The Phylogeny and a New Classification of the Gingers (Zingiberaceae): Evidence from Molecular Data1

Home , Aframomum, Burbidgea, Gagnepainia, Geocharis (plant), Geostachys, Hitchenia

American Journal of Botany 89(11): 1682±1696. 2002.

THE PHYLOGENY AND A NEW CLASSIFICATION OF THE GINGERS (ZINGIBERACEAE): EVIDENCE FROM MOLECULAR DATA1

W. J OHN KRESS,2,5 LINDA M. PRINCE,2,4 AND KYLE J. WILLIAMS3

2Botany, MRC-166, United States National Herbarium, National Museum of Natural History, Smithsonian Institution, P.O. Box 37012, Washington, D.C. 20013-7012 USA; 3Department of Biology, Duke University, Durham, North Carolina 27708 USA

The pantropical Zingiberaceae is the largest family in the order Zingiberales with 53 genera and over 1200 species. Classi®cations of the family ®rst proposed in 1889 and re®ned by others since that time recognize four tribes (Globbeae, Hedychieae, Alpinieae, and Zingibereae) based on morphological features, such as number of locules and placentation in the ovary, development of staminodia, modi®cations of the fertile anther, and rhizome-shoot-leaf orientation. New phylogenetic analyses based on DNA sequences of the nuclear internal transcribed spacer (ITS) and plastid matK regions suggest that at least some of these morphological traits are homo- plasious and three of the tribes are paraphyletic. The African genus Siphonochilus and Bornean genus Tamijia are basal clades. The former Alpinieae and Hedychieae for the most part are monophyletic taxa with the Globbeae and Zingibereae included within the latter. The results of these phylogenetic investigations are used to propose a new classi®cation of the Zingiberaceae that recognizes four subfamilies and four tribes: Siphonochiloideae (Siphonochileae), Tamijioideae (Tamijieae), Alpinioideae (Alpinieae, Riedelieae), and Zingiberoideae (Zingibereae, Globbeae). Morphological features congruent with this classi®cation and the taxonomic status of various monotypic genera are discussed.

Key words: classi®cation; gingers; ITS; matK; phylogeny; tropical; Zingiberaceae; Zingiberales.

The primarily tropical Zingiberales are phylogenetically em- Nagamasu, 2000; Kress and Larsen, 2001; Larsen and Jenjit- bedded within the derived eumonocots (Arecales, Commelin- tikul, 2001) newly described in the last several years. ales, Poales) and include many conspicuous taxa, such as the The currently accepted classi®cation of the Zingiberaceae bananas (Musaceae), birds of paradise (Strelitziaceae), heli- (Petersen, 1889; Schumann, 1904; Holttum, 1950; Burtt and conias (Heliconiaceae), and gingers (Zingiberaceae). With 53 Smith, 1972; Larsen et al., 1998) includes four tribes (Hedy- genera and over 1200 species, the Zingiberaceae is the largest chieae: 22 genera, Figs. 2±3; Alpinieae: 25 genera, Fig. 4; of the eight families of the order (Kress, 1990). In earlier clas- Zingibereae: one genus, Fig. 5; and Globbeae: four genera, si®cations (e.g., Petersen, 1889; Schumann, 1904) the family Fig. 6) and is based on both vegetative and ¯oral character- Costaceae (Fig. 1) was included in the Zingiberaceae, but with istics (Table 1). Although a number of morphological features a number of distinctive characters (e.g., lack of aromatic oils, have been used to distinguish the four tribes, in most cases branched aerial stems, and spiral monostichous phyllotaxy; the de®ning characters are either not unique to any one tribe Specht et al., 2001) it is now accepted as the sister clade to or are not universal for all taxa within any tribe (Table 1). For the gingers (Kress, 1990, 1995; Kress et al., 2001). The Zin- example, the plane of distichy of the leaves is perpendicular giberaceae is distributed pantropically with one genus (Re- to the rhizome in the Alpinieae (except in the genera Rhyn- nealmia) found in the Neotropics, four genera (Aframomum, chanthus and Pommereschea) and is parallel to the rhizome in Aulotandra, Siphonochilus, and Renealmia) found in Africa, the other three tribes (except for Siliquamomum in the Hedy- and the rest of the genera distributed in east Asia and the chieae). Similarly, the ovary is unilocular in all members of Paci®c Islands. The family is still poorly known taxonomically the Globbeae, but trilocular in the other three tribes with some with many species (Theilade and Mood, 1997; Sakai and Na- exceptions, e.g., unilocular in Tamijia and partially unilocular gamasu, 1998; Poulsen, Mood, and Ibrahim, 1999; Williams, in Caulokaempferia, Paracautleya, Scaphochlamys, Siphono- Kress, and Thet Tun, 2002) and even genera (Newman, 1995; chilus, and Riedelia (Larsen et al., 1998; Sakai and Nagamasu, Mood and Larsen, 1997; Larsen and Mood, 1998; Sakai and 2000). The lateral staminodes are well developed in Hedy- chieae, Globbeae, and Zingibereae (Figs. 2±3, 5±7) and gen- 1 Manuscript received 15 February 2002; revision accepted 3 May 2002. The authors thank Ray Baker, Josef Bogner, Mike Bordelon, Alan Carle, erally absent in the Alpinieae (Fig. 4; but all ®ve staminodes Mark Collins, David Harris, Yin Yin Kyi, Kai Larsen, Qing-Jun Li, Jing-Ping are absent in Rhynchanthus). The fusion of the lateral stami- Liao, Ida Lopez, Paul Manos, John Mood, Mark Newman, Chatchai Ngam- nodes to the central labellum (Fig. 7) in Siphonochilus (He- riabsakul, Dan Nicolson, David Orr, Louise Pedersen, Kai Rangsiruji, Shoko dychieae; Fig. 2) and Zingiber (Zingibereae; Fig. 5) has now Sakai, Henk van der Werf, Tom Wood, Qi-Gen Wu, and Yong-Mei Xia for been reported in the narrow endemic Tamijia (Alpinieae) from discussion, assistance, and tissue samples that made this investigation possible. The comments of two anonymous reviewers signi®cantly improved the Borneo (Sakai and Nagamasu, 2000). Most of the characters manuscript. This work was funded by the Smithsonian Scholarly Studies Pro- that are used to de®ne the tribes are often inconsistent and gram, the Biotic Surveys and Inventories Program of the National Museum variable. of Natural History, and the A.W. Mellon Training Grant for Plant Systematics The circumscription of various genera of the Zingiberaceae to Duke University. has also been debated by investigators (e.g., Etlingera: Burtt 4 Current address: Rancho Santa Ana Botanic Garden, 1500 North College Avenue, Claremont, California 91711-3157 USA. and Smith, 1986; Curcumorpha: Larsen, 1997; Paramomum: 5 Author for reprint requests (phone: 202-357-2534; FAX: 202-786-2563; Wu, 1997), and a consensus has not been reached on the actual e-mail: [email protected]). number of genera in the family. Nearly one-quarter of the gen- 1682 October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1683

