Near telomere-to-telomere genome assembly of the rainbow trout (Oncorhynchus mykiss)

Abstract

The rainbow trout (Oncorhynchus mykiss) exhibits extensive karyotypic diversity (2n = 58–64) driven by Robertsonian translocations, yet widely used reference genomes are derived from North American lineages, leaving Chinese aquaculture populations underrepresented. Here, we present a near telomere-to-telomere (T2T) genome assembly of a farmed rainbow trout from China. Integrating PacBio HiFi, ONT ultra-long reads, and Hi-C data, we assembled a 2.29 Gb genome with 99.04% anchored to 30 chromosomes. Notably, the genome contains only 18 gaps, with 16 gap-free chromosomes and 13 achieving T2T status. Comparative synteny analysis revealed a third chromosomal fission/fusion iteration in which Swanson Omy14 splits into Arlee Omy14 and Omy32. Annotation identified 43,137 protein-coding genes, with a BUSCO completeness of 98.9%. This dataset provides a valuable resource for resolving lineage-specific structural variation, supporting pangenome construction and facilitating molecular breeding in rainbow trout.

Background & Summary

Oncorhynchus mykiss belongs to the family Salmonidae and has a complex genome shaped by the salmonid-specific fourth round of whole-genome duplication (Ss4R)¹. This event, coupled with the proliferation of transposable elements (TEs), has resulted in a genome characterized by high repetitive content and retained paralogous regions derived from delayed rediploidization^1,2. In addition, the rainbow trout genome is relatively large, with an estimated size of approximately 2.4 Gb³, about 2–3 times larger than many common fish species, such as Oreochromis niloticus⁴, Ictalurus punctatus⁵, and Micropterus salmoides⁶. These genomic complexities pose substantial challenges for de novo genome assemblies.

Rainbow trout are native to the Pacific coast of North America and the Kamchatka Peninsula in Russia^7,8. Many domesticated rainbow trout used for aquaculture worldwide are thought to trace to hatchery propagation and stock transfers initiated in western North America in the late nineteenth century^8,9. Rainbow trout were first introduced into China from North Korea in 1959, but the exact origin of this initial introduction remains unclear¹⁰. Subsequently, rainbow trout aquaculture in China involved additional introductions and germplasm exchanges from different sources, likely contributing to complex genetic backgrounds among Chinese farmed populations. These features highlight the need for genomic resources that better represent Chinese aquaculture germplasm.

Rainbow trout exhibit remarkable karyotypic diversity (2n = 58–64) across geographic populations due to Robertsonian translocations, while the number of chromosome arms remains stable (NF = 104)¹¹. The Swanson (2n = 58) and Arlee (2n = 64) assemblies differ in three major chromosomal fission/fusion events, involving Swanson chromosomes Omy04, Omy14 and Omy25^12,13. Two additional 2n = 60 de novo assemblies, from the Hayspur hatchery strain in Idaho and the interior redband rainbow trout from Keithley Creek, correspond to two of these iterations: the Hayspur genome reflects the fission of Swanson Omy25 into Arlee Omy25 and Omy31, and the Keithley Creek assembly the fission of Swanson Omy04 into Arlee Omy04 and Omy30^14,15. Beyond this karyotypic variation, many current assemblies still contain unresolved gaps in telomeric and other complex regions, limiting the characterization of these evolutionarily dynamic regions. A high-quality assembly is therefore still needed to improve the resolution of these previously unassembled complex regions.

In this study, we integrated PacBio HiFi, ONT ultra-long reads and Hi-C data to generate a near-T2T genome assembly of a farmed rainbow trout from China. The resulting assembly substantially improves chromosome-scale continuity over existing rainbow trout references. Comparative synteny analysis revealed a third 30-chromosome fission/fusion iteration, in which Swanson Omy14 corresponds to Arlee Omy14 and Omy32, distinct from the previously reported Hayspur and Keithley Creek iterations. This assembly provides a valuable resource for comparative genomics, structural variation analysis, pangenome construction, germplasm evaluation and molecular breeding in rainbow trout.

