Background

Peach (Prunus persica) is the third most important of the deciduous fruit crop in the world, and is widely cultivated in temperate and subtropical regions, ranking only after apple and pear. Benefiting from the small genome size, peach has been extensively used for comparative and functional genomic researches of the Rosaceae family [1]. Nine years ago, the first high-quality reference genome sequence of peach was released [2]. Subsequently, the evolution [3, 4], domestication regions [3, 5, 6], and genes associated with important traits [7, 8] for peach had been successfully dissected.

Adaptability is the basis of plant survival and reproduction, especially under extreme environmental conditions. Among the four wild relatives of P. persica, P. mira originate from the Qinghai-Tibet Plateau (QTP) and show strong adaptability to cold and ultraviolet rays (UV) resistance. P. davidiana is widely distributed in the slopes and valleys of mountains in northern China and is resistant to cold and drought environments. P. kansuensis and P. ferganensis grow in one specific region, Gansu and Xinjiang province of China, respectively [9]. However, genomic basis of the response of these wild species to different geographical environments remains unclear.

Genome assembly is a powerful tool to discover elite genes and the method have used frequently in peach in recently years. For example, Guan et al. [10] produced a comprehensive structure variations map for peach to identify a large 1.67-Mb heterozygous inversion that segregates perfectly with flat-fruit shape after assembling a flat-fruit peach cultivar “Rui You Pan 1.” Cao et al. [11] analyzed the molecular basis of the reduction of volatile compounds during the evolution of peaches and identified the key gene regulating linalool content through the assembly of a cultivated peach variety “Chinese Cling” and the association analysis of 256 peach varieties. Moreover, some wild related species of peach were also assembled. Tan et al. [12] reported de novo genome assemblies for five species, including P. persica, P. mira, and P. davidiana. Next, the author performed an association analysis of 417 peach accessions to identify loci associated with fruit shape, fruit development period, and floral morphology, respectively. In general, researchers mainly focused on gene identification of fruit quality characters in these studies. To dissect the genetic basis of adaptation to high plateau of P. mira, Wang et al. [13] de novo assembled three high-quality genomes of Tibetan Prunus species, including P. mira. The results showed that the expansion of SINE retrotransposons helped Tibetan Prunus species adapt to the harsh environment of the Himalayan plateau by promoting the accumulation of beneficial metabolites.

In this study, to dissect the mechanism by which regulate peach evolution and species adaptation at the genome-wide level, we de novo assembled the genomes of four wild relatives of P. persica. Comparative genomic analysis showed that P. davidiana was evolved from P. mira directly and crossed P. dulcis during its evolutionary history. Then, a nematode resistance gene was identified in P. kansuensis using the identified genetic variations based on new assemblies. Finally, comprehensive characterization of selective sweeps revealed the new mechanisms underlying the high-altitude adaptability in P. mira. Compared with previous studies, the genomes of four wild species assembled in the study are of higher quality, providing more detailed basic data for other researchers to study peach evolution and dissect the genetic mechanism of resistance traits.

Results

Assembly of the genomes of four wild peach species

The high-quality genome of P. mira, P. davidiana, P. kansuensis, and P. ferganensis were assembled using a combination of PacBio, Illumina, and Hi-C (High-throughput chromosome conformation capture) platforms. Using the k-mer method (Additional file 1: Table S1; Additional file 2: Fig. S1), we estimated the size of four genomes to be 242.94, 237.29, 238.06, and 237.24 with a heterozygous ratio of 0.76%, 1.10%, 0.56%, and 0.53% in P. mira, P. davidiana, P. kansuensis, and P. ferganensis, respectively. Then, a total of 600.37 ×, 574.84 ×, 343.99 ×, and 387.90 × coverage of sequences were generated for above species and used for genome assembly, respectively (Additional file 1: Table S2). Among the ultimately obtained 253.63, 259.27, 253.17, and 261.3 Mb assemblies (Table 1; Fig. S2), 92.29%, 89.46%, 94.03%, and 90.53% of them were anchored and allocated to eight pseudochromosomes for P. mira, P. davidiana, P. kansuensis, and P. ferganensis, respectively (Additional file 1: Table S3). The contig N50 sizes of the four final assemblies were 25.95, 22.64, 24.42, and 26.47 Mb, respectively (Table 1), which were higher than that of P. persica (250 kb), which produced by the Sanger technology [14]. Compared with the previously published genomes of P. mira and P. davidiana with contig N50 of ~ 8.3 Mb and 2.3 Mb, respectively [12], our final assembly showed an ~ 3.1 and 10.0-fold increase in the length of the contig N50. In addition, compared with another released P. mira assembly with a contig N50 of 12.14 Mb [13], our P. mira assembly showed an ~ 2.14-fold increase in the length of the contig N50. We also performed Benchmarking Universal Single-Copy Orthologs (BUSCO) analysis [15] (Additional file 1: Table S4) and RNA-Seq read mapping rate calculation (Additional file 1: Table S5) to evaluate the quality of the assembly. The results indicated that more than 95% of the highly conserved embryophyta genes were completely present in the four genomes, and more than 85% RNA-Seq reads could be mapped to the corresponding genomes. Subsequently, repeat sequence were annotated and the results showed that long terminal repeat (LTR) retrotransposons made up the majority of the transposable elements (TEs), comprising 35.61%, 42.44%, 41.75%, and 41.86% of the P. mira, P. davidiana, P. kansuensis, and P. ferganensis genome, respectively (Additional file 1: Table S6). Gene prediction and annotation were performed resulting in 28,519, 27,236, 26,986, and 28,587 protein-coding genes in P. mira, P. davidiana, P. kansuensis, and P. ferganensis, respectively (Additional file 2: Fig. S2; Additional file 1: Table S7 and 8). About 96.82-97.73% of the protein-coding genes of the four wild species could be functionally annotated (Additional file 1: Table S9). In addition, we also identified 8171-15,770 ribosomal RNA, 1226-1564 transfer RNA, 514-753 small nuclear RNA, and 563-693 microRNA genes in the four wild peach species (Additional file 1: Table S10).

