Background & Summary

Belonging to the order Perciformes and the family Sparidae, blackhead seabream (Acanthopagrus schlegelii) has been among the most economically important and well-liked marine fish species. It has quick development, outstanding flesh quality, and good environmental adaptability1,2. Its artificial breeding and aquaculture is expanding quickly as a result of the growing demand in the market. Implementing artificial breeding or selection has been employed to facilitate the increased production of blackhead seabream. Sex control is therefore applied as one of the most crucial strategies.

Interestingly, blackhead seabream is a protodromous hermaphrodite with an impressive life cycle of natural sex change from male to female, making it a valuable model for studying molecular mechanisms of fish sex development3.The fish exhibit sexual differentiation during its juvenile stage, possessing a bisexual gonad. This can lead to a fact that male and female reproductions are not synchronized and the ratio of female to male is severely disproportional. Without the ability to regulate sexual differentiation, maturation, and reproduction, farmers indeed have little control over practical breeding processes. Interestingly, previous studies showed that age and season play an combined impact on the dynamic process of bisexual gonad development3. Meanwhile, development of vitellogenic oocytes in the ovary serves as one of the indicators for natural sex change in blackhead seabream4,5. Although some genes such as wnt4 (Wnt family member 4), foxl2 (Forkhead Box L2), cyp19a1a (cytochrome P450, family 19, subfamily A, polypeptide 1a), dmrt1 (doublesex and mab-3 related transcription factor 1), amh (anti-Mullerian hormone), and amhr2 (anti-Mullerian hormone receptor type 2), potentially related to sexual differentiation and sex controlling, have been reported1,4,5; however, the detailed molecular mechanisms of natural sex change is still unknown in blackhead seabream.

For investigations of functional, ecological, and evolutionary genomics in this species as well as other hermaphrodite fishes, it is necessary for researchers to have a well-assembled genome in hand. A draft genome assembly of a female blackhead seabream was released by us in 2018, containing 89% of the full BUSCOs (actinopterygii_odb9 database) and numerous contigs (115, 091) with a low N50 value of 17.2 kb6. Although this assembly provided an important genetic resource for comparative genomics studies on various Perciformes, its fragmentation and incompleteness has restricted its comprehensive applications in fish research. In our present work, we assembled a gap-free telomere-to-telomere (T2T) genome of a male blackhead seabream by integration of PacBio (Pacific Biosciences) HiFi long reads, ONT (Oxford Nanopore Technologies) ultra-long reads, MGI short reads, and Hi-C (High-through chromosome conformation capture) sequencing reads. This high-quality genome dataset will greatly enable further works on understanding of biological characteristics (especially sex change) of blackhead seabream.

Methods

Sample collection

A two-year-old male adult blackhead seabream (Fig. 1a) was collected from Guangdong Marine Fisheries Experimental Centre, which belongs to the Agro-Tech Extension Center of Guangdong Province with an offsite location in Huizhou city, Guangdong province, China. Muscle tissue was pooled for whole genome sequencing, and muscle, liver, brain, gill, and gonad were collected for additional transcriptome sequencing. The sampling procedure and practical pipeline were conducted according to the recommendations and approval of the Animal Ethics Committee of Shenzhen University (Shenzhen, China; license number: A202301455).

Fig. 1
figure 1

A T2T genome assembly of the protandrous hermaphrodite blackhead seabream. (a) A photo of the sequenced fish. (b) A GenomeScope k-mer plot. (c) A total of 24 distinct blocks in the Hi-C contact matrixes. (d) A Circos plot of the main genome features. From outside to inside: (I) the 24 chromosomes, (II) repeats, (III) Long terminal repeat (ltr), (IV) gene density, and (V) GC content. Links inside the Circos refer to internal syntenic blocks among different chromosomes within the assembled genome.

Full size image

DNA extraction and genome sequencing

A blood & cell culture DNA kit (Qiagen, USA) was used for extraction and purification of genomic DNA (gDNA) from the muscle in accordance with the manufacturer’s instructions. The extracted gDNA was used for construction of a library (insert size of 350 bp) with the MGIEasy Universal DNA Library Prep Kit (MGI, China), which was then sequenced on an MGISEQ. 2000 platform (MGI). A total of 100.47 Gb of raw reads (150 bp in length) were generated, among them low-quality reads and adaptor sequences were filtered using SOAPfilter (v2.2)7 with default settings. Finally, we obtained 95.98 Gb of clean reads for estimating genome size and assembling sequences.

Additionally, long-read libraries were created utilizing a PacBio Sequel II System and a SMRTbell Express Template Prep Kit 2.0 for HiFi sequencing, following PacBio’s standard technique (Pacific Biosciences, USA). The consensus sequences were then produced using the CCS software (SMRT Link v9.0)8. Approximately 62.60 Gb of consensus reads with a mean length of 15.25 kb were obtained.

