Introduction

The genus Pyrus in Pyrinae is widely distributed in Asia, Europe, and North Africa and has a long history of utilization of its fruits as food. The classification of Pyrus is often very difficult due to natural or artificial interspecific hybrids, which arise easily because of self-incompatibility and the lack of distinguishable characters between species. Pyrus can generate fertile progeny with ease, and repetitious cross hybridization generates numerous progeny harbouring different heterogeneity. Some of these could be given the rank of form, variety, or species (Rubtsov 1944; Browicz 1993). Rehder (1940) described 15 principal species, 6 varieties, and 5 related species in Pyrus. Rubstov (1944) considered that 14 species found in Central Asia should be added, and he proposed that the genus Pyrus consisted of c.a. 35 species in total. However, some of these are suspected of being hybrid so that Challice and Westwood (1973) tried to reconsider speciation in Pyrus and recognized about 22 species according to numerical taxonomy using morphological characters and phenolic compounds. Bell (1990) recognized 22 primary species of pear, whose distribution covers most of Europe, temperate Asia, and mountain areas of northern Africa. In addition, he recognized at least nine natural or artificial interspecific hybrids between primary species; those were classified into different species.

Rubstov (1944) reported that European pear cultivars of Pyrus communis exhibit the characteristics of at least four species: Pyrus elaegrifolia Pall., Pyrus salicifolia Pall., Pyrus korshinskyi Litv., and Pyrus syriaca Boiss. Wild populations of P. communis var. pyraster and/or Pyrus caucasica Fed. are possible ancestors of cultivated P. communis, and there is also some evidence indicating an involvement of Pyrus nivalis Jacq. (Challice and Westwood 1973). However, it is still difficult to determine the origin of European pear cultivars phenotypically or even by DNA markers (Aldasoro et al. 1996; Volk et al. 2006).

Nakai (1919) described that four species in Pyrus native to Japan, i.e., Pyrus pyrifolia [Burm.] Nakai (Japanese pear), Pyrus dimorphophylla Makino (pea pear), Pyrus hondoensis [Nakai et Kikuchi] Rehder, and Pyrus aromatica [Nakai et Kikuchi] Rehder. P. hondoensis and P. aromatica are included in Pyrus ussuriensis Maxim. and treated as its varieties (P. ussuriensis var. hondoensis (Nakai et Kikuchi) Rehder and P. ussuriensis var. aromatica (Nakai et Kikuchi) Ohwi) by Rehder (1920) and Ohwi (1965), respectively. Although almost all modern Japanese pear cultivars are included in P. pyrifolia, the origin of P. pyrifolia is still unclear. Generally, Japanese pear cultivars were considered to have been domesticated from native P. pyrifolia occurring in Japan (Kikuchi 1948). However, candidate trees of native P. pyrifolia were only found near human habitation and therefore presumed to be escapes. The progenitor of Japanese pear cultivars may have come prehistorically from China (Shirai 1929; Kajiura 1983). Recently P. ussuriensis var. aromatica has been reported as native and prehistorically introduced into Northern Tohoku region in Japan by means of taxonomical and molecular analyses (Iketani et al. 2010).

In Chinese pear, 15 pear species have been identified (Gu and Sponberg 2003). Pear cultivars native to China consist of the following four groups: Chinese sand pear (P. pyrifolia Nakai), Ussurian pear (P. ussuriensis Maxim.), Chinese white pear (Pyrus bretschneideri), and Xinjiang pear (Pyrus sinkiangensis Yu) (Yu 1979). Kikuchi (1946) proposed that P. bretschneideri might be generated by hybridization between P. ussuriensis and P. pyrifolia according to their geographical distribution. A recent study using RAPD and SSR markers could not distinguish P. bretschneideri and P. pyrifolia (Teng et al. 2002; Bao et al. 2007). Teng et al. (2002) suggested that, from molecular data, P. pyrifolia might be a common progenitor of P. bretschneideri and Japanese pear cultivars.

