Molecular phylogeny of Dioscorea ( Dioscoreaceae ) in East and Southeast Asia

You are free to share to copy, distribute and transmit the work, under the following conditions: Attribution: You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). Non-commercial: You may not use this work for commercial purposes. No derivative works: You may not alter, transform, or build upon this work. For any reuse or distribution, you must make clear to others the license terms of this work, which can be found at http://creativecommons.org/licenses/by-nc-nd/3.0/legalcode. Any of the above conditions can be waived if you get permission from the copyright holder. Nothing in this license impairs or restricts the author’s moral rights. Blumea 58, 2013: 21–27 www.ingentaconnect.com/content/nhn/blumea http://dx.doi.org/10.3767/000651913X669022 RESEARCH ARTICLE

Since Dioscorea is a large genus, many researchers have proposed infrageneric classifications of Dioscorea. Knuth (1924) has proposed 58 sections in Dioscorea, most of which are still used today. Prain & Burkill (1936, 1938 presented several new sections for the Asian members of Dioscorea. In comparison to Kuth (1924) they placed greater emphasis on seed characters, underground organ and male inflorescence morphology. Burkill (1960) proceeded to the arrangement of the Old World sections of the genus Dioscorea, dividing 220 species into 23 sections. Recently, Huber (1998) has proposed 28 sections of Dioscorea including Borderea, Epipetrum, Rajania, Tamus and Testudinaria. However, more detailed studies of the infrageneric classification of Dioscorea have revealed that several sections seem to be artificial groupings, and many species are not fit to their section boundaries. For example, compound-leaved yams in the Old World (D. sect. Lasiophyton, sect. Trieuphorostemon and sect. Botryosicyos) have been treated as one to three sections by different systematists (Knuth 1924, Prain & Burkill 1936, Ding & Gilbert 2000. Phylogenetic relationships of Dioscorea have presented a challenge to systematists for many years because of the difficulties in species identification, which is due to a continuous variability of morphological characters, especially of aerial parts, such as leaves (Pavan . Further, many morphological characters are shared by different species, which make the identification and classification of the genus a rather difficult task. For example, some classifications have considered D. batatas, D. doryphora and D. potanini as synonyms of D. polystachya, because those species have many morphological characters in common (Ding & Gilbert 2000).
A further question is whether these morphological groups correctly reflect their genetic relationships within Dioscorea. Recent studies have analysed molecular datasets to provide additional indications of the relationships within this genus. The phylogenetic relationships of six species (D. gracillima, D. nipponica, D. quinqueloba, D. septemloba, D. tenuipes and D. tokoro) in D. sect. Stenophora were investigated based on DNA sequences of the phosphoglucose isomerase (Kawabe et al. 1997). It was reported that D. tenuipes and D. tokoro were clustered into a clade, while the rest species formed a separate clade.
Furthermore, chloroplast sequence data has been used to examine the phylogenetic relationships within Dioscorea. Wilkin & Caddick (2000) found that the palaeotropical compound-leaved yams were classified into two monophyletic groups based on a combined analysis of chloroplast sequence data and morphological characters. Later, the phylogenetic relationships of 67 Dioscorea taxa were reconstructed based on chloroplast rbcL and matK sequence data . They found that the main Old World groups (such as the left-twining D. sect. Stenophora and the right-twining D. sect. Enantiophyllum) were monophyletic. However, these studies included a limited sampling of Asian species and the obtained phylogenetic resolution  was relatively low. Therefore the phylogenetic relationships among the species of Asian Dioscorea have not been well established (e.g., in D. sect. Shannicorea). Further studies to resolve both the limits of their species and the phylogenetic relationships between them are necessary.
A complete plastid genome of a Dioscorea species is available (Hansen et al. 2007) and this provides a rich source of phylogenetic tools to unravel the genetic relationships within Dioscorea. Based on chloroplast genes including trnL-F, matK, rbcL and atpB-rbcL sequence data, the objectives of this study are to further clarify infrageneric classification of Asian Dioscorea and provide information for the genetic conservation of wild and cultivated yams. We examine currently recognized species within seven sections (sect. Botryosicyos, Combilium, Enantiophyllum, Lasiophyton, Opsophyton, Shannicorea and Stenophora) from East and Southeast Asia and investigate the relationships amongst these sections. We compare our results to recent studies of Dioscorea and the molecular phylogeny of Dioscorea in East and Southeast Asian is discussed.

