Background & Summary

Sweet orange (SWO, Citrus sinensis L. Osbeck) is the most important citrus species1. SWO varieties are typically categorized into three subgroups based on their agronomical characteristics: common orange, navel orange, and blood orange (BO)2. BO stands out for brilliant red coloration of both flesh and rinds3, which is not usually found in Citrus L.4,5.

Anthocyanins, which belong to a large family of flavonoids, are accountable for the characteristic red color of BO3. In addition to contributing to pigmentation3, anthocyanins have various health-promoting benefits in humans, such as their antioxidant capacity and potential for cancer prevention6. As consumers become increasingly health-conscious, the popularity of BO has been growing worldwide7 because of its exceptional nutraceutical attributes, including vitamins, sugars, dietary fiber, minerals, and flavonoids, particularly anthocyanins8.

Moro, Tarocco, and Sanguinello are the three most important commercial BO types9. Moro has the deepest red color among the three varieties, followed by Sanguinello and Tarocco4,9. Tarocco is a medium-sized seedless variety famous for its peelability and sweetest taste2. In our long-term BO breeding program, we have discovered an unexpected and natural bud mutation of Tarocco, which we have named ‘内秀’ (Neixiu, NX). In Chinese wisdom, ‘内秀’ is used to describe a person who looks pretty ordinary, but he is intelligent in an understated way. Based on more than 5 years of careful observation, we found that NX surpasses common Tarocco in terms of both sweetness and redness in the Southwest region of China (Fig. 1). Consequently, we consider NX to be a highly promising BO cultivar.

Fig. 1
figure 1

Morphological and genomic characteristics of Neixiu blood orange. (a) Fruit phenotypes of Neixiu (left) and Tarocco (right) blood oranges. (b) Genomic landscape of Neixiu blood orange, including chromosome ideogram, transposon element density, gene density, GC content, and intra-genome collinear blocks.

Full size image

Recent advancements in sequencing technology and associated bioinformatic tools have significantly expedited citrus genomic studies. To date, three genomes of the SWO variety have been released. In 2013, the first draft of a di-haploid SWO genome was complied using short Illumina reads10. Subsequently, Wang et al.11 successfully generated a de novo reference genome of the di-haploid SWO using the Nanopore ultra-long and PacBio long-read sequencing platforms. More recently, Wu et al. 12 accomplished the assembly of a diploid SWO genome at the chromosome level, specifically for the ‘Valencia’ variety. However, it is worth noting that genomic data for this important BO in the citrus industry is currently unavailable. In the investigation of BO functional genomics and genetics, the initial task involves the interpretation of genomic data.

Therefore, in the present study, we constructed a high-quality chromosome-scale genome assembly of BO by combining Illumina sequencing, third-generation circular consensus sequencing (CCS), and high-throughput chromosome conformation capture (Hi-C) sequencing. This integrated methodology resulted in a genome size of approximately 336.63 Mb, with a contig N50 value of 30.6 Mb. A total of 96% of the assembled sequences were successfully anchored to nine pseudo-chromosomes (Table 1). To investigate the evolutionary patterns of genes and genomes, comparative genomic analyses were performed on the BO genome and 11 other genomes representing various plant species. The study presents the first high-quality chromosome-scale genome of BO. The dataset generated from this research will significantly contribute to the advancement of our knowledge in BO functional genomics and the trajectory of citrus genomes.

Table 1 Assembly and assessment of Neixiu blood orange genome.
Full size table

Methods

Plant materials

For genome sequencing, young leaf samples were randomly collected from five-year-old NX trees. Samples were immediately frozen in liquid nitrogen, followed by preservation at −80 °C until DNA and RNA extraction. For RNA extraction, fresh plant tissues including leaves, fruits, buds, roots, and branches were obtained from the same tree. The ‘Valencia’ SWO11 was used in the bioinformatics analysis.

