Phylogenetic Relationships of Citrus and Its Relatives Based on matK Gene Sequences

The genus Citrus includes mandarin, orange, lemon, grapefruit and lime, which have high economic and nutritional value. The family Rutaceae can be divided into 7 subfamilies, including Aurantioideae. The genus Citrus belongs to the subfamily Aurantioideae. In this study, we sequenced the chloroplast matK genes of 135 accessions from 22 genera of Aurantioideae and analyzed them phylogenetically. Our study includes many accessions that have not been examined in other studies. The subfamily Aurantioideae has been classified into 2 tribes, Clauseneae and Citreae, and our current molecular analysis clearly discriminate Citreae from Clauseneae by using only 1 chloroplast DNA sequence. Our study confirms previous observations on the molecular phylogeny of Aurantioideae in many aspects. However, we have provided novel information on these genetic relationships. For example, inconsistent with the previous observation, and consistent with our preliminary study using the chloroplast rbcL genes, our analysis showed that Feroniella oblata is not nested in Citrus species and is closely related with Feronia limonia. Furthermore, we have shown that Murraya paniculata is similar to Merrillia caloxylon and is dissimilar to Murraya koenigii. We found that “true citrus fruit trees” could be divided into 2 subclusters. One subcluster included Citrus, Fortunella, and Poncirus, while the other cluster included Microcitrus and Eremocitrus. Compared to previous studies, our current study is the most extensive phylogenetic study of Citrus species since it includes 93 accessions. The results indicate that Citrus species can be classified into 3 clusters: a citron cluster, a pummelo cluster, and a mandarin cluster. Although most mandarin accessions belonged to the mandarin cluster, we found some exceptions. We also obtained the information on the genetic background of various species of acid citrus grown in Japan. Because the genus Citrus contains many important accessions, we have comprehensively discussed the classification of this genus.


Introduction
The genus Citrus, which includes mandarin, orange, lemon, grapefruit and lime, has high economic and nutritional value. This genus belongs to the subfamily Aurantioideae, which is one of the 7 subfamilies of the family Rutaceae. Therefore, phylogenetic study of both the genus Citrus and of the subfamily Aurantioideae is important.
The Aurantioideae consists of 2 tribes with 33 genera [1]. These 2 tribes are each composed of 3 subtribes: the tribe Clauseneae, which includes Micromelinae, Clauseninae, and Merrillinae; and the tribe Citreae, which includes Triphasiinae, Citrinae, and Balsamocitrinae. None of the Clauseneae species develop axillary spines, and the odd-pinnate leaves have alternately attached leaflets. The fruits are usually small and carry semi-dry or juicy berries, except in Merrillia. In contrast, nearly all the species develop axillary spines in the Citreae. The simple leaves are easily distinguished from those of the tribe Clauseneae. The subtribe Citrinae, in the tribe Citreae, is distinct from all the other subtribes in the Aurantioideae because of the presence of pulp vesicles in the fruit. In this subtribe, ''true citrus fruit trees'' are considered the most advanced genera based on morphological traits [1]. The genus Citrus belongs to the ''true citrus fruit trees.'' The characteristics of Citrus species include asexual reproduction, high mutation frequency, and cross compatibility between species. Because of these characteristics, there is great morphological and ecological diversity among Citrus species.
Recent progress in DNA sequencing techniques has allowed the extensive use of short DNA fragments, especially those of the chloroplast genome, in the study of phylogenetic relationships.
We have also previously reported the phylogenetic relationships among the Aurantioideae, including Citrus and its relatives, based on rbcL gene sequences [25]. The rbcL gene, located on the chloroplast DNA (cpDNA), encodes the large subunit of ribulose 1, 5-bisphosphate carboxylase/oxygenase, an enzyme that catalyzes carbon fixation in photosynthesis. Compared to most genes encoded in the cpDNA, the rbcL gene has a relatively slow nucleotide substitution rate [26][27][28]. A characteristic feature of our previous study [25] is that it included several accessions, which had not been examined in other studies [15][16][17][18][19][20][21][22][23][24]. However, the power of discrimination in our previous study was not high.