Oxford Nanopore Technologies (ONT) was applied for construction of an ultra-long library and then one flow cell was sequenced on a PromethION platform (Oxford Nanopore Technologies Co., UK). The raw reads were first refined to remove those with quality value (QV) below 7. Subsequently, Porechop (https://github.com/rrwick/Porechop) was applied to eliminate adaptors, and Filtlong (https://github.com/rrwick/Filtlong) was employed to filter out those reads shorter than 30 kb and mean read quality scores less than 90%. Finally, a total of 1.171 million clean reads were retained, accumulating a substantial count of 22.44 Gb. The average read length was 79.63 kb, with an N50 length of 100 kb.

For the Hi-C sequencing, DNA libraries were prepared with a GrandOmics Hi-C kit (GrandOmics, China; using DpnII as the restriction enzyme) as per the manufacturer’s instructions. The Hi-C libraries were then sequenced on an Illumina Novaseq system (Illumina, USA), generating 76.46 Gb of 150-bp paired-end raw reads. Trimmomatic (V0.39)9 with optimized parameters “SLIDINGWINDOW: 4:20, LEADING: 3, TRAILING:3, ILLUMINACLIP: adapter.fa:2:30:10:8:true” was used to filter out adaptor sequences, low-quality reads (quality scores < 20), and those reads shorter than 36 bp. This filtering process retained 63.53 Gb of clean data for subsequent construction of chromosomes.

RNA extraction and transcriptome sequencing

Using a normal Trizol methodology (Invitrogen, USA), RNA samples were isolated from muscle, liver, brain, gill, and gonad tissues, and then purified using a Qiagen RNeasy Mini Kit (Qiagen). Following the manufacturer’s instructions, equivalent amount of RNA from each tissue were combined to create an Illumina cDNA library, which was subsequently sequenced on a HiSeq X Ten platform (Illumina). After generating 31.12 Gb of paired-end raw reads, adaptor sequences and low-quality reads were removed using SOAPfilter (v2.2)7 with default settings. In the end, 27.1 Gb of clean reads were retained for annotation of gene structures.

Genome assembly

Genome-size estimation

Using Jellyfish (v2.2.6)10 and GenomeScope (v2.0)11, a K-mer frequency distribution of the MGI clean reads was determined. Accordingly, the genome size of blackhead seabream is calculated to be 671.32 Mb, and the rate of genomic heterozygosity was predicted to be 0.59% (Fig. 1b).

De novo genome assembly

In this study, HiFiasm (v0.19.5)12 was employed for assembling into contigs using HiFi + ONT long reads. Subsequently, these contigs were refined by T2T-polish13 with the optimized parameter set to task = best using the MGI short reads.This primary genome assembly was 731.09 Mb in length, and it is anchored into 41 contigs (with a contig N50 of 31.8 Mb).

Construction of chromosomes and gap filling

This high-quality genome assembly served as the foundation for subsequent construction of chromosomes using the Hi-C reads. Initially, Bowtie2 (v2.3.2)14 was employed to map clean reads to the primary genome assembly. Subsequently, the HiC-Pro (v2.8.1)15 pipeline was employed to detect valid contact paired reads. Using these Hi-C valid reads, the assembled contigs were anchored onto chromosomes via the 3D-DNA pipeline16 with the parameter set to -r 0. Manual correction of the chromosome-level scaffolds was then performed using JuiceBox (v1.11.08)17. Based on the HiFi and ONT long reads, we applied TGS-GapCloser (v1.1.1)18 with default parameters to resolve any remaining gaps in the chromosome-level genome assembly. As a result, we obtained the final genome assembly, which has a total size of 714.71 Mb and is anchored into 24 chromosomes with 97.78% of the primary assembly sequences (Fig. 1c, d and Table 1). Its contig N50 reaches 31.8 Mb.

Table 1 Telomere and centromere positions in the assembled blackhead seabream genome.
Full size table

Identification of centromere and telomere sequences

We identified telomere sequences through searching for repeating sequences (TTAGGG/CCCTAA) in telomeric regions. Centromere identification was performed by using the Centromics program (https://github.com/ShuaiNIEgithub/Centromics), which dealt with raw HiFi sequencing data, Hi-C sequencing data, and genome assembly data. Ultimately, we discovered that blackhead seabream chromosomes owned 24 centromeres and 47 telomeres (see more details in Fig. 2 and Table 1).

Fig. 2
figure 2

An overview of the T2T gap-free reference genome of blackhead seabream. The yellow areas at both ends of each chromosome represent the telomere regions, and the gully area within each chromosome represents the centromere region.

Full size image

Genome annotation

Repetitive sequence annotation

Both ab initio and homology-based strategies were employed to detect repetitive sequences in the blackhead seabream genome. In detail, using RepeatModeler (v2.0.1)19 and MITE-Hunter20 program with default settings, we constucted an ab initio repeat sequence library. This library was then aligned to the Repbase 24.021 for classification of different repetitive elements via the TEclass tool22. For the homolog-based prediction based on the Repbase database21, DNA and protein transposable elements (TEs) were detected by RepeatMasker (v4.1.2) and RepeatProteinMasker (v4.0.7)23, respectively. Tandem repeats were identified with Tandem Repeat Finder (v4.10.0)24. After removal of redundant data from both methods, our results showed that a total of 221.23 Mb of repetitive sequences were identified, accounting for 30.95% of the assembled blackhead seabream genome (Table 2).