Library construction and sequencing

Genomic DNA and total RNA were extracted using DNeasy Plant Mini Kit and RNeasy Plus Mini Kit (Qiagen, Valencia, CA, USA), respectively, according to the manufacturer’s instructions. After extraction, short-read (350-bp) libraries were constructed using a library construction kit (Illumina, San Diego, CA, USA) and then sequenced on a Novaseq 6000 platform (Illumina), which finally generated a total of 24.21 Gb of raw data, covering 74.66 × of the genome. The resulting clean reads were used for genome surveys, including the evaluation of genome size, GC content, and heterozygosity.

PacBio sequencing libraries were constructed by Biomarker Technologies Corporation (Beijing, China) using the SMRTbell® express template prep kit 2.0 (PacBio, Menlo Park, CA, USA). Before library preparation, genomic DNA was sheared into 15 kb fragments using Megaruptor® 3 (Diagenode, Denville, NJ, USA). A total of 21.21 Gb high-fidelity (HiFi) clean data with an N50 value of 19.36 kb and an average read length of 18.88 kb were produced using the CCS mode on a PacBio Sequel II platform with the Sequel sequencing kit 2.0 (PacBio). These data are equivalent to 65 × coverage of the entire genome.

Hi-C libraries with 300~700-bp insert size were prepared following Rao et al.13 and sequenced on a NovaSeq 6000 platform (Illumina). This sequencing generated approximately 55.6548 Gb reads.

Genome survey and assembly

Illumina short reads were filtered using fastp14 to remove low-quality reads and adapters before genome size estimation. SOAP v.2.2115 was used for the initial assembly. The frequencies of 19 K-mers were determined using Jellyfish v.2.1.416. Based on these analysis, the genome size was estimated to be 324.21 Mb, with a heterozygosity rate of 1.82%, a repeat element ratio of 43.81%, and a GC content of 35.63% (Fig. 2).

Fig. 2
figure 2

Frequency distribution of the 19-mer analysis. The x-axis represented the K-mer depth and y-axis represented the frequency of K-mer correspond to the depth.

Full size image

The HiFi long reads were subjected to genome assembly using Hifiasm v.0.1617, resulting in a contig length of 494.34 Mb and a contig N50 value of 30.18 Mb. Redundant contigs caused by heterozygosity were removed using Purge_dups18, resulting in a contig length of 336.63 Mb and a contig N50 value of 35.13 Mb (Table 1).

Adaptors and low-quality Hi-C reads were filtered using HiC-Pro v.2.10.019, retaining only uniquely mapped paired-end reads with a mapping quality greater than 20. The scaffolds/contigs underwent clustering, ordering, and orientation onto chromosomes using LACHESIS20. Subsequently, any placement or orientation errors that displayed distinct chromatin interaction patterns were manually adjusted. These scaffolds were anchored to nine pseudo-chromosomes, which accounted for 96% of the assembled genome (Fig. 3). The Hi-C scaffolding process ultimately achieved the final chromosome-scale genome assembly of BO (336.63 Mb) with contig and scaffold N50 values of 30.6 Mb (Table 1).

Fig. 3
figure 3

Hi-C interaction heatmap for Neixiu blood orange. The map shows scaffolded and independently assembled chromsomes at high resolution.

Full size image

Repeat element identification

Transposon elements (TEs) were identified by combining de novo and homology-based strategies using RepeatModeler2 v.2.0.421. This involved in the automated execution of two repeat-finding programs (RECON v.1.0.8 and RepeatScout v.1.0.6) and the classification of prediction results using RepeatClassifier21, which entailed a search of Dfam v.3.522. LTRharvest v.1.0623 and LTR_finder v.1.5.1024 were used identify the full-length repeat retrotransposons (LTR-RTs). High-quality intact full-length LTR-RTs and non-redundant LTR libraries were produced from the outputs of LTR_retriever v.2.9.025. By combining the de novo TE library with known TEs in RepBase v.19.0626, REXdb v.3.027, and Dfam v.3.522, a non-redundant species-specific TE library was obtained. The final TEs were identified and classified through a homology search against the library using RepeatMasker v.4.1.428. Tandem repeats were annotated using Tandem Repeats Finder29 and MISA v.2.130. In the BO genome, we identified 127.82 Mb (37.97%) of TEs and 25.72 Mb (7.64%) of tandem repeats. The majority of repeats (28.06%) were Class I retrotransposons, dominated by gypsy (13.04%) and copia (7.52%) elements. Class II DNA transposons accounted for 9.91% of the BO genome (Table 2).

