Background & Summary

Food security is a fundamental challenge in the context of human survival and development. Aquatic foods, known as blue foods, are abundant in essential micronutrients and fatty acids while imposing lower environmental burdens. These foods offer protein and valuable nutrients to billions of people, particularly in developing countries1,2,3,4. In comparison to capture fisheries, aquaculture continues to dominate global blue food production and holds promise for meeting food demand and addressing malnutrition2,5. Inland aquaculture, excluding mariculture, significantly contributes to global food security, particularly in the global south6,7,8. However, freshwater resources and arable land pose primary constraints to the growth of inland aquaculture industries. Arid regions, covering approximately 6.1 billion hectares or 41% of the Earth’s land area, constitute a substantial part of the planet’s landmass9. Expanding inland aquaculture in arid areas represents a crucial pathway for industry development. Advancements in cultivation techniques have facilitated the robust development of aquaculture in arid regions, particularly in Africa. This not only mitigates food crises to some extent but also drives overall national and societal progress10.

The Tarim River, China’s longest inland river, serves as the main river of the southern Xinjiang Autonomous Region11. The Tarim River Basin is characterized by arid conditions, including limited precipitation, high evaporation rates, sparse vegetation, minimal runoff, severe water salinization, and a simple native fish fauna12. To utilize saline-alkaline water resources and diversify animal protein sources, several euryhaline non-native fish species, such as Ctenopharyngodon della, Cyprinus carpio, and Carassius auratus, have been introduced for saline-alkaline fisheries. Unfortunately, this practice poses significant ecological risks13,14,15. Furthermore, various water storage and diversion projects have been implemented along the upstream to downstream axis of the Tarim River Basin to alleviate water scarcity16,17. The combination of hydraulic engineering projects and the introduction of non-native fish species for aquaculture can facilitate fish invasions and lead to a drastic decline in indigenous fish populations and faunal homogeneity18,19,20.

Breeding euryhaline native fish species may offer an effective solution for mitigating the ecological impacts resulting from biological invasions and conserving native fish populations through artificial propagation and release15,21. T. yarkandensis, exclusively distributed in the Tarim River Basin, is an euryhaline indigenous fish species with aquaculture potential22,23,24. This species is currently under serious threat and has been included in the list of key protected wild animals for the Xinjiang Uygur Autonomous Region (https://www.xinjiang.gov.cn/xinjiang/zfgbml/202301/2dff780e69894c2cbe56a7b7866e58ca.shtml). Elucidating the complete genome of T. yarkandensis not only provides insights into breeding techniques but also offers valuable suggestions for its protection. Therefore, this study combines PacBio long-read sequencing and high-throughput chromosome conformation capture (Hi-C) technology to generate a high-quality, chromosome-level reference genome of T. yarkandensis. This achievement will assist in developing effective protection strategies for this species and serve as a basis for exploring adaptive evolution in arid regions.

Methods

Sample collection and sequencing

In August 2021, a female T. yarkandensis (Fig. 1) was captured at the sampling location of Yarkant River (76°30′56′′ E, 37°59′5′′ N). This species was identified according to their morphological features as described in Fauna Sinica (Osteichthyes: Cyprinifores) and The fishes of the Qinghai-Xizang Plateau. Following anesthesia with MS-222 and disinfection, white muscle tissue was preserved in liquid nitrogen for genomic DNA sequencing. The QIAGEN Genomic-tip 100/G kit was employed for genomic DNA extraction from T. yarkandensis, and high-quality DNA was utilized for subsequent library preparation and high-throughput sequencing.

Fig. 1
figure 1

The morphological image of female Triplophysa yarkandensis collected in the Yarkant River.

Full size image

To construct a 20 kb long-read sequencing library (SMRT bell library), 10 µg of DNA was utilized. Once the library passed the quality assessment, PacBio Sequel was used for sequencing, following the desired data volume requirements25. Sequencing was conducted using the Sequel Binding Kit 2.0, Sequel Sequencing Kit 2.1, and Sequel SMRT Cell 1 M v2, and the resulting data was processed using SMRT LINK 5.0 software. For Hi-C sequencing, the process began with Hi-C biotin labeling and genomic DNA extraction26. The captured DNA was subjected to end repair, poly A tailing, adapter ligation, evaluation of PCR amplification cycles, and purification. After qualifying the library inspection, the library was pooled based on the effective concentration and the target offline data volume for HiSeq sequencing.

