2019 Volume 88 Issue 2 Pages 222-231
Japanese apricot (Prunus mume Sieb. et Zucc.) is one of the major fruit tree crops in Japan. However, a paucity of molecular tools has limited studies on the species’ genetic diversity and clone identification. Therefore, we newly designed 201 microsatellite markers using the P. mume reference genome and selected 20 highly polymorphic markers. The markers showed higher polymorphism detectability than those previously developed using peach and apricot genomes. They were used successfully for fingerprinting most of the Prunus cultivars examined (124 P. mume accessions and one accession each of P. armeniaca, P. salicina, P. persica, and P. dulcis), and the resulting genotype data were used to examine the genetic differentiation of six Japanese apricot cultivar groups, including those producing normal fruit, small-fruit, and ornamental flowers, as well as Taiwanese cultivars, putative hybrids of P. armeniaca and P. mume, and putative hybrids of P. salicina and P. mume. Phylogenetic cluster analysis showed three clades with high support values; one clade comprised the putative P. armeniaca × P. mume hybrids, and the two others included Taiwanese and ornamental cultivars. The rest of the accessions were grouped into two wide clusters, but not clearly divided into the respective cultivar groups. These complex relationships were supported by the principal coordinate and STRUCTURE analyses. Since Japanese apricot is thought to have originated in China, many factors such as human preference, geographical separation, introgression, and local breeding, may have been involved to form the present complex genetic structure in Japanese apricot.
Japanese apricot (Prunus mume Sieb. et Zucc.) is a stone fruit species that is found in the same genus as peach [P. persica (L.) Batsch], almond [P. dulcis (Mill.) D. A. Webb.], apricot (P. armeniaca L.), Japanese plum (P. salicina Lindl.), European plum (P. domestica L.), and sweet cherry (P. avium L.). Japanese apricot is thought to have originated in China and has been cultivated for over 3000 years with over 500 cultivars (Horiuchi et al., 1996). It is believed that the species was introduced to Japan ~2000 years ago (Horiuchi et al., 1996; Mega et al., 1988), and the annual production of the fruit in Japan recently reached 111,400 tones (MAFF, 2016). The fruits of the Japanese apricot are generally processed into pickles (i.e., “Umeboshi”), fruit liqueurs (i.e., “Ume-shu”), syrups, jams, and other processed foods.
Conventionally, Japanese apricot is divided into two major groups: fruit and flower ornamental cultivars (Horiuchi et al., 1996; Mega et al., 1988). The two groups differ in a variety of morphological traits, tree architecture, petal number, flower color, and fruit size, but clear genetic differences have yet to be reported. In addition, there are some minor groups with characteristic traits. For example, cultivars in the small-fruit group literally bear very small-sized fruits (approximately less than 10 g), and mostly exhibit self-compatibility (Yaegaki et al., 2003). Taiwanese cultivars generally show a very weak chilling requirement for endodormancy release, which is a suitable survival strategy in subtropical climates (Kitamura et al., 2018). Interspecific hybrid groups derived from the crosses with apricot or Japanese plum also have very specific morphological and physiological characteristics (Mehlenbacher et al., 1991; Yaegaki et al., 2012; Yoshida and Yamanishi, 1988).
In general, the direction of domestication is influenced by human preference (Zeder et al., 2006). In peach, for example, a distinct genetic differentiation was detected between fruit and ornamental cultivars based on genome-wide SNPs analysis (Akagi et al., 2016). However, there are not enough molecular markers for genetic analysis in Japanese apricot. Previously, microsatellite markers were designed using the peach and cherry nucleotide sequences (Gao et al., 2004), but most of them do not give informative band patterns. Hayashi et al. (2008) examined genetic diversity in Japanese apricot using microsatellite markers based on the peach and apricot genomic sequences, but the marker numbers and their genomic information are limited. Recently, Zhang et al. (2012) reported the genome sequences of eight linkage groups from a Chinese P. mume accession.
