Genome sequencing, assembly and annotation

Sequencing of the ‘Junzao’ genome resulted in a 351-Mb assembly with contig and scaffold N50 sizes of 34 kb and 754 kb, respectively (Table 1; Table A in S2 File). A k-mer analysis of ‘Junzao’ sequences suggested an estimate genome size of ~350 Mb, consistent with the size estimated from the flow cytometry analysis (Fig C in S1 File; Table B in S2 File). The GC content of the assembled ‘Junzao’ genome was 32.6% (Fig D in S1 File). Approximately 98.3% of the 2,901 expressed sequence tag (EST) sequences and 98.9% of the assembled transcriptome contigs could be mapped to the ‘Junzao’ genome (Table 1; Table C in S2 File). In addition, 99.6% of the core eukaryotic genes were mapped to the ‘Junzao’ genome using CEGMA [9] (Fig E in S1 File) and 93.2% were completely mapped to the assembled ‘Junzao’ genome using BUSCO [10] (Table 1), indicating a high quality of the ‘Junzao’ genome assembly.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of assembled genomes between ‘Junzao’ and ‘Dongzao’

https://doi.org/10.1371/journal.pgen.1006433.t001

Using two high-density genetic linkage maps, we anchored 600 assembled scaffolds to the 12 linkage groups, covering 83.6% (293 Mb) of the assembled ‘Junzao’ genome (Table D in S2 File; Fig F in S1 File). We predicted a total of 27,443 protein-coding genes with an average coding sequence length of 1,136 bp and an average of 4.83 exons (Table E in S2 File), of which 91.2% were mapped to the 12 pseudo-chromosomes. A total of 2.1 million single-nucleotide polymorphisms (SNPs) were detected in the ‘Junzao’ genome, and therefore the heterozygosity level of the genome was calculated as 0.72% (Table F in S2 File). In addition, 2,309 small insertions and deletions (indels) were found to be located in the exonic regions (Table G in S2 File).

Comparison of the two jujube (‘Junzao’ vs. ‘Dongzao’) genomes

The assembled ‘Junzao’ genome was 86.5 Mb smaller than the reported genome of ‘Dongzao’ (437.7 Mb), which was assembled by sequencing the in vitro cultured plantlet [8]. One notable difference between the ‘Junzao’ genome and the reported ‘Dongzao’ genome was the abundance of transposable elements (TEs). A total of 136 Mb of TEs were identified, accounting for 38.8% of the assembled ‘Junzao’ genome, while the reported genome of ‘Dongzao’ contained 204 Mb of TEs (46.8%) (Table 1; Fig G in S1 File; Table H and Table I in S2 File). In addition, a more recent accumulation of TEs was found in ‘Dongzao’ (<1.2 million years ago) (Fig H(a) in S1 File), and a greater proportion of genes were close to the TEs in ‘Dongzao’ than in ‘Junzao’ (Fig H(b) in S1 File). Phylogenetic analysis also indicated a greater expansion of specific LTR retrotransposon clades in the ‘Dongzao’ genome (Fig H(c) in S1 File).

Collinear genome regions between ‘Dongzao’ and ‘Junzao’ were identified (Table J in S2 File). The syntenic blocks in the ‘Dongzao’ genome (326.3 Mb) were 34.1 Mb larger than those in the ‘Junzao’ genome (292.2 Mb). We found that 26.0 Mb (77%) of the 34.1 Mb were repetitive sequences, further supporting that transposons are one of the major factors contributing to the genome size difference between ‘Junzao’ and ‘Dongzao’.

We found that unanchored scaffolds in the reported ‘Dongzao’ genome had many syntenic blocks with the anchored scaffolds, much higher than those in the assembled ‘Junzao’ genome (Fig I in S1 File). Furthermore, read coverage distribution of coding regions in the ‘Dongzao’ genome displayed a heterozygous peak at the half depth of the major homozygous peak, while no heterozygous peak was found in ‘Junzao’ (Fig J. and Fig K in S1 File). These findings suggest that sequences of heterozygous alleles from the same loci (redundant sequences) were sometimes failed to be assembled into consensus sequences in ‘Dongzao’, partially contributing to the larger genome assembly size of ‘Dongzao’ than that of ‘Junzao’. In addition, we identified ~4.9 Mb bacterial sequences in the ‘Dongzao’ genome assembly. Taken together, we suggest that higher levels of repetitive sequences, redundant sequences and bacterial contaminated sequences in the assembled ‘Dongzao’ genome have contributed to the larger genome assembly of ‘Dongzao’ than ‘Junzao’.