Figs. 1±6. Representatives of the major groups of the family Zingiberaceae and the sister family Costaceae. 1. Costaceae (Monocostus). 2. Hedychieae (Siphonochilus). 3. Hedychieae (Hedychium). 4. Alpinieae (Alpinia). 5. Zingibereae (Zingiber). 6. Globbeae (Mantisia). era in the Zingiberaceae are monospeci®c, but no criteria for molecular analyses, to evaluate the status of monotypic genera the recognition of these species at the generic level have been where possible, and to construct a new phylogenetic classi®- established. One important question in the classi®cation of the cation of the family. family is how to determine the taxonomic status of these often morphologically very distinctive monospeci®c genera. MATERIALS AND METHODS Recently several papers have used molecular data to explore the phylogenetic relationships within the family Zingiberaceae TaxaÐWe sampled 104 species in 41 genera representing all four tribes of (Searle and Hedderson, 2000; Wood, Whitten, and Williams, the Zingiberaceae (http://ajbsupp.botany.org/v89/). In the larger genera (e.g., 2000) as well as within several genera (Hedychium: Wood, Alpinia, Curcuma, Hedychium, Globba, Zingiber) several species representing Whitten, and Williams, 2000; Alpinia: Rangsiruji, Newman, infrageneric morphological diversity were included. Unfortunately, for several important genera (e.g., Elettaria, Geocharis, Geostachys) we either did not and Cronk, 2000a, b; Roscoea: Ngamriabsakul, Newman, and have adequate tissue samples or we were not able to amplify DNA. We also Cronk, 2000; Aframomum: Harris et al., 2000). These analyses lacked adequate samples of a number of rare monospeci®c genera (e.g., Lep- have succeeded in clarifying the patterns of evolutionary re- tosolena, Parakaempferia, Paracautleya, Nanochilus, and Cyphostigma). lationships to varying degrees, but in general have been limited in breadth of taxon sampling as well as resolution. OutgroupÐWe initially used representative taxa of the seven other families Our goal in the present study is to use molecular sequence of the Zingiberales as outgroups in our molecular analyses (http://ajbsupp. data to generate hypotheses on the phylogenetic relationships botany.org/v89/). However, because of the dif®culty of aligning the sequences among the genera of the Zingiberaceae in order to evaluate of the internal transcribed spacer (ITS) loci and matK-trnK ¯anking intergenic past classi®cations of the family, to identify morphological spacer regions in these other families with our ingroup taxa, we ®rst conducted features that characterize the various clades detected by the more focused analyses of the coding region of matK, which is easier to align 1684 AMERICAN JOURNAL OF BOTANY [Vol. 89 Larsen et al., 1998). (100) around style petaloid, fused to labellum Zingiber (3) (3) (4) (100) sometimes connate to ®lament Hemiorchis Mantisia Gagnepainia 1-locular 3-locular not exserted beyond anther sacs exserted beyond anther sacs Globba Tribes (2) (10) (3) (1) (30) (1) (4) (15) (1) (1) (60) (1) (4) (7) (40) (1) (50) (2) (3) (50) (2) (17) (2) pletely so) sacs Hitchenia Kaempferia Laosanthus Paracautleya Parakaempferia Pyrgophyllum Roscoea Scaphochlamys Siliquamomum Siphonochilus Smithatris Stahlianthus Cornukaempferia Curcuma Curcumorpha Distichochlamys Hanif®a Haplochorema Hedychium Camptandra Caulokaempferia Cautleya 3-locular (sometimes incom- not exserted beyond anther Boesenbergia (25) (1) (2) (6) (20) (1) (1) (10) (1) (50) (1) (50) (1) (18) (1) (6) (75) (8) (7) (70) (150) (6) (60) (1) (225) Alpinieae A. Rich. (1841) Hedychieae Horan. (1862) Globbeae Meisn. (1842) Zingibereae Meisn. (1842) pletely so) sacs Burbidgea Cyphostigma Elettaria Elettariopsis Etlingera Geocharis Geostachys Hornstedtia Leptosolena Nanochilus Paramomum Plagiostachys Pleuranthodium Pommereschea Renealmia Rhynchanthus Riedelia Siamanthus Stadiochilus Tamijia Vanoverberghia Aframomum Alpinia Amomum Aulotandra a Character 1. Characteristics and genera of the previously recognized tribes of the Zingiberaceae (after Schumann, 1904; Holttum, 1950; Burtt and Smith, 1972; Authorities follow Reveal (2002). Placement according to references cited above. Number of species follows each genus. a ABLE 3) Labellum4) Stamen5) Ovary not connate to ®lament medium length 3-locular (sometimes incom- not connate to ®lament short length connate to ®lament in slender tube not connate to ®lament long with arching ®lament anther crest elongated and wrapped 6) Placentation7) Style axial or free central not exserted beyond anther axial, basal, or free columnar parietal axial T 1) Plane of distichy2) of Lateral leaves staminodes perpendicular to rhizome small or absent, never petaloid parallel to rhizome petaloid, free8) from Stigma labellum9) Placement of genera petaloid, free from labellum and parallel to rhizome expanded parallel to rhizome expanded not expanded not expanded October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1685

under vacuum, and run on an ABI 377 Automated Sequencer (Applied Bio- systems, Foster City, California, USA) at the Smithsonian Institution's Lab- oratory for Molecular Systematics. Raw sequences were assembled and edited using Sequencher 3.1.1 (Gene Codes Corporation, Ann Arbor, Michigan, USA) and manually aligned in Se-Al 2.0a3 (Rambaut, 2000). Results of ex- ploratory analyses (not shown) that excluded dif®cult-to-align regions (104 base pairs [bp] in ITS and 254 bp in matK) produced topologies very similar to those including the entire alignment. For this reason, we are presenting here only results from analyses of the full data set.

Phylogenetic analysesÐMaximum parsimony analyses of the ITS and matK sequence data were conducted using PAUP*4.0 (Swofford, 1998) with unweighted characters and 500 random-sequence-addition replicates, saving all shortest trees under tree bisection-reconnection (TBR) branch swapping, STEEPEST DESCENT off, MULTREES on, COLLAPSE branches if maximum length is zero. Multiple random-sequence additions were chosen to min- imize the likelihood of being trapped on any particular tree island (Maddison, 1991). Bootstrap analyses (Felsenstein, 1985; Mort et al., 2000) were conducted using PAUP*4.0 with ten random addition replicates, TBR branch swapping, for 100 bootstrap replicates. Bootstrap support was categorized as strong (Ͼ85%), moderate (70±85%), weak (50±70%), or poor (Ͻ50%). The data sets for each gene region were analyzed separately and then, following the total evidence approach for multiple data sets (de Queiroz, Donoghue, and Fig. 7. Floral diagram of the Zingiberaceae with perianth whorls, fertile stamen, lateral staminodes, and labellum indicated. (Not drawn to scale; based Kim, 1995; Nixon and Carpenter, 1996), we combined the sequence data. For on Kress, 1990.) the combined analysis, all 104 species were included, although six species (Caulokaempferia coenobialis, Curcuma aeruginosa, Distichochlamys citrea, Kaempferia parvi¯ora, K. pulchra, and Zingiber corallinum; http://ajbsupp. among all of the families. Our results provide strong support for the placement botany.org/v89/) lacked sequence data for either ITS or matK. of the African genus Siphonochilus as the basal lineage to the remaining Zingiberaceae. In addition, all of the other members of the Zingiberaceae we RESULTS sequenced except Siphonochilus share a 12-base deletion at the 3Ј-end of matK; this deletion is absent in all other Zingiberales. We interpret this de- Internal transcribed spacerÐThe ITS-1 had a total aligned letion as a synapomorphy of the Zingiberaceae excluding Siphonochilus, length of 277 bp (unaligned sequences ranged from 206 to which retains this 12-base sequence as a symplesiomorphy with the rest of 247 bp) with a mean guanine-cytosine (GC) content of the Zingiberales. We therefore used Siphonochilus as the outgroup for the rest 54.06%; the 5.8S region had an aligned length of 165 bp of the Zingiberaceae in the subsequent analyses of the complete ITS and matK (range of 163±165 bp) and a GC content of 51.18%; and the regions in order to facilitate sequence alignment within the family and provide maximum resolution among genera. ITS-2 aligned length was 372 bp (range of 242±292 bp) with a GC content of 59.53%. Molecular methodsÐTotal genomic DNAs were extracted using a minor The analysis of the ITS sequence data resulted in 48 equally modi®cation of the Doyle and Doyle (1987) hexadecyltrimethylammonium parsimonious nearly fully resolved trees of 1603 steps (number bromide (CTAB) method. The aqueous phase was extracted with 24 parts of parsimony-informative characters ϭ 323; consistency index chloroform : 1 part isoamyl alcohol. DNA was resuspended in tris and eth- [CI] ϭ 0.366; retention index [RI] ϭ 0.724; rescaled consis- ylenediaminetetraacetic acid (TE) buffer following isopropyl alcohol precip- tency index [RC] ϭ 0.265; Fig. 8). Tamijia is strongly sup- itation. Ampli®cation of ITS was accomplished using ITS4 and ITS5 primers ported (bootstrap value ϭ 90%) as a stem lineage one node (White et al., 1990). The chloroplast matK region was ampli®ed with trnK1F above Siphonochilus. The remaining taxa comprise two major (Manos and Steele, 1997) and trnK2R (Steele and Vilgalys, 1994). All am- monophyletic lineages: one includes most of the Alpinieae pli®cations used Gibco BRL Taq DNA polymerase (Carlsbad, California, (bootstrap value ϭ 86%) and the other includes the Hedy- USA) according to the manufacturer's directions with annealing temperatures chieae plus the Globbeae and Zingibereae, although bootstrap of 54±58ЊC. Ampli®ed products were puri®ed using the polyethylene glycol support for the latter clade is weak (bootstrap value ϭ 52%). 8000 (PEG) precipitation protocol (Johnson and Soltis, 1995) with the prod- Within the Alpinieae clade, Pleuranthodium ϩ Riedelia ϩ ucts sequenced directly using automated sequencing methodology of the ABI Burbidgea ϩ Siamanthus are strongly supported as a mono- Prism௢ Big Dye Terminator Cycle Sequencing Ready Reaction Kit (Foster phyletic group (bootstrap value ϭ 98%). Among the larger City, California, USA). Sequencing primers included the ampli®cation primers genera, Aframomum and Renealmia are strongly supported as plus ITS2 (White et al., 1990) and ITS3G (5Ј-GCA TCG ATG AAG AAC monophyletic. However, several genera are not monophyletic: GTA GT-3Ј; K. J. Williams) for the ITS region. Zingiberales-speci®c internal Alpinia, with four nonmonophyletic groups of species referred sequencing primers were designed by L. M. Prince for the matK region including mSP2F (5Ј-TGG GTT AGA GAC GAA TGT GT-3Ј), mIF (5Ј-GTT to as Alpinia I(A. conchigera and A. galanga), Alpinia II (A. CAG TAC TTG TGA AAC GTT-3Ј), m5Fa (5Ј-CTC TAT GGG TCT TCA elegans, A. luteocarpa, and A. vittata), Alpinia III (A. ble- AGG AT-3Ј), m8Fa (5Ј-TAC TTC GAC TTT CCT GTG CC-3Ј), mSP2R (5Ј- pharocalyx, A. intermedia, A. pumila, A. calcarata, A. of®ci- TTT AAC GTC TAA TTA GAT CGG-3Ј), mIR (5Ј-CGT TTC ACA AGT narum, and A. foxworthyi), and Alpinia IV (A. carolinensis); ACT GAA CTA-3Ј), m5R (5Ј-AGG ATC CTT GAA AAT CCA TAG A-3Ј), Amomum, with two groups of species referred to as Amomum and m8R (5Ј-AGC ACA AGA AAG TCG AAG-3Ј). Primers m5Fa, m8Fa, I(A. glabrum and A. longipetiolatum) and Amomum II (A. and m8R are modi®cations or complements of the primers of Steele and Vil- villosum and A. sp.); and Etlingera, with two groups of species galys (1994). Products were cleaned in Sephadex G-50 (®ne) Centri-Sep spin referred to as Etlingera I(E. elatior and E. yunnanensis) and columns (Princeton Separations P/N 901, Adelphia, New Jersey, USA), dried Etlingera II (E. littoralis). 1686 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 8. One of the 48 equally parsimonious trees of the Zingiberaceae in the analysis of the ITS sequence data (length ϭ 1603; consistency index ϭ 0.366 excluding uninformative characters; retention index ϭ 0.724; and rescaled consistency index ϭ 0.265) showing branch lengths (above the line) and bootstrap values (below the line if Ն50%; values for branches within genera are excluded). Asterisks indicate nodes that collapse in the strict consensus tree. Polyphyletic species groups in the genera Alpinia (Alp), Amomum (Amo), Etlingera (Etl), and Curcuma (Cur) are indicated. October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1687

