Everything About Wood

Alpinia Roxb. is the largest and most widespread genus in the Zingiberaceae with some 230 species occurring from Sri Lanka and the Western Ghats of India to China, Japan, all of southeast Asia, the Pacific as far as Fiji, Samoa, and the Caroline Islands, and Australia as far south as northern New South Wales (Larsen et al., 1998⇓; Smith, 1990⇓). Most species grow in low- to mid-elevation forests and form clumps with stems from 1–3 m high, although species east of Wallace's Line tend to grow much larger. Alpinia regia R.M. Sm. of the Moluccas and A. boia Seem. of Fiji, for example, reach over 8 m in height. Some species are found in montane forests up to 2000 m above sea level in New Guinea and Sulawesi. However, very few are tolerant of frost. The most northerly species is Alpinia japonica (Thunb.) Miq., which survives north of Tokyo where the winters can be severe. Several species are important ornamentals (e.g., A. purpurata (Vieill.) K. Schum.) as potted plants, landscape accents, and cut flowers, and at least one (A. zerumbet (Pers.) B.L. Burtt & R.M. Sm.) is naturalized in tropical regions around the world. In Asia, especially China (Wu and Larsen, 2000⇓), alpinias are used as medicinals (e.g., A. officinarum Hance) and in cooking [A. galanga (L.) Willd.].

Alpinias play an important ecological role in the understory of tropical and subtropical forests where many species are quite common. In some cases (e.g., A. kwangsiensis T.L. Wu & S.J. Chen and A. blepharocalyx K. Schum.) the plants form large stands in the understory, along forest margins, and in light gaps, while other species are dominant in wetlands and along water courses [e.g., A. nigra (Gaertn.) B.L. Burtt]. Although most alpinias are pollinated by large bees, some species attract birds and even bats as pollinators (Zhang et al., 2003⇓; Kress and Specht, in press⇓). Flexistyly, a novel floral mechanism promoting outcrossing in which styles move up or down depending on the timing of anther dehiscence, has been described in a number of species of Alpinia (Li et al., 2001⇓, 2002⇓; Zhang et al., 2003⇓).

The generic name Alpinia was first used by Linnaeus for Alpinia racemosa, a neotropical species. Many Asiatic species were added to Alpinia, while later authors tended to refer American species to Renealmia L.f. Schumann (1904)⇓ finalized these taxonomic concepts and subsequently Alpinia Roxb. was conserved for the Asiatic species with Alpinia galanga (L.) Willd. as its type.

Alpinia is the type genus of the tribe Alpinieae A. Rich. of the family Zingiberaceae. This tribe consists of evergreen herbs, in which an abscission layer between the rhizome and the leafy shoots is lacking, the plane of distichy of the leaves is transverse to the direction of growth of the rhizome, and the lateral staminodes of the flowers are small, reduced to swellings at either side of the base of the labellum, or are entirely absent. Extrafloral nectaries are absent, and the fruit is usually spherical and indehiscent or fleshy (Kress et al., 2002⇓).

Within the tribe Alpinieae, generic limits are difficult to discern. While some genera may be easily recognized by their respective morphological characters and/or geographic distribution (e.g., Aframomum, Elettaria, Hornstedtia, Burbidgea), it is hard to identify an apomorphy or universal character for species currently assigned to Alpinia. Virtually all species flower terminally on the leafy shoots and all are Asiatic. These characters distinguish Alpinia from the Afro-American Renealmia, in which most species produce inflorescences on a separate, leafless shoot from the rhizome, but do not uniquely separate it from other members of the Alpinieae. Therefore, to a large degree, one is forced to recognize Alpinia only by eliminating other genera, i.e., it is distinguished only by the plesiomorphic characters of the tribe.

Several attempts have been made to divide Alpinia into smaller genera by elevating some of the more coherent groups of species to the generic rank. Holttum (1950)⇓ applied the name Alpinia to a small group of species with funnel-shaped bracteoles and allocated the remaining species to Catimbium, Cenolophon, and Languas. Several nomenclatural problems were present in this system, but its principal failing was that it only worked for the species of Malaysia. Later authors, therefore, returned to the concept of Alpinia sensu Schumann until Smith (1990)⇓ recognized a group of 22 species in New Guinea that she segregated under the generic name Pleuranthodium(K. Schum.) R.M. Sm.

