary
Table 4. The contiguity of this assembly is much better than the
previous assembly (GCA_000233375.4) (Lien et al. 2016) which
had scaffold N50 and L50 of ∼0.6 Mb and 594, respectively.
However, the contiguity of the new reference for the European
Atlantic salmon lineage genome assembly (GCA_905237065.2)
with scaffold N50 and L50 of 28 Mb and 33, respectively, is better
than the assembly we present here. It also has a nearly perfect
BUSCO score showing 99.3% of complete protein-coding genes de-
tected compared to 96.2% in this genome assembly for the N.A.
Atlantic salmon lineage.
SNP array positions on the chromosome
sequences
The genome positions of the SNPs from the Axiom 50K array that
were mapped to unique positions of the N.A. Atlantic salmon
chromosome sequences with no indels and including 60 bp flank-
ing sequences for each SNP are available in Supplementary File 3.
The distribution of the 50K SNPs on the chromosomes of the gen-
ome assembly is shown in Supplementary Fig. 1. The average
number of SNP markers per chromosome is 1,725 with a min-
imum of 920 and a maximum of 2,885. Overall, the total number
8 | G3, 2023, Vol. 00, No. 0
Fig. 3. Distribution of the linkage map markers in chromosomes. The chromosome axis is in the order of linkage groups from LG1 to LG27. Chromosomes
are labeled to match the nomenclature of the European Atlantic salmon karyotype.
of SNPs per chromosome is highly correlated with the physical
size of the chromosomes (Pearson r = 0.96) in the new genome as-
sembly we generated for N.A. Atlantic salmon. An additional 188
SNPs were mapped to unplaced scaffolds or contigs. The SNP
marker density per 1 Mb in each of the 27 chromosomes is dia-
grammed in Fig. 6. The marker density range is from zero to 40
per 1 Mb. The marker density fluctuates within the chromosomes
with lower density typically near the centromeres and telomeres.
This detailed characterization of marker density on each chromo-
some is important for assessing the utility of the array in genome
scans for QTL and for other genomic analyses. The overall distri-
bution of the array SNPs on chromosomes is highly correlated
with the chromosomes’ length, which is similar to what was re-
ported for high-density arrays that were developed for the
European linage (Houston et al. 2014; Yáñez et al. 2016).
Comparison between European and N.A. genome
assemblies:
Similar to rainbow trout’s plasticity in chromosome number be-
tween populations which correlates with geographic distribution
(Thorgaard 1983; Gao et al. 2021), the Atlantic salmon genome
also exhibits tolerance to chromosomal rearrangements. While
other karyotypes have been reported for N.A. Atlantic salmon
from other populations with chromosome numbers varying
from 2N = 54 to 2N = 58 (see discussion in Brenna-Hansen et al.
2012), here, we found unequivocal evidence that supports 2N =
54 in the SJR strain individual fish that we used for the genome
assembly presented in this report. The results of our linkage ana-
lysis coupled with the de novo assembly of large sequence
scaffolds confirmed the chromosomal rearrangements between
N.A. and European Atlantic salmon proposed previously by
Brenna-Hansen et al. based on their own linkage analysis and kar-
yotyping (Brenna-Hansen et al. 2012).
The chromosome sequences from this assembly of the N.A. sal-
mon genome (USDA_NASsal_1.1; GCA_021399835.1, N = 27) were
aligned with the chromosome sequences of the reference genome
from the European lineage (Ssal_v3.1; GCA_905237065.2, N = 29) to
identify and confirm the chromosomal rearrangements. Graphic
representations of the alignments for all 27 chromosomes can
be found in Supplementary File 4. For 23 of the 27 chromosomes,
we found simple collinear sequence alignments indicating highly
similar order and direction of those chromosome sequences and
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
the genes that they carry as illustrated in Supplementary Fig. 2.
The rearrangements we found in the remaining 4 chromosomes
can be explained by 4 major separate events. The first was fission
in the centromere region of chromosome Ssa01 resulting in
the separation of chromosome arms Ssa01q and Ssa01p. The
second was the translocation and fusion of Ssa01p with Ssa23
resulting in Chr 23 of the N.A. lineage genome assembly.
The evidence from sequence alignments for those 2 chromo-
somal rearrangements is illustrated in Fig. 7. According to
Brenna-Hansen et al. (2012), this is the most common rearrange-
ment between salmon of European and N.A. origins as all the
chromosome karyotypes that they have observed from N.A. sal-
mon were homozygous for the translocation of Ssa01p to Chr 23.
However, more recent studies have documented differences in
the Ssa01/23 translocation in North America in specific geo-
graphic regions where secondary contact between European
and N.A. Atlantic salmon has occurred (Lehnert et al. 2019;
Watson et al. 2022). The third rearrangement event is the forma-
tion of Chr 08 in the N.A. genome from the fusion of the acro-
centric chromosomes Ssa08 and Ssa29 (Fig. 8). The fourth is
the fusion of Ssa26 and Ssa28 to form Chr 26 in the N.A. genome
assembly (Fig. 9). Karyotype preparations and other evidence
from molecular markers revealed that variation in the latter 2
chromosomal fusions (Ssa08/29 and Ssa26/28) can be found in
a wider range of Atlantic salmon populations from the N.A. lin-
eage (Brenna-Hansen et al. 2012; Lehnert et al. 2019; Wellband
et al. 2019). Hence, it is likely that those 2 chromosomal fusion
events are more recent than the translocation event of Ssa01p
to Chr 23 in N.A. Atlantic salmon.
Smaller scale rearrangements in the form of inversions were
recently reported in Atlantic salmon (Stenløkk et al. 2022). Based
on our DNA sequence alignments between the genome assembly
reported here and the European reference genome assembly, we
detected 2 of those inversions and found them to be larger than
what was reported by Stenløkk et al. (2022), with the size of 7
and 4 Mb for the inversions on Ssa09 (just below 100M) and
Ssa18 (just below 80M), respectively (Supplementary File 4). We
were able to confirm that those 2 inversions between the 2 gen-
ome assemblies are likely to be real as they are encompassed
within a single contig in both assemblies, and thus not likely to
be due to scaffolding errors. In addition, there is an optical appear-
ance of a very small inversion on Ssa16 at ∼60 Mb as shown in
 G. Gao et al. | 9

Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023

Fig. 4. Genetic distance is plotted against physical distance to illustrate the differences in the typical recombination patterns between males and females
in salmonids. In the acrocentric chromosome, a) the paternal chromosome displays almost no recombination except for the area proximal to the q-arm
telomere. Conversely, the maternal pattern displays a higher rate of recombination near the centromere, a continuous even level of recombination
throughout the rest of the chromosome, and gradual decrease to a plateau near the telomere. In the metacentric chromosome, b) the maternal
chromosome displays strong recombination interference in the centromeric region and then a similar pattern to the acrocentric chromosome in the
direction of either telomere, while the paternal pattern is consistent with high interference in recombination except for elevated levels of recombination
near both telomeres.

Supplementary File 4. However, this is not a true inversion, but ra-
ther an optical illusion caused by the proximity (∼150 kb) of 2 du-
plicated sequences on both genomes.
Contiguity of scaffolds in homeologous
chromosome arms
The Atlantic salmon genome contain 9 pairs of chromosome arms
with very high sequence similarity due to residual tetraploidy
(Lien et al. 2016). It is nearly impossible to distinguish alleles
from homeologous tetraploid regions, and all salmonid genomes
share this problem (Allendorf et al. 2015). In past de novo genome
assemblies of salmonid genomes that were based on short reads
(Berthelot et al. 2014; Lien et al. 2016; Pearse et al. 2019), the home-
ologous or duplicated chromosome arms were more fragmented
than the rest of the genome because of those nearly identical
stretches of sequence that could not be resolved with short reads.
Hence, the expectation would be that the average number of scaf-
folds in those hard to resolve chromosome arms will be higher
than the rest of the genome. However, the average number of
scaffolds in the 18 duplicated chromosome arms in this assembly
was 15 (Supplementary Table 5) compared to the overall average
of 18 for all chromosome arms in the assembly, suggesting similar
contiguity with the rest of the genome. The total number of scaf-
folds in chromosome sequences in this assembly was 633, and the
overall average was calculated based on the number of 50 chromo-
some arms identified by Lien et al. (2016). Therefore, it appears that
the problem of assembling those homeologous genome regions
can largely overcome with reads that are long enough to cover
10 | G3, 2023, Vol. 00, No. 0

Table 2. Statistics of the N.A. Atlantic salmon genome assembly at each step of the assembly pipeline.