In order to assist in genomic annotation, total RNA was extracted from six tissues involving kidney, liver, gonad, muscle, brain and gill. The cDNA library was constructed using mixed RNA samples, and the Illumina HiSeq X-Ten platform was employed for sequencing.

Genome assembly

The SMRTbell libraries were subjected to sequencing on a PacBio Sequel II system. The consensus reads, also known as HiFi reads, were generated using the ccs software (https://github.com/pacificbiosciences/unanimity) with the parameter ‘-minPasses 3’. To enhance the quality and validate the assemblies, we generated 24.6 Gb of PacBio HiFi reads for this specific sample (Table 1). These HiFi reads, which are long (approximately 15 kb) and highly accurate ( > 99%), were assembled using Hifiasm27 (https://github.com/chhylp123/hifiasm). To rectify any errors in the primary assembly, Illumina-derived short reads were employed, and remaining errors were corrected using pilon28,29 (v1.23). As a result, the T. yarkandensis genome assembly reached a total length of approximately 520.6 Mb, consisting of 1707 contigs, with a ContigN50 value of 1.3 Mb (Table 2).

Table 1 Sequencing data used for the genome T. yarkandensis assembly.
Full size table
Table 2 Assembly and annotation statistics of the T. yarkandensis genome.
Full size table

For the anchored contigs, a total of 130.5 Gb of clean read pairs was generated from the Hi-C library (Table 1). These reads were mapped to the polished T. yarkandensis genome using BWA (bwa-0.7.17) with default parameters. Paired reads in which the mate was mapped to a different contig were utilized for Hi-C-associated scaffolding. Various types of invalid reads, including self-ligation, non-ligation, Start NearRsite, PCR amplification, random break, Large Small Fragments, and Extreme Fragments, were filtered out using Hicup software. Subsequently, we successfully clustered 1354 contigs into 25 groups using the agglomerative hierarchical clustering method in 3d-DNA (Fig. 2). 3d-DNA was further employed to order and orient the clustered contigs30,31,32,33. Finally, we obtained the first high-quality assembly at the chromosomal level, with chromosomal lengths ranging from 13.6 Mb to 31.1 Mb, the T. yarkandensis genome was obtained with 245 scafolds and a total length of 521,366,357 bp, encompassing 93% of the total sequence (Table 2).

Fig. 2
figure 2

Triplophysa yarkandensis genome contig contact matrix using Hi-C data. LG1–25 represented for the 25 pseudo-chromosomes. The depth of red color shows the contact density.

Full size image

Repetitive sequence annotation

In our study, we employed a combination of two methods, namely homology-based and de novo prediction, to identify repeat contents in the T. yarkandensis genome. For the homology-based analysis, we used RepeatMasker (open-4.0.9) with the Repbase TE library to identify known TEs within the genome. In addition, we employed RepeatModeler (http://www.repeatmasker.org) for de novo prediction. RepeatModeler integrates two core de novo repeat-finding programs, RECON (v1.08) and RepeatScout (v1.0.5), to comprehensively discover, refine, and classify consensus models of potential interspersed repeats in the T. yarkandensis genome34,35. Moreover, we conducted a de novo search specifically for long terminal repeat (LTR) retrotransposons using LTR FINDER (v1.0.7), LTR harvest (v1.5.11), and LTR_retriever (v2.7) against the T. yarkandensis genome sequences36,37. Tandem repeats were identified using the Tandem Repeat Finder38 (TRF) package, and Simple Sequence Repeats (SSRs) were detected using MISA (v1.0). Finally, we merged the library files generated from both methods and utilized RepeatModeler to determine the repeat contents. Based on these analyses, we identified a total of 149.99 Mb of repeats in the T. yarkandensis genome (Table 3).

Table 3 Repeat sequence results statistics of Triplophysa yarkandensis genome.
Full size table

Non-coding RNA annotation

To identify specific gene categories in the T. yarkandensis genome, we utilized various algorithms and databases. The tRNAscan-SE (v1.3.1) algorithm with default parameters was employed to detect tRNA genes. tRNA molecules act as adaptors in biological processes, bridging the three-letter genetic code in messenger RNA (mRNA) with the twenty-letter amino acid code in proteins. For the identification of rRNA genes, we used RNAmmer (v1.2) with the parameters “-S euk -m lsu,ssu,tsu”. rRNAs are integral components of the ribosome and play a crucial role in protein synthesis. snoRNAs, a class of small RNA molecules, guide chemical modifications of other RNAs, including ribosomal RNAs, transfer RNAs, and small nuclear RNAs. MiRNAs and snRNAs were identified by CMSAN (v1.1.2) software against the Rfam (v14.0) database with default parameters (Table 4)39,40,41.