A presence-absence variation (PAV) analysis identified 7.8 Mb of ‘Dongzao’-specific sequences containing 354 genes and 14.2 Mb of ‘Junzao’-specific sequences containing 432 genes. Gene Ontology (GO) terms including DNA recombination and DNA integration were found to be significantly enriched in ‘Dongzao’-specific genes (Table K in S2 File). In addition, we identified 131 expanded gene families (930 genes) and 232 contracted families (702 genes) in ‘Junzao’ in comparison with ‘Dongzao’ (Fig L in S1 File).

‘Junzao’ and ‘Dongzao’ are representative cultivars of dry and fresh jujubes, respectively (Fig B in S1 File). Their fruits contain highly different levels of crude fiber, which is derived from the fruit primary cell walls (Table L in S2 File). We found that several families of genes involved in cell wall modification were substantially expanded in the dry jujube ‘Junzao’ compared with ‘Dongzao’, including those encoding glycosyl hydrolases (beta-glucosidases, xyloglucan endotransglucosylase-hydrolases, endoglucanases and polygalacturonases) and those encoding pectin esterases and rhamnogalacturonate lyases (Table M in S2 File).

Evolutionary scenario of genome rearrangements within the Rhamnaceae

Eight putative proto-chromosomes of the common ancestor of Rhamnaceae and Rosaceae, two sister families in the order Rosales, were inferred based on the available genome sequences of jujube, peach (Prunus persica) and apple (Malus × domestica) (Fig 1), and they are similar to the nine putative proto-chromosomes of the ancestor of Rosaceae [11,12]

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Evolutionary scenario of genome rearrangements from the ancestor of Rosales to jujube, peach and apple.

In the top-right diagram, different colors in each chromosome represent the origin from the seven common ancestral chromosomes of eudicot, and the eight chromosomes filled with different colors represent putative paleo-chromosomes for the common ancestor of Rhamnaceae and Rosaceae.

https://doi.org/10.1371/journal.pgen.1006433.g001

No recent whole-genome duplication (WGD) events were detected in jujube [8] and peach [13] after their divergence, while one such event was identified in apple [14]. Although the numbers of proto-ancestral chromosomes in the Rosaceae increased from eight to nine after its divergence from the Rhamnaceae [11], we were still able to identify a one-to-two relation between the jujube and apple genomes. Considering the intergenomic relations among jujube, peach and apple, we determined the two largest chromosome synteny pairs as follows: 1) jujube chromosome 3, peach chromosome 2 and apple chromosomes 1 and 7, and 2) jujube chromosome 10, peach chromosome 3 and apple chromosomes 9 and 17 (Fig 1), which reflected the recent diploidization of the apple genome [14]. A conserved block was also identified among jujube chromosome 3, peach chromosome 2 and apple chromosome 7, which did not undergo any rearrangements, fissions or fusions and is thus likely derived directly from ancient chromosome III (Fig 1). These results showed that larger syntenic blocks were retained in jujube chromosomes, and illustrated that fewer chromosome fissions, fusions and rearrangements occurred in the jujube genome compared with the peach and apple genomes (Table N in S2 File).

Jujube population structure

Resequencing the genomes of 31 accessions, including 10 wild jujube individuals (6 typical wild jujubes and 4 semi-wild accessions) and 21 jujube cultivars (Table O in S2 File; Fig M in S1 File), generated a total of 344 Gb of sequences, representing an average depth of 27.8× and an average coverage of 92.5% (Table O in S2 File). After mapping the reads of each accession to the genome of ‘Junzao’, we detected a total of 5,300,355 SNPs. The parameter θπ values [15] indicated that wild jujubes, although represented in our analysis by half the number of accessions (10) as cultivated accessions (21), exhibited greater diversity (θπ = 2.60×10⁻³) than cultivated jujubes (θπ = 2.19×10⁻³). The neighbor-joining phylogenetic tree illustrated the domestication process as a transition from wild to cultivated jujubes via certain semi-wild accessions (Fig 2A). In addition, the cultivated jujube group could be further divided into two subgroups that were generally correlated with their geographical distributions in West China and East China (Fig 2A; Fig N in S1 File). A principle component analysis (PCA) generated a similar pattern (Fig 2B) to that of the phylogenetic analysis in that the jujube cultivars formed a tight cluster that was distant from the wild jujube accessions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Jujube population structure.