Table 1 Genomic sequencing, assembly, and annotation statistic for four wild peach species
Full size table

Genome variations between species were also identified by anchoring the four assembled wild genomes onto the reference genome of P. persica [14]. A total of 547,998-2,971,987 single nucleotide polymorphisms (SNPs; Additional file 1: Table S11), 106,436-490,421 small insertions and deletions (indels; Additional file 1: Table S12), 3870-10,994 large structural variants (SVs; ≥ 50 bp in length; Additional file 1: Table S13), and 3344-6840 copy number variations (CNVs; Additional file 1: Table S14) were identified in the four species (Fig. 1a; Additional file 2: Fig. S3). A total of 20 SVs that differed in P. mira and P. persica was randomly selected (Additional file 1: Table S15) to design primers (Additional file 1: Table S16) for amplification in cultivar “2010-138” and “Shenzhou Li He Shui Mi” to verify the accuracy of these variations. The results showed that among the 20 variants, only the electrophoresis results of two primers might be inconsistent with the sequencing results (Additional file 2: Fig. S4). Meanwhile, IGV software (www.broadinstitute.org/igv) was used to observe the above 20 variant sites manually. We analyzed the function of the genes which comprised different variations (Additional file 2: Fig. S5-8) and found that plant-pathogen interaction pathways were enriched in genes containing small indels and CNVs in all four wild relatives of peach (p = 0.01 ~ 1.74e−22), showing that responding with environmental stress was very important for the speciation of four species.

Fig. 1.
figure 1

Variation screening of four wild species and evolutionary analysis of peach genome. a SNPs, small indels, SVs, and CNVs in P. mira, P. davidiana, P. kansuensis, and P. ferganensis compared to P. persica. b Core and dispensable gene families of four wild peaches and P. persica. c Gene Ontology annotation of genes specific in each species. d Estimation of divergence times of 15 species and identification of gene family expansions and contractions. Numbers on the nodes represent the divergence times from present (million years ago, Mya). MCRA, most recent common ancestor. e Distribution of Ks (synonymous mutation rate) values of orthologous genes between six genomes of the Prunus species and strawberry

Full size image

Peach genome evolution and species divergence

Combining the genome of P. mira, P. davidiana, P. kansuensis, P. ferganensis, and P. persica, all of the genes in the genomes of these five peach species were classified into 22,747 families on the basis of the homology of their encoded proteins (Additional file 2: Fig. S9). Presence/absence variations (PAVs) analysis of gene families revealed 8,138 (35.78%) dispensable gene-families distributed across all five genomes, and 368, 303, 303, 388, and 763 families specific to each of the above species, respectively (Fig. 1b). Single-organism metabolic process (p = 1.36e−6) and oxidation-reduction (p = 2.53e−15) were found to be enriched in gene families specific to P. mira. Cellular macromolecule biosynthetic process (p = 0.05) and gene expression (p = 0.01) pathways were enriched in specific gene families of P. davidiana (Fig. 1c).

A total of 830 single-copy gene families in the 15 species were identified (Additional file 2: Fig. S10) and used to construct a phylogenetic tree of P. persica and its wild related species as well as other representative plant species (Fig. 1d). Based on the known divergence time between Arabidopsis thaliana and strawberry, the age estimate for the split of the ancestor of P. mume, P. armeniaca, P. salicina, and the common ancestor of peach and almond was around 21.1 million years ago (Mya), later than that in a previous report, presumably 44.0 Mya [16]. Hereafter, the divergence time of P. dulcis and P. mira was about 13.3 Mya, which was obviously earlier than that of Yu et al. [4] and Alioto et al. [17] who found the divergence time of the two species was 4.99 and 5.88 Mya, respectively. Furthermore, we found that P. mira split with the common ancestor of P. davidiana, P. kansuensis, and P. ferganensis approximately 11.5 Mya. The event occurred around the drastic crustal movement of QTP [18] where P. mira originated.

Lineage-specific gene family expansions may be associated with the emergence of specific functions and physiology [19]. With the genome evolution of wild peach relatives, the number of expanded gene families was decreased from P. mira to P. kansuensis while increased in P. ferganensis or P. persica compared with that in the most recent common ancestor (MCRA; Fig. 1d). We found that the expanded gene families were highly enriched with those related to aminobenzoate degradation (or arginine and proline metabolism, styrene degradation) in P. mira (p = 5.53e−21; Additional file 2: Fig. S11) which originated in QTP, same as in Polylepis tarapacana [20] and Populus kangdingensis [21] where lived in high altitude regions.

To validate the speciation events of peach species, fourfold synonymous third-codon transversion (4DTv) rates were calculated and the results showed that all 4DTv values (Additional file 2: Fig. S12) among paralogs in four wild peach species peaked at around 0.50 to 0.55, consistent with the whole-genome triplication event (γ event) shared by all eudicots and indicating that no recent whole-genome duplication occurred. A peak 4DTv value at around 0 for the orthologs between four wild related species of peach and P. persica highlighted a very recent diversification of Prunus species [16]. To estimate the time of species divergence of the four wild species, we calculated the Ks (rate of synonymous mutation) values of orthologous genes between these species. As shown in Fig. 1e, the peaks at a Ks mode for orthologs between P. persica-P. dulcis was higher than that of other peaches. In addition, among the wild related species of P. persica, the similar Ks value of P. mira, P. davidiana, and P. kansuensis showed significantly higher than that of P. ferganensis, indicating similar results with phylogenetic tree.

Variations screening provided identification of new genes regulating resistance

To assess the advantage of genome assemblies of peach using five species, we analyzed the molecular mechanism for regulating the resistance traits of different species one by one. According to previous reports, P. fergenansis was recognized as only a geographic type of P. persica [3], so it was not discussed in the following analysis. We found that PAV may explain the multi-resistance of P. davidiana to diverse stresses and the deletions in the promoter region could enhances the resistance of P. kansuensis to nematodes and P. mira to low temperature by regulating the expression of target genes, respectively. Specific cases are described below.