Table 4 The number of the annotated non-coding RNA in the Triplophysa yarkandensis.
Full size table

Protein-coding gene prediction and annotation

To predict protein-coding genes in the T. yarkandensis genome, we employed three methods: ab initio gene prediction, homology-based gene prediction, and RNA-Seq-guided gene prediction. Prior to gene prediction, the assembled genome underwent hard and soft masking using RepeatMasker. Ab initio gene prediction was performed using Augustus (v. 3.3.3)42. The gene predictors’ models were trained using a set of high-quality proteins derived from the RNA-Seq dataset. For homology-based gene prediction, we utilized MAKER (v. 2.31.10)43. Protein and transcript sequences were aligned to the genome assembly, and coding genes were predicted using maker with default parameters. RNA-Seq-guided gene prediction involved aligning clean RNA-Seq reads to the genome using hisat2 (v2.0.0). Gene structures were generated using Trinity (v2.3.2), Transdecoder (v2.01), and MAKER. To integrate the predictions from the three methods and generate gene models, we employed EVidenceModeler (EVM, v1.1.1)44. The resulting output comprised consistent and non-overlapping sequence assemblies, which described the gene structures. Overall, a total of 25,505 protein-coding genes with an average length of 158,469 bp were predicted in the assembled T. yarkandensis genome. The predicted protein-coding gene BUSCO integrity using the Actinopterygii odb9 database was 91.5%. The number of predicted proteins was 30,673.

For inferring gene functions, we conducted alignments to various protein databases, including the National Center for Biotechnology Information (NCBI) Non-Redundant (NR), TrEMBL, KOG, and SwissProt, using BLASTP (NCBI BLAST v2.6.0+). Additionally, we utilized the Kyoto Encyclopedia of Genes and Genomes (KEGG) database with an E-value threshold of 1E-5. Protein domains were annotated using PfamScan (v1.6) based on the PFAM and InterPro protein databases. Gene Ontology (GO) IDs for each gene were obtained from Blast2GO. In total, approximately 23,288 (about 91%) of the predicted protein-coding genes in T. yarkandensis could be functionally annotated with known genes, conserved domains, and Gene Ontology terms (Table 5).

Table 5 Protein-coding gene prediction for T. yarkandensis genome.
Full size table

Data Records

All raw data of the whole genome have been deposited into the National Center for Biotechnology Information (NCBI) SRA database under BioProject accession number PRJNA995909 The genomic PacBio sequencing data were deposited in the SRA at NCBI SRR2535771245 and the Hi-C sequencing data were deposited in the SRA at NCBI SRR2534350746. The RNA sequencing data were deposited in the SRA at SRR2637750347. The assembled genome was deposited in the NCBI Genome with the accession number GCA_033220385.148. Genome annotations, along with predicted coding sequences and protein sequences, can be accessed through the Figshare49.

Technical Validation

Evaluation of the quality of genomic DNA and RNA

Before constructing the DNA library, we assessed the purity (OD260/280 and OD260/230) and concentration of the genomic DNA using Nanodrop (LabTech, USA). To precisely measure the concentration of genomic DNA, we employed Qubit (Thermo Fisher Scientific, USA). By comparing the Qubit concentration with the Nanodrop concentration, we determined the sample purity. Additionally, the integrity of the DNA was assessed through agarose gel electrophoresis (1%). RNA purity and integrity were determined using NanoPhotometer spectrophotometer (Implen, USA) and Agilent 2100 bioanalyzer (Agilent Technologies, USA).

Genome assembly integrity assessment

The assembled genome was subjected to BUSCO (v3.1) analysis using OrthoDB to assess its completeness50. Overall, the assembled genome identified 94.1% completeness of the BUSCOs (Actinopterygii odb9) (Fig. 3).

Fig. 3
figure 3

BUSCO assessment results of Triplophysa yarkandensis genome. C represents complete BUSCOs, S represents complete and single-copy BUSCOs, D represents complete and duplicated BUSCOs, F represents fragmented BUSCOs, M represents missing BUSCOs, n represents total BUSCO groups searched.

Full size image