Although it is well known that the structure and gene content of the chloroplast genome are conserved among divergent plant species, structural alterations such as insertions, deletions (gaps), inversions, and translocations have been found in certain plant lineages by comparing the genome structure of chloroplast DNA (cpDNA). Therefore, using mutational events in chloroplast genomes as DNA markers enables us to reconstruct plant phylogeny at higher taxonomic levels (Downie and Palmer 1992; Doyle et al. 1992; Katayama and Ogihara 1996).

Three hypervariable regions, which may represent intra-molecular recombination hotspots, have been detected in the cpDNA of Poaceae (Ogihara et al. 1988; Hiratsuka et al. 1989; Morton and Clegg 1993; Maier et al. 1995; Katayama and Ogihara 1996). Tracing of indels in one hypervariable region led to an understanding of the phylogenetic relationships at interspecific level and at lower levels (Guo and Terachi 2005).

Although the family Rosaceae contains many commercially important fruit trees and ornamental trees, the chloroplast (cp) genome structure in rosaceous plants is still not well known. Recently, the cp genome structure was characterized at the physical map level for Pyrus and Prunus in Spiraeoideae (Katayama and Uematsu 2003, 2005). These reports indicated the extremely low cpDNA diversity in Pyrus and the existence of a hypervariable region in Prunus cpDNAs. At the DNA sequence level, Shaw et al. (2007) reported 21 variable noncoding chloroplast regions among angiosperms including Prunus.

Iketani et al. (1998) highlighted inconsistencies between the cpDNA types and the morphological classification when analysing relationships between Occidental and Oriental pears by RFLP in cpDNAs. They supposed hybridization or lineage sorting and proposed that a large number of materials including pure natural population would be necessary to solve this issue. Using comparative sequence analysis in six cpDNA regions such as atpB-rbcL intergenic spacer, trnL-Fspacer, accD-psaI spacer, ndhA intron, rpl16 intron, and rpoC1 intron in cultivated pear varieties from Asia, Kimura et al. (2003) noted that the phylogenetic relationships among Asian pears was quite mingled and incongruent, because of lack of polymorphisms from the result of high conservation in the cp genome in Pyrus.

We tried to develop the hypervariable regions harbouring structural alterations as a cpDNA marker to trace the evolution of the cp genome and to understand the phylogenetic position at the inter and/or intraspecific level in Pyrus. In this report, the phylogenetic utility of hypervariable regions in the extremely conserved chloroplast genome is evaluated and effectiveness of haplotype network analysis in Occidental and Oriental pear cpDNAs, using both gaps and base changes found in two hypervariable regions, is determined.

Materials and methods

Eighty-one accessions from 21 species in Pyrus originating from Asia, Europe, and Africa were selected for this study (Table 1). Total DNA of European, West Asian, and North African pears was obtained from the DNA bank in the Royal Botanic Gardens, Kew, UK. Twenty-one accessions of Chinese pear varieties are maintained at the Nanjing Agricultural University, China. Two accessions of P. ussuriensis Maxim. var. hondoensis (Nakai & Kikuchi) Rehder endemic to Japan are maintained at the National Institute of Fruit Tree Science. Other Japanese pear cultivars and wild pears in Japan were selected from the Pyrus germplasm collection at the Food Resources Education and Research Center, Kobe University, or selected from the collection of the Botanical Gardens, Osaka City University. Classification of these Pyrus followed Rehder (1940), Yu (1979), and Ohwi (1965).

Table 1 Plant materials used in this study and cpDNA haplotypes determined by two large gaps
Full size table

Total DNA was isolated from young leaves of pear according to the method reported by Hosaka (1995). Oligonucleotide primers designed by Small et al. (1998) were used to detect a hypervariable region in the intergenic region between accD-psaI genes in Pyrus cpDNA. Another hypervariable region between rps16-trnQ was amplified using primers rps16-F and trnQ-R (Table 2). PCR amplification of intergenic region between rbcL-accD was performed by primers rbcL-F and accD-R (Table 2). Amplified fragments were fractionated by agarose gel electrophoresis. Primers used for sequencing of the two hypervariable regions were shown in Table 2. Positions of primers corresponded to nucleotide numbers of tobacco and pear cpDNA complete sequences (accession numbers Z00044 and AP012207 in GenBank/EMBL/DDBJ).

