The telomere-to-telomere genome of flowering cherry (Prunus campanulata) reveals genomic evolution of the subgenus Cerasus

Abstract

Background

Prunus campanulata, a species of ornamental cherry, holds significant genetic and horticultural value. Despite the availability of various cherry genomes, a fully resolved telomere-to-telomere (T2T) assembly for this species has been lacking. Recent advancements in long-read sequencing technologies have made it possible to generate gap-free genome assemblies, providing comprehensive insights into genomic structures that were previously inaccessible.

Findings

We present the first T2T genome assembly for P. campanulata “Lianmeiren” (v2.0), achieved through the integration of PacBio HiFi, ultra-long Oxford Nanopore Technologies, Illumina, and Hi-C sequencing. The assembly resulted in a highly contiguous genome with a total size of 266.23 Mb and a contig N50 of 31.6 Mb. The genome exhibits remarkable completeness (98.9% BUSCO) and high accuracy (quality value of 48.75). Additionally, 13 telomeres and putative centromere regions were successfully identified across the 8 pseudochromosomes. Comparative analysis with the previous v1.0 assembly revealed 336,943 single nucleotide polymorphisms, 107,521 indels, and 1,413 structural variations, along with the annotation of 1,402 new genes.

Conclusions

This T2T genome assembly of P. campanulata “Lianmeiren” provides a critical reference for understanding the genetic architecture of the species. It enhances our ability to study structural variations, gene function, and evolutionary biology within the Prunus genus.

Data Description

Context

Cherry trees, known for their beautiful blossoms and abundant yield, have significant ornamental and economic value. Species of cherry belong to the subgenus Cerasus of Prunus that originated in China and is now distributed across the Northern Hemisphere [1–3]. The subgenus Cerasus encompasses approximately 50 to 60 species and varieties, with the majority found in China. However, only a select few species are cultivated. One such species, Prunus campanulata Maxim.(NCBI:txid136465), native to southern China, is highly valued as an ornamental [4, 5]. P. campanulata present a set of favorable traits such as early flowering, vibrant colored flowers, disease resistance, self-compatibility, and abundant seed production, making it a prime genetic resource for developing superior cultivars [6–8]. Additionally, the low ploidy level (2n = 16), small genome size, and low heterozygosity render P. campanulata an ideal model for studies of the cherry genome.

The advent of accurate long-read sequencing technology has brought the telomere-to-telomere (T2T) concept to the forefront of plant genomics research [9]. This approach marks a revolutionary shift in genomic sequencing and assembly, focusing on the creation of complete, continuous sequences from the end of one chromosome to another [10]. The T2T genome offers insights into the structure of centromeres and telomeres by accurately resolving repeat sequences [11–13]. The method also facilitates the annotation of additional protein-coding genes, thereby providing avenues for advances in comparative genomics and evolutionary biology, and it provides precise genome sequences for applications in genetic domestication and breeding [9, 14–17]. In the future, the T2T genome is expected to become the standard reference. While the T2T genomes of many horticultural plants have been released in recent years, the T2T genome of cherries remains unsequenced.

At present, the genomes of several cherry species and varieties have been sequenced using next-generation or third-generation sequencing platforms. These species include Prunus avium [18, 19], Prunus yedoensis [20], Cerasus × yedoensis [21], Cerasus serrulata [22], Cerasus × kanzakura [23], Prunus fruticosa [24], P. campanulata [4, 5], Prunus pusilliflora [25], Prunus cerasus [26], and Prunus conradinae [27]. In February 2023, we reported a chromosome-level assembly of P. campanulata “Lianmeiren” (v1.0) that was achieved using a combination of PacBio, 10x Genomics, and Illumina sequencing technologies [4]. This was closely followed by the publication of another chromosome-scale genome of P. campanulata “Plena” [5]. Despite these advancements, challenges remain in resolving gaps and highly repetitive regions within the cherry genome, highlighting the need for ongoing refinement and improvement.

To tackle the existing challenges, we have assembled a high-quality T2T genome of P. campanulata. This assembly was achieved through the integration of ultra-long ONT, PacBio HiFi, Illumina, and Hi-C sequencing. The completion of a gap-free P. campanulata genome significantly advances our understanding of the cherry genome and paves the way for new opportunities in the utilization of cherry germplasm resources.

Methods

Plant materials and sequencing

P. campanulata “Lianmeiren,” a double-flowered cherry variety, was cultivated at the Zhejiang Academy of Forestry nursery in Hangzhou, China (Fig. 1A). During the fruit-ripening stage, fresh young leaves were harvested and immediately preserved in liquid nitrogen for DNA extraction. We employed the CTAB method to prepare high-molecular-weight genomic DNA that was subsequently purified using a Qiagen genomic kit (Qiagen, 13,343) for PacBio HiFi sequencing [28]. Additionally, DNA for ultra-long Oxford Nanopore Technologies (ONT) sequencing was extracted via the sodium dodecyl sulfate method [29]. The quality of the DNA was assessed using a NanoDrop One spectrophotometer (NanoDrop Technologies) and a Qubit 3.0 Fluorometer (Life Technologies). An Ultra-Long DNA Sequencing Kit v14 (SQK-ULK114; ONT) was used to create the ONT sequencing library, and a SMRTbell express template prep kit 2.0 (Pacific Biosciences) was employed for the preparation of the PacBio HiFi sequencing library. The sequencing of the ONT and PacBio libraries was performed on a Nanopore PromethION sequencer and the PacBio Sequel II platform (RRID:SCR_017990), respectively. Previous studies provided Illumina and Hi-C reads for supplementary analysis.

