The Molecular Phylogeny of Alpinia (Zingiberaceae): a Complex and Polyphyletic Genus of Gingers1

http://www.paper.edu.cn

American Journal of Botany 92(1): 167±178. 2005.

THE MOLECULAR PHYLOGENY OF ALPINIA (ZINGIBERACEAE): A COMPLEX AND POLYPHYLETIC GENUS OF GINGERS1

W. J OHN KRESS,2,3,5 AI-ZHONG LIU,2 MARK NEWMAN,4 AND QING-JUN LI3

2Department of Botany, MRC-166, United States National Herbarium, National Museum of Natural History, Smithsonian Institution, PO Box 37012, Washington, D.C. 20013-7012 USA; 3Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Mengla, Yunnan 666303 China; and 4Royal Botanic Garden, 20A Inverleith Row, Edinburgh EH3 5LR, Scotland, UK

Alpinia is the largest, most widespread, and most taxonomically complex genus in the Zingiberaceae with 230 species occurring throughout tropical and subtropical Asia. Species of Alpinia often predominate in the understory of forests, while others are important ornamentals and medicinals. Investigations of the evolutionary relationships of a subset of species of Alpinia using DNA sequence- based methods speci®cally test the monophyly of the genus and the validity of the previous classi®cations. Seventy-two species of Alpinia, 27 non-Alpinia species in the subfamily Alpinioideae, eight species in the subfamily Zingiberoideae, one species in the subfamily Tamijioideae, and three species in the outgroup genus Siphonochilus (Siphonochiloideae) were sequenced for the plastid matK region and the nuclear internal transcribed spacer (ITS) loci. Parsimony analyses of both individual and combined data sets identi®ed six polyphyletic clades containing species of Alpinia distributed across the tribe Alpinieae. These results were supported by a Bayesian analysis of the combined data set. Except in a few speci®c cases, these monophyletic groupings of species do not correspond with either Schumann's (1904) or Smith's (1990) classi®cation of the genus. Here we build on previous molecular analyses of the Alpinioideae and propose the next steps necessary to recognize new generic boundaries in the Alpinieae.

Key words: Alpinia; ¯existyly; gingers; ITS; matK; phylogeny; tropical; Zingiberaceae.

Alpinia Roxb. is the largest and most widespread genus in & S.J. Chen and A. blepharocalyx K. Schum.) the plants form the Zingiberaceae with some 230 species occurring from Sri large stands in the understory, along forest margins, and in Lanka and the Western Ghats of India to China, Japan, all of light gaps, while other species are dominant in wetlands and southeast Asia, the Paci®c as far as Fiji, Samoa, and the Car- along water courses [e.g., A. nigra (Gaertn.) B.L. Burtt]. Al- oline Islands, and Australia as far south as northern New South though most alpinias are pollinated by large bees, some spe- Wales (Larsen et al., 1998; Smith, 1990). Most species grow cies attract birds and even bats as pollinators (Zhang et al., in low- to mid-elevation forests and form clumps with stems 2003; Kress and Specht, in press). Flexistyly, a novel ¯oral from 1±3 m high, although species east of Wallace's Line tend mechanism promoting outcrossing in which styles move up or to grow much larger. Alpinia regia R.M. Sm. of the Moluccas down depending on the timing of anther dehiscence, has been and A. boia Seem. of Fiji, for example, reach over 8 m in described in a number of species of Alpinia (Li et al., 2001, height. Some species are found in montane forests up to 2000 2002; Zhang et al., 2003). m above sea level in New Guinea and Sulawesi. However, The generic name Alpinia was ®rst used by Linnaeus for very few are tolerant of frost. The most northerly species is Alpinia racemosa, a neotropical species. Many Asiatic species Alpinia japonica (Thunb.) Miq., which survives north of To- were added to Alpinia, while later authors tended to refer kyo where the winters can be severe. Several species are im- American species to Renealmia L.f. Schumann (1904) ®nal- portant ornamentals (e.g., A. purpurata (Vieill.) K. Schum.) as ized these taxonomic concepts and subsequently Alpinia Roxb. potted plants, landscape accents, and cut ¯owers, and at least was conserved for the Asiatic species with Alpinia galanga one (A. zerumbet (Pers.) B.L. Burtt & R.M. Sm.) is naturalized (L.) Willd. as its type. in tropical regions around the world. In Asia, especially China Alpinia is the type genus of the tribe Alpinieae A. Rich. of (Wu and Larsen, 2000), alpinias are used as medicinals (e.g., the family Zingiberaceae. This tribe consists of evergreen A. of®cinarum Hance) and in cooking [A. galanga (L.) Willd.]. herbs, in which an abscission layer between the rhizome and Alpinias play an important ecological role in the understory the leafy shoots is lacking, the plane of distichy of the leaves of tropical and subtropical forests where many species are is transverse to the direction of growth of the rhizome, and quite common. In some cases (e.g., A. kwangsiensis T.L. Wu the lateral staminodes of the ¯owers are small, reduced to swellings at either side of the base of the labellum, or are 1 Manuscript received 26 March 2004; revision accepted 16 September entirely absent. Extra¯oral nectaries are absent, and the fruit 2004. The authors thank Ray Baker, Mike Bordelon, Jiang-Yun Gao, Mary Gibby, is usually spherical and indehiscent or ¯eshy (Kress et al., David Harris, Kai Larsen, Jing-Ping Liao, Ida Lopez, Achariya Rangsiruji, Chel- 2002). sea Specht, and Yong-Mei Xia, for discussion, assistance, and tissue samples Within the tribe Alpinieae, generic limits are dif®cult to dis- that made this investigation possible. This work was funded by the key project cern. While some genera may be easily recognized by their of the Ministry of Science and Technology of China (2001CCA00300), the respective morphological characters and/or geographic distri- National Natural Science Foundation of China Grants 30225007 and 30170069, the Smithsonian Scholarly Studies Program and the Biotic Surveys and Inven- bution (e.g., Aframomum, Elettaria, Hornstedtia, Burbidgea), tories Program of the National Museum of Natural History. it is hard to identify an apomorphy or universal character for 5 E-mail: [email protected]. species currently assigned to Alpinia. Virtually all species

167 转载 中国科技论文在线 http://www.paper.edu.cn

168 AMERICAN JOURNAL OF BOTANY [Vol. 92

Figs. 1±9. Representative ¯oral types of the major groups of species of Alpinia. 1. Alpinia nigra. 2. A. galanga. 3. A. conchigera. 4. A. carolinensis. 5. A. zerumbet. 6. A. guangdongensis. 7. A. calcarata. 8. A. oxyphylla. 9. A. elegans.

