Bull. Natn . Sci. Mus., Tokyo, Ser. B, 28(4), pp. 129- 139, December 20, 2002

Molecular Phylogeny and Character Evolution of ()

1 2 3 Tomohisa Yukawa , Kazumitsu Miyoshi and Jun Yokoyama

1 Tsukuba Botanical Garden, National Science Museum , 1- 1, Amakubo 4, Tsukuba, 305- 0005 Japan E-mail : [email protected] 2 Department of Biological Production, Akita Prefectural University, Shimoshinjo-nakano, Akita 0 I 0- 01 95 Japan E-mail: mi [email protected] 3 Department of Ecology & Evolutionary Biology, Tohoku University, Aramaki, Aoba-ku, Sendai 980- 8578 Japan E-mail: [email protected]

Abstract Molecular phylogenetic analyses using two data sets, deri ved from DNA sequences of matK, the maturase-encoding gene located in an intron of the chloroplast gene trnK, and ITS (in­ ternal transcribed spacer) region of the 18S- 26S nuclear ribosomal DNA, were performed to ex­ amine relationships among 37 taxa in Cymbidium. Although each data set did not provide conclu­ sive evidence in itself, the data sets combining the two regions yi elded the following insights: (I) Cymbidium subgenus Cyperorchis and section Pachyrhizanthe, often treated as independent gen­ era, were nested within the remaining members of Cymbidium. (2) All the three subgenera of Cym­ bidium, namely, Cymbidium, Cyperorchis, and Jensoa, turned to be para-/polyphyletic. (3) Among sections of Cymbidium, Austrocymbidium, Cyperorchis, lridorchis, and Maxillarianthe did not show monophyly. (4) At specific level, our results indicated the polyphyly of Cymbidium bicolor, although bootstrap supports for the clades including C. bicolor were not robust. Key words: Cymbidium, ITS, matK, molecular, Orchidaceae, phylogeny

generic level , there have been inconsistencies on Introduction the delimitation. For example, Blume ( 1848, The genus Cymbidium Sw. comprises about 50 1849, 1858), Reichenbach ( 1852), Hooker species. Geographic distribution extends from ( 1890), and Schlechter ( 1924) treated Cyper­ the northwestern Himalaya to Japan and south orchis as an independent genus. Another exam­ through Indochina and Malesia to northern and ple is Pachyrhizanthe. As mentioned above, this eastern Australia. This group exhibits distinctive group shows very distinctive vegetative morphol­ ecological diversification and occurs as terrestrial ogy and nutritional habit, and Nakai ( 1931) thus and epiphytic life forms. Moreover, one section, elevated Pachyrhizanthe to generic rank. Pachyrhizanthe Schltr. lacks foliage leaves and The first major revision of Cymbidium was un­ has a strongly mycoparasitic existence. dertaken by Schlechter ( 1924) where he pro­ Swartz ( 1799) established the genus Cymbidi­ posed sectional limits within the genus. His um with a broad generic delimitation. Circum­ framework has been modified by Hunt ( 1970), scription of the genus has been variously defined and Seth & Cribb ( 1984). So far, the most exten­ by later workers: Lindley ( 1833), Blume ( 1848, sive infrageneric classification of Cymbidium is 1849, 1858), Reichenbach (1852, 1864), Hooker that of Du Puy & Cribb ( 1988), because their ( 1890), Schlechter ( 1924), Hunt ( 1970), Seth & system was based on the most recent information Cribb (1984), and Du Puy & Cribb (1988). At on morphological and anatomical characters. 130 Tomohisa Yukawa et al.

