Everything About Wood

With over 100 species, Globba L. is the third largest genus of the Zingiberaceae (53 genera, 1200+ species), ranking in number behind only the polyphyletic genera Alpinia and Amomum (Kress et al., 2002⇓). Globba (Figs. 1–3), along with Mantisia (four species; Fig. 4), Gagnepainia (three species; Fig. 5), and Hemiorchis (three species; Fig. 6) comprise tribe Globbeae; one of the four traditionally recognized tribes of the Zingiberaceae (e.g., Petersen, 1889⇓; Schumann, 1904⇓; Holttum, 1950⇓; Burtt and Smith, 1972⇓; Larsen et al., 1998⇓). The unifying feature of the Globbeae is a unilocular ovary with parietal placentation (vs. a trilocular ovary with axile placentation in most of the family). Members of all four genera are relatively small (<1 m) understory herbs, although at least one species (G. racemosa Sm.) can reach 3 m. Characters of use in distinguishing the genera include inflorescence position, fruit shape, stamen length, and lateral staminode position (Table 1).

View this table:

In this window

Table 1. Comparison of Globbeae genera (following Schumann, 1904)

View larger version:

Download as PowerPoint Slide

Figs. 1–7. Representatives of tribe Globbeae. Classification follows Schumann (1904)⇓ and Larsen (1972)⇓ 1. Globba winitii (sect. Globba). Note pendulous inflorescence and conspicuous bracts common to sect. Globba. 2. Globba insectifera (sect. Nudae). Note the erect inflorescence and lack of large, persistent bracts; both characteristics are found throughout the traditional sections Ceratanthera and Nudae and most of sect. Haplanthera. 3.Globba sessiliflora (sect. Haplanthera) flower. Note lateral staminodes (the thin, upcurved, petaloid structures) are attached below the top of the labellum. 4. Mantisia radicalis flower. Note lateral staminodes are positioned on filament above labellum. 5. Gagnepainia thoreliana flowers. Note the yellow central point on the labellum. 6. Hemiorchis rhodorrachis flower. Note short stamen and yellow central point on labellum. 7. Anther appendage types in Globba. (A) Sect. Haplanthera (no appendages). (B) Sect. Ceratanthera ser. Basicalcaratae (two basally attached appendages). (C) Sect. Ceratanthera ser. Mediocalcaratae (two broadly attached appendages). (D) Sects. Globba and Nudae (four appendages)

Globba species are distributed throughout tropical (and parts of subtropical) Asia, ranging from India to southern China, south and east to the Philippines and New Guinea (Schumann, 1904⇓), with the center of distribution in monsoonal Southeast Asia, especially Thailand (Larsen, 1996⇓) and Myanmar (Kress et al., 2003⇓). Virtually all species distributed north of the Isthmus of Kra (most species of Globba and all species of the remaining genera) enter dormancy from approximately November through April, while most species south of that point remain evergreen throughout the year. The other three genera of Globbeae are more restricted in distribution and fall completely within the range of Globba itself. Gagnepainia is found primarily in Thailand, Laos, Vietnam, and Cambodia, while Hemiorchis and Mantisia are distributed in northeastern India, Myanmar, and Bangladesh.

Flowers in the Globbeae (Figs. 1–6), like all Zingiberaceae, are among the most highly derived in angiosperms (Endress, 1994⇓; Kress et al., 2002⇓). Calyces in the Globbeae are highly reduced, with petals replacing most of their protective function. Standard petal function (i.e., pollinator attraction and mechanical assistance to pollination) has been co-opted by elaborate staminodes that have replaced four of the six stamens that were fertile in ancestral species of Zingiberales (the fifth stamen is aborted in the Zingiberaceae and the sixth remains fertile; Kirchoff, 1988⇓). Globba flowers (Fig. 3) are distinctive in having a relatively small staminodal labellum and a greatly elongated, arched stamen that is as long or longer than the floral tube and staminodes (also seen in Gagnepainia, Fig. 5; Mantisia, Fig. 4; and the unrelated Pommereschia). However, the hallmark of most (∼90%) Globba species are the small linear to triangular appendages along the sides of the anther (Fig. 7). The colorful bracts and flowers seen in many species are useful taxonomically and have attracted horticultural interest, especially for G. winitii C. H. Wright (Fig. 1) and the as yet undescribed G. magnifica (K. J. Williams and M. F. Newman, unpublished data). Most, if not all, species of Globba can reproduce through the production of asexual vegetative bulbils in the inflorescence, a rare occurrence in the rest of the family (Larsen et al., 1998⇓). In some species (e.g., G. marantina L. and G. bulbifera Roxb.) seeds are rarely produced and plants produce bulbils as their primary means of reproduction.