Within the Hedychieae clade the Globbeae are monophyletic nealmia clade'') and Alpinia II ϩ Vanoverberghia ϩ Etlingera and sister to the remaining taxa, but with poor bootstrap sup- I ϩ Hornstedtia ϩ Etlingera II ϩ Amomum II ϩ Alpinia III port (Ͻ50%). Globba ϩ Mantisia and Hemiorchis ϩ Gag- ϩ Plagiostachys form the second (hereafter referred to as the nepainia are strongly supported as sister genera (bootstrap val- ``Etlingera clade''). As in the ITS results, Pommereschea and ue ϭ 100%). The Zingibereae (the genus Zingiber) are placed Rhynchanthus of the Alpinieae are embedded within the He- as a crown clade within the Hedychieae. Although the topol- dychieae, and Siliquamomum of the Hedychieae is placed ogy within this clade is fully resolved there is only poor sup- within the Alpinieae. port for all nodes connecting more than four genera. Most branches on the tree are short. A clade including Stahlianthus, Combined data setÐThe analysis of the combined ITS and Hitchenia, Smithatris, and Curcuma is strongly supported matK sequence data resulted in 980 equally parsimonious trees (bootstrap value ϭ 98%), and the latter genus is paraphyletic of 2647 steps (number of parsimony-informative characters ϭ with three groups of species referred to as Curcuma I(C. com- 806; CI ϭ 0.450; RI ϭ 0.805; RC ϭ 0.362; Figs. 10±11). A osa, C. aeruginosa, and C. roscoeana), Curcuma II (C. tho- strict consensus of these 980 shortest trees (Figs. 10±11) pro- relii), and Curcuma III (C. bicolor and C. sp.). Boesenbergia vides strong support (bootstrap values Ͼ99%) for the two ma- is also paraphyletic with two groups of species referred to as jor clades identi®ed by the ITS and matK data analyzed sep- Boesenbergia I(B. pulcherrima) and Boesenbergia II (B. ro- arately: (1) the Hedychieae, Globbeae, and Zingibereae, and tunda). All of the remaining genera in the Hedychieae are (2) the traditional Alpinieae. Pommereschea and Rhynchan- strongly supported as monophyletic. Pommereschea and Rhyn- thus are embedded within the former, and Siliquamomum is chanthus of the Alpinieae are embedded within the Hedy- placed in the latter. Tamijia is strongly supported (bootstrap chieae clade and Siliquamomum of the traditional Hedychieae value ϭ 100%) as sister to these two major clades. There is is placed with other Alpinieae. only weak resolution and support among genera in the He- dychieae clade, except for Globba ϩ Mantisia, Hemiorchis ϩ matKÐThe 5Ј trnK-matK intergenic spacer region had a Gagnepainia, Cautleya ϩ Roscoea, and Distichochlamys ϩ total aligned length of 1101 bp (unaligned sequences ranged Scaphochlamys (bootstrap values ϭ 95±100%). The Curcuma from 767 to 884 bp) with a mean GC content of 30.01%; the clade, including Hitchenia, Stahlianthus, Smithatris, and a par- matK coding region had an aligned length of 1641 bp (range aphyletic Curcuma, is strongly supported (bootstrap value ϭ of 1542±1572 bp) and a GC content of 29.69%; and the matK- 100%), whereas the Kaempferia clade is weakly supported 3Ј trnK intergenic spacer region aligned length was 507 bp (bootstrap value ϭ 54%). (range of 279±316 bp) with a GC content of 29.98%. Excluding Tamijia, three of the four major clades in Alpi- The analysis of the matK region (coding and noncoding) nieae are resolved as in the matK analysis. Siliquamomum is resulted in more than 40 000 equally parsimonious trees (at unresolved and Siamanthus is strongly supported (bootstrap which point the analysis was stopped) of 1104 steps (number value ϭ 88%) as part of the Riedelia clade (bootstrap value of parsimony-informative characters ϭ 482; CI ϭ 0.630; RI ϭ 100%). The remaining Alpinieae are moderately supported ϭ 0.893; RC ϭ 0.562; Fig. 9). A strict consensus of these as a monophyletic group (bootstrap value ϭ 85%) with strong shortest trees (Fig. 9) provides strong support (bootstrap values support for the Renealmia clade and Etlingera clade (bootstrap Ͼ99%) for the same two major clades as de®ned by the ITS values ϭ 86 and 91%, respectively) within it. data (the traditional Alpinieae and the traditional Hedychieae plus Globbeae and Zingibereae), but does not resolve the po- DISCUSSION sition of the genus Tamijia with respect to these two clades. There is little resolution among genera in the latter clade. Molecular evidence and the current classi®cationÐThe Globba ϩ Mantisia and Hemiorchis ϩ Gagnepainia are two molecular data sets used in our analysis of the evolution- strongly supported as sister genera (bootstrap values ϭ 100%), ary relationships of the genera in the Zingiberaceae resulted but there is only poor support for uniting these four genera of in generally congruent and compatible trees. The ITS sequence the traditional Globbeae into a single clade. The paraphyletic data produced an almost fully resolved tree, but in general nature of Globba with Mantisia is strongly supported. As in lacked even moderate bootstrap support for many of the main the ITS tree, a ``Curcuma clade,'' including Hitchenia, Stah- clades, especially among the Hedychieae-Globbeae-Zingiber- lianthus, Smithatris, and a paraphyletic Curcuma, is moder- eae taxa. The phylogenetic analysis of the matK coding and ately supported (bootstrap value ϭ 74%). The matK data noncoding regions resulted in stronger bootstrap support for weakly support (bootstrap value ϭ 54%) a ``Kaempferia relationships among the genera of the Alpinieae, but support clade'' consisting of Zingiber, Boesenbergia, Curcumorpha, within the Hedychieae was weak. The combined data sets pro- Kaempferia, Scaphochlamys, Distichochlamys, Cornukaemp- duced a more fully resolved tree with strong support for many, feria, and Hanif®a. but not all, groups of genera. Within the Alpinieae, four major clades are resolved. Sia- The taxonomic concepts of earlier workers (e.g., Schumann, manthus and Siliquamomum are weakly supported as sister 1904; Holttum, 1950; Burtt and Smith, 1972; Larsen et al., genera (bootstrap value ϭ 53%), but unresolved with the other 1998) on the classi®cation of the Zingiberaceae are partially Alpinieae. Pleuranthodium ϩ Riedelia ϩ Burbidgea are congruent with the results of our phylogenetic analyses of mo- strongly supported as a clade (bootstrap value ϭ 96%; here- lecular sequence data. Members of the two largest tribes, Al- after referred to as the ``Riedelia clade''), but likewise unre- pinieae and Hedychieae, for the most part form well-supported solved in the Alpinieae. Among the remaining Alpinieae, two monophyletic groups with the major rearrangements being the primary and almost fully resolved clades are strongly sup- basal position of Siphonochilus and Tamijia in the family, the ported (bootstrap value Ն87%). Renealmia ϩ Aframomum ϩ inclusion of Pommereschea and Rhynchanthus in the Hedy- Elettariopsis ϩ Paramomum ϩ Amomum I ϩ Alpinia I form chieae clade, and the position of Siliquamomum in the Alpin- one monophyletic group (hereafter referred to as the ``Re- ieae clade. We have con®rmed that the former Globbeae are 1688 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 9. One of over 40 000 equally parsimonious trees of the Zingiberaceae in the analysis of the matK region (coding and noncoding) sequence data (length ϭ 1104; consistency index ϭ 0.630 excluding uninformative characters; retention index ϭ 0.893; and rescaled consistency index ϭ 0.562) showing branch lengths (above the line) and bootstrap values (below the line if Ն50%; values for branches within genera are excluded). Asterisks indicate nodes that collapse in the strict consensus tree. Polyphyletic species groups in the genera Alpinia (Alp), Amomum (Amo), Etlingera (Etl), and Curcuma (Cur) are indicated. October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1689