However, satisfying conclusions have not been neighbor-joining (NJ) method (Saitou & Nei, affirmed for the infrageneric relationships and 1987) with a Kimura two-parameter correction circumscriptions of several species in Cymbi­ (Kimura, 1980). To assess the relative robustness dium. for branches, the bootstrap method (Felsenstein, Recently, DNA sequences of matK, the mat­ 1985) was used with 1000 replicates. urase-encoding gene located in an intron of the chloroplast gene trnK, and ITS (internal tran­ Results scribed spacer region) of the 18S- 26S nuclear ri­ bosomal DNA have proven to be powerful tools Phylogenetic analyses for phylogenetic reconstruction within an­ Figure 1 shows a strict consensus of 28,800 giosperm families and genera (e.g., Baldwin, most parsimonious (MP) trees derived from the 1992; Johnson and Soltis, 1994; Steele and Yil­ matK sequences. The tree had a consistency galys, 1994). In this study we compared and index (Cl) of 0.7726 (0.6115 excluding uninfor­ combined DNA sequences of matK and ITS to mative characters) and a retention index (RI) of establish more concrete relationships in Cymbidi­ 0.8033. The NJ topology with more than 50% um. boots trap support was identical (Fig. 1). A strict consensus of 325 MP trees on the basis of ITS data set is shown in Fig. 2. Cl was 0.6848 Material and Methods (0.5606 excluding uninformative characters) and In this study we selected 37 representative taxa RI was 0.7790. In this data set the NJ topology from all the subgenera and sections in Cymbidi­ with more than 50% bootstrap support was iden­ um. Results of matK and ITS analyses for tribe tical (Fig. 2). On the other hand, analyses of have indicated that the sister group matK sequences provided results substantially of Cymbidium becomes the clade comprising concordant with those achieved via analyses of Acriopsis, Crammatophyllum, and Th ecostele ITS sequences except for positions of Cymbidi­ (Yukawa, unpublished). Consequently, these gen­ um erythrostylum and Cymbidium sanderae. In era were chosen as outgroup of this study. Mate­ such cases, analyses of combined data sets ex­ rials used in the phylogenetic analyses are shown cluding problematic taxa provide more resolution in Table 1. Voucher specimens are deposited at and internal support for relationships than do the TNS. individual data sets (e.g., Olmstead and Sweere, Experimental methods were described in 1995). We thus conducted a combined analysis of Yukawa et al. ( 1993, 1996). Sequences were de­ the information derived from matK and ITS se­ termined by amplifying the matK (including quences. In this analysis C. erythrostylum and C. parts of trnK introns) and ITS regions (including sanderae were excluded. The length of aligned 5.8S rDNA and parts of 18S and 26S rDNA) via sequence was 2829 base pairs. Figure 3 shows a the polymerase chain reaction (PCR) from a total strict consensus of 168 MP trees from the com­ DNA extract. All of the primers were from bined data set. Cl was 0.7283 (0.5931 excluding Yukawa et al. ( 1999) for matK and Douzery et al. uninformative characters) and RI was 0.7788. In ( 1999) for ITS. DNA sequences were aligned this data set the clades revealed by NJ analyses manually. Gaps were treated as missing charac­ with more than 50% bootstrap support was also ters. The aligned data file is available from the identical (Fig. 3). first author upon request. Parsimony and distance In the combined analysis, the following in­ analyses were conducted with PAUP* version sights were provided (Fig. 3): (I) Cymbidium 4.0b6 (Swofford, 200 I). The huristic option was subgenus Cyperorchis and section Pachyrhizan­ used to perform Fitch parsimony analyses (Fitch, th e, often treated as independent genera, were 1971 ). Distance trees were obtained using the nested within the remaining members of Cymbid- Molecular Phylogeny of Cymbidium 131

Table 1. Taxa of Cymbidium and outgroups used for sequencing.

Species Voucher

Subgenus Cymbidium Section Cymbidium Cymbidium a/oifolium (L.) Sw. TBGII8253* C. bicolor Lindl. subsp. obtusum Du Puy & P J. Cribb TBGI24328 C. bicolor Lindl. subsp. pubescens (Lindl.) Du Puy & P J. Cribb TBGII8509 C.finlaysonianum Lindl. TBGI22826 C. rectum Rid!. Leiden Bot. Gard. 913298 Section Hymantophyllum Schltr. C. dayanum Rchb. f. TBGI27899 Section Borneense Du Puy & P. J. Cribb C. borneense J. J. Wood Suzuki s.n. Section Austrocymbidium Schltr. C. cana/iculatum R. Br. TBGI35991 C. chloranthum Lindl. TBGI27438 C. hartinahianum J. B. Comber & Nasution Cibodas Bot. Gard. s.n. C. madidum Lindl. TBG119076 Section Floribundum Seth & P J. Cribb C..floribundum Lindl. TBG127357 C. suavissimum Sander ex C. Curtis TBG 128817 Section Bigibbarium Schltr. C. devonianum Paxton TBGII9073 Subgenus Cyperorchis (Biume) Seth & P J. Cribb Section lridorchis (Biume) P. Hunt C. e1y thraeum Lindl. TBGI35399 C. hookerianum Rchb. f. TBG 133783 C. insigne Rolfe TBG135526 C. iridioides D. Don TBG 118251 C. lowianum (Rchb. f.) Rchb. f. TBG54943 C. sanderae (Ro1fe) Du Puy & P J. Cribb Hiroshima Bot. Gard. 5138 C. tracyanum L. Castle TBGI24406 Section Eburnea Seth & P. J. Cribb C. eburneum Lindl. Suzuki s.n. C. roseum J. J. Sm. TBGI33724 Section Annamaea (Schltr.) P Hunt C. e1y throstylum Rolfe TBGI26650 Section Cyperorchis (Biume) P. Hunt C. elegans Lindl. Karasawa s.n. C. cochleare Lindl. Ohba s.n. Section Parishiella (Schltr.) P Hunt C. tigrinum Parish ex Hook. TBGI33781 Subgenus Jensoa (Raf.) Seth & P. J. Cribb Section Jensoa (Raf.) Schltr. C. ensifolium (L.) Sw. TBGII8255 C. kanran Makino TBGI33782 C. sinense (Jackson in Andr.) Willd. lwagawa s.n. Section Maxil/arianthe Schltr. C. aliciae Quisumb. TBGII9071 C.faberi Rolfe TBGI33593 C. goeringii (Rchb. f.) Rchb. f. TBGII5919 C. tortisepalum (Fukuyama) Y. S. Wu & S. C. Chen TBGI33595 Section Geocymbidium Schltr. C. lancifolium Hook. TBG56111 Section Pachyrhizanthe Schltr. C. aberrans (Finet) Sch1tr. Yagame s.n. C. macrorhizon Lindl. Yagame s.n. Acriopsis liliifo/ia (Koen.) Ormerod Kyoto Bot. Gard. s.n. Grammatophyllum scriptum (L.) Blume TBGII8874 G. 5peciosum Blume Shinjukugyoen s.n. G. stapeliiflorum (Teijsm. & Binn.) J. J. Sm. TBGI24396 Th ecostele alata (Roxb.) Par. & Rchb. f. TBGII8540

