Introduction

The genus Malus Mill. (Rosaceae) includes economically important apple species with cultivated apple fruits and wild crabapples. The domesticated apple, M. domestica Borkh., is one of the most widely cultivated fruit crops in temperate regions worldwide, and wide varieties of apple cultivars are bred for various tastes and use (Korban and Skirvin 1984; Morgan and Richards 2003). Its wild relatives, known as crabapples, usually having fruits less than 5 cm in diameter, offer a useful source of genetic diversity for apple breeding (Brown 2012), and are widely planted as ornamental and landscaping trees. The use of these natural resources in breeding and the development of effective conservation programs for apples require a good understanding of the genetic relationships as well as the genetic polymorphisms within and between cultivated apples and related wild crabapples (Cornille et al. 2014). Wild crabapples also provide habitats for wildlife and serve as a direct source of food for both human and wildlife or as components of hedges in agricultural landscapes.

The genus Malus comprises approximately 25 to 55 species, divided into five sections including six series (Harris et al. 2002; Phipps et al. 1990; Rehder 1974; Robinson et al. 2001), although recent phylogenetic studies have suggested elevating several series to sections as specified in Table 1 (Langenfelds 1991; Jiang et al. 1996; Qian et al. 2006, 2008). This study follows Phipps et al.’s treatment (1990) of infrageneric Malus classification (Table 1), adapted from Rehder (1974), Huckins (1972), and Williams (1982). The section Malus is traditionally sub-divided into two series, Malus and Baccatae (Rehder) Rehder, based on differences in fruit size and deciduous or persistent calyces. In the nomenclature of the taxa in the series, Malus of section Malus accommodating the cultivated apple, M. domestica, is complex. With few discrete characters to differentiate species, the morphological characters used to delimit the species are continuous and overlapping. Moreover, the intimate association that humans have with apples has blurred the distinction between wild and cultivated species, which has been further complicated by hybridization. The origin of the cultivated apple from its wild progenitors is relevant; however, difficulty in the delimitation of progenitor species has hampered investigations of parental contribution to its origin (Harris et al. 2002; Robinson et al. 2001). Additionally, the genetic identification of cultivars of artificial cross hybridization is difficult to determine their phylogenetic relationships or population genetic inference. The artificial selection that has been repeated through vegetative reproduction and re-crossing for a long period must have yielded a phylogeny like a network or disturbed the genetic structure of populations (Forte et al. 2002). Recently, several studies have provided insight into the origin of the cultivated apple and pointed out that several different wild species could have contributed organelle and nuclear genomes to the domesticated apple. The current most widely accepted theory based on morphological (Forte et al. 2002), phylogenetic (Forte et al. 2002; Harris et al. 2002; Robinson et al. 2001), population genetic (Cornille et al. 2012; Coart et al. 2006), and genome-wide (Nikiforova et al. 2013; Velasco et al. 2010) evidence suggests that M. sieversii (Ldb.) Roem in the Tian Shan Mountains of Central Asia was initially domesticated, and subsequently dispersed to West Europe along the great trade route known as the Silk Route, allowing hybridization and introgression of other wild crabapples from Siberia (M. baccata), Caucasus (M. orientalis Uglitz.), and Europe (M. sylvestris Mill) (Cornille et al. 2014). In addition to the initial progenitor, M. sieversii, the wild European crabapple, M. sylvestris, was, specifically, identified by population genetic study using microsatellite markers (Cornille et al. 2012), to be a major secondary contributor to the gene pool of current varieties of cultivated apple.