Here, we designed 201 new microsatellite markers using the P. mume reference genome and selected 20 highly polymorphic markers that were more capable of detecting polymorphisms than markers previously developed using peach and apricot genomes. The markers were used to fingerprint 124 P. mume accessions (mainly Japanese cultivars), as well as four other Prunus species (P. armeniaca, P. salicina, P. persica, and P. dulcis). The resulting genotype data were used to evaluate the genetic differentiation of Japanese apricot cultivars.
The plant materials used in this study are listed in Tables 1 and S1. A total of 124 P. mume accessions were used, including 46 fruit (F) cultivars, 10 small-fruit (FS) cultivars, 49 ornamental (O) cultivars, five Taiwanese (T) cultivars, 10 putative P. armeniaca × P. mume (AM) hybrids, and four putative P. salicina × P. mume (SM) hybrids. In addition, one accession each of apricot (Pa), Japanese plum (Ps), peach (Pp), and almond (Pd) were used as outgroup species. Of these, 61 accessions had multiple entries of different trees, and 79 accessions were the same as previously used by Hayashi et al. (2008) (Table S1). All the plant materials were maintained in the experimental orchards of the Japanese Apricot Laboratory, Wakayama Fruit Tree Experiment Station, Minabe, Wakayama, Japan. Genomic DNA was extracted from leaves using a DNeasy plant mini kit (Qiagen, Hilden, Germany).
Number of the Prunus accessions and trees used in this study.
Microsatellite primers were designed using the P. mume reference genome (Zhang et al., 2012). The nucleotide sequences of the eight linkage groups (NC_024126.1 for LG1, NC_024127.1 for LG2, NC_024128.1 for LG3, NC_024129.1 for LG4, NC_024130.1 for LG5, NC_024131.1 for LG6, NC_024132.1 for LG7, and NC_024133.1 for LG8) were downloaded from the National Center for Biotechnology Information website (www.ncbi.nlm.nih.gov/genome/13911), and surveyed for microsatellite regions using a Tandem repeats finder (Benson, 1999). Microsatellite regions with ~20 GA/CT or AG/TC repeats were selected at each 1 Mb interval (a total of 201 regions). Primer pairs were then designed using Primer 3 (Untergasser et al., 2012) with the following parameters: 17–22 nucleotides long, Tm approximately 60°C, and product size in the range of 100–250 bp.
Microsatellite marker screeningThe 201 microsatellite markers (designated as JAM, Table S2) were subjected to the following three screenings for the marker availability: 1) amplification ability with a standard cultivar of ‘Nanko’, 2) a wide range amplification check using eight Prunus accessions (three F cultivars and one each of FS, O, T, AM, and Pa), 3) polymorphism examination with 16 P. mume cultivars (10 F, three FS, and three O cultivars). For all screenings, PCR was performed in a 20 μL volume using Ex Taq (Takara Bio, Shiga, Japan) according to the manufacturer’s instructions. The PCR conditions were as follows: an initial denaturation of 5 min at 94°C, followed by 35 cycles of 30 s at 94°C, 30 s at 55°C, 1 min at 72°C, and 7 min at 72°C for final extension. PCR products were separated by 4% polyacrylamide gel electrophoresis and visualized by a silver staining method (Panaud et al., 1996).
Microsatellite marker analysisUsing a selected set of 20 highly polymorphic microsatellite markers, a total of 128 Prunus accessions (199 tree entries) were genotyped. In addition, 11 microsatellite markers previously reported by Hayashi et al. (2008) were also employed. To precisely determine the lengths of amplified fragments, PCR was performed using the post-labeling method described by Schuelke (2000) with minor modifications. To generate fluorescent PCR products, the U-19 universal primer (5'-GTTTTCCCAGTCACGACGT-3') was labeled with four kinds of fluorescent molecules (6-FAM, VIC, NED, and PET), as well as with 2-bp barcodes at the 3' ends (TG for 6-FAM-labeled U-19, AC for VIC-labeled U-19, CA for NED-labeled U-19, and GT for PET-labeled U-19). Non-labeled forward primers, with 7-bp 5' pig-tails (5'-GTTTCTT-3'), and U-19-fused reverse primers were also synthesized and used to perform multiplex PCR. The 20 μL PCR mixtures were prepared using a Type-it Microsatellite PCR kit (Qiagen), with 50–100 ng template DNA, 0.2 μM each labeled U-19 primer, 0.2 μM each pig-tailed forward primer, and 0.04 μM each U-19-fused reverse primer. The PCR conditions were as follows: initial denaturation of 5 min at 95°C; followed by 10 cycles of 30 s at 95°C, 90 s at 60°C with a decrease of 0.5°C in each cycle, and 30 s at 72°C; followed by 22 cycles of 30 s at 95°C, 90 s at 55°C, and 30 s at 72°C, and a final extension step of 30 min at 60°C. Finally, the fragments were analyzed using an ABI 3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) using LIZ-600 size standard (Applied Biosystems). Alleles were defined on the basis of fragment size (in nucleotides) and scored using Genemapper 4.1 (Applied Biosystems).