*TBG series indicate accession numbers in living collection database at Tsukuba Botanical Garden. 132 Tomohisa Yukawa et al.

Subgenus Section C.aloifolium Cymbidium Cymbidium C.finlaysonianum Cymbidium Cymbidium 100 C.bicolor subsp.pubescens Cymbidium Cymbidium 97 C.bicolor subsp.obtusum Cymbidium Cymbidium C.rectum Cymbidium Cymbidium

72 C.aliciae Jensoa Maxillarianthe 51 C.borneense Cymbidium Borneense 98 C.canaliculatum Cymbidium Austrocymbidium 92 C.madidum Cymbidium Austrocymbidium 90 C.dayanum Cymbidium Himantophyllum 84 C.erythrostylum Cyperorchis Annamaea C.cochleare Cyperorchis Cyperorchis C.elegans Cyperorchis Cyperorchis C.sanderae Cyperorchis lridorchis C.insigne Cyperorchis lridorchis C.erythraeum Cyperorchis lridorchis C.iridioides Cyperorchis lridorchis C.lowianum Cyperorchis lridorchis C.hookerianum Cyperorchis lridorchis 100 C.eburneum Cyperorchis Eburnea 100 C.roseum Cyperorchis Eburnea C.tracyanum Cyperorchis lridorchis C.tigrinum Cyperorchis Parishiella C.devonianum Cymbidium Bigibbarium C.hartinahianum Cymbidium Austrocymbidium 93 C.chloranthum Cymbidium Austrocymbidium 93 C.floribundum Cymbidium Floribundum C.suavissimum Cymbidium Floribundum C.goeringii Jensoa Maxillarianthe C.faberi Jensoa Maxillarianthe C.tortisepalum Jensoa Maxillarianthe C.ensifolium Jensoa Jensoa C.kanran Jensoa Jensoa C.sinense Jensoa Jensoa C.lancifolium Jensoa Geocymbidium C.macrorhizon Jensoa Pachyrhizanthe C.aberrans Jensoa Pachyrhizanthe Grammatophyllum speciosum Grammatophyllum scriptum Grammatophyllum stapeliaeflorum 100 99 Thecostele alata Acriopsis lilifolia

Fig. I. Strict consensus of 28,800 most-parsimonious Fitch trees based upon matK sequences: length =4 75, con- sistency index =0. 7726 (0.611 5 excluding uninformative characters), retention index of 0.8033. Numbers above internodes indicate bootstrap values from I ,000 replicates of Fitch parsimony analysis (maxtree limit of 200 per replicate). Numbers below intern odes indicate bootstrap va lues from I ,000 replicates of neighbor- joining distance analysis. Molec ular Phylogeny of Cymbidium 133