Methods

Sample collection and sequencing

All experimental procedures were conducted in accordance with institutional animal care and use guidelines, with approval from the Institutional Animal Care and Use Committee (IACUC) of the Yellow Sea Fisheries Research Institute, Chinese Academy of Fishery Sciences (YSFRI-2026075). A healthy adult male rainbow trout introduced from Benxi, Liaoning Province, was collected from Kunming Tanghao Aquaculture Co., Ltd. in Yunnan, China. To ensure genomic uniformity, ten distinct tissues were harvested from a single individual, including muscle, liver, kidney, spleen, fin, heart, brain, gill, testis, and intestine. Samples were flash-frozen in liquid nitrogen immediately upon collection and stored at −80 °C to preserve nucleic acid integrity. To construct a high-quality reference genome, we integrated data from PacBio Revio, Nanopore PromethION, and MGI DNBSEQ-T7 platforms.

For HiFi sequencing, muscle tissue was selected as the DNA source for its high yield of high-quality genomic DNA (gDNA) and low contamination, both favorable for long-read library preparation. gDNA was extracted using the NucleoBond HMW DNA Kit (Macherey-Nagel), with purity, concentration, and integrity assessed by NanoDrop, Qubit fluorometry, and Agilent 4200 TapeStation, respectively. The gDNA was sheared and used to construct PCR-free SMRTbell libraries, which were sequenced on the PacBio Revio platform. CCS reads with QV ≥ 20 were retained, yielding 173.86 Gb (~76×) of HiFi data (Table 1).

Table 1.

Statistics of the sequencing data.

Type	Platform	Tissue	Data size (Gb)	Average depth (×)	Average Length (bp)
WGS	MGI DNBSEQ T7	Muscle	166.33	73	150
Hi-C	MGI DNBSEQ T7	Liver	187.84	82	150
HiFi	PacBio Revio	Muscle	173.86	76	18,179
ONT ultra-long	Nanopore PromethION	Muscle, Liver	98.30	43	92,786
RNA-Seq	MGI DNBSEQ T7	Heart	6.84	—	150
RNA-Seq	MGI DNBSEQ T7	Kidney	5.76	—	150
RNA-Seq	MGI DNBSEQ T7	Spleen	6.55	—	150
RNA-Seq	MGI DNBSEQ T7	Brain	6.49	—	150
RNA-Seq	MGI DNBSEQ T7	Gill	6.63	—	150
RNA-Seq	MGI DNBSEQ T7	Fin	6.17	—	150
RNA-Seq	MGI DNBSEQ T7	Testis	6.40	—	150
RNA-Seq	MGI DNBSEQ T7	Intestine	6.84	—	150
RNA-Seq	MGI DNBSEQ T7	Muscle	6.28	—	150
RNA-Seq	MGI DNBSEQ T7	Liver	6.72	—	150

For WGS sequencing, paired-end libraries (~350 bp insert size) were constructed from the same gDNA preparation used for HiFi sequencing following standard MGI protocols, and sequenced on the MGI DNBSEQ-T7 platform (PE150). Raw reads were filtered with fastp (v0.23.0) using ‘-q 19 -u 50 -n 5’, yielding 166.33 Gb (~73×) of clean data for genome assembly quality evaluation (Table 1).

For ONT ultra-long sequencing, gDNA was independently extracted from muscle and liver tissues to maximize long-read yield. gDNA was extracted using a sodium dodecyl sulfate (SDS)-based method and assessed by agarose gel electrophoresis, NanoDrop, and Qubit fluorometry. Libraries were constructed using the SQK-LSK114 kit (Oxford Nanopore Technologies) and sequenced on the Oxford Nanopore Technologies PromethION platform with FLO-PRO114M (R10.4.1) flow cells. Reads with mean Q-score ≤ 9 were filtered out, and reads from all libraries were combined, yielding 98.30 Gb (~43×) of data (Table 1).

For Hi-C sequencing, liver tissue was selected for its efficient formaldehyde cross-linking and chromatin recovery. Cross-linked chromatin from the same individual was digested with DpnII, biotin-labeled at the cleavage sites, proximity-ligated, and purified. The recovered DNA was sheared, biotin-enriched, and processed for adapter ligation to construct paired-end libraries, with library quality assessed by Qubit and Agilent 2100 Bioanalyzer. Libraries were sequenced on the MGI DNBSEQ-T7 platform (PE150), yielding 187.84 Gb (~82×) of clean data for anchoring (Table 1).