¯ower terminally on the leafy shoots and all are Asiatic. These in¯orescence and ¯ower characters. Much variation exists in characters distinguish Alpinia from the Afro-American Re- these features (Figs. 1±9), from species with branched in¯o- nealmia, in which most species produce in¯orescences on a rescences and long cincinni subtended by bracts in which the separate, lea¯ess shoot from the rhizome, but do not uniquely ¯owers are each subtended by bracteoles, to other species with separate it from other members of the Alpinieae. Therefore, to no bracts or bracteoles and cincinni of only a single ¯ower. In a large degree, one is forced to recognize Alpinia only by the Flora of British India, Baker (1894) considered the species eliminating other genera, i.e., it is distinguished only by the of Alpinia that occur from Sri Lanka to Singapore. His account plesiomorphic characters of the tribe. included descriptions of 17 species from a known total of 30 Several attempts have been made to divide Alpinia into at that time and divided them into two subgenera and two smaller genera by elevating some of the more coherent groups sections according to the presence of an anther crest, the pos- of species to the generic rank. Holttum (1950) applied the session of large bracteoles, and the position of the in¯ores- name Alpinia to a small group of species with funnel-shaped cence. Schumann (1904) treated Alpinia throughout its range bracteoles and allocated the remaining species to Catimbium, in his account of the Zingiberaceae for Das P¯anzenreich (Ta- Cenolophon, and Languas. Several nomenclatural problems ble 1) dividing it into ®ve subgenera and 27 sections. Eight were present in this system, but its principal failing was that of Schumann's sections have now been placed in entirely dif- it only worked for the species of Malaysia. Later authors, ferent genera, leaving Alpinia with ®ve subgenera and 19 sec- therefore, returned to the concept of Alpinia sensu Schumann tions. Characteristics of the bracts and bracteoles are the most until Smith (1990) recognized a group of 22 species in New important diagnostic features in Schumann's classi®cation. Guinea that she segregated under the generic name Pleuran- Valeton (1913) later divided section Eubractea into subsection thodium (K. Schum.) R.M. Sm. Eustales and subsection Kolowratia and added a new section Infrageneric classi®cations of Alpinia have been based on Monanthocrater to Schumann's 1904 system. He admitted that 中国科技论文在线 http://www.paper.edu.cn

January 2005] KRESS ET AL.ÐPHYLOGENY OF ALPINIA 169

TABLE 1. Infrageneric classi®cation of Alpinia according to Schumann (1904).

Accepted generic Corresponding clades of Subgenus Section position molecular analyses Autalpinia Pycnopyramis ϭ Plagiostachys Zerumbet (IV) Leptosolenia ϭ Leptosolena Eubractea (V) Hellenia ϭ Alpinia Galanga (II), Zerumbet (IV), Eubractea (V) Psychanthus ϭ Pleuranthodium Tribe Riedelieae Cenolophon ϭ Alpinia not represented Pleuranthodium ϭ Pleuranthodium Tribe Riedelieae Guillania ϭ Alpinia Eubractea (V) Probolocalyx ϭ Alpinia Zerumbet (IV) Catimbium Flos Paradisi ϭ Alpinia Zerumbet (IV) Boniophyton ϭ Alpinia Zerumbet (IV) Dieramalpinia Allughas ϭ Alpinia Eubractea (V) Pycnanthus ϭ Alpinia Carolinensis (III) Amomiceps ϭ Alpinia not represented Medusula ϭ Alpinia not represented Eubractea ϭ Alpinia Eubractea (V) Cylindrobotrys ϭ Alpinia Carolinensis (III) Myriocrater ϭ Alpinia not represented Strobidia ϭ Alpinia Galanga (II) Brachybotrys ϭ Alpinia Raf¯esiana (VI) Javana ϭ Alpinia Raf¯esiana (VI) Oligocincinnus ϭ Alpinia Carolinensis (III) Monopleura ϭ Alpinia Carolinensis (III) Rhizalpinia Coralliophyton ϭ Riedelia Tribe Riedelieae Botryamomum ϭ Alpinia not represented Geocharis ϭ Geocharis Tribe Alpinieae Didymanthus ϭ Alpinia Zerumbet (IV) Cylindrostachys ϭ Amomum & Plagiostachys Zerumbet (IV) Bintalua ϭ Amomum & Plagiostachys Zerumbet (IV)

section Monanthocrater was not sharply distinct from Pyc- tigations by Rangsiruji et al. (2000a, b) and Kress et al. (2002) nanthus. The following year, Valeton (1914) added another are most pertinent. In the former study, in which 47 species new section, subgenus Autalpinia section Presleia. of Alpinia and a small number of outgroup taxa were sampled, By contrast, Smith (1990) only recognized two subgenera, the authors demonstrated signi®cant statistical support for sev- based on characters of the labellum (Figs. 1±9). Subgenus Al- eral monophyletic groups of species of Alpinia, but suggested pinia was divided into seven sections and 10 subsections, that the genus may not be monophyletic. In a broader analysis while subgenus Dieramalpinia had four sections and two sub- of the genera in the Alpinioideae, Kress et al. (2002) identi®ed sections (Table 2). Bract and bracteole characters were used four separate groups of alpinias (Alpinia I±IV) in the 11 spe- to delimit the sections and subsections. Stigma types also pro- cies they sampled (Fig. 10). These four groups did not form vided some support for this classi®cation system. Smith's clas- a monophyletic assemblage, were scattered throughout the si®cation is the one most often recognized and used today. tribe, and corresponded to at least some of the clades recog- Recently, several papers have used molecular data to ex- nized in the molecular analyses of Rangsiruji et al. (2000b). plore phylogenetic relationships within the family (Searle and Neither the results of the Rangsiruji et al. (2000a, b) nor the Hedderson, 2000; Wood et al., 2000; Kress et al., 2002; Ped- Kress et al. (2002) investigations supported the classi®cation ersen, 2004) as well as within several genera (Hedychium: of Alpinia proposed by Smith (1990). Alpinia galanga (see Wood et al., 2000; Alpinia: Rangsiruji et al., 2000a, b; Ros- Fig. 2) and A. conchigera (see Fig. 3; forming Alpinia Iof coea: Ngamriabsakul et al., 2000; Aframomum: Harris et al., Kress et al. [2002; their ®g. 10] in a clade with Aframomum, 2000; Globba: Williams et al., 2004; Amomum: Xia et al., Renealmia, Amomum, Elettariopsis, and Paramomum; part of 2004). the A. galanga clade of Rangsiruji et al. [2000a, b]) are placed The study by Kress et al. (2002) is the most thorough paper in two separate sections of subgenus Alpinia by Smith. Alpinia to date addressing the relationships among genera in the Zin- II of Kress et al. (2002; their ®g. 10), which includes A. ele- giberaceae. In that study, sequence data from both the Internal gans (see Fig. 9), A. luteocarpa, and A. vittata; (part of the A. Transcribed Spacer (ITS) loci and matK regions were used to eubractea clade of Rangsiruji et al., 2000a, b), as well as the establish, for the most part, well-resolved phylogenetic rela- genus Vanoverberghia (from the Philippines and Taiwan), is tionships among the genera, and a new classi®cation of the allied with species of Etlingera, Hornstedtia, and some Amo- Zingiberaceae was proposed that recognized four subfamilies mum, also encompasses both of Smith's subgenera and two of and four tribes. They also demonstrated that a number of the her sections. Alpinia III of Kress et al. (2002, their ®g. 10; six larger genera in the family (Amomum, Alpinia, Etlingera, Boe- species included in four of Rangsiruji et al.'s [2000a, b] clades) senbergia, and Curcuma) may be para- or polyphyletic and is united with the genus Plagiostachys and species of three of suggested that more extensive sampling is necessary for these Smith's sections; Alpinia IV of Kress et al. (2002, their ®g. taxa, which has subsequently been carried out in some of them 10; the single species A. carolinensis [see Fig. 4] found in (Pedersen, 2004; Xia et al., 2004). Rangsiruji et al.'s [2000a, b] A. carolinensis clade) is unre- With respect to the genus Alpinia, the results of the inves- solved with Alpinia II and III. The highly polyphyletic nature 中国科技论文在线 http://www.paper.edu.cn

170 AMERICAN JOURNAL OF BOTANY [Vol. 92

TABLE 2. Infrageneric classi®cation of Alpinia according to Smith (1990).

