Systematic Botany (2005), 30(4): pp. 736–749 ᭧ Copyright 2005 by the American Society of Plant Taxonomists A Plastid Gene Phylogeny of the Yam Genus, Dioscorea: Roots, Fruits and Madagascar PAUL WILKIN,1,6 PETER SCHOLS,2 MARK W. C HASE,1 KONGKANDA CHAYAMARIT,3 CAROL A. FURNESS,1 SUZY HUYSMANS,2 FRANCK RAKOTONASOLO,4 ERIK SMETS,2 and CHIRDSAK THAPYAI5 1Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AB, U.K.; 2Laboratory of Plant Systematics, Institute of Botany and Microbiology, K.U.Leuven, Kasteelpark Arenberg 31, B-3001 Leuven, Belgium; 3Forest Herbarium, 61 Phahonyothin Rd., Chatuchak, Bangkok 10900 Thailand; 4Parc Botanique et Zoologique de Tsimbazaza, BP 4096, Antananarivo, Madagascar; 5Biology Department, Faculty of Science, Naresuan University, Muang District, Phitsanulok 65000, Thailand 6Author for correspondence ([email protected]) Communicating Editor: Alan W. Meerow ABSTRACT. Following recent phylogenetic studies of the families and genera of Dioscoreales, the identification of mono- phyletic infrageneric taxa in the pantropical genus Dioscorea is a priority. A phylogenetic analysis based on sequence data from the plastid genes rbcL and matK is presented, using 67 species of Dioscorea and covering all the main Old World and selected New World lineages. The analysis used 14 outgroup taxa, including Trichopus Gaertn., Tacca J.R. & G. Forster, Stenomeris Planch., Burmannia L. and Thismia Griff. The main findings are: a) that a clade of rhizomatous taxa is sister to the rest of Dioscorea; b) the main Old World groups (such as the right-twining D. sect. Enantiophyllum) are monophyletic and c) there are two distinct lineages among the endemic Malagasy taxa. The consequences of the results for infrageneric classification of Dioscorea is considered, in particular the possibility of greatly simplifying the classifications of Knuth and Burkill. The results are also used to present novel hypotheses of character evolution in selected underground storage organ, inflorescence, fruit and seed characters and to discuss the origins of diversity in Dioscorea. Dioscoreales are one of the most critical taxa in trilocular ovary with two ovules per locule like Tricho- monocot systematics (Chase 2004 and references there- pus, but it lacks the complex arrangement of expanded in). Since the revolution in that field stimulated by the stamen connectives and the ‘‘umbrella-like’’ stigma of works of Huber (1969) and Dahlgren et al. (1985), the that genus. With the generic and familial limits now composition and relationships of the order have been better understood, the main systematic challenge in the subject of an increasing level of study. In particu- terms of biodiversity in Dioscoreales is Dioscorea. Dios- lar, the recent research of Caddick et al. (2002a; 2002b), corea is also by far the most geographically widespread using morphological and sequence data (from three taxon, being almost ubiquitous in tropical and sub- genes), has demonstrated conclusively that the order tropical regions, with a few species being found in comprises three families: Nartheciaceae, Burmanni- temperate areas. Species of Dioscorea are of significant aceae, and Dioscoreaceae. All are relatively small fam- importance as food (mainly in the form of their starchy ilies (Kubitzki 1998). The largest of these in terms of tubers or ‘‘yams’’) and pharmaceuticals (e.g., cortico- number of species, Dioscoreaceae, comprise four gen- steroids and the contraceptive pill; Coursey 1967). era (Caddick 2002a, 2002b). Three are monoecious and Dioscorea has presented a challenge to systematists contain relatively few species. Tacca J.R. & G. Forst. (at for many years due to its great morphological diver- least ten species) has a unilocular ovary with many sity, dioecy, and small flowers. The first taxonomic ovules; unlike most taxa in the family it does not climb treatments of significant numbers of species were but is a herb with a short stem. Stenomeris Planch. (two those of Kunth (1850) and Uline (1898). The last com- species) is a climber with a trilocular ovary with many plete monograph was published by Knuth (1924). Us- ovules; it forms a three-winged, dehiscent capsule at ing a typically narrow ‘‘Pflanzenreich’’ species concept, least 25 cm long. Its flowers have an urceolate torus, he recognized ca. 600 species and divided them into with the stamens inserted towards the torus mouth four subgenera based on seed wing position, and then and reflexed into it. The third monoecious genus, Tri- into 60 sections. However, when studied by a contem- chopus Gaertn. (two species), possesses an ovary with porary systematist, it is clear that many of Knuth’s in- two ovules in each of three locules, with up to five frageneric taxa are clearly para- or even polyphyletic, ovules