Table 2 PCR and sequencing primers used in this study
Full size table

DNA sequences of amplified fragments were determined using ABI3100 Genetic Analyser (Applied Biosystems, USA). The nucleotide sequences reported here have been assigned GenBank/EMBL/DDBJ accession numbers AB603891 to AB603936 for rbcL-accD, AB604678 to AB604722 for rps16-trnQ, and AB604874 to AB604916 for accD-psaI. The sequence of ‘Chojuro’ for accD-psaI was previously assigned to GenBank accession no. AB204878 (Kimura et al. 2003).

DNA sequences were aligned using multiple alignment, GENETYX-MAC ver.12 (GENETYX CO. Japan) with minor manual adjustments. Mutation rates among haplotypes in different regions were calculated as the average number of nucleotide substitutions and length mutations per site according to Jukes and Cantor (1969) using the PAUP version 4.0b10 (Swofford 1998). A median-joining network (MJ) including potential median vectors was performed with Network. 4.5.1.6 (Bandelt et al. 1999) using DNA nucleotide data based on the gaps (indels) and base changes found in two hypervariable regions. Gaps and base changes were scored as independent single character.

Results

Two hypervariable regions in the conserved pear cp genome

A total of 40 mutations including 30 gaps (indels) and 9 base changes were found and characterized into three intergenic regions (accD-psaI, rps16-trnQ, and rbcL-accD) for 45 accessions. The length of an intergenic region between rbcL-accD was determined for 45 accessions from 21 species in Pyrus by sequencing. A base change was found at an intergenic region between rbcL-accD among 45 accessions. The base change and length mutation rates revealed with the Jukes and Cantor method among haplotypes in this region were quite low (ranging from 0.0009 for 3 accessions of Pyrus pashia, P. korshinskyi, and P. ussuriensis var. hondoensis to 0.001 for 21accessions with an average of 0.001, and from 0.0006 for 25 accessions to 0.0024 for an accession of Pyrus gharbiana with an average of 0.0009) for the 21 species in Pyrus (Table 3). The base change rate was almost equivalent to the length mutation rate in an intergenic region between rbcL-accD. The length of an intergenic region between accD-psaI was determined for 21 species in Pyrus by sequencing and found to vary from 585 bp for P. korshinskyi and P. bretschneideri ‘Pingli’ to 857 bp for P. pyraster. Sixteen haplotypes were determined by 15 gaps and 6 base changes found in this region. Although the base change rate among 45 accessions in this region was quite low (ranging from 0.0003 for 34 accessions to 0.0021 for an accession of Pyrus cossonii with an average of 0.0006) for the 21 species in Pyrus, the length mutation rate ranged from 0.0037 for seven accessions to 0.013 for an accession of P. pyraster with an average of 0.0056 and was approximately 9.3 times higher than the base change rate (Table 3). The length of the gaps varied from 1 to 228 bp (Table 4). Ten motifs of short direct repeats varied from 6 to 919 bp in length, and one inverted repeat motif of 15 bp in length was found in association with length variations in each repeat (Table 6).

Table 3 Mutation rates of three intergenic regions among cpDNA haplotypes in Pyrus
Full size table
Table 4 Characterization of 16 haplotypes by insertions (+)/deletions (−) and base changes found in the 857-bp intergenic region between accD and psaI in Pyrus species by DNA sequencing
Full size table