Recent molecular work on the Zingiberaceae has allied the Globbeae with Zingibereae (formally tribes Zingibereae and Hedychieae) into the new subfamily Zingiberoideae (Kress et al., 2002⇓). Although the Globbeae was maintained (Kress et al., 2002⇓), the monophyly of the tribe was only supported in the strict consensus tree based on nuclear internal transcribed spacer (ITS1 and 2) and 5.8S coding region data. Combined trnK-matK (matK) and ITS data from the Kress et al. (2002)⇓ analyses suggest a paraphyletic grade of Globbeae and Caulokaempferia successively sister to a monophyletic Zingibereae. Our current study attempts to further clarify relationships of Globbeae within the family and to resolve relationships within Globba, one of the most distinctive genera in the Zingiberaceae.

Since the first infrageneric classification by Horaninow (1862)⇓, circumscription in Globba (Table 2) has focused primarily on the number of pointed to triangular appendages on the anther: zero, two, or four (Fig. 7). This infrageneric classification is based upon the work of Lestiboudois (1841)⇓, who split Globba into three genera (Colebrookia, Ceratanthera, and Globba) based on the form of the anther appendages. As Horaninow's (1862)⇓ sect. Marantella contains the type of the name Globba (G. marantina), it will be called sect. Globba in this paper. Schumann (1904)⇓ provided the most recent and comprehensive monograph of the genus, recognizing three sections based upon anther appendages (combining the two zero-appendaged sections, Haplanthera and Careyana of Horaninow, 1862⇓; Table 2). He further subdivided section Ceratanthera into three series based on the attachment point of the anther appendages on the anther (i.e., basal in Basicalcaratae, medial in Mediocalcaratae, or apical in the monotypic Apicalcaratae). Section Globba was divided into two series based on bract and bulbil characters by Schumann (1904)⇓. The series of sect. Globba were based mainly on poorly preserved herbarium specimens and inadequate descriptions, and most species in the section could be placed in either series depending on the opinion of the collector and the condition of the plant. Larsen (1972)⇓ created the monotypic section Nudae for a northern Thailand species (G. nuda K. Larsen) that has highly reduced bracts and leaves, but which would otherwise be placed in section Globba (i.e., four anther-appendages). Other characters for section Nudae include erect, open inflorescences, and elongate, ridged fruit (K. J. Williams, personal observation). Several species, including G. expansa Horan., G. insectifera Ridl., G. flagellaris K. Larsen, and G. lithophytica (K. J. Williams and M. F. Newman, unpublished data), have the sectional characteristics of sect. Nudae but have not previously been placed in the group. Besides anther appendage number, characters useful for classification within Globba include inflorescence shape (erect, arching, or pendulous), bract size, color, and persistence, and fruit morphology (i.e., globose and rugose or ellipsoid and longitudinally ridged). Numerous cytological studies conducted on Globba and Mantisia (e.g., Larsen, 1972⇓; Lim, 1972a⇓, b⇓, c⇓, 1973a⇓, b⇓; Newman and Jong, 1986⇓; Takano, 2001⇓) suggested that x = 8 is the base number for Globba and x = 10 for Mantisia.

View this table:

In this window

Table 2. Traditional sectional classification of Globba (following Schumann, 1904, and Larsen, 1972)

The only published phylogenetic study that has included more than three species of Globba is Takano and Okada (2002)⇓. Their study focused primarily on the Sumatran species of Globba and used molecular data to document multiple occurrences of “triploidy” (= hexaploidy; see Lim, 1972a⇓) in the genus. Their results showed sect. Ceratanthera to be polyphyletic, but sampling in this group was quite low. A number of the members of sect. Globba native to Sumatra form a well-supported clade. As the focus of their study was the Sumatran and Malaysian species, only seven species of Globba from the monsoonal region (all from Thailand) were sampled, and Gagnepainia and Mantisia were not sampled at all.