Corresponding clades of molecular Subgenus Section Subsection analyses Alpinia Alpinia Alpinia Galanga (II), Zerumbet (IV) Catimbium Zerumbet (IV) Cenolophon Zerumbet (IV) Paniculatae Zerumbet (IV) Presleia Galanga (II), Zerumbet (IV) Probolocalyx Zerumbet (IV) Didymanthus Zerumbet (IV) Kolowratia Eubractea (V) Fax Fax (I) Guillania Eubractea (V) Arcti¯orae Eubractea (V) Allughas Allughas Galanga (II), Raf¯esiana (VI) Caeruleae Eubractea (V) Odontychium not represented Strobidia Galanga (II) Dieramalpinia Dieramalpinia Carolinensis (III), Zerumbet (IV), Eubractea (V) Eubractea Eubractea (V) Myriocrater Carolinensis (III), Zerumbet (IV) Pycnanthus Pycnanthus Carolinensis (III) Amomiceps not represented

of Alpinia as demonstrated by these molecular systematic in- (2002) were included in the analyses for a total of 112 species (Appendix, vestigations is congruent with the absence of any recognized see Supplemental Data accompanying online version of this article). In our morphological apomorphies for the genus as mentioned earlier. selection of Alpinia species, both Schumann's (1904) and Smith's (1990) clas- Our goals in the present study, which uses additional mo- si®cations were well sampled. All ®ve of Schumann's subgenera are repre- lecular sequence data from an expanded taxon sampling of the sented in our analyses, as are 14 of the 19 sections still considered to be genus Alpinia, together with a wide range of taxa of the Al- alpinias. Eight of Schumann's sections have been transferred to different gen- pinioideae included in the investigation of Kress et al. (2002), era in the family, and we have sampled seven of those eight sections. Rep- are (1) to obtain a better understanding of the phylogenetic resentatives of both subgenera and all 11 sections of Smith's classi®cation relationships of the species now taxonomically placed in this were sampled. As demonstrated by Kress et al. (2002) Siphonochilus (three genus, (2) to further test the monophyly of the genus as well species) was designated as the outgroup to all remaining taxa in the Zingi- as of the groups of species identi®ed in the earlier analyses, beraceae in our analyses. In addition, representatives of the other two sub- families were included to polarize character states in the Alpinioideae. and (3) to evaluate both Schumann's (1904) and Smith's (1990) classi®cations with respect to our phylogenetic results. Molecular methodsÐSequences for the plastid matK-trnK ¯anking inter- genic spacer regions and the nuclear internal transcribed spacer (ITS) loci MATERIALS AND METHODS were obtained for each taxon either from GenBank or generated according to the following method. Total genomic DNAs were extracted from fresh or TaxaÐSeventy-two taxa of Alpinia (including 47 species previously ana- silica dried tissue using either a minor modi®cation of Doyle and Doyle lyzed by Rangsiruji et al. [2000a, b] and Kress et al. [2002]), together with (1987) CTAB (hexadecyltrimethylammonium bromide) method or a DNeasy 27 non-Alpinia species in the subfamily Alpinioideae (including the three new Plant Mini kit (Qiagen, Valencia, California, USA) extraction protocol. The taxa Pleuranthodium ¯occosum, P. trichocalyx, and Leptosolena haenkei), aqueous phase was extracted with 24 : 1 chloroform/isoamyl alcohol, and eight species in the subfamily Zingiberoideae, one species in the subfamily DNA was resuspended in Tris-ethylenediaminetetraacetic acid (TE) buffer fol- Tamijioideae, and three species in the Siphonochiloideae from Kress et al. lowing isopropyl alcohol precipitation with the CTAB method. DNeasy ex- traction followed the manufacturer's protocols. Ampli®cation of ITS was ac- complished using either primer pair ITS4 and ITS5 (White et al., 1990) or ITS4 and ITS5a (Stanford et al., 2000). The plastid matK region was ampli®ed with trnK1F (Manos and Steele, 1997) and trnK2R (Steele and Vilgalys, 1994). All ampli®cations used Taq DNA polymerase (Carlsbad, California, USA) according to the manufacturer's direction with annealing temperatures of 54±58ЊC. Ampli®ed products were puri®ed using the Qiagen Qiaquick (Valencia, California, USA) puri®cation protocol with the products sequenced directly using automated sequencing methodology of the ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction kit (Foster City, California, USA). Sequencing primers included the ampli®cation primers plus ITS2 (White et al., 1990) and ITS3G (Kress et al., 2002) as necessary for the ITS region. Zingiberaceae speci®c internal matK primers used were mSP2F, mIF, m5Fa, m8Fa, mSP2R, mIR, m5R, and m8R (Steele and Vilgalys, 1994; Kress et al., 2002). Products were cleaned in Sephadex G-50 (®ne) Centri-Sep spin Fig. 10. The phylogenetic relationships among the genera of the subfam- columns (Princeton Separations P/N 901, Adelphia, New Jersey, USA), dried ily Alpinioideae of the Zingiberaceae based on a parsimony analysis of ITS under vacuum, and run on ABI 3100 Automated Sequencer (Perkin Elmer, and matK sequence data (modi®ed from Kress et al., 2002). Note the two Applied Biosystems, Inc., Foster City, California, USA) at the Smithsonian tribes and four polyphyletic clades of Alpinia. Institution's Laboratory for Analytic Biology. 中国科技论文在线 http://www.paper.edu.cn