Table 2 Repetitive elements and their proportions in Neixiu blood orange.
Full size table

Protein-coding genes prediction

A total of 30,395 protein-coding genes have been annotated by incorporating de novo, homology, and transcript-based predictions (Table 3). The de novo gene models were predicted using Augustus v.3.2.231 and SNAP v.2006-07-2832. GeMoMa v.1.733 was used for homology-based predictions by annotating the gene models in BO with amino acid sequences from C. grandis, SWO, Poncirus trifoliata, and Arabidopsis thaliana genomes. For transcript-based prediction, RNA-seq data was mapped to the reference genome using HISAT v.2.2.134 and quantified with StringTie v.2.1.435. Genes were predicted from the assembled transcripts using GeneMarkS-T v.5.136. Another transcript-based prediction method was performed using Trinity v.2.1.137. Program to Assemble Spliced Alignments (PASA) v.2.4.138 was used to predict gene models based on the unigenes. The genes predicted in the aforementioned three annotation files were merged using EVidenceModeler v.1.1.139, and the final gene set was updated using PASA v.2.4.138. Each gene exhibited an average of 5.02 exons, with a mean gene length of 3489.94 bp and a coding sequence size of 1152.21 bp. The average lengths of exons and intros were 1440.51 and 2049.43 bp, respectively (Table 3).

Table 3 Genome annotation of Neixiu blood orange.
Full size table

Gene function annotation

To ascertain the functional characteristics, the predicted genes underwent annotation by aligning them with the gene ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), eukaryotic orthologous groups (KOG), protein families (Pfam), SwissProt, TrEMBL, evolutionary genealogy of genes, non-supervised orthologous groups (eggNOG), and NCBI non-redundant protein (Nr) databases. Additionally, the motifs and domains were annotated using InterProscan v.5.27.6640. Based on the aforementioned multiple databases, a total of27,223 genes, accounting for 89.56% of the predicted protein-coding genes, were successfully annotated. Specifically, the GO, KEGG, KOG, Pfam, SwissProt, TrEMBL, Eggnog, and Nr databases annotated approximately 72.6%, 63.79%, 45.6%, 71.71%, 68.2%, 88.79%, 71.99%, and 87.57% of genes, respectively (Table 3).

Non-coding RNA annotation

Transfer RNA (tRNA) and ribosomal RNA (rRNA) were identified using tRNAscan-SE v.1.3.141 and Barmap v.0.9.042, respectively. Furthermore, other non-coding RNAs (ncRNAs), including microRNA (miRNA), small nucleolar RNA (snoRNA), and small nuclear RNA (snRNA), were identified using Infernal v.1.1.243 by searching against Rfam v.14.144. In total, 8,248 ncRNAs (5,339 rRNAs, 475 tRNAs, 162 miRNAs, 905 snRNAs, and 1,367 snoRNAs) were identified in the BO genome (Table 3).

Comparative genomics analysis

An all-against-all protein sequence similarity search was conducted between the BO genome and 11 representative plant species (P. trifoliata, Malus domestica, Arabidopsis thaliana, Solanum lycopersicum, C. sinensis, Oryza sativa, Ziziphus jujuba, C. clementina, Amborella trichopoda, Vitis vinifera, and C. unshiu) using Orthofinder v.2.3.845 with the diamond alignment method. The resulting gene families were then annotated using Panther v.1546. Unique gene families in BO were subjected to GO and KEGG enrichment analysis using ClusterProfiler v.3.14.047.