Table 1 Infrageneric classifications of Malus used in this study
Full size table

Malus baccata (L.) Borkh. in the series Baccatae of section Malus is a 10–14 m tall tree commonly called Siberian crabapple. It is native to Bhutan, China, India, Kashmir, Korea, Mongolia, Nepal, and Russia (Siberia) in northern Asia (Gu and Spongberg 2003), and has been introduced to Japan, Europe, northeastern USA, and Canada (USDA Natural Resources Conservation Service n.d.). It is one of the wild relatives of M. domestica that can be readily hybridized with varieties of cultivated apples (Cornille et al. 2014). Therefore, it is widely used as a rootstock and breeding resource in high-latitude apple-producing areas because of its disease resistance, and cold tolerance (Chen et al. 2019). Malus toringo (Siebold) Siebold ex de Vriese (Toringo or Siebold crabapple) is a 2–6 m tall shrub distributed naturally in China, Japan, and Korea (Iketani and Ohashi 2001; Lee 2007), and introduced to the USA and Europe as high horticultural value ornamental trees with semi-weeping branches (Dickson 2015); it is sometimes referred to by its illegitimate name, Malus sieboldii (Regel) Rehd. (Akiyama et al. 2014). M. toringo was traditionally classified in the series Sieboldianae of section Sorbomalus based on its lobed young leaves (among older entire leaves) and conduplicate bud vernation; however, M. baccata was placed in the series Baccatae of section Malus owing to its entire leaves and convolute or involute vernation (Fig. 1) (Phipps et al. 1990; Rehder 1974). The necessity for additional studies to clarify the systematic position of series Sieboldianae has been recognized, especially concerning its relationship to the series Baccatae. Despite different sectional assignments in Malus and Sorbomalus, the series of Baccatae and Sieboldianae displayed genetic similarity from molecular evidence as well as in morphology. Morphologically, they share several characters in common: flowers with deciduous calyces, umbellate inflorescences, and no or relatively few sclereids in fruits (Phipps et al. 1990; Rehder 1974; Forte et al. 2002). In nuclear DNA internal transcribed spacer (ITS) phylogeny, section Sorbomalus was polyphyletic, and the species belonging to series Sieboldianae (M. sieboldii and M. sargentii) were nested in the clade comprising the species of series Baccatae and Sieboldianae, suggesting their common origin (Forte et al. 2002; Harris et al. 2002; Robinson et al. 2001). Such phylogenetic similarity has also been revealed by amplified fragment length polymorphism (AFLP) and retrotransposon-based polymorphism analyses (Savelyeva and Kudryavtsev 2015; Savelyeva et al. 2017). However, chloroplast DNA (cpDNA) matK phylogeny and randomly amplified polymorphic DNA (RAPD) analyses did not provide sufficient resolution to infer robust species relationships (Forte et al. 2002; Harris et al. 2002; Robinson et al. 2001). Haplotype analyses of expanded cpDNA regions (trnH-psbA, trnS-trnG, trnL-trnF, and trnT-trnL) were also inconclusive owing to low resolution (Volk et al. 2015). Simple sequence repeat (SSR) markers have turned out to be inadequate in resolving interspecific relationships concordant with taxonomic classification or geographic origin, or both, although they have proven to be quite robust for many germplasm management applications (Hokanson et al. 2001). Disentangling the relationships between these two series has been exceptionally problematic owing to the high degree of hybridization between them and the application of the name “wild apple” to these hybrids, which blurs the boundaries between the two series. Robinson et al. (2001) claimed that the series Sieboldianae could be of hybrid origin, formed by hybridization between a series Baccatae taxon and a section Sorbomalus taxon.

Fig. 1
figure 1

a Gene map of the chloroplast genomes of Malus baccata and Malus toringo collected from China, Japan, and Korea and sequenced in this study. b Flowers and leaves of Malus baccata. c Flowers and leaves of Malus toringo. Photo credit: Min Sung Cho, Sungkyunkwan University, Republic of Korea

Full size image

To investigate their phylogenetic relationships, we sequenced and assembled four whole chloroplast genomes of two species representing the series Baccatae and Sieboldianae: one wild accession of M. baccata (series Baccatae; section Malus) from Korea and three accessions of M. toringo (series Sieboldianae; section Sorbomalus) from China, Japan, and Korea. The genetic and morphological similarity between M. baccata and M. toringo has been reported in previous studies despite taxonomic assignments into different sections of Malus and Sorbomalus, respectively, but their phylogenetic relationship has never been determined clearly yet. Specifically, we sampled M. toringo from natural environments in three countries of Korea, China, and Japan to examine the plastome variation among allopatric populations of M. toringo. With the advent of high-throughput sequencing technologies for next-generation sequencing (NGS), massive amounts of data are now available which improves the poor resolution in previous cpDNA phylogenies, to reveal considerable genome-wide variation in the sequences and structures of entire plastomes. Comparative genomic analysis of whole plastomes is now available as one of the effective markers to infer the phylogenetic relationships and evolutionary histories of numerous plant groups, including Rosaceae (Cheng et al. 2017; Daniell et al. 2016; Jansen et al. 2011; Jeon and Kim 2019; Njuguna et al. 2013; Parks et al. 2009; Terakami et al. 2012; Yang et al. 2018). Genome-wide data of Malus plastomes could provide vital information regarding genetic variation among wild crabapple species, not only increasing phylogenetic resolutions but also enhancing our understanding of organelle genome evolution. Based on the complete plastome sequences, we tested the previous phylogenetic hypotheses, focusing particularly on the relationship between Baccatae and Sieboldianae. Comparative plastome analyses allowed us to determine the structure, gene content, and rearrangements in the chloroplast genomes of wild Malus crabapples. Furthermore, highly variable chloroplast regions were identified as potentially useful markers for crabapples. Lastly, this study provided a glimpse into the infraspecific plastome variation of one of East Asian crabapple species, M. toringo.