Interspecific hybridization is essential for P. davidiana to obtain the multi-resistance traits

P. davidiana is a very important wild species of P. persica because of its resistance to various abiotic stresses such as drought and cold and biotic stresses such as aphids, which may be of great significance in the use of resistance breeding systems in the future [22]. Here, we first screened the R genes from four wild species and found a total of 295, 341, 310, and 322 genes were identified in P. mira, P. davidiana, P. kansuensis, and P. ferganensis genomes, respectively (Additional file 1: Table S17). The largest number of R genes in P. davidiana might explain its strong and multiple resistances to different pathogens, such as aphid, Agrobacterium tumefaciuns, etc. We found that most of the disease resistances QTLs/genes were located in genome regions containing candidate R genes (Additional file 2: Fig. S13).

Next, we want to analyze the source of these R genes. Considering that the k-mer frequency of P. davidiana presents a characteristic of a high heterozygosity genome (Additional file 2: Fig. S1b), we speculated that there was a hybridization event in the evolutionary history of P. davidiana. To further verify this inference, one hundred and twenty-six accessions of wild peach species (Additional file 1: Table S18; accessions 101-126) were resequenced, and the obtained sequences (Additional file 1: Table S19) were aligned to the P. mira genome to obtain a total of 839,431 high-quality SNPs. Using these SNPs, we constructed a phylogenetic tree and found that cherries, plums, and apricots had a close relationship with P. persica and its wild related species. And during the latter category, P. dulcis, P. tangutica, P. davidiana, and P. mira formed a small subgroup (Fig. 2a). Principal component analysis (PCA) also demonstrated that P. dulcis or P. tangutica is located between P. mira and P. davidiana in the diagram of PC2-PC3 (Fig. 2b). We then phenotyped the stone streak of different species and found that abundant dot streaks were present in the P. davidiana but absent in the more recent species, P. kansuensis. However, the dot streaks could trace back to the ancient species of this family, such as P. dulcis (Fig. 2c). According to the above evidence, we proposed that P. davidiana showed intermediate genomic characteristics between P. mira and P. dulcis and might originate from the cross between these two species. A previous study [4] reported that P. tangutica has a closer relationship with P. davidiana. However, since no reference genome is available for P. tangutica, a direct comparison of genes from P. davidiana and P. tangutica is not feasible. Therefore, we first aligned the sequences of P. davidiana to its putative ancestor, P. mira, and assembled the unmapped sequences to obtain a partial reference genome. Genome resequencing data of different species were then aligned to this partial genome. We found that the highest mapping rate was an accession belonging to P. dulcis, and next was P. tangutica (Fig. 2d), which again proved that the introgression in P. davidiana came from P. dulcis or P. tangutica. In addition, all the genes in P. davidiana were aligned with those from the putative ancestors who had an assembled genome, including P. mira and P. dulcis (Additional file 2: Fig. S14). The genes were then assigned as originating from a specific species if the highest score was observed from the alignments of orthologous genes between P. davidiana and the other species. The results showed that although about 50.74% of the genes (13,719) were specific to P. mira, 26.87% (7,266) were thought to be derived from P. dulcis, and no more than 8% from other species.

Fig. 2.
figure 2

The evolution of P. davidiana. a Phylogenetic tree of 126 peach accessions which included P. persica and its wild relative species. b Principal component analysis (PCA) of above accessions. c Stone steak morphology of P. dulcis, P. mira, P. davidiana, P. kansuensis, and P. ferganensis. d Percent of P. davidiana-specific contigs covered by reads from different Prunus species. e Regional collinearity between P. davidiana with P. dulcis or with P. mira genomes. The detailed location of these three regions were at Chr. 1: 4.79 Mb, Chr. 1: 5.08 Mb, and Chr. 2: 7.50 Mb in P. davidiana genome. Green and blue boxes indicate the positive and negative direction of genes. “P” and “A” letters showed the gene was presence and absence in P. davidiana genome

Full size image

To further study the evolutionary events leading to the genome structure of P. davidiana, we investigated the chromosome-to-chromosome relationships based on 16 (P. davidiana versus P. mira, including 21,186 genes) and 28 (P. davidiana versus P. dulcis, including 18,480 genes) identified syntenic blocks. The mosaic syntenic patterns again demonstrated that P. davidiana might have arisen during the evolutionary process of P. mira but with a cross of other species (Additional file 2: Fig. S15). Regional collinearity between P. davidiana with P. dulcis or P. mira indicated that some regions contained PAVs in P. davidiana genome showed a high structural similarity with them in P. dulcis (Fig. 2e). Although P. dulcis is mainly distributed in Georgia, Azerbaijan, Turkey, Syria, and Xinjiang province (China), it can also be found in Sichuan province of China where some P. mira grow in this region at the same time, indicating the hybridization event is highly possible (Additional file 2: Fig. S16).

We finally analyzed the origin of R genes in P. davidiana (Additional file 2: Fig. S12) and found 57.25% of all R genes were categorized to originate from P. mira, followed by P. dulcis (23.05%). Therefore, we believed that the cross between P. mira and P. dulcis or P. tangutica enhanced the adaptation of P. davidiana when it was spread to new environments.

An indel in the promoter of an NBS-LRR gene regulates the nematode resistance in P. kansuensis

P. kansuensis shows high resistance to root-knot nematodes [23]. To discover candidate causal sequence variations involved in the resistance, we constructed a backcross (BC1) population from the cross between “Hong Gen Gan Su Tao 1#” (P. kansuensis) and a cultivated peach “Bailey” (P. persica). “Hong Gen Gan Su Tao 1#” harbored high resistance to root-knot nematodes (Meloidogyne incognita), whereas other accessions including P. mira, P. davidiana, and P. ferganensis all showed low resistance to M. incognita [23]. Using this BC1 population, a nematode resistance locus was mapped at the top region of chromosome 2 (5.0-7.0 Mb) based on bulk segregation analysis (BSA) (Fig. 3a). The most significant locus was located at 6,671,644 bp of chromosome 2. In other species, the R genes to nematodes generally encode proteins containing the NBS-LRR domain [24]. Moreover, previous reports showed that the expression level of R gene which conferring root-knot nematode resistance was significantly higher in the roots of resistant plants than that of susceptible one and its value drastically increased when the plants were inoculated with nematodes [25, 26]. Therefore, transcriptome data [27] of two cultivars (resistant and susceptible to nematode infection, respectively) were analyzed again. We found that three of 19 expressed NBS-LRR genes (Prupe.2G046000, Prupe.2G053600, and Prupe.2G055500) have higher expression levels in “Hong Gen Gan Su Tao 1#” than those of “Bailey” during most inoculation periods (Fig. S17). However, only Prupe.2G053600 was induced by root-knot nematodes in 6 or 12 h of resistant accession than that of 0 h (Additional file 2: Fig. S17; Fig. 3c). In the regions, we further focused on the genome variations mainly including small indels and SVs which occurred in the 24 R genes and their promoter. We found that there were 92 variations which only present in P. kansuensis but not in other species (Additional file 1: Table S20). Among them, Prupe.2G053600, which comprised a large deletion (Chr. 2: 6,212,387..6212,422 bp) in the promoter was further analyzed (Fig. 3b).