Figure 1:

P. campanulata “Lianmeiren” morphology and T2T genome assembly. (A) Floral morphology of P. campanulata “Lianmeiren.” Scale bar corresponds to 1 cm. (B) Hi-C interaction heatmap for 8 pseudochromosomes of the P. campanulata genome. (C) Circos plot of the P. campanulata genome assembly. From outer to inner ring: chromosomes, gene density, repeat sequence density, GC content, and gene collinearity, with a window length of 500 K.

Genome assembly and assessment

After discarding reads with Q-scores less than 7, we processed the ONT ultra-long reads by removing adapters and filtering out short reads (less than 10 kb) using Filtlong v0.2.4 (RRID:SCR_024020) and Porechop v0.2.4 (RRID:SCR_016967), respectively. We retained high-quality reads that were at least 30 kb in length and had a Q-score above 9 for further analysis. The quality of the raw HiFi reads was evaluated using CCS v6.0.0 (RRID:SCR_024379); this step involved filtering out sequences with fewer than 3 rounds of sequencing and low-quality subreads with a signal-to-noise ratio (SNR) below 2.5, thereby ensuring the validity of the data for subsequent analyses. Similarly, raw Illumina reads were processed with FastQC v0.21.0 (RRID:SCR_014583) [30] to eliminate low-quality reads. To facilitate chromosome-level assembly of the Hi-C data, only valid interaction pairs were extracted using HiCUP v0.8.0 (RRID:SCR_005569) [31].

Preliminary assembly of the ONT ultra-long sequencing data was conducted using NextDenovo v2.5.0 (RRID:SCR_025033) [32] with the read_cutoff set to 1k, blocksize at 1 g, and nextgraph_options configured as −a1. To polish the assemblies, we utilized 2 iterative rounds of Racon v1.4.11 (RRID:SCR_017642) and Pilon v1.23 (RRID:SCR_014731) [33] for the ONT and Illumina reads, respectively, adhering to their default settings. For the PacBio HiFi draft genome assembly, 2 distinct approaches were adopted: one using Hifiasm v0.16.1-r375 (RRID:SCR_021069) [34] exclusively for PacBio HiFi data and the other combining ONT ultra-long with PacBio HiFi data via Hifiasm v0.18.2-r467 (RRID:SCR_021069) [34]. During the assembly process, we utilized Purge_haplotigs v1.0.4 (RRID:SCR_017616) [35] and Purge_dups v1.2.5 (RRID:SCR_021173) [36] to process the T2T HiFi draft genome. These tools were employed to identify and remove redundant contigs or haplotigs resulting from heterozygous regions, ensuring a clean and representative assembly. This process yielded 52 contigs from the ONT genome assembly, 354 from the PacBio assembly, and 249 from the hybrid assembly. To screen for and remove nontarget sequences, Minimap2 v2.17-r941 (RRID:SCR_018550) [37] was used to identify mitochondrial and chloroplast data. Sequences with a base alignment of over 50% were excluded. Additionally, bacterial contamination was identified and eliminated by comparing the sequences to those in the RefSeq library (RRID:SCR_003496) [38], and contigs with insufficient read support were discarded.

ALLHiC v0.9.8 (RRID:SCR_022750) [39] was employed for clustering the contig sequences into distinct chromosome groups utilizing a bottom-up hierarchical clustering algorithm. This was followed by ordering and orienting the contigs within each chromosome group. The pairwise interactions between contigs were then transformed into “.hic” files using 3D-DNA v180419 (RRID:SCR_017227) [40] and Juicer v1.6 (RRID:SCR_017226) [41]. Juicebox v1.11.08 (RRID:SCR_021172) [42] was used for manual ordering and orientation. The gap-free ONT genome sequence was used as a reference to fill in the gaps in the genome assembled by Hifiasm v0.18.2-r467 (RRID:SCR_021069) [34]. The heterozygous sequences identified through the pairwise interaction of contigs were removed, and gaps were filled with a sequence of 100 Ns to construct the final chromosome-level genome sequence. Finally, to visualize the genomic interactions, a heatmap was generated using HiCExplorer v3.6 (RRID:SCR_022111) [43].

The ONT ultra-long reads were mapped to the genome assembly using Winnowmap v1.11 (RRID:SCR_025349) [44], focusing on collecting reads at the terminals of each pseudochromosome within a 50-bp screening window. We searched for the numbers of telomere repeats (CCCATTT at the 5′ end and TTTAGGG at the 3′ end) across all reads. The read with the highest count of these repeats was designated as “ref” (reference) and the others as “query.” Both reference and query sequences were then reassembled to obtain consensus sequences using medaka_consensus v1.2.1. The consensus sequences were then used to replace the terminal sequences on each pseudochromosome, a process carried out using MUMmer’s nucmer v3.1 (RRID:SCR_018171) [45]. For gap filling, we compared the data and genome gap intervals, prioritizing gap filling with the sequence hierarchy of “genome version after error correction > HiFi data > ONT Ultra-long data.” Sequences that spanned the entire gap on the alignment were chosen, with preference being given to the best-aligned sequence that covered the longest length of the region. This sequence was then used to fill the gap in the genome.

For error correction, HiFi reads with a length of at least 10 kb were aligned to the gapped version of the genome using Winnowmap2 (RRID:SCR_025349) [44]. The alignment parameters included k = 15, greater than, distinct = 0.9998, –MD, and -ax map-pb. The aligned fragments underwent filtering through SAMtools v1.10 (RRID:SCR_002105) [46], using the parameter -F 256. To remove chimeric alignments, we applied falconc bam-filter-clipped, setting parameters -t and -F 0×104. Utilizing the information from these filtered alignments, a specialized branch of Racon v1.6.0 (RRID:SCR_017642) was employed for error correction.