For RNA-seq, transcriptome sequencing was performed on ten tissues to support gene prediction and functional annotation (Table 1). Total RNA was extracted using the MagPure Universal RNA Kit (Magen Biotech), with purity, concentration, and integrity assessed by NanoDrop 2000 and LabChip GX Touch. Poly(A)-enriched mRNA libraries were constructed using the MGIEasy RNA Library Prep Set (MGI Tech) and assessed by Qubit 3.0, Agilent 2100 Bioanalyzer, and qPCR. Qualified libraries were sequenced on the MGI DNBSEQ-T7 platform (PE150), and raw reads were filtered with fastp, yielding an average of 6.47 Gb of clean data per tissue (Table 1).

Genome assembly

A comprehensive strategy was employed to construct the near-T2T genome. Initial assembly was performed with hifiasm¹⁶ (v0.25.0) with the parameter ‘-u 0 --telo-m CCCTAA’ by jointly integrating PacBio HiFi, ONT and Hi-C data. The primary contig set (p_ctg) was used as the main backbone for the T2T consensus assembly. Potential errors in the primary contigs were corrected using CRAQ¹⁷ (v1.0.9) with HiFi and WGS data by splitting chimeric fragments based on clipping signals, followed by redundancy removal via KmerDedup¹⁸ with the parameter ‘-mpr 0.5’.

Subsequently, HapHiC¹⁹ (v1.0.6) was used to anchor contigs to 30 chromosomes with Hi-C data, consistent with the previously reported 2n = 60 karyotypes of rainbow trout²⁰ (Fig. 1A). To ensure representation of the male-specific locus in the final assembly, the haplotype-resolved contigs generated by hifiasm were additionally examined. These contigs were independently scaffolded using HapHiC, and screened for sdY using BLASTN²¹ (v2.16.0) against the published O. mykiss sdY coding sequence. An sdY-bearing scaffold was identified, with support from Hi-C contact patterns. This sdY-bearing scaffold was incorporated into the final assembly in place of the corresponding primary scaffold.

Fig. 1.

Hi-C interaction heatmap and snail plot of rainbow trout genome. (A) The heatmap illustrates the interaction frequencies among 30 chromosomes. Chromosomes are represented by distinct squares along the diagonal. The color scale indicates interaction strength from purple (low) to yellow (high). (B) Snail plot summarizing scaffold count, scaffold length, N50 and N90 lengths, BUSCO completeness and GC/AT content.

All scaffolds were manually curated in Juicebox²² (v2.15.0). Then, we filled gaps using Quartet²³ (v1.1.1) with ONT and HiFi reads, and polished the assembly using NextPolish2²⁴ (v0.2.1) with HiFi and WGS data. The final genome size was 2.29 Gb with a scaffold N50 of 86.07 Mb (Fig. 1B and Table 2). Comparative analysis with previously published rainbow trout assemblies showed high continuity, with 99.04% of sequences anchored to chromosomes and only 18 gaps remaining across the genome, including 16 gap-free chromosomes (Table 2).

Table 2.

Assembly statistics of Rainbow trout.

Assembly	USDA_OmykA_1.1	Omyk_2.0	This Study
Total length	2.33 Gb	2.32 Gb	2.29 Gb
Number of chromosomes	32	29	30
Anchoring rate	95.37%	93.62%	99.04%
Scaffold N50	39.16 Mb	52.39 Mb	86.07 Mb
BUSCO completion rate	99.89%	98.92%	99.72%
Number of gaps (chromosomes)	438	491	18
Number of gap-free chromosomes	0	0	16
Number of telomeres (chromosomes)	27	29	53

To facilitate comparative analyses across rainbow trout assemblies, chromosome names were reassigned according to the established rainbow trout nomenclature convention based on whole-genome syntenic relationships with the Arlee and Swanson assemblies^12,15. Accordingly, the 30 chromosomes in this study assembly were named Omy01–Omy29 and Omy32. Under this nomenclature, a full-length sdY coding sequence was identified on Omy29 by BLASTN alignment.

