Background & Summary

Large yellow croaker (Larimichthys crocea) is a marine fish that inhabits the nearshore seas and estuaries of the northwestern Pacific Ocean, typically in temperate areas. The Chinese mariculture industry currently yields over 281,000 tons of large yellow croaker, indicating the important economic value of this fish1. Abundant genomic resources have been established for this species, comprising six genetic maps2,3,4,5,6,7, three draft genomes utilizing Illumina technology8,9,10, a draft genome using the combination of Illumina and PacBio sequencing technologies11, and a chromosome-level reference genome generated based on PacBio and Hi-C technologies12.

In the 1960s, some studies divided the large yellow croaker along the Chinese coastline into three stocks based on morphological data: Naozhou stock (NZ, located in the west South China Sea), Min-Yuedong stock (MYD, located in the eastern South China Sea and Taiwan Strait), and Daiqu stock (DQ, located in the East China Sea)13,14. In recent years, with the application of molecular genetic markers, more studies have reported the analysis of the population structure of large yellow croaker. Lin et al. have utilized a UPGMA tree based on eight strictly chosen simple sequence repeats (SSRs) to classify the species into NZ, MY (i.e. MYD), and DQ stocks15. Single nucleotide polymorphism (SNP), as a new generation of genetic markers, has also been used to evaluate the genetic structure of populations. Principal component analysis (PCA) based on SNP data reveals that large yellow croaker populations can be divided into three stocks, i.e., Nanhai (NH, distributed in the South China Sea and the Taiwan Strait), Mindong (MD, distributed near the Taiwan Strait, i.e. MYD or MY) and Daiqu (DQ, distributed in the East China Sea)16. Chen et al. also reported that climate change drove the boundary between Naozhou (i.e. Nanhai) stock and Min-Yuedong (i.e. Mindong) stock might have moved northwards from the Pearl River Estuary to the northern part of the Taiwan Strait, accompanied by highly asymmetric introgression16. A series of studies have shown that there are differences in genetic information among the three geographical stocks of large yellow croaker. However, so far, only the chromosome-level reference genome of Mindong stock has been reported12. The lack of high-quality reference genomes of the other two stocks hinders the assessment of the fine-scale genetic structure of large yellow croaker and adversely affects the population genetic and evolutionary studies, wild stock conservation, fishery management, and germplasm exploitation.

To fill the gap, we sequenced and constructed reference genomes for L. crocea from three geographical stocks of Nanhai (NH), Mindong (MD), and Daiqu (DQ). Using a combination of the PacBio single-molecule real-time sequencing technique (SMRT) and high-through chromosome conformation capture (Hi-C) technologies, we assembled each genome at the chromosome level with a total length of 706.71–722.73 Mb, a contig N50 of 19.46–29.71 Mb, a scaffold N50 of 27.00–27.83 Mb, and a complete BUSCO value of 98.5%-98.7%. To supply materials for gene annotation and conduct functional analysis, we additionally sequenced the transcriptomes of 4 tissues for each of the three stocks of L. crocea. A total of 209.84–219.64 Mb (29.69%-30.39% of the assemblies) of repeat content, 26,851–28,133 protein-coding genes, and 21,156–33,583 ncRNAs were identified. In conclusion, this study reports high-quality chromosome-level reference genomes of different geographical stocks of large yellow croaker for the first time, serving as valuable genomic resources for large yellow croaker and providing a vital reference for future research on the conservation and utilization of genetic diversity.

Methods

Sample collection and nucleic acid extraction

Three healthy female L. crocea from NH, MD, and DQ stocks were obtained from Dongshan Sea (Huizhou, Guangdong), Sandu Bay (Ningde, Fujian), Xiangshan Bay (Ningbo, Zhejiang) in China, respectively (indicated in Fig. 1). Muscle, brain, liver, and spleen were sampled from each fish. All samples were snap-frozen in liquid nitrogen and stored at −80 °C to preserve nucleic acid integrity. Genomic DNA (gDNA) of L. crocea was extracted from muscle tissues using an SDS-based DNA extraction method17, while total RNA was extracted from the brain, spleen, liver, and muscle by a TRIzoL kit (Invitrogen, CA, USA) and mixed at an equal concentration for transcriptome sequencing. The quality of gDNA was assessed by 1.5% agarose gel electrophoresis and DNA was quantified by a Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA). The quality of RNA was assessed by the Fragment Analyzer 5400 (Agilent Technologies, CA, USA).

Fig. 1
figure 1

Geographical distribution of three large yellow croaker stocks. The sampling points of the three fish sequenced in this paper are marked as red dots. NH, Nanhai; MD, Mindong; DQ, Daiqu.