Infrageneric classifications of Alpinia have been based on inflorescence and flower characters. Much variation exists in these features (Figs. 1–9), from species with branched inflorescences and long cincinni subtended by bracts in which the flowers are each subtended by bracteoles, to other species with no bracts or bracteoles and cincinni of only a single flower. In the Flora of British India, Baker (1894)⇓ considered the species of Alpinia that occur from Sri Lanka to Singapore. His account included descriptions of 17 species from a known total of 30 at that time and divided them into two subgenera and two sections according to the presence of an anther crest, the possession of large bracteoles, and the position of the inflorescence. Schumann (1904)⇓ treated Alpinia throughout its range in his account of the Zingiberaceae for Das Pflanzenreich (Table 1) dividing it into five subgenera and 27 sections. Eight of Schumann's sections have now been placed in entirely different genera, leaving Alpinia with five subgenera and 19 sections. Characteristics of the bracts and bracteoles are the most important diagnostic features in Schumann's classification. Valeton (1913)⇓ later divided section Eubractea into subsection Eustales and subsection Kolowratia and added a new section Monanthocrater to Schumann's 1904⇓ system. He admitted that section Monanthocrater was not sharply distinct from Pycnanthus. The following year, Valeton (1914)⇓ added another new section, subgenus Autalpinia section Presleia.

View this table:

In this window

Table 1. Infrageneric classification of Alpinia according to Schumann (1904)

View larger version:

Download as PowerPoint Slide

Figs. 1–9. Representative floral 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

By contrast, Smith (1990)⇓ only recognized two subgenera, based on characters of the labellum (Figs. 1–9). Subgenus Alpinia was divided into seven sections and 10 subsections, while subgenus Dieramalpinia had four sections and two subsections (Table 2). Bract and bracteole characters were used to delimit the sections and subsections. Stigma types also provided some support for this classification system. Smith's classification is the one most often recognized and used today.

View this table:

In this window

Table 2. Infrageneric classification of Alpinia according to Smith (1990)

Recently, several papers have used molecular data to explore phylogenetic relationships within the family (Searle and Hedderson, 2000⇓; Wood et al., 2000⇓; Kress et al., 2002⇓; Pedersen, 2004⇓) as well as within several genera (Hedychium: Wood et al., 2000⇓; Alpinia: Rangsiruji et al., 2000a⇓, b⇓; Roscoea: Ngamriabsakul et al., 2000⇓; Aframomum: Harris et al., 2000⇓; Globba: Williams et al., 2004⇓; Amomum: Xia et al., 2004⇓).

The study by Kress et al. (2002)⇓ is the most thorough paper to date addressing the relationships among genera in the Zingiberaceae. In that study, sequence data from both the Internal Transcribed Spacer (ITS) loci and matK regions were used to establish, for the most part, well-resolved phylogenetic relationships among the genera, and a new classification of the Zingiberaceae was proposed that recognized four subfamilies and four tribes. They also demonstrated that a number of the larger genera in the family (Amomum, Alpinia, Etlingera, Boesenbergia, and Curcuma) may be para- or polyphyletic and suggested that more extensive sampling is necessary for these taxa, which has subsequently been carried out in some of them (Pedersen, 2004⇓; Xia et al., 2004⇓).

With respect to the genus Alpinia, the results of the investigations by Rangsiruji et al. (2000a⇓, b⇓) and Kress et al. (2002)⇓ are most pertinent. In the former study, in which 47 species of Alpinia and a small number of outgroup taxa were sampled, the authors demonstrated significant statistical support for several monophyletic groups of species of Alpinia, but suggested that the genus may not be monophyletic. In a broader analysis of the genera in the Alpinioideae, Kress et al. (2002)⇓ identified four separate groups of alpinias (Alpinia I–IV) in the 11 species they sampled (Fig. 10). These four groups did not form a monophyletic assemblage, were scattered throughout the tribe, and corresponded to at least some of the clades recognized in the molecular analyses of Rangsiruji et al. (2000b)⇓.

View larger version:

Download as PowerPoint Slide

Fig. 10. The phylogenetic relationships among the genera of the subfamily Alpinioideae of the Zingiberaceae based on a parsimony analysis of ITS and matK sequence data (modified from Kress et al., 2002⇓). Note the two tribes and four polyphyletic clades of Alpinia

Neither the results of the Rangsiruji et al. (2000a⇓, b⇓) nor the Kress et al. (2002)⇓ investigations supported the classification of Alpinia proposed by Smith (1990)⇓. Alpinia galanga (see Fig. 2) and A. conchigera (see Fig. 3; forming Alpinia I of Kress et al. [2002⇓; their fig. 10] in a clade with Aframomum, Renealmia, Amomum, Elettariopsis, and Paramomum; part of the A. galanga clade of Rangsiruji et al. [2000a⇓, b⇓]) are placed in two separate sections of subgenus Alpinia by Smith. Alpinia II of Kress et al. (2002⇓; their fig. 10), which includes A. elegans (see Fig. 9), A. luteocarpa, and A. vittata; (part of the A. eubractea clade of Rangsiruji et al., 2000a⇓, b⇓), as well as the genus Vanoverberghia (from the Philippines and Taiwan), is allied with species of Etlingera, Hornstedtia, and some Amomum, also encompasses both of Smith's subgenera and two of her sections. Alpinia III of Kress et al. (2002⇓, their fig. 10; six species included in four of Rangsiruji et al.'s [2000a⇓, b] clades) is united with the genus Plagiostachys and species of three of Smith's sections; Alpinia IV of Kress et al. (2002⇓, their fig. 10; the single species A. carolinensis [see Fig. 4] found in Rangsiruji et al.'s [2000a⇓, b] A. carolinensis clade) is unresolved with Alpinia II and III. The highly polyphyletic nature of Alpinia as demonstrated by these molecular systematic investigations is congruent with the absence of any recognized morphological apomorphies for the genus as mentioned earlier.