Telomere and candidate centromere identification

Telomeres were systematically identified using the TeloExplorer module of QuarTeT²³ (v1.1.1). The parameter -c animal was employed to screen for the canonical vertebrate motif (TTAGGG/CCCTAA) with a minimum of 100 repeats. The analysis identified ‘AACCCT’ as the specific motif; notably, 24 chromosomes possessed telomeric sequences at both ends, while the remaining 6 exhibited them at a single end (Fig. 2 and Table 3). Notably, 13 chromosomes were gap-free and flanked by telomeric repeats at both ends, indicating that the assembly has reached a near telomere-to-telomere level (Table 2 and Table 3).

Fig. 2.

Circos plot of the Rainbow trout genome assembly. Chromosomes (arranged clockwise) are displayed with tracks for DNA transposons, LINEs, SINEs, LTRs, gene density, and GC content. Gray infill indicates gapless chromosomes, and black dots mark telomeric repeats on the chromosome backbone.

Table 3.

Assembly statistics of chromosomes.

Chromosome	Number	Length (bp)	Telomere count	Gap count	Candidate Centromere Region (bp)
Omy01	1	91,198,505	1	1	58,179,228–58,280,571
Omy02	2	105,143,900	2	1	44,090,061–44,243,632
Omy03	3	97,901,038	2	1	45,431,080–45,497,992
Omy04	4	96,693,822	2	0	48,398,376–49,429,577
Omy05	5	98,439,701	2	1	52,638,904–53,030,215
Omy06	6	86,070,840	1	0	835,260–954,406
Omy07	7	86,081,750	2	1	24,243,462–24,493,010
Omy08	8	93,834,071	2	0	42,081,270–42,148,484
Omy09	9	85,079,425	2	2	37,417,792–37,736,271
Omy10	10	89,517,026	2	1	53,746,840–54,156,239
Omy11	11	88,491,901	2	0	39,260,023–40,059,741
Omy12	12	94,940,302	1	1	N/A
Omy13	13	85,821,800	2	1	N/A
Omy14	14	43,912,043	2	0	1,654–364,722
Omy15	15	86,670,873	2	1	40,942,789–41,121,839
Omy16	16	81,810,240	2	0	29,474,326–30,481,011
Omy17	17	79,384,037	1	0	26,769,670–27,013,632
Omy18	18	83,070,729	2	1	6,349,456–6,602,464
Omy19	19	76,573,737	2	1	32,107,109–32,261,142
Omy20	20	51,449,739	1	3	N/A
Omy21	21	59,054,681	1	0	25,377,433–26,317,128
Omy22	22	59,009,247	2	2	N/A
Omy23	23	64,933,454	2	0	61,259,727–61,340,125
Omy24	24	47,224,275	2	0	40,098,387–40,188,631
Omy25	25	94,296,089	2	0	44,562,798–45,062,563
Omy26	26	53,084,125	2	0	24,040,670–24,183,025
Omy27	27	52,905,079	2	0	48,739,472–48,799,076
Omy28	28	45,670,262	2	0	41,081,750–41,226,940
Omy29	29	48,245,569	2	0	41,509,904–41,645,162
Omy32	30	42,417,835	2	0	911,702–1,078,631
Unplaced	—	22,022,562	—	—	—

Candidate centromere regions were predicted with reference to the method described for Ctenopharyngodon idella²⁵. Tandem repeat arrays were identified using TRF²⁶ (v4.10) with the parameters ‘2 7 7 80 10 50 500’. The TRF output was then processed with BSLtool²⁷ for monomer decomposition, alignment and clustering to generate a species-specific satellite monomer library. This library was used as a custom repeat library in RepeatMasker²⁸ (v4.1.7) to identify the genomic distribution of satellite monomers. RepeatMasker annotations were intersected with TRF-derived tandem repeat arrays using BEDTools²⁹ (v2.31.1). Adjacent records assigned to the same monomer within 2 kb were merged, and merged satellite-rich blocks with a length of ≥50 kb were retained as candidate regions. Representative candidate centromeric repeat regions were selected based on satellite enrichment and inspection of Hi-C contact maps (Table 3). Functional centromere positions remain to be experimentally validated.