(A) Neighbor-joining phylogenetic tree of jujube accessions based on whole-genome SNP data. The jujube accessions are marked with E (east) or W (west), indicating their geographical distributions. (B) PCA of cultivated and wild jujube populations using whole-genome SNP data. (C) Population structure analysis of jujube accessions using FRAPPE at multiple kinship levels (K = 2, 3, 4 and 5). Each vertical bar represents one jujube accession. The length of each colored segment in each vertical bar represents the proportion contributed by ancestral populations. The accession marked with * in part A corresponds to the wild jujube ‘Yanchuandasuanzao’.

https://doi.org/10.1371/journal.pgen.1006433.g002

Population structure analysis indicated that wild and cultivated jujubes could be divided into two groups when K = 2, although admixed features were observed in 17 accessions covering both cultivated and wild jujubes (Fig 2C). With K = 3, the cultivated populations were further divided into two subgroups corresponding to their geographical distributions (West and East China), whereas the wild population remained relatively uniform. When the K value increased progressively from 3 to 5, new subgroups emerged in the wild jujube group and further differentiation was found within the cultivated jujubes (Fig 2C).

Domestication of jujube and population differentiation

As shown in Fig 3A, jujube fruit had a much higher content of soluble sugars, and lower levels of organic acids than wild jujube fruits (Table P in S2 File), indicating both sweetness and acidity are important traits under human selection. Selective sweep regions covering 1,372 genes were identified in the jujube genome (Fig 3B; Table Q in S2 File). These included four genes, which encode an NADP-dependent malic enzyme (NADP-ME), a pyruvate kinase (PK), an isocitrate dehydrogenase (IDH), and an aconitate hydratase (ACO), all of which play key roles in organic acid metabolism in fruit (Fig 3C; Table R in S2 File). In addition, three vacuolar proton pumps (V-type proton ATPase), transporting H⁺ into vacuolar, were also in the putative sweep regions. On the other hand, three genes involved in sugar metabolism in fruit, encoding a sucrose synthase (SUSY), a phosphoglucomutase and a 6-phosphofructokinase, and 13 sugar transporters were also identified in the regions of putative selective sweeps.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Jujube sugar and acid metabolism associated with domestication.

(A) Sugar and acid accumulation in fruits at various developmental stages from cultivated and wild jujubes. YF, young fruit; EF, expanding fruit; WMF, white mature fruit; HRF, half-red fruit; and FRF, full-red fruit. (B) Number of genes detected in the putative selective regions using different methods. (C) Transcript abundance of genes involved in sugar and acid metabolism in cultivated and wild jujubes. Stars indicate the genes that were located in the regions putatively detected as selective sweeps. Scaled log₂ expression values (RPKM) are shown in the heat map legend. The six boxes in one row of each heat map (left to right) correspond to the expression levels at stages EF, HRF and FRF of the wild accession (‘Qingjiansuanzao’) and WMF, HRF and FRF of the ‘Junzao’ cultivar. Each row in the heat map corresponds to one gene. SUSY: sucrose synthase; SPS: sucrose phosphate synthase; SPP: sucrose-phosphatase; HK: hexokinase; FK: fructokinase; PK: pyruvate kinase; MDH: malate dehydrogenase; ACO: aconitate hydratase; IDH: isocitrate dehydrogenase; SDH: succinate dehydrogenase; CS: citrate synthesis; NINV: neutral invertase; CWINV: cell wall invertase; vAINV: vacuolar acid invertase; PFK: 6-phosphofructokinase.