Materials and methods

Plant sampling, DNA isolation, and plastome sequencing/annotation

The silica-gel dried leaves of four Malus wild crabapples were collected in the field; we collected one sample of M. baccata from a forest in Gangwon-do, Korea (37° 49′ 29.9″ N, 128° 21′ 46.5″ E, altitude 535 m), and three samples of M. toringo from three countries, i.e., Zhejiang, China (30° 18′ 01.2″ N, 119° 07′ 02.0″ E, 1117 m), Nagano-ken, Japan (35° 54′ 14.4″ N, 138° 09′ 43.0″ E, 1724 m), and Jeollanam-do, Korea (34° 57′ 50.3″ N, 127° 25′ 59.4″ E, 441 m). The voucher specimens were collected and deposited in the Ha Eun Herbarium (SKK) at Sungkyunkwan University, Republic of Korea. The total genomic DNA was isolated using a DNeasy Plant Mini Kit (Qiagen, Carlsbad, CA, USA) following the manufacturer’s protocol. An Illumina paired-end (PE) genomic library was constructed and sequenced using the Illumina HiSeq platform (Illumina, Inc., San Diego, CA, USA) at Macrogen Corporation (Seoul, Korea). By the de novo genomic assembler, Velvet 1.2.10 (Zerbino and Birney 2008), four Malus plastid genomes were assembled from the produced paired-end sequence reads (i.e., total 42,661,706 reads for M. baccata; 113,097,408, M. toringo China; 123,360,708, M. toringo Japan; 87,614,680, M. toringo Korea) with coverages of 364x (M. baccata), 289x (M. toringo China), 921x (M. toringo Japan), and 1054x (M. toringo Korea), respectively. The programs of Velveth and velvetg in Velvet were run using the optimized parameters of various hash length (k-mer) and coverage values to assemble each plastome. Annotation was performed using the Dual Organellar GenoMe Annotator (Wyman et al. 2004), ARAGORN v1.2.36 (Laslett and Canback 2004), and RNAmmer 1.2 Server (Lagesen et al. 2007). Using Geneious R10 (Biomatters, Auckland, New Zealand) (Kearse et al. 2012), annotation was inspected and corrected manually by comparison with other Malus plastomes. The annotated plastome sequences were deposited in the GenBank databank under the accession numbers MK571561 for M. baccata, and MK571562, MK571563, and MK571564 for M. toringo from China, Japan, and Korea, respectively. The annotated GenBank (NCBI, Bethesda, MD, USA) format sequence file was used to draw a circular map (Fig. 1) using the OGDRAW software v1.2 (CHLOROBOX, Postdam-Golm, Germany) (Lohse et al. 2009).

Comparative plastome analysis

We performed several comparative plastome analyses among the eight Malus plastomes including the previously reported four Malus species representing its major sections, i.e., M. angustifolia from section Chloromeles (NC_045410); M. sieversii from section Malus (NC_045390); M. tschonoskii from section Docyniopsis (NC_035672); and M. trilobata from Eriolobus (NC_035671); and those of M. baccata and M. toringo assembled in this study. The codon usage frequency was calculated using MEGA7 (Kumar et al. 2016) with relative synonymous codon usage (RSCU) value, which is the relative frequency of occurrence of the synonymous codon for a specific amino acid. The online program predictive RNA editor for plants (PREP) suite (Mower 2009) was used to predict the potential RNA editing sites for annotated protein-coding genes with 35 reference genes available with known edit sites, based on a cutoff value of 0.8 (suggested as optimal for PREP-Cp). Overall sequence divergence was estimated using the LAGAN alignment mode (Brudno et al. 2003) in mVISTA (Frazer et al. 2004). The nucleotide diversity (Pi) was calculated using sliding window analysis (window length = 1000 bp and step size = 200 bp, excluding sites with alignment gaps) to detect the most divergent regions (i.e., mutation hotspots) in DnaSP (Librado and Rozas 2009). Two types of repeat sequences were identified and compared among eight plastid genomes. REPuter (Kurtz et al. 2001) was used to detect the various types of repetitive sequences with search parameters of maximum computed repeats = 50, minimum repeat size = 8 bp, and hamming distance=1. Simple sequence repeats (SSRs) were identified using MISA web (http://pgrc.ipk-gatersleben.de/misa/) with search parameters of 1–15 (unit size-minimum repeats, i.e., mono-nucleotide motifs with 15 minimum numbers of repetition), 2-5, 3-3, 4-3, 5-3, and 6-3 with 100 interruption (maximum difference for two SSRs). To evaluate natural selection pressure in the protein-coding genes of the eight plastomes, the rates of nonsynonymous to synonymous substitution (ω=dN/dS) were estimated using EasyCodeML (Gao et al. 2019) based on PAML (Phylogenetic Analysis by Maximum Likelihood) algorithms (Yang 1997). Seven codon substitution models with heterogeneous ω across sites (Yang et al. 2000) implemented in EasyCodeML were employed to investigate the aligned sequences of protein-coding genes of eight Malus plastomes, M0 (one ratio), M1a (nearly neutral), M2a (positive selection), M3 (discrete), M7 (beta), M8 (beta and ω>1), and M8a (beta and ω=1). The fit of these models to the sequence data was compared in preset running mode using likelihood ratio test (LRT), and the pairwise comparison of codon models, M7 vs. M8, was effective for identifying amino acid residues that have potentially evolved under selection among Malus plastomes. The site potentially evolving under positive selection was presented based on the posterior probability higher than the standard threshold (0.95) (Scheffler and Seoighe 2005) calculated by the Bayes empirical Bayes (BEB) method (Yang et al. 2005).

