Achieving Chromosome-Level Genome Assembly of Onychostoma macrolepis with PacBio Sequencing and Hi-C Technologies

Abstract

Onychostoma macrolepis is an emerging commercial fish with high edible value. Due to overfishing and environmental pollution, the wild population of Onychostoma macrolepis is declining sharply. In this study, we successfully assembled a chromosome-level genome for Onychostoma macrolepis using a comprehensive approach that integrated Illumina sequencing, PacBio sequencing, and Hi-C technology. The genome spans is 944.02 MB, the contig N50 is 32.36 Mb, 99.15% of the assembled sequences are anchored on 25 chromosomes. A total of 26,294 protein-coding genes were predicted, among which 96.30% were functionally annotated. The high-quality genome data of O. macrolepis obtained in this study provide reliable resources for further research and restoration of wild germplasm resources.

Background & Summary

Largescale shoveljaw fish (Onychostoma macrolepis, Bleeker, 1871, NCBI Taxonomy ID: 369639), Cypriniformes, Cyprinidae, Barbinae, Onychostoma¹, is an emerging commercial fish, distributed mainly in the river system of the Yangtze River, the Huaihe River, the Weihe River and the Haihe River in China. Onychostoma macrolepis is popular among children, the elderly and especially pregnant women because of its tender meat, delicious taste and medicinal value. It is twenty times more expensive than common carp (Cyprinus carpio). In recent decades, because of overfishing and environmental pollution, wild populations of O. macrolepis have declined sharply, and the wild population of O. macrolepis is listed as a second-class protected animal in the List of Key Protected Wild Animals of China². Therefore, to meet the market demand, O. macrolepis farming has received more and more attention.

In order to protect the germplasm resources of this fish and develop its industrialization, there are many fundamental research areas including immunity^3,4, metabolism⁵, feed nutrition^6,7, reproduction mechanism^8,9 and so on. In addition, we found that the fish grew slowly, with the body length of adult fish being only about 20 cm, which showed sexual dimorphism. Adult female O. macrolepis grow twice as large as their male siblings¹⁰. For the further development of relevant studies, especially in molecular breeding fields like all-female breeding and the establishment of fast-growing strains, possessing high-quality assembled genomes of O. macrolepis serves as a crucial and indispensable foundation.

In our current study, a comprehensive strategy was adopted, which combined PacBio long-read sequencing, chromosome conformation capture (Hi-C), and Illumina paired-end sequencing for initial assembly and polishing, with Oxford Nanopore Technologies (ONT) ultra-long reads used for gap filling. As a result, we successfully assembled a high-quality reference genome of O. macrolepis at the chromosomal level. The assembly spans 944.02 Mb with contig N50 of 32.36 Mb and scaffold N50 of 36.04 Mb, organized into 118 contigs and 79 scaffolds, of which 936.02 Mb (99.15%) were anchored onto 25 chromosomes in size from 26.85 to 61.86 Mb. According to the BUSCO assessment (97.9%) and CEGMA assessment (95.56%), along with the QV value of 48.86 and the re-mapping coverage of 99.59% for Illumina paired reads, it was shown that the assembly of O. macrolepis exhibited good completeness. In total, 26,294 protein-coding genes were predicted in the genome, and ~96.30% of the genes were functionally annotated. In addition, we identified 444.62 Mb (47.07% of the genome) of repetitive sequence, and the non-coding RNA prediction enabled the identification of 25,255 tRNAs, 19,839 rRNAs, 2,926 miRNAs, and 2,695 snRNAs, in the genome. The quality of our O. macrolepis genome shows notable enhancement compared with the genome of the previous study, and more protein-coding genes and non-coding RNAs have been annotated¹¹. The genomic data in this study provide a reliable resource for the further study of fish biology and genetics, restoration of wild germplasm resources and other aspects.