Taxon sampling
Our analysis of chloroplast trnL-F, matK, rbcL and atpB-rbcL covered a total of 72 accessions of 48 ingroup species and five outgroup species (Table 1). These five outgroup taxa were part of Tacca and Stenomeris in Dioscoreaceae and Stemona in Stemonaceae (Caddick et al. 2002).

Sequence analyses
The sequences were aligned and edited using BioEdit 7.0.1 (Hall 1999). The alignments of the concatenated sequence datasets were obtained by using CLUSTAL-X version 1.83 (Thompson et al. 1997) with manual adjustments for accuracy. Statistical analyses of the alignments were performed using MEGA v. 4 (Tamura et al. 2007).

Phylogenetic analyses
After alignment, phylogenetic analyses were conducted with PAUP* 4.0b10 (Swofford 2002) using the methods of distance and maximum parsimony (MP). Bayesian inference (BI) analyses were conducted with MrBayes 3.1.2 (Ronquist & Huelsenbeck 2003). The optimal model of nucleotide substitution was evaluated by a likelihood ratio test with MODELTEST 3.7 (Posada & Crandall 1998). The K81uf+I+G model with proportion of invariable sites (I) = 0.3661 and gamma distribution shape parameter (G) = 0.9624 was selected as the best model for the concatenated DNA sequence of trnL-F, matK, rbcL and atpB-rbcL genes.
Based on this model, a distance tree was constructed with the neighbor-joining (NJ) algorithm. In the MP analysis, characters were equally weighted and a heuristic search option with tree bisection reconnection (TBR) branch-swapping and 10 random stepwise additions was used (gaps were treated as missing data). All bootstrap values were based on 1 000 replicates performed for NJ and MP. The BI analysis was run for 2× 106 generations, with a sample frequency of 100. The first 2 000 trees were discarded and 18 000 trees were applied in the final consensus tree. The posterior probabilities (calculated with MrBayes) were recorded to represent the support for nodes.

Sequence characteristics and variations
For all Dioscorea and the outgroup species, the sequenced trnL-F region was 640-745 bp, the matK region 895-901 bp, the rbcL region 1 159 bp and the atpB-rbcL region 690-838 bp. The lengths of the alignments are given in

Opsophyton
A B D C compound-leaved Macroura Fig. 1 Bayesian tree of Dioscorea specimens reconstructed with combined chloroplast genome trnL-F, matK, rbcL and atpB-rbcL DNA sequences. Statistical supports for each node (node numbers on the branches of the tree) in NJ, MP and BI analyses is shown in the table on the left. An asterisk (*) indicates a node value < 50 %.

Phylogenetic analyses
The MP analysis of the combined dataset resulted in a single tree of 1 601 steps with CI = 0.78 and RI = 0.90. The phylo genetic tree based on the cpDNA combined datasets as reconstructed by the Bayesian method with statistical supports for each node in NJ, MP and BI analyses is shown in Fig. 1. There were no supported contradictions between the topologies of NJ, MP and Bayesian consensus tree. The Dioscorea species formed a monophyletic group with maximum support at node 4. Within Dioscorea, there were two strongly supported clades, clade A (node 5, 100/100/1.00) and clade B (node 21, 100/100/1.00). Clade B was further divided into two clades (C and D). Clade C included two strongly supported sections, D. sect. Combilium (node 10, 98/98/1.00) and D. sect. Shannicorea (node 16, 100/99/1.00), which were moderately supported (node 12, 85/65/0.99) as sister to each other. Clade D includes five strongly supported sections. Within D. sect. Opsophyton, 6 individuals of D. bulbifera and two individuals of D. sansibarensis were not clustered together, but formed monophyletic clades with strong support (node 20 and 34, 100/100/1.00) individually. Next, the D. sect. Botryosicyos clade was strongly supported as monophyletic group (node 27, 97/93/1.00), and its sister D. sect. Lasiophyton was also strongly supported as monophyletic group (node 32, 100/100/1.00). Finally, D. sect. Enantiophyllum was also strongly supported as monophyletic (node 47, 100/99/1.00).