Our goals in the present study, which uses additional molecular sequence data from an expanded taxon sampling of the genus Alpinia, together with a wide range of taxa of the Alpinioideae included in the investigation of Kress et al. (2002)⇓, are (1) to obtain a better understanding of the phylogenetic relationships of the species now taxonomically placed in this genus, (2) to further test the monophyly of the genus as well as of the groups of species identified in the earlier analyses, and (3) to evaluate both Schumann's (1904)⇓ and Smith's (1990)⇓ classifications with respect to our phylogenetic results.

MATERIALS AND METHODS

Taxa

Seventy-two taxa of Alpinia (including 47 species previously analyzed by Rangsiruji et al. [2000a⇓, b] and Kress et al. [2002]⇓), together with 27 non-Alpinia species in the subfamily Alpinioideae (including the three new taxa Pleuranthodium floccosum, P. trichocalyx, and Leptosolena haenkei), eight species in the subfamily Zingiberoideae, one species in the subfamily Tamijioideae, and three species in the Siphonochiloideae from Kress et al. (2002)⇓ were included in the analyses for a total of 112 species (Appendix, see Supplemental Data accompanying online version of this article). In our selection of Alpinia species, both Schumann's (1904)⇓ and Smith's (1990)⇓ classifications were well sampled. All five of Schumann's subgenera are represented in our analyses, as are 14 of the 19 sections still considered to be alpinias. Eight of Schumann's sections have been transferred to different genera in the family, and we have sampled seven of those eight sections. Representatives of both subgenera and all 11 sections of Smith's classification were sampled. As demonstrated by Kress et al. (2002)⇓ Siphonochilus (three species) was designated as the outgroup to all remaining taxa in the Zingiberaceae in our analyses. In addition, representatives of the other two subfamilies were included to polarize character states in the Alpinioideae.

Molecular methods

Sequences for the plastid matK-trnK flanking intergenic spacer regions and the nuclear internal transcribed spacer (ITS) loci were obtained for each taxon either from GenBank or generated according to the following method. Total genomic DNAs were extracted from fresh or silica dried tissue using either a minor modification of Doyle and Doyle (1987)⇓ CTAB (hexadecyltrimethylammonium bromide) method or a DNeasy Plant Mini kit (Qiagen, Valencia, California, USA) extraction protocol. The aqueous phase was extracted with 24 : 1 chloroform/isoamyl alcohol, and DNA was resuspended in Tris-ethylenediaminetetraacetic acid (TE) buffer following isopropyl alcohol precipitation with the CTAB method. DNeasy extraction followed the manufacturer's protocols. Amplification of ITS was accomplished using either primer pair ITS4 and ITS5 (White et al., 1990⇓) or ITS4 and ITS5a (Stanford et al., 2000⇓). The plastid matK region was amplified with trnK1F (Manos and Steele, 1997⇓) and trnK2R (Steele and Vilgalys, 1994⇓). All amplifications used Taq DNA polymerase (Carlsbad, California, USA) according to the manufacturer's direction with annealing temperatures of 54–58°C. Amplified products were purified using the Qiagen Qiaquick (Valencia, California, USA) purification 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 amplification primers plus ITS2 (White et al., 1990⇓) and ITS3G (Kress et al., 2002⇓) as necessary for the ITS region. Zingiberaceae specific 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 (fine) Centri-Sep spin columns (Princeton Separations P/N 901, Adelphia, New Jersey, USA), dried under vacuum, and run on ABI 3100 Automated Sequencer (Perkin Elmer, Applied Biosystems, Inc., Foster City, California, USA) at the Smithsonian Institution's Laboratory for Analytic Biology.

Raw forward and reverse sequences for each sample were assembled, ambiguous bases were corrected, and consensus sequences were edited using Sequencher 4.1 (Gene Codes Corporation, Ann Arbor, Michigan, USA). Consensus sequences for ITS and matK were manually aligned in Se-Al 2.0a11 (Rambaut, 2000⇓). All regions of ambiguous alignment within the ITS regions were excluded and gaps were treated as missing data.