Phylogenetic analysis

Phylogenetic relationships of the newly sequenced accessions of M. baccata and M. toringo assembled in this study were investigated in the context of their relationships with other Malus species. We analyzed 23 plastome sequences of major sections of the genus Malus: eight accessions of section Malus including two accessions of M. baccata (MK571561 and NC045389), M. domestica cultivar M9 (MK434916), M. halliana (MT246302), M. hupehensis (NC040170), M. micromalus (NC036368), M. prunifolia (NC031163), and M. sieversii (NC045390); nine accessions of section Sorbomalus, i.e., five accessions of M. torigno (MT268884, NC050059, MK571562, MK571563, and MK571564), M. florentina (NC035625), M. toringoides (NC049113), M. transitoria (MK098838), and M. yunnanensis (NC039624); three species of section Chloromeles, i.e., M. coronaria (NC045308), M. ioensis (NC045393), and M. angustifolia (NC045410); two species of section Docyniopsis, M. doumeri (NC045343) and M. tschonoskii (NC035672); one species of section Eriolobus, M trilobata (NC035671). The analysis included Pyrus pyrifolia (NC015996) from the same tribe (Maleae) as an outgroup species. The sequences of concatenated sequences of 79 common protein-coding genes (excluding duplicated ones in IR regions) among the Malus species were aligned using MAFFT v. 7 (Katoh and Standley 2013), and the ML phylogenetic tree was constructed using IQ-TREE v. 1.4.2, with 1000 replicate bootstrap analysis (Nguyen et al. 2015). The best-fit evolutionary model was chosen as “TVM+F+I,” which was scored according to the Bayesian information criterion (BIC) scores and weights by testing 88 DNA models of ModelFinder (Kalyaanamoorthy et al. 2017) implemented in IQ-TREE v. 1.4.2.

Results and discussion

Genome features, content, order, and organization of wild Malus plastomes

The plastome of M. baccata contained 160,149 base pairs (bp), and consisted of four typical regions: a large single-copy (LSC) region of 88,260 bp, a small single-copy (SSC) region of 19,181 bp, and a pair of inverted repeat regions (IRs) of 26,354 bp. The total lengths of the three M. toringo plastomes were 160,105 (Japan), 160,138 (Korea), and 160,168 bp (China), and comprised of the same four regions of LSC, SSC, and a pair of IRs (Fig. 1, Table 2). The overall guanine-cytosine (GC) content of M. baccata was 36.5%, while that of M. toringo, 36.5% (China) or 36.6% (Japan and Korea), respectively (Table 2). Each of the four plastomes contained 129 genes, including 84 protein-coding (excluding pseudogenes), eight rRNA, and 37 tRNA genes. Eighteen genes contained introns, including seven tRNA genes. Three genes, clpP, rps12, and ycf3, exhibited two introns. The trnK-UUU gene harbored the largest intron that contained the matK gene within it (Table 3). In total, 16 genes were duplicated in the IR regions, including seven tRNAs, four rRNAs, and six protein-coding genes. The trans-splicing gene rps12, consisting of three exons, was located in the LSC region of exon 1, but exon 2 and exon 3 of the gene were embedded in the IR regions. The infA gene located in the LSC region became a pseudogene, and part of each ycf1 and rps19, duplicated in the IR region, also became pseudogenes.

Table 2 Summary of the genomic characteristics of eight chloroplast genomes of wild Malus species used for comparative genomic analyses in this study
Full size table
Table 3 Genes encoded by eight Malus chloroplast genome
Full size table

A genomic comparison of eight wild Malus plastomes, including M. baccata and M. toringo (sequenced in this study), and four other crabapple species (M. angustifolia, M. sieversii, M. tschonoskii, and M. trilobata) revealed high conservation in their plastome organization. They shared the most common genomic features of sequences, and gene content and numbers, demonstrating a 99.2% pairwise sequence identity, despite their different sectional assignments. Generally, the length of the chloroplast genome and its quadripartite regions vary among plant lineages due to the contraction and expansion of inverted repeat regions. Evaluating their contraction and expansion by comparing the location of the boundaries among the four chloroplast regions (two IRs, LSC, and SSC) can provide some insights into plastome evolution (Menezes et al. 2018). All eight Malus plastomes shared exactly the same genes and similar gene contents at all boundaries among the four regions, with only slight changes in the length of intergenic regions. They all contained the functional protein-coding gene of ycf1 at SSC/IR with its pseudogene copy, ycf1Ψ at IR/SSC, and functional rps19 at LSC/IR with pseudogene copy rps19Ψ at IR/LSC endpoints (Fig. 2).