January 2005] KRESS ET AL.ÐPHYLOGENY OF ALPINIA 171

Raw forward and reverse sequences for each sample were assembled, am- (bootstrap value ϭ 97%), whereas the remaining taxa com- biguous bases were corrected, and consensus sequences were edited using prising the tribe Alpinieae have only poor bootstrap support Sequencher 4.1 (Gene Codes Corporation, Ann Arbor, Michigan, USA). Con- (Ͻ50%) as a monophyletic group. sensus sequences for ITS and matK were manually aligned in Se-Al 2.0a11 Within the Alpinieae, the genus Alpinia forms six polyphy- (Rambaut, 2000). All regions of ambiguous alignment within the ITS regions letic clades of species each with varying statistical support. were excluded and gaps were treated as missing data. Clades I, III, and VI are each strongly supported as monophy- letic (bootstrap values Ͼ 90%), whereas Clade V (bootstrap Phylogenetic analysesÐMaximum parsimony analyses of the ITS and values ϭ 59%) and Clades II and IV (bootstrap values Ͻ 50%) matK sequence data were conducted using PAUP* 4.0b10 (Swofford, 2002) are weakly to poorly supported groups. The ITS results pro- with equally weighted characters and 1000 random-sequence-addition repli- vide no resolution of the relationships among these six clades cates, saving all shortest trees under ACCTRAN optimization, with the op- of Alpinia. tions tree bisection-reconnection (TBR) branch swapping, STEEPEST DE- SCENT off, MULTREES on, and COLLAPSE branches if maximum length is zero (AMB). Multiple random-sequence additions were used to search for matKÐThe 5Ј trnK-matK intergenic spacer region had a multiple tree islands (Maddison, 1991). The data sets for each gene region total aligned length of 1005 bp (unaligned sequences ranged were analyzed separately (111 taxa in the ITS analysis and 105 taxa in the from 789 to 854 bp) with a mean GC content of 29.99%; the matK analysis; see Appendix) and then, following the total evidence approach matK-coding region had an aligned length of 1636 bp (range for multiple data sets (de Queiroz et al., 1995; Nixon and Carpenter, 1996), of 1551±1564 bp) and GC of 29.69%; and the aligned length the sequence data were combined. Incongruence between the ITS and matK of matK-3Ј trnK intergenic spacer region was 402 bp (range data sets was assessed using the incongruence length difference (ILD) test of 257±279 bp) with a GC of 29.97%. (Farris et al., 1994) as implemented in PAUP*. The analysis of the matK region (coding and noncoding) Support for the nodes resolved in the strict consensus of the most parsi- resulted in 14 433 equally parsimonious trees of 1202 steps monious trees was evaluated with bootstrap analyses (Felsenstein, 1985; Mort (number of parsimony-informative characters ϭ 486; CI ϭ et al., 2000) using PAUP* with TBR branch swapping on 1000 bootstrap 0.522; RI ϭ 0.808; rescaled CI ϭ 0.422; Fig. 12). A strict replicates. Bootstrap support was categorized according to Kress et al. (2002) criteria, i.e., strong (Ͼ85%), moderate (70±85%), weak (50±70%), or poor consensus of these shortest trees (Fig. 12) provides strong sup- (Ͻ50%) support. port for the monophyly of both subfamilies Zingiberoideae and Alpinioideae (bootstrap values ϭ 100%) with Tamijia placed A Bayesian analysis using MRBAYES, version 3.0 (Huelsenbeck and Ron- ϭ quist, 2001) was performed using the same combined ITS-matK parsimony as outgroup to the latter subfamily (bootstrap value 62%). matrix. The most appropriate molecular model for each data set was deter- Although strong support is provided for the tribe Riedelieae mined with Modeltest, version 3.06 (Posada and Crandall, 1998). A general (minus Siamanthus), the Alpinieae (minus Siliquamomum as time reversible model (rates ϭ gamma, nst ϭ 6) was used for both ITS and in the ITS analysis) is only weakly supported (bootstrap value matK. Data from ITS and matK were partitioned (using the ``lset apply to'' ϭ 63%). Within the Alpinieae, the identical six clades of spe- command) in order to accommodate differing evolutionary rates for the re- cies of Alpinia vary in support from weak (clade V) to mod- spective data sets. Four Markov chain Monte Carlo (MCMC) chains, one cold erate (clade II) to strong (clades I, III, IV, and VI). Clade VI and three heated, were performed. Four MCMC runs of one million genera- is placed with clades I and II, and clades III, IV, and V are tions each, starting from different random points in parameter space, were united, but all with only poor statistical support (bootstrap val- performed in order to more fully explore tree space and stationarity of param- ue Ͻ 50%). eters (e.g., Miller et al., 2002; Jordan et al., 2003) to verify consistency in our results. Trees were sampled every 100th cycle from the chain. All sample points that occurred before stationarity of negative log likelihood (ϪlnL) Combined data setÐThe two separate molecular data sets scores was achieved were discarded as part of the burn-in period (Huelsen- passed the ILD test (P ϭ 0.07) supporting the total evidence beck and Ronquist, 2001). Nodes with posterior probability values Ն95% approach to combining the molecular evidence. The analysis were retained in the 50% majority rule consensus tree. of the combined ITS and matK sequence data resulted in 3390 equally parsimonious trees of 2759 steps (number of parsi- RESULTS mony-informative characters ϭ 799; CI ϭ 0.432; RI ϭ 0.763; rescaled CI ϭ 0.330; Figs. 13±14). A strict consensus of these Internal transcribed spacerÐITS-1 had a total aligned 3390 shortest trees (Fig. 13) provides strong support (bootstrap length of 244 bp (unaligned sequences ranged from 201 to values ϭ 92±100%) for the placement of Tamijia as sister to 231 bp) with a mean GC content of 54.06%, the 5.8S region all other Zingiberaceae (except the outgroup genus Siphono- had an aligned length of 164 bp (range of 163±164 bp) and chilus) as well as the monophyly of the Zingiberoideae, the GC of 51.18%, and the ITS-2 aligned length was 304 bp Alpinioideae, the Riedelieae (including Siamanthus), and the (range of 253±281 bp) with GC of 59.52%. Alpinieae (including Siliquamomum). Clades I, II, III, and VI The analysis of the ITS sequence data resulted in 36 300 of Alpinia are strongly supported as monophyletic (bootstrap equally parsimonious nearly fully resolved trees of 1461 steps value ϭ 94±100%). Clade IV is only poorly supported if A. (number of parsimony-informative characters ϭ 313; consis- oxymitra is included, yet moderately supported (bootstrap val- tency index [CI] ϭ 0.386; retention index [RI] ϭ 0.752; re- ue ϭ 83%) internal to this taxon. The combined data only scaled CI ϭ 0.290; Fig. 11). In a strict consensus of these weakly support the species of clade V as monophyletic (boot- 36 300 shortest trees, Tamijia is strongly supported (bootstrap strap value ϭ 61%), but strongly support this clade (bootstrap value ϭ 100%) as a stem lineage one node above the outgroup ϭ 100%) when united with the genera Etlingera, Hornstedtia, Siphonochilus. The Zingiberoideae (here represented by eight and Amomum. Clades I and II are moderately supported (boot- taxa) is poorly supported as monophyletic (bootstrap value Ͻ strap value ϭ 76%) as belonging to the same monophyletic 50%), but separate from the Alpiniodieae with weak support group with Aframomum, Renealmia, Amomum, Elettariopsis, (bootstrap value ϭ 69%). The tribe Riedelieae (here repre- and Paramomum, whereas clades III, IV, and V are only weak- sented by eight taxa) is strongly supported as monophyletic ly to poorly united with each other. 中国科技论文在线 http://www.paper.edu.cn

172 AMERICAN JOURNAL OF BOTANY [Vol. 92

Fig. 11. The strict consensus of 36 300 equally parsimonious trees of the Zingiberaceae with an emphasis on the Alpinieae in the analysis of the ITS sequence data (length ϭ 1461; CI ϭ 0.386 excluding uninformative characters; RI ϭ 0.752; and rescaled CI ϭ 0.290) showing bootstrap values (below the line if Ն50%). The six major clades of Alpinia and non-Alpinieae tribes and subfamilies of the Zingiberaceae are indicated. Abbreviations: ITS, internal transcriber spacer; CI, consistency index; RI, retention index. 中国科技论文在线 http://www.paper.edu.cn