A total of 40,592 gene families were identified, including 2,571 gene families that shared among these species and 123 that were specific to BO (Fig. 4a). Notably, a significant proportion of the genes in BO and the other 11 species were found to be single-copy genes (Fig. 4b). Among the Rutaceae species, including BO, C. sinensis, C. clementina, C. unshiu, and P. trifoliata, a total of 11,808 gene families were shared with 278 gene families specific to BO (Fig. 4c). Further KEGG analysis revealed that these BO specific genes were significantly enriched in various pathways, such as protein processing in the endoplasmic reticulum, monoterpenoid biosynthesis, and starch and sucrose metabolism (Fig. 4d).

Fig. 4
figure 4

Comparative genomic analysis of Neixiu blood orange and other 11 representative plant species. (a) Gene family cluster petal map of Neixiu blood orange and other 11 representative plant species. The central circle represents common gene families, and the outer petals represent specific gene families. (b) Venn diagram showing gene family clusters of five Rutaceae species. (c) The number of gene copies and their distribution among 12 plant species. (d) KEGG enrichment analysis of genes specific to Neixiu blood orange.

Full size image

Phylogenetic and evolutional analyses

A phylogenetic tree was constructed using IQ-Tree48 based on 594 single-copy gene sequences obtained from these 12 species. The alignment of orthologous gene sequence was performed independently using MAFFT v.7.49049, followed by the conversion of protein alignments to nucleotide sequence alignments using PAL2NAL v.1450. The alignments were then refined using the Gblocks 0.91b51. Clean super-alignments were used to construct a maximum likelihood phylogenetic tree using IQ-Tree48 with a fitted model of GTR + F + I + G4 suggested by ModelFinder52. The resulting tree revealed BO is a sister clade to C. sinensis, indicating a closer relationship with SWO than with mandarins (C. unshiu and C. clementina) (Fig. 5a).

Fig. 5
figure 5

Evolution analyses of Neixiu blood orange and other 11 representative plant species. (a) Phylogenetic tree showing the relationships among 12 species with divergence time. The top and bottom of the tree represent the absolute age (millions of years) and geological time (Neogence, Ne.; Paleogence, Pa.; Cretaceous, Cr.; and Jurassic, Ju.). All the nodes have 100% boostrap support. (b) Phylogenetic tree showing the relationships among 12 species with gene family expansion (green color) and contraction (pink color). MRCA, most recent common ancestor. (c) Genome synteny among Neixiu blood orange, sweet orange, Citrus clementina, Vitis vinifera, and Malus domestica. (d) Distribution of the 4DTv rates among the paralogs of the studied species.

Full size image

The divergence time among the 12 plant species was calculated using MCMCTree in the PAML v.4.953 with 95% confidence intervals. TimeTree54 calibration points were used to infer the divergence time. The calculated divergence times were as follows: C. sinensis-Amborella trichopoda, 179.0–199.1 million years ago (mya); C. sinensis-C. clementina, 1.5–5.7 mya; C. sinensis-O. sativa, 143.0–174.8 mya; C. sinensis-S. lycopersicum, 112.4–125.0 mya; C. sinensis-M. domestica, 102.0–113.8 mya; and C. sinensis-Arabidopsis thaliana, 90.0–99.9 mya. These estimates were subsequently used to correct the fossil time obtained from the software algorithm. Amborella trichopoda was used as the outgroup for conducting maximum-likelihood-based phylogenetic analyses. The divergence time between the SWO and BO (2.24–4.83 mya) was comparatively more recent compared than that of C. unshiu and C. clementina (2.33–4.96 mya), while the divergence time of oranges and mandarins (2.98–5.94 mya) was found to be the earliest among the four Citrus species (Fig. 5a). The gene expansion and contraction of the gene families were determined using Computation Analysis of gene Family Evolution (CAFE)55 v.3.1. In total, 920 and 1,313 gene families expanded and contracted in the BO genome, respectively (Fig. 5b).