The continuity of the genome was evaluated by identifying the location and number of gaps in the assembly. To estimate the genome consensus, we mapped Illumina and Hi-C reads to the final assembly using BWA v0.7 (RRID:SCR_010910) [47]. Additionally, ONT and PacBio HiFi reads were aligned with Minimap2 v2.17-r941 (RRID:SCR_018550) [37]. The completeness of the genome assembly was assessed using BUSCO v5.3.0 (RRID:SCR_015008) [48]. To evaluate the quality and accuracy of the genome assembly, we compared the k-mer spectrum of Illumina sequencing reads with the assembled genome. This comparison was expressed through the consensus quality value (QV), which is a logarithmic score representing the accuracy of the genome assembly.

Genome annotation

To identify and classify repeat sequences, we initially utilized RepeatModeler v1.0.11 (RRID:SCR_015027) for de novo prediction. This was complemented by the use of LTR_Finder (RRID:SCR_015247) [49] and LTR_retriever (RRID:SCR_017623) [50] to identify nonredundant long terminal repeat (LTR) sequences. We then combined these sequences to form a de novo repeat sequence library. This library was merged with the Repbase v20181026 library (RRID:SCR_021169) [51] to obtain a comprehensive database. This combined library was then used in RepeatMasker v4.0.9 (RRID:SCR_012954) to predict repeat sequences throughout the genome. Additionally, RepeatProteinMask v4.0.9 was used specifically for the prediction of transposable element (TE) proteins. We compiled the final set of repeat sequences after the removal of redundant sequences.

For the prediction of gene structure, we employed a combined strategy encompassing ab initio, homology-based, and RNA sequencing (RNA-seq)–based methods. For ab initio prediction, we used Augustus v3.3.2 (RRID:SCR_008417) [52] and GlimmerHMM v3.0.4 (RRID:SCR_002654) [53], focusing on the genomic regions masked for repeat sequences. BUSCO v5.2.2 (RRID:SCR_015008) [48] was then used to derive training sets for this purpose. In the homology-based approach, protein sequences of P. avium [18], C. serrulata [22], P. mume [54], P. persica [55], and P. campanulata (v1.0) [4] were mapped to the reference genome using TBLASTN v2.7.1 (RRID:SCR_011822). This was followed by the use of Exonerate v2.4.0 (RRID:SCR_016088) [56] to predict transcripts and coding regions. For RNA-seq–based prediction, RNA-seq reads, filtered using fastp v0.21.0 (RRID:SCR_016962) [30], were initially aligned to the genome via HISAT2 v2.1.0 (RRID:SCR_015530) [57]. The resulting alignment data were then utilized to acquire transcripts with StringTie v2.1.4 (RRID:SCR_016323) [58], and these transcripts aided in predicting gene models using TransDecoder v5.1.0 (RRID:SCR_017647). Finally, we integrated all data to form the final set of gene models via MAKER v2.31.10 (RRID:SCR_005309) [59].

Gene functions were predicted based on sequence and motif similarities. This involved comparing their protein sequences against several databases: UniProt (RRID:SCR_002380) [60], Nr [61], GO (RRID:SCR_002811) [62], KOG [63], Pfam (RRID:SCR_004726) [64], InterPro (RRID:SCR_006695) [65], and KEGG (RRID:SCR_012773) [66]. For the KEGG annotations, we utilized DIAMOND v0.9.30 (RRID:SCR_016071) [67] and KOBAS v3.0 (RRID:SCR_006350) [68]. Protein domain and GO term annotations were derived using InterProScan v5.52–86.0 (RRID:SCR_005829) [69], while protein family annotations were obtained by searching the Pfam database (RRID:SCR_004726) [64] with hmmscan v3.3.2 [70]. In addition, transfer RNAs (tRNAs) in the genome were identified with tRNAscan-SE v1.23 (RRID:SCR_008637) [71], focusing on their structural characteristics. Ribosomal RNAs (rRNAs) were predicted using the rRNA database, and small nuclear RNA (snRNA) and microRNA (miRNA) sequences were annotated based on the Rfam database (RRID:SCR_007891) [72] using Infernal v1.1.2 (RRID:SCR_011809) [73].

Genomic comparison between v2.0 and v1.0 assemblies

The complete T2T genome assembly was aligned pairwise with the v1.0 genome using SyRI v1.63 (RRID:SCR_023008) [74], enabling us to identify syntenic regions and various structural variants (SVs), including inversions, translocations, and duplications. For visual comparison between the T2T and v1.0 genomes, we employed OrthoVenn2 (RRID:SCR_022504) [75] to create a Venn diagram, setting an e-value threshold of 1e−10. To annotate genes newly identified in the T2T assembly, Gene Ontology (GO) analysis was conducted using InterProScan (RRID:SCR_005829) [69]. This analysis focused on characterizing gene functions across biological process, cellular component, and molecular function terms, as defined by the GO knowledgebase (RRID:SCR_017505) [62]. The R package clusterProfiler (RRID:SCR_016884) [76] was then used to perform the GO enrichment analysis and to visualize the results. Additionally, we utilized JCVI v0.9.13 [77] to identify genes that were newly annotated in the T2T genome relative to v1.0, particularly those located in inversions, duplications, and translocations.