Systematic implications of the molecular phylogeny
Based on their twining stems, compound leaves, underground organ morphology, hairs, male flowers, capsule and seed characters, the species of Asian Dioscorea can be divided into nine sections (sect. Botryosicyos, Combilium, Enantiophyllum, Lasiophyton, Opsophyton, Paramecocarpa, Shannicorea, Stenocorea and Stenophora). A total of seven out of nine sections (except D. sect. Stenocorea and Paramecocarpa) were included in our analysis and the phylogenetic tree of Dio scorea was reconstructed by cpDNA combined datasets. Our results show general support for the infrageneric classification of Dioscorea.

The Stenophora clade
As shown in Fig. 1, all Dioscorea species formed a monophyletic group with two distinct, strongly supported clades (clade A and B). This confirms that D. sect. Stenophora (clade A) is sister to the rest of Dioscorea (clade B) in the systematics of the genus as reported in Wilkin et al. (2005). Many ancestral characteristics of the genus are also present in D. sect. Stenophora including rhizome, diploid chromosome number and single pollen aperture (Pei et al.1979, Chin et al.1985, Schols et al. 2003. Because its fossil record is the earliest of the genus Dioscorea, sect. Stenophora has been proposed as the oldest section in Dioscorea (Burkill 1960).
Furthermore, D. collettii was reported as a Sino-Himalayan species in Thapyai et al. (2005). Burkill (1960) had distinguished an additional species from D. collettii, which he called D. hypoglauca. However, in the most recent treatment of this species, Ding & Gilbert (2000) defined D. hypoglauca as a variety of D. collettii, D. collettii var. hypoglauca. These two taxa exhibit continuous morphological variations and show sympatric distribution in China. Gao et al. (2008) suggested that D. collettii var. collettii and D. collettii var. hypoglauca were sister to each other with only weak support. In this study, these two taxa were also sister to each other, but with strong support (Fig. 1 node 7). In addition, the specimens of D. collettii var. collettii sampled from Taiwan and Lanyu Island showed three stable transversions within cpDNA trnL-F and matK regions. Thus, denser sampling is required to evaluate the intraspecific classification of D. collettii var. collettii in the future.

The Combilium and Shannicorea clades
Dioscorea sect. Combilium and D. sect. Shannicorea show some morphological characters in common, such as producing one or several annually renewed storage tubercles, capsules which are longer than their wide and distally-winged seeds. In the arrangement of the Old World sections of the genus Dioscorea, Burkill's (1960) has divided 220 species into 23 sections. He has emphasized on the seed characters, underground organ morphology and development, and male inflorescence morphology as the defining characteristics in his report. Describing the relationships among these 23 sections, he indicated that D. sect. Combilium and D. sect. Shannicorea were closely related. This is also supported by our result in which a novel sister relationship of D. sect. Combilium to D. sect. Shannicorea was found with moderate support (Fig. 1  node 12). Furthermore, our study is the first analysis showing the internal topology of the Shannicorea with strong support ( Fig. 1 node 16). Within this monophyletic clade, four taxa are endemic to southern China (D. martini, D. nitens, D. subcalva and D. yunnanensis), one is distributed in Northern Thailand, Myanmar and southern China (D. velutipes) and one is distributed from central China to Indochina (D. hemsleyi). These five species plus one variety (D. subcalva var. submollis) are grouped together and sister to D. hemsleyi ( Fig. 1 node 16). Within this clade, a major branch is found in the NJ and BI tree, but not in the MP tree ( Fig. 1 node 13). Dioscorea martini and D. nitens were grouped together and sister to D. yunnanensis and D. velutipes, these four species were closer to D. subcalva var. submollis than to D. subcalva. Dioscorea sect. Shannicorea comprises eight species, of which a total of six species and one variety were investigated for their phylogenetic relationships in this study. With regard to the species within D. sect.
Shannicorea not included in this study D. pseudonitens Prain & Burkill was not sister to D. tentaculigera Prain & Burkill in the phylogenetic tree presented by Wilkin et al. (2005). Recently, Wilkin & Thapyai (2011) have reported that D. pseudonitens is conspecific with D. nitens. In summary, this study should be able to provide a framework for Shannicorea clade but it would need further study to evaluate the phylo genetic position of D. tentaculigera in the future.