Methods

Sample collection and construction of sequencing libraries

The fish was dissected promptly following anesthesia with MS-222. Muscle sample was collected from a healthy O. macrolepis. Genomic DNA was extracted using Qiagen DNeasy Kit (Qiagen, Shanghai, China). In line with the protocols corresponding to different sequencing strategies, one Illumina paired-end sequencing library with an insert size of 350 bp, one SMRT bell sequencing library with a 15 kb insert size, and one Hi-C library were constructed. Subsequently, the Illumina paired-end sequencing library and the Hi-C library were sequenced on the Illumina HiSeq platform, while the SMRT bell sequencing library was sequenced on the PacBio Sequel IIe platform. The raw sequence reads generated by the Illumina platform, which had adapters or contained N bases exceeding 10% or had low-quality bases ($\le$5) accounting for more than 20%, were filtered out^12,13. Finally, a total of 55.10 Gb Illumina paired-reads and 112.29 Gb Hi-C data were obtained. The subreads of PacBio data were filtered using default parameters. Subsequently, the ccs software (https://github.com/PacificBiosciences/ccs) was employed to generate high-precision HiFi reads. In total, 2,794,360 reads with an average read length of 17,051 bp and N50 of 16,905 bp were obtained and used for further assembly of O. macrolepis genome (Table 1). Nine tissues including blood, brain, kidney, gonad, skin, gill, swim bladder, intestine and spleen were sampled for the same fish individual. Subsequently, RNA was extracted from all nine samples with the RNAiso Pure RNA Isolation Kit (TaKaRa, Japan), after which DNaseI treatment was carried out. The concentration and quality of the obtained RNAs were evaluated using Nanodrop (Thermo Fisher Scientific, USA). Afterwards, all the RNA samples were sequenced on the Illumina NovaSeq. 6000 to generate 150 bp paired-end reads (Table 1).

Table 1.

The output quality of O. macrolepis transcriptome data.

Sample	Raw Reads	Clean Reads	Raw Base (G)	Clean Base (G)	Q20 (%)	Q30 (%)	GC Content (%)
SM1-sp-A	25,370,225	23,729,684	7.61	7.12	97.22	92.38	44.12
SM1-G-A	39,240,210	35,746,754	11.77	10.72	97.16	92.31	45.48
SM1-br-A	51,598,038	47,140,625	15.48	14.14	96.91	91.73	45.16
SM1-Sk-A	48,783,650	43,486,340	14.64	13.05	96.83	91.5	46.33
SM1-Gi-A	21,513,134	21,018,129	6.45	6.31	97.16	92.13	46.27
SM1-Sw-A	21,539,751	21,049,229	6.46	6.31	97.48	92.85	48.34
SM1-In-A	37,950,784	35,102,149	11.39	10.53	97.09	92.14	47.14
SM1-K-A	24,020,715	22,059,731	7.21	6.62	97.41	92.78	45.27
SM1-bl-A	24,832,153	21,761,636	7.45	6.53	97.78	93.61	47.49

For Oxford Nanopore Technologies (ONT) ultra-long reads (which are based on the Single molecule nanopore DNA sequencing method developed by ONT), DNA required for Nanopore sequencing was isolated from muscle samples of a healthy O. macrolepis. The Qiagen DNeasy kit was used for extraction, following the manufacturer’s provided instructions. The extracted genomic DNA underwent quality assessment for purity, concentration, and integrity using Nanodrop, Qubit, and 1% agarose gel electrophoresis. Subsequently, the DNA was fragmented with the Diagenode Megaruptor. Fragment selection was performed using BluePippin. Finally, the constructed library was sequenced on the PromethION-specific Flow Cell (R9.4). Fastq files were generated from raw Nanopore data using Guppy and quality control was subsequently performed on the data using Nanoplot (v1.38.1). A total of 1,207,111 reads (42.67 Gb) with an average read length of 35,349 bp and N50 of 52,217 bp were passed for further assembly (Table 2).

Table 2.

Statistics of genome sequence of O. macrolepis.

platform	PacBio-HiFi	ONT ultra-long	HiC	Illumina
Total read bases (Gb)	46.04	42.67	112.29	55.10
Read number	2,794,360	1,207,111		—
Read N50 (bp)	16,905	52,217		—
Mean read length (bp)	17,051	35,349		—
Coverage depth (×)	48.64	45.08	118.63	58.27