Identification of telomere and centromere

To identify telomeres, all ONT reads were first aligned to the reference genome using Winnowmap v1.11 (RRID:SCR_025349) (parameters: k = 15, –MD) [39], specifically targeting reads that aligned singularly within 50 bp of chromosomal ends. We then calculated the frequency of telomere repeat sequences (“CCCTAAA”/“TTTAGGG”) in each read, referencing the Telomere database [78]. The read with the most telomere repeats was designated as the reference and the others as queries. Following this, medaka_consensus v1.2.1 (parameters: -m r941_min_high_g360) was employed to reassemble the reference and query the telomere reads, yielding a consensus sequence. This consensus sequence was then aligned to each chromosome using MUMmer’s nucmer v3.1 (RRID:SCR_018171) [45] to replace terminal telomere sequences using the best alignment results. However, replacement was not conducted if the identity fell below an 80% threshold or if the aligned region was not within 20 kb of the chromosomal end. Leveraging the distinct features of high-density short tandem repeat finders (TRFs) and low-density gene distribution in centromere regions, we employed BEDTools (RRID:SCR_006646) [79] to compute TRF and gene coverage, utilizing a 10-bp window. This analysis led to the prediction of 8 centromeric regions within the chromosomes of the P. campanulata genome.

Evolutionary analysis

In addition to P. campanulata, we included 13 other plant species—Arabidopsis thaliana [80], P. persica [55], P. mume [54], P. avium [18], C. serrulata [22], C. × yedoensis [21], P. yedoensis [20], P. salicina [81], Malus domestica (v3.0.a1) [82, 83], Pyrus pyrifolia [84], Fragaria vesca [85], Rubus argutus [86], and Rosa chinensis [87]—for gene family clustering. This clustering was performed using BLASTP v2.6.0 (RRID:SCR_001010) [88] and OrthoFinder v2.3.12 (RRID:SCR_017118) [89]. After clustering, the R package clusterProfiler (RRID:SCR_016884) [76] was employed to conduct GO and KEGG analyses. For the analysis of single-copy orthologous gene families, protein sequences were aligned using MUSCLE v3.8.31 (RRID:SCR_011812) [90]. The alignment results were then refined with trimAl v1.2rev59 (RRID:SCR_017334) [91] and amalgamated to create a comprehensive super-alignment matrix.

Using the super-alignment matrix, we constructed a maximum likelihood (ML) phylogenetic tree with RAxML v8.2.10 (RRID:SCR_006086) [92] using the GTRGAMMA substitution model. MCMCTree from PAML v4.9 (RRID:SCR_014932) [93] was used to estimate divergence times. We incorporated 3 calibration priors from TimeTree (RRID:SCR_021162) [94] in our analysis. These included divergence time estimates between P. campanulata and A. thaliana (102.0–112.5 million years ago [Mya]), M. domestica and P. pyrifolia (2.30–54.83 Mya), and R. chinensis and F. vesca (21.12–57.76 Mya).

Using the insights gained from the phylogenetic tree and gene family clustering, gene family expansions and contractions were identified using CAFE v3.1 [95]. To further characterize the genetic variation, clusterProfiler (RRID:SCR_016884) [76] was employed to conduct GO and KEGG enrichment analyses, providing a deeper understanding of the functional implications of the gene family dynamics.

To detect whole-genome duplication (WGD) events, our initial step involved aligning the protein sequences from P. campanulata with those of other related species using BLAST v2.6.0+ [88]. This was followed by identifying collinear segments both within P. campanulata and between P. campanulata and related species (P. avium, P. mume, and P. persica) using MCScanX v0.8 (RRID:SCR_022067) [96] under the default settings. The frequency of synonymous (Ks) and nonsynonymous (Ka) mutations, as well as their ratio (Ka/Ks), in these collinear gene pairs was calculated using the yn00 module of PAML v4.9 (RRID:SCR_014932) [93]. The resulting data were then visually represented in a density map created using ggplot2 v2.2.1 (RRID:SCR_014601) [97].

For the analysis of positive selection, we utilized the CodeML module in PAML v4.9 (RRID:SCR_014932) [93]. MAFFT (RRID:SCR_011811) [98] was initially employed to align protein sequences from single-copy gene families among the selected species. The protein sequences were then converted into codon sequences using PAL2NAL v14 [99]. Then, CodeML using the Branch-site model was used to perform likelihood ratio tests between model A (which assumes that the foreground branches ω are under positive selection, i.e., ω > 1) and the null model (where no site is permitted to have an ω value greater than 1). These tests were conducted using the chi2 program in PAML v4.9 (RRID:SCR_014932) [93]. Genes exhibiting significant differences (P < 0.05) were classified as being subject to positive selection.

To identify similar gene pairs, we utilized LAST v1170 (RRID:SCR_006119) [100] to compare gene sequences between 2 species. Following this, JCVI v0.9.13 [77] was employed to ascertain the chromosomal positions of these similar gene pairs. We then plotted a collinear map to visually represent the relationships and alignments of these gene pairs across chromosomes.

Results

T2T assembly of P. campanulata genome

The initial genome survey utilizing Illumina reads estimated the genome size of P. campanulata “Lianmeiren” to be approximately 295.31 Mb, with a heterozygosity rate of 0.60% (Supplementary Fig. S1). To construct a T2T gap-free genome assembly of P. campanulata, we generated approximately 16.19 Gb (∼54× coverage) of ultra-long sequencing reads using the ONT platform and approximately 30.54 Gb (∼108× coverage) of PacBio HiFi reads using the PacBio Sequel II platform (Supplementary Table S1). The N50 lengths for the HiFi and ONT ultra-long reads exceeded 15.58 kb and 100 kb, respectively (Supplementary Table S1). Additionally, around 63.83 Gb (∼209× coverage) of Illumina paired-end sequencing data were utilized to correct the genome assembly and for QV evaluation. Three draft assemblies were created using ONT ultra-long reads, PacBio HiFi reads, and a hybrid assembly combining both. The assembly utilizing PacBio HiFi reads exhibited superior performance, resulting in 354 highly continuous contigs, with a contig N50 of 29.86 Mb, a QV of 53.8, and 98.9% completeness (Supplementary Table S2). This assembly was selected as the T2T genome framework. Following the exclusion of nonnuclear and contaminated sequences and contigs with low support, anchoring of contigs was performed using approximately 35.34 Gb (∼117× coverage) of Hi-C sequencing data (Supplementary Table S1), organizing all contigs into 8 pseudochromosomes (Fig. 1B). A complete, gap-free reference genome (v2.0) was subsequently produced by filling all remaining gaps with the gap-free ONT genome data. Most of the gaps were located in chromosomes 1, 4, 5, and 7, with sequence lengths ranging from 99 to 95,957 bp (Supplementary Table S3). The finalized genome size was 266.23 Mb, a value that was slightly lower than the estimate derived from flow cytometry (∼295 Mb) (Supplementary Fig. S1), with a contig N50 of 31.6 Mb (Fig. 1C, Table 1, Supplementary Table S4).