Phylogenetic analyses

Maximum parsimony analyses of the ITS and matK sequence data were conducted using PAUP* 4.0b10 (Swofford, 2002⇓) with equally weighted characters and 1000 random-sequence-addition replicates, saving all shortest trees under ACCTRAN optimization, with the options tree bisection-reconnection (TBR) branch swapping, STEEPEST DESCENT off, MULTREES on, and COLLAPSE branches if maximum length is zero (AMB). Multiple random-sequence additions were used to search for multiple tree islands (Maddison, 1991⇓). The data sets for each gene region were analyzed separately (111 taxa in the ITS analysis and 105 taxa in the matK analysis; see Appendix) and then, following the total evidence approach for multiple data sets (de Queiroz et al., 1995⇓; Nixon and Carpenter, 1996⇓), the sequence data were combined. Incongruence between the ITS and matK data sets was assessed using the incongruence length difference (ILD) test (Farris et al., 1994⇓) as implemented in PAUP*.

Support for the nodes resolved in the strict consensus of the most parsimonious trees was evaluated with bootstrap analyses (Felsenstein, 1985⇓; Mort et al., 2000⇓) using PAUP* with TBR branch swapping on 1000 bootstrap replicates. Bootstrap support was categorized according to Kress et al. (2002)⇓ criteria, i.e., strong (>85%), moderate (70–85%), weak (50–70%), or poor (<50%) support.

A Bayesian analysis using MRBAYES, version 3.0 (Huelsenbeck and Ronquist, 2001⇓) was performed using the same combined ITS-matK parsimony matrix. The most appropriate molecular model for each data set was determined with Modeltest, version 3.06 (Posada and Crandall, 1998⇓). A general time reversible model (rates = gamma, nst = 6) was used for both ITS and matK. Data from ITS and matK were partitioned (using the “lset apply to” command) in order to accommodate differing evolutionary rates for the respective data sets. Four Markov chain Monte Carlo (MCMC) chains, one cold and three heated, were performed. Four MCMC runs of one million generations each, starting from different random points in parameter space, were performed in order to more fully explore tree space and stationarity of parameters (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 (−lnL) scores was achieved were discarded as part of the burn-in period (Huelsenbeck and Ronquist, 2001⇓). Nodes with posterior probability values ≥95% were retained in the 50% majority rule consensus tree.

RESULTS

Internal transcribed spacer

ITS-1 had a total aligned length of 244 bp (unaligned sequences ranged from 201 to 231 bp) with a mean GC content of 54.06%, the 5.8S region had an aligned length of 164 bp (range of 163–164 bp) and GC of 51.18%, and the ITS-2 aligned length was 304 bp (range of 253–281 bp) with GC of 59.52%.

The analysis of the ITS sequence data resulted in 36 300 equally parsimonious nearly fully resolved trees of 1461 steps (number of parsimony-informative characters = 313; consistency index [CI] = 0.386; retention index [RI] = 0.752; rescaled CI = 0.290; Fig. 11). In a strict consensus of these 36 300 shortest trees, Tamijia is strongly supported (bootstrap value = 100%) as a stem lineage one node above the outgroup Siphonochilus. The Zingiberoideae (here represented by eight taxa) is poorly supported as monophyletic (bootstrap value < 50%), but separate from the Alpiniodieae with weak support (bootstrap value = 69%). The tribe Riedelieae (here represented by eight taxa) is strongly supported as monophyletic (bootstrap value = 97%), whereas the remaining taxa comprising the tribe Alpinieae have only poor bootstrap support (<50%) as a monophyletic group.

View larger version:

Download as PowerPoint Slide

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 = 1461; CI = 0.386 excluding uninformative characters; RI = 0.752; and rescaled CI = 0.290) showing bootstrap values (below the line if ≥50%). 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

Within the Alpinieae, the genus Alpinia forms six polyphyletic clades of species each with varying statistical support. Clades I, III, and VI are each strongly supported as monophyletic (bootstrap values > 90%), whereas Clade V (bootstrap values = 59%) and Clades II and IV (bootstrap values < 50%) are weakly to poorly supported groups. The ITS results provide no resolution of the relationships among these six clades of Alpinia.

matK

The 5′ trnK-matK intergenic spacer region had a total aligned length of 1005 bp (unaligned sequences ranged from 789 to 854 bp) with a mean GC content of 29.99%; the matK-coding region had an aligned length of 1636 bp (range of 1551–1564 bp) and GC of 29.69%; and the aligned length of matK-3′ trnK intergenic spacer region was 402 bp (range of 257–279 bp) with a GC of 29.97%.