Fig. 3.
figure 3

Identification of nematode resistance genes in P. kansuensis. a Bulked segregation analysis to locate the nematode resistance gene using a backcross population. b The Indels in the promoter and mRNA regions of one NBS-LRR genes on Chr. 2. “Ins” and “Del” indicate an insertion and deletion, respectively. And the number after the variation type indicates the length (bp) of Indels. c Expression of Prupe.2G053600 in two accessions (“Hong Gen Gan Su Tao 1#” and “Bailey”) inoculated with nematode. e Functional validation of Prupe.2G053600 through analysis of transgenic tomato plants expressing Prupe.2G053600 under nematode treatment. e Promoter activity assay. Promoters with different lengths were fused to the GUS gene in plasmid pBI101. GUS was dyed, and its activity was measured using protein extracts of tobacco. pBI101 and pBI121 indicate negative control and positive control, respectively

Full size image

Then, the coding sequence of this gene was inserted into the plant expression vectors and transformed into tomato (cv. Micro-Tom). The transgenic lines were validated by genomic polymerase chain reaction (PCR) and qRT-PCR. We found the most transgenic lines showed remarkable nematode resistance with less root knots compared to control plants (Fig. 3d; Additional file 2: Fig. S18).

We validated this promoter deletion (35 bp) and found that it co-segregated with resistant phenotype of the seedlings in the BC1 population. To identify the active region of the promoter in Prupe.2G053600, we amplified 161, 282, 693, 1497, and 2063 bp of the 5′ flanking region of the gene in “Hong Gen Gan Su Tao 1#’ and linked the amplified products with the β-glucuronidase (GUS) coding sequence to transiently transform into Nicotiana tabacum. Leaves from the transgenic lines were analyzed for GUS activity by histochemical GUS staining and GUS quantitative enzyme activity determination. The lines carrying the various Prupe.2G053600 promoters displayed remarkable but lesser GUS activity in comparison with the CaMV35S transformed one (pBI121 vector). An increase in GUS expression was observed with promoters longer than 693 bp, indicating the deletion (310 bp ahead of start codon) could drive the expression of Prupe.2G053600 (Fig. 3e). We also constructed a mutation by recovery of 35 bp deletion to the promoter of “Hong Gen Gan Su Tao 1#” peach and linked it with GUS gene. The result found that the GUS activity of this mutation was significantly inhibited compared with control (Additional file 2: Fig. S19). In addition, we analyzed the difference of cis-elements in the promoter region of the resistant accession “Hong Gen Gan Su Tao 1#” and the susceptible one “Bailey.” The result showed that although the former accession was less than the latter by four TATA-box Elements, three more CAAAT-box elements in the former than that of the latter may be the main reason for its higher activity in regulating the expression of the target gene (Additional file 1: Table S21).

Combined RNA-seq and selection sweep analysis to dissect the high-altitude adaptability in P. mira

P. mira is an attractive model for studying high-altitude adaptability of perennial plants because it originated in the QTP in China. The region has an average elevation of ∼4000 m above the sea level, and the oxygen concentration is ∼40% lower and UV radiation is ∼30% stronger than those at the sea level [28]. Up to date, knowledge on the mechanism of high-altitude adaptability has been reported in pig [29], yak [30], human [28, 31], snake [32], hulless barley [33], and the herbaceous plant Crucihimalaya himalaica [34]. However, little is known in perennial crops about the genetic basis of response to harsh conditions, such as low temperature and high UV radiation in high-altitude environments.

Population genomic analyses were performed to analyze the high-altitude adaptability of P. mira. A total of 49 accessions (Additional file 1: Table S18; Accessions 127-175) belonging to the species with an altitude ranging from 2293 to 4390 m were resequenced to an average depth of 24.96×, and the sequencing reads were aligned to the P. mira genome to identify a total of 1,187,611 SNPs. Based on the altitude of these samples (Additional file 2: Fig. S20), the phylogenetic and structure analyses using the identified SNPs, 12 accessions thought to have not corresponded to their altitude categories or in an admixture subgroup between high and low altitude subgroups were excluded in the following analysis. Finally, ten accessions were classified into a high-altitude subgroup and 27 into the low-altitude subgroup (Additional file 2: Fig. S21). We calculated and compared the nucleotide diversity (π; Fig. 4a), Tajima’s D (Fig. 4b), and fixation index (FST, Fig. 4c) values using SNPs across the genome of high- and low-altitude groups, resulting in the identification of selective sweeps of a total of 1.9 Mb and containing 231 genes (Additional file 1: Table S22). These genes were mostly involved in resistance to a series of stresses, such as cold, UV light, and DNA damage (Additional file 1: Table S23; Additional file 2: Fig. S22).