The length of the intergenic region between rps16-trnQ genes varied from 662 bp for P. cossonii to 904 bp for P. korshinskyi. Fourteen haplotypes were determined by 12 gaps and 3 base changes found in this region (Table 5). The base change rates were also quite low, ranging from 0.00008 for 42 accessions to 0.0016 for 2 accessions of P. gharbiana and P. ussuriensis var. hondoensis with an average of 0.0001 for the 21 species in Pyrus (Table 3). The length mutation rate ranged from 0.0018 for eight accessions to 0.0052 for an accession of P. cossonii with an average of 0.0028 and was approximately 28 times higher than the base change rate. The length of the 12 gaps varied from 1 to 141 bp (Table 5). This region also contained six direct repeats ranging from 16 to 42 bp in each gap (Table 6; Fig. 1). Two chloroplast simple sequence repeat loci (repetition of the single nucleotide ‘A’ between 9 and 21 times) were found adjacent to the borders of the 141-bp large gap in the region between the rps16-trnQ genes (Table 7). These two SSR loci were removed from the data set for phylogenetic analysis, because the length of the SSR loci was too variable among accessions. Finally, 25 haplotypes, H1-H25, derived from the two intergenic regions accD-psaI and rps16-trnQ, were identified among 21 species in Pyrus by 36 mutations including 27 gaps and 9 base changes (Tables 4 and 5). The mutation rates including both base changes and length mutations for two intergenic regions accD-psaI and rps16-trnQ among 21 species in Pyrus were estimated ranging from 0.004 to 0.0132 with an average of 0.0062 and from 0 to 0.0044 with an average of 0.0027. Two regions evolved 1.4–3.2 times faster than another intergenic region between the rbcL-accD genes ranging 0 to 0.0035 with an average of 0.0019.

Table 5 Characterization of 14 haplotypes by insertions (+)/deletions (−) and base changes found in the 904-bp intergenic region between rps16 and trnQ in Pyrus species by DNA sequencing
Full size table
Table 6 Direct and inverted repeats detected in a gap
Full size table
Fig. 1
figure 1

Sequence alignment of intergenic region between rps16-trnQ among four species in Pyrus. Numbering starts from the first base of P. korshinskyi. Arrowheads and dashed lines mean short direct repeat regions and gaps

Full size image
Table 7 Length of mononucleotide found in breakpoints of G3 between rps16 and trnQ in Pyrus
Full size table

Haplotype network inferred from gap characters

Analysis of genetic relatedness of 21 species in Pyrus was based on the 36 mutations of 27 gaps and 9 base changes from 25 haplotypes (H1-H25) with the MJ network (Bandelt et al. 1999). Validity of gaps coded as phylogenetic characters was verified. The MJ network enabled the identification of three major groups (types A–C) from two large deletions, (the 228-bp deletion in accD-psaI and the 141-bp deletion in rps16-trnQ) (Fig. 2). The 141-bp deletion allowed Occidental pears to be categorised as type C and distinguished them from oriental pears (categorised as types A and B). Type A was grouped by seven haplotypes, H18–H25, including wild ussurian pears (P. ussuriensis), pea pears (Pyrus betulaefolia, Pyrus fauriei, and P. dimorphophylla), two Japanese pear cultivars (P. pyrifolia), and two Chinese cultivars (P. ussuriensis), and P. regelii. H18 and H19 were the most common haplotypes in type A. Apple, H19, and H20–H23 were branched out from median vector 1 (mv1). H18 was branched out from mv2, and H25 branched out from mv3, 4, and 5. The origin of haplotypes derived from median vectors is still unclear. Finally, 16 accessions (62%) from ussurian pears including wild and cultivated ones and all of the pea pears were included in type A by sequencing and/or PCR analyses (Table 1).

Fig. 2
figure 2

Median-joining network for 25 cpDNA haplotypes in Pyrus. The haplotypes are indicated by the circles, the size of each circle being proportional to the observed frequency of each haplotype. Each node of haplotype and median vectors are labelled as H and mv. Two large deletions found in intergenic regions of accD-psaI and rps16-trnQ are indicated by black rectangles. Three dashed circles (types A, B, and C) are grouped by two large deletions

Full size image

Type B was composed of six haplotypes, H12–H16, including Japanese pear cultivars (P. pyrifolia), Chinese sand pear and white pear (P. pyrifolia and P. bretschneideri), P. pashia, P. korshinskyi (H13), and P. ussuriensis var. aromatica (H16) (Fig. 2). Although P. korshinskyi belonged to Occidental pears according to their distribution, the 141-bp deletion typical of Occidental pears was not detected although the 228-bp deletion was. Thus, P. korshinskyi was involved in the type B. H15 was the most common in type B. H16, a local variety ‘Sotoorihime’ known as a derivative of P. ussuriensis var. aromatica, was branched out from H15. Type B was separated from type A by the large deletion of 228 bp in accD-psaI. There was no median vector in type B. Therefore, type B might be derived from type A, because H17 in type B connected to H24 in type A.