https://doi.org/10.1371/journal.pgen.1006433.g003

Expression profiling analysis of sugar- and acid-related metabolism genes showed that a gene encoding a vacuole acid invertase (VAINV), an enzyme that irreversibly catalyzes the hydrolysis of sucrose to glucose and fructose, was expressed at a significantly lower level in the ripe fruits of cultivated jujubes than in those of wild jujubes, possibly contributing to higher sucrose accumulation in the vacuoles of cultivated jujube fruits (Fig 3C). In addition, most genes involved in acid metabolism pathways, including those encoding NADP-ME, PK, phosphoenolpyruvate carboxylase (PEPC), malate dehydrogenase (MDH), shikimate dehydrogenase (SD) and citrate synthesis (CS), were expressed at much higher levels in wild than in cultivated jujube fruits (Table S in S2 File). This trend was also the case for a neutral invertase in the sucrose biosynthesis pathway, which supplies glucose and fructose for organic acid metabolism (Fig 3C). On the contrary, genes involved in decomposing citrate, such as ACO and IDH, were expressed at lower levels in wild than in the cultivated jujubes.

Furthermore, a population differentiation analysis based on a population fixation index (Fst) between dry and fresh jujube groups (Table T in S2 File) uncovered four genes encoding beta-galactosidases and one encoding endo-1,4-beta-xylanase in the highly differentiated regions (Table U in S2 File).

Identification of jujube S-locus genes

We identified a candidate S-RNase gene (Zj.jz035833030; chromosome 1) and two S-like RNases (Zj.jz026761011 and Zj.jz022467042; chromosomes 7 and 9, respectively) that belong to the T2-RNase family in the ‘Junzao’ genome (Table V in S2 File). We also identified the three T2-RNase genes in the ‘Dongzao’ genome (Table V in S2 File). Phylogenetic analysis further confirmed that Zj.jz035833030 was the S-RNase and that Zj.jz026761011 and Zj.jz022467042 were S-like RNases (Fig 4A). A transcriptome analysis revealed that Zj.jz035833030 (named S1) was specifically expressed in flowers, and the two S-like RNase genes were expressed in all tested tissues (Fig 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Characterization of the jujube S locus.

(A) Phylogenetic analysis of three T2-RNase sequences from each ‘Junzao’ (Zj.jz) and ‘Dongzao’ (Zj.dz) with other known S genes. (B) Expression (RPKM) of S-RNase and S-like RNase genes in different tissues. (C) Physical positions of S-RNase and SFB genes on chromosome 1 of the ‘Junzao’ genome. The scaled log₂ expression values (RPKM) are shown in the heat map. From left to right, leaf (L), flower (F), phloem (P), expanding fruit (EF), half-red fruit (HR) and full-red fruit (FR). (D) Frequency of S-locus genotypes in 32 jujube accessions.

https://doi.org/10.1371/journal.pgen.1006433.g004

We identified four candidate SFB genes near the S1 S-RNase gene on chromosome 1 and inferred that the jujube S locus is likely localized within a narrow region (from 3.8–4.3 Mb) on chromosome 1. However, none of these four SFB genes was specifically expressed in flowers (Fig 4C). A phylogenetic analysis showed that all four SFB genes were clustered together in the same clade with the Prunus SFB (Fig O in S1 File). The structure of the jujube S locus was similar to that of pear [16].

By mapping genome sequencing reads that were generated from the 31 jujube accessions to the putative S1 region, we identified 21 SNPs (9 in the first exon, 11 in the second exon, and one in the intron) and 3 indels (one in the first exon, one in the second exon, and one in the intron). Among the 21 SNPs, 13 were located in the ribonuclease domain of T2-RNase (Fig P in S1 File). According to the SNP pattern in the S-RNase gene, we assigned each accession a corresponding haplotype. We found that S1 was the most common haplotype, and there were 10 accessions with a homozygous S1 locus and 14 with a heterozygous locus (Fig 4D).