The analysis of the matK region (coding and noncoding) resulted in 14 433 equally parsimonious trees of 1202 steps (number of parsimony-informative characters = 486; CI = 0.522; RI = 0.808; rescaled CI = 0.422; Fig. 12). A strict consensus of these shortest trees (Fig. 12) provides strong support for the monophyly of both subfamilies Zingiberoideae and Alpinioideae (bootstrap values = 100%) with Tamijia placed as outgroup to the latter subfamily (bootstrap value = 62%). Although strong support is provided for the tribe Riedelieae (minus Siamanthus), the Alpinieae (minus Siliquamomum as in the ITS analysis) is only weakly supported (bootstrap value = 63%). Within the Alpinieae, the identical six clades of species of Alpinia vary in support from weak (clade V) to moderate (clade II) to strong (clades I, III, IV, and VI). Clade VI is placed with clades I and II, and clades III, IV, and V are united, but all with only poor statistical support (bootstrap value < 50%).

View larger version:

Download as PowerPoint Slide

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 = 1202; CI = 0.522 excluding uninformative characters; RI = 0.808; and rescaled CI = 0.422) showing bootstrap values (below the line if ≥50%). The six major clades of Alpinia and non-Alpinieae tribes and subfamilies of the Zingiberaceae are indicated. Abbreviations: CI, consistency index; RI, retention index

Combined data set

The two separate molecular data sets passed the ILD test (P = 0.07) supporting the total evidence approach to combining the molecular evidence. The analysis of the combined ITS and matK sequence data resulted in 3390 equally parsimonious trees of 2759 steps (number of parsimony-informative characters = 799; CI = 0.432; RI = 0.763; rescaled CI = 0.330; Figs. 13–14). A strict consensus of these 3390 shortest trees (Fig. 13) provides strong support (bootstrap values = 92–100%) for the placement of Tamijia as sister to all other Zingiberaceae (except the outgroup genus Siphonochilus) as well as the monophyly of the Zingiberoideae, the Alpinioideae, the Riedelieae (including Siamanthus), and the Alpinieae (including Siliquamomum). Clades I, II, III, and VI of Alpinia are strongly supported as monophyletic (bootstrap value = 94–100%). Clade IV is only poorly supported if A. oxymitra is included, yet moderately supported (bootstrap value = 83%) internal to this taxon. The combined data only weakly support the species of clade V as monophyletic (bootstrap value = 61%), but strongly support this clade (bootstrap = 100%) when united with the genera Etlingera, Hornstedtia, and Amomum. Clades I and II are moderately supported (bootstrap value = 76%) as belonging to the same monophyletic group with Aframomum, Renealmia, Amomum, Elettariopsis, and Paramomum, whereas clades III, IV, and V are only weakly to poorly united with each other.

View larger version:

Download as PowerPoint Slide

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 = 2759; CI = 0.432 excluding uninformative characters; RI = 0.763; and rescaled CI = 0. 330) showing bootstrap values from the parsimony analysis (below the line if ≥50%) and posterior probability values resulting from the Bayesian analysis (above the line if ≥95). 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 flexistyly are designated with an asterisk. Abbreviations: ITS, internal transcriber spacer; CI, consistency index; RI, retention index

View larger version:

Download as PowerPoint Slide

Fig. 14. Condensed tree of the Alpinioideae resulting from the analysis of the combined ITS and matK region sequence data (see Fig. 13) in which monophyletic genera have been collapsed into single branches for clarity. Note the six clades of Alpinia (compare to Fig. 10). The Alpinia Zerumbet clade includes the genus Plagiostachys; the Alpinia Eubractea clade includes the genera Vanoverberghia and Leptosolena. Clades in which flexistyly has been demonstrated are designated with an asterisk. Abbreviations: ITS, Internal Transcriber Spacer

Bayesian analysis results

The 95% majority rule consensus of 9600 trees (10 000 trees minus 400 burn-in trees) resulting from the Bayesian analysis of the combined data set is highly congruent with the strict consensus of the parsimony analysis of the combined data sets (Fig. 13). Each of the six Alpinia clades had a posterior probability value of 98–100%. The only significant exceptions being the placement of Siliquamomum as sister to the tribe Alpinieae and the unresolved placement of Alpinia oxymitra in either clade IV or clade V. At almost all nodes, posterior probability values were equal to or greater than bootstrap values in the parsimony results.

DISCUSSION

The six clades of Alpinia

Alpinia has always been considered to be a taxonomically difficult and complex genus, both in defining the characters that distinguish the genus from other genera in the Alpinioideae and in classifying its species (Schumann, 1904⇓; Smith, 1990⇓; Larsen et al., 1998⇓). The results of our molecular analyses provide new insights into the phylogenetic basis for these taxonomic problems. Although we have sampled only approximately one-third of the described species of Alpinia, the phylogenetic patterns that emerge strongly suggest that the genus is highly polyphyletic in the tribe Alpinieae. These results support the earlier phylogenetic investigations based on the ITS and the trnL-F spacer region of Rangsiruji et al. (2000a⇓, b⇓) in which a more limited number of taxa were sampled in both the ingroup and the outgroup. They suggested that the genus may not be monophyletic and recognized nine major clades. However, because their outgroup taxa were restricted to only two genera in the Alpinieae, Renealmia and Elettariopsis, the wide distribution of species of Alpinia throughout the tribe was not evident. Our results, in which we sampled extensively (though not completely) among the genera in the Alpinieae, strongly support the close phylogenetic relationship of species of Alpinia to other genera in the tribe, such as Renealmia, Aframomum, Amomum, Etlingera, Hornstedtia, Leptosolena, Plagiostachys, Siliquamomum, and Vanoverberghia. In fact, we believe that a major realignment of genera in the tribe, including the description of new taxa, will be necessary if the taxonomy is to accurately reflect evolutionary history (see later discussion).

