Background & Summary

Siluriformes, better known as catfishes, constitute a significant portion of the teleost orders, making up approximately 11% of all species. With 39 families and around 4,094 species documented to date, this order is among the largest in existence1. Owing to their delectable meat, minimal intermuscular bone, and impressive feed conversion ratio, catfishes have risen to become one of the top three farmed fish and shellfish, boasting a production of 5,519 kilotons in 2017 alone2,3.

Yellow catfish (Pelteobagrus fulvidraco) is an important aquaculture fish species in China, the production of which was about 565 thousand tons in 20204. Darkbarbel catfish (Pelteobagrus vachelli), a close relative of yellow catfish within the family Bagridae, has a faster growth rate than yellow catfish5. Recently, hybrid yellow catfish (P. fulvidraco♀ × P. vachelli♂) has been widely cultured in China due to its faster growth rate and better emergence rate6. Unfortunately, specific research on the darkbarbel catfish genome proved difficult to access, but the general principles of genomics in aquaculture suggest that having a high-quality reference genome for this species will have significant benefits for breeding programs and other genetic studies. The insights gained from the darkbarbel catfish genome can also support the successful production of the hybrid yellow catfish, by contributing to a better understanding of their genetic makeup and the genetic factors influencing their advantageous traits.

In this research, we have employed a combination of PacBio long-read sequencing and Hi-C technology to generate a high-quality, chromosome-level assembly of the darkbarbel catfish genome. With the development of this high-quality reference genome, we foresee a significant propulsion in the field of population genetics and the identification of functional genes associated with critical economic traits in the darkbarbel catfish. The elucidation of these genomic underpinnings is expected to provide deeper insights into the hybrid vigor observed in catfish hybrids, thereby contributing to the optimization of hybrid breeding strategies.

Methods

Sample collection and sequencing

An adult female darkbarbel catfish was collected from the Yangtze River in Wuhan, Hubei, China. High-molecular weight (HMW) genomic DNA was extracted from muscle for Illumina sequencing and PacBio SMRT sequencing. The quality and quantity of the extracted DNA was assessed using standard agarose gel electrophoresis and a Qubit fluorometer (Thermo Fisher Scientific, USA).

For Illumina sequencing, the genomic DNA was randomly sheared to ~350 bp fragments, and a paired-end genomic library was prepared following the manufacturer’s protocol. Then, the library was sequenced on an Illumina HiSeq X-Ten platform using a paired-end 150 bp layout. For PacBio sequencing, the genomic DNA was used to construct SMRTbell libraries following the manufacturer’s protocol. After that, the libraries were sequencing on a PacBio Sequel platform with SMRT technology. Finally, we generated 96.33 Gb Illumina short-read data and 109.76 Gb of raw PacBio continuous long reads (CLR) with an average read length of 13.8 kb and a N50 read length of 22.4 kb (Table 1).

Table 1 Statistics of the sequencing data.
Full size table

For genome scaffolding, a Hi-C library was prepared using blood sample from the same darkbarbel catfish used for genomic DNA sequencing. The Hi-C library construction, including cell crosslinking, cell lysis, chromatin digestion, biotin labelling, proximal chromatin DNA ligation and DNA purification, was performed as previously described7, and the resulting Hi-C library was then subjected to paired-end sequencing with 150 bp read lengths on an Illumina HiSeq X-Ten platform. As a result, 89.12 Gb of Hi-C read data was generated (Table 1).

To aid in genome annotation, total RNA was extracted from multiple tissues, including spleen, kidney, brain, muscle, ovary and liver, and the quality were evaluated using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA) and an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). The mixed RNA sample was used to construct a cDNA library using the TruSeq Stranded mRNA Library Prep Kit (Illumina, USA) following the manufacturer’s protocol. The library was then sequenced on an Illumina HiSeq X-Ten platform using a paired-end 150 bp layout, and 13.11 Gb of data was obtained (Table 1).