Type C was composed of 11 haplotypes, H1–H11, and mainly contained European pear cultivars and wild relatives native to WC Asia, Europe, and Russia (Fig. 2). The main haplotypes in European pear cultivars (P. communis) were H1 and H2 in type C. H2 included most of the European pear cultivars (P. communis), but some cultivars dispersed in H1, H3, and H4. H1 was composed from Pyrus cordata, Pyrus amygdaliformis, and P. communis ‘General leclerc’. H3 (P. communis ‘Le lectier’), H5 (P. pyraster), H8 (P. elaeagrifolia), and H9 (P. cossonii) were branched out from H1. Therefore, H1 might be ancestral to these haplotypes. P. caucasica (H7) was branched out from H2 (Pyrus balansae and five cultivars of P. communis). H6 (P. nivalis and P. salicifolia), H10 (P. gharbiana), and H11 (P. mamorensis) were branched out from H4 (P. communis ‘Marguerite Marillat’). The haplotype of P. nivalis was coincident with that of P. salicifolia. There was no median vector within type C. Type C connected to type A via mv3, 4, and 2.

Discussion

Two hypervariable regions in the conserved pear cp genome

In the previous study, we reported high conservatism of pear cp genome in comparison with those of other angiosperms (Katayama and Uematsu 2003). In contrast to the previous result, our present study revealed the two intergenic regions of accD-psaI and rps16-trnQ were extremely hypervariable. The intergenic region between rbcL-cemA (ca. 3 kb for Poaceae and ca. 5 kb for tobacco) which includes accD-psaI has been reported as a hypervariable region and might represent an intra-molecular recombinational hotspot mediated by short direct repeats in the cpDNA of Poaceae (Ogihara et al. 1988; Hiratsuka et al. 1989; Maier et al. 1995; Guo and Terachi 2005). The 228-bp largest deletion, occurring in the intergenic region between accD-psaI, was reported by RFLP analysis based on physical mapping and DNA sequencing in Japanese pears (Katayama and Uematsu 2003; Kimura et al. 2003). A 13-bp direct repeat was detected only in the left border of the 228-bp deletion in this study. This deletion might be a result of intra-molecular recombination mediated by the short direct repeat.