Fig. 4.
figure 4

Selective regions associated with high-altitude adaptation in P. mira. a–c Domestication signals in accessions originating in high-altitude region compared to those in low-altitude. The signals were defined by the top 5% of πratio (a), Tajima’s D (b), and FST values (c). d Distribution of expression of genes induced by low temperature and UV of P. mira. Grey dots indicate the background genes and red dots indicate selective genes associated with high-altitude adaptation. e Detailed π ratio and FST values in the genome region of the candidate gene, Pmi02g3025 (pointed by the dashed line), which was substantially induced by low temperature. f Detailed Tajima’s D in the genome region of the candidate gene Pmi02g3025. g A. thaliana plants expressing Pmi02g3025 gene (OE) and the control (WT) treated with low temperature. h Genotypes (K indicates G/T) of a variation (Chr. 2: 28,612,439 bp) located at the promoter of Pmi02g3025 in accessions from different altitude regions

Full size image

Furthermore, using the young seedlings of P. mira treated under low temperature (0 °C) and UV-light (~ 100 μW/cm2) for 10 h, we found that most genes presented stronger induction by UV than by low temperature based on the RNA-Seq data (Fig. 4d). Gene expression analysis showed that genes encoding C-repeat binding transcription factor (CBF) and late embryogenesis abundant protein (LEA) are strongly expressed after being induced by low temperature, while the enrichment of thaumatin-like protein plays a key role in the UV response in P. mira. One CBF gene, Pmi02g3025, showed more than 240-fold induction of expression by cold and about 6-fold induction by UV-light. According to the FST and Tajima’s D values, this gene was indeed under selection by altitude (Fig. 4e, f). In addition, we heterologously expressed the Pmi02g3025 gene in Arabidopsis and treated the T2 transgenic plants under 0 °C for 48 h. The transgenic Arabidopsis seedlings showed increased resistance to low temperature compared to the wild type (Fig. 4g). We found that the gene expression of Pmi02g3025 at 2 h after cold induction were higher than 0 h in the transgenic plants (Additional file 2: Fig. S23). And the change of proline contents in wild-type and transgenic plants after cold induction also proved transgenic ones have a higher resistance to low temperature (Additional file 2: Fig. S24).

Based on resequencing data, we found seven SNPs and an indel in the promoter regions having a strong association with the phenotype, such as a SNP (Fig. 4 h) and a 19 bp deletion (Additional file 2: Fig. S25) located at 1222 bp and 212 bp upstream of the start codon, respectively. The minor allele frequency of the former was 0.72 and 0 in low and high altitude groups, respectively. And the resequencing results showed that all high-altitude varieties had a 19 bp deletion in promoter of Pmi02g3025. All these results indicated that an obvious divergence existed between high and low altitude groups. Subsequent experiments also proved the relationship between the deletion and promoter activity.

Discussion

In peach, a high-quality reference genome of P. persica was released and has since widely used as a valuable resource for effectively mining candidate genes for important traits [14]. However, this genome sequence alone is not adequate to uncover wild-specific sequences which might have been lost during domestication or artificial selection [35]. In this study, we de novo assembled the genomes of four high-quality wild relatives of P. persica. We analyzed the divergence time of peach and its relatives and found that the divergence time of Prunus genus from others was 56.6 Mya, later than that in other studies, 66.2 Mya [16]. However, the divergence time between the wild relative of peach and almond was 13.3 Mya, which was significantly earlier than previous report, 4.99 Mya [4]. We believe that the more number of species analyzed in our study and the differential software used in previous two studies and our (BEAST vs PAML software) was the main reasons for the differences between different reports.

Furthermore, using this large-scale comprehensive dataset, millions of genomic variations including SNPs and SVs were identified. In addition, hundreds of specific gene families in each of the wild peach species were also identified. The above gene sets represent a useful source for in-depth functional genomic studies including the identification of a nematode resistance gene from P. kansuensis. We identified the genotype of the 35 bp deletion located in the promoter of Prupe. 2G053600 in other accessions and found that it has no correlation with resistance (Additional file 1: Table S24). In addition, the putative interspecific hybridization of P. davidiana identified in this study expands our understanding of the evolutionary path of the species and the molecular mechanisms underlying multiple resistance traits. Moreover, based on expanded gene families and comparative genomic analysis using different accessions of P. mira originating from low- and high-altitude regions, we found that the CBF genes play an important role in high-altitude adaptation of P. mira. However, analyzing the distribution of the 19 bp deletion in varieties with different low temperature resistance indicated that the variation was specific in P. mira which lived in high altitude region (Additional file 1: Table S25).

In addition, we found that gene families involved in vitamin B6 metabolism were significantly expanded in P. mira and P. ferganensis. Further analysis indicated that among the above expanded gene families, a total of 148 genes encoding pyridoxal 5′-phosphate synthase were identified in P. mira and the corresponding gene numbers decreased to 6, 19, and 92 in P. davidiana, P. kansuensis, and P. ferganensis, respectively. Enzyme encoded by this gene family catalyzes the hydrolysis of glutamine to glutamate and ammonia as part of the biosynthesis of pyridoxal 5'-phosphate [36]. It was reported that pyridoxal 5'-phosphate was an essential compound to improve salt-stress tolerance in Arabidopsis [37], wheat [38] and rice [39]. The component also responded to several stresses such as chilling, UV radiation, and drought in tobacco [40]. However, there were few reports on the involvement of the vitamin B6 metabolic pathway in altitude adaptation in plants. Interestingly, the researcher compared gene families in the genome of Triplophysa bleekeri, an endemic fish inhabiting high-altitude regions, to those fish species living in non-QTP regions and found that the expanded gene families in former species were primarily enriched in categories of vitamin B6 metabolism [41]. Furthermore, after sequencing the 16S rRNA genes of oral [42] or skin [43] microbiomes which sampled from Tibetans at different altitude in Tibet, the vitamin B6 metabolic pathway was all found to be enriched in the higher altitude group. Therefore, the expansion of gene involved in vitamin B6 metabolism in P. mira might explain its high-altitude adaptability.

In recent years, we and other groups have made important progress in peach evolution, divergence, and excellent genes identification. However, the main ideas of these studies were different from that of this study. For example, by performing large-scale resequencing of 10 wild and 74 cultivated peach varieties, we show that P. mira is the most primitive wild species of peach, and P. ferganensis can be classified as a subgroup of P. persica [3]. Obviously, gene introgression analysis was not performed in the above study. However, in the study, we constructed the phylogenetic trees of different species and speculated that the specific genes in P. davidiana may be derived from the hybridization of P. mira and almond (Fig. 2). To verify this speculation, P. kansuensis and P. davidiana genomes were firstly aligned to P. dulcis. The results showed that the mapping rate of the latter (65.15%) was slight higher than that of the former (63.73%). At the same time, these two genomes were also aligned with P. mira and found that the rate was 68.52% in the latter, which was also slight higher than the former (66.91%), confirming that P. kansuensis was more evolutional than P. davidiana during peach evolution. Different from this result, Yu et al. [4] found that an ancient introgression that existed between P. mira and the common ancestor of P. kansuensis and P. persica might lead to the development of fruit edibility in peach. Genome assembly and identification and functional verification of several excellent genes are not involved in the above studies. Here, we first completed the assembly of four wild species of P. persica and identified the nematode-resistant genes of P. kansuensis and the altitude adaptability-related genes of P. mira. However, there are still significant differences between the discovery of important genes in the study and that reported before [8] which mainly focused on fruit quality traits.