Feature Canu contigs Purge duplicates Bionano scaffolds Hi-C scaffolds Chr. sequencesa

Number of scaffolds/contigs 14,845 14,240 2,012 1,633 1,755

Total length (Mb) 3,970 3,889 2,835 2,834 2,829
Maximum length (kb) 24,768 24,768 70,926 115,072 92,895
Minimum length (bp) 1,003 1,012 3,771 3,768 3,768
N50 (kb) 1,452 1,534 11,936 37,905 16,202
BUSCOb 95.8% 96.0% 95.8% 96.2% 96.2%

Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023

a
Includes the unplaced scaffolds (N = 1,122). The total number of scaffolds mapped to chromosomes was 633. Some of the largest scaffolds were broken by the
linkage map information (e.g. 2 parts were mapped to 2 different chromosomes).
b
BUSCO: Benchmarking Universal Single-Copy Orthologs. Showing the percent of complete protein-coding genes detected (using actinopterygii_odb9 as the
lineage dataset).

Fig. 5. Distribution of chromosome lengths in base pairs in the assembly from N.A. Atlantic salmon (accession GCA_021399835.1) presented in this report
a) vs. b); the distribution in the reference genome assembly from the European lineage (Ssal_v3.1; GCA_905237065.2).

stretches of DNA sequence that contain enough sequence variants
to distinguish between those nearly identical genome regions.
Gene annotation
Gene annotation for this genome assembly accession is not avail-
able currently from NCBI. Chromosome sequences or specific re-
gions of interests from the genome assembly can be aligned by
Blast searches with the Refseq-annotated version of the
European lineage genome assembly (GCF_905237065.1). For ex-
ample, the T cell receptor beta 01 region which is found on
Ssa01p in the European lineage reference genome has been shown
to be translocated to Chr 23 (Ssa01p_Ssa23) in the assembly of the
N.A. Atlantic salmon genome presented here (Grimholt et al.
2022). In addition, the alignments of USDA_NASsal_1.1 (GCA_
 G. Gao et al. | 11

Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023

Fig. 6. Density of markers from the N.A. Atlantic salmon Axiom 50K SNP array per million base pairs throughout the genome assembly chromosomes.

Fig. 7. Chromosome fusions and fissions between the European and N.A. Atlantic salmon lineages. Fission of the European chromosome Ssa01q and p
arms and fusion of Ssa01p with Ssa23 in the N.A. Atlantic salmon genome. Sequences were compared using data from the reference assembly for the
European lineage in NCBI (Ssal_v3.1; GCA_905237065.2).