Genome size estimation and assembly

According to the k-mer analysis (k-mer = 17) conducted using the software Jellyfish 2.2.7¹⁴, the genome size of O. macrolepis was estimated based on the paired-end Illumina sequencing data. The results indicated that the genome size of O. macrolepis was 945.67 Mb, the heterozygosity rate of the genome was 0.28%, and the proportion of repetitive sequences was 51.55%. Utilizing the long HiFi reads produced by the PacBio Sequel lle platform, the contig assembly of O. macrolepis genome was carried out with Hifiasm 0.16.1¹⁵. Subsequently, the primary contigs underwent polishing with PacBio reads via Arrow Smartlink 8.0¹⁶. Next, the Purge Haplotigs software¹⁷ was employed to eliminate redundant contigs from the initial assembly, yielding a non-redundant contig assembly. In the end, the short reads from Illumina were utilized to rectify any remaining errors in the contig assembly through Pilon 1.22 software¹⁸. These processes would yield a final draft genome of O. macrolepis assembly with a total length of 944.02 Mb and contig N50 of 32.36 Mb (Table 3). To generate a chromosome-level genome assembly, Hi-C technology was employed to anchor the contigs onto chromosomes. By means of the hierarchical clustering algorithm, the contigs were clustered, ordered and oriented. Subsequently, they were anchored into 25 pseudo-chromosome linkage groups using ALLHiC 0.9.8, and a total of 936.02 Mb assembled genome sequences were anchored on 25 chromosomes, with an average chromosome length of 37.44 Mb. Finally, gap filling was carried out using the ultra-long read length of Oxford Nanopore Technologies (ONT) through the TGS-GapCloser¹⁹, and as a result, the number of gaps was reduced from 39 to 4. (Table 3, Fig. 1).

Table 3.

Summary of our O. macrolepis genome and published O. macrolepis genome assembly.

Parameter	O. macrolepis (this study)	O. macrolepis (previous study)
Assembly strategy	PacBio-HiFi + ultra-ONT + Hi-C	PacBio + Hi-C
Total Size (bp)	944,024,602	894,945,186
Anchored onto chromosomes (bp)	936,022,107	889,754,504
Sequences assigned to chromosemes (%)	99.15	99.42
Gap Number	4	—
No. of contig	118	689
Contig N50 length (bp) /rank	32,362,784/13	20,336,701/16
Contig N90 length (bp) /rank	11,335,856/30	4,120,503/56
No. of scaffold	79	241
Scaffold N50 length (bp) /rank	36,044,644/11	35,346,331/11
Scaffold N90 length (bp) /rank	32,328,402/22	31,713,302 /21
BUSCO (%)	98.20	97.20
CEGMA (%)	95.56	94.35
Quality Value (QV)	48.86	—
Mapping rate of Illumina data (%)	99.59	98.49
Mapping covarge of Illumina data (%)	99.95	99.23
Protein coding genes	26294	25779
Percentage of functionally annotated genes	96.3	94.70
Repeat bases (Mb)	444.62	424.37
Percentage of repeats (%)	47.07	47.42
tRNAs	25255	3098
rRNAs	19839	410
miRNAs	2926	2654
snRNAs	2695	708

Fig. 1.

Hi-C assembly of chromosome interactive heatmap of O. macrolepis genomes. Chromosome interactive heatmap: Chr1 to Chr25 are the abbreviations for the 25 Chromosomes. The abscissa and ordinate represent the order of each bin on the corresponding chromosome group; The colour block illuminates the intensity of interaction from yellow (low) to red (high). Window size = 500 kb.

The quality of the assembled genome was evaluated through four methods. Specifically, the Benchmarking Universal Single-Copy Orthologs analysis (BUSCO, v5.2.2)²⁰ and the Core Eukaryotic Genes Mapping Approach analysis (CEGMA, V2.5)²¹ were utilized to assess the integrity of the genome. For BUSCO, this was accomplished by searching against 3640 BUSCOs of the actinopterygii_odb10, while for CEGMA, it was performed by selecting the conserved genes (248 genes) existing in six eukaryotic model organisms. The Illumina paired reads were realigned to the assembled genome using the BWA 0.7.8²² software. The realignment ratio and coverage depth were then computed to evaluate the integrity of the assembly and the uniformity of the sequencing; and the package Merqury²³ was employed for assessing the accuracy of genome assembly using a reference-free and k-mer-based approach.

Genome annotation of O. macrolepis

In our repeat annotation pipeline, a combined strategy relying on homology alignment and de novo search was applied to identify the whole genome repeats. Tandem Repeat was extracted using TRF v4.09²⁴ by ab initio prediction. The homolog prediction utilized Repbase database²⁵ employing RepeatMasker v4.1.0²⁶ software and its in-house scripts (RepeatProteinMask v4.1.0) with default parameters to extracted repeat regions. For ab initio prediction, a de novo database of repetitive elements was built using LTR_FINDER v1.06²⁷, RepeatScout v1.0.5²⁸, and RepeatModeler v2.0.1²⁹ with default parameters. Subsequently, all repeat sequences with lengths > 100 bp and gap ‘N’ less than 5% constituted the raw transposable element (TE) library. A custom library, which combined Repbase with our de novo TE library (processed by uclust to make it non-redundant), was provided to RepeatMasker v4.1.0 for identifying DNA-level repeats. Based on these analyses, approximately 444.62 Mb of repeat sequences were ultimately uncovered, making up 47.07% of the entire O. macrolepis genome (Table 4).