Table 1:

Genomic statistics of P. campanulata v2.0 assembly and previous assemblies

Feature	“Lianmeiren” v2.0	“Lianmeiren” v1.0	“Plena”
Genome size (Mb)	266.23	299.15	280.20
Contig N50 (Mb)	31.6	2.02	18.31
Number of contigs	8	687	41
Gaps	0	/	/
Number of telomeres	13	0	0
Number of centromeres	8	0	0
Number of gene models	28,961	28,319	27,181
BUSCOs (%)	98.90	96.60	98.70

Extensive evaluation of the T2T P. campanulata assembly

The quality of the P. campanulata genome was assessed across 4 dimensions: consistency, contiguity, completeness, and accuracy. The assembly demonstrated near-perfect consistency, as evidenced by the absence of gaps and mismatches (N) across all chromosomes, and the count of contigs precisely matched the number of chromosomes. For contiguity, 96.1% of the Illumina short reads, 99.11% of the ONT ultra-long reads, and 99.98% of the HiFi reads could be aligned to the assemblies, covering 100%, 99.99%, and 99.99% of the respective assembly regions (Supplementary Table S5). Completeness was evaluated using BUSCO, with 98.9% (N = 1,614) of conserved plant genes identified as complete (Supplementary Table S6). The k-mer statistical analysis indicated a QV value of 48.75 for the genome, with individual chromosomes ranging from 46.08 to 51.67, reflecting the high accuracy of the assembly (Supplementary Table S7). Considering all these factors, the T2T P. campanulata genome presented here is of the highest reliability and quality.

Genome annotation analysis

Various prediction methods were employed to annotate repeat sequences in the P. campanulata T2T genome, yielding results for TE proteins and a combination of de novo and Repbase methods. After synthesizing the prediction results and eliminating redundancy, a total of 130.84 Mb of repeat sequences were identified, constituting 49.14% of the entire genome. This included 23.92% LTR retrotransposons, 15.47% DNA transposons, 4.63% long interspersed nuclear elements, 0.47% short interspersed nuclear elements, and 7.01% uncharacterized TEs (Fig. 2A, Supplementary Table S8). Employing a combination of de novo, homology-based, and transcriptome prediction methods, we identified 28,961 protein-coding genes in the P. campanulata genome (Table 2). The average lengths of transcripts, coding sequences (CDSs), exons, and introns were approximately 3,724 bp, 1,141 bp, 320 bp, and 523 bp, respectively, with an average of 5.02 exons per gene (Table 2). We analyzed the length distribution of genes, CDSs, exons, and introns among C. serrulata, P. avium, P. mume, P. persica, and P. campanulata v1.0. The exon and intron length distributions were consistent across species, with some variation in the gene and CDS length distributions, particularly in the P. avium genome (Fig. 2B). BUSCO analysis revealed that 98.2% (1,585 of 1,614) of the core conserved plant gene orthologs were fully detected, confirming the high-confidence annotation of these genes in P. campanulata (Supplementary Table S6). A significant majority (27,934; 96.45%) of the predicted protein-coding genes were successfully annotated by at least 1 gene function database, slightly higher than the 93.1% in the v1.0 assembly (Supplementary Table S9). Additionally, we identified a total of 2,414 noncoding RNAs, including 287 miRNAs, 668 tRNAs, 886 rRNAs, and 573 snRNAs, exceeding the numbers in the v1.0 assembly (Supplementary Table S10).

Figure 2:

Genomic features of the P. campanulata genome. (A) Composition and divergence of transposable elements in the P. campanulata genome. (B) Distribution of gene length, CDS length, exon length, and intron length among the assemblies of P. campanulata v2.0, C. serrulata, P. avium, P. mume, P. persica, and P. campanulata v1.0. The y-axis represents the percentage of genes, CDSs, exons, and intron.

Table 2:

Statistics of protein-coding gene annotation for the P. campanulata genome

Method	Software	Gene number	Average gene length (bp)	Average CDS length (bp)	Average exon per gene	Average exon length (bp)	Average intron length (bp)
Ab initio	GlimmmerHMM	36,212.00	4,849.50	911.42	3.76	242.55	1,428.07
Ab initio	AUGUSTUS	26,704.00	2,963.24	1,209.35	5.07	238.4	430.64
Homology	Exonerate	25,022.50	2,582.15	1,191.79	4.21	281.78	430.74
RNA-seq	TransDecoder	26,816.00	4,481.45	999.95	5.63	397.82	484.19
Integration	Maker	27,457.00	4,111.52	1,130.70	5	286.72	668.32
Final set	Anno-self	28,961.00	3,724.42	1,140.73	5.02	320.28	523.14