Fig. 10. One of 980 equally parsimonious trees of the Zingiberaceae in the analysis of the combined ITS and matK region sequence data (length ϭ 2647; consistency index ϭ 0.450 excluding uninformative characters; retention index ϭ 0.805; and rescaled consistency index ϭ 0.362) showing branch lengths (above the line) and bootstrap values (below the line if Ն50%; values for branches within genera are excluded). Asterisks indicate nodes that collapse in the strict consensus tree. Polyphyletic species groups in the genera Alpinia (Alp), Amomum (Amo), Etlingera (Etl), and Curcuma (Cur) are indicated. 1690 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 11. One of 980 equally parsimonious trees of the combined analysis of the Zingiberaceae in which monophyletic genera have been collapsed into a single branch for clarity. Branch lengths (above the line) and bootstrap values (below the line if Ն50%) are shown for clades encompassing two or more genera. Asterisks indicate nodes that collapse in the strict consensus tree. Subfamilies and tribes of the new classi®cation of the family are indicated (see Fig. 10 and Tables 2±4). October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1691 closely allied with the Hedychieae, but we have not resolved TABLE 2. A key to the subfamilies and tribes of a new Linnaean clas- whether the four genera of this tribe constitute a monophyletic si®cation of the Zingiberaceae (see Figs. 10 and 11 and Table 4). or paraphyletic group. It is clear that the genus Zingiber is a Authorities follow Reveal (2002). derived taxon within the Hedychieae and should not be recognized at the tribal level. 1. Plane of distichy of leaves perpendicular to rhizome. 2. Lateral staminodes well developed and fused to labellum. (Sta- Our results provide some insight into the long-standing tax- minodia laterales bene evoluta et ad labellum conjuncta.) onomic problems that have been encountered in some of the 3. Plants evergreen with ®brous rhizomes; ovary unilocular with largest genera of the family, e.g., Alpinia, Amomum, Etlingera, parietal placentation. (Plantae sempervirens rhizomatibus ®bro- Curcuma, and Globba. Even though we only included a mod- sis; ovarium uniloculare placentatio parietalis.) ....Tamijioideae est sampling of each genus in our analysis, the polyphyletic W. J. Kress, subfam. nov. et Tamijieae W. J. Kress, tribus nov. status of each of these ®ve genera is clear. Other large genera 3. Plants with seasonal dormancy period and ¯eshy rhizomes; ovary trilocular with axial placentation. (Plantae dormientes such as Hedychium, Kaempferia, Zingiber, Aframomum, and seasonaliter et rhizomatibus carnosis; ovarium triloculare pla- Renealmia are strongly supported as monophyletic. centatio axilis.) .... Siphonochiloideae W. J. Kress, subfam. nov. The present study is the most exhaustive investigation to et Siphonochileae W. J. Kress, tribus nov. date of the evolutionary relationships within the family Zin- 2. Lateral staminodes reduced or absent. (Staminodia laterales re- giberaceae. Searle and Hedderson (2000) analyzed the rela- ducta vel absens.)...... Alpiniodeae Link (1829) tionships among 12 genera (including Siphonochilus) in the 4. Extra¯oral nectaries present on leaf blades; fruits silique-like opening by longitudinal slits. (Laminae foliorum nectariis ex- traditional Hedychieae plus Zingiber and ®ve genera of the tra¯oralibus; fructibus similibus siliquarum, a rimis longitu- Alpinieae using ITS sequence data; Globba was used to root dinalibus aperiens.)...... Riedelieae W. J. Kress, tribus nov. the tree. Their results were congruent with our analyses in 4. Extra¯oral nectaries absent; fruits ¯eshy or indehiscent. uniting 11 of the 12 genera of the Hedychieae in a single clade (Nectariis extra¯oralibus absens; fructibus carnosis vel and recognizing a ``Kaempferia group'' consisting of Kaemp- indehiscentibus.) ...... Alpinieae A. Rich. (1841) feria, Boesenbergia, Zingiber, Cornukaempferia, Scaphochla- 1. Plane of distichy of leaves parallel to rhizome .. Zingiberoideae Haask. (1844) mys, and Distichochlamys. However, only weak jackknife sup- 5. Ovary trilocular with axial, basal or free columnar placentation; port was present for any branches connecting more than two labellum usually not connate to the ®lament .....Zingibereae Meisn. genera. They also noted that Siphonochilus occupied an iso- (1842) lated position exterior to the Hedychieae and sister to the tra- 5. Ovary unilocular with parietal placentation; labellum often con- ditional Alpinieae. This position sister to the Alpineae was nate to ®lament in a slender tube ...... Globbeae Meisn. (1842) clearly an anomaly resulting from selection of Globba as the outgroup in their study. Wood, Whitten, and Williams (2000) focused on the genus Hedychium in their analysis and included results of our analyses that do have strong support, we propose 17 other genera primarily in the traditional Hedychieae. Their a realignment of the genera of the Zingiberaceae into four results were similar to Searle and Hedderson's in de®ning a subfamilies (Tables 2±4): the Siphonochiloideae (the genus Si- Kaempferia group (without Scaphochlamys, but with Zingiber) phonochilus only), the Tamijioideae (the single genus Tami- and placing Siphonochilus near Alpinia. As resolved by our jia), the Alpinioideae (most of the former Alpinieae), and the data, the analysis by Wood, Whitten, and Williams provided Zingiberoideae (including the former tribes Hedychieae, Zin- support for a Curcuma clade (including the genera Curcuma, gibereae, and Globbeae). Hitchenia, and Stahlianthus) and the position of Pommeres- Within the Alpinioideae, we recognize the two main clades chea and Rhynchanthus in the Hedychieae. as tribes Alpinieae and Riedelieae; the placement of Siliqua- The investigations by Rangsiruji, Newman, and Cronk momum is equivocal and requires further study (Fig. 11). The (2000a, b) and Harris et al. (2000) concentrated on the tradi- polyphyletic nature of three of the major genera in the Alpi- tional Alpinieae, especially infrageneric relationships within nieae (Alpinia, Amomum, and Etlingera) prevents us from for- Alpinia and Aframomum, respectively, using ITS sequence mally recognizing the Renealmia clade and the Etlingera clade data and included several additional genera as outgroups. Our at this time, even though each clade has strong bootstrap sup- analyses con®rmed the close relationship of Pleuranthodium port. Within the Zingiberoideae we recognize two tribes to and Burbidgea, the polyphyly of Amomum, and the monophy- distinguish most of the old Hedychieae (here called the Zin- ly of Aframomum as shown in their studies. Rangsiruji, New- gibereae) from the Globbeae (Fig. 11). Although strong mo- man, and Cronk suggested that Alpinia may be paraphyletic lecular support is lacking for either tribe within the Zingiber- with Renealmia, but because of their limited outgroup selec- oideae, we believe that it is better to recognize the Zingibereae tion they did not have suf®cient data to support any major and to retain the Globbeae until more evidence is accumulated. taxonomic changes in these taxa (see discussion below under We also refrain from formally recognizing either the Kaemp- Alpinioideae). feria clade or the Curcuma clade at this time. We have not been able to place the genus Caulokaempferia into either tribe A new classi®cation of the ZingiberaceaeÐAlthough some with any degree of certainty. A forthcoming investigation of congruence exists between the former classi®cations of the the Globbeae and related taxa will address some of these issues Zingiberaceae and our current phylogenetic results (Fig. 11), in more depth (K. J. Williams, W. J. Kress, and P. S. Manos, we believe that a revised classi®cation is needed that better unpublished data). The relationships among the genera are dis- represents the new data on evolutionary relationships. We ac- cussed below within the framework of this revised classi®ca- cept that our results do not fully resolve all of the relationships tion. among genera, that some major clades have only weak bootstrap support, that some ambiguity still exists in the placement Basal lineagesÐThe strong support from the coding region of certain genera, and that several critical large genera have of matK for the position of Siphonochilus as the basal lineage now been shown to be polyphyletic. Nonetheless, based on the in the family was at ®rst surprising. The subsequent discovery 1692 AMERICAN JOURNAL OF BOTANY [Vol. 89

TABLE 3. Placement of genera in the new classi®cation of the family Zingiberaceae. (Asterisks indicate taxa not sampled in the molecular phylogenetic analyses; these genera are tentatively placed based on morphological features.)