The primary goal of this study is to generate a hypothesis of phylogenetic relationships based on nuclear and chloroplast molecular sequence data for Globba and tribe Globbeae. The ITS and matK regions were chosen based in large part on the utility of these regions in other phylogenetic studies within the Zingiberaceae (Harris et al., 2000⇓; Ngamriabsakul et al., 2000⇓; Rangsiruji et al., 2000⇓; Searle and Hedderson, 2000⇓; Wood et al., 2000⇓; Kress et al., 2002⇓; and Takano and Okada, 2002⇓) and to have representation from both nuclear (ITS) and chloroplast (matK) DNA. Our phylogenetic results will be compared to previous classification systems, principally that of Schumann (1904)⇓. The relationship between anther appendage morphology (i.e., number and shape) and phylogeny will be examined. The homologies of different fruit and inflorescence morphologies will be compared in light of this phylogenetic hypothesis as well. A new infrageneric classification of Globba, based on these results, will be presented.

MATERIALS AND METHODS

Taxa

One hundred and nine accessions of tribe Globbeae were sampled for ITS (generic and infrageneric classification follows Schumann, 1904⇓, and Larsen, 1972⇓). Over 70% of the species in the Globbeae were sampled, representing 64 species of Globba, all four species of Mantisia, two species of Gagnepainia, and two species of Hemiorchis, with multiple samples of several species. Multiple species were sampled for all sections and series of Globba (sect. Ceratanthera ser. Apicalcaratae is unispecific). Due to the clear phylogenetic pattern seen in the ITS phylogeny (Williams et al., 1999⇓) that corresponded closely to identifiable morphological characters, a subset of 55 Globbeae accessions (42 Globba species representing all major clades in the ITS analysis, all four species of Mantisia, and two species each from Gagnepainia and Hemiorchis) were sampled for matK. Twenty-two ITS and 15 matK sequences from Takano and Okada (2002)⇓ representing primarily Sumatran taxa were included in the analysis. All sequences newly generated for this study were submitted to Genbank (see Supplementary Data accompanying the online version of this paper). Sequencing of the chloroplast regions trnL-F, rpl16, rps16, and the atpBE-rbcL spacer were also attempted in a pilot study of 10 taxa from across the tribe, all of which showed too little sequence variation to be considered useful for this study (results not shown).

Outgroups

Thirteen taxa, representing 12 genera, of Zingiberaceae were used as outgroups in this analysis. Outgroup choice was based on the results of Kress et al. (2002)⇓. Three members of subfamily Alpinioideae were used to root the tree. Eight diverse members of subfamily Zingiberoideae tribe Zingibereae (and two accessions of the unplaced Caulokaempferia) were also included. Caulokampferia and the Zingibereae are of interest as they formed a paraphyletic grade with two clades of Globbeae in the Kress et al. (2002)⇓ analysis. All outgroup sequences were obtained from the data matrix of Kress et al. (2002)⇓.

Molecular methods

Total genomic DNAs were extracted from fresh, silica dried, or herbarium material using either a DNeasy Plant Mini Kit (Qiagen, Valencia, California, USA) extraction protocol or a minor modification of the Doyle and Doyle (1987)⇓ hexadecyltrimethylammonium bromide (CTAB) method. DNeasy extractions followed the manufacturer's protocols. Amplification of ITS (ITS 1, 5.8s, and ITS 2) was accomplished using ITS4 (White et al., 1990⇓) and ITS5a (Stanford et al., 2000⇓) primers. The chloroplast matK region was amplified with trnK1F (Manos and Steele, 1997⇓) and trnK2R (Steele and Vilgalys, 1994⇓). All amplifications used Qiagen Taq DNA polymerase according to the manufacturer's directions with annealing temperatures of 50°C. Amplified products were purified using the Qiagen Qiaquick purification protocol with the products sequenced directly using automated sequencing methodology of the ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, 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 (Kress et al. 2002⇓). The additional primers used (m5Fa, m8Fa, and m8R) are modifications or complements of the primers of Steele and Vilgalys (1994)⇓. 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 either an ABI 373, 377, or 9700 Automated Sequencer at the Duke University Department of Biology Sequencing Facility. Raw sequences were assembled and edited using Sequencher 3.1.1 (Gene Codes, Ann Arbor, Michigan, USA) and manually aligned.

Phylogenetic analyses

The first 62 and last 79 bases of matK were excluded from this analysis. Uninformative characters were also excluded from the parsimony analysis. Eight ITS indels were coded as binary characters and included in the large ITS analysis. The combined ITS-matK analysis used 19 matK binary indel characters in addition to those scored for the ITS data set. For each gap coded in this analysis, the length was uniform in all taxa sampled (variable length gaps were excluded). Uneven length gaps, and gaps where exact placement was uncertain, were not coded.