Genome-wide identification of variation

We conducted a comparative analysis with the v1.0 assembly, focusing on various sequences and SVs. A collinearity analysis revealed 270.82 Mb (97.5%) of syntenic regions between the v2.0 and v1.0 genomes (Fig. 3A). Within these syntenic regions, we identified 336,943 single nucleotide polymorphisms (SNPs), of which 166,274 were distributed in gene regions and 170,669 in intergenic areas (Supplementary Table S11). Furthermore, we detected 107,521 insertions–deletions (indels) ranging from 2 to 50 bp, comprising 62,058 insertions and 45,463 deletions (Supplementary Table S11). A total of 1,413 SVs were identified, comprising 1,212 duplications, 163 translocations, and 38 inversions (Fig. 3B, Supplementary Fig. S2A–C). Structural annotation analysis indicated that most SVs were located 2 kb upstream and downstream of genes, in introns, and in intergenic regions, with a length distribution primarily centered on 1,001–2,000 bp and >9,000 bp (Fig. 3C). Additionally, presence–absence variation (PAV) revealed 928 presence variants and 1,223 absence variants (Fig. 3B, Supplementary Fig. S2D). GO functional and KEGG pathway-enrichment analyses of these SV and PAV sequences showed significant enrichment in defense responses, including plant–pathogen interactions and secondary-metabolite biosynthesis (Supplementary Fig. S3). These findings underscore the importance of high-quality genome assembly in advancing plant research.

Figure 3:

Structural variations analysis between v2.0 and v1.0 P. campanulata genomes. (A) Collinearity between v2.0 and v1.0 P. campanulata genomes, with v2.0 as the reference. The gray lines show the syntenic regions between v2.0 and v1.0. Nonsyntenic regions represent gaps in the v1.0 assembly. The orange, green, and bluish-green lines represent structural variations, specifically inversion, translocation, and duplication, respectively. (B) Number and proportion distribution of each type of structural variations. (C) Distribution of structural variations counts across different regions of the genome.

Newly annotated genes in the P. campanulata T2T genome

By comparing gene models from the v2.0 assembly with those from v1.0, we identified 1,402 genes that were present in v2.0 but absent from v1.0, representing 4.9% of the protein-coding genes in v2.0 (Supplementary Fig. S4A, Supplementary Table S12). We conducted Gene Ontology (GO) and KEGG pathway enrichment analyses to determine the functions of these newly annotated genes. The GO annotation results showed significant enrichment of genes involved in defense response within the biological process category, integral components of membranes within the cellular component category, and nucleic acid binding and ATP binding within the molecular function category (Supplementary Fig. S4B). The KEGG annotations indicated that these genes were associated with processes detailed in the global and overview maps (Supplementary Fig. S4C) and with transcription.

Telomere and centromere characteristics

Telomeres are essential conserved structures in plant genomes that differentiate natural chromosome ends from double-stranded breaks in DNA, thereby protecting the chromosome ends from degradation or end-to-end fusion with other chromosomes [101]. Typically, they are tandemly arranged minisatellites, following the formula (TxAyGz)n [78]. Utilizing telomere repeats as queries, we successfully identified 13 telomeres located at the ends of the 8 pseudochromosomes (Table 3). Notably, chromosomes 4, 5, and 7 each had a telomere at only one end. The number of motif repeats ranged from a minimum of 158 to a maximum of 612 (Table 3). To predict potential centromere regions of the P. campanulata chromosomes, we utilized short tandem repeats, integrating these data with Hi-C interaction heat maps, large blank regions, areas of low gene density, and regions with high LTR/Gypsy density (Fig. 4). This approach successfully identified a presumptive centromere for each chromosome, with lengths ranging from 1.98 to 2.99 Mb (Table 3). However, verifying the actual locations of these centromeres will require further research, for example, using fluorescence in situ hybridization and chromatin immunoprecipitation sequencing methods.

Table 3:

Telomeres and centromeres in P. campanulata chromosomes

Chromosome	Telomeres		Centromeres
	Upstream	Downstream	Start	End	Length
Chr1	332	536	26,000,366	27,989,631	1,989,265
Chr2	204	330	22,001,493	23,999,214	1,997,721
Chr3	390	541	15,001,219	16,998,339	1,997,120
Chr4	0	248	4,000,087	5,998,911	1,998,824
Chr5	0	565	6,002,969	7,994,235	1,991,266
Chr6	199	306	15,000,384	17,999,009	2,998,625
Chr7	240	0	22,503,972	24,999,143	2,495,171
Chr8	158	612	18,001,542	19,995,827	1,994,285

Figure 4:

Predicted centromeric regions in the chromosomes of P. campanulata. The blue and orange bars represent the coverage of LTR/Copia and LTR/Gypsy, respectively. The blue bars represent the coverage of TRF and indicate the predicted centromeric regions. The bluish-green bars represent the coverage of genes. The coverages were calculated using a 10k window.

Comparative genomic analysis

To examine the evolutionary dynamics of the flowering cherry genome, we conducted a comparative genomic analysis between the P. campanulata genome and those of 13 other species to identify homologous genes, performed gene family clustering analysis, and assessed the distribution of single-copy and multiple-copy genes. A total of 74,894 orthologous gene families comprising 504,527 genes were detected across all species, with 7,893 gene families (encompassing 189,790 genes) shared by all (Fig. 5A, Supplementary Table S13). In addition, 182 single-copy gene families were identified. The P. campanulata v2.0 genome contained 1,198 unique gene families, including 1,446 unique paralogs (Supplementary Tables S13 and S14). GO enrichment analysis revealed that these gene families were predominantly involved in hexosyltransferase activity, fucose metabolic processes, and histidine biosynthesis (Supplementary Fig. S5A). KEGG pathway analysis indicated significant involvement in protein export, fructose and mannose metabolism, and nucleotide excision repair, among others (Supplementary Fig. S5B). Comparative analysis with the 3 closely related Prunus species P. avium [18], C. serrulata [22], and P. persica [55] showed that 13,844 gene families were shared among these species, while 1,636 gene families were unique to P. campanulata (Fig. 5B). These findings offer significant insights into the genetic uniqueness and evolutionary trajectory of the flowering cherry genome.