Subfamily Subfamily Subfamily Subfamily

Siphonochiloideae Tamijioideae Alpinioideae Zingiberoideae Globbeae W. J. Kress W. J. Kress Link Haask. Zingiberoideae:

Tribe Tribe Tribe Tribe labellum and sometimes connate to ®lament slender tube times arching spurred or absent 1-locular parietal Siphonochileae Tamijieae Alpinieae Zingibereae Meisn. petaloid, free from

W. J. Kress W. J. Kress A. Rich. rsen et al., 1998; Sakai and Siphonochilus Tamijia Aframomum Boesenbergia Alpinia Camptandra e 1). Amomum Cautleya *Aulotandra Cornukaempferia *Cyphostigma Curcuma *Elettaria Curcumorpha

Elettariopsis Distichochlamys Zingibereae Etlingera Hanif®a Zingiberoideae: well-developed and wrapped around style incompletely so) columnar *Geocharis *Haplochorema fused to labellum short to long short to long,axial, some- basal, or free petaloid, free from or *Geostachys Hedychium parallel to rhizome parallel to rhizome Hornstedtia Hitchenia *Leptosolena Kaempferia Paramomum *Laosanthus Plagiostachys *Nanochilus Renealmia *Paracautleya Vanoverberghia *Parakaempferia

Pommereschea Alpinieae

Tribe Pyrgophyllum Alpinioideae: arching petaloid Riedelieae Rhynchanthus rhizome indehiscent or ¯eshy ¯eshy and dehiscent globbose and dehiscent small or absent, never W. J. Kress Roscoea perpendicular to Burbidgea Scaphochlamys Pleuranthodium Smithatris Riedelia *Stadiochilus Siamanthus Stahlianthus Zingiber Subfamilies and Tribes Incertae Sedis Siliquamomum Tribe

Globbeae Meisn. Riedelieae Gagnepainia Alpinioideae: longitudinal slits petaloid Globba rhizome small or absent, never Hemiorchis perpendicular to Mantisia

Incertae Sedis Caulokaempferia Tamijieae

of a 12-base insertion at the 3Ј end of matK that this genus Tamijioideae: labellum shares with all other Zingiberales, but not the rest of the Zin- rhizome 1-locular 1- or 3-locular 3-locular 3-locular (sometimes petaloid, fused to giberaceae, provided unequivocal support for this basal place- perpendicular to ment. Our morphological interpretation of this topology is that the partial to almost complete fusion of the lateral staminodes to the large labellum in Siphonochilus is a plesiomorphic character of the basal Zingiberaceae shared with the sister family Costaceae (Figs. 1±2, 7). The recent discovery of the monotypic genus Tamijia, a narrow endemic from northern Borneo

(Sakai and Nagamasu, 2000), is equally interesting. Although Siphonochileae Siphonochiloideae: incompletely so) labellum the in¯orescence structure and vegetative morphology are un- rhizome petaloid, fused to like Siphonochilus, Tamijia shares the same fusion of the lat- perpendicular to eral staminodes and labellum, and some relationship between the two genera was suggested by Sakai and Nagamasu (2000). The independent results of our molecular analysis place Tam- ijia as sister to the remaining Zingiberaceae, excluding the basal Siphonochilus, and con®rm that this ¯oral character is a

feature shared with the sister taxa of the Costaceae. Because Character of the distinctness of these two genera in the family and their 4. Characteristics of the subfamilies and tribes of the new classi®cation of the Zingiberaceae presented here (Burtt, 1972; Burtt and Smith, 1972; La Nagamasu, 2000; W. J. Kress, personal observation). Some nondiagnostic features are included as a reference to the previous classi®cation (see Tabl

basal positions, we have chosen to recognize each genus as a leaves ABLE 6) Labellum7) Filament8) Anther crest not connate to ®lament9) Ovary short not connate10) to petaloid Placentation ®lament not11) connate Capsule to ®lament axial not connate to 3-locular ®lament (sometimes not connate to petaloid short ®lament ¯eshy connate to ®lament in parietal petaloid or absent unknown medium petaloid or absent axial or parietal silique-like, opening absent, by petaloid, medium, or sometimes axial or free central 4) Extra¯oral nectaries5) Lateral staminodes absent absent present on leaf blades absent absent absent subfamily of the Zingiberaceae, the Siphonochiloideae and the T 1) Seasonality2) Rhizomes3) Plane of distichy of dormancy period ¯eshy evergreen ®brous evergreen ®brous evergreen dormancy period ®brous dormancy period ¯eshy ¯eshy October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1693