Moreover, when analyzing the genetic variations in key genes controlling various adaptive traits of different peach species, we found that the source of variations was diverse, including gene expansion, hybridization, and mutation. In previous studies, genome expansion and contraction of repetitive elements within the genome was reckoned to fuel evolutionary diversity in many plant species [44]. And adaptive introgression or sex might represent more efficient management option to conserve evolutionary potential in a changing environment [45, 46]. In view of the different origin of peach species, we speculated that the different sources of the variations were related to the intensity of the environmental change of the species. But, more evidences are needed to prove the precise mechanism underlying plant adaptation to changing environment during evolution.

Conclusions

In summary, the study reported four high-quality genome assemblies and found different wild species contained a large number of specific genetic variations, including PAV and indels. To demonstrate their utility for basic genetics research and plant breeding applications, the role of some genetic variations in the resistance formation in wild peach species were dissected. We believe these genomic resources will increase our understanding of the evolution of fruit crops and be available to accelerating crop improvement in future.

Methods

Plant materials

In this study, different samples were used for DNA sequencing. First, genomes of four wild accessions (2010-138, Zhou Xing Shan Tao 1#, Hong Gen Gan Su Tao 1#, and Ka Shi 1#) were sequenced and de novo assembled. Second, 175 peach accessions were selected for genome resequencing. The above accessions were conserved in the National Germplasm Resource Repository of Peach at Zhengzhou Fruit Research Institute, Chinese Academy of Agricultural Sciences, China. Among them, most accessions belonging to P. mira were sampled from Tibet with different altitudes. Third, one BC1 population was constructed between “Hong Gen Gan Su Tao 1#” (P. kansuensis) and “Bailey” (P. persica) to identify QTLs linked to nematode resistance. Resistance to nematode in this BC1 population was evaluated previously by our group [3]. Genomic DNA was extracted from young leaves using the Plant Genomic DNA kit (Tiangen, Beijing, China).

Moreover, different samples were used for RNA sequencing (RNA-Seq). First, young leaves, mature fruits, seeds, phloem, and roots (obtained through asexual reproduction) of P. persica (Shang Hai Shui Mi), P. ferganensis (Kashi 1#), P. kansuensis (Hong Gen Gan Su Tao 1#), P. davidiana (Hong Hua Shan Tao), and P. mira (2010-138) were collected. Second, roots of “Hong Gen Gan Su Tao 1#” and “Bailey” infected with Meloidogyne incognita for 3, 6, 9, 12 h were collected for RNA-Seq analysis which was reported before [27].

Genome sequencing of wild peach species

The genomes of four wild species of peach were sequenced using different platforms including PacBio Sequel and Illumina according to the manufacturers’ protocols. Library construction and sequencing were performed at Novogene Bioinformatics Technology Co., Ltd (Tianjin, China). For short-read sequencing, two short-insert libraries (230 bp and 500 bp) and 4 large-insert libraries (2 kb, 5 kb, 10 kb, and 20 kb) were constructed for P. mira and P. davidiana, while two short-insert libraries and 2 large-insert libraries (2 kb and 5 kb) were constructed for P. kansuensis and P. ferganensis. These libraries were sequenced on an Illumina HiSeq X Ten platform (Illumina, San Diego, CA).

For single molecule real-time (SMRT) sequencing, a 15-kb library was constructed and sequenced on the PacBio Sequel II platform (Pacific Biosciences, Menlo Park, CA). On the basis of subreads, the ccs software (https://github.com/PacificBiosciences/ccs) was used to generate HiFi reads. The parameter was set as below: min-passes = 3, min-rq = 0.99.

For chromosome conformation capture (Hi-C) sequencing, leaves fixed in 1% (vol/vol) formaldehyde were used for library construction. Cell lysis, chromatin digestion, proximity-ligation treatments, DNA recovery, and subsequent DNA manipulations were performed as previously described [47]. DpnII was used as the restriction enzyme in chromatin digestion. The Hi-C library was sequenced on the Illumina HiSeq X Ten platform (Illumina, San Diego, CA) to generate 150 bp paired-end reads.

RNA-Seq data generation

To assist protein-coding gene predictions, we performed RNA-Seq using five different tissues for each species of 15 years old, and for each sample, three independent biological replicates were generated. Total RNA was extracted with the RNA Extraction Kit (Aidlab, Beijing, China), following the manufacturer’s protocol. RNA-Seq libraries were prepared with the Illumina standard mRNA-seq library preparation kit and sequenced on a HiSeq 2500 system (Illumina, San Diego, CA) with paired-end mode.

Genome assembly of wild peach species

The genome sizes of the four wild peach species were estimated by K-mer analysis. The occurrences of K-mer with a peak depth were counted using Illumina paired-end reads, and genome sizes were calculated according to the formula: total number of K-mers/depth at the K-mer peak, using JELLYFISH 2.1.3 software [48] with K set to 17. Firstly, all HiFi reads which obtained from the above procedure were de novo assembled using Hifiasm (v0.13-r308) [49]. Then, Hi-C data were used to correct superscaffolds and cluster the scaffolds into pseudochromosomes through ALLHiC [50] pipeline.

To evaluate the quality of the genome assemblies, we performed BUSCO v4.0.0 [15] analysis on the four assembled genomes with the 1,614 conserved plant single-copy orthologs. Moreover, RNA-Seq data with different tissues were used to validate the results of gene prediction (Additional file 1: Table S5).