The recombination frequencies of large gaps were lower than those of short gaps (Hamilton et al. 2003; Ingvarsson et al. 2003; Yamane et al. 2006). The 228-bp deletion was highly stable and seemed to arise at the interspecific level. The other gaps found in this region were rearranged in a unit of repeat sequences (Table 6). In contrast to the length mutation rate, ranging from 0.004 to 0.0132 with an average of 0.0062, the base change rate was relatively low ranging from 0.0003 to 0.0021 with an average of 0.0006 for the 21 species in Pyrus (Table 3). Although the intergenic region between accD-psaI was hypervariable with respect to length mutation, the intergenic region between rbcL-accD was highly conservative in Pyrus unlike that in Poaceae (Ogihara et al. 1991). The base change rate, ranging from 0.0006 to 0.0024 with an average of 0.0009, in the region between rbcL-accD was quite low and equivalent to those of the other two intergenic regions, but the length mutation rate, ranging from 0.0009 to 0.0016 with an average of 0.001, was extremely low (Table 3). It is consistent with results indicating low genetic diversity of cpDNA in Pyrus by our previous study (Katayama and Uematsu 2003). It was lower than those of other taxa, i.e. 0.046–0.09 of rbcL-pseudo rpl23 hypervariable region in Triticum–Aegilops, 0.0055–0.333 of trnL-trnF intergenic region in Crassulaceae, and 0.0193 of rpl20-rps18 intergenic region in Pinus (Ogihara et al. 1991; Van Ham et al. 1994; Wang et al. 1999). Therefore, the hypervariable region that was observed in Poaceae may exist in the intergenic region between accD-psaI in Pyrus. An intergenic region between rps16-trnQ genes was reported as one of the hypervariable regions in cpDNA of angiosperms (Shaw et al. 2007). Earlier we reported a hypervariable region in cpDNAs of Prunus (Rosaceae), which was about 9.1 kb and included psbA to atpA near the left border of LSC (Katayama and Uematsu 2005). This highly variable region in the Prunus cpDNA may contain a recombinational hotspot region between the psbA-atpA genes, since short direct repeats of an 8-bp specific motif (GTTATTTA) and an 11-bp length T-stretch (microsatellite sequence in cpDNA) were found inside the breakpoint of the 277-bp deletion found in all Prunus species examined except for P. persica. As a result of a survey for the same region in Pyrus, a number of gaps were detected in the intergenic region between rps16-trnQ (Table 5). Mononucleotide repeats (A stretches from 9 to 21 bp) were detected in the left and right borders of the 141-bp largest deletion (Table 7). The large deletion might have been mediated through recombination across mononucleotide repeats. The 141-bp deletion was highly stable and seemed to arise at the interspecific level. Twelve gaps of various lengths in this region seemed to have resulted from rearrangement in a unit of repeat sequences (Table 6, Fig. 1). In contrast with the high rate of length mutation (average of 0.0026), the base change rate was quite low at an average of 0.0001 in this region (Table 3). As the hypervariable region in cpDNA, Van Ham et al. (1994) found a total of 50 small gaps (partly due to mononucleotide repeat variation) in a intergenic trnL-trnF spacer of 15 species belonging to the families Crassulaceae, Saxifragaceae, and Solanaceae. Twenty-seven gaps were found in two regions in 21 species in Pyrus in this study. Both of the two intergenic regions (accD-psaI and rps16-trnQ) were concluded to be hypervariable regions in Pyrus cpDNAs.

Haplotype network inferred from gap characters

After haplotype network analysis using all gap and base change characters in two regions, all cpDNA haplotypes were divided into three groups of A, B, and C by two large deletions, one being 228 bp in accD-psaI and the other being 141 bp in rps16-trnQ. These two deletions were stable and thus could be applicable to phylogenetic utility (Fig. 2).

Group A was composed basically of pea pears (P. dimorphophylla, P. fauriei, P. betulaefolia) and wild ussurian pears (P. ussuriensis Maxim. native to East Asia and in Japan). Exceptionally, two Japanese pear cultivars ‘Niitaka’ and ‘Nangetsu’ from P. pyrifolia were also included in type A. This could be explained by an interspecific hybridization between P. pyrifolia and P. ussuriensis such as the Chinese ussurian pear. Katayama et al. (2007) have already reported that three Japanese pear cultivars ‘Nansui’, ‘Niitaka’, and ‘Nangetsu’ have type A cytoplasm which might have resulted from interspecific hybridization between P. pyrifolia and P. ussuriensis.

Type B basically included pear cultivars in East Asia and India (P. pyrifolia, P. bretschneideri, and P. pashia). P. pashia was described in the literature as an intermediate in morphology between the Occidental and Oriental pear groups (Ghora and Panigrahi 1995). In the present study, P. pashia was found to have a close phylogenetic relationship to Oriental pears. Exceptionally, P. korshinskyi (H13) native to SC Asia and Afghanistan and P. ussuriensis var. aromatica ‘Sotoorihime’ (H16) were also included. Rehder (1940) considered that P. korshinskyi (synonym P. bucharica Litv.) was a related species of P. communis native to Central Asia. However, in the present study, an accession from P. korshinskyi belonged to type B; P. communis belonged to type C. Therefore, the cytoplasmic donor for P. korshinskyi remains unclear. Population structure analysis of cpDNA haplotype for P. korshinskyi will be required, because haplotype determination using a limited number of accessions from species or population has often led incorrect results.