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 (100 for the large ITS analysis), saving all shortest trees under ACCTRAN optimization, tree bisection-reconnection (TBR) branch swapping, STEEPEST DESCENT off, MULTREES on, 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 and then, following the total evidence approach for multiple data sets (de Queiroz et al., 1995⇓; Nixon and Carpenter, 1996⇓), we combined the sequence data. Incongruence between the ITS and matK data sets was assessed using the incongruence length difference (ILD) test (Farris et al., 1994⇓) as implemented in the “partition homogeneity” test of PAUP* (10 000 replicates, Mulpars off, outgroups included).

Support for nodes resolved in the strict consensus of most parsimonious trees was evaluated with bootstrap (Felsenstein, 1985⇓; Mort et al., 2000⇓) and Bremer support (Bremer, 1988⇓) analyses. Bootstrap analyses were conducted using PAUP* with TBR branch swapping on 1000 bootstrap replicates. Bootstrap values ≥85% support were considered strong, 70–84% moderate, 50–69% weak, and <50% lacking. Bremer support values were generated using AutoDecay version 4.0 (Eriksson, 1998⇓).

A Bayesian phylogenetic analysis using MrBayes version 3.0 (Huelsenbeck and Ronquist, 2001⇓) was performed using the same combined ITS-matK parsimony matrix (except that gaps were not coded and uninformative characters were included). 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 applyto” command) to accommodate differing evolutionary rates for the respective data sets. Four Markov Chain Monte Carlo (MCMC) chains, one cold and three heated, were performed. Three MCMC runs of 1 × 106 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 majority rule consensus tree.

RESULTS

Internal transcribed spacer

Total aligned sequence length for the ITS region in the 122 taxon data set was 675 bp. The unaligned ITS1 region varied from 162 to 202 base pairs (bp) in length, unaligned ITS2 was 225–256 bp, and 5.8s was 187 bp in all taxa. Analysis of the 122 taxon ITS data set resulted in 3979 most parsimonious trees (MPTs) of length 739 steps (231 parsimony informative characters, consistency index [CI] = 0.478, retention index [RI] = 0.843, rescaled consistency index [RC] = 0.403, mean GC content 52.2%). All MPTs were found in the first two addition sequence replicates. In the strict consensus (Fig. 8) Gagnepainia and Hemiorchis form a strongly supported clade whose position is unresolved. Mantisia formed a strongly supported monophyletic clade that is part of a basal polytomy with four clades of Globba species. Section Ceratanthera is polyphyletic with sect. Ceratanthera ser. Basicalcaratae forming a distinct clade that is part of the basal polytomy, while sect. Ceratanthera ser. Mediocalcaratae (+ Apicalcaratae) shares a close affinity to section Nudae. Section Globba forms three clades (two large clades plus G. fragilis S. N. Lim) in a polytomy with the Nudae + Mediocalcaratae clade. Section Haplanthera is not resolved as monophyletic, with the two resulting clades forming the remaining members of the basal polytomy. Globba + Mantisia has strong support as a monophyletic group, but bootstrap support for the topology of the major lineages relative to each other is generally lacking.

View larger version:

Download as PowerPoint Slide

Fig. 8. Strict consensus of 3739 most parsimonious trees (MPTs) from the 122 taxon internal transcribed spacer (ITS) data set. Globba is abbreviated “G.” Sectional and series names correspond to Schumann (1904)⇓ and Larsen (1972)⇓. Numbers listed above the horizontal lines are bootstrap values ≥50%. Numbers in parentheses are the number of anther appendages found on the included taxa. Species whose specific epithet is in quotes are unpublished, but these are the intended names for these taxa (K. J. Williams and M. F. Newman, unpublished data)

Analysis of the reduced ITS data set (identical taxa to the matK matrix) of 213 potentially parsimony informative characters resulted in 368 MPTs (649 steps, CI = 0.505, RI = 0.775, RC = 0.391, mean GC content 52.2%; not shown). All MPTs were found during the first two addition sequence replicates. Aside from a single clade, which lacked bootstrap support in the Globba sect. Globba lineage, no differences exist between the topologies of the bootstrap consensus trees of the large and small ITS matrices (for taxa represented in both analyses).