Subgenus Section

C.aloifolium Cymbidium Cymbidium C.finlaysonianum Cymbidium Cymbidium C.bicolor subsp.pubescens Cymbidium Cymbidium C.bicolor subsp.obtusum Cymbidium Cymbidium 66 52 C.rectum Cymbidium Cymbidium C.canaliculatum Cymbidium Austrocymbidium C.madidum Cymbidium Austrocymbidium 100 C.aliciae Jensoa Maxillarianthe 100 C.borneense Cymbidium Borneense C.dayanum Cymbidium Himantophyllum C.cochleare Cyperorchis Cyperorchis C.elegans Cyperorchis Cyperorchis C.insigne Cyperorchis lridorchis C.erythraeum Cyperorchis lridorchis C.eburneum Cyperorchis Eburnea C.roseum Cyperorchis Eburnea C.sanderae Cyperorchis lridorchis C.tracyanum Cyperorchis lridorchis C.erythrostylum Cyperorchis Annamaea C.iridioides Cyperorchis lridorchis 86 C.hookerianum Cyperorchis 89 lridorchis C.lowianum Cyperorchis lridorchis C.tigrinum Cyperorchis Parishiella C.devonianum Cymbidium Bigibbarium 17 C.goeringii Jensoa Maxillarianthe 16 C.tortisepalum Jensoa Maxillarianthe C.faberi Jensoa Maxillarianthe C.ensifolium Jensoa Jensoa C.kanran Jensoa Jensoa 39 59 C.sinense Jensoa Jensoa C.lancifolium Jensoa Geocymbidium C.macrorhizon Jensoa Pachyrhizanthe C.aberrans Jensoa Pachyrhizanthe C.hartinahianum Cymbidium Austrocymbidium 58 52 C.chloranthum Cymbidium Austrocymbidium C.floribundum Cymbidium Floribundum C.suavissimum Cymbidium Floribundum Grammatophyllum speciosum Grammatophyllum scriptum 100 Grammatophyllum stapeliaeflorum 100 Thecostele alata Acriopsis lilifolia

Fig. 2. Strict consensus of 325 most-parsimonious Fitch trees based upon ITS sequences: length=368, consis­ tency index=0.6848 (0.5606 excluding uninformative characters), retention index of0.7790. Numbers above intern odes indicate bootstrap values from I ,000 replicates of Fitch parsimony analysis (maxtree limit of 200 per replicate). Numbers below intern odes indicate boots trap values from I ,000 replicates of neighbor-joining distance analysis. 134 Tomohisa Yukawa er al.

Subgenus Section C.aloifolium Cymbidium Cymbidium C.finlaysonianum Cymbidium Cymbidium C.bicolor subsp.obtusum Cymbidium Cymbidium C.bicolor subsp.pubescens Cymbidium Cymbidium C.rectum Cymbidium Cymbidium

79 100 C.aliciae Jensoa Maxillarianthe 72 100 C.borneense Cymbidium Borneense

100 C.canaliculatum Cymbidium Austrocymbidium 100 C.madidum Cymbidium Austrocymbidium C.dayanum Cymbidium Himantophyllum C.cochleare Cyperorchis Cyperorchis C.elegans Cyperorchis Cyperorchis C.insigne Cyperorchis lridorchis C.erythraeum Cyperorchis lridorchis C.iridioides Cyperorchis lridorchis C.lowianum Cyperorchis lridorchis C.hookerianum Cyperorchis lridorchis 60 94 100 C.eburneum Cyperorchis Eburnea 100 C.roseum Cyperorchis Eburnea C.tracyanum Cyperorchis lridorchis C.tigrinum Cyperorchis Parishiella 59r------C.devonianum Cymbidium Bigibbarium 81 C.hartinahianum Cymbidium Austrocymbidium C.chloranthum Cymbidium Austrocymbidium C.floribundum Cymbidium Floribundum C.suavissimum Cymbidium Floribundum C.goeringii Jensoa Maxillarianthe 82 65 C.tortisepalum Jensoa Maxillarianthe C.faberi Jensoa Maxillarianthe C.ensifolium Jensoa Jensoa C.kanran Jensoa Jensoa C.sinense Jensoa Jensoa C.lancifolium Jensoa Geocymbidium C.macrorhizon Jensoa Pachyrhizanthe C.aberrans Jensoa Pachyrhizanthe Grammatophyllum speciosum Grammatophyllum scriptum Grammatophyllum stapeliaeflorum 100 100 Thecostele alata Acriopsis lilifolia