Plant materials, sequencing and assembly

A 9-year-old diploid, highly heterozygous dry jujube cultivar, ‘Junzao’ (voucher number: NWAFU-Junzao001), grown at the Jujube Experimental Station (N 37.13, E 110.09) of Northwest A&F University, Qingjian, Shaanxi Province, China, was used for genome sequencing. Genome sizes of jujubes including ‘Junzao’, ‘Dongzao’ and other 11 accessions were estimated by flow cytometry analyses on the young leaves (Table B in S2 File, S1 File Supplementary notes). The ‘Junzao’ genome was sequenced using a whole-genome shotgun strategy [30]. High-quality genomic DNA was extracted from young leaves using the Qiagen DNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA). A total of 3 μg of DNA was used for each library construction. Short-insert paired-end libraries (180 bp and 500 bp) were generated using the NEB Next Ultra DNA Library Prep Kit for Illumina (NEB, USA) according to the manufacturer’s instructions. Large-insert (2 kb, 5 kb, 10 kb, 15 kb and 20 kb) DNA sequencing libraries were prepared through circularization by Cre-Lox recombination [31]. These libraries were sequenced on the Illumina HiSeq 2000 system. A total of 79 Gb of high-quality cleaned sequences (approximately 227x coverage of the genome) was generated and used for de novo genome assembly (Table X in S2 File). A modified version of SOAPdenovo was developed specifically for the de novo assembly of the highly heterozygous jujube genome (S1 File Supplementary Notes).

Gene prediction

Augustus [32], Geneid [33], Genscan [34], GlimmerHMM [35] and SNAP [36] were used for ab initio gene predictions. We also aligned the protein sequences of Arabidopsis thaliana, Capsicum annuum, Citrus clementina, Eucalyptus grandis, Malus × domestica, Oryza sativa, Populus trichocarpa, and Vitis vinifera to the ‘Junzao’ genome using TBLASTN with an E-value cutoff of 1e^-5. The homologous genome sequences were then aligned to the matched proteins for accurate spliced alignments using GeneWise [37]. Finally, a total of 36 Gb of high-quality RNA-Seq reads was aligned to the ‘Junzao’ genome using TopHat [38] with default parameters. Based on the RNA-Seq read alignments, Cufflinks [39] was then used for transcriptome-based gene structure predictions. Outputs from ab initio gene predictions, homologous protein alignments and transcript mapping were integrated using EVM [40] to form a comprehensive and non-redundant reference gene set and then filtered by removing the genes with incorrect coding sequences and putative repeat elements (80% coverage).

Genetic map construction and scaffold anchoring

We combined two genetic maps, described below, to anchor the assembled scaffolds of ‘Junzao.’ First, we used a previously published restriction site-associated DNA (RAD)-based high-density genetic map generated from an inter-specific F₁ population to anchor the genome assembly [41]. Second, we constructed a genetic map by using a different F₁ population (‘Dongzao’ × ‘Yingshanhong’, 96 progenies), which was also based on the RAD strategy according to Baird et al [42]. High-quality SNP and SSR markers were used to construct a linkage map (Table D in S2 File). The resulting genetic map was used to further anchor the assembled scaffolds of ‘Junzao.’

Genome evolution analysis

To better understand the evolutionary processes that shaped the genome structures of jujubes, we reconstructed the putative proto-chromosomes of the common ancestor of Rhamnaceae and Rosaceae, which are sister families in the order Rosales [43]. Protein sequences from 13 plant species (A. thaliana, C. annuum, C. sinensis, M. × domestica, O. sativa, P. trichocarpa, V. vinifera, Cucumis sativus, Pyrus × bretschneideri, Actinidia chinensis, Cypripedium arietinum, Z. jujuba ‘Dongzao’ and Z. jujuba ‘Junzao’) were extracted for building gene families. For alternatively spliced isoforms, only the longest proteins were used in the analysis. An all-to-all BLASTP was used to compare protein sequences with an E-value cutoff of 1e^-7, and OrthoMCL [44] was then used to cluster genes from these species into families with the parameter “-inflation 1.5.” MUSCLE [45] was used to generate multiple sequence alignments of proteins in single-copy gene families with default parameters. RAxML [46,47] and a ‘supermatrix’ of protein sequences were used to construct the phylogenetic tree with the maximum likelihood algorithm. A molecular clock model was implemented to estimate the divergence time of these 13 species using McMctree in PAML [48]. To obtain a more accurate result, ‘r8s’ was used to estimate the divergence time based on the constructed tree. Café [49] was used to identify gene families that have undergone significant expansion or contraction in the Z. jujuba ‘Junzao’ genome with a p-value cutoff of 0.05.