Fig. 2
figure 2

Comparison of the border positions of the large single-copy (LSC), small single-copy (SSC), and inverted repeat (IR) regions among eight wild Malus chloroplast genomes. Gene names are indicated in colored arrow boxes, and their lengths in the corresponding regions are displayed beside the boxes. Ψ indicates a pseudogene

Full size image

Comparative phylogenomic analyses of wild Malus plastomes

The frequency of codon usage in the eight wild Malus plastomes was calculated based on the sequences of protein-coding genes (Fig. S1, Table S1). The average codon usage in all plastomes was identical at 26,527, except for M. trilobata (26,532), and the patterns of frequently used codons were also consistent among them. The genetic code encoding protein in a mode of triplet codon is said to be redundant in that the same amino acid residue can be encoded by more than one, so-called synonymous codons. Most amino acids are encoded by several synonymous codons, as 64 different codons are translated into 20 amino acids and termination of translation (three stop signals). Synonymous codons are not used in equal frequency, but specific codons are used more often than other synonymous codons during translation of genes. This feature of preferential use of codons is known as codon usage bias. The codon usage bias and variation in codon usage within and among species suggest some selective constraint on codon choice. The frequency of codon usage varies by factors in species-specific ways, showing different preferences for codons used to encode specific amino acids, probably as a result of evolution in the presence of mutational biases, selection for translation rate and accuracy, and possibly other factors (Orešič and Shalloway 1998). Codon usage values are described by the relative synonymous codon usage (RSCU), which is a reflection of how often a particular codon is used relative to the expected number of times that codon would be used in the absence of codon usage bias. In our analyses, all RSCU values for each amino acid considered were similar among the eight plastomes. The highest RSCU value was indicated in the usage of the UUA codon for leucine (1.94–1.95) followed by GCU for alanine (1.84) and AGA for arginine (1.83–1.84), while the lowest were AGC for serine (0.38) and GAC for aspartic acid (0.38). We found that codon usage was biased toward a high RSCU value of U and A at the third codon position as found in other Rosaceae species (Yang et al. 2020).

RNA editing alters the nucleotide sequence of transcribed RNA molecules from that of the DNA template encoding it, which usually results in a change in the amino acid sequence of the translated protein. In plants, RNA editing affects mitochondrial and plastid transcripts of all major lineages of land plant, and the site-specific modification of cytidines to uridines (C-to-U conversion) is prevalent in organellar genomes of all land plants (Chateigner-Boutin and Small 2011). Comparison of editing frequencies and editing patterns shows that RNA editing is a transcript- and species-specific process, but its frequencies and patterns are not correlated with the phylogenetic position of the species, sometimes revealing extensive species-specific divergence among closely related species. Species specificity of the editing frequencies and gene-specific editing patterns suggest multiple independent acquisitions and occasional losses of editing at a specific site (Freyer et al. 1997). This raises questions about the selection pressures acting to maintain editing in the evolution of angiosperms that are yet to be completely resolved. Editing tends to correct the effect of DNA mutations that would otherwise compromise the synthesis of functional proteins, and its additional function could be generating protein diversity or regulating gene expression (Chateigner-Boutin and Small 2011). The extent of variation in the number of RNA editing sites is currently less known at shallow taxonomic levels (i.e., among congeneric species or across multiple genera belonging the same family), although several studies demonstrated that the number of editing sites can vary widely among large taxonomic groups of land plants and also between the two organellar genomes (Corneille et al. 2000; Guo et al. 2015; Tsudzuki et al. 2001). We have found that the RNA editing patterns across the eight Malus plastomes were similar in gene location and codon conversion type of the predicted RNA editing sites; only slight changes were observed in the numbers of editing sites for several codon conversions. The total number of RNA editing sites identified among them ranged from 62 to 64 for 25 of the 35 protein-coding genes. These genes included photosynthesis-related genes (atpA, atpB, atpF, atpI, ndhA, ndhB, ndhD, ndhF, ndhG, petB, petG, psaI, psbE, psbF, and psbL), self-replication genes (rpoA, rpoB, rpoC1, rpoC2, rps2, rps14, and rps16), and others (accD, clpP, and matK). We detected no RNA editing sites in the ccsA, petD, petL, psaB, psbB, rpl2, rpl20, rpl23, rps8, and ycf3 genes. The highest numbers of potential editing sites were found in the NADH dehydrogenase genes, which was consistent with the previous findings in tobacco, maize, rice, and other plants (Corneille et al. 2000; Kim et al. 2019; Tsudzuki et al. 2001; Yang et al. 2020), i.e., ndhB gene was the highest in frequency at 11–12 sites, followed by the ndhD gene at eight sites. Most editing sites were distributed at the 1st and 2nd codon positions (Table S2) as observed in the chloroplast genome of the hornwort Anthoceros formosae (Kugita et al. 2003) and the mitochondrial genes of Arabidopsis (Giegé and Brennicke 1999), whereas the mitochondrial genes of Physarum polycephalum showed the different pattern of codon bias with more editing at the 3rd codon position than at the 1st and 2nd positons within coding regions (Mahendran et al. 1991). The highest conversions in the editing frequencies of codons associated with the corresponding amino acid changes were represented by the changes from serine (S) to leucine (L) (average confidence score of 22.935) followed by proline (P) to leucine (L) (average confidence score of 8.86) (Fig. S2).