Fig. 3. Strict consensus of 168 most-parsimonious Fitch trees based upon matK and ITS sequences: length= 828, consistency index =0. 7283 (0.5731 excluding uninformative characters), retention index of 0. 7788. Numbers above internodes indicate bootstrap va lues from I ,000 replicates of Fitch parsimony analy­ sis (maxtree limit of I ,000 per replicate). Numbers below internodes indicate bootstrap values from 1,000 replicates of neighbor-joining distance analysis. Molecular Phylogeny of Cymbidium 135 ium . (2) All the three subgenera of Cymbidium, monophyletic group, if Cymbidium aliciae is re­ namely, Cymbidium, Cyperorchis, and Jensoa, moved. The four pollinia is a principal synapo­ turned to be para-/polyphyletic. (3) Among sec­ morphic character of subgenus Jensoa, but C. tions of Cymbidium, Austrocymbidium, Cyper­ aliciae and Cymbidium borneense, distantly re­ orchis, lridorchis, and Maxillarianthe did not lated to subgenus Jensoa, also share this charac­ show monophyly. ( 4) At specific level, our results ter. Taking the phylogenetic relationships and the indicated the polyphyly of Cymbidium bicolor, morphological characters into consideration, sub­ although bootstrap supports for the clades in­ divisions of Cymbidium at subgeneric level are cluding C. bicolor were not robust. not useful. In other words, we did not find any stable synapomorphic characters that feature each major clade clarified in this study. Discussion Our results further indicated that sectional de­ Classification limitation of Cymbidium also has several prob­ Generic circumscription of Cymbidium by lems. Among members of subgenus Cymbidium, modern taxonomists was supported in this study. section Austrocymbidium exhibits paraphyly and Cyperorchis, sometimes treated as an indepen­ comprises the following three clades: (I) Cym­ dent genus (e.g. Blume, 1848, 1849, 1858; Re­ bidium canaliculatum and Cymbidium madidum; ichenbach, 1852; Hooker, 1890; and Schlechter, (2) Cymbidium hartinahianum; (3) Cymbidium 1924) was nested within other clades of Cymbidi­ chloranthum. The first clade includes the type of um. Pachyrhizanthe, another problematic group this section (C. canaliculatum) and corresponds with reduced vegetative morphology and strongly to a revised circumscription of section Austro­ mycoparasitic nutritional existence, was elevated cymbidium. Since the remaining two clades fur­ to a new genus by Nakai ( 1931 ). The results of ther form another major clade together with sec­ thi s study showed that Pachyrhizanthe makes a tion Floribundum and morphological characters sister group relationship with section Geocym­ of the two cl ades are consistent with those of sec­ bidium and that it is reasonable to treat this group tion Floribundum, we extend the limits of this as a section of Cymbidium. Moreover, these two section to include members of the two clades, sections have the following synapomorphies: (I) namely, C. hartinahianum and C. chloranthum. underground rhizomes that survive throughout In subgenus Cyperorchis, the monophyly of their life history, (2) whitish perianth lobes with the clade including sections Cyperorchis, purple striations and markings. Eburnea, and lridorchis was supported by high We clarified that all the three subgenera of bootstrap values. However, support on several re­ Cymbidium proposed by Seth & Cribb ( 1984) lationships among members of sections Cyper­ represent para-/polyphyly. Subgenus Cymbidium orchis and lridorchis was weak due to few comprises four clades and these clades become synapomorphic nucleotide substitutions for these polyphyletic assemblage. It is apparent that the groups. It is likely that section lridorchis was es­ subgenus was defined by plesiomorphic charac­ tablished on the basis of plesiomorphic charac­ ters such as a lip without any fusing structures ters such as widely-opened perianth lobes and an and two pollinia. The monophyly of subgenus operculum without an obvious backwards-point­ Cyperorchis was broken by a nested position of ing beak, but explicit rel ationships on the basis subgenus Cymbidium section Himantophyllum. of more molecular signals are needed to re­ Transfer of section Himantophyllum to a member arrange the classification for these taxa. of subgenus Cyperorchis results in the loss of a As mentioned above, Cymbidium aliciae broke prominent synapomorphic character of subgenus the monophyly of subgenus Jensoa, more specifi­ Cyperorchis, that is, a fused basal part of the lip cally section Maxillarianthe. Du Puy & Cribb with the . Subgenus Jensoa will become a ( 1988) placed C. aliciae into subgenus Jensoa 136 Tomohisa Yukawa et al. section Maxillarianthe as a synonym of Cymbidi­ Pollinia number um cyperifolium Wall. ex Lindl. Our results did The most controversial morphological charac­ not support this placement and suggested a sister ter of Cymbidium is different numbers of pollinia group relationship of C. aliciae to Cymbidium (two or four) in the genus. Most members of borneense, a sole representative of subgenus Cymbidium have two pollinia, but subgenus Jen­ Cymbidium section Borneense. Observations soa and subgenus Cymbidium section Borneense of morphological and anatomical characters exhibit four pollinia. On the other hand, the sister (Yukawa & Stern, 2002; Yukawa, unpublished) group of Cymbidium (A criopsis, Grammatophyl­ of C. aliciae also supported the affinity of these lum, and Th ecostele) and the sister group to these two species and a distant relationship of section taxa (Bromh eadia: Yukawa, unpublished) share Maxillarianthe to them. For example, C. aliciae two pollinia. Since the pollinia number of Orchi­ and C. borneense have hypodermal fiber strands daceae is usually considered to be stable and has both in abaxial and adaxial surfaces, but mem­ been treated as a cardinal character among taxon­ bers of section Maxillarianthe lack this character omists, variations in this genus have provoked at least in abaxial surface. Du Puy & Cribb much discussion (e.g. Du Puy & Cribb, 1988). ( 1988) established section Borneense, because C. The results of this study showed that the four borneense shows mixed diagnostic characters of pollinia are an apomorphic character in Cymbidi­ subgenus Cymbidium section Cymbidium and um and likely evolved twice on the lineages lead­ subgenus Jensoa. The present results corroborat­ ing to subgenus Jensoa and subgenus Cymbidium ed the soundness of section Borneense and the section Borneense (both in ACCTRAN and two species, C. borneense and C. aliciae, consti­ DELTRAN optimizations). Our results also indi­ tute the section. cated that the conservative nature of pollinia Relationships at species level also may include number is not a rule, but it is a kind of "dogma" problems. Our results indicated the polyphyly of in orchid classification. Actually, variations of Cymbidium bicolor, although bootstrap supports pollinia numbers are found in other small- to for the clades including C. bicolor were not ro­ middle-sized monophyletic groups such as Thu­ bust. On the other hand, a species complex in­ nia (e.g. Seidenfaden, 1986) and Phalaenopsis cluding Cymbidium ensifolium, Cymbidium kan­ (e.g. Christenson, 200 I). ran , and Cymbidium sinense has low resolution. Accumulation of more molecular signals is need­ Life form ed to clarify these relationships. As mentioned before, Cymbidium exhibits a On the other hand, we found discrepant posi­ great variation of life forms. In this study we de­ tions for Cymbidium erythrostylum and Cymbidi­ fine epiphytes in a broad sense, that is, a group of um sanderae between matK and ITS trees. Since species inhabiting trees and/or rocks. In Cymbid­ maternal inheritance of plastid DNA was con­ ium, many species grow on trees and some of firmed in Orchidaceae (Corriveau & Coleman, them also have potentials to live on rocks. How­ 1988; Sears, 1980), inconsistent phylogenetic po­ ever, accurate figures on habitats of each species sitions for these two species in matK trees can be are not known due to the lack of observations in caused by introgressions. Interestingly, collection situ. Besides, there are no Cymbidium species records of C. erythrostylum and C. sanderae are that grow exclusively on rocks. We thus use a very few, and this fact also may account for their broad concept of epiphyte. hybrid origin. Moreover, Du Puy and Cribb The sister group taxa of Cymbidium (A cri­ ( 1988) also indicated a hybrid nature of C. opsis , Grammatophyllum, and Thecostele) are sanderae. Further studies by use of more markers epiphytes and the sister group to these taxa both from nuclear and plastid genomes can con­ (Bromheadia : Yukawa, unpublished) comprises firm this hypothesis. both epiphytic and terrestrial species. The most Molecular Phylogeny of Cymbidium 137