021399835.1) to the annotated Ssal_v3.1 (GCF_905237065.1) gen-
ome assembly are now available in the NCBI remap menu at
https://www.ncbi.nlm.nih.gov/genome/tools/remap. The USDA
assembly can be used as the source assembly and the Ssal_v3.1
as the target assembly. After pasting or uploading the chromo-
some coordinates of interest from the source assembly, the user
can choose the output format of interest. The output file in gen-
ome workbench format can be used to view the aligned chromo-
some region or regions of choice in a genome browser. In
addition, a first pass gene annotation can be downloaded from
Ensembl Rapid Release at https://rapid.ensembl.org/Salmo_salar_
GCA_021399835.1/Info/Index.
12 | G3, 2023, Vol. 00, No. 0
Fig. 8. Fusion of the European lineage chromosomes Ssa08 and Ssa29. Sequences were compared using data from the reference assembly for the
European lineage in NCBI (Ssal_v3.1; GCA_905237065.2).
Fig. 9. Fusion of the European lineage chromosomes Ssa26 and Ssa28. Sequences were compared using data from the reference assembly for the
European lineage in NCBI (Ssal_v3.1; GCA_905237065.2).
Location of sdY in the genome assembly
The sequence of the sex determination gene (sdY) can be mapped
by sequence alignment to the last scaffold on the telomeric
end of chromosome 3 in the reverse strand between positions
104,938,219–104,933,632. Hence, like the current European salmon
reference genome assembly (GCA_905237065.2), this genome as-
sembly also provides a reference to the sdY lineage located on chro-
mosomes 3 or 6, in contrast to the sdY lineage located on
chromosome 2 (Kijas et al. 2018).
Conclusions
We report here on the de novo assembly of the first chromosome-
level reference genome and the development and validation of
the first publicly available high-density SNP array for the N.A.
lineage of Atlantic salmon. These genomic resources are re-
quired for implementing genomic selection for aquaculture pro-
duction and are crucial to current research on the genetics,
biology, evolution, and ecology of this subspecies that holds
high economic and social value in Eastern United States,
Canada, Chile, and Australia.
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
Data availability
The reference genome is available for downloading and browsing
at NCBI (GenBank: GCA_021399835.1; BioProject: PRJNA759632).
The PacBio (wgs) and the Illumina (wgs and Hi-C) raw sequence
data used for the genome assembly are in NCBI-SRA (Accessions
SRX17524033, SRX17524034, and SRX17524035). The new SNP dis-
covery dataset, the linkage map, and the genome positions of the
SNPs from the 50K Axiom array are available as supplemental
files. Supplemental Material available at figshare: https://
doi.org/10.25387/g3.22540894.
Acknowledgements
We thank Kristy Shewbridge for her technical assistance with the
preparation of the DNA sample used for genome sequencing. We
thank Dr. Serap Gonen and Dr. Matthew Baranski from Mowi for
sharing SNP polymorphism information from the European SNP
array and for contributing samples of farmed salmon from
European origin. We thank Dr. Matthew Kent for sharing the
flanking sequence data of the markers used in the linkage map
published by Brenna-Hansen et al. (2012). Mention of trade names
or commercial products in this publication is solely for the