Repetitive element identification

A combined strategy based on homology alignment and de novo search was used to identify repeat elements in the four wild peach genomes. For de novo prediction of TEs, we used RepeatModeler (http://www.repeatmasker.org/RepeatModeler.html), RepeatScout (http://www.repeatmasker.org/), Piler [51], and LTR-Finder [52] with default parameters. For alignment of homologous sequences to identify repeats in the assembled genomes, we used RepeatProteinMask and RepeatMasker (http://www.repeatmasker.org) with the repbase library [53]. TEs overlapping with the same type of repeats were integrated, while those with low scores were removed if they overlapped more than 80 percent of their lengths and belonged to different types.

Gene prediction and functional annotation

Gene prediction was performed using a combination of homology, ab initio and transcriptome based approaches. For homology-based prediction, protein sequences from P. persica, Pyrus bretschneideri, P. mume, Malus domestica, and Fragaria vesca (Genome Database for Rosaceae; https://www.rosaceae.org) and Vitis vinifera (http://www.genoscope.cns.fr/externe/GenomeBrowser/Vitis/) and Arabidopsis thaliana (https://www.arabidopsis.org) were downloaded and aligned to the peach assemblies. Augustus [54], GlimmerHMM [55], and SNAP [56] were used for ab initio predictions. For transcriptome-based prediction, RNA-Seq data derived from root, phloem, leaf, flower, and fruit were mapped to the assemblies using HISAT2 software [57] and assembled into the transcripts using Cufflinks (version 2.1.1) with a reference-guided approach [58]. Moreover, RNA-Seq data were also de novo assembled using Trinity v2.0 [59] and open reading frames in the assembled transcripts were predicted using PASA [60]. Finally, gene models generated from all three approaches were integrated using EvidenceModeler [60] to generate the final consensus gene models.

The predicted genes were functionally annotated by comparing their protein sequences against the NCBI non-redundant (nr), Swiss-Prot (http://www.uniprot.org/), TrEMBL (http://www.uniprot.org/), Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/), InterPro, and Gene ontology (GO) databases.

tRNAscan-SE [61] was used with default parameters to identify tRNA sequences in the genome assemblies. rRNAs in the genomes were identified by aligning the reference rRNA sequence of relative species to the assemblies using BLAST with E-values < 1e−10 and nucleotide sequence identities > 95%. Finally, the INFERNAL v1.1 (http://infernal.janelia.org/) software was used to compare the genome assemblies with the Rfam database (http://rfam.xfam.org/) to predict miRNA and snRNA sequences.

Genome alignment and collinearity analysis

Orthologous genes within the P. mira and P. persica genomes were identified using BLASTP (E value <1e−5), and MCScanX [62] was used to identify syntenic blocks between the two genomes. The collinearity of the two genomes was then plotted according to the identified syntenic blocks.

Four wild peach genomes were aligned to the P. persica genome using LASTZ [63] with the parameters of “M=254K=4500 L=3000 Y=15000 --seed=match 12 --step=20 --identity=85.” In order to avoid the interference caused by repetitive sequences in alignments, RepeatMasker (http://www.repeatmasker.org) and RepBase library [53] were used to mask the repetitive sequences in genomes of P. persica and four wild species. The raw alignments were combined into larger blocks using the ChainNet algorithm implemented in LASTZ [63].

Variation identification

We identified SNPs and small indels (< 50 bp) between the four wild and reference peach genomes using SAMtools (http://samtools.sourceforge.net/) and LAST (https://gitlab.com/mcfrith/last) with parameters “-m20 -E0.05” and SNP and indel filtering criteria “minimum quality = 20, minimum depth = 5, maximum depth = 200.” SVs were identified from genome alignments by LAST with parameters “-m20 -E0.05.” The software has been used in many studies [64, 65] for long read or genome alignments. CNVs were identified using CNVnator-0.3.3 [66]. And observation by IGV software and PCR amplification (Additional file 2: Fig. S4; Additional file 1: Table S15 and 16) was used to confirm the accuracy of SVs.

BSA to locate nematode resistance gene

Using a previously constructed BC1 population [24], we extracted genomic DNA from fresh leaves of its parents and two bulked populations which mixed by 20 resistance seedlings and 20 susceptible ones. The samples were used for resequencing on an Illumina HiSeq 2500 platform (Illumina, San Diego, CA) to generate 125 bp paired-end reads. Then BWA software [67] was used to align the clean reads of each sample against the reference genome of Prunus persica v2.0.a1 [14].

Variants calling were performed for all samples by using GATK3.3 software [68] to identify a total of 2,442,036 SNPs on all 8 chromosomes. The homozygous SNPs between two parents were extracted from the vcf files. The reads depth information for homozygous SNPs above in the offspring pools was gained to calculate the SNP index [69]. The loci whose SNP index was less than 0.3 or greater than 0.7 in both the pools were filtered. The difference of the SNP index of the two pools was calculated as the Δ(SNP index). Finally, we made a graph of Δ(SNP index) of the whole-genome by the sliding window method.

Promoter activity measurement

A total of five primers upstream and one downstream of the start codon of Prupe.2G053600 were synthesized and used to amplify a series of 5 indel regions in the Prupe.2G053600 promoter using PCR amplification. The amplified PCR products were ligated into pGEM-T easy vector and cloned into pBI101 binary vector after being digested by XbaI and BamHI. Furthermore, each of the 5 amplified products was transformed into A. tunefaciens (GV1301) cells and collected and resuspended in infiltration buffer and then transformed into 6-week-old tobacco leaves using sterilized syringes. The transiently transformed tobacco plants were grown in a growth chamber for 48 h and the infection sites were cut to measure GUS activity as previously described in Jefferson et al. [70]. The pBI121 vector was used as a positive control.

Transgenic analysis

The full-length open reading frame of the Prupe.2G053600 gene was amplified through PCR using cDNA synthesized from RNA that was isolated from root of the “Hong Gen Gan Su Tao 1#” (P. kansuensis). The amplified product was cloned into the pEASY vector driven by the cauliflower mosaic virus (CaMV) 35 S promoter. The resulting vector was transformed into Solanum lycopersicum cv. Micro-Tom by A. tumefaciens C58. The T0 plants were generated and inoculated with M. incognita to observe resistance and measure gene expression.