Life form .---- C.aloifolium epiphyte C.finlaysonianum epiphyte C.bicolor subsp.obtusum epiphyte C.bicolor subsp.pubescens epiphyte C.rectum epiphyte C.aliciae terrestrial C.borneense terrestrial C.canaliculatum epiphyte C.madidum epiphyte .------C.dayanum epiphyte C.cochleare epiphyte C.elegans epiphyte '--lt-- C.insigne terrestrial c______C.erythraeum epiphyte C.iridioides epiphyte C.lowianum epiphyte '---- C.hookerianum epiphyte ....______C.tracyanum epiphyte C.eburneum epiphyte C.roseum epiphyte '------C.tigrinum epiphyte .------C.devonianum epiphyte .----- C.hartinahianum terrestrial .---- C.chloranthum epiphyte C.floribundum epiphyte C.suavissimum epiphyte C.goeringii terrestrial C.tortisepalum terrestrial '---- C.faberi terrestrial C.ensifolium terrestrial '----+-- C.kanran terrestrial C.sinense terrestrial .---- C.lancifolium terrestrial C.macrorhizon terrestrial C.aberrans terrestrial Grammatophyllum speciosum epiphyte Grammatophyllum scriptum epiphyte Grammatophyllum stapeliaeflorum epiphyte Thecostele alata epiphyte .______Acriopsis lilifolia epiphyte