Full size image

Library construction and sequencing

For HiFi read generation, high-molecular-weight (HMW) gDNA was sheared to 15,000–20,000 bp, and the PacBio HiFi library was constructed using the SMRTbell Express Template Prep Kit 2.0 (Pacbio, USA). The genomic library was sequenced in CCS mode on the PacBio Sequel II system at Novogene (Tianjin, China). A total of 52.18–56.59 Gb of CCS clean reads were obtained with an average read length of 7,947–8,284 bp, resulting in 73.84–79.94-fold coverage of the three L. crocea genomes. For Hi-C library construction, the MboI restriction enzyme was used to digest the cross-linked high-molecular-weight (HMW) gDNA. DNA was purified by the phenol-chloroform extraction and randomly sheared into 300–500 bp fragments. After the addition of A-tails to the fragment ends and the subsequent ligation by the Illumina paired-end (PE) sequencing adapters, Hi-C sequencing libraries were amplified by PCR (12–14 cycles) and sequenced on the Illumina platform to obtain PE150 reads. Finally, 43.16–45.78 Gb of paired-end clean reads were generated from the Hi-C library. The RNA-seq library was constructed using the NEBNext UltraTM RNA Library Prep Kit for Illumina (NEB, USA) following the manufacturer’s recommendations and sequenced on the Illumina Novaseq 6000 platform. A total of 24.65–25.58 Gb of paired-end clean reads were generated from the RNA-seq library. For the PacBio Iso-Seq library, RNA was converted into cDNA using the SMARTer PCR cDNA Synthesis Kit (Clontech, USA). The library was then sequenced on the PacBio Sequel II system at Novogene (Tianjin). As shown in Table 1, 30.10–60.44 Gb of long reads were obtained from the sequencing of Iso-Seq libraries.

Table 1 Summary of genome sequencing data generated with multiple sequencing technologies.
Full size table

Genome assembly

HiFiasm18 (v0.18.5) was used to generate contig-level genomes based on the HiFi long reads with default parameters, resulting in three preliminary assemblies of each of the three stocks. For NH stock, the assembly contained 100 contigs, with a total length of 722.73 Mb and a contig N50 of 29.71 Mb. For MD stock, the assembled genome size was 707.86 Mb, including 87 contigs, with a contig N50 of 22.39 Mb. For DQ stock, we gained the preliminary assembly with a total length of 706.71 Mb, including 101 contigs, with a contig N50 of 18.46 Mb. The complete BUSCO value ranged from 98.5% to 98.7%, indicating eximious assembly integrity (Table 2).

Table 2 Statistics of the genome assemblies of three geographical stocks of L. crocea.
Full size table

Hi-C sequencing data was provided to Juicer19 (v1.11.08) and 3D-DNA pipeline20 for chromosome-level genome assemblies of the three L. crocea stocks. We mapped Hi-C clean reads to the preliminary genomes by Juicer. Then, the genomic proximity signal in the Hi-C datasets was used to obtain the chromosome-level scaffolds. Subsequently, the 3D-DNA pipeline was used for scaffolding the genomes. Ultimately, chromosome-level genomes were finalized by using Juicebox21 to adjust misjoins, translocations, inversions, and chromosome boundaries. The size of each chromosome-level genome assembly was estimated to be 636.97 Mb (NH stock, 88.13% of the total length of contigs), 656.91 Mb (MD stock, 92.80% of the total length of contigs), and 644.04 Mb (DQ stock, 91.13% of the total length of contigs). All three assemblies contained 24 chromosomes, with an average chromosome length of 26.54 Mb, 27.37 Mb, and 26.83 Mb, respectively (Table 3).

Table 3 Assembly summary of the 24 chromosomes in each geographical stock of L. crocea.
Full size table

Annotation of repetitive sequences

The repetitive sequences of three L. crocea genomes were identified using both homology-based and de novo strategies. First, RepeatModeler22 (v2.0.1) was utilized to detect repetitive sequences and generate a de novo repeat library. Then, unknown repeats were classified by TEclassTest.pl in TEclass23 (v2.1.3). In another way, repeat elements were forecasted based on the library from RepeatMasker24 (v4.1.2). Finally, all repetitive regions were integrated and masked. Totally, 219.63 Mb (30.39% of the assembled genome), 211.82 Mb (29.92%), and 209.84 Mb (29.69%) of consistent and non-redundant repeat sequences were obtained from genomes of NH, MD, and DQ stocks, respectively. The most abundant repetitive elements for three genomes were DNA transposons. Notably, 47.74 Mb LTRs (6.6% of the assembled genome) were identified from the genome of NH stock, which was higher than the other two stocks (Table 4, Fig. 2A–C).

Table 4 Classification of repetitive elements in three geographical stocks of L. crocea genomes.
Full size table
Fig. 2
figure 2

Gene and repetitive element annotations of three L. crocea genomes. A–C: Distribution of divergence rate for TEs in Nanhai (NH) genome (A), Mindong (MD) genome (B), and Daiqu (DQ) genome (C). D–F: Venn diagrams of functional annotation based on different databases in NH genome (D), MD genome (E), and DQ genome (F).

Full size image

Genome annotation

For noncoding RNA (ncRNA) annotation, Infernal25 (v1.1.4) was utilized based on the Rfam database (http://eggnogdb.embl.de/). Five types of ncRNA were identified from the L. crocea genomes. For NH stock, 7,110 tRNAs, 1,699 miRNAs, 23,465 rRNAs, 1,305 snRNAs, and 4 lncRNAs were identified. For MD stock, there were 6,206 tRNAs, 2,032 miRNAs, 11,805 rRNAs, 1,109 snRNAs and 4 lncRNAs. For DQ stock, 6,331 tRNAs, 2,085 miRNAs, 15,114 rRNAs, 960 snRNAs, and 3 lncRNAs were identified (Table 5, Fig. 3).