January 2005] KRESS ET AL.ÐPHYLOGENY OF ALPINIA 173

Fig. 12. The strict consensus of 14 433 equally parsimonious trees of the Zingiberaceae with an emphasis on the Alpinieae in the analysis of the matK region (coding and noncoding) sequence data (length ϭ 1202; CI ϭ 0.522 excluding uninformative characters; RI ϭ 0.808; and rescaled CI ϭ 0.422) showing bootstrap values (below the line if Ն50%). The six major clades of Alpinia and non-Alpinieae tribes and subfamilies of the Zingiberaceae are indicated. Abbreviations: CI, consistency index; RI, retention index. 中国科技论文在线 http://www.paper.edu.cn

174 AMERICAN JOURNAL OF BOTANY [Vol. 92

Fig. 13. The strict consensus of 3390 equally parsimonious trees of the Zingiberaceae with an emphasis on the Alpinieae in the analysis of the combined ITS and matK region sequence data (length ϭ 2759; CI ϭ 0.432 excluding uninformative characters; RI ϭ 0.763; and rescaled CI ϭ 0. 330) showing bootstrap values from the parsimony analysis (below the line if Ն50%) and posterior probability values resulting from the Bayesian analysis (above the line if Ն95). The six major clades of Alpinia and non-Alpinieae tribes and subfamilies of the Zingiberaceae are indicated. Species (or non-Alpinia genera) that have been documented to possess the plant mating system ¯existyly are designated with an asterisk. Abbreviations: ITS, internal transcriber spacer; CI, consistency index; RI, retention index. 中国科技论文在线 http://www.paper.edu.cn

January 2005] KRESS ET AL.ÐPHYLOGENY OF ALPINIA 175

Bayesian analysis resultsÐThe 95% majority rule consen- African/neotropical Renealmia. A third species, A. rufescens sus of 9600 trees (10 000 trees minus 400 burn-in trees) re- (not sampled here), with similar characteristics that may be sulting from the Bayesian analysis of the combined data set is appropriately included in the Fax clade, is only known from highly congruent with the strict consensus of the parsimony the type and is in need of further study. Members of this small analysis of the combined data sets (Fig. 13). Each of the six clade are geographically well circumscribed, occurring in Sri Alpinia clades had a posterior probability value of 98±100%. Lanka and a small part of SW India. The occurrence of the The only signi®cant exceptions being the placement of Sili- closest relatives of this Indian subcontinent clade in Africa quamomum as sister to the tribe Alpinieae and the unresolved suggests that the common ancestor of the Fax clade may have placement of Alpinia oxymitra in either clade IV or clade V. ``drifted'' across the Indian Ocean with the breakup of Gond- At almost all nodes, posterior probability values were equal to wana. or greater than bootstrap values in the parsimony results. Clade II, the Galanga clade, includes species from three of Schumann's sections and two of Smith's sections (Figs. 1±3). DISCUSSION The four species included in this clade are mostly found in continental Asia although A. bilamellata comes from the Bo- The six clades of AlpiniaÐAlpinia has always been con- nin Islands. From the illustration in the Flora of the Bonin sidered to be a taxonomically dif®cult and complex genus, Islands (Toyoda, 1981, plate 104), A. bilamellata appears to both in de®ning the characters that distinguish the genus from be very similar to A. galanga. The Galanga clade consists of other genera in the Alpinioideae and in classifying its species Rangsiruji et al.'s A. galanga clade with the addition of A. (Schumann, 1904; Smith, 1990; Larsen et al., 1998). The re- bilamellata. The four species of our Galanga clade form a sults of our molecular analyses provide new insights into the relatively coherent group possessing cincinni made up of phylogenetic basis for these taxonomic problems. Although we many, small ¯owers with a similar labellum shape. The rela- have sampled only approximately one-third of the described tionship of A. galanga with A. conchigera and A. nigra is also species of Alpinia, the phylogenetic patterns that emerge supported by Liao and Wu (1996), who studied fruit wall anat- strongly suggest that the genus is highly polyphyletic in the omy. The con¯ict between our results and Schumann's and tribe Alpinieae. These results support the earlier phylogenetic Smith's classi®cations arises from the nature of the bracteoles, investigations based on the ITS and the trnL-F spacer region which are tubular in A. conchigera and A. nigra, but open to of Rangsiruji et al. (2000a, b) in which a more limited number the base in A. galanga and A. bilamellata. Our results suggest of taxa were sampled in both the ingroup and the outgroup. that this character can reverse states within a clade. The dis- They suggested that the genus may not be monophyletic and parate position of the two accessions of A. nigra is somewhat recognized nine major clades. However, because their out- problematic and may be due to the widespread distribution of group taxa were restricted to only two genera in the Alpinieae, this species with signi®cant local differentiation. Alpinia gal- Renealmia and Elettariopsis, the wide distribution of species anga, the type of the genus, is contained in this clade. Further of Alpinia throughout the tribe was not evident. Our results, taxonomic revisions may require that the name ``Alpinia'' be in which we sampled extensively (though not completely) restricted to members of this relatively small group of species. among the genera in the Alpinieae, strongly support the close Clade III, the Carolinensis clade, is made up of species that phylogenetic relationship of species of Alpinia to other genera share the characters of Smith's subgenus Dieramalpinia, such in the tribe, such as Renealmia, Aframomum, Amomum, Etlin- as the narrow, ¯eshy labellum adpressed to the stamen; tubular, gera, Hornstedtia, Leptosolena, Plagiostachys, Siliquamo- tightly clasping bracteoles; ¯owers in cincinni; and the persis- mum, and Vanoverberghia. In fact, we believe that a major tent primary bract (Fig. 4). Members of this clade belong to realignment of genera in the tribe, including the description of three of the four sections in Smith's classi®cation, namely My- new taxa, will be necessary if the taxonomy is to accurately riocrater (A. aenea, A. cylindocephala, A. monopleura, A. er- re¯ect evolutionary history (see later discussion). emochlamys, and A. coeruleoviridis), Pycnanthus (A. boia), The four clades of Alpinia recognized by Kress et al. (2002; and Dieramalpinia (A. carolinensis). Schumann also placed Fig. 10) and the nine clades described by Rangsiruji et al. these species, at least those that were known at the time, in (2000b) correspond closely to the six major clades de®ned in four sections of subgenus Dieramalpinia. These species are our analysis (Figs. 13±14). In most cases, our increased taxon disjunct between Fiji (A. boia), the Caroline Islands (A. car- sampling and additional sequence data from the matK region olinensis), and Sulawesi (remaining ®ve species). Subgenus have provided stronger bootstrap support for each of their Dieramalpinia sensu Smith occurs east of Wallace's Line, with clades. The current analyses, building on their earlier work, very few exceptions in Borneo. The species of this area are now include representative taxa from all of Smith's 11 sections far less well known than those taxa west of Wallace's Line. and 14 of Schumann's 19 sections now considered to be al- Clade IV, the Zerumbet clade (Figs. 5±8), is the largest pinias (although a few sections are only sparsely sampled). group of species in our analyses and includes four of Rang- We, therefore, believe that additional species not yet sampled siruji et al.'s clades (the A. zerumbet clade, A. polyantha clade, will most likely be contained within one of these six clades. A. glabra clade, and A. aquatica clade). The large Zerumbet Clade I, hereafter called the Fax clade (following the no- clade contains members of Smith's subgenus Alpinia section menclature of Rangsiruji et al., 2000a, b), contains two species Alpinia and section Didymanthus (A. pumila), subgenus Dier- that share capitate, usually radical in¯orescences surrounded amalpinia section Dieramalpinia and section Myriocrater (A. by sterile bracts with each lateral cincinnus composed of up vulcanica), and two species of the genus Plagiostachys. Al- to seven ¯owers. This combination of characters is not found though only poor bootstrap support exists for the Zerumbet in any other genus with which we are familiar. The strong clade as a whole, the lack of support may be due to the poorly bootstrap support for this clade and the distinctive morpholog- resolved position of Alpinia oxymitra, which is sister to all ical features may warrant that these species be recognized as remaining species in the clade. The Bayesian analysis also a distinct genus in a clade with the African Aframomum and suggests an unresolved position for this species in the Alpi- 中国科技论文在线 http://www.paper.edu.cn