Tamijioideae, to indicate their phylogenetic position in the are placed in two separate sections of subgenus Alpinia by Zingiberaceae (Tables 2±4). Smith, but here form a strongly supported clade basal to the other members of the Renealmia clade (Aframomum, Reneal- AlpinioideaeÐThe former tribe Alpinieae and new subfam- mia, Amomum I, Elettariopsis, and Paramomum; including Af- ily Alpinioideae are de®ned by a number of features (Tables rican, Asian, and American species). The rest of Alpinia is 1, 2, 4). The most conspicuous ¯oral feature is the great re- scattered in the Etlingera clade. Alpinia II (A. elegans, A. lu- duction or absence of the two lateral staminodes (Figs. 4, 7). teocarpa, and A. vittata; part of the A. eubractea clade of These staminodes represent two modi®ed stamens of the outer Rangsiruji, Newman, and Cronk) includes the genus Vanov- staminal whorl of the Zingiberaceae ¯oral plan; the third mem- erberghia (from the Philippines and Taiwan) and is allied with ber of this whorl is absent. The ``labellum,'' which is nearly the polyphyletic Etlingera, Hornstedtia, and Amomum II (all ubiquitous in the family, is a petaloid structure formed by the Asian). Alpinia III (six species included in four of Rangsiruji, fusion of the two lateral sterile stamens of the inner staminal Newman, and Cronk's clades) is united with the genus Pla- whorl; the median member of this whorl is the single tetras- giostachys whereas Alpinia IV (single species A. carolinensis porangiate fertile stamen. The lateral staminodes are also ab- in Rangsiruji, Newman, and Cronk's A. carolinensis clade) is sent in the genera Rhynchanthus and Pommereschea, usually unresolved with Alpinia II and III. A more extensive analysis placed in the traditional Alpinieae. However, in these two gen- of Alpinia is in progress (A.-Z. Liu and W. J. Kress, unpub- era the labellum is also reduced or absent and our molecular lished data). results place them within the new Zingiberoideae. These two Taxonomic matters are made worse by the polyphyletic na- genera are clear examples of independent reductions in lateral ture of the genus Amomum. Although we only sampled four staminodes in the family. species in this genus, these four species form two separate The second apomorphy of the Alpinioideae is the perpen- lineages: one within the Renealmia clade as sister to the genus dicular orientation of the plane of distichy of the shoots with Elettariopsis, which has a striking similarity in ¯oral structure respect to the direction of growth of the rhizome (Burtt, 1972). and is often dif®cult to distinguish from Amomum (Holttum, This character is present in all of the genera formerly placed 1950; Kiew, 1982), and the other in the Etlingera clade, which in the Alpinieae, except for Rhynchanthus and Pommereschea is strongly supported to include a species of Etlingera. The in which the orientation is parallel. In accordance with the polyphyly of Amomum was also suggested by Harris et al. suggestions of others (e.g., K. Larsen [AAU], T. Wood (2000) in their investigation of the molecular systematics of [FLAS], and M. Newman [E], personal communication; the genus Aframomum. In a separate analysis of over 30 spe- Smith, 1980), these two genera should be included in the Zin- cies of Amomum using ITS and matK sequence data, we have giberoideae. Sakai and Nagamasu (2000) assigned their new identi®ed four separate lineages of Amomum scattered among genus Tamijia to the Alpinieae based on the perpendicular the other genera of the Alpinieae (Y.-M. Xia, W. J. Kress, and shoot orientation, but recognized that their genus was some- L. M. Prince, unpublished data). Needless to say, the generic what anomalous in this tribe. Indeed, according to our molec- boundaries and de®nitions of all the taxa in this tribe require ular results, Tamijia deserves recognition as a separate sub- more extensive and careful molecular and morphological study family altogether. before a new classi®cation can be proposed. For this reason Within the subfamily Alpinioideae are two major groups of we have chosen not to give formal taxonomic status to either genera, here recognized as tribes Alpinieae and Riedelieae (Ta- the Renealmia clade or the Etlingera clade, even though strong bles 2±4). The ®rst tribe contains the majority of genera in the subfamily and includes taxa distributed in Asia, Africa, and molecular support exists for both lineages. the Americas. The taxa in the Alpinieae fall into two main The Riedelieae (Tables 2±4) include four genera united by subclades, the Renealmia clade and the Etlingera clade, each both molecular and morphological characters: Burbidgea (Bor- of which is supported by strong bootstrap values. However, neo), Pleuranthodium (primarily New Guinea and Bismarck three of the largest genera in this tribe, Etlingera, Amomum, Archipelago), Riedelia (New Guinea), and Siamanthus (Thai- and Alpinia, are polyphyletic, and the latter two have species land). The af®nities of these four genera are indicated by their in both the Renealmia clade and the Etlingera clade. Because common possession of long, slender silique-like capsules that our sampling of Etlingera was modest (three species) we sug- open by longitudinal slits to the base (Smith, 1972, 1990b; gest that taxon sampling be increased before any ®rm conclu- Larsen and Mood, 1998). Larsen and Mood (1998) also point- sions be drawn about the status of this genus. However, it is ed out that the close relationship of at least three of these clear from this investigation as well as others (Harris et al., genera (Burbidgea, Riedelia, and Siamanthus) is suggested by 2000; Rangsiruji, Newman, and Cronk, 2000b) that major ad- unique ``nectary pits'' on the dorsal midrib of the leaf blade. ditional taxonomic studies and realignments are necessary in We have now con®rmed the presence of these extra¯oral nec- both Alpinia and Amomum.InRenealmia, the one genus in taries in Pleuranthodium as well (W. J. Kress and M. Borde- the Alpinioideae that is distributed across continents, our lim- lon, unpublished data). These vegetative and reproductive ited sampling of six species provided good support for a dis- characters taken together make the Riedelieae a readily rec- tinct African clade and a tropical American clade (Figs. 8±10). ognizable tribe in the Alpinioideae. For Alpinia the results of our analysis identi®ed four sepa- The long silique-like fruits of Siliquamomum (Southern Chi- rate groups (Alpinia I±IV) in the 12 species we sampled. Each na and Vietnam) are similar to the fruits found in the Riede- of these four groups corresponds to one of the clades recog- lieae, but the presence of a well-developed labellum and lateral nized in the molecular analysis of Rangsiruji, Newman, and staminodes as well as the absence of the extra¯oral nectaries Cronk (2000b) and, as in their analysis, our results do not on the leaf midribs separate this genus from other members support the classi®cation of Alpinia proposed by Smith of that tribe. The placement of Siliquamomum in either the (1990a). Alpinia galanga and A. conchigera (Alpinia I; part Riedelieae or the Alpinieae is unresolved by our molecular of the A. galanga clade of Rangsiruji, Newman, and Cronk) analysis. For these reasons we prefer to place this genus as 1694 AMERICAN JOURNAL OF BOTANY [Vol. 89 incertae sedis until further evidence con®rms its position in (Burtt, 1972; Smith, 1980; Wood, Whitten, and Williams, the Alpinioideae. 2000). A second vegetative trait that characterizes the Zingi- Within the Alpinioideae, we included in our molecular anal- beroideae (and Siphonochilus), brie¯y discussed by Wood, ysis four of the six monospeci®c genera. Cyphostigma pul- Whitten, and Williams (2000), is the capacity for members of chellum (Thw.) Benth., restricted in distribution to Sri Lanka, this subfamily to go into dormancy during the dry season in is poorly known (Dassanayake, 1983), and no tissue was avail- monsoonal climates. Unlike the Alpinioideae and Tamijioi- able for our investigation. Leptosolena haenkei Presl, endemic deae, in which individuals are evergreen throughout the year, to the lower slopes of Mt. Pinatubo in Northern Luzon, Phil- the Zingiberoideae, including taxa of the former Hedychieae, ippines, has not been collected since it was described in the Globbeae, and Zingibereae, have a forced dormancy period early 1900s and is presumed extinct (Madulid, 1996). Based during which all aboveground parts are shed and the plant on morphological features we have tentatively placed these overwinters as a thick, often ¯eshy underground rhizome small genera (along with the African Aulotandra and Asian (sometimes with starch-®lled roots or tubers). Either just prior Geocharis and Geostachys) in the Alpinieae. Vanoverberghia, to or at the earliest sign of the wet season, individuals will also from the Philippines, was only known from a single spe- break dormancy with either vegetative shoots or reproductive cies, V. sepulchrei Merr., until a second species, V. sasakiana shoots (e.g., Hemiorchis, Gagnepainia, Mantisia, some H. Funak. & H. Ohashi, was described from the Lanyu Islands kaempferias and many curcumas). However, some species in southeast of Taiwan (Funakoshi and Ohashi, 2000). Merrill the Zingiberoideae can also be evergreen in the wet forest (1912), in his original description of the genus, compared it habitats where they occur (e.g., some globbas, zingibers, and to Riedelia, but also recognized its af®nity to Alpinia as did hedychiums), but these same species can be forced into dor- Smith (1990a). Our results show that this genus is a member mancy under stress or in greenhouse environments (M. Bor- of the Alpinia II clade that includes other species from the delon [US] and T. Wood [FLAS], personal communication). Philippines (e.g., A. elegans and A. luteocarpa). As noted These latter taxa may be descendents of ancestors from mon- above, a phylogenetic and taxonomic revision of Alpinia and soonal climates that have secondarily colonized evergreen hab- its closely related genera, including Vanoverberghia and Pla- itats. As far as we know, members of Alpinioideae cannot be giostachys, is needed. forced into dormancy, possess tough ®brous rather than ¯eshy The monotypic Paramomum, endemic to Yunnan, China, rhizomes, and are always evergreen in their natural habitats. was recently moved by Wu (1997) into the genus Amomum. Within the Zingiberoideae, we have chosen, for historical In the original description of Paramomum, Tong (1985) com- consistency, to recognize two tribes even though our molecular pared it to Costus because of its supposed spiral phyllotaxy. data do not provide strong support for these clades (Tables 2± Our molecular results place this genus as sister to Elettariopsis 4). The Zingibereae include most of the former Hedychieae as in a clade with species of Amomum I, thus supporting its sep- well as the genus Zingiber. All members of this tribe possess aration from Amomum and clear distance from Costus. The a