Table 5 Classification of ncRNAs in three geographical stocks of L. crocea genomes.
Full size table
Fig. 3
figure 3

Circos plot of the reference genomes of three geographical stocks of L. crocea. Circos plot of 24 chromosome-level scaffolds, representing annotation results of genes and ncRNA of (A) Nanhai (NH) stock, (B) Mindong (MD) stock, (C) Daiqu (DQ) stock. The tracks from inside to outside are ncRNA abundance of the positive strand, ncRNA abundance of the negative strand, GC content, gene abundance of the positive strand, gene abundance of the negative strand, and 24 chromosome-level scaffolds.

Full size image

For gene structure annotation, both ab intio and transcriptome-based strategies were used for gene prediction in the genome of each stock after soft-masking all repeat regions. RNA-seq data was assembled into transcripts using Trinity26 (v2.8.5). For the ab intio approach, BRAKER227 was executed based on the transcripts assembled from RNA-seq and known genes of L. crocea and Cyprinus carpio. For the transcriptome-assisted approach, RNA-seq data was aligned to the genomes to assemble into transcriptome by HISAT228 (v2.2.1) and StringTie29 (v2.1.4). After that, the open reading frame (ORF) was predicted via TransDecoder (https://github.com/TransDecoder/TransDecoder) (v5.5.0). Eventually, EvidenceModeler (v1.1.1) was adopted to produce comprehensive gene sets, which was further annotated for protein-coding gene structure by PASA30 (v2.4.1). As a result, we predicted 36,151 (NH stock), 36,151 (MD stock), and 33,013 (DQ stock) protein-coding genes, respectively, which were subsequently used for functional annotation (Table 6).

Table 6 Gene structure and function annotation in three geographical stocks of L. crocea genomes.
Full size table

For functional annotation of protein-coding genes, Diamond31 (v2.0.6) was applied to align protein-coding genes to the NCBI nr, TrEMBL (http://www.uniprot.org/), and Swiss-Prot (http://www.uniprot.org/) protein databases with the threshold of E-values less than 1 × 10−5. The annotation of GO and KEGG pathways was performed using InterProScan32 (v5.53) and the online website KEGG Automatic Annotation Server33 (KAAS, https://www.genome.jp/tools/kaas/). After integration and de-redundancy, a total of 28,133 (NH stock), 26,851 (MD stock), and 27,279 (DQ stock) protein-coding genes were annotated (Table 6, Figs. 2D–F, 3).

Data Records

All the PacBio long DNA reads, Hi-C reads, Illumina short RNA reads, and PacBio long RNA reads are available from NCBI via the accession numbers SRR29302595-SRR2930260634. The assembled genomes of NH, MD, and DQ have been deposited at Genbank under the accession numbers of JBEDUZ00000000035, JBEDUY00000000036, and JBEDUX00000000037, respectively. Moreover, the assembled genomes and genome annotation are available on Figshare38,39,40.

Technical Validation

Evaluation of genome assemblies and annotation

To ensure the accuracy and integrity of the assemblies, we assessed the completeness of the final genome assemblies using Benchmarking Universal Single-Copy Orthologues (BUSCO)41 with the Actinopterygii_odb10 lineage database. Out of 3,640 single-copy orthologues, all of the three assemblies have > 98.5% BUSCO completeness, which are comparable to those of NH (98.7%), MD (98.5%), DQ (98.6%) (Supplementary Table 1). We additionally used merqury42 (version 1.3) to assess the quality of the three assemblies. The results revealed exceptionally low error rates across all three genomes, with QV values exceeding 60 (Supplementary Table 2), which is sufficient to indicate that all three assemblies are of good quality.

We plotted the Hi-C interaction heatmaps of the chromosomes in each of the three assemblies (Supplementary Fig. 1). The contigs were anchored on 24 chromosomes in all three genomes, which formed squared boxes along the main diagonal of the heatmap matrix. The assembled chromosomes in each of the three assembled genomes display high collinearity with those of the large yellow croaker reference genome published in 201912, indicating that the structures of the assembled genomes are consistent. Meanwhile, for all three genomes, the number of gaps in the assembled chromosomes was significantly reduced compared with the genome released in 201912 (Supplementary Fig. 2). No gap was observed in chr9 and chr10 of the NH genome and chr6 of the MD genome. These results demonstrate improvements in the assembly and anchoring completeness of the new genomes.

Moreover, we successfully obtained a total of 36,151, 36,151, and 33,013 protein-coding genes by combining ab initio strategies and transcriptome-assisted approaches in NH, MD, and DQ genomes, respectively. A total of 28,133, 26,851, and 27,279 genes were functionally annotated in at least one of these databases in three genomes (Fig. 2D–F, Table 6). Taken together, these results suggest that the three assembled L. crocea genomes were of superior quality.