The four clades of Alpinia recognized by Kress et al. (2002⇓; Fig. 10) and the nine clades described by Rangsiruji et al. (2000b)⇓ correspond closely to the six major clades defined in our analysis (Figs. 13–14). In most cases, our increased taxon sampling and additional sequence data from the matK region have provided stronger bootstrap support for each of their clades. The current analyses, building on their earlier work, now include representative taxa from all of Smith's 11 sections and 14 of Schumann's 19 sections now considered to be alpinias (although a few sections are only sparsely sampled). We, therefore, believe that additional species not yet sampled will most likely be contained within one of these six clades.

Clade I, hereafter called the Fax clade (following the nomenclature of Rangsiruji et al., 2000a⇓, b⇓), contains two species that share capitate, usually radical inflorescences surrounded by sterile bracts with each lateral cincinnus composed of up to seven flowers. This combination of characters is not found in any other genus with which we are familiar. The strong bootstrap support for this clade and the distinctive morphological features may warrant that these species be recognized as a distinct genus in a clade with the African Aframomum and African/neotropical Renealmia. A third species, A. rufescens (not sampled here), with similar characteristics that may be appropriately included in the Fax clade, is only known from the type and is in need of further study. Members of this small clade are geographically well circumscribed, occurring in Sri Lanka and a small part of SW India. The occurrence of the closest relatives of this Indian subcontinent clade in Africa suggests that the common ancestor of the Fax clade may have “drifted” across the Indian Ocean with the breakup of Gondwana.

Clade II, the Galanga clade, includes species from three of Schumann's sections and two of Smith's sections (Figs. 1–3). The four species included in this clade are mostly found in continental Asia although A. bilamellata comes from the Bonin Islands. From the illustration in the Flora of the Bonin Islands (Toyoda, 1981⇓, plate 104), A. bilamellata appears to be very similar to A. galanga. The Galanga clade consists of Rangsiruji et al.'s A. galanga clade with the addition of A. bilamellata. The four species of our Galanga clade form a relatively coherent group possessing cincinni made up of many, small flowers with a similar labellum shape. The relationship of A. galanga with A. conchigera and A. nigra is also supported by Liao and Wu (1996)⇓, who studied fruit wall anatomy. The conflict between our results and Schumann's and Smith's classifications arises from the nature of the bracteoles, which are tubular in A. conchigera and A. nigra, but open to the base in A. galanga and A. bilamellata. Our results suggest that this character can reverse states within a clade. The disparate position of the two accessions of A. nigra is somewhat problematic and may be due to the widespread distribution of this species with significant local differentiation. Alpinia galanga, the type of the genus, is contained in this clade. Further taxonomic revisions may require that the name “Alpinia” be restricted to members of this relatively small group of species.

Clade III, the Carolinensis clade, is made up of species that share the characters of Smith's subgenus Dieramalpinia, such as the narrow, fleshy labellum adpressed to the stamen; tubular, tightly clasping bracteoles; flowers in cincinni; and the persistent primary bract (Fig. 4). Members of this clade belong to three of the four sections in Smith's classification, namely Myriocrater (A. aenea, A. cylindocephala, A. monopleura, A. eremochlamys, and A. coeruleoviridis), Pycnanthus (A. boia), and Dieramalpinia (A. carolinensis). Schumann also placed these species, at least those that were known at the time, in four sections of subgenus Dieramalpinia. These species are disjunct between Fiji (A. boia), the Caroline Islands (A. carolinensis), and Sulawesi (remaining five species). Subgenus Dieramalpinia sensu Smith occurs east of Wallace's Line, with very few exceptions in Borneo. The species of this area are far less well known than those taxa west of Wallace's Line.