aborting during the development of an irregu- for example the Old World compound-leafed species larly dehiscent to indehiscent fruit. Its hypanthium is studied by Wilkin and Caddick (2000) and Wilkin relatively small and not urceolate. Dioscorea L., the only (1999). Knuth divided these taxa into three sections dioecious genus, comprises ca. 450 species and has a and 51 species based on geography. The research cited 736 2005] WILKIN ET AL.: PHYLOGENY OF DIOSCOREA 737 above showed that there are approximately 18 species sist plant breeders, pathologists, phytochemists, and that can be be divided into two groups with clear mor- other applied biologists. A partial dataset for plastid phological differences between them. Knuth even man- rbcL already existed from those taxa sampled by Cad- aged to place a single species in different sections un- dick et al. (2002b), and matK was selected to provide der two different names (D. cochleari-apiculata De Wild. further resolution following its successful use in other in sect. Botryosicyos (Hochst.) Uline and the synony- studies of monocot taxa (e.g., Fuse and Tamura 2000; mous D. stolzii R. Knuth. in sect. Lasiophyton Uline). Gravendeel et al. 2001; Ge et al. 2002). The main sys- Therefore, a reappraisal of Knuth’s classification is tematic hypotheses to be evaluated were the mono- needed. phyly of Dioscorea and Knuth and Burkill’s sections of The taxonomic ideas of Knuth were to some extent Dioscorea, towards the aim of constructing a classifi- refined and improved by Burkill (Prain and Burkill cation based on the principle of monophyly. 1936, 1938; Burkill 1939, 1951, 1952, 1960; Burkill and The generation of a phylogenetic tree for Dioscorea Perrier de la Baˆthie 1950). He had a thorough knowl- based on sequence data, and therefore independent of edge of the tropical African and Asian species of Dios- morphology, also allowed morphological character corea and based his classification on an intimate knowl- evolution to be explored. Similar studies have been un- edge of their morphology and ecology derived from dertaken, by, for example Cameron et al. (2001) for pol- many years of study of herbarium specimens and liv- len morphology in Malphigiaceae and Allen et al. ing plants. However, his knowledge of species from (2003) and Lamb Frye and Kron (2003) for selected China, Madagascar, and the New World was restricted macromorphological characters in Erythronium L. and to herbarium material. In his infrageneric classification Polygonaceae, respectively. Tuber morphology, stem of the Old World taxa, Burkill avoided the rank of sub- twining direction, dioecy, and fruit/seed wing shape genus, and instead divided some 220 species into 23 are among the most important characters in the sys- sections. Like Knuth, he emphasized seed characters, tematics of Dioscorea. A broadly sampled tree also al- but he added underground organ morphology and de- lows biogeographical hypotheses to be evaluated, such velopment and male inflorescence morphology to the as Burkill’s (1960) suggestion that there are two inde- character set used. This resulted in a classification pendent lineages in Madagascar; a group with deeply more complex than that of Knuth, but one with a ten- buried tubers and seeds winged at the base only and dency to separate taxa using what now appear to be another with shallowly buried tubers at the end of long autapomorphies. Since 1960, the genus has been the roots and seeds winged all round the margin. Overall, subject of piecemeal floristic studies (e.g., Mie`ge 1968; this study represents the first step in understanding Milne-Redhead 1975; Tellez and Schubert 1994; the diversity of Dioscorea in a phylogenetic context. N’Kounkou 1993; Mie`ge and Sebsebe 1998; Ding and Gilbert 2000). The only complete taxonomic treatment MATERIALS AND METHODS was that of Huber (1998), in which the Knuth/Burkill Taxon Sampling. This study sampled 50 paleotropical species system of classification was recapitulated, with all of of Dioscorea, covering all of the main lineages evident from the the dioecious taxa of Dioscoreaceae included in sub- infrageneric classifications of Knuth (1924) and Prain and Burkill family Dioscoreoideae as ‘‘genera and genus-equiva- (1960) and from the authors’ own study of morphology (Table 1). lent sections’’ (Huber’s terminology). Although this Material from Thailand was particularly important in achieving this aim; the Thai flora contains almost all of the Asian species idea was novel, the flaws inherrent in the treatments groups. There is also a sampling bias towards Madagascar. Both of Knuth and Burkill meant that many of the taxa used of these areas have high numbers of Dioscorea species per unit 2 were not systematically sound, and the decision to area: approximately one