purpose of providing specific information and does not imply
recommendation or endorsement by the U.S. Department of
Agriculture. USDA is an equal opportunity provider and employer.
Funding
This study was supported by the USDA Agricultural Research
Service in-house project numbers 8030-31000-004/005 and
8082-31000-012/013 and 6066-31000-016.
Conflicts of interest statement
The authors declare no conflict of interest.
Authors’ contributions
Guangtu Gao: writing—original draft, conceptualization, method-
ology, formal analysis, data curation, and writing—review & edit-
ing. Geoffrey C. Waldbieser: investigation, methodology,
resources, and writing—review & editing. Ramey C. Youngblood:
investigation, methodology, and formal analysis. Dongyan Zhao:
writing—original draft, methodology, and formal analysis.
Michael R. Pietrak: project administration, investigation, and val-
idation. Melissa S. Allen: validation and data curation. Jason
A. Stannard: project administration, methodology, and validation.
John T. Buchanan: methodology and supervision. Roseanna
L. Long: investigation and validation. Melissa Milligan: investiga-
tion and validation. Gary Burr: investigation and validation.
Katherine Mejía-Guerra: formal analysis. Moira J. Sheehan: re-
sources and supervision. Brian E. Scheffler: resources and supervi-
sion. Caird E. Rexroad III: conceptualization and resources. Brian
C. Peterson: conceptualization, supervision, and resources. Yniv
Palti: writing—original draft, conceptualization, methodology,
supervision, project administration, resources, and writing—
review & editing. All authors read and approved the manuscript.
Literature cited
Allendorf FW, Bassham S, Cresko WA, Limborg MT, Seeb LW, Seeb JE.
Effects of crossovers between homeologs on inheritance and
population genomics in polyploid-derived salmonid fishes. J
Hered. 2015;106(3):217–227. doi:10.1093/jhered/esv015.
Allendorf FW, Thorgaard GH. Tetraploidy and the evolution of sal-
monid fishes., pp. 1–46. In: Turner BJ, editors. Evolutionary
Genetics of Fishes. New York: Plenum Press; 1984.
Berthelot C, Brunet F, Chalopin D, Juanchich A, Bernard M, Noël 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):
3657. doi:10.1038/ncomms4657.
Bradbury IR, Hamilton LC, Dempson B, Robertson MJ, Bourret V,
Bernatchez L, Verspoor E. Transatlantic secondary contact in
Atlantic salmon, comparing microsatellites, a single nucleotide
polymorphism array and restriction-site associated DNA sequen-
cing for the resolution of complex spatial structure. Mol Ecol.
2015;24(20):5130–5144. doi:10.1111/mec.13395.
Brenna-Hansen S, Li J, Kent MP, Boulding EG, Dominik S, Davidson
WS, Lien S. Chromosomal differences between European and
North American Atlantic salmon discovered by linkage mapping
and supported by fluorescence in situ hybridization analysis.
BMC Genomics. 2012;13(1):432. doi:10.1186/1471-2164-13-432.
G. Gao et al. | 13
Council NR. Genetic Status of Atlantic Salmon in Maine: Interim
Report from the Committee on Atlantic Salmon in Maine.
Washington (DC): The National Academies Press; 2002.
Council NR. Atlantic Salmon in Maine. Washington (DC): The
National Academies Press; 2004.
Gao G, Magadan S, Waldbieser GC, Youngblood RC, Wheeler PA,
Scheffler BE, Thorgaard GH, Palti Y. A long reads-based de-novo
assembly of the genome of the Arlee homozygous line reveals
chromosomal rearrangements in rainbow trout. G3 (Bethesda).
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
2021;11(4):jkab052. doi:10.1093/g3journal/jkab052.
Gao G, Pietrak MR, Burr GS, Rexroad CE, Peterson BC, Palti Y. A new
single nucleotide polymorphism database for North American
Atlantic salmon generated through whole genome resequencing.
Front Genet. 2020;11:85. doi:10.3389/fgene.2020.00085.
Gao G, Waldbieser GC, Youngblood RC, Zhao D, Pietrak M, Allen MS,
Stannard JA, Buchanan JT, Long RL, Milligan M, et al. The generation
of the first chromosome-level de-novo genome assembly and the
development and validation of a 50K SNP array for North