Almost all the P. ussuriensis accessions belonged to type A, but ‘Sotoorihime’ was an exception. ‘Sotoorihime’ (H16) is known as a local pear variety derived from wild pear (P. ussuriensis var. aromatica) in the Northern Tohoku Region in Japan. Previously, we reported that about 80% of the 58 accessions of pear, including P. ussuriensis var. aromatica collected from Northern Tohoku region in Japan, had a 228-bp deletion like the haplotype in type B (Katayama et al. 2007). This suggests the coexistence of P. ussuriensis var. aromatic and hybrid progeny with P. pyrifolia. Recently, Iketani et al. (2010) represented the introgression from P. pyrifolia to P. ussuriensis var. aromatica by population structure analysis using SSR markers. CpDNA in P. ussuriensis var. aromatica in type A might be changed to that of P. pyrifolia in type B by chloroplast capture through interspecific crosses.

Type C included Occidental pears native to Europe, West and Central Asia, Russia, and Africa (P. communis including European pear cultivars, P. nivalis, P. amygdaliformis, P. elaeagrifolia, P. cossonii, P. gharbiana, P. mamorensis, P. salicifolia, P. cordata, P. balansae, P. pyraster, and P. caucasica). The main haplotypes in European pear cultivars (P. communis) were H1 and H2. P. pyraster (H5) and P. caucasica (H7), candidates for the ancestor of European pear cultivars, branched out from H1 and H2 individually. Aldasoro et al. (1996) reported that, phenotypically, P. pyraster was similar to P. caucasica. But the present study reveals that even though P. pyraster (H5) and P. caucasica (H7) share a large 228-bp gap (G12 for accD-psaI), there are still four gap differences (G7, G8, G10, G13 for accD-psaI, and G12 for rps16-trnQ) between the two accessions. Thus, in view of cpDNA structure, P. pyraster and P. caucasica might be distantly related. According to Challice and Westwood (1973), P. cordata native to the UK and France was grouped with P. cossonii and East Asian pears such as pea pears. But in the present study, P. cordata (native to Plymouth, UK) was included in H1 with P. communis ‘General leclerc’ and P. amygdaliformis both native to Europe. P. cordata may have diverged from a common ancestor of P. communis such as ‘General leclerc’ and P. amygdaliformis. At least it seems geographically congruent.

The haplotype network revealed that two long deletions were key to distinguishing three types, A, B, and C. The141-bp-long deletion in rps16-trnQ, particularly, could divide all pear accessions into two groups, i.e. Occidental pears and Oriental pears. Oriental pears were then divided into types A and B by another long deletion of 228 bp in accD-psaI. In type A, pea and ussurian pears were involved. In contrast, type B included more domesticated ones, i.e. local or modern cultivars. Therefore, pears in type A are considered to have been more wild and primitive compared to pears in type B. This result agrees with the consideration described by Challice and Westwood (1973) that pea pears were phylogenetically primitive in Pyrus based on a chemotaxonomical study using phenolics. Although group B (P. pyrifolia, P. bretschneideri, and P. pashia) may have diverged from ancestral primitives harbouring haplotype H24 in type A via a 228-bp deletion, the origin of type C (Occidental pear) is still unclear because there are many median vectors (mv2, 3, 4, and 5) between type A and C, indicating there are still unknown or disappearing intermediate types. Genetic variations between Occidental and Oriental pears were quite clear (Iketani et al. 1998; Oliveira et al. 1999; Kimura et al. 2003; Zheng et al. 2008). Occidental pears are geographically and genetically distinct from Asian pears and might have evolved and diversified independently. Recurrent mutations such as gaps found in cpDNA easily result in homoplasy and often lead to erroneous phylogenetic relationships (Golenberg et al. 1993; Graham et al. 2000). Hence, it would be better to evaluate the gaps in comparison to the base change in order to understand whether gaps found in the hypervariable regions are homoplasious mutations or not. In general, longer gaps seemed to be more stable and less homoplasious. This is only a preliminary report that suggests the phylogenetic utility of gaps in the hypervariable region as a cpDNA marker in Pyrus which has an extremely conserved chloroplast genome. These gaps could be used as a powerful tool to estimate population structure and gene flow in Pyrus.