Fi g. 4. Reconstruction of diversification in life form in Cymbidium under ACCTRAN optimization. The tree is one of 168 most-parsimonious Fitch trees based upon matK and ITS sequences. Black bars = change of states from epiphytic to terrestrial habitat; grey bars= change of states from terrestrial to epiphytic habitat. 138 Tomohisa Yukawa et al. parsimonious reconstruction of character evolu­ of rhizomes and inflorescences represents a case tion on life forms indicated changes of character of heterochrony. states at internodes shown by bars (Fig. 4). The Our molecular data further showed low genetic ancestor of Cymbidium was likely an epiphyte. divergence between the mycotrophic clade and Under ACCTRAN optimization, the epiphytic its autotrophic sister clade, namely, Cymbidium Cymbidium evolved terrestrial three times; lancifolium (section Geocymbidium). The num­ moreover, a single reversal from terrestrial to epi­ ber of base differences and sequence divergences phytic condition likely occurred in Cymbidium. within the ITS region was as follows: Cymbidium Under DELTRAN optimization, the epiphytic an­ macrorhizon-C. lancifolium: 9, 0.0118; Cymbidi­ cestor of Cymbidium evolved into terrestrial um aberrans-C. lancifolium : 9, 0.0 I 06. This fact plants four times, that is, (I) in the common an­ indicated that these apparently drastic shifts of cestor of Cymbidium section Borneense, (2) in nutritional habit and morphological characters the ancestor of Cymbidium insigne, (3) in the an­ were established within relatively short period of cestor of Cymbidium hartinahianum, (4) in the time. ancestor of Cymbidium subgenus Jensoa (results not shown in figures). Acknowledgements Recent phylogenetic analyses clarified that Or­ chidaceae had a terrestrial ancestry (Neyland and would like to thank Dedy Darnaedi, Ed de Urbatsch, 1995 ; Cameron et al., 1999). There­ Vogel, Genjiro lshida, Fumihi ro lwagawa, Kohji fore, evolution from epiphytic to terrestrial habits Karasawa, Jun' ichi Nagasawa, and Takahiro in Cymbidium represents reversal in Orchidaceae Yagame for providing the material, and as a whole. These results indicated that the Osamu Miikeda, Koichi Kita, and Hideaki changes of habitats appeared repeatedly in the Shimizu for technical assistance. This study is lineage of Cymbidium and are not rare evolution­ partly supported by a Grant-in-Aid to Scientific ary events in Orchidaceae. Research from the Japan Society for the Promo­ tion of Science (13640708). Nutritional habit Cymbidium section Pachyrhizanthe Schltr. References keeps strong mycotrophy throughout life history and lacks foliage leaves as well as roots. Albeit Bald win, B. G., 1992. Phylogenetic utility of the internal transcribed spacers of nuclear ribosomal DNA in Nakai (1931 )'s suggestion on its separate taxo­ plants: an example from the Compositae. Mol. Phyla. nomic status from Cymbidium, we found that this Evol., 1:3- 16. group belongs to Cymbidium. The present results Blume, C. L., 1848. Orchideae. Rumphia, 4: 38- 56. also showed that the state, mycotrophy through­ Blume, C. L., 1849. Cyperorchis. Ann. Mus. Bot. Lug­ out li fe history, evolved only once within a clade duno-Batavum , 1: 48. Blume, C. L., 1858. Collection des Orchidees les plus re­ of terrestrial Cymbidium species. All the terres­ marquables de I' Archipel lndien et du Japan. 190 pp. trial species of Cymbidium have obli gate my­ C. G. Sulpke, Amsterdam. cotrophy in their juvenile stage of life history. Cameron, K. M. , M. W. Chase, M. W. Whitten, P J. During this stage, the rhizome only represents Kores, D. C. Jarrell, V A. Albert, T. Yukawa, H. G. the body and functions as the home of symbiotic Hills & D. H. Goldman, 1999. A phylogenetic analysis fungi. Undoubtedly, the existence of obligate my­ of the Orchidaceae: Evidence from rbcL nucleotide se­ quences. Amer. J Bot. , 86: 208- 224. cotrophy during juvenile stage in terrestrial Chri stenson, E. A. , 200 I. Phalaenopsis. A Monograph. species of Cymbidium acted as preadaptation of 330 pp. Timber Press, Portland, USA. complete mycotrophy throughout life history in Corriveau, J. L. & A. W. Coleman, 1988. Rapid screening section Pachyrhizanthe. Furthermore, the growth method to detect potential biparental inheritance of habit of section Pachyrhizanth e only composed plastid DNA and results for over 200 angiospenn Molecul ar Phylogeny of Cymbidium 139