Collinear region and PAV detection between ‘Junzao’ and ‘Dongzao’ genomes

The one-to-one collinear regions between ‘Junzao’ (accession number: PRJNA306374) and ‘Dongzao’ [8] were detected using the MUMmer package [50] with the parameters ‘-maxmatch -c 90 -l 40 -d 0.05’. The sequence alignments were performed on the scaffold level between these two jujube genomes. The best reciprocal alignments with length less than 300 bp or an identity less than 90% were discarded, and then the aligned regions within the same scaffold were connected together. Regions were identified as syntenic blocks if there were more than 5 adjacent alignment regions between the two genome sequences.

In addition, we use the “show-snps” program in MUMmer package to detect homozygous SNPs and small indels from the one-to-one alignments. RepeatMasker (http://www.repeatmasker.org/RepeatModeler.html) was used to find repeat elements in sequences that could not be aligned to the ‘Junzao’ genome. Sequences shorter than 100 bp were removed. ‘Dongzao’-specific sequences were obtained after realigning them with the ‘Junzao’ genome and discarding sequences with identities of greater than 95% and gap lengths of less than 100 bp. Potential bacterial sequences, which were identified on the basis of BLAST searches against the GenBank NT database, were excluded. Genes with at least 90% of the CDS regions covered by ‘Dongzao’-specific sequences were defined as ‘Dongzao’-specific genes.

Reconstruction of common ancestral chromosomes

Syntenic blocks shared by the seven species (V. vinifera, P. trichocarpa, Theobroma cacao, A. thaliana, Prunus persica, M. × domestica, and Z. jujuba ‘Junzao’) were identified with MCscanX [51] using the grapevine genome as the reference. Syntenic blocks containing at least 3 gene pairs were retained to reconstruct the genome structure of the seven selected species. Based on the syntenic and overlapping relations of Z. jujuba, P. persica and M. × domestica genomes, we reconstructed the paleo-chromosomes of the common ancestor of Rhamnaceae and Rosaceae using a previously described method [52, 53]. The structures of the V. vinifera, P. trichocarpa, T. cacao and A. thaliana genomes were reconstructed by comparing them with the seven proto-chromosomes of the eudicot ancestor.

Resequencing and SNP calling

Thirty-one jujube accessions were chosen for genome resequencing analysis, including 10 wild jujube individuals (6 typical wild jujubes and 4 semi-wild accessions) and 21 jujube cultivars (Table O in S2 File; Fig M in S1 File). Illumina paired-end genome libraries were constructed for each accession following the manufacturer’s instructions and then sequenced on Illumina HiSeq 2500/4000 platforms, which yielded a total of 363 Gb of raw paired-end sequences. The raw data were processed to remove low-quality bases, adapter sequences, and putative PCR duplicates, resulting in a total of 344 Gb of high-quality paired-end sequences. The cleaned reads were mapped to the ‘Junzao’ genome using BWA [54] with the parameters “mem -t 4 -k -M” [29]. BAM files were processed for SNP calling using the SAMtools mpileup function [55] with the parameters “-m 2 -F 0.002 -d 1000 -u.” High-quality SNPs, which were supported by a coverage depth of 5–1,000, mapping quality >20, distance of adjacent SNPs >10 bp and missing ratio of samples within each group <50%, were retained for subsequent analyses.

Phylogenetic analysis and population structure of jujube accessions

Peach was used as the outgroup to construct the phylogenetic tree. The NUCmer program in MUMmer [50] was used to align the peach genome (GenBank accession no. AKXU00000000) with the ‘Junzao’ genome with default settings. SNPs within the best-hit regions were extracted, and then the genotypes of peach were used to provide outgroup information at corresponding positions. The neighbor-joining phylogenetic tree was constructed using Treebest-1.9.2 (http://treesoft.sourceforge.net/treebest.shtml) on the basis of the p-distance. We used the parameter θπ [15] to assess the level of genetic diversity for cultivated and wild jujube populations by scanning whole-genome SNP sites, respectively. PCA was performed using EIGENSOFT [56]. The eigenvectors were obtained from the covariance matrix using the R function ‘eigen.’ The population structure was further inferred using the program FRAPPE [57] with kinship (K) set from 2 to 5 and the maximum iteration of expectation-maximization set to 10,000.