Figure 5:

Comparative genomic analysis for P. campanulata. (A) Orthologous gene families between P. campanulata and the other 13 angiosperm species. Atha: A. thaliana; Cser: C. serrulata; Cyed: C. × yedoensis; Fves: F. vesca; Mdom: M. domestica; Pavi: P. avium; Pbre: P. bretschneideri; Pcam: P. campanulata; Pmum: P. mume; Pper: P. persica; Psal: P. salicina; Pyed: P. yedoensis; Rarg: R. argutus; Rchi: R. chinensis. (B) Unique and shared gene families among P. campanulata and the other 3 closely related species genomes. (C) Phylogenetic tree for P. campanulata and the other 13 angiosperm species. The green numbers indicate expanded gene families, while the red numbers indicate contracted gene families. (D) Genome synteny analysis among P. campanulata, C. serrulata, and P. avium. The number above the bars indicates the chromosome index.

A phylogenetic tree was constructed using 177 single-copy orthologous genes from 14 species, with A. thaliana [80] serving as the outgroup. The resulting ML tree showed that P. campanulata forms a monophyletic clade with P. yedoensis [20], C. × yedoensis [21], C. serrulata [22], and P. avium [18], collectively classified under the subgenus Cerasus (Supplementary Fig. S6). This clade is sister to the subgenus Prunus clade that includes P. salicina [81], P. mume [54], and P. persica [55]. Phylogenetic analysis and fossil calibration indicated that P. campanulata and P. avium diverged approximately 23.4 Mya, with a 95% highest probability density (HPD) of 12.9–36.6 Mya (Supplementary Fig. S6, Supplementary Table S15). The divergence between the subgenus Cerasus and the subgenus Prunus occurred around 30.5 Mya, with a 95% HPD of 17.5–46.1 Mya. Employing the likelihood model in CAFE v3.1 with default parameters [95], we identified 377 gene family expansions and 1,272 gene family contractions affecting 1,543 and 1,473 genes, respectively (Fig. 5C). GO functional enrichment and KEGG pathway analyses revealed that the expanded gene families were primarily associated with pentose and glucuronate interconversions, fatty acid degradation, and tyrosine metabolism, whereas the contracted gene families were predominantly involved in plant–pathogen interactions, homologous recombination, and glutathione metabolism (Supplementary Fig. S7). WGD events in the P. campanulata genome were estimated by analyzing the synonymous mutation rates of homologous genes among P. campanulata, P. avium, C. serrulata, P. persica, and M. domestica [82, 83], based on their orthologous gene pairs. The analysis revealed distinct peaks at 4-fold degenerate synonymous sites of the third codons (4DTv) values of approximately 0.02 and 0.55 on the map (Supplementary Fig. S8A). The peak at approximately 0.02 4DTv highlighted divergence events between P. campanulata and P. avium, P. campanulata, and P. persica, as well as P. campanulata and P. mume. The second peak at approximately 0.55 4DTv suggested a whole-genome or large-fragment duplication event in the common ancestor of these 4 species. Collinearity analysis between P. campanulata and P. avium, as well as between P. campanulata and C. serrulata, was conducted, showing an overall syntenic depth ratio of 1:1 for both comparisons (Fig. 5D, Supplementary Fig. S8B). This indicates that neither P. campanulata nor these related species experienced WGD events.

The study of positive selection of genes in plants is crucial to understanding their adaptive evolution. In this study, treating P. campanulata as the foreground branch and P. avium, C. serrulata, and P. persica as background branches, we identified several genes as candidates for positive selection, genes that may have contributed to the species’ evolutionary adaptation (Supplementary Table S16).

Reuse potential

The T2T genome of P. campanulata “Lianmeiren” represents a comprehensive and versatile genomic resource with significant reuse potential in multiple research and applied domains. As a high-quality reference genome, it provides an unparalleled opportunity for comparative genomics studies. Researchers can use this dataset to explore structural and functional genomic variations within the Prunus genus and among related species, enhancing our understanding of evolutionary dynamics, synteny, and chromosomal evolution.

The discovery of 1,402 new genes in the T2T genome assembly provides significant opportunities for comparative genomics. For example, candidate genes related to defense response, transcription regulation, and stress tolerance found in other Prunus species, such as P. avium [18] and P. persica [55], could be cross-referenced with this dataset to identify homologous or novel functional variants. Although these newly identified genes have not yet been experimentally validated, we have outlined their potential roles in breeding and applied research. For instance, genes involved in pathways for stress resistance—traits critical for ornamental and fruit crop breeding—can serve as targets for genome editing or marker-assisted selection. The T2T genome provides a comprehensive reference to identify structural or sequence variations affecting these traits.

We have also expanded on how this genome can serve future studies, such as generating a haplotype-resolved genome for P. campanulata to investigate allelic variation and its effects on phenotype. Additionally, this resource could aid in building a Prunus pan-genome, enabling the identification of core and dispensable genomic regions critical for adaptive traits and species divergence.

Additional Files

Supplementary Fig. S1. The k-mer analysis of P. campanulata with GenomeScope (K = 19). Aa: homozygosity rate; Ab: heterozygosity rate; Dup: duplication rate; Err: error rate; Kcov: k-mer coverage for the heterozygous bases; Len: estimated total genome length; P: peak number; Uniq: unique portion of the genome (not repetitive).

Supplementary Fig. S2. Distribution of number and length of SVs. (A) Duplication. (B) Translocation. (C) Inversion. (D) PAV.