American Atlantic salmon. bioRxiv: 2022.09.28.509896. https://doi.
org/10.1101/2022.09.28.509896, 2022. preprint: not peer reviewed.
Garrison E, Marth G. Haplotype-based variant detection from short-
read sequencing. arXiv:1207.3907v2. https://doi.org/10.48550/
arXiv.1207.3907, 2012. preprint: not peer reviewed.
Ghurye J, Rhie A, Walenz BP, Schmitt A, Selvaraj S, Pop M, Phillippy
AM, Koren S. Integrating Hi-C links with assembly graphs for
chromosome-scale assembly. PLoS Comput Biol. 2019;15(8):
e1007273. doi:10.1371/journal.pcbi.1007273.
Gonzalez-Pena D, Gao G, Baranski M, Moen T, Cleveland BM, Kenney
PB, Vallejo RL, Palti Y, Leeds TD. Genome-wide association study
for identifying loci that affect fillet yield, carcass, and body
weight traits in rainbow trout (Oncorhynchus mykiss). Front
Genet. 2016;7(203):203. doi:10.3389/fgene.2016.00203.
Grimholt U, Sundaram AYM, Bøe CA, Dahle MK, Lukacs M.
Tetraploid ancestry provided Atlantic salmon with two paralogue
functional T cell receptor beta regions whereof one is completely
novel. Front Immunol. 2022;13:930312. doi:10.3389/fimmu.2022.
930312.
Guan D, McCarthy SA, Wood J, Howe K, Wang Y, Durbin R.
Identifying and removing haplotypic duplication in primary gen-
ome assemblies. Bioinformatics. 2020;36(9):2896–2898. doi:10.
1093/bioinformatics/btaa025.
Hardie DC, Hebert PDN. The nucleotypic effects of cellular DNA con-
tent in cartilaginous and ray-finned fishes. Genome. 2003;46(4):
683–706. doi:10.1139/g03-040.
Houston RD, Taggart JB, Cézard T, Bekaert M, Lowe NR, Downing A,
Talbot R, Bishop SC, Archibald AL, Bron JE, et al. Development
and validation of a high density SNP genotyping array for
Atlantic salmon (Salmo salar). BMC Genomics. 2014;15:90. doi:10.
1186/1471-2164-15-90.
Kijas J, Elliot N, Kube P, Evans B, Botwright N, King H, Primmer CR,
Verbyla K. Diversity and linkage disequilibrium in farmed
Tasmanian Atlantic salmon. Anim Genet. 2017;48(2):237–241.
doi:10.1111/age.12513.
Kijas J, McWilliam S, Naval Sanchez M, Kube P, King H, Evans B,
Nome T, Lien S, Verbyla K. Evolution of sex determination loci
in Atlantic salmon. Sci Rep. 2018;8(1):5664. doi:10.1038/s41598-
018-23984-1.
King T, Eackles M, Letcher B. Microsatellite DNA markers for the
study of Atlantic salmon (Salmo salar) kinship, population struc-
ture, and mixed-fishery analyses. Mol Ecol Notes. 2005;5(1):
130–132. doi:10.1111/j.1471-8286.2005.00860.x.
King TL, Kalinowski ST, Schill WB, Spidle AP, Lubinski BA. Population
structure of Atlantic salmon (Salmo salar L.): a range-wide
14 | G3, 2023, Vol. 00, No. 0
perspective from microsatellite DNA variation. Mol Ecol. 2001;
10(4):807–821. doi:10.1046/j.1365-294X.2001.01231.x.
Koren S, Walenz BP, Berlin K, Miller JR, Bergman NH, Phillippy AM.
Canu: scalable and accurate long-read assembly via adaptive
k-mer weighting and repeat separation. Genome Res. 2017;
27(5):722–736. doi:10.1101/gr.215087.116.
Lehnert SJ, Bentzen P, Kess T, Lien S, Horne JB, Clément M, Bradbury
IR. Chromosome polymorphisms track trans-Atlantic divergence
and secondary contact in Atlantic salmon. Mol Ecol. 2019;28(8):
2074–2087. doi:10.1111/mec.15065.
Lehnert SJ, Kess T, Bentzen P, Clément M, Bradbury IR. Divergent and
linked selection shape patterns of genomic differentiation be-
tween European and North American Atlantic salmon (Salmo sal-
ar). Mol Ecol. 2020;29(12):2160–2175. doi:10.1111/mec.15480.
Li H. Aligning sequence reads, clone sequences and assembly contigs
with BWA-MEM. arXiv:1303.3997. https://doi.org/10.48550/arXiv.
1303.3997, 2013. preprint: not peer reviewed.
Li H. Minimap2: pairwise alignment for nucleotide sequences.
Bioinformatics. 2018;34(18):3094–3100. doi:10.1093/bioinformatics/
bty191.
Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G,
Abecasis G, Durbin R. The sequence alignment/map format and
SAMtools. Bioinformatics. 2009;25(16):2078–2079. doi:10.1093/
bioinformatics/btp352.
Lien S, Koop BF, Sandve SR, Miller JR, Kent MP, Nome T, Hvidsten TR,
Leong JS, Minkley DR, Zimin A, et al. The Atlantic salmon genome