trilocular ovary, which in some genera may be incompletely monotypic and distinct nature of the remaining two genera in closed towards the apex (e.g., Scaphochlamys and Paracau- the Alpinioideae, Siliquamomum and Siamanthus, is also sup- tleya), and axial or rarely basal placentation. Within the Zin- ported here by the long branch lengths and isolated positions gibereae several groups of genera are supported by varying of these taxa in or near the Riedelieae. bootstrap values. A Kaempferia clade, including Hanif®a, Zin- giber, Kaempferia, Distichochlamys, Scaphochlamys, Boesen- ZingiberoideaeÐOur subfamily Zingiberoideae contains bergia, Curcumorpha, and Cornukaempferia (and probably the remaining genera in the family, which were formerly Haplochorema and Parakaempferia not included in our anal- placed within the three tribes Zingibereae, Hedychieae, and ysis) is only weakly supported here, but has also been recog- Globbeae (Table 3). The most recognizable ¯oral features of nized by others (Newman, 1995; Searle and Hedderson, 2000; this subfamily are the conspicuous and often well-developed Wood, Whitten, and Williams, 2000). These taxa lack the dis- lateral staminodes (Figs. 3, 5, 6) that are generally absent in tinctive pseudostem (except for Zingiber and Hanif®a) found the Alpinioideae (Tables 2, 4). These lateral staminodes are in other members of the subfamily. Although it has been sug- fused to the corolla tube and free from the labellum in most gested that the genus Cornukaempferia is allied with either of the former Hedychieae, whereas they are fused to the la- Kaempferia (Mood and Larsen, 1997) or Zingiber (Searle and bellum in the genus Zingiber (former Zingibereae; Fig. 5) and Hedderson, 2000; Wood, Whitten, and Williams, 2000), our connate to the ®lament in Globba and Mantisia (part of the results do not support either of these hypotheses and resolution former Globbeae; Fig. 6). In a few genera the lateral stami- among the genera of the Kaempferia clade is generally low. nodes and/or the labellum are reduced or absent (Rhynchan- The close relationship between Distichochlamys and Sca- thus, Pommereschea, Nanochilus, Stadiochilus, Laosanthus, phochlamys as well as their generic boundaries have been rec- Hedychium hors®eldii, and H. bordelonianum; Smith, 1980; ognized (Newman, 1995). However, the taxonomic position of Newman, 1990; Larsen and Jenjittikul, 2001; Williams, Kress, Curcumorpha and Boesenbergia requires some discussion. and Thet Tun, 2002). Hence, the presence of well-developed Boesenbergia is a genus of about 60 species found throughout lateral staminodes is a feature not shared by all members of tropical Asia, including India, Myanmar, Thailand, Malaysia, the Zingiberoideae as recognized here. and Borneo (Sirirugsa, 1992), and is distinguished by its two- The most prominent vegetative feature that is universal in ranked in¯orescence bracts, saccate labellum, and phenologi- this subfamily is the parallel orientation of the plane of disti- cal pattern of ¯owers opening from the apex toward the base chy of the leafy shoots with respect to the rhizome (perpen- (Smith, 1987). Curcumorpha longi¯ora was described as a dicular orientation in all other Zingiberaceae). As pointed out monotypic genus from northeastern India (now known to also earlier, this character was the basis for some authors to suggest occur in Myanmar and Thailand) distinctive in its radical in- that genera such as Rhynchanthus and Pommereschea with ¯orescences with spirally arranged bracts and the presence of parallel orientation were not properly placed in the Alpinieae a staminodial cup (Rao and Verma, 1971). Larsen (1997) October 2002] KRESS ET AL.ÐPHYLOGENY OF THE GINGERS 1695 placed Curcumorpha into Boesenbergia stating that the char- beae as earlier recognized until forthcoming evidence proves acters cited by the original authors did not suf®ciently distin- otherwise. guish it from that genus. Our results suggest that Boesenbergia Caulokaempferia, ®rst included within the genus Kaempferia, itself may be polyphyletic with one group of species (here was subsequently recognized as distinct by Larsen (1964). Its represented by B. pulcherrima) allied to Curcumorpha. This possible relationship to both Camptandra and Boesenbergia have taxonomic complex is yet another example of the problems been noted (Larsen and Smith, 1972). Our results to not closely that exist in de®ning clear generic boundaries in the Zingi- ally Caulokaempferia to any of the above three genera, and the beraceae. combined analyses of the two molecular data sets leaves the The Curcuma clade has strong bootstrap support and in- placement of the genus within the Zingiberoideae unresolved cludes Curcuma, Stahlianthus, Hitchenia, and Smithatris, with respect to the Zingibereae and the Globbeae. We therefore which share cone-like in¯orescences of few-¯owered, con- at present recognize Caulokaempferia as incertae sedis within the gested bracts (although much reduced in Stahlianthus). Cur- Zingiberoideae in our revised classi®cation. cuma, with about 50 species, has been circumscribed by its We have been able to address the phylogenetic placement ``pouched'' in¯orescence bracts and versatile, usually spurred and distinctiveness of a number of the monotypic genera with- anthers. However, these features are neither unique nor uni- in the Zingiberoideae. Taxa such as Hanif®a, Cornukaempfer- versal in the genus, and our results along with those of others ia, and Pyrgophyllum are characterized by relatively long (Wood, Whitten, and Williams, 2000) strongly suggest that branch lengths in the combined analyses (32±41 bp), while Curcuma is paraphyletic with Hitchenia and Stahlianthus, and others, such as Curcumorpha, Hitchenia, Stahlianthus, and maybe Smithatris. Expanded taxon sampling within these gen- Smithatris, are less distinct (9±20 bp). As exploration of new era is necessary to realign species and distinguish boundaries habitats in Southeast Asia increases, additional species are be- among the taxa (T. Rehse, W. J. Kress, and P. S. Manos, un- ing found in many genera that were previously only known published data). Rather than accepting an enlarged concept of from a single species (e.g., Curcumorpha, Cornukaempferia, Curcuma to include these associated genera, the strong support Hanif®a, Distichochlamys, and Smithatris). In addition, several of several clades within the Curcuma complex may necessitate monotypic genera such as Nanochilus, Paracautleya, Parak- the splitting up of Curcuma into several smaller genera. aempferia, Laosanthus, and Stadiochilus are yet to be sampled Among the remaining genera in the Zingibereae two pri- for inclusion in molecular investigations. We have tentatively marily Himalayan clades have strong support: the Roscoea ϩ placed these genera in the Zingibereae based on their mor- Cautleya clade in the western end of the range from Kashmir phology alone. A more thorough sampling and expanded phy- to southwestern China and the Pommereschea ϩ Rhynchan- logenetic analyses should provide additional answers to the thus clade in the eastern end straddling the borders of Yunnan, status of the many small genera in the family. Myanmar, and Thailand. Both of these clades have moderate to strong support and at least in the ITS analysis form a weakly Future directionsÐWe recommend that more species of the supported Himalayan alliance of four genera. The placement large paraphyletic genera Curcuma, Alpinia, Amomum, and Etlin- of Hedychium (a large, strongly supported monophyletic ge- gera be added to future molecular analyses to better delimit genus), the monotypic Pyrgophyllum endemic to southwestern neric boundaries in these taxa and their allied genera. Taxa from China, and the few-species Camptandra from the lowland wet important, but undersampled, genera such as Hornstedtia, Aulo- forests of Thailand and Malaysia is still unresolved, although tandra, Elettaria, Geocharis, Geostachys, Riedelia, Burbidgea, the latter two genera are allied with the Curcuma clade in the and Scaphochlamys should be included in future investigations. ITS analysis. We are especially interested in the placement and taxonomic sta- The four genera in the traditionally recognized Globbeae tus of the remaining monotypic genera (Stadiochilus, Nanochilus, (Tables 2±4; Fig. 6) are united by several morphological char- Cyphostigma, Laosanthus, Paracautleya, Parakaempferia, and acters, such as the unilocular ovary with parietal placentation Leptosolena), some of which are exceedingly rare and may be (convergent with Tamijia) and the long-arching ®lament to extinct in the wild. We plan to continue our taxon sampling which the labellum and/or lateral staminodes are fused (in throughout the family with the hopes of recognizing appropriate Globba and Mantisia only). Our molecular analysis strongly monophyletic genera and eventually proposing a more detailed places the four genera of the Globbeae within the Zingibero- classi®cation of the Zingiberaceae. We are also currently utilizing ideae, but only weakly outside the taxa comprising the Zin- our phylogenetic results to investigate patterns of biogeographic gibereae (see discussion below on Caulokaempferia). A num- history and character evolution in this interesting and diverse ber of vegetative and ¯oral characters unite Globba with Man- pantropical family. tisia and Hemiorchis with Gagnepainia, a result that is strongly supported by both the ITS and matK sequence data as well. LITERATURE CITED Our analysis shows that the Indo-Burmese Mantisia is sister to an Indo-Burmese clade of Globba suggesting that the latter BURTT, B. L. 1972. General introduction of papers on Zingiberaceae. Notes genus is paraphyletic; the relationships within this clade are from the Royal Botanic Garden Edinburgh 31: 155±165. BURTT,B.L.,AND R. M. SMITH. 1972. Tentative keys to the subfamilies, currently under investigation (K. J. Williams, W. J. Kress, and tribes and genera of the Zingiberales. Notes from the Royal Botanic Gar- P. S. Manos, unpublished data). Our ITS data provide only den Edinburgh 31: 171±176. weak support to unite the four genera, and the combined anal- BURTT,B.L.,AND R. M. SMITH. 1986. Etlingera: the inclusive name for ysis does not resolve this relationship. However, a more com- Achasma, Geanthus and Nicolaia (Zingiberaceae). Notes from the Royal prehensively sampled data set (K. J. Williams, W. J. Kress, Botanic Garden Edinburgh 43: 235±241. and P. S. Manos, unpublished data) suggests that the tribe, as DASSANAYAKE, M. D. 1983. A revised handbook to the ¯ora of Ceylon. Amerind Publishing, New Delhi, India. traditionally recognized, is in fact monophyletic. Because of DE QUEIROZ, A., M. J. DONOGHUE, AND J. KIM. 1995. Separate versus com- this additional evidence and in the interest of maintaining sta- bined analysis of phylogenetic evidence. Annual Review of Ecology and bility with past classi®cations we advocate retaining the Glob- Systematics 26: 657±681. 1696 AMERICAN JOURNAL OF BOTANY [Vol. 89