Supplementary Fig. S3. GO and KEGG enrichment analysis of the SV and PAV genes. (A) Significantly enriched GO terms of the SV genes. (B) KEGG pathway enrichment of the SV genes. (C) Significantly enriched GO terms of the PAV genes. (D) KEGG pathway enrichment of the PAV genes.

Supplementary Fig. S4. Prediction and annotation of new genes in P. campanulata v2.0 genome. (A) Venn diagram of new gene prediction using de novo, homology-based, and RNA-seq–based strategies. (B) GO analysis of new gene set, including biological process, cellular component, and molecular function. (C) KEGG pathway analysis of new gene set.

Supplementary Fig. S5. GO and KEGG pathway enrichment analysis of unique gene families between P. campanulata and the other 13 angiosperm species. (A) Significantly enriched GO terms of unique gene families. (B) KEGG pathway enrichment of unique gene families.

Supplementary Fig. S6. Phylogenetic tree and divergence time estimation. The numbers outside the square brackets indicate the average divergence time, and the numbers inside the square brackets indicate the 95% confidence interval of divergence time.

Supplementary Fig. S7. GO and KEGG pathway enrichment analysis of expansion and contraction gene families in P. campanulata v2.0 genome. (A) GO enrichment for expansion gene families. (B) KEGG pathway enrichment for expansion gene families. (C) GO enrichment for contraction gene families. (D) KEGG pathway enrichment for contraction gene families.

Supplementary Fig. S8. (A) Ka/Ks distribution of each pair of the 5 species genomes. Cser: C. serrulata; Mdom: M. domestica; Pavi: P. avium; Pcam: P. campanulata; Pper: P. persica. (B) Ratio of syntenic depth between P. campanulata and C. serrulata, as well as P. avium.

Supplementary Table S1. Summary of sequencing data of P. campanulata assembly.

Supplementary Table S2. Statistics and assessment of different assembly strategies.

Supplementary Table S3. Gap region and length in P. campanulata assembly.

Supplementary Table S4. The length and contig number of chromosomes in P. campanulata genome.

Supplementary Table S5. Assessment of P. campanulata assembly contiguity.

Supplementary Table S6. BUSCO analysis of P. campanulata genome completeness.

Supplementary Table S7. QV analysis of P. campanulata genome accuracy.

Supplementary Table S8. Summary statistics of repetitive sequences in P. campanulata v2.0 genome.

Supplementary Table S9. Statistics of gene function annotation.

Supplementary Table S10. Statistics of noncoding RNAs in P. campanulata genome.

Supplementary Table S11. Statistics of SNP and indel in P. campanulata genome.

Supplementary Table S12. Functional annotation of new genes.

Supplementary Table S13. Statistics of gene families of P. campanulata and the other 13 angiosperm species.

Supplementary Table S14. Statistics of ortholog genes between P. campanulata and the other 13 angiosperm species.

Supplementary Table S15. Fossil calibration points used to calibrate the phylogenetic tree.

Supplementary Table S16. GO annotation of genes under positive selection.

giaf009_Reviewer_1_Report_Original_Submission

ThomasWolfgangWöhner -- 10/27/2024

giaf009_Reviewer_1_Report_Revision_1

ThomasWolfgangWöhner -- 1/3/2025

giaf009_Reviewer_2_Report_Original_Submission

DiegoMicheletti --11/8/2024

Abbreviations

BLAST: Basic Local Alignment Search Tool; bp: base pairs; BUSCO: Benchmarking Universal Single-Copy Orthologs; CDS: coding sequence; CTAB: cetyltrimethylammonium bromide; Gb: gigabase pairs; GO: Gene Ontology; Hi-C: high-throughput chromosome conformation capture; HPD: highest posterior density; indel: insertion–deletion; KEGG: Kyoto Encyclopedia of Genes and Genomes; Kb: kilobase pairs; KOG: EuKaryotic Orthologous Groups; LINE: long interspersed nuclear element; LRT: likelihood ratio test; LTR: long terminal repeat; Mb: megabase pairs; miRNA: micro RNA; ML: maximum likelihood; Mya: million years ago; Nr: NCBI’s nonredundant database; ONT: Oxford Nanopore Technologies; PacBio HiFi: Pacific Biosciences high fidelity; PAV: presence–absence variation; RNA-seq: RNA sequencing; rRNA: ribosomal RNA; SINE: short interspersed nuclear element; snRNA: small nuclear RNA; SNP: single nucleotide polymorphism; SNR: signal-to-noise ratio; SV: structural variation; T2T: telomere to telomere; TE: transposable element; tRNA: transfer RNA; TRF: tandem repeat; WGD: whole-genome duplication; 4DTv: 4-fold degenerate synonymous sites of the third codons.

Author Contributions

Conceptualization and supervision: D.J. and X.L. Methodology: X.S. and D.J. Software: X.S. and Y.L. Investigation: Q.Z. and F.Z. Writing—original draft preparation: X.S. and D.J. Writing—review and editing: X.S. and Y.L. Visualization: W.Z. and Q.Z. Funding acquisition: F.Z, X.S. and X.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Zhejiang Science and Technology Major Program on Agricultural New Variety Breeding (grant number 2021C02071-4). The research was supported by the Special Support Funds of Zhejiang for Scientific Research Institutes (grant number 2023F1068-2) and the National Natural Science Foundation of China (grant number 32101585).

Data Availability

All raw sequencing data presented in this study have been deposited at NCBI under BioProject accession number PRJNA1162277 (ONT, PacBio HiFi, and Illumina) and PRJNA884816 (Hi-C and Transcripts). The genome assembly and annotation data are available at JBKKFS000000000. All additional supporting data are available in the GigaScience repository, GigaDB [102].

Competing Interests

All authors declare that they have no competing interests.