Synteny and whole-genome duplication (WGD) analysis

To better understand the evolutionary history of BO, we performed a genomic collinearity analysis of BO, SWO, C. clementina, V. vinifera, M. domestica, and Z. jujube. Homologous gene pairs were identified through a comparison of the genomic sequences of two species using the DIAMOND v.0.9.29.13056. Subsequently, JCVI v.0.9.13 was used to visualize collinear blocks identified using homologous gene pairs in MCScanX57. A significant level of synteny was observed between the genomes of BO and SWO. The BO chromosomes were mapped with more fragments in the SWO than in C. clementina (Fig. 5c).

To determine the occurrence of WGD events, a combination of the synonymous mutation rate (Ks) and fourfold synonymous third-codon transversion (4DTv) was employed. This analysis was conducted using WGD v.1.1.158 and a publicly available script (https://github.com/JinfengChen/Scripts). The 4DTv values of BO, SWO, and C. clementina reached a peak of 0.5, indicating the occurrence of WGD events in Citrus. The Citrus speciation event took place prior to the duplication event observed in Citrus species, evidenced by the pairwise 4DTv distribution of BO compared to M. domestica, V. vinifera, Z. jujuba, and Arabidopsis thaliana (Fig. 5d).

Genome-wide variation analysis

To investigate the genomic differences between BO and SWO, we used the assembled NX as the reference genome and the most recent chromosome-level phased diploid Valencia SWO genome, as published by Wu et al.12, for conducting genome-wide alignments with the nucmer, delter-filter, and show-coord programs from MUMmer v.4.059. This analysis yielded a total of 1,275,362 single-nucleotide polymorphism (SNP) differences and 295,024 insertion-deletions (InDels), including 170,365 insertions and 124,659 deletions. Subsequently, the filtered delta files were subjected to SyRI60 for the identification of structural variations (SVs) in. A total of 694 copy number variations (CNVs) were found in SWO genome compared to the BO genome, with 362 copies increased and 332 copies lost in number in the SWO genome. Presence-absence variations (PAVs) are major contributors to genome structural variations, impacting both phenotypic and genomic variability61. We detected 1,081 present and 1,340 absent variations. GO and KEGG enrichment analyses were conducted using clusterProfiler v.3.14.1047 for genes where mutations were detected. ANNOVAR62 was used for the functional annotation of genetic variants.

Data Records

The genome sequences, PacBio raw data, and Hic-C raw data have been deposited to the NCBI SRA database63,64 and the genome gff annotation file was uploaded to65. Genome estimation, statistics of assembled genome sequences, integrated function annotation, statistics of gene family clustering, and list of the expanded and constracted gene families were submitted at the Figshare66.

Technical Validation

The assessment of the final assembled genome completeness and quality involved the implementation of (1) Core Eukaryotic Genes Mapping Approach (CEGMA) v. 2.567, (2) Benchmarking Universal Single-Copy Orthologs (BUSCO) v. 5.2.168, (3) alignment using Burrows–Wheeler Aligner (BWA)69 with Illumina data, and (4) alignment using Minimap 270 with HiFi reads.

The evaluation of the final assembled genome’s integrity was performed by referencing the CEGMA database, which contains 458 core eukaryotic genes (CEGs) and 248 highly conserved CEGs, and by employing tblastn, genewise, and geneid software67. The assembled genome contained 98.25% (450) of CEGs and 95.16% (236) of highly conserved CEGs, suggesting that it contained most CEGs. To evaluate the integrity of the assembly, BUSCO68 analysis was conducted using the Embryophyta database OrthoDB v. 10 (http://cegg.unige.ch/orthodb), which encompasses 1,614 orthologous single-copy genes. The assembled genome contained 1,585 (98.20%) of these genes. Mapping of Illumina short reads and HiFi long reads to the assembled genome revealed that approximately 97.66% and 99.58% of the reads, respectively, aligned successfully (Table 1).

To ensure the reliability of the MCMCTree analyses, the correlated molecular clock and JC69 model were employed, and all relevant computations were performed twice to ensure consistency. The correlation between two iterations in this test is 1.

In order to evaluate the reliability of the inference in constructing the phylogenetic tree, 1000 bootstrap replicates were performed for each branch.