The matK gene is also located on the cpDNA and encodes a maturase involved in splicing type II introns from RNA transcripts. The matK gene is encoded by the chloroplast trnK intron. Since matK has a relatively fast mutation rate, it evolves faster than the rbcL gene [26][27][28]. Therefore, matK analysis should be useful for studying the phylogeny of the genera included in Aurantioideae.
To comprehensively analyze the phylogenetic relationships of the superfamily Aurantioideae, we determined the matK sequences of 135 accessions from 22 genera of the Aurantioideae. In this study, we used matK sequences derived from basic and major species of the Aurantioideae. Similar to our previous study [25], our current study included several accessions that had not been examined in other studies [15][16][17][18][19][20][21][22][23][24]. Furthermore, we increased the number of accessions that were analyzed, in order to focus on interspecific relationships among the mandarin varieties of the Citrus genus because several studies have suggested the existence of great genetic variation among mandarin varieties [5,6,29,30]. Our study included many kinds of mandarin varieties grown in Japan, many of which have not been studied previously at the DNA level. The genetic background of various species of acid citrus (considered to be of hybrid origin) grown in Japan has also been investigated.

Plant Materials
The 135 accessions from 22 genera of the Rutaceae subfamily Aurantioideae that were used in this study as well as the sources of the materials are shown in Tables 1 and 2

Polymerase Chain Reaction Amplification and DNA Sequencing
Crude extracts from approximately 10-mm 2 regions of leaves were prepared by incubating the leaf tissue with 100 ml of a solution containing 100 mM Tris-HCl (pH 9.5), 1 M KCl, and 10 mM EDTA, at 95uC for 20 min [31]. The primers used for polymerase chain reaction amplification of the matK gene were matK1F (59-ACCGTATCGCACTATGTATC-39) and matK1R (59-GAACTAGTCGGATGGAGTAG-39). Using the crude extract as template, the matK gene was amplified by PCR with proofreading KOD FX or KOD FX Neo DNA polymerase (Toyobo, Osaka, Japan). The amplified DNA fragments were purified using the MonoFas DNA Purification Kit I (GL Sciences, Tokyo, Japan). The primers used for sequencing of the matK gene were matK1F, matK2F (59-ACGGTTCTTTCTCCACGAGT-39), matK3F (59-GGTCCGATTTCTCTGATTCT-39), matK1R, matK2R (59-AGAATCAGAGAAATCGGACC-39), and matK3R (59-ACTCGTGGAGAAAGAACCGT-39). The purified DNA fragments were sequenced in both directions in an Applied Biosystems 3130 Genetic Analyzer (Applied Biosystems) with a BigDye Terminator Cycle Sequencing Ready Reaction Kit v. 3.1 (Applied Biosystems), as described previously [32]. Sequence data were submitted to DDBJ/GenBank/EBI and were assigned accession numbers ranging from AB626749 to AB626802 and from AB762316 to AB762396. The DNA templates used in this study are distinct from those used in the previous study [25].

Phylogenetic Analyses
The maximum likelihood (ML) and neighbor-joining (NJ) methods from the MEGA (version 5.05) program [33] were used to create phylogenetic trees. The reliability of each branch was tested by bootstrap analysis with 1,000 replications. The sequences of Zanthoxylum sp. Clayton 15 or Triphasia trifolia were used as an outgroup.

Results and Discussion
We constructed multiple sequence alignments of DNA sequences containing the matK gene from different accessions. The typical length of the protein-coding sequences and 39 UTRs was 1,530 bases and 100 bases, respectively. However, some indels were present in several accessions. Because no indel was observed in the rbcL gene of this subfamily [25], we concluded that matK had a relatively fast mutation rate. The matK sequence used in this study was different from the partial matK sequence used in other published studies [16,34,35]. It is also different from the partial matK sequence used in the unpublished DNA barcoding projects. In addition, although other studies [24,36] used the entire matK sequence, the number of accessions tested was small.