DOYLE,J.J.,AND J. L. DOYLE. 1987. A rapid DNA isolation procedure for In A. Engler and K. Prantl [eds.], Die NatuÈrlichen P¯anzenfamilien, 1st small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11±15. ed., vol. 2 (6), 1±43. Verlag von Wilhelm Engelmann, Leipzig, Germany. FELSENSTEIN, J. 1985. Con®dence intervals on phylogenies: an approach us- POULSEN, A. D., J. MOOD, AND H. IBRAHIM. 1999. Three new species of ing the bootstrap. Evolution 39: 783±791. Etlingera (Zingiberaceae) from Borneo. Nordic Journal of Botany 19: FUNAKOSHI, H., AND H. OHASHI. 2000. Vanoverberghia sasakiana H. Funk. 139±147. & H. Ohashi (Zingiberaceae), a new species and a new generic record RAMBAUT, A. 2000. Se-Al v2.0a3 Sequence Alignment Editor. University of for the Flora of Taiwan. Taiwania 45: 270±275. Oxford, Oxford, UK. HARRIS, D. J., A. D. POULSEN,C.FRIMODT-MOLLER,J.PRESTON, AND Q. C. RANGSIRUJI, A., M. F. NEWMAN, AND Q. C. B. CRONK. 2000a. Origin and B. CRONK. 2000. Rapid radiation in Aframomum (Zingiberaceae): evi- relationships of Alpinia galanga (Zingiberaceae) based on molecular dence from nuclear ribosomal DNA internal transcribed spacer (ITS) data. Edinburgh Journal of Botany 57: 9±37. sequences. Edinburgh Journal of Botany 57: 377±395. RANGSIRUJI, A., M. F. NEWMAN, AND Q. C. B. CRONK. 2000b. A study of HOLTTUM, R. E. 1950. The Zingiberaceae of the Malay Peninsula. Garden- the infrageneric classi®cation of Alpinia (Zingiberaceae) based on the er's Bulletin of Singapore 13: 1±249. ITS region of nuclear rDNA and the trnL-F spacer of chloroplast DNA. JOHNSON, L. A., AND D. E. SOLTIS. 1995. Phylogenetic inference in Saxifra- In K. L. Wilson and D. A. Morrison [eds.], MonocotsÐsystematics and gaceae sensu stricto and Gilia (Polemoniaceae) using matK sequences. evolution, 695±709. CSIRO Publishing, Collingwood, Australia. Annals of the Missouri Botanical Garden 82: 149±175. RAO,A.S.,AND D. M. VERMA. 1971. CurcumorphaÐa new genus of Zin- KIEW, K. Y. 1982. The genus Elettariopsis (Zingiberaceae) in Malaya. Notes giberaceae. Bulletin of the Botanical Survey of India 13: 339±341. from the Royal Botanic Garden Edinburgh 40: 139±152. REVEAL, J. 2002. Index Nominum Supragenericorum Plantarum Vascular- KRESS, W. J. 1990. The phylogeny and classi®cation of the Zingiberales. ium. http://matrix.nal.usda.gov:8080/cgi-bin/star®nder/0?pathϭsuprag. Annals of the Missouri Botanical Garden 77: 698±721. txt&idϭanon&passϭ&OKϭOK. KRESS, W. J. 1995. Phylogeny of the Zingiberanae: morphology and mole- SAKAI, S., AND H. NAGAMASU. 1998. Systematic studies of Bornean Zingi- cules. In P. Rudall, P. J. Cribb, D. F. Cutler, and C. J. Humphries [eds.], beraceae: I. Amomum in Lambir Hills, Sarawak. Edinburgh Journal of Monocotyledons: systematics and evolution, 443±460. Royal Botanic Botany 55: 45±64. Gardens, Kew, UK. SAKAI, S., AND H. NAGAMASU. 2000. Systematic studies of Bornean Zingi- KRESS,W.J.,AND K. LARSEN. 2001. Smithatris, a new genus of Zingibera- beraceae: III. Tamijia: a new genus. Edinburgh Journal of Botany 57: ceae from Southeast Asia. Systematic Botany 26: 226±230. 245±255. KRESS, W. J., L. M. PRINCE,W.J.HAHN, AND E. A. ZIMMER. 2001. Unrav- SCHUMANN, K. 1904. Zingiberaceae. In A. Engler [ed.], Das P¯anzenreich eling the evolutionary radiation of the families of the Zingiberales using 4/46, 1±458. Leipzig, Germany. morphological and molecular evidence. Systematic Biology 51: 926±944. SEARLE,R.J.,AND T. A. J. HEDDERSON. 2000. A preliminary phylogeny of LARSEN, K. 1964. Studies on Zingiberaceae IV: Caulokaempferia, a new ge- the Hedychieae tribe (Zingiberaceae) based on ITS sequences of the nu- nus. Botaniska Tidsskrift 60: 165±179. clear rRNA cistron. In K. L. Wilson and D. A. Morrison [eds.], Mono- LARSEN, K. 1997. Further studies in the genus Boesenbergia (Zingiberaceae). cotsÐsystematics and evolution, 710±718. CSIRO Publishing, Colling- Nordic Journal of Botany 17: 361±366. wood, Australia. LARSEN, K., AND T. J ENJITTIKUL. 2001. Laosanthus, a new genus of Zingi- SIRIRUGSA, P. 1992. A revision of the genus Boesenbergia Kuntz (Zingiberaceae) beraceae from Laos. Nordic Journal of Botany 21: 135±138. in Thailand. Natural History Bulletin of the Siam Society 40: 67±90. LARSEN, K., J. M. LOCK,H.MAAS, AND P. J. M. MAAS. 1998. Zingiberaceae. SMITH, R. M. 1972. The genus Burbidgea. Notes from the Royal Botanic In K. Kubitzki [ed.], The families and genera of vascular plants, vol. IV, Garden Edinburgh 31: 297±306. 474±495. Springer-Verlag, Berlin, Germany. SMITH, R. M. 1980. A new genus of Zingiberaceae from N. Burma. Notes LARSEN, K., AND J. MOOD. 1998. Siamanthus, a new genus of Zingiberaceae from the Royal Botanic Garden Edinburgh 38: 13±17. from Thailand. Nordic Journal of Botany 18: 393±397. SMITH, R. M. 1987. A review of the Bornean Zingiberaceae: III. (Hedy- LARSEN, K., AND R. M. SMITH. 1972. Notes on Caulokaempferia. Notes from chieae). Notes from the Royal Botanic Garden Edinburgh 44: 203±232. the Royal Botanic Garden Edinburgh 31: 287±295. SMITH, R. M. 1990a. Alpinia (Zingiberaceae): a proposed new infrageneric classi®cation. Edinburgh Journal of Botany 47: 1±75. MADDISON, D. R. 1991. The discovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40: 315±328. SMITH, R. M. 1990b. Psychanthus (K. Schum.) Ridley (Zingiberaceae): its acceptance at generic level. Edinburgh Journal of Botany 47: 77±82. MADULID, D. A. 1996. The family Zingiberaceae and the ¯ora of the Phil- SPECHT, C. D., W. J. KRESS,D.W.STEVENSON, AND R. DESALLE. 2001. A ippines project. In T.-L. Wu, Q.-G. Wu, and Z.-Y. Chen [eds.], Proceed- molecular phylogeny of Costaceae (Zingiberales). Molecular Phylogeny ings of the Second Symposium on the Family Zingiberaceae, 1±9. South and Evolution 21: 333±345. China Institute of Botany, Guangzhou, People's Republic of China. STEELE,K.P.,AND R. VILGALYS. 1994. Phylogenetic analyses of Polemon- MANOS,P.S.,AND K. P. STEELE. 1997. Phylogenetic analyses of ``higher'' iaceae using nucleotide sequences of the plastid gene matK. Systematic Hamamelididae based on plastid sequence data. American Journal of Botany 19: 126±142. Botany 84: 1407±1419. SWOFFORD, D. L. 1998. PAUP*: phylogenetic analysis using parsimony (* MERRILL, E. D. 1912. Sertulum Bontocense. New or interesting plants col- and other methods), version 4.0. Sinauer, Sunderland, Massachusetts, lected in Bontoc Subprovince, Luzon, by Father Morice Vanoverbergh. USA. Philippines Journal of Science, C. Botany 7: 71±75. THEILADE, I., AND J. MOOD. 1997. Five new species of Zingiber (Zingiber- MOOD, J., AND K. LARSEN. 1997. Cornukaempferia, a new genus of Zingi- aceae) from Borneo. Nordic Journal of Botany 17: 337±347. beraceae from Thailand. Natural History Bulletin of the Siam Society 45: TONG, S. Q. 1985. Paramomum, a new genus of Zingiberaceae from Yunnan. 217±221. Acta Botanica Yunnanica 7: 309±312. MORT, M. E., P. S. SOLTIS,D.E.SOLTIS, AND M. L. MABRY. 2000. Com- WHITE, T. J., T. BRUNS,S.LEE, AND J. TAYLOR. 1990. Ampli®cation and parison of three methods for estimating internal support for phylogenetic direct sequencing of fungal ribosomal RNA genes for phylogenies. In trees. Systematic Biology 49: 160±170. M. A. Innis, D. H. Gelfand, J. J. Sninsky, and T. J. White [eds.], PCR NEWMAN, M. F. 1990. A reconsideration of Brachychilum Peterson (Hedy- protocols: a guide to methods and applications, 315±322. Academic chicieae: Zingiberaceae). Edinburgh Journal of Botany 47: 83±87. Press, San Diego, California, USA. NEWMAN, M. F. 1995. Distichochlamys, a new genus from Vietnam. Edin- WILLIAMS, K. J., W. J. KRESS, AND THET TUN. 2002. A striking new epi- burgh Journal of Botany 52: 65±69. phytic Hedychium (Zingiberaceae) from Myanmar with a discussion of NGAMRIABSAKUL, C., M. F. NEWMAN, AND Q. C. B. CRONK. 2000. Phylog- several anomalous related genera. Edinburgh Journal of Botany, in press. eny and disjunction in Roscoea (Zingiberaceae). Edinburgh Journal of WOOD, T. H., W. M. WHITTEN, AND N. H. WILLIAMS. 2000. Phylogeny of Botany 57: 39±61. Hedychium and related genera (Zingiberaceae) based on ITS sequence NIXON, K. C., AND J. M. CARPENTER. 1996. On simultaneous analysis. Cla- data. Edinburgh Journal of Botany 57: 261±270. distics 12: 221±241. WU, T.-L. 1997. Notes on Lowiaceae, Musaceae, and Zingiberaceae for the PETERSEN, O. G. 1889. Musaceae, Zingiberaceae, Cannaceae, Marantaceae. Flora of China. Novon 7: 440±442.