Genetic diversity and population structure analysisStatistical analyses were performed after excluding two triploid cultivars (‘Horyukaku’ and ‘Takasago’) and the four accessions of other Prunus species (P. armeniaca ‘Heiwa’ was employed as an outgroup only for phylogenetic tree construction). Different genotypes found in the same accessions were independently examined in the analyses. For accessions that had identical genotypes at all loci, only one representative accession was used in the analyses, except for phylogenetic tree construction. Marker scores with null or more than three alleles were treated as missing data. However, for only the phylogenetic analysis, the score sets of loci with missing data were excluded. The number of alleles, observed heterozygosity (Ho), expected heterozygosity (He), number of effective alleles (Ne), fixation index (FIS), Nei’s genetic distance (Nei et al., 1983), and G′st (Hedrick, 2005) were calculated using GenAlEx v. 6.502 (Peakall and Smouse, 2012). Here, G′st is defined as a standardized genetic differentiation measure of Gst (Nei, 1972) for highly polymorphic markers (Hedrick, 2005). Polymorphism information content (PIC) was calculated using POLYSAT (Clark and Jasieniuk, 2011) in R (R Development Core Team, 2008). Power of discrimination (PD) (Kloosterman et al., 1993) was computed manually using the following formula: PD = 1 − Σgi2, where gi is the frequency of the ith genotype. Meanwhile, within-group genetic variation was assessed using the gene diversity value (D) (Nei, 1987). Statistical significance for FIS and G′st was tested using GENEPOP 4.2 (Raymond and Rousset, 1995; Rousset, 2008) and GenAlEx v. 6.502 (Peakall and Smouse, 2012), respectively. The levels of linkage disequilibrium (LD) were calculated as the squared allele-frequency correlation (r2) values using MIDAS (Gaunt et al., 2006). The r2 values were calculated between the most frequent alleles at each two markers, and among all combinations of alleles at each two markers and summed up with weight of allele frequencies, according to Iwata et al. (2013).
The population structure of the P. mume accessions was analyzed using three methods, namely principal coordinate analysis using GenALEx v. 6.502 (Peakall and Smouse, 2012), neighbor-joining (NJ) tree construction using Poptree 2 with 1000 bootstrap replicates (Takezaki et al., 2010), and individual-based Bayesian clustering using STRUCTURE 2.3.4 with Markov chain Monte Carlo simulations (Pritchard et al., 2000). The NJ tree was visualized using MEGA6 (Tamura et al., 2013). In the STRUCTURE analysis, the AM and SM hybrids were excluded in order to simplify the simulations. In addition, the number of genetic clusters (K) was set from 1–15, and each analysis was allowed to run for 200,000 simulations after a burn-in period of 200,000 iterations. Ten runs were performed for each K value, and the optimal value was determined using L(K), |L"(K)|, and ΔK (Evanno et al., 2005), which were calculated using Structure Harvester (Earl and vonHoldt, 2012). A representative bar plot for each K was selected based on LnP(D) values generated by STRUCTURE software, and visualized using Structure Plot v2.0 (Ramasamy et al., 2014).