Clade IV, the Zerumbet clade (Figs. 5–8), is the largest group of species in our analyses and includes four of Rangsiruji et al.'s clades (the A. zerumbet clade, A. polyantha clade, A. glabra clade, and A. aquatica clade). The large Zerumbet clade contains members of Smith's subgenus Alpinia section Alpinia and section Didymanthus (A. pumila), subgenus Dieramalpinia section Dieramalpinia and section Myriocrater (A. vulcanica), and two species of the genus Plagiostachys. Although only poor bootstrap support exists for the Zerumbet clade as a whole, the lack of support may be due to the poorly resolved position of Alpinia oxymitra, which is sister to all remaining species in the clade. The Bayesian analysis also suggests an unresolved position for this species in the Alpinieae. The node interior to A. oxymitra, which includes all of the remaining species of the Zerumbet clade, is moderately supported in the parsimony analysis and strongly supported in the Bayesian analysis. Yet despite this statistical support for the molecular results, it is difficult to find any morphological apomorphies of the Zerumbet clade. Similarly, no biogeographical patterns are apparent for this clade as a whole.

Within the Zerumbet clade only two of the four clades of Rangsiruji et al. have significant bootstrap support (the A. glabra clade and A. aquatica clade with bootstrap values of 98% and 81%, respectively). The A. glabra clade includes species found in Borneo and also encompasses the genus Plagiostachys in our analysis. The A. aquatica clade is primarily restricted to the Philippines (except for A. aquatica) and is generally characterized by small flowers with a petaloid four-lobed labellum. A third, strongly supported subclade consists of three species, A. oxyphylla (Fig. 8), A. calcarata (Fig. 7), and A. officinarum, the latter two of which are included in the A. zerumbet clade of Rangsiruji et al. A more thorough analysis of the morphology and biogeography is needed for all the subclades of the large and somewhat amorphous Zerumbet clade.

Clade V, the Eubractea clade, similar to the Zerumbet clade, is difficult to characterize morphologically. No good characters define this clade as a whole, although it contains several smaller clades with circumscribed geographical ranges in the Philippines, Australia, the Bismarck Archipelago, and the tropical Pacific. The well-supported clade of A. arundelliana, A. caerulea, and A. modesta contains only Australian species. A clade found primarily in the Philippines is composed of three species of Alpinia (A. elegans [Fig. 9], A. pinetorum, and A. luteocarpa) plus Vanoverberghia sepulchrei and may encompass the poorly resolved Leptosolena haenkei also from the Philippines. The latter monotypic genus, which was not included in the analysis by Kress et al. (2002)⇓, was previously considered to be extinct in the wild, but is now known to be quite common (Funakoshi et al., in press⇓). A small Pacific Ocean clade (A. oceania, A. vittata, and A. purpurata) is also strongly supported.

Clade VI, the Rafflesiana clade, includes just two species, A. javanica and A. rafflesiana, distributed on the Sunda Shelf in southern Thailand, peninsular Malaysia, Sumatra, and Java. Smith placed both of these taxa in her subgenus Alpinia section Allughas subsection Allughas, while Schumann recognized each of the two species as separate sections Brachybotrys and Javana in the same part of his key sharing the feature of short cincinni with no more than six flowers. Although the Rafflesiana Clade is sister to the monotypic Vietnamese/Chinese Siliquamomum, the bootstrap support for this relationship is poor, and the Bayesian analysis places Siliquamomum basal to the Alpinieae.

Previous classifications of Alpinia

The genus Alpinia has assumed varying significance in previous classifications of the Zingiberaceae. Schumann's (1904)⇓ comprehensive treatment of the family included in Alpinia a large portion of what we now recognize as subfamily Alpinioideae, including all or parts of the genera Pleuranthodium, Riedelia, Geocharis, Plagiostachys, Leptosolena, and Amomum. The first two genera are now placed in tribe Riedelieae, and the latter four, along with Alpinia and 10 other genera, are included in the tribe Alpinieae (Kress et al., 2002⇓; Fig. 10; Table 1). We have sampled species currently placed in Alpinia in 14 of Schumann's 19 sections contained in his five subgenera. Two to five of Schumann's sections are contained within five of the six clades currently recognized in our molecular analyses (Table 1; Appendix).

Smith's (1990)⇓ elegant and intricate classification of Alpinia was an attempt to provide a modern interpretation of the complex array of species placed in this genus. Her two subgenera, 11 sections, and 12 subsections encompassed the 221 species known at that time. Four of our clades include taxa placed in from two to six of her sections. Only our Fax clade and Rafflesiana clade correspond to a single section (section Fax) or subsection (section Allughas subsection Allughas) of Smith's classification.

It is clear that a significant disparity exists between our phylogenetic results and the taxonomic concepts of generic and infrageneric classifications of Schumann (1904)⇓ and Smith (1990)⇓. Unfortunately, at this time we do not yet have adequate morphological characters to support all of the results of our molecular analyses. Yet, some of the morphological characters of previous taxonomists are clearly at odds with some of our well-supported DNA sequence results. For example, both Schumann and Smith gave great weight to the nature of the bracteoles in delimiting genera. It is very rare in their classifications to find infrageneric sections or subsections with both tubular and open bracteoles. Our results in the Galanga clade confirm those of Rangsiruji et al. (2000a)⇓ who found A. galanga with open bracteoles to be most closely related to A. conchigera and A. nigra with tubular ones. Clearly, this inflorescence character must be used with caution.