The Opsophyton clade
Dioscorea bulbifera is the main species of D. sect. Opsophyton distributed in the wild state in both Asia and Africa. The formation of many axillary tubers (bulbils) is the distinct characteristic of D. bulbifera, but intraspecific classifications are still diverse. Prain & Burkill (1936) have treated the African form with angular bulbils as a single variety, D. bulbifera var. anthropophagorum, and the Asian form with globular bulbils has been divided into nine varieties according to highly variable characters such as the colour, shape, and dimension of bulbils and leaves. In this study, we found six different haplotypes of the Asian form of D. bulbifera (data not shown). Two accessions from Bangladesh and Indonesia (D. bulbifera-2 and D. bulbifera-6) were grouped together and sister to the rest of D. bulbifera in clade D (Fig. 1).

The Botryosicyos and Lasiophyton clades
These two sections show many morphological characters in common including perennial crown with annual tubers, lefttwining, usually pubescent and spiny, compound leaves and capsules that are longer than their wide. Prain & Burkill (1936) have combined these two sections and treated the members of D. sect. Botryosicyos within D. sect. Lasiophyton. However, these two sections show clear morphological differences to each other, such as the variations in leaflet venation, male bracts, and stamen number. Thus, the obtained phylogenetic relationships seem to be well-supported by morphological characters. As shown in Fig. 1, D. sect. Botryosicyos and D. sect. Lasiophyton were both identified in the tree as well-supported clades within the compound-leaved clade ( Fig. 1 node 27 and 32). The members of D. sect. Botryosicyos, characterized by one main vein per leaflet, were sister to those of D. sect. Lasiophyton, which had several veins per leaflet.

The Enantiophyllum clade
In Wilkin et al. (2005), twelve species of D. sect. Enantiophyllum were sampled and found to form a monophyletic clade with strong support. Our study was based on a sampling of 24 taxa of sect. Enantiophyllum and obtained a similar result with the monophyly of the section also strongly supported ( Fig. 1 node  47). Dioscorea sect. Enantiophyllum is consistently defined by right-twining stems and usually opposite leaves. This section is the largest in terms of the number of species, with about 120 species, distributed mainly in tropical Asia and Africa (Prain & Burkill 1938), but still many species are often not clearly distinguished. There are two main groups under Enantiophyllum section, an Asian-Oceanian group and an African group. Wilkin et al. (2005) reported that the African species D. schimperiana Hochst. ex Kunth and the Asian species were clearly separated. In addition, it was suggested by Tostain et al. (2006) that the haplotypes of African species were different from those of the Asian-Oceanian species based on data derived from SSR markers. In this study, Asian species of this section were investigated and several groups of which relationships were not clear in previous studies were clearly identified (Fig. 1). Malapa et al. (2005) proposed that D. alata, the most important cultivated yam in Asia, should be grouped with D. nummularia and D. transversa together representing a southeast Asian-Oceanian gene pool, rather than to D. persimilis (as a synonym of D. hamiltonii) as reported in Wilkin et al. (2007). However, our study has surveyed three typical species, D. alata, D. nummularia and D. hamiltonii, and the result showed that D. alata and D. hamiltonii were grouped together with strong support (Fig. 1 node 60) and sister to the rest of Asian-Oceanian species.
Many species identification and nomenclatural problems of the group, from D. japonica to D. potanini, have been mentioned in previous studies. For example, Ding & Gilbert (2000) considered that D. batatas, D. doryphora and D. potanini should be regarded as a synonym of D. polystachya. In this study, it was shown that they could be distinguished from each other ( Fig. 1 nodes 40, 42, 44). However, further experiments with population-based sampling would be necessary to verify clearly the phylogenetic relationships among D. batatas, D. doryphora and D. potanini. In addition, it is interesting to notice that three different haplotypes of D. japonica are found and do not form a monophyletic group within the Enantiophyllum clade (Fig. 1). Additional synonyms and varieties of D. japonica were also reported in Prain & Burkill (1938). Therefore, a denser sampling is required to evaluate the intraspecific classification of D. japonica in the future. Finally, Fig. 1 shows that eight right-twining species (D. sp. A-H) fall within the Enantiophyllum clade. The results also are congruent with those of Wilkin et al. (2005), the right-twining habit has clearly only evolved once in Asia.
In summary, this study shows that the molecular phylogenetic results are generally congruent with past morphology-based infrageneric classifications of Dioscorea. The resolution of the available phylogeny within Dioscorea was improved by adding information from the cpDNA trnL-F, matK, rbcL and atpB-rbcL combined datasets in our results. The low levels of molecular divergence within some clades (as measured by the short branch lengths) indicated that radiation might be relatively recent or at a slower rate. This hypothesis warrants further evaluation with a more extensive sample and even a higher resolution.