Genomic synteny and structural variant analysis

To assess genome-wide synteny and chromosomal organization, our rainbow trout assembly was aligned against two representative rainbow trout reference genomes, Arlee (GCF_013265735.2) and Swanson¹³ (GCA_025558465.1), using Minimap2³⁰ (v2.17) with the asm5 preset. The alignments were visualized as dot plots using pafCoordsDotPlotly.R (https://github.com/tpoorten/dotPlotly) to examine chromosome-scale collinearity and fission/fusion differences. The assembly showed high synteny with both references, while also revealing distinct fission/fusion patterns. Compared with the Arlee genome, Arlee OmyA25/OmyA31 corresponded to Omy25 in our assembly, and Arlee OmyA04/OmyA30 corresponded to Omy04 (Fig. 3A). Compared with the Swanson genome, Swanson Omy14 corresponded to two chromosomes in our assembly, Omy14 and Omy32 (Fig. 3B). This pattern is consistent with a third 30-chromosome fission/fusion iteration, in which Swanson Omy14 corresponds to Arlee Omy14 and Omy32, distinct from the Hayspur iteration involving Swanson Omy25 and the Keithley Creek iteration involving Swanson Omy04.

Fig. 3.

Synteny of the Rainbow trout genome against different strains. Dot plots comparing the assembly in this study (y-axis) against (A) Arlee (USDA_OmykA_1.1) and (B) Swanson (Omyk_2.0) reference genomes (x-axis).

To further characterize chromosome-scale structural variation, our assembly was used as the reference and compared with five rainbow trout genomes, including Arlee, Swanson, Whale Rock (GCA_029834435.1), Keithley Creek (GCA_034753235.1) and Hayspur (GCA_049308825.1)^12–15. Homologous chromosome pairs were aligned with NUCmer³¹ (MUMmer v4.0.1) using the parameters ‘--mum -l 100 -c 500 -b 500’, filtered with delta-filter using ‘-i 90 -l 1000’, and converted with show-coords using -THrd for SyRI³² (v1.7.1) input. Query sequences oriented opposite to the reference were reverse-complemented and realigned. Among the structural variants identified by SyRI, we highlight three large inversions on Omy05, Omy17 and Omy29. The most prominent event was a double inversion on Omy05 relative to the Swanson assembly, comprising two adjacent inverted segments of approximately 58 Mb in total (Fig. 4A). An approximately 7 Mb inversion on Omy17 was detected relative to both the Arlee and Swanson assemblies, but not relative to Whale Rock, Keithley Creek or Hayspur (Fig. 4B). An approximately 5.8 Mb inversion on Omy29 was detected relative to Arlee, Swanson, Whale Rock and Keithley Creek, but not relative to Hayspur (Fig. 4C). The complete SyRI outputs for all five comparisons are provided on Figshare³³.

Fig. 4.

Dot plots for Omy05, Omy17 and Omy29. Pairwise whole-chromosome dot plots were generated between our rainbow trout assembly and five rainbow trout genomes for Omy05 (A), Omy17 (B) and Omy29 (C). Our assembly is shown on the x-axis, and the corresponding chromosomes from each compared genome are shown on the y-axis. Reverse-oriented alignment blocks indicate inverted regions identified by SyRI.

Repetitive sequence annotation

Repetitive sequences were identified using a combination of homology-based and de novo approaches. RepeatMasker and RepeatProteinMask (v4.1.4) were employed to screen the genome against the FishTEDB 2.0³⁴ database, while tandem repeats were identified using TRF. The analysis revealed that repetitive sequences occupy ~60.79% of the genome. Specifically, long interspersed nuclear elements (LINEs), DNA transposons, long terminal repeats (LTRs), and short interspersed nuclear elements (SINEs) account for 33.74%, 22.71%, 9.88%, and 0.51% of the genome assembly, respectively (Fig. 2 and Table 4).

Table 4.

Statistics of repeat content.