Another character used at the generic level is the position of the inflorescence. Schumann (1904)⇓ recognized only three radical-flowering species of Alpinia, namely, A. chrysogynia and A. melichroa in section Botryamomum and A. pumila in section Didymanthus, along with the majority of species with terminal inflorescences; both of these sections were placed in subgenus Rhizalpinia. Smith (1975)⇓ transferred two more species with radical inflorescences, A. abundiflora and A. fax, from Schumann's Amomum section Geanthus series Polyanthae into her section Fax, arguing that the presence of a showy involucre of sterile bracts and cincinni of two to seven flowers made it impossible to place them in Amomum. She also added A. rufescens from Schumann's subgenus Autalpinia section Cenolophon to her section Fax. Our results indicate that it may be possible to recognize section Fax at the generic rank, although we have not been able to include a sample of A. rufescens in our phylogenetic anaylses.

Of Schumann's three radical-flowering species, we now know from plants cultivated in our research greenhouses (Appendix) that the inflorescence position of A. pumila is terminal on a leafy shoot (W. J. Kress and M. Bordelon, Smithsonian Institution, unpublished data). This observation is confirmed by Wu and Larsen (2000)⇓ who have seen living plants and do not refer to the inflorescence of A. pumila as radical. The final two species of Schumann's section Botryamomum remain very poorly known and are only tentatively placed in Alpinia. Smith (1990)⇓ thought they might belong to Amomum. This may be correct in the case of Alpinia chrysogynia, which has bracts and bracteoles with single-flowered cincinni. However, A. melichroa lacks bracts and bracteoles, and may be closer to the radical-flowering species of Riedelia. If both these species are excluded from Alpinia, then all remaining species of the genus are terminal flowering.

In the search for new morphological characters useful in classification, a lead has been given by the careful study of the fruit wall of Chinese Alpinia by Liao and Wu (1996)⇓ who demonstrated a link between three species of the Galanga clade. Their study should be extended to encompass the entire geographical range of the genus.

A new classification of the tribe Alpinieae

Based on our current results and those of Rangsiruji et al. (2000a⇓, b⇓) for Alpinia as well as the investigations of Kress et al. (2002)⇓ for the family Zingiberaceae, it is tempting to propose here a new classification of the tribe Alpinieae. The congruence between the major clades of Alpinia in these analyses provides nearly unequivocal evidence that this genus is polyphyletic and that groups of species are more closely related to other genera in the tribe than they are to each other (Fig. 14). At least four of the six clades are strongly supported in both the parsimony and Bayesian analyses, yet the relationships among the clades are not fully resolved. It is quite unlikely that major suites of species of Alpinia will be available for analysis in the near future due to their widespread distribution and the difficulty of obtaining tissue samples. Yet, we expect that a few additional critical species may still be added to this data set. Moreover, the monophyly and phylogenetic position of a number of genera in the Alpinieae have not yet been established. Although 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 polyphyletic and in need of further taxon sampling (Xia et al., 2004⇓). In addition, the taxonomic uniqueness of such genera as Vanoverberghia, Leptosolena, Plagiostachys, and Elettariopsis needs to be resolved. For all these reasons, we are reluctant at this time to propose a new classification with redefined generic boundaries. However, our results together with the others previously listed will provide the foundation for a revised classification of the Alpinieae in the near future.

The ecological and evolutionary distribution of flexistyly in Alpinia

One purpose for determining phylogenetic history is to understand patterns of evolution of various morphological and ecological characteristics of taxa. Flexistyly, a unique floral mechanism in plants that appears to promote outcrossing in the populations where it is found (Li et al., 2001⇓; Renner, 2001⇓; Barrett, 2002⇓), was first described in Amomum and Alpinia, two genera that we now understand to be polyphyletic in the Zingiberaceae. Although our species sampling for molecular phylogenetic analyses of both of these genera is modest, we have sufficient data to make initial interpretations of the evolutionary origin of this angiosperm mating system. Flexistyly has now been documented in 24 species of the Zingiberaceae (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 Amomum (A. koenigii, A. tsaoko, and A. paratsako) and Etlingera (E. yunnanensis) and possibly Plagiostachys and Paramomum (Cui et al., 1996⇓). The distribution of flexistyly in the family, based on these preliminary results, suggests that this mating system may have evolved in the common ancestor of the tribe Alpinieae or independently at least three to five times in the tribe (Figs. 13, 14). Both of these hypotheses are significant in terms of understanding differential patterns of species diversification in the Zingiberaceae. As additional field documentation of this floral mechanism becomes available, a more thorough understanding of the prevalence of flexistyly and its evolutionary origin in the gingers will be possible.