176 AMERICAN JOURNAL OF BOTANY [Vol. 92

nieae. The node interior to A. oxymitra, which includes all of 19 sections contained in his ®ve subgenera. Two to ®ve of the remaining species of the Zerumbet clade, is moderately Schumann's sections are contained within ®ve of the six clades supported in the parsimony analysis and strongly supported in currently recognized in our molecular analyses (Table 1; Ap- the Bayesian analysis. Yet despite this statistical support for pendix). the molecular results, it is dif®cult to ®nd any morphological Smith's (1990) elegant and intricate classi®cation of Alpinia apomorphies of the Zerumbet clade. Similarly, no biogeo- was an attempt to provide a modern interpretation of the com- graphical patterns are apparent for this clade as a whole. plex array of species placed in this genus. Her two subgenera, Within the Zerumbet clade only two of the four clades of 11 sections, and 12 subsections encompassed the 221 species Rangsiruji et al. have signi®cant bootstrap support (the A. gla- known at that time. Four of our clades include taxa placed in bra clade and A. aquatica clade with bootstrap values of 98% from two to six of her sections. Only our Fax clade and Raf- and 81%, respectively). The A. glabra clade includes species ¯esiana clade correspond to a single section (section Fax)or found in Borneo and also encompasses the genus Plagios- subsection (section Allughas subsection Allughas) of Smith's tachys in our analysis. The A. aquatica clade is primarily re- classi®cation. stricted to the Philippines (except for A. aquatica) and is gen- It is clear that a signi®cant disparity exists between our phy- erally characterized by small ¯owers with a petaloid four- logenetic results and the taxonomic concepts of generic and lobed labellum. A third, strongly supported subclade consists infrageneric classi®cations of Schumann (1904) and Smith of three species, A. oxyphylla (Fig. 8), A. calcarata (Fig. 7), (1990). Unfortunately, at this time we do not yet have adequate and A. of®cinarum, the latter two of which are included in the morphological characters to support all of the results of our A. zerumbet clade of Rangsiruji et al. A more thorough anal- molecular analyses. Yet, some of the morphological characters ysis of the morphology and biogeography is needed for all the of previous taxonomists are clearly at odds with some of our subclades of the large and somewhat amorphous Zerumbet well-supported DNA sequence results. For example, both clade. Schumann and Smith gave great weight to the nature of the Clade V, the Eubractea clade, similar to the Zerumbet clade, bracteoles in delimiting genera. It is very rare in their classi- is dif®cult to characterize morphologically. No good characters ®cations to ®nd infrageneric sections or subsections with both de®ne this clade as a whole, although it contains several small- tubular and open bracteoles. Our results in the Galanga clade er clades with circumscribed geographical ranges in the Phil- con®rm those of Rangsiruji et al. (2000a) who found A. gal- ippines, Australia, the Bismarck Archipelago, and the tropical anga with open bracteoles to be most closely related to A. Paci®c. The well-supported clade of A. arundelliana, A. ca- conchigera and A. nigra with tubular ones. Clearly, this in¯o- erulea, and A. modesta contains only Australian species. A rescence character must be used with caution. clade found primarily in the Philippines is composed of three Another character used at the generic level is the position species of Alpinia (A. elegans [Fig. 9], A. pinetorum, and A. of the in¯orescence. Schumann (1904) recognized only three luteocarpa) plus Vanoverberghia sepulchrei and may encom- radical-¯owering species of Alpinia, namely, A. chrysogynia pass the poorly resolved Leptosolena haenkei also from the and A. melichroa in section Botryamomum and A. pumila in Philippines. The latter monotypic genus, which was not in- section Didymanthus, along with the majority of species with cluded in the analysis by Kress et al. (2002), was previously terminal in¯orescences; both of these sections were placed in considered to be extinct in the wild, but is now known to be subgenus Rhizalpinia. Smith (1975) transferred two more spe- quite common (Funakoshi et al., in press). A small Paci®c cies with radical in¯orescences, A. abundi¯ora and A. fax, Ocean clade (A. oceania, A. vittata, and A. purpurata) is also from Schumann's Amomum section Geanthus series Polyan- strongly supported. thae into her section Fax, arguing that the presence of a showy Clade VI, the Raf¯esiana clade, includes just two species, involucre of sterile bracts and cincinni of two to seven ¯owers A. javanica and A. raf¯esiana, distributed on the Sunda Shelf made it impossible to place them in Amomum. She also added in southern Thailand, peninsular Malaysia, Sumatra, and Java. A. rufescens from Schumann's subgenus Autalpinia section Smith placed both of these taxa in her subgenus Alpinia sec- Cenolophon to her section Fax. Our results indicate that it may tion Allughas subsection Allughas, while Schumann recog- be possible to recognize section Fax at the generic rank, al- nized each of the two species as separate sections Brachybo- though we have not been able to include a sample of A. ru- trys and Javana in the same part of his key sharing the feature fescens in our phylogenetic anaylses. of short cincinni with no more than six ¯owers. Although the Of Schumann's three radical-¯owering species, we now Raf¯esiana Clade is sister to the monotypic Vietnamese/Chi- know from plants cultivated in our research greenhouses (Ap- nese Siliquamomum, the bootstrap support for this relationship pendix) that the in¯orescence position of A. pumila is terminal is poor, and the Bayesian analysis places Siliquamomum basal on a leafy shoot (W. J. Kress and M. Bordelon, Smithsonian to the Alpinieae. Institution, unpublished data). This observation is con®rmed by Wu and Larsen (2000) who have seen living plants and do Previous classi®cations of AlpiniaÐThe genus Alpinia has not refer to the in¯orescence of A. pumila as radical. The ®nal assumed varying signi®cance in previous classi®cations of the two species of Schumann's section Botryamomum remain very Zingiberaceae. Schumann's (1904) comprehensive treatment poorly known and are only tentatively placed in Alpinia. Smith of the family included in Alpinia a large portion of what we (1990) thought they might belong to Amomum. This may be now recognize as subfamily Alpinioideae, including all or correct in the case of Alpinia chrysogynia, which has bracts parts of the genera Pleuranthodium, Riedelia, Geocharis, Pla- and bracteoles with single-¯owered cincinni. However, A. mel- giostachys, Leptosolena, and Amomum. The ®rst two genera ichroa lacks bracts and bracteoles, and may be closer to the are now placed in tribe Riedelieae, and the latter four, along radical-¯owering species of Riedelia. If both these species are with Alpinia and 10 other genera, are included in the tribe excluded from Alpinia, then all remaining species of the genus Alpinieae (Kress et al., 2002; Fig. 10; Table 1). We have sam- are terminal ¯owering. pled species currently placed in Alpinia in 14 of Schumann's In the search for new morphological characters useful in 中国科技论文在线 http://www.paper.edu.cn