Phylogenetic trees of Aurantioideae were created using the ML ( Figure 1) and NJ ( Figure 2) methods. Except for 7 species (Glycosmis citrifolia, Glycosmis pentaphylla, Murraya koenigii, Micromelum minutum, Clausena anisata, Clausena harmandiana, and Clausena lansium), both phylogenetic trees showed the same topology. Among these 7 species, 3 species of genus Clausena belonged to the same cluster, and 2 species of the genus Glycosmis belonged to the same cluster. A characteristic feature of both trees is that the ''true citrus fruit trees'' were clearly distinguished from other species. Another feature is that the ''wood apple group'' of Balsamocitrinae (Feronia limonia and Feroniella oblata), ''primitive citrus fruit trees'' (Severinia buxifolia) and ''near citrus fruit trees'' (5 species of the genus Atalantia) also belonged to the same cluster in both trees. The remaining species were divided into a few groups in both phylogenetic trees. One group contained ''primitive citrus fruit trees'' (Hesperethusa crenulata), ''near citrus fruit trees'' (3 species of Citropsis), and the tabog group (Swinglea glutinosa). Another group contained 2 species of the Bael fruit group (Aegle marmelos and Afraegle paniculata), and yet another group contained Murraya paniculata and Merrillia caloxylon. Members of the tribe Citreae belonged to the same large cluster, whereas members of the tribe Clauseneae did not.
The trees showed that the tribe Citreae is monophyletic, that is, they clearly discriminated Citreae from Clauseneae. This result supports Swingle and Reece's system of tribes. Previously, Morton et al. [20] and our group [25] reported the difficulty encountered in discriminating Citreae from Clauseneae in the analyses of rps19 and rbcL sequences, respectively. Our current results may be attributable to the fact that the matK data have high discrimination power. Although a previous study using 9 cpDNA sequences also discriminated Citreae from Clauseneae [16], we succeeded in this discrimination by using only 1 chloroplast DNA sequence. In contrast, the members of the tribe Clauseneae did not belong to the same cluster, i.e., Clauseneae is not monophyletic. Rather, the  C. medica var. sarcodactylis Swing.
Citron Fingered Citron A 5008  members of Clauseneae appeared to be an outgroup of Citreae. This result did not support Swingle and Reece's system of tribes. Thus, the matK data only partially supported Swingle and Reece's system of tribes. The previous study using 9 cpDNA sequences [16] also showed that the members of the tribe Clauseneae did not belong to the same cluster. Next, we focused on Swingle and Reece's system of subtribes. The matK data did not completely supported Swingle and Reece's system of subtribes. For example, the Balsamocitrinae (Swinglea glutinosa, Aegle marmelos, Afraegle paniculata, Feronia limonia, and Feroniella oblata) were not placed in 1 cluster. In contrast, Swingle and Reece [1] considered that the Balsamocitrinae paralleled the Citrinae and that both had evolved from a common ancestor. Among Balsamocitrinae, the Bael fruit group of Balsamocitrinae (Aegle marmelos and Afraegle paniculata) is clustered together, which is consistent with previous reports [16,20,25]. Similarly, the ''wood apple group'' of Balsamocitrinae (Feronia limonia and Feroniella oblata) is clustered together, which is consistent with the previous reports by us [25] and Morton et al. [20].
Interestingly, Bayer et al. [16] concluded that Feroniella oblata is nested in Citrus species. The leaves of Feronia and Feroniella are oddpinnate with paired opposite leaflets on a rachis. In addition, they are morphologically different from other genera of the orange subfamily; the core or axis of the ovary has disappeared and an entirely new placentation has developed in this group. The 2 genera were distinguished on the basis of a few flower, seed, and fruit traits. Thus, Feronia and Feroniella are believed to be closely related to each other but they are not closely related to the remaining genera of the orange subfamily [1]. To illustrate this further we have include photographs of the leaves from these species in Figure 3. Thus, we have provided evidence that the conclusion made by Bayer et al. [16] is misleading.