matK

The total aligned matK data set was 3350 bp in length. Unaligned sequence length for the 5′ trnK-matK spacer region was 805–893 bp in length, while the matK coding region was 1534–1600 bp, and the 3′ trnK-matK spacer was 281–383 bp. This data set has 260 parsimony informative characters resulting in 1500 MPTs of length 489 steps (CI = 0.640, RI = 0.858, RC = 0.549, mean GC content 48.7%). All MPTs were found during the first addition sequence replicate. The strict consensus tree from the matK analysis (Fig. 9) supports each of the primary clades seen in the ITS analyses, except that support for a monophyletic four-appendaged sect. Globba (plus sect. Ceratanthera ser. Mediocalcaratae) is lacking. Two clades of Globbeae form a polytomy with Caulokaempferia and the Zingibereae. Gagnepainia and Hemiorchis are resolved as distinct sister lineages with strong bootstrap support. Globba + Mantisia are resolved as a monophyletic lineage with Mantisia strongly supported as sister to section Haplanthera, except for G. substrigosa Baker, which forms a distinct lineage of its own. The Haplanthera + Mantisia clade forms a polytomy along with a Nudae + Mediocalcaratae clade, sect. Ceratanthera ser. Basicalcaratae clade, two distinct clades of section Globba, and G. substrigosa (Fig. 9).

View larger version:

Download as PowerPoint Slide

Fig. 9. Strict consensus of 1500 most parsimonious trees (MPTs) from the matK data set. Sectional and series names correspond to Schumann (1904)⇓ and Larsen (1972)⇓. Numbers listed above the horizontal lines are bootstrap values ≥50%. Numbers in parentheses are the number of anther appendages found in the included taxa

Combined analysis

The parsimony analysis of the combined data set resolves four MPTs of 1157 steps (473 parsimony informative characters, CI = 0.554, RI = 0.806, RC = 0.447; Fig. 10) and has better resolution and more well-supported nodes than the results of either data set alone. All MPTs were found in the first addition sequence replicate.

View larger version:

Download as PowerPoint Slide

Fig. 10. Strict consensus of four most parsimonious trees (MPTs) from the combined ITS-matK data set. Numbers above the horizontal lines and left of the slashes indicate bootstrap values (if ≥50%). Numbers above the horizontal lines and to the right of the slashes indicate Bayesian posterior probability values (PPV) ≥95%. Numbers below the horizontal lines are Bremer support values. An asterisk indicates bootstrap value <50% or PPV <95% (if node is supported in the alternate analysis). A pound sign indicates that the topology of this clade in the Bayesian analysis is identical to that of the matK analysis (Fig. 9; the clade of G. bulbifera + G. “magnifica” has 97% PPV)

The combined data set failed the ILD test (P = 0.0036). The position of G. xantholeuca Craib and G. insectifera were found to vary (with strong bootstrap support) within their respective clades between the ITS and matK analyses (Figs. 8, 9). When G. xantholeuca and G. insectifera were removed from the analysis, the combined data set easily passed the ILD test (P = 0.1094). This result suggests that the nuclear and chloroplast data sets are congruent for the data set as a whole. The effect on combinability attributable to the conspicuous difference in the sect. Globba + Nudae topology between ITS (Fig. 8) and matK (Fig. 9) was also tested (by removing the lower sect. Globba clade), but did not significantly affect ILD results (P = 0.1023). Problems with the ILD test are well known (e.g., Dolphin et al., 2000⇓; Yoder et al., 2001⇓; Darlu and Lecointre, 2002⇓; Dowton and Austin, 2002⇓), so failures due to localized cases of incongruence are not surprising (Thornton and DeSalle, 2000⇓). Both G. xantholeuca and G. insectifera were left in the analysis, as in each case the incongruence affected only the relationships of three closely related species.

The 95% majority rule consensus of 9801 trees (10 001 trees minus 200 burn-in trees) resulting from the Bayesian analysis of the combined data set is congruent with the strict consensus of the parsimony analysis. The only exception being that the relationship of G. bulbifera to G. magnifica is identical to that of the matK parsimony analysis (Fig. 9). In all but one case, nodes with ≥50% bootstrap support garnered a posterior probability value of ≥95% (Fig. 10). In four cases, posterior probability values of ≥95% occurred for nodes with <50% bootstrap support. Unless otherwise stated, a “strongly supported” clade in the combined analysis refers to a node with ≥85% bootstrap support and a posterior probability value ≥95%.