January 2005] KRESS ET AL.ÐPHYLOGENY OF ALPINIA 177

though genera such as Aframomum, Renealmia, and Etlingera appear to be monophyletic (Harris et al., 2000; Kress et al., 2002; Pedersen, 2004), others such as Amomum are polyphy- letic and in need of further taxon sampling (Xia et al., 2004). In addition, the taxonomic uniqueness of such genera as Van- overberghia, Leptosolena, Plagiostachys, and Elettariopsis needs to be resolved. For all these reasons, we are reluctant at this time to propose a new classi®cation with rede®ned generic boundaries. However, our results together with the others pre- viously listed will provide the foundation for a revised clas- si®cation of the Alpinieae in the near future.

The ecological and evolutionary distribution of ¯existyly in AlpiniaÐOne purpose for determining phylogenetic history is to understand patterns of evolution of various morphological and ecological characteristics of taxa. Flexistyly, a unique ¯o- ral mechanism in plants that appears to promote outcrossing in the populations where it is found (Li et al., 2001; Renner, 2001; Barrett, 2002), was ®rst described in Amomum and Al- pinia, two genera that we now understand to be polyphyletic in the Zingiberaceae. Although our species sampling for mo- lecular phylogenetic analyses of both of these genera is mod- est, we have suf®cient data to make initial interpretations of the evolutionary origin of this angiosperm mating system. Flexistyly has now been documented in 24 species of the Zin- giberaceae (Cui et al., 1996; Li et al., 2001, 2002; Zhang et al., 2003; Q.-J. Li and W. J. Kress, Xishuangbanna Tropical Botanical Garden and Smithsonian Institution, unpublished data; Figs. 13, 14) and occurs in the Alpinia Galanga clade, the Alpinia Zerumbet clade, as well as several species of Amo- Fig. 14. Condensed tree of the Alpinioideae resulting from the analysis mum (A. koenigii, A. tsaoko, and A. paratsako) and Etlingera of the combined ITS and matK region sequence data (see Fig. 13) in which (E. yunnanensis) and possibly Plagiostachys and Paramomum monophyletic genera have been collapsed into single branches for clarity. (Cui et al., 1996). The distribution of ¯existyly in the family, Note the six clades of Alpinia (compare to Fig. 10). The Alpinia Zerumbet based on these preliminary results, suggests that this mating clade includes the genus Plagiostachys; the Alpinia Eubractea clade includes system may have evolved in the common ancestor of the tribe the genera Vanoverberghia and Leptosolena. Clades in which ¯existyly has been demonstrated are designated with an asterisk. Abbreviations: ITS, Inter- Alpinieae or independently at least three to ®ve times in the nal Transcriber Spacer. tribe (Figs. 13, 14). Both of these hypotheses are signi®cant in terms of understanding differential patterns of species di- versi®cation in the Zingiberaceae. As additional ®eld docu- classi®cation, a lead has been given by the careful study of mentation of this ¯oral mechanism becomes available, a more the fruit wall of Chinese Alpinia by Liao and Wu (1996) who thorough understanding of the prevalence of ¯existyly and its demonstrated a link between three species of the Galanga evolutionary origin in the gingers will be possible. clade. Their study should be extended to encompass the entire geographical range of the genus. LITERATURE CITED

A new classi®cation of the tribe AlpinieaeÐBased on our BAKER, J. G. 1894. Order CXLIX. Scitamineae. In J. D. Hooker [ed.], Flora current results and those of Rangsiruji et al. (2000a, b) for of British India, vol. VI, Orchideae to Cyperaceae, 198±264. L. Reeve Alpinia as well as the investigations of Kress et al. (2002) for & Co., London, UK. BARRETT, S. C. H. 2002. Sexual inference of the ¯oral kind. Heredity 88: the family Zingiberaceae, it is tempting to propose here a new 154±159. classi®cation of the tribe Alpinieae. The congruence between CUI, X.-L., R.-C. WEI, AND R.-F. HUANG. 1996. A study on the breeding the major clades of Alpinia in these analyses provides nearly system of Amomum tsao-ko. In T.-L. Wu, Q.-G. Wu, and Z.-Y. Chen unequivocal evidence that this genus is polyphyletic and that [eds.], Proceedings of the Second Symposium on the Family Zingiber- groups of species are more closely related to other genera in aceae, 288±296. South China Institute of Botany, Guangzhou, P. R. Chi- the tribe than they are to each other (Fig. 14). At least four of na. DE QUEIROZ, A., M. J. DONOGHUE, AND J. KIM. 1995. Separate versus com- the six clades are strongly supported in both the parsimony bined analysis of phylogenetic evidence. Annual Review of Ecology and and Bayesian analyses, yet the relationships among the clades Systematics 26: 657±681. are not fully resolved. It is quite unlikely that major suites of DOYLE,J.J.,AND J. L. DOYLE. 1987. A rapid DNA isolation procedure for species of Alpinia will be available for analysis in the near small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11±15. future due to their widespread distribution and the dif®culty FARRIS, J. S., M. KAÈ LLERSJOÈ ,A.G.KLUGE, AND C. BULT. 1994. Testing of obtaining tissue samples. Yet, we expect that a few addi- signi®cance of incongruence. Cladistics 10: 315±319. FELSENSTEIN, J. 1985. Con®dence intervals on phylogenies: an approach us- tional critical species may still be added to this data set. More- ing the bootstrap. Evolution 39: 783±791. over, the monophyly and phylogenetic position of a number FUNAKOSHI, H., W. J. KRESS,J.SKORNICKOVA, A.-Z. LIU, AND K. INOUE.In of genera in the Alpinieae have not yet been established. Al- press. Return from the lost: rediscovery of the presumed extinct Lep- 中国科技论文在线 http://www.paper.edu.cn