Similarly, one candidate gene, Pmi02g3025, was cloned from the leaf of “2210-198” (P. mira) and ligated to the vector and transformed into A. thaliana “Columbia.” When the transformed plants were grown to about 5 leaves, low temperature (0 °C) treatment was applied and samples were collected at 0, 24, and 48 h post treatment. Phenotype was observed and the drooping leaves were used to indicate that the accession was susceptible to low temperature. The proline content of leaves in Arabidopsis was measured according to previous reference as previously described [71].

Comparative analysis

Protein sequences from 15 plant species including P. persica (phytozomev10), P. ferganensis, P. kansuensis, P. davidiana, P. mira, Prunus dulcis (https://www.rosaceae.org/species/prunus/prunus_dulsis/lauranne/genome_v1.0), P. armeniaca (https://www.rosaceae.org/species/prunus_armeniaca/genome_v1.0), Prunus mume (http://prunusmumegenome.bjfu.edu.cn/index.jsp), P. salicina (NCBI accession: PRJNA574159), P. avium (NCBI accession: PRJNA395588), M. domestica (NCBI accession: PRJNA534520), P. bretschneideri (NCBI accession: PRJNA259338), Fragaria vesca (phytozome v10), A. thaliana (phytozome v10), and Vitis vinifera (phytozome v10) were used to construct orthologous gene families. To remove redundancy caused by alternative splicing, we retained only the gene model at each gene locus that encoded the longest protein. To exclude putative fragmented genes, genes encoding protein sequences shorter than 50 amino acids were filtered out. All-against-all BLASTp was performed for these protein sequences with an E-value cut-off of 1e−5. OrthoMCL V1.4 [72] was then used to cluster genes into gene families with the parameter “-inflation 1.5.”

Protein sequences from 830 single-copy gene families were used for phylogenetic tree construction. MUSCLE [51] was used for multiple sequence alignment for protein sequences in each single-copy family with default parameters. The alignments from all single-copy families were then concatenated into a super alignment matrix, which was used for phylogenetic tree construction using the maximum likelihood (ML) method implemented in the RAxML software (http://cme.h-its.org/exelixis/web/software/raxml/index.html). Divergence times between the 15 species were estimated using MCMCTree in PAML software (http://abacus.gene.ucl.ac.uk/software/paml.html) with the options “correlated rates” and “JC69” model. A Markov Chain Monte Carlo analysis was run for 10,000 generations, using a burn-in of 10,000 iterations and sample-frequency of 2. Five calibration points were applied according to the TimeTree database (http://www.timetree.org): A. thaliana and V. vinifera (106.9-134.7 Mya), A. thaliana and the common ancestor of F. vesca, M. domestica, and other Prunus species (97.4-109.3 Mya), F. vesca and common ancestor of M. domestica, and other Prunus species (64.5-75.5 Mya), M. domestica, P. bretschneideri, and other Prunus species (51.7-61.8 Mya), M. domestica and P. bretschneideri (11.3-26.4 Mya).

To detect the whole genome duplication events, we first identified collinearity blocks using paralogous gene pairs with software MCScanX [62]. Using the sum of transversion of fourfold degenerate site divided by the sum of fourfold degenerate site, we then calculated 4dTv values of each block. In addition, Ks values of homologous gene pairs were also calculated using PAML [73] based on the sequence alignments by MUSCLE [51], to validate speciation times.

Gene family expansions and contractions

Expansion and contractions of orthologous gene families were determined using CAFE [74], which uses a birth and death process to model gene gain and loss over a phylogeny. Significance of changes in gene family size in a phylogeny was tested by calculating the p-value on each branch using the Viterbi method with a randomly generated likelihood distribution. This method calculates exact p-values for transitions between the parent and child family sizes for all branches of the phylogenetic tree. Enrichment of GO terms and KEGG pathways in the expanded gene families of each of the four wild peach species were identified using the R package clusterProfiler [75].

Identification of an interspecific hybridization of P. davidiana

Using SNPs identified from 126 peach accessions, we performed PCA to evaluate the evolution path using the software GCTA [76].

To trace the origin of genes in P. davidiana, each gene was aligned to other genomes to calculate the alignment score. Genes were classified as putatively originating from the specie which had the highest alignment scores. Collinearity among P. dulcis, P. davidiana, and P. mira was plotted based on the identified syntenic blocks.

Resistance genes

Hidden Markov model search (HMMER; http://hmmer.janelia.org) was used to identify R genes in the four wild peach genomes according to the nucleotide-binding site (NBS) domain (PF00931), Toll and interleukin-1 receptor-like (TIR) model (PF01582), and several leucine-rich repeat (LRR) models (PF00560, PF07723, PF07725, PF12799, PF13306, PF13516, PF13504, PF13855, and PF08263) in the Pfam database (http://pfam.sanger.ac.uk). Coiled-coil (CC) motifs were detected using the COILS prediction program 2.2 (https://embnet.vital-it.ch/software/COILS_form.html) with a p score cut-off of 0.9.

Identification of selective sweeps associated with high-altitude adaptation in P. mira

Raw genome reads of the 49 accessions of P. mira from Tibet, China with different altitudes were processed to remove adaptor, contaminated and low quality sequences, and the cleaned reads were mapped to the assembled P. mira genome using BWA v0.7.8 [67]. Based on the alignments, the potential PCR duplicates were removed using the SAMtools command “rmdup.” SNP calling at the population level was performed using SAMtools [77]. The identified SNPs supported by at least five mapped reads, mapping quality ≥ 20, and Phred-scaled genotype quality ≥ 5, and with less than 0.2 missing data were considered high-quality SNPs (1,394,483) and used for subsequent analyses.

A phylogenetic tree constructed using TreeBeST 1.9.2 [78] with 1000 bootstraps and structure analysis was performed for the 49 accessions of P. mira using the program frappe [79] with the number of assumed genetic clusters (K) ranging from two to three, and 10,000 iterations for each run.

To identify genome-wide selective sweeps associated with high-altitude adaptation, we scanned the genome in 50-kb sliding windows with a step size of 10 kb and calculated the reduction inπ based on the P. mira accessions originating in high-altitude to low-altitude regions (πhigh/πlow). In addition, Tajima’s D and FST between the two groups were also calculated. Windows with the top 5% of the π ratios, Tajima’s D ratio, and FST values were considered as selective sweeps.