In the Citrinae, our analysis did not clearly support the distinction between ''primitive citrus fruit trees'' and ''near citrus fruit trees,'' made by Swingle and Reece [1]. The tree strongly supported the polytomous clade containing Hesperethusa crenulata and 3 Citropsis species (Citropsis gabunensis, Citropsis gilletiana, and Citropsis schweinfurthii), which is consistent with our previous report [25]. Similarly, Morton et al. [20] reported that Hesperethusa crenulata and Citropsis gilletiana are clustered together, and Bayer et al. [20] reported that Hesperethusa crenulata, Citropsis schweinfurthii, and Citropsis daweana are clustered together. Hesperethusa and Citropsis have some morphological similarities such as odd-pinnate leaves with broadly winged petioles and seeds with a hard testa. In addition, both are graft compatible with Citrus. However, Hesperethusa is native to Southeast Asia, whereas Citropsis is present only in Africa [1].
Our matK data showed that Murraya paniculata and Merrillia caloxylon are clustered together, which is consistent with the previous reports by us [25] and Bayer et al. [16]. However, Murraya koenigii belongs to an independent cluster, which is consistent with our previous report [25]. Although the flowers of Merrillia caloxylon, which are very long (55-60 mm long) and trumpet shaped, are unique to the orange subfamily, this species resembles Murraya paniculata in its general growth habit, leaf shape, and wood texture. These 2 species grown in the wild are sometimes confused by natives of Malay Peninsula. As shown by the results from a series of studies [1], Swingle considered that Merrillia had probably evolved from a Murraya-like ancestral form. On the other hand, the leaf shapes of Murraya paniculata and Murraya koenigii are different. The leaf number of Murraya koenigii is larger than that of Murraya paniculata, although both species have odd-pinnate leaves (see Figure 4).
Our matK data showed that Clausena anisata, Clausena harmandiana, and Clausena lansium formed a monophyletic group, which is consistent with our rbcL data [25]. Bayer et al. [16] reported that Clausena harmandiana is clustered with Clausena excavate. Furthermore, our matK data showed that Glycosmis citrifolia and Glycosmis pentaphylla formed a monophyletic group, which is consistent with our rbcL data [25]. Bayer et al. [16] reported that Glycosmis pentaphylla is clustered with Glycosmis trichanthera and Glycosmis mauritiana. Thus, each of Clausena and Glycosmis forms a monophyletic group. Both phylogenetic trees produced a well-supported (BS, 99%) clade that contained all members of the ''true citrus fruit trees,'' thus supporting their monophyletic origin (Figures 1   and 2). All the genera belonging to ''true citrus fruit trees'' were incorporated into Citrus in accordance with the results reported by Mabberley [37,38], Zhang et al. [39], Bayer et al.  [16], and Tshering Penjor et al. [25]. Our results support their concept.
These ''true citrus fruit trees'' contain the genus Citrus having high economic and nutritional value. Therefore, we phylogenet-ically analyzed ''true citrus fruit trees'' by creating ML and NJ trees with Triphasia trifolia as an outgroup (Figures 5 and 6, respectively). The topologies of both trees were essentially identical and could be classified into 2 clusters. One large cluster included   [3]. The analysis of the rbcL sequence [25] and 9 cpDNA sequences [16] also showed a close relationship between both genera. It is generally believed that the species of Microcitrus and Eremocitrus arose over millions of years of slow evolution in geographically isolated land masses (Australia and New Guinea), separate from the other genera of the true citrus group in Southeast Asia [1]. Thus, together with other studies [16,25], our current study showed that Microcitrus species formed a monophyletic group. However, Lu et al. [23] showed that Microcitrus australasica was nested in Citrus species, although Microcitrus australis is clustered with Eremocitrus glauca. Furthermore, Li et al. [22] showed that Microcitrus australasica and Microcitrus australis are not clustered together. We suspect that the Microcitrus australasica used in the latter studies might have been of hybrid origin and that its maternal parent might be different from Microcitrus. Further study is required to confirm these conflictions.