178 AMERICAN JOURNAL OF BOTANY [Vol. 92

tosolena (Zingiberaceae) in the Philippines and its phylogenetic place- RAMBAUT, A. 2000. Se-Al, version 2.0a3, Sequence alignment editor. Uni- ment in the gingers. Acta Phytotaxonomica et Geobotanica. versity of Oxford, Oxford, UK. HARRIS, D. J., A. D. POULSEN,C.FRIMODT-MùLLER,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- relationshiups 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 Alpina (Zingiberaceae) based on the ITS er's Bulletin of Singapore 13: 1±249. region of nuclear rDNA and the trnL-F spacer of chloroplast DNA. In HULSENBECK,J.P.,AND F. R. RONQUIST. 2001. MRBAYES: Bayesian infer- K. L. Wilson and D. A. Morrison [eds.], MonocotsÐsystematics and ence of phylogenetic trees. Bioinformatics 17: 754±755. evolution, 695±709. CSIRO Publishing, Collingwood, Australia. JORDAN, S., C. SIMON, AND D. POLHEMUS. 2003. Molecular systematics and RENNER, S. S. 2001. How common is heterodichogamy? Trends in Ecology adaptive radiation of Hawaii's endemic damsel¯y genus Megalagrion and Evolution 16: 595±597. (Odonata: Coenagrionidae). Systematic Biology 52: 89±109. SCHUMANN, K. 1904. Zingiberaceae. In A. Engler [ed.], Das P¯anzenreich KRESS,W.J.,AND C. D. SPECHT. In press. Between Cancer and Capricorn: 4, 1±458. Leipzig, Germany. phylogeny, evolution, and ecology of the tropical Zingiberales. In I. Friis SEARLE,R.J.,AND T. A. J. HEDDERSON. 2000. A preliminary phylogeny of and H. Balslev [eds.], Plant diversity and complexity patternsÐlocal, the Hedychieae tribe (Zingiberaceae) based on ITS sequences of the nu- regional and global dimensions. Biologiske Skrifter, The Royal Danish clear rRNA cistron. In K. L. Wilson and D. A. Morrison [eds.], Mono- Academy of Sciences and Letters, Copenhagen, Denmark. cotsÐsystematics and evolution, 710±718. CSIRO Publishing, Colling- KRESS, W. J., L. M. PRINCE, AND K. J. WILLIAMS. 2002. The phylogeny and wood, Australia. a new classi®cation of the gingers (Zingiberaceae): evidence from mo- SMITH, R. M. 1975. A preliminary review of the large bracteate species of lecular data. American Journal Botany 89: 1682±1696. Alpinia. Notes from the Royal Botanic Garden Edinburgh 34: 149±182. LARSEN, K., J. M. LOCK,H.MAAS, AND P. J. M. MAAS. 1998. Zingiberaceae. SMITH, R. M. 1990. Alpinia (Zingiberaceae): a proposed new infrageneric In K. Kubitzki [ed.], The families and genera of vascular plants, vol. 4, classi®cation. Edinburgh Journal of Botany 47: 1±75. 474±495. Springer-Verlag, Berlin, Germany. STANFORD, A. M., R. HARDEN, AND C. R. PARKS. 2000. Phylogeny and LI, Q.-J., Z.-F. XU,W.J.KRESS, Y.-M. XIA,L.ZHANG, X.-B. DENG, J.-Y. biogeography of Juglans (Juglandaceae) based on matK and ITS se- GAO, AND Z.-L. BAI. 2001. Flexible style that encourages outcrossing. quence data. American Journal of Botany 87: 872±882. Nature 410: 432. STEELE,K.P.,AND R. VILGALYS. 1994. Phylogenetic analyses of Polemon- LI, Q.-J., W. J. KRESS, Z.-F. XU, Y.-M. XIA,L.ZHANG, X.-B. DENG, AND J.- iaceae using nucleotide sequences of the plastid gene matK. Systematic Botany 19: 126±142. Y. G AO. 2002. Stigmatic behavior and the pollination biology of Alpinia kwangsiensis (Zingiberaceae). Plant Systematics and Evolution 232: SWOFFORD, D. L. 2002. PAUP*: phylogenetic analysis using parsimony 123±132. (*and other methods), version 4.0b10. Sinauer, Sunderland, Massachu- setts, USA. LIAO, J.-P., AND Q.-G. WU. 1996. Fruit anatomy of Chinese Alpinia and its TOYODA, T. 1981. Flora of Bonin Islands. Aboc-Sha, Kamakura, Japan. taxonomic signi®cance. In T.-L. Wu, Q.-G. Wu, and Z.-Y. Chen [eds.], VALETON, T. 1913. Zingiberaceae. Nova Guinea 8: 923±988. Proceedings of the 2nd Symposium on the Family Zingiberaceae, 82± VALETON, T. 1914. Die Zingiberaceen Deutsch-Neu-Guineas. Botanische 90. Zhongshan University Press, Guangzhou, P. R. China. JahrbuÈcher fuÈr Systematik, P¯anzengeschichte und P¯anzengeographie MADDISON, D. R. 1991. The discovery and importance of multiple islands 52: 41±100. of most-parsimonious trees. Systematic Zoology 40: 315±328. WHITE, T. J., T. BRUNS,S.LEE, AND J. TAYLOR. 1990. Ampli®cation and MANOS,P.S.,AND K. P. STEELE. 1997. Phylogenetic analyses of ``higher'' direct sequencing of fungal ribosomal RNA genes for phylogenies. In Hamamelididae based on plastid sequence data. American Journal of M. A. Innis, D. H. Gelfand, J. J. Sninsky, and T. J. White [eds.], PCR Botany 84: 1407±1419. protocols: a guide to methods and applications, 315±322. Academic MILLER, R. E., T. R. BUCKLEY, AND P. S. MANOS. 2002. An examination of Press, San Diego, California, USA. the monophyly of morning glory taxa using Bayesian phylogenetic in- WILLIAMS, K. J., W. J. KRESS, AND P. S. MANOS. 2004. The phylogeny, ference. Systematic Biology 51: 740±753. evolution, and classi®cation of the genus Globba and tribe Globbeae MORT, M. E., P. S. SOLTIS,D.E.SOLTIS, AND M. L. MABRY. 2000. Com- (Zingiberaceae): appendages do matter. American Journal of Botany 91: parison of three methods for estimating internal support for phylogenetic 100±114. trees. Systematic Biology 49: 160±170. WOOD, T. H., W. M. WHITTEN, AND N. H. WILLIAMS. 2000. Phylogeny of NGAMRIABSAKUL, C., M. F. NEWMAN, AND Q. C. B. CRONK. 2000. Phylog- Hedychium and related genera (Zingiberaceae) based on ITS sequence eny and disjunction in Roscoea (Zingiberaceae). Edinburgh Journal of data. Edinburgh Journal of Botany 57: 261±270. Botany 57: 39±61. WU, T.-L., AND K. LARSEN. 2000. Family Zingiberaceae. In Z.-G. Wu and P. NIXON, K. C., AND J. M. CARPENTER. 1996. On simultaneous analysis. Cla- H. Raven [eds.], Flora of China, vol. 24, 322±377. Science Press, Beijing, distics 12: 221±241. China, and Missouri Botanical Garden Press, St. Louis, Missouri, USA. PEDERSEN, L. B. 2004. Phylogenetic analysis of the subfamily Alpinioideae XIA, Y.-M., W. J. KRESS, AND L. M. PRINCE. 2004. A phylogenetic analysis (Zingiberaceae) with special emphasis on Etlingera Giseke, based on of Amomum (Alpinioideae: Zingiberaceae) using ITS and matK DNA nuclear and plastid DNA. Plant Systematics and Evolution 245: 239± sequence data. Systematic Botany 29: 334±344. 258. ZHANG, L., Q.-J. LI, X.-B. DENG,P.-Y.REN, AND J.-Y. GAO. 2003. Repro- POSADA, D., AND K. A. CRANDALL. 1998. MODELTEST: testing the model ductive biology of Alpinia blepharocalyx (Zingiberaceae): another ex- of DNA substitution. Bioinformatics 14: 817±818. ample of ¯existyly. Plant Systematics and Evolution 241: 67±76.