Tribe Globbeae was resolved as monophyletic in the strict consensus (Fig. 10). Hemiorchis + Gagnepainia formed a strongly supported clade that is sister to Globba + Mantisia. As in the matK parsimony analysis (Fig. 9), Mantisia is strongly supported as derived out of Globba and is sister to Globba sect. Haplanthera (exclusive of G. substrigosa; Fig. 10). Globba substrigosa again forms a distinct lineage of its own and is placed sister to the Mantisia + Haplanthera clade. Section Ceratanthera ser. Basicalcaratae is resolved as sister to the rest of Globba in the strict consensus but this placement lacks support. Section Ceratanthera ser. Mediocalcaratae is strongly supported as sister to sect. Nudae, and together these taxa belong to a strongly supported four-appendaged clade. The Nudae + Mediocalcaratae clade makes the two well-supported clades of section Globba paraphyletic in the parsimony and Bayesian analyses.

In order to compare a topology where all species of Globba with anther appendages (i.e., every section but Haplanthera), except Mantisia, belong to a single clade, and within this clade sect. Globba sensu Schumann is monophyletic (altered topology), to the most parsimonious results from the combined data set (Fig. 10), a significantly less parsimonious test (SLP_T; Templeton, 1983⇓) was performed as implemented in PAUP*. A constraint tree uniting all members of sections Ceratanthera, Nudae, and Globba into a monophyletic clade was constructed. Within this clade all members of sections Globba were constrained to be a monophyletic group. A parsimony search with the same parameters as the unconstrained combined analysis was conducted using this constrained topology. The resulting nine MPTs (1160 steps) were each compared to all four MPTs of the unconstrained data set (1157 steps). SLP_T values for all comparisons were nonsignificant (P = 0.3173 or 0.3657), suggesting the altered topology is not in conflict with molecular data. A similar SLP_T was conducted on a constraint tree reflecting the classification of Schumann (1904⇓; i.e., Globba exclusive of Mantisia is monophyletic as are the sections of Globba based exclusively on anther appendage number). This topology (24 MPTs, 1181 steps) is significantly noncongruent as compared to the unconstrained data set (P ≤ 0.0001). A less constrained topology uniting all Globba species exclusive of sect. Mantisia (with no other constraints) resulted in eight MPTs of 1167 steps and also failed the SLP_T (P = 0.0253 or 0.0330). Thus, while we can accept the altered topology as a reasonable possibility, we must reject the hypothesis that Globba is monophyletic exclusive of sect. Mantisia and that anther appendage number alone can delineate monophyletic groups within Globba (as would be necessary to retain these groups in the sense of Schumann, 1904⇓).

DISCUSSION

The status of tribe Globbeae

Although all members of the Globbeae are outside the Zingibereae clade, the relationship between these groups cannot be resolved with a high degree of confidence using these data alone. The combined ITS-matK parsimony data resolve the Globbeae as monophyletic and sister to tribe Zingibereae (Fig. 10). The two Caulokaempferia taxa sampled are placed sister to the Globbeae-Zingibereae clade. This result is essentially the same as Kress et al. (2002)⇓ except that the strict consensus trees in that analysis showed these four lineages forming a resolved grade (bootstrap trees were identical to the present analysis). Likewise, increasing taxon sampling by combining the entire Kress et al. (2002)⇓ data set with the entire Globbeae data set (results not shown) brings no further resolution to the issue. Resolving the status and position of the Globbeae and Caulokaempferia awaits further sequencing and analysis of additional DNA regions. As the strict consensus resolves the Globbeae as monophyletic, along with the strong apomorphy of the unilocular ovary in all taxa, we will continue to recognize the tribe. Retaining the Globbeae will also preserve nomenclatural stability within the family, a desirable result unless paraphyly can be unambiguously demonstrated. The infrageneric and subfamilial relationships of Caulokaempferia are currently under investigation (K. J. Williams, W. J. Kress, and K. Larsen, unpublished data), and preliminary results suggest the genus is polyphyletic, with some taxa being related to Boesenbergia. For these reasons, Caulokaempferia will remain Zingiberoideae incertae sedis until that study is complete.

Gagnepainia and Hemiorchis

The small and poorly understood genera Gagnepainia and Hemiorchis are strongly supported as monophyletic genera and as sister taxa