Identification of selective sweeps and population differentiation

Four methods, i.e., the θπ ratios, pairwise population differentiation (Fst) levels [58], Tajima’s D test [59] and the cross-population composite likelihood ratio test (XP-CLR) [60], were used to identify the selective sweeps associated with jujube domestication events. Briefly, θπ ratios (θπ, wild/θπ, cultivated), Fst and TajimaD were calculated using a sliding window analysis with a window size of 20 kb and a step size of 10 kb. XP-CLR test was performed with the following parameters: sliding window size of 0.6 cM, grid size of 10 kb, maximum number of SNPs within a window of 300, and correlation value for 2 SNPs weighted with a cutoff of 0.95. Genome regions with the top 5% of scores in each of the four methods were identified and those detected by at least two of the methods were identified as selective sweeps. In addition, we used the top 5% highest F_ST values to characterize the population differentiation between dry and fresh jujubes. Genes within these regions were subjected to GO enrichment analysis using EnrichPipline [61].

S locus gene identification and analysis

We compared the ‘Junzao’ genes to the information in a local database containing known S-RNase gene sequences collected from NCBI with an E-value cutoff of 1e^-10 using BLASTN. We then screened for genes belonging to the T2 RNase gene family from the BLAST results because S-RNase genes are members of the T2 RNase family [62, 63]. Candidate S-RNase genes were further screened according to two criteria: the absence of the amino acid pattern 4 ([CG] P [QLRSTIK][DGIKNPSTVY]) [63] and the presence of a maximum of two introns [64]. The S-RNase genes from four families (Rosaceae, Fabaceae, Solanaceae, and Plantaginaceae) and the candidate jujube S-RNases were used to construct the phylogenetic tree using RAxML with a generalized time-reversible (GTR) model of sequence evolution.

Pollen-determinant S-haplotype-specific genes belong to the F-box family. We used a similar BLAST strategy to that described above to search for the F-box genes in the chromosome region in which the candidate S-RNase gene was located. A phylogenetic analysis was performed for those candidate SFB genes together with the known Prunus SFB, Petunia SLFs, Prunus SLF1 and Malus SFBs.

To investigate the expression of the candidate S-RNase and SFB genes, we used RNA-Seq data from leaves, phloem, flowers and fruits of ‘Junzao.’ The SNP calling results derived from the resequencing of 31 accessions were used to reconstruct the two haplotypes of S-RNase gene using HapCUT [65].

Transcriptome sequencing and expression analysis

Phloem, mature leaves, flowers, and fruits at different stages (expanding fruit, half-red, and full-red) of the ‘Junzao’ cultivar and the wild jujube ‘Qingjiansuanzao’ (8 years old) were collected in 2013 and 2014, respectively. All the samples were immediately frozen in liquid nitrogen. Total RNAs were isolated using a modified CTAB method and then treated with RNase-free DNase I (Promega, USA). First-strand cDNAs were synthesized using a Clontech kit. RNA-Seq libraries were constructed using the NEB Next UltraTM RNA Library Prep Kit (NEB, USA) and sequenced on a HiSeq 2000/2500 system. RNA-Seq reads were mapped to the ‘Junzao’ genome using TopHat [38]. The total numbers of aligned reads (read counts) for each gene were normalized to the reads per kilobase exon model per million mapped reads (RPKM) [66]. DESeq [67] was used to identify differentially expressed genes.

Determination of sugar and organic acid levels in jujube fruits

Fruits were collected at three developmental stages: expanding fruit, half-red fruit and full-red fruit. Sugars (fructose, glucose and sucrose) and acids (malic acid, citric acid and succinic acid) were quantified using high-performance liquid chromatography (HPLC, Shimadzu) as described previously [68]. A total of 1 g of the edible part of the dried jujubes was ground and incubated in 50 mL of 80% ethanol in an ultrasonic bath (40 kHz, 45°C, 20 min). The samples were centrifuged at 3,500 ×g for 10 min, and the supernatant was collected in a new tube. The pellet was re-extracted by repeating the above steps. The combined supernatants were evaporated in a rotary evaporator at 45°C and then diluted with deionized water to 10 mL. The diluted extracts were filtered through a 0.45-μm membrane filter prior to HPLC analysis.

Data availability

Accession codes: Sequence data have been deposited in the GenBank/EMBL/DDBJ nucleotide core database under the accession number LPXJ00000000 (PRJNA306374) and all sequence reads have also been deposited in the online database. The version described in this paper is the first version.