species. Amet: J Bot., 75: 1443- 1458. Saitou, H. & M. Nei, 1987. The neighbor-joining method: Douzery, E. J. P, A. M. Pridgeon, P Kores, H. P Linder, a new method for reconstructing phylogeneti c trees. H. Kurzweil & M. W. Chase, 1999. Molecular phyloge­ Mol. Bioi. Evol., 4 : 406-425. neti cs of Diseae (Orchidaceae): A contribution from Schlechter, R., 1924. Die Gattungen Cymbidium Sw. und nuclear ribosomal ITS sequences. Amer J Bot., 86: Cyperorchis Bl. Feddes Repert. Spec. Nov. Regni Veg., 887- 899. 20 : 96-- 11 0. Du Puy, D. & P. J. Cribb, 1988. The Genus Cymbidium. Sears, B. B., 1980. Elimination of pl astids during sper­ 236 pp. Timber Press, Portland, USA. matogenesis and fertilization in the plant kingdom. Fe lsenstein , J. , 1985. Confi dence limits on phylogenies: Plasmid, 4: 23 3- 255. an approach using the bootstrap. Evolution, 39: 783- Seidenfaden, G., 1986. Orchid genera in Thail and XIII. 791. Thirty-three epidendroid genera. Opera Bot., 89: 1- Fitch, W. M., 197 1. Toward defining the course of evolu­ 2 16. tion: minimum change fo r a specifi c tree topology. Syst. Seth, C. J. & P. J. Cribb, 1984. A reassessment of the sec­ Zoo/., 20 : 406-416. tional limits in the genus Cymbidium. In J. Arditti (ed.), Hooker, J. D., 1890. Cymbidium and Cyperorchis. Flora of Orchid Bi ology, Reviews and Perspectives, Ill, pp. British 6. pp. 8-15. Reeve & Co., London. 283-332. Cornell Uni versity Press, lthaca, USA. Hunt, P. F.. 1970. Notes on As iati c orchids 5. Kew Bull., Steele, K. P. & R. Yil galys, 1994. Phylogeneti c analyses 24 : 93- 94. of Polemoni aceae using nucleotide sequences of th e Johnson, L. A. & D. E. Soltis, 1994. marK DNA se­ pl astid gene marK. Syst. Bot. , 19 : 126- 142. quences and phylogeneti c reconstruction in Saxifra­ Swartz, 0 ., 1799. Dianome Epidendri generi s Linn. Nova gaccaes. s/J : Syst. Bot., 19 : 143-156. Acta Regiae Soc. Sci. Upsal., 6: 6 1- 88. Kimura, M., 1980. A simple method fo r estimating evolu­ Swofford, D. L. , 200 I. PAU P*: phylogeneti c analysis ti onary rates of base substitutions th rough comparati ve using parsimony (* and other meth ods), version 4.0b6. studies of nucleotide sequences. J Mo l. Evol .. 16: 111 - Sin auer, Sunderl and, Massachusetts, USA. 120. Yukawa, T. , S. Kurita, M. Nishida & M. Hasebe, 1993. Lindley, J. , 1833. Cymbidium. The Genera and Species of Phylogeneti c implicati ons of chloropl ast DNA restric­ Orchidaceous Pl ants, pp. 16 1- 172. Ridgways, London. ti on site vari ati on in subtribe Dendrobiinae (Orchi­ Nakai. T. , 193 1. Notulae ad plantas Japoniae & Koreae daceae). Lindlayana, 8: 2 11 - 22 1. XL. Bot. Mag. (Tof..yo), 45:91- 137, 146-- 160. Yukawa, T. , H. Ohba, K. M. Cameron & M. W. Chase, Ncyland, R. & L. E. Urbatsch, 1995. A terrestrial ori gin 1996. Chl oroplast DNA phylogeny of subtribe Dendro­ for th e Orchidaceae suggested by a phylogeny in fe rred biinae (Orchidaceae): in sights from a combined analy­ from ndhF chl oroplast gene sequences. Lindleyana, 10: sis based on rbcL sequences and restri ction site varia­ 244- 25 1. tion. J Plant Res., I 09 : 169- 176. O lm stead, G. R. & J. A. Sweere, 1995. Combined data in Yukawa, T., S. Koga, & T. Handa, 1999. DNA un covers phylogeneti c systematics: an empirical approach using paraphyly of Dendrobium (Orchidaceae). In S. An­ th ree molec ul ar data sets in the Solanaceae. Syst. Bioi., drews, A. G. Leslie & C. Alexander (eds.), 43 : 467-48 1. of Cultivated Pl ants, pp. 35 1-354. Royal Botani c Gar­ Re ichenbach H. G., 1852. Orchidaceae (Cymbidium and dens, Kew. Cyperorchis). Ann. Bot. Svsr., 3: 547- 548. Yuk awa, T. & W. L. Stern, 2002. Comparative vegetati ve Reichenbach H. G., 1864. Orchidaceae (Cymbidium). anatomy and systemati cs of Cymbidium (Cymbidieae: Ann. Bot. Syst., 6: 622- 626. Orchidaceae). Bot. J Lin n. Soc., 138: 383-419.