Table 2 Repetitive elements and their proportions in the assembled blackhead seabream genome.
Full size table

Protein-coding genes and functional annotations

Three methods were combined to annotate protein-coding genes, including ab initio gene prediction, homology-based annotation, and transcriptome-based annotation. Using the HISAT2 (v2.2.1)25, refined transcriptome (RNA-seq) reads were aligned to the assembled genome for the transcriptome-based annotation. PASA26 was then applied to detect open reading frames (ORFs), and Stringtie27 was used for gene structure annotation by assembling corresponding transcripts. For the homology prediction, GeMoMa (v2.3)28 was employed to map protein sequences from four representative species (including yellowfin seabream Acanthopagrus latus, sharksucker Echeneis naucrates, zebrafish Danio rerio, and large yellow croaker Larimichthys crocea) to our assembly for prediction of gene structures. For the ab initio prediction, a training model library was constructed from a protein set generated from the RNA-seq reads by Trinity (v2.8.5)29. Augustus (v3.4.0)30 was then employed to annotate genes with the–noInFrameStop = true–strand = both setting based on the training data.

Integration of protein-coding genes annotated by these three approaches was carried out using the EVidenceModeler (EVM) pipeline (v1.1.1)26. Consequently, 24,081 protein-coding genes were identified, with an average gene length of 16.34 kb and an average coding sequence (CDS) length of 1,721.43 bp. The average number of exons per gene was 10.28, while exons averaged 271.91 bp in length and introns averaged 1,437.93 bp (see Table 3).

Table 3 Gene structures and the functional annotation.
Full size table

The protein-coding genes were functionally annotated by aligning them with several routine protein databases using Blastp (v2.2.26)31 and Diamond (v2.0.7)32. This alignment included comparisons with the NCBI nonredundant (NR) protein database (v5; released on September 29, 2020), as well as Swissprot (released on October 07, 2020)33, KEGG (released on October 1, 2019)34, TrEMBL (released on October 07, 2020)34, InterPro (v5.50–84.0)35 and Gene Ontology (GO; v1.2, released on July 27, 2023)36 databases. Of the total gene models, 24,179 (98.36%) were annotated with at least one homologous hit from these public databases (refer to Table 3).

Annotation of non-coding RNA genes

For annotation of non-coding RNA genes, tRNAscan-SE (v2.0.9)37 with default settings was employed to annotate tRNA-associated genes; rRNAs annotation was carried out using RNAmmer (v1.2)38; miRNAs and snRNAs were detected using Infernal (v1.1.2)39 against the Rfam (v1.4.1) database40 with default parameters. As a result, annotations predicted 956 rRNAs, 2,175 tRNAs, 874 miRNAs, and 1,212 snRNAs (see more details in Table 4).

Table 4 Statistics of the non-coding RNA annotations.
Full size table

Data Records

All genome data have been uploaded to the NCBI SRA database with the BioProject accession PRJNA1134337, including the specific accessions SRR29908548 to SRR2990855141,42,43,44. The genome assembly have been deposited in the GenBank database under the accession number JBGQWW00000000045. Furthermore, detailed documents on genome assembly, gene structures, gene functions, and repeat annotations for blackhead seabream have been shared on the Figshare46.

Technical Validation

Evaluation of the genome assembly and annotation

To evaluate the quality of our genome assembly, we employed several approaches. First, we employed BUSCO (v5.2.2)47 to examine completeness. The BUSCO analysis revealed that, based on the 3640 single-copy orthologs in the actinopterygii_odb10 database, 99.3% of our annotated genes were correctly classified as complete (99.1% single-copy genes and 0.2% duplicated genes), with 0.6% being fragmented (see Table 5). Second, Merqury (v1.328)48 estimated the assembly quality value (QV) to be 52.95. Third, by aligning the sequencing data to the assembled genome, we evaluated the accuracy rate, which showed mapping rates of 96.53% for RNA-Seq data, 99.43% for the MGI data, 99.74% for the PacBio data, and 99.99% for the ONT data. These results collectively indicate high quality of the blackhead seabream genome assembly. Moreover, a BUSCO analysis was performed to assess the completeness of the gene structure annotation, revealing that 96.2% of our annotated genes were correctly classified as complete, with 0.8% being fragmented.

Table 5 Statistics of BUSCO evaluation results of previously published6 and T2T genome assembly.
Full size table

Collinearity analysis

Whole-genome synteny analysis was performed using the GenomeSyn (v1.2.7)49 by aligning the chromosome-level genome assemblies between blackhead seabream (this study) and its relative yellowfin seabream (A. latus)50. Our results prove that they had excellent one-to-one correspondences among their chromosomes (Fig. 3). This good similarity also underlines the high quality of our sequencing and assembly of the blackhead seabream genome.

Fig. 3
figure 3

Good synteny of chromosomes between blackhead seabream (Acanthopagrus schlegelii) and its relative yellowfin seabream (Acanthopagrus latus)46.

Full size image