The divergence level of nucleotide diversity among the eight plastomes of wild Malus species was visualized by plotting with mVISTA (Frazer et al. 2004), using the plastome of M. tschonoskii of section Docyniopsis as a reference. The results exhibited a high degree of synteny and gene order conservation in the mVISTA graph (Fig. 3). The LSC region was the most divergent, whereas the two IR regions were highly conserved. Most noncoding and intron regions were found to be more divergent and variable than the coding regions; however, several protein-coding regions of accD, rpoA, ycf1, and ndhF were relatively divergent. The overall nucleotide diversity (Pi) among eight plastomes showed an average Pi value of 0.00167 with 938 polymorphic sites, ranging from 0 to 0.01546, which was quite low, albeit similar to other genera of Rosaceae (i.e., Rosa at 0.00154) (Jeon and Kim 2019). The genetic polymorphisms in different regions of the chloroplast genome vary substantially, and wild Malus species harbored relatively higher nucleotide polymorphisms in both the LSC and SSC regions compared to those in the IR regions as observed similarly in Panax species (Jiang et al. 2018). They showed higher Pi values in the LSC (Pi = 0.002192) and SSC (Pi = 0.002394) regions, while obviously low values were found in the IR regions (Pi = 0.00045). Seven divergence hotspots among eight wild Malus plastomes are suggested as potential chloroplast markers: six intergenic regions (trnK-rps16, trnR-atpA, petN-psbM, trnT-psbD, psbZ-trnG, and ndhC-trnV) and one protein-coding region (ycf1) (Fig. 4).

Fig. 3
figure 3

Comparison of the chloroplast genomes of eight Malus species visualized by mVISTA. Gray arrows indicate genes with their orientation and position. Genome regions are color coded as pink blocks for the conserved coding genes (exon), blue blocks for introns, and peach blocks for noncoding sequences in intergenic regions (CNS). Thick lines below the alignment indicate the quadripartite regions of genomes; the LSC region is green, IR regions, aqua blue, and SSC region, orange

Full size image
Fig. 4
figure 4

Seven most variable hotspot regions found in eight plastomes of wild Malus species by sliding window analysis. Six intergenic regions of trnK-rps16, trnR-atpA, petN-psbM, trnT-psbD, psbZ-trnG, ndhC-trnV, and one coding gene of ycf1

Full size image

All of eight Malus cp genomes contained comparable numbers and distribution patterns of repeated sequences. Simple sequence repeats (SSRs) have high polymorphisms due to large variations in motifs and number of repetitions. Because of their high level of polymorphisms and genome-wide distribution, they have been used for powerful tools to measure genetic diversity and address the population genetic issues, such as gene flow, parentage, and population structure (Wang et al. 2009). In this study, we detected 103-114 SSRs by MISA based on search parameters set for 1-15 (mono-nucleotide motifs with 15 minimum numbers of repetition), 2-5, 3-3, 4-3, 5-3, and 6-3. The majority of the SSRs were tri-nucleotide motifs (61–68 SSRs, 60%) followed by di-nucleotide (17–20, 18%), and mono-nucleotide (14–18, 15%) (Fig. S3A). The most abundant repeat motif was “AAT/ATT” (22%) followed by “AAG/CTT” (21%) in all eight genomes (Fig. S3B, Table S3). SSRs were distributed most frequently in the intergenic regions (62%), followed by coding regions (31%), with much lower numbers found in the noncoding introns (7%) in each cp genome (Table S4). The coding regions with highest number of SSRs were ycf genes, eight in ycf1 (two pseudogenized, six coded in SSC and IR), and six SSRs (three duplicated in each IR) in ycf2. Considering the quadripartite regional occupancy of SSRs, the IR and SSC regions were lower in overall SSR frequency compared with the LSC region, 16% from the SSC region and 11% from each of both IR regions versus 62% from the LSC region (Table S4). Additionally, we found 49 pairs of large repeats in each cp genome (excluding duplicated IR region) using the parameters of maximum computed repeats = 50, minimum repeat size = 8 bp, and hamming distance = 1 by REPuter. They contained 23–31 forward, 2–12 reverse, and 13–20 palindromic matches of repeats (Fig. S4A). Lengths of 21–25 repeats were the most frequent (49%) followed by lengths of 26–30 repeats (21%), while longer repeats of 31–35 (8%), 36–40 (14%), and 41< (8%) were rarer than shorter ones (Fig. S4B).