The isolated genus Clymenia is native to the Bismarck Archipelago of New Guinea and is considered to be the most primitive of all the genera of ''true citrus fruit trees'' owing to its morphological traits [1]. Another study [16] also suggested that Clymenia was isolated from other species.
Our matK data showed that Fortunella species formed a monophyletic group, which is consistent with previous reports [16,23,25]. Swingle and Reece [1] placed Fortunella at the genus level only because it has 2 collateral ovules near the top of each locule, whereas Citrus has 4-12. However, Fortunella was placed within the Citrus group in the present study. Fortunella has also been classified within the Citrus group in previous studies on DNA analysis [10][11][12][13][14]16,22,23,25,40,41]. It is considered to be difficult to distinguish Fortunella from Citrus at the DNA level, based on these results.
Poncirus was also placed within the Citrus group in the present study. However, Swingle and Reece [1] placed Poncirus at the genus level mainly because it is deciduous, its flowering period differs from Citrus, it has trifoliate leaves, and its geographical distribution differs from that of Citrus. Several DNA analysis studies on citrus phylogeny [10,12,40] have revealed that Poncirus is distant from Citrus. However, some studies involving cpDNA analysis have suggested a close relationship between Poncirus and Citrus [15,16,20,22,23,25]. These studies, along with the present study, strongly suggest that Poncirus is closely related to Citrus at the DNA level.  Next, we focused on the members of the genus Citrus that are economically and nutritionally important fruit trees. Our current study is the most extensive phylogenetic study of Citrus species among other studies (e.g., [16,22,23,25]) because we studied 93 accessions of the Citrus species. The phylogenetic tree showed that Citrus can be classified into 3 clusters: the citron cluster, the pummelo cluster, and the mandarin cluster (Figures 5 and 6). This finding is consistent with the results of previous studies using not only cpDNA but also nuclear and mitochondrial DNA (mtDNA) [10][11][12][13][14][15][16]23,42]. We found that Poncirus and Fortunella belonged to the parent cluster containing the members of the genus Citrus, but did not belong to these 3 clusters.
The mandarin cluster can be divided into 1 major cluster and some minor subclusters. The major subcluster includes, amongst others, C. reticulata, C. unshiu, C. clementina, C. kinokuni, C. deliciosa, C. reshni, and C. sunki. One of the minor subclusters includes C. tachibana and C. depressa. The major subcluster also contains some C. depressa accessions. C. ichangensis and C. junos belonged to the mandarin cluster, but they were not grouped with either of the subclusters. Some Japanese acid citrus (C. sp. speciosa, C. hanaju, and C. acidoglobosa) were included in the C. junos subcluster. Similarly, Bayer et al. [16] reported that C. reticulata, C. junos, C. tachibana, and C. ichangensis were clustered together, and Lu et al. [23] reported that C. reticulata, C. sunki, and C. tachibana were clustered together. However, because the number of tested accessions is larger in our study, we addressed the classification of the mandarin cluster extensively, as discussed below.
Our matK analysis showed that most mandarin accessions belonged to the mandarin cluster. However, our extensive analysis showed that there are some exceptions. The mandarin accessions, Citrus keraji, Citrus oto, Citrus tarogayo, Citrus platymamma, and Citrus yatsushiro belonged to the pummelo cluster, and not to the mandarin cluster. Interestingly, Citrus nobilis (King) belonged to the mandarin cluster, and Citrus nobilis (Kunenbo) belonged to the pummelo cluster. Similarly, the previous study [22] showed that most mandarin landraces formed a monophyletic clade and some exceptions were similar to pummelo, although their tested accessions were different from ours.