Genome assembly and polishing

To assemble the genome, we utilized two different assemblers: Wtdbg2 v2.58 and Flye v2.99 (Fig. 1a). The assembly generated by each assembler with default parameters was then polished using Arrow, a consensus algorithm that can generate highly accurate consensus sequences from PacBio subreads. The two polished assemblies were then merged using Quickmerge10, a tool that combine multiple genome assemblies into a single consensus assembly. The resulting merged assembly was then polished twice using two rounds of Arrow and two rounds of NextPolish with default parameters (Fig. 1a). We used PacBio subreads for Arrow and Illumina short reads for NextPolish. The resulting assembly consists of 318 contigs and has a total length of 691.96 Mb (Table 2).

Fig. 1
figure 1

Overview of the genome assembly for darkbarbel catfish. (a) Flow chart illustrating the genome assembly pipeline for darkbarbel catfish. (b) Hi-C heatmap (200-kb resolution) showcasing the interaction frequencies between different chromosomes of darkbarbel catfish.

Full size image
Table 2 Assembly statistics of darkbarbel catfish.
Full size table

Hi-C scaffolding

The raw Hi-C reads were processed to remove adapters and low-quality bases using Fastp v0.20.111 with parameters -q 20 -l 50. The processed reads were then aligned to the assembly using the Juicer pipeline12. Then the 3D-DNA pipeline13 was used to group the contigs into chromosomes, orient and order the contigs within each chromosome. To further improve the quality of the assembly, we manually corrected the errors using the Juicebox Assembly Tools12. Following the scaffolding procedure, 691.13 Mb were successfully anchored to the 26 chromosomes (Fig. 1b), encompassing an impressive 99.9% of the total assembly size of 692.10 Mb (Table 2). The observed chromosome number concurs with the karyotype analysis reported in the previous study14. The scaffold N50 reached a substantial 27.55 Mb for the final assembly (Table 2). Notably, among the 26 chromosomes, 16 of them exhibited exceptional contiguity with no more than 10 gaps observed (Table 3).

Table 3 Assembly statistics for chromosomes.
Full size table

Repeat annotation

RepeatModeler v2.0.215 firstly identified repetitive sequences in the genome assembly using several tools, including RECON, RepeatScout, TRF, Ltr_retriever, and LTRharvest. The identified sequences were then clustered and classified into families using RepeatModeler. The classified libraries were combined with the Teleostei library from Repbase16. RepeatMasker v4.1.417 was performed to mask repetitive sequences in the genome assembly using the combined library generated by RepeatModeler. A significant portion of the genome, approximately 35.79%, is masked, resulting in 247,692,317 bp being identified as repetitive elements. Retroelements, including long terminal repeats (LTRs, 6.90%), long interspersed nuclear elements (LINEs, 5.87%), and short interspersed nuclear elements (SINEs, 1.01%), collectively comprise the largest proportion, occupying 13.77% of the genome (Fig. 2). Furthermore, DNA transposons occupy a notable 11.48% of the genome, while unclassified elements add a nuanced layer, constituting 4.69% of the genomic landscape (Fig. 2).

Fig. 2
figure 2

Genomic landscape of darkbarbel catfish. Circos plot of darkbarbel catfish illustrating from outside to inside, gene density (a), GC content (b), and the densities of DNA transposons (c), LTRs (d), LINEs (e), and SINEs (f), all represented in 200-kb genomic windows.