Among the 201 microsatellite markers designed, 188 (93.5%) were successfully amplified using the template DNA of ‘Nanko’ (Table S2). Of these, 128 markers were randomly selected for second amplification screening. As a result, 59 markers gave clear bands for all eight Prunus accessions. They were preliminarily checked for the level of polymorphism using 16 P. mume cultivars. Their PIC values were compared and a total of 20 highly polymorphic markers were selected: at least two markers were selected from those giving the highest PIC values in each linkage group. They were renamed as PMKS markers (Table S2).
Allele combinations at 20 polymorphic microsatellite loci among Prunus accessionsIn order to examine the allelic diversity, a total of 128 Prunus accessions (199 tree entries) were genotyped with a set of 20 highly polymorphic markers (Table S3). Identical genotypes for all the loci were observed among the trees in 56 out of 61 accessions having multiple entries, whereas different allele combinations were observed in the other five accessions: ‘Fudono’ (F4), ‘Naniwa’ (F24), ‘PM1-1’ (SM1), ‘Tsuyuakane’ (SM4), and ‘Heiwa’ (Pa) (Table S3).
Among the 124 P. mume accessions, a total of 107 allele combinations were observed, with 92 of the combinations being unique to single accessions and 15 being shared by multiple accessions. Thirteen of the redundant genotypes were shared by two accessions each, and genotypes of ‘B’ and ‘J’ were observed for seven and three accessions, respectively (Table S3). In addition to the triploid cultivars (‘Horyukaku’ and ‘Takasago’), several other accessions also gave more than three alleles at certain loci.
Efficacy of microsatellite markers in P. mumeThe efficacy (i.e., polymorphism detection ability) of the 20 microsatellite markers was assessed with the allele combinations found in P. mume (Table 2). The PIC values ranged from 0.63 to 0.90 (mean = 0.79) and the PD values ranged from 0.84 to 0.97 (mean = 0.93). Using the same P. mume accessions, 11 microsatellite markers previously reported by Hayashi et al. (2008) were also evaluated (Tables 2 and S2). The average values of all the polymorphism indices (number of alleles, Ho, He, PIC, and PD) for the newly developed marker set were greater than those for previously reported, thereby reflecting the powerful detection ability of the new marker set. For example, ‘Hakuo’ (FS2), ‘Koshu Saisho’ (FS4), and ‘Purple Queen’ (FS8) shared a single genotype when using the previously developed set, whereas the newly developed markers were able to distinguish ‘Hakuo’ from the other two. Similarly, the new marker set was also able to distinguish ‘Ikuyonezame’ (O14) and ‘Kinko’ (O19). The FIS values ranged from 0.01 to 0.46 (mean = 0.13) and 16 of 20 loci showed statistical significance at the 5% level (Table 2). The levels of LD were low among all combinations of 20 loci (Table S4).
Polymorphism indices calculated for the present (A) and the previous marker sets (B).
Genetic diversity indices were calculated to evaluate variation within the six cultivar groups (F, FS, O, T, AM, and SM; Table 3), using the same set of allele combinations without duplicates. Although the Ho and D values of the six P. mume groups were similar, a relatively high Ne values were observed for the F and O groups (Table 3).
Average genetic diversity indices for six cultivar groups in P. mume.
Genetic differentiation among the P. mume cultivar groups was first evaluated using principal coordinate analysis with individual genotype data. The first principal coordinate axis (PC1) seemed to separate the F and O cultivars (Fig. 1), whereas the second axis (PC2) distinguished the T cultivars from the others, and also seemed to explain within-group variation. However, even though the analysis revealed genetic differentiation among the P. mume accessions, the percentages of variation explained by PC1 and PC2 were relatively low (9.0% and 6.7%, respectively).
Principal coordinate analysis of the six Japanese apricot cultivar groups. Values in parentheses indicate the proportion of the total variance explained by each principal coordinate. The abbreviations AM, F, FS, O, SM, and T indicate P. armeniaca × P. mume hybrids, fruit, small-fruit, ornamental cultivars, P. salisina × P. mume hybrids, and Taiwanese cultivars, respectively.