provides insights into rediploidization. Nature. 2016;533(7602):
200–205. doi:10.1038/nature17164.
Liu L, Ang KP, Elliott JAK, Kent MP, Lien S, MacDonald D, Boulding EG.
A genome scan for selection signatures comparing farmed
Atlantic salmon with two wild populations: testing colocalization
among outlier markers, candidate genes, and quantitative trait
loci for production traits. Evol Appl. 2017;10(3):276–296. doi:10.
1111/eva.12450.
Lubieniecki KP, Jones SL, Davidson EA, Park J, Koop BF, Walker S,
Davidson WS. Comparative genomic analysis of Atlantic salmon,
Salmo salar, from Europe and North America. BMC Genet. 2010;
11(1):105. doi:10.1186/1471-2156-11-105.
Meuwissen T, Hayes B, Goddard M. Genomic selection: a paradigm
shift in animal breeding. Anim Front. 2016;6(1):6–14. doi:10.
2527/af.2016-0002.
Nash CE, Waknitz FW. Interactions of Atlantic salmon in the Pacific
Northwest: I. Salmon enhancement and the net-pen farming indus-
try. Fish Res. 2003;62(3):237–254. doi:10.1016/S0165-7836(03)00063-8.
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 Res. 2015;15(3):
662–672. doi:10.1111/1755-0998.12337.
Pearse DE, Barson NJ, Nome T, Gao G, Campbell MA, Abadía-Cardoso
A, Anderson EC, Rundio DE, Williams TH, Naish KA, et al.
Sex-dependent dominance maintains migration supergene in
rainbow trout. Nat Ecol Evol. 2019;3(12):1731–1742. doi:10.1038/
s41559-019-1044-6.
Poplin R, Ruano-Rubio V, DePristo MA, Fennell TJ, Carneiro MO, Van
der Auwera GA, Kling DE, Gauthier LD, Levy-Moonshine A,
Roazen D, et al. Scaling accurate genetic variant discovery to
tens of thousands of samples. bioRxiv:201178. https://doi.org/
10.1101/201178. 2018. preprint: not peer reviewed.
Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender
D, Maller J, Sklar P, de Bakker PIW, Daly MJ, et al. PLINK: a toolset
for whole-genome association and population-based linkage
analysis. Am J Hum Genet. 2007;81(3):559–575. doi:10.1086/
519795.
Rastas P. Lep-MAP3: robust linkage mapping even for low-coverage
Downloaded from https://academic.oup.com/g3journal/advance-article/doi/10.1093/g3journal/jkad138/7202260 by guest on 28 August 2023
whole genome sequencing data. Bioinformatics. 2017;33(23):
3726–3732. doi:10.1093/bioinformatics/btx494.
Roach MJ, Schmidt SA, Borneman AR. Purge haplotigs: allelic contig
reassignment for third-gen diploid genome assemblies. BMC
Bioinformatics. 2018;19(1):460. doi:10.1186/s12859-018-2485-7.
Stenløkk K, Saitou M, Rud-Johansen L, Nome T, Moser M, Árnyasi M,
Kent M, Barson NJ, Lien S. The emergence of supergenes from in-
versions in Atlantic salmon. Philos Trans R Soc B Biol Sci . 2022;
377(1856):20210195. doi:10.1098/rstb.2021.0195.
Thorgaard GH. Chromosomal differences among rainbow trout po-
pulations. Copeia. 1983;3:650–662. doi:10.2307/1444329.
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.
2017a;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. 2017b;8:156. doi:10.3389/
fgene.2017.00156.
Watson KB, Lehnert SJ, Bentzen P, Kess T, Einfeldt A, Duffy S,
Perriman B, Lien S, Kent M, Bradbury IR. Environmentally asso-
ciated chromosomal structural variation influences fine-scale
population structure of Atlantic salmon (Salmo salar). Mol Ecol.
2022;31(4):1057–1075. doi:10.1111/mec.16307.
Wellband K, Mérot C, Linnansaari T, Elliott JAK, Curry RA,
Bernatchez L. Chromosomal fusion and life history-associated
genomic variation contribute to within-river local adaptation of
Atlantic salmon. Mol Ecol. 2019;28(6):1439–1459. doi:10.1111/
mec.14965.
Yáñez JM, Naswa S, López ME, Bassini L, Correa K, Gilbey J,
Bernatchez L, Norris A, Neira R, Lhorente JP, et al. Genomewide
single nucleotide polymorphism discovery in Atlantic salmon
(Salmo salar): validation in wild and farmed American and
European populations. Mol Ecol Resour. 2016;16(4):1002–1011.
doi:10.1111/1755-0998.12503.
Zhang Z, Schwartz S, Wagner L, Miller WA. A greedy algorithm for
aligning DNA sequences. J Comput Biol. 2000;7(1–2):203–214.
doi:10.1089/10665270050081478.
Editor: J. Yáñez