Selective pressure in genes or genomic regions is inferred by the proportion of amino acid substitutions driven by natural selection during chloroplast genome evolution. Purifying selection removes deleterious variations, while positive selection fixes beneficial variation in the population and promotes the emergence of new phenotypes, offering fitness advantages in adaptation to the environment (Choudhuri 2014). Comparison of synonymous and nonsynonymous substitution rates can reveal the direction and strength of natural selection acting on the protein level. The rate of synonymous substitutions (dS), which is similar for many different genes, is significantly higher than that of nonsynonymous substitutions (dN), and the genes under positive selection are considered to have an evolutionary character in that dN is greater than dS (Endo et al. 1996). Therefore, the ratio of nonsynonymous substitution and synonymous substitution rates (denoted as ω=dN/dS) has been widely used as a genomic signature of selective pressure acting on a protein-coding gene, with ω=1 indicating neutral mutations; ω < 1, purifying selection; and ω > 1, diversifying positive selection (Yang et al. 2000). We identified that one of NADH dehydrogenase subunit genes of photosynthesis, ndhD gene, potentially evolved under positive selection in eight Malus plastomes by calculating the dN/dS ratio using various site-specific substitution models implemented in EasyCodeML (Gao et al. 2019; Yang 1997). Support for the gene under positive selection was identified, as codon substitution alternative model M8 (beta and ω > 1) provides a better fit than the null model M7 (beta in the interval 0 < ω < 1) from the pairwise comparison of likelihood ratio test (LRT) at significant level with p-value below 0.05 (Yang et al. 2000). Positively selected site in ndhD gene was suggested based on the posterior probability calculated by the Bayes empirical Bayes (BEB) method (Yang et al. 2005) with cutoff = 0.95 indicated with asterisk (*) in Table 4. The ndhD gene was included in the previously reported six genes (accD, rbcL, rps3, ndhB, ndhD, and ndhF) as undergoing positive selection in other Rosaceae plants (Yang et al. 2020). The critical importance of the genes that a plastome carries and its high conservativeness has contributed to the traditional view that purifying selection is the predominant force shaping chloroplast evolution due to functional limitations; however, the latest empirical evidence is no longer supportive of this hypothesis, and is now pointing to adaptive plastome variation (Bock et al. 2014). Recent advances in sequencing technology have contributed to an increase in the interest in taking advantage of the study of plastomes in phylogenetics, phylogeography, and population genetics. Variable genes potentially evolving under positive selection have occurred in the plastomes of a few other plant groups; three genes (rps2, rbcL, and ndhG) have been identified in Paulownia (Li et al. 2020a), five (rbcL, clpP, atpF, ycf1, and ycf2) in Panax (Jiang et al. 2018), three (clpP, ycf1, and ycf2) in the tribe Sileneae (Sloan et al. 2014), and in other angiosperms (Park et al. 2018; Piot et al. 2018; Wang et al. 2019). The genes identified as positively selected might undergo certain functional diversification in local adaptation during their evolutionary history that, in previous studies, has been discussed mainly as photosynthetic performance under variable environments of temperature and moisture (Bock et al. 2014).

Table 4 Positively selected sites having dN/dS values > 1 detected in eight Malus plastomes
Full size table

Phylogenetic analysis

Maximum likelihood (ML) analysis computed by IQ-TREE (Nguyen et al. 2015) enabled us to build robust phylogenetic relationships among wild Malus crabapple species based on the best-fit model, “TVM+F+I” (Fig. 5). The ML tree was reconstructed using the aligned sequences of 79 protein-coding genes from 23 representative Malus plastomes with Pyrus pyrifolia as outgroup. We selected plastomes of wild crabapple collections (rather than germplasm resources) deposited in the GenBank database, in addition to the four plastomes assembled in this study, to elucidate the phylogenetic relationships among wild Malus species. Previous studies revealed many unresolved branches and low bootstrap supports within Malus due to the partial usage of conservative chloroplast genomes (Forte et al. 2002; Harris et al. 2002; Robinson et al. 2001; Volk et al. 2015). Our current plastome phylogenomic analysis provided greater resolution with high bootstrap values in the relationships among major sections of the genus Malus. In the ML tree, Malus was divided into two clades; the first clade included the species of sections Chloromeles, Eriolobus, and M. florentina of section Sorbomalus, while the other clade comprised primarily the species of sections Docyniopsis, Malus, and Sorbomalus. M. florentina is a single species in the series Florentinae of section Sorbomalus, and has been placed under Sorbomalus owing to its morphological similarity of lobed leaves. It has previously been suggested to raise the taxonomic rank of M. florentina to section Florentinae, because, based on phytochemical and molecular studies, it showed greater similarity to the sections of Docyniopsis and Eriolobus than to other Sorbomalus species (Qian et al. 2008).

Fig. 5
figure 5

Maximum likelihood tree inferred from 79 protein-coding genes of 23 Malus (Rosaceae) taxa using Pyrus pyrifolia as outgroup. Bootstrap values over 50%, based on 1000 replicates, are shown on each node. The species indicated in red are the four Malus plastomes newly sequenced in this study. The subclade of red square contains the species belonging to the series Baccatae (section Malus) and series Sieboldianae (section Sorbomalus) except four species specified in blue