The major subcluster of the mandarin cluster includes edible mandarins such as C. reticulata, C. unshiu, and C. clementina. In the present study, it was difficult to further classify members of this subcluster. Similarly, the previous researchers who conducted analyses based on cpDNA [12,22,43,44] and mtDNA [14,41] did not subdivide these members. Thus, our analysis confirmed the similarities of these mandarins with respect to organellar DNA. The major subcluster of the mandarin also includes C. reshni, C. sunki, and C. depressa, which are small-fruited mandarins mainly used as rootstocks. In contrast, the previous study using 3 cpDNA sequences [23] separated Citrus sunki from Citrus reticulata. One of the minor subclusters of the mandarin consists of C. tachibana and C. depressa, which are native to Japan. Hence, C. tachibana and some C. depressa accessions are different from other members of the mandarin cluster. This finding is consistent with the findings of previous studies that used cpDNA analyses [12,15,19,23,40,43]. Furthermore, the analyses of mtDNA [14], isozymes [6], and chromosomes [30] also separated C. tachibana from the other members of the mandarin cluster. Thus, our analysis confirms that C. tachibana differs from other members of the mandarin cluster. Members of C. depressa are placed in both the major and the minor subclusters. Consistent with our analysis, the analyses by Urasaki et al. [44] and Yamamoto et al. [45] showed the genetic diversity in cpDNA among the members of C. depressa. These results suggest that different C. depressa accessions have different origins.
The pummelo cluster contained, amongst others, C. maxima, papeda, C. sinensis, C. limon, and C. aurantifolia. As described above, this cluster also contained some mandarin accessions such as C. nobilis (Kunenbo) and C. keraji. This cluster can be divided into 7 subclusters, which are represented by the species C. sinensis, C. maxima, C. latipes, C. aurantifolia (plus papeda), C. aurantium, and C. limon and the Citrus hybrid cultivar 'Tosu'. A number of mandarins and Japanese acid citrus belonged to the C. sinensis subcluster. Similarly, Bayer et al. [16] reported that C. sinensis, C. maxima, C. aurantifolia, papeda, C. aurantium, and C. limon were clustered together. However, because the numbers of tested accessions are larger in our study, we addressed the classification of the pummelo cluster extensively, as discussed below.
Some of the mandarins are clustered exclusively with pummelo accessions. This result strongly suggests that their maternal origins are members of the pummelo cluster. Previous studies showed that C. nobilis (Kunenbo) has organellar DNA derived from a member of the pummelo cluster [14,46]. Its maternal predecessor is probably a sweet orange with cytoplasmic DNA originally derived from a member of the pummelo cluster. C. nobilis (Kunenbo) is thought to be the maternal ancestor of C. keraji (Keraji and Kabuchii) and C. oto [46,47]. Consistent with that hypothesis, the results of the current study show that the matK sequences from C. keraji (Keraji and Kabuchii), C. oto, and C. nobilis (Kunenbo) are identical. In agreement with the results of a previous morphological characterization [48], the results of our analysis suggest that C. nobilis (Kunenbo) is also the maternal origin of C. tarogayo.
We also studied the nucleotide variation in the matK genes from accessions that were grouped in the pummelo cluster, although the diversity is low. By using SSR analysis, Deng et al. [43] showed that cpDNAs of pummelos and their relatives have many types of nucleotide polymorphisms. Except for pummelo (C. maxima) and papeda, all members of the pummelo cluster originate from hybrids. Therefore, the nucleotide variation among the maternal ancestors can contribute to the variation in the pummelo cluster. Our results show that the matK sequences of C. aurantium, C. sinensis, C. limon, and C. limettioides are derived from members of the pummelo cluster, and this finding is consistent with the results of previous studies using cpDNA analysis [12,15,16,40,47,49]. C. wilsonii and C. latifolia also have pummelo-type cpDNA sequences, and to our knowledge, this is the first study to highlight this point. The matK sequences from C. aurantium and C. limon in the pummelo cluster differ by single-base pair mismatches. Previous analyses of cytoplasmic DNA suggested that C. aurantium is the maternal ancestor of C. limon [14,50]. Thus, further work is required to confirm this result.