Table 4.

Summary of the TEs in O. macrolepis genome.

Type	Repbase + Denovo		TE Protiens		Combined TEs(all without trf)
	Length (Bp)	% in genome	Length (Bp)	% in genome	Length (Bp)	% in genome
DNA	70,667,895	7.48193	6,797,579	0.71969	73,713,681	7.804401
LINE	20,795,259	2.201688	26,746,203	2.831742	38,099,063	4.03372
SINE	858,828	0.090928	0	0	858,828	0.090928
LTR	331,980,268	35.148255	15,783,509	1.671072	333,349,095	35.293179
Satellite	2,349,268	0.248728	0	0	2,349,268	0.248728
Simple_repeat	3,552	0.000376	0	0	3,552	0.000376
Unknown	18,131,052	1.919617	0	0	18,131,052	1.919617
Total	430,957,298	45.627402	49,317,394	5.221456	434,587,509	46.011749

Three methods (ab initio prediction, homology-based prediction and RNA-Seq assisted prediction) were employed to predict the protein-coding genes in the O. macrolepis genome. For homology-based prediction, the reference protein sequences of four species (Table 5) were downloaded from the NCBI database. Subsequently protein sequences were aligned to the O. macrolepis genome using TblastN(v2.2.26; E-value ≤ 1*e^{−5 30}), the gene statistics, including gene length, coding sequence (CDS), intron length, and exon length, were comparable between these species and the O. macrolepis genomes (Fig. 2).

Table 5.

Statistics of gene structure in O. macrolepis genomes.

Gene set		Number	Average transcript length(bp)	Average CDS length(bp)	Average exons per gene	Average exon length(bp)	Average intron length(bp)
De novo	Augustus	38567	9562.43	1149.98	6.53	176.23	1522.45
	GlimmerHMM	66212	9490.28	620.71	4.22	147.23	2757.95
	SNAP	63070	13538.17	712.09	5.32	133.92	2970.85
	Geneid	30989	18493.87	1361.98	6.42	212.12	3160.48
	Genscan	31303	20538.98	1539.5	8.23	187.09	2628.42
Homolog	Drer	22811	13389.39	1592.77	8.73	182.35	1525.17
	Ccar	23679	13539.57	1579.98	8.74	180.74	1544.83
	Anig	28372	11209.33	1310.14	7.48	175.21	1528.28
	Amex	22060	13271.45	1573.65	8.55	183.97	1548.62
RNAseq	PASA	74320	10969.31	1263.94	7.75	163.13	1438.24
	Transcripts	51883	27841.98	4318.61	12.79	337.74	1995.75
EVM		35920	12348.78	1269.24	7.37	172.3	1740.28
Pasa-update*		35627	12688.85	1291.61	7.43	173.75	1771.53
Final set*		26294	15557.19	1575.97	9.13	172.7	1720.71

Fig. 2.

Comparisons of the lengths of genes, CDSs, exons, and introns between O. macrolepis and closely related species, along with a Venn diagram showing the number of genes annotated by different gene databases in the O. macrolepis genome. Amex represents A. mexicanus, Anig represents Ancherythroculter nigrocauda, Ccar represents C. carpio, Drer represents D. rerio, and duolinbaijiayu represents the studied species O. macrolepis.

The matching proteins were aligned to the homologous genome sequences via GeneWise (v2.4.1)³¹ software for accurate spliced alignments and to predict the gene structure within each protein region. Regarding ab initio gene prediction, five tools, namely Augustus (v3.2.3)³², Geneid (v1.4)³³, Genescan (v1.0)³⁴, GlimmerHMM (v3.04)³⁵ and SNAP (V2013.11.29)³⁶ were utilized in the gene prediction pipeline. For RNA-Seq assisted prediction, transcriptome reads assemblies were generated with Trinity (v2.1.1) for the genome annotation. To optimize the genome annotation, the RNA-Seq reads from nine different tissues were aligned to the genome fasta using TopHat (v2.0.11) with default parameters³⁷ to identify exons region and splice positions. Subsequently, Cufflinks (v2.2.1) was used for reconstruction and assembly of transcripts³⁸. The non-redundant reference gene set was generated by merging genes predicted by three methods with EvidenceModeler (EVM,v1.1.1) using PASA(Program to Assemble Spliced Alignment)³⁹ terminal exon support and including masked transposable elements as input into gene prediction. Overall, 26,294 protein-coding genes were successfully identified in the O. macrolepis genome, respectively (Table 3).