Full size image

The traditional classification of several sections and series of the genus Malus was not supported by chloroplast genome-wide phylogeny in this study. First, section Chloromeles was not monophyletic, as M. coronaria and M. ioensis were nested within clade I, whereas M. angustifolia was included in the subclade of Baccatae/Sieboldianae within clade II. This unexpected placement of M. angustifolia was also reported in previous study (Liu et al. 2019) owing to its close relationship to M. baccata. Section Docyniopsis was not monophyletic either, although the species of section Docyniopsis were nested in the same clade II. The occurrence of non-monophyly in the core Malus group (i.e., section Malus and Sorbomalus) was more complex, as both sections of Malus and Sorbomalus were polyphyletic, and their species were intermixed. Specifically, the species of series Baccatae (section Malus) and series Sieboldianae (section Sorbomalus) were closely related to each other within the strongly supported subclade Baccatae/Sieboldianae (100% bootstrap value), even though they are traditionally placed under different sections. Within the subclade Baccatae/Sieboldianae, the 1st group included the species of Baccatae of section Malus (M. halliana and M. hupehensis) and the species of section Sorbomalus (M. toringo of series Sieboldianae from China and M. toringoides of series Kansuenses). M. toringoides has been suggested to have a hybrid origin with paternal contribution most likely from M. transitoria (Feng et al. 2007; Tang et al. 2014), for which this study is quite supportive, based on the fact that M. toringoides was not genetically close to M. transitoria in maternally inherited chloroplast phylogeny, but was closely related to M. toringo. This study could not decisively determine its maternal parent, as the plastomes of the species from section Malus (M. sikkimensis of series Baccatae or M spectabilis of series Malus) that had been previously suggested as its maternal parent (Tang et al. 2014) were not available for comparison with M. toringo in our analyses. The 2nd group in subclade Baccatae/Sieboldianae primarily included M. baccata and M. toringo (collected from Korea and Japan) together with the species of series Malus of section Malus (M. micromalus and M. prunifolia), and M. angustifolia of section Chloromeles. There was a weakly supported interrelationship among the members of the 2nd group in subclade Baccatae/Sieboldianae.

The genetic similarity of Baccatae and Sieboldianae in addition to their morphological similarity has raised questions concerning their systematic positioning (Forte et al. 2002; Harris et al. 2002; Robinson et al. 2001; Savelyeva and Kudryavtsev 2015; Savelyeva et al. 2017). Even from the highly resolved chloroplast genome-based phylogeny in this study, they were not separated from each other, although they taxonomically belong to two different sections of Malus and Sorbomalus. However, the taxonomic requirements to merge the series of Baccatae and Sieboldianae remain a subject of future studies considering both the maternally inherited chloroplast and the biparently inherited nuclear DNA-based phylogenies. Despite its high resolution, plastome phylogeny represents only the maternal line in Rosaceae as reported previously (Brettin et al. 2000; Kaneko et al. 1986; Matsumoto et al. 1997; Raspé 2001); therefore, nuclear DNA-based tree topology should be compared with chloroplast phylogeny to yield the most plausible conclusion with an unequivocal explanation of the new classification based on the congruency between them. One novel finding of the current whole plastome phylogenomic study is that M. toringo exhibited the geographic pattern of its plastome diversity, exposing two distinct chlorotypes. The accession of M. toringo sampled from China (MK571562) was clustered with the Chinese accessions of M. sieboldii (synonym of M. toringo) and M. toringoides within the 100% supported clade, while the other chlorotype was displayed by the accessions of M. toringo collected from Japan and Korea. They were nested into another clade (90% bootstrap value) separate from the Chinese chlorotype and were the most closely related to M. baccata sampled from the sympatric region in Korea. The geographic location of M. sieboldii (NC050059), which was included in the same clade, is not known, as it was sampled from the Royal Botanic Gardens, Kew, in the UK without specified source information (Li et al. 2020b). Although we identified two types of chloroplast genomes in M. toringo, questions remain as to its monophyly and species delimitation, requiring further in-depth investigation. Given the maternal inheritance of plastid genomes and the high frequency of hybridization and gene flow in Malus, it is plausible that our current results are due to past gene flow among congeneric species in sympatric regions. To disentangle the complex evolutionary history of the crabapple genus and the unexpected findings in M. toringo, it would be necessary to carry out detailed population genetic or phylogeographic studies, or both.

Conclusion

In this study, we determined the complete plastome sequences of two wild crabapple species in the genus Malus (Rosaceae). As expected, we found highly conserved plastomes within the genus, including gene order and content, and a slow rate of evolution. The frequency of codon usage was biased toward high RSCU values of U and A at the third codon position, and we found that the highest numbers of potential editing sites were found in the ndhB gene followed by the ndhD gene with most editing sites at the 1st and 2nd codon positions. Comparative analysis among the eight wild Malus plastomes revealed seven divergence hotspots of six intergenic regions (trnK-rps16, trnR-atpA, petN-psbM, trnT-psbD, psbZ-trnG, and ndhC-trnV), and one protein-coding gene (ycf1) as potential chloroplast markers for phylogenetic studies of Malus species. We also identified that ndhD gene in eight Malus plastomes potentially evolved under positive selection. The results of phylogenetic analyses based on the aligned sequences of 79 protein-coding genes of 23 representative Malus plastomes provided high resolutions with strong bootstrap support in the relationships among major sections of the genus Malus. The genetic similarity between the series Baccatae and Sieboldianae from two different sections (Malus and Sorbomalus, respectively) was confirmed in this study. Interestingly, M. toringo exhibited the geographic pattern of its plastome diversity, revealing two distinct chlorotypes distributed in China and Japan/Korea.