We next calculated pairwise Nei’s genetic distance and G′st values for the six cultivar groups (Table 4). Significant (P < 0.01) G′st values were obtained for all pairs. Among the six groups, the T cultivars showed relatively high genetic distance values (1.053–2.042), followed by the two putative hybrid groups (SM and AM). However, in contrast to the results of the principal coordinate analysis, the genetic distance between the F and O cultivars was quite low (0.238). A similar low value (0.275) was also observed between the F and FS groups, although they have major phenotypic differences in terms of fruit size and harvesting time.
Pairwise genetic differentiation measure, G′st (below diagonal) and Nei’s genetic distance value (above diagonal) between six cultivar groups.
Phylogenetic analysis was performed with all the P. mume accessions with no missing data and P. armeniaca ‘Heiwa’ as an outgroup (Fig. 2). Three of the clades were strongly supported (bootstrap value > 70). One, which was clearly separated from the others, included the AM hybrids and the P. armeniaca ‘Heiwa’, whereas the two others contained T (T clade) and O cultivars, including ‘China mume’ (O-1 clade). The rest of the accessions belonged to two wider clusters that mainly included F and O cultivars (Fig. 2).
Neighbor joining tree of P. mume accessions based on the allele frequency of 20 polymorphic microsatellite markers. P. armeniaca ‘Heiwa’ was used as an outgroup. Solid lines indicate clades (AM, putative P. armeniaca × P. mume hybrids; T, Taiwanese cultivars; and O-1, ornamental cultivars). Dotted lines indicate groups that mainly consist of fruit cultivars (F) and ornamental cultivars (O).
STRUCTURE analysis was performed using 94 genotypes from the four P. mume cultivar groups (F, FS, O, and T). The highest ΔK value was obtained at K = 2, followed by K = 7 (Fig. 3A). In the bar plot for K = 2, all T cultivars showed a single orange-colored cluster (Fig. 3B), whereas the other three groups gave a mixture of two clusters. Yellow and orange clusters seemed to be dominant in the F and O cultivar groups, respectively. On the other hand, for K = 7, T cultivars could be clearly separated, and most FS cultivars showed a single red cluster. A purple cluster seemed to be dominant in the F group. O cultivars consisted of many admixture types, but cultivars belonging to the O-1 clade in the phylogenetic tree shared the same navy-colored cluster.
STRUCTURE analysis of P. mume cultivars. Fruit (F), small-fruit (FS), flower ornamental (O), and Taiwanese (T) cultivars were employed for the simulations. (A) Values of mean L(K), |L"(K)|, and ΔK. The highest value was obtained at K = 2, implying the existence of two clusters. The ΔK value was also high at K = 7, indicating the existence of 7 clusters. (B) Bar plots for each value of K, at K = 2 and 7. Representative bar plots were selected based on LnP(D) values generated by STRUCTURE software. Cultivars belonging to the O-1 clade (Fig. 1) are indicated by asterisks.
Studies on genetic diversity and clone identification in P. mume have been relatively limited owing to a lack of effective molecular tools. Therefore, we designed 201 new microsatellite markers using the P. mume reference genome reported by Zhang et al. (2012), and selected 20 highly polymorphic markers for further analysis. As shown in Tables 2 and S2, the present marker set exhibited greater polymorphism detection ability than those previously developed using the peach and apricot genomes (Hayashi et al., 2008) and was successfully used to fingerprint most of the Prunus cultivars (128 accessions, 199 trees). However, 15 redundant genotypes (designated as A to O in Table S3) were shared by two or more accessions. Hayashi et al. (2008) also reported that identical genotypes were detected among cultivars having very similar morphological characteristics, namely, among the three cultivars of ‘Shirokaga’ (F36), ‘Gojiro’ (F7), and ‘Gyokuei’ (F8), two of ‘Kodama’ (F20) and ‘Kinyuji’ (F19), and two of ‘Hanakami’ (F10) and ‘Rinshu (Nara)’ (F32). Since these accessions may be recent somaclonal variants, resulting from bud sports with slight mutations, it is quite difficult to distinguish them using microsatellite markers. Therefore, next generation sequencing will likely be needed to detect cultivar-specific SNPs.