Gene functions were assigned according to the best match by aligning the protein sequences to the Swiss-Prot using Blastp (with a threshold of E-value ≤ 1e⁻⁵). The motifs and domains were annotated using InterProScan70 (v5.31)⁴⁰ by searching against publicly available databases, including ProDom, PRINTS, Pfam, SMRT, PANTHER, and PROSITE. The Gene Ontology (GO) IDs⁴¹ for each gene were assigned according to the corresponding InterPro entry. The protein function was predicted by transferring annotations from the closest BLAST hit (E-value < 10⁻⁵) in the Swissprot database⁴² and BLAST hit (E-value < 10⁻⁵) in the NR database. Additionally, the gene set was mapped to a KEGG pathway and identified the best match for each gene⁴³. In summary, a total of 25,327 genes (96.30%) were successfully annotated by function implications or the conserved functional motifs in O. macrolepis genomes (Table 6, Fig. 2).

Table 6.

Summary of functional annotations for predicted genes of O. macrolepis genome.

Annotation Database	Number of Annotated Genes	Percentage (%)
Swissprot	21,539	81.9
Nr	24,617	93.6
KEGG	21,609	82.2
InterPro	24,533	93.3
GO	16,620	63.2
Pfam	20,509	78
Annotated	25,327	96.3
Unannotated	967	3.7
Total	26,294	—

For non-coding RNA annotation, the tRNAs were predicted using the program tRNAscan-SE v1.4⁴⁴. Due to rRNAs are highly conserved, the relative species’ rRNA sequences were chosen as references to predict rRNA sequences by using Blast with an E-value of 1*E⁻¹⁰⁴⁵. Other ncRNAs, including miRNAs and snRNAs, were identified by searching against the Rfam database with default parameters using the infernal software⁴⁶.

Data Records

The genomic data of O. macrolepis have been submitted to NCBI with the respective accession numbers, PRJNA1212228⁴⁷ (BioProject) and SAMN46295721⁴⁸ (BioSample). Furthermore, various data from Illumina sequencing, Hi-C sequencing, PacBio sequencing, RNA-seq data and ONT reads have been submitted to NCBI SRA with the accession numbers SRR32040837⁴⁹, SRR32040935⁵⁰, SRR32355807⁵¹, SRR32040922⁵²-SRR32040930⁵³, SRR33977354⁵⁴ respectively. Additionally, the final chromosome-level genome assembly of O. macrolepis has been uploaded to GenBank in NCBI, bearing the accession number JBMGJU000000000⁵⁵.

Technical Validation

The quality of the assembled genome was evaluated through four methods. Specifically, the Benchmarking Universal Single-Copy Orthologs analysis (BUSCO, v5.2.2) and the Core Eukaryotic Genes Mapping Approach analysis (CEGMA, V2.5) were utilized to assess the integrity of the genome. For BUSCO, this was accomplished by searching against 3640 BUSCOs of the actinopterygii_odb10, while for CEGMA, it was performed by selecting the conserved genes (248 genes) existing in six eukaryotic model organisms, the results showed that 98.20% and 95.56% of BUSCOs and CEGMAs recovered in genome assemblies of O. macrolepis, respectively (Table 2). The Illumina paired reads were realigned to the assembled genome using the BWA 0.7.8 software. 99.59% Illumina short-reads were mapped the genome assembly of O. macrolepis and the coverage reached to 99.95%. The package Merqury was employed for assessing the accuracy of genome assemblies using a reference-free and k-mer-based approach, which resulted in a high assembly quality value (QV) of 48.86 for O. macrolepis genome and exceed the Vertebrate Genome Project (VGP) standard of QV40⁵⁶. Overall, the genome assemblies of O. macrolepis in the present study showed excellent completeness and superior accuracy.

Author contributions

Chao Li conceived the research. Lanzhu Zou was involved in sample collection. Lu Zhang and Ying Zhu contributed to genome assembly and annotation, data analysis and wrote the manuscript. Min Cao and Chao Li revised the manuscript. All authors read and approved the final manuscript.

Code availability

All data processing commands and pipelines were carried out in strict accordance with the manuals and protocols of the relevant bioinformatics software. No custom script or code was used in the study.

Competing interests

The authors declare no competing interests.