Swingle and Reece [1] divided the genus Citrus into the subgenera Citrus and Papeda. However, in the present study, papeda belonged to the pummelo cluster, and it is difficult to Figure 6. Neighbor-joining tree of the matK genes from accessions belonged to ''true citrus fruit trees.'' Numbers at the nodes indicate bootstrap values (% over 1,000 replicates). Numbers in parenthesis indicates the number of accessions. Citrus depressa z contains 6 accessions (Kaachi, Mikanguwa, Shiikunin, Shiikuribu, Ishikunibu, and Okitsu strains). Citrus depressa y contains 2 accessions (Fusubuta and Kabishi). doi:10.1371/journal.pone.0062574.g006 discern the subgenera Citrus and Papeda. Thus, our analysis does not support the classification proposed by Swingle and Reece [1]. Similar results were reported based on the cpDNA analyses [12,16]. The matK sequences from C. hystrix, C. micrantha, and C. macroptera are identical, which is consistent with the results of previous analyses of cpDNA by Nicolosi et al. [12]. They suggested that C. micrantha was the maternal ancestor of C. aurantifolia. Our present results show that the matK sequence from C. aurantifolia is identical to that from C. hystrix, C. micrantha, and C. macroptera, which confirms the results of the study by Nicolosi et al. [12]. However, the rbcL genes and chloroplast SSRs of C. micrantha and C. aurantifolia differ [25,43]. Therefore, their maternal relationship is unclear.
Most of miscellaneous acid citrus species grown in Japan belonged to the mandarin major cluster, the C. junos cluster, or the C. sinensis cluster. Although this result almost agrees with that of Asadi Abkenar et al. [51] who analyzed cp and mtDNA using PCR-RFLP, our analysis showed that matK sequences of some accessions (C. sp. tenuissima and C. nanseiensis) were identical with that of C. maxima. As above mentioned, matK sequence of C. sinensis was similar to that of C. maxima but not identical with it. Our result was more informative than the previous study [51] and it is considered to be useful information for estimation of genetic background of Japanese acid citrus. These data obtained from cytoplasmic DNA analysis were not coincident with the RAPD data [40]. Furthermore, our analysis clearly separated Citrus hybrid cultivar 'Tosu' from the other citrus. Thus, these accessions appear to have arisen from complicated cross and combined results of nuclear and cytoplasmic genomes; further studies are required to elucidate their exact their origin.
Swingle and Reece [1] have classified C. ichangensis into the subgenera Papeda. However, our analysis shows that C. ichangensis belongs to the mandarin cluster. There are conflicting reports regarding the cytoplasmic relationships between C. ichangensis and other Citrus species. Consistent with our current result, based on the analyses of cpDNAs, Asadi Abkenar et al. [40], Bayer et al. [16], Deng et al. [43], and Nicolosi et al. [12] showed that C. ichangensis is closely related to the mandarins. However, according to the phylogeny of rbcL [25], C. ichangensis is most closely related to the Poncirus species. On the other hand, cpSSR analysis [49] failed to show a relationship to other species. A comparison of mtDNA sequences [41] suggested that C. ichangensis was not related to the mandarins and was identical to C. hystrix and C. aurantifolia. A complete understanding of the cytoplasmic relationships of C. ichangensis with the Citrus species and its relatives including the Poncirus species requires further analyses.

Conclusions
Based on the chloroplast matK sequences, the present study provides novel information that resolves the genetic relationships among members of the Aurantioideae, especially of the genus Citrus, and confirms previous observations. Because the matK gene has a relatively fast rate of nucleotide substitutions, our study provides more information on interspecific relationships within the genus Citrus than the analysis of the rbcL gene [25]. Our extensive classification of 135 accessions from 22 genera of the Aurantioideae could be useful for the breeding of these trees.