On the other hand, different allele combinations were observed among tree entries in the five accessions: ‘Fudono’ (F4), ‘Naniwa’ (F24), ‘PM1-1’ (SM1), ‘Tsuyakane’ (SM4), and ‘Heiwa’ (Pa) (Table S3). Two different genotypes between trees were observed in ‘Fudono’ and ‘Naniwa’. Since they are old local varieties, some chance seedlings showing similar phenotypes may have been cultivated as the same varieties. For the modern varieties ‘PM1-1’ and ‘Tsuyuakane’, length mutations were observed between two tree entries only at the PMKS113 locus, suggesting that they were generated by recent somaclonal variation. Two ‘Heiwa’ trees were identical at 15 of the 20 loci. ‘Heiwa’ is quite an old variety. Probably, the difference was also caused by somaclonal variation.
In recent years, Japanese fruit crops (strawberry, sweet cherry, and grapevine) have been taken overseas and cultivated. To control such dissemination, domestic fruit cultivars should be protected by international patent or registration with fingerprinting data. The highly polymorphic marker set developed here may be suitable for identifying cultivars of Japanese apricot, and moreover, 201 markers designed to cover whole genome linkage groups in P. mume (Table S2) can be useful for linkage and QTL analyses to improve Japanese apricot.
Core collections have been reported for many crops, such as rice (Ebana et al., 2008), soybean (Kaga et al., 2012), and strawberry (Wada et al., 2017). They allow us to maintain a minimum number of genetic resources with maximum genetic variation. The present marker set can be utilized to develop a core collection of Japanese apricot to enhance efficient breeding.
Population structure in P. mume accessionsAmong the P. mume accessions, the average observed heterozygosity of 20 PMKS markers was 0.70, which was lower than that of expected heterozygosity (0.81, Table 2), and most of the FIS values showed significance at the 5% level. These results are due to the fact that the present group of Japanese accessions is not a natural population with random mating. Probably, Japanese accessions went through strong artificial selection, introgression and specific breeding. The levels of LD were low (Table S4), suggesting no association of alleles at 20 loci.
Generally, DNA markers with non-significant FIS values are suitable for population structure analyses. In this study, we preliminarily used all the 20 PMKS markers to clarify the present genetic differentiation of P. mume accessions through three analyses (principal coordinate analysis, phylogenetic cluster analysis, and STRUCTURE analysis).
Among six cultivar groups, putative Taiwanese (T) cultivars were clearly distinguished from others in both phylogenetic and STRUCTURE analyses (Figs. 2 and 3B). The results are consistent with those of Hayashi et al. (2008) using SSRs derived from peach and apricot, and Shimada et al. (1994) using RAPD markers. The T cultivars are uniquely characterized by their weak bud dormancy, which enables them to adapt in subtropical climates (Kitamura et al., 2018). The AM cultivars except ‘Taihei’ (AM10) were clearly distinguished in the phylogenetic analysis (Fig. 2). They were derived from interspecific hybrids between P. mume and P. armeniaca (Mehlenbacher et al., 1991; Yoshida and Yamanashi, 1988). Although ‘Taihei’ (AM10) showed evidence for inheritance of P. armeniaca characteristics (Hayashi, 2009; Hayashi et al., 2008), it was included in the huge cluster of Japanese apricot. Probably, ‘Taihei’ went through a subsequent cross with Japanese apricot cultivars. Other F, FS, and O cultivars were not clearly divided into the respective cultivar groups. They formed a complex genetic structure in the Japanese population.
In this study, the genetic differentiation among Japanese apricot was examined by three analyses. However, we do not know the exact sequence of the differentiation process because Japanese apricot is thought to have originated in China and many cultivars may have been occasionally introduced to Japan. Many factors, such as human preference, geographical separation, introgression, and local breeding, may be involved to form the complex genetic structure in Japanese apricot. More comprehensive analytical methods (e.g., genome-wide SNP survey with wider genetic resources including Chinese cultivars and wild relatives) will shed light on the details of this species’ evolution, domestication, and improvement history.
We are deeply grateful to the researchers at the Japanese Apricot Laboratory, Wakayama Fruit Tree Experiment Station, for maintaining the Japanese apricot cultivars.