Type	Length (bp)	Percent of Genome (%)
DNA transposons	520,368,747	22.71
LINEs	772,881,697	33.74
SINEs	11,632,141	0.51
LTRs	226,286,894	9.88
Others	566,407	0.03
Unknown	19,806,533	0.87
Total (exclude nested/overlapping TEs)	1,392,665,831	60.79

Protein-coding gene prediction and annotation

Protein-coding genes were predicted by integrating de novo prediction, homology-based search, and transcriptome evidence. De novo prediction was performed using Augustus³⁵ (v3.3). Homology-based evidence was generated by aligning proteins from four related salmonids (Oncorhynchus clarkii, Oncorhynchus gorbuscha, Oncorhynchus kisutch, and the O. mykiss Arlee strain) using Miniprot³⁶ (v0.15) and GeneWise³⁷ (v2.2.0). Transcriptome evidence was derived from ten tissues, processed via HISAT2³⁸ (v2.1.0), StringTie³⁹ (v2.1.4), TACO⁴⁰ (v0.7.3), and TransDecoder⁴¹ (v5.5.0), with assembly refinement using Trinity⁴² (v2.15.1) and PASA⁴³ (v2.5.2). These results were integrated using EvidenceModeler⁴⁴ (EVM; v2.0.0) and combined with EGAPx (v0.4.0-alpha; https://github.com/ncbi/egapx) predictions. The final gene set contained 43,137 genes, achieving 98.9% completeness according to BUSCO⁴⁵ (v6.0.0) assessment with the actinopterygii_odb10 dataset.

Functional annotation was performed by aligning the predicted protein sequences against public databases using Diamond⁴⁶ with an E-value threshold of 1e-5. The databases included NR⁴⁷, EggNOG⁴⁸, SwissProt⁴⁹, KEGG⁵⁰, Pfam⁵¹, and Kofam⁵². In total, 98.46% of the genes were successfully annotated by at least one database (Fig. 5 and Table 5).

Fig. 5.

Gene functional annotation summary of the rainbow trout gene set. UpSet plot displaying the overlap of gene functional assignments among KEGG, Kofam, SwissProt, Pfam, NR, and EggNOG databases.

Table 5.

Statistics of gene function annotation.

Type	Number	Percentage (%)
NR	42,467	98.45
EggNOG	41,050	95.16
SwissProt	30,227	70.07
KEGG	30,377	70.42
Pfam	37,899	87.86
Kofam	30,741	71.26
Total*	42,472	98.46

*Total: Non-redundant genes annotated in at least one database, as a percentage of all predicted protein-coding genes.

Data Records

The primary sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA1392806 and SRA accession number SRP658351⁵³, comprising PacBio HiFi reads (SRR36613569), ONT ultra-long reads (SRR36613570), Hi-C reads (SRR36613567), WGS reads (SRR36613568), and RNA-seq reads from ten tissues (SRR36613571–SRR36613580). The genome assembly has been deposited in NCBI GenBank under accession number GCA_057628025.1⁵⁴, with the Whole Genome Shotgun project accession JBTHGX000000000. The version described in this paper is JBTHGX010000000. The genome annotation files and SyRI outputs are available via Figshare³³.

Technical Validation

The quality of the genome assembly was comprehensively evaluated using multiple complementary approaches. Genome completeness was first assessed using compleasm⁵⁵ (v0.2.6) with the actinopterygii_odb10 dataset, yielding an overall completeness of 99.72% (Fig. 1B and Table 2). Base-level accuracy was evaluated using Merqury⁵⁶ (v1.3) with WGS data, resulting in a quality value (QV) of 40.93. Additionally, the assembly exhibited high read mapping rates, with 99.78% of WGS reads aligned using BWA⁵⁷ (v0.7.17) and 99.97% of HiFi reads aligned using Minimap2.

Author contributions

C.S. conceived and designed the study. S.J., S.L., S.H. and C.L. analyzed the data. S.J. drafted the manuscript. X.C., C.L., C.D. and Y.Z. performed sample collection and coordinated the sequencing efforts. S.J. and C.M. conducted the data submission to NCBI and Figshare. S.L., K.L., Y.L., H.W. and Q.W. revised the manuscript. All authors have reviewed and approved the final version of the manuscript.

Code availability

All bioinformatics analyses in this study were performed using publicly available software. Software versions and parameter details are provided in the Methods section; unless otherwise specified, default parameters were used. No custom scripts or software were used in this study.

Competing interests

The authors declare no competing interests.