Full size image

Gene prediction and function assignment

In this research, we employed a comprehensive approach combining transcriptome-based, de novo, and homology-based methods to predict genes within the genome. For transcriptome-based prediction, RNA-seq reads underwent stringent quality filtering using Fastp v0.20.1, with specific parameters set at -q 20 -l 50. These filtered reads were then aligned to the genome assembly using HISAT2 v2.2.118, followed by assembly using StringTie v2.2.119. Gene structures were subsequently predicted utilizing TransDecoder v5.5.0 (https://github.com/TransDecoder/TransDecoder). For de novo prediction, RNA-seq aligned BAM files served as input for training the AUGUSTUS v3.4.020 gene prediction tool via BRAKER21. This trained model was then employed to predict gene structures within the genome. In the homology-based prediction, we utilized miniport v0.1122 to align protein sequences from P. fulvidraco7, Silurus meridionalis23, and Ictalurus punctatus24 to the genome assembly, enabling the prediction of gene structures based on homologous evidence. To consolidate the results from these three methods, EvidenceModeler25 was employed, enabling the merging and integration of gene predictions. Following the gene prediction, the finalized gene sets derived from preceding methods underwent functional annotation through matching with a variety of databases. In particular, we utilized BLASTP v2.9.026 to align the anticipated genes with SwissProt27, TrEMBL28, eggNOG29, and the NCBI non-redundant (NR) protein databases.

In total, we successfully predicted 22,109 protein-coding genes within the genome (Fig. 2). These predicted genes displayed an average coding sequence length of 1,695.04 bp, an average gene length of ~15 kb, and an average exon number of 10. Further, 96.1% of the total predicted genes, which equates to 21,243 genes, were successfully assigned with at least one functional annotation (Table 4 and Fig. 3).

Table 4 Statistics of functional annotation result.
Full size table
Fig. 3
figure 3

Venn diagram of function annotations from various databases. The Venn diagram displays the overlap and uniqueness of functional gene annotations derived from 4 databases: TrEMBL, NR, Siwss-Prot, and eggNOG.

Full size image

Genome synteny analysis

To compare the whole genome synteny, two chromosome-level genomes of Bagridae including yellow catfish30 and Chinese longsnout catfish (Leiocassis longirostris)31 were aligned to the genome assembly of darkbarbel catfish using LAST v135432 with default parameters. The synteny were visualized using Circos v0.69.933. A high degree of synteny conservation between the compared genomes was observed (Fig. 4).

Fig. 4
figure 4

Chromosome sequence synteny comparisons. (a) Syntenic relationship between the darkbarbel catfish genome and the Chinese longsnout catfish genome. (b) Syntenic relationship between the darkbarbel catfish genome and the yellow catfish genome. Each line connects a pair of homologous sequences between the two species.

Full size image

Data Records

All the raw sequencing data utilized in this study, including WGS, RNA-Seq, and Hi-C, have been deposited in the NCBI database under the BioProject accession number PRJNA819563. Specifically, the Illumina WGS data was archived with the accession number SRR2492634334, while the PacBio WGS data was deposited with the accession number SRR2235495735. The RNA-Seq and Hi-C data sets were archived under the accession numbers SRR2492826336 and SRR2179906337, respectively. The genome assembly is available for public access at the NCBI GenBank under the accession number GCA_030014155.138. Genome annotations, along with predicted coding sequences and protein sequences, can be accessed through the Figshare39.

Technical Validation

To evaluate the completeness of the genome assembly, we used the BUSCO v5.4.240 with the Actinopterygii database (actinopterygii_odb10) to assess the presences of conserved sing-copy genes in the assembly. Out of the total 3,640 BUSCO groups searched, an impressive 99.07% were identified as complete, indicating a high level of gene content preservation. Among these, 98.54% were both complete and present as single-copy genes, further emphasizing the quality of the assembly. Additionally, only 0.33% of the BUSCOs were fragmented, and 0.60% were missing from the assembly (Table 5). This demonstrates the remarkable completeness and conservation of gene content in the darkbarbel catfish genome assembly, achieving one of the best BUSCO scores observed among reported catfish genomes.

Table 5 BUSCO assessment result.
Full size table

To ensure the quality and accuracy of the genome assembly, we employed a two-step validation process. Firstly, the assembly’s Quality Value (QV) was quantified using Merqury41, resulting in a QV score of 40.89, reflecting a high-grade assembly. Then, we mapped the raw sequencing data back to the assembly. For WGS short reads, we utilized BWA v0.7.1742, which resulted in a high mapping rate of 99.79%. For mapping RNA-Seq reads, we used HISAT2 v2.2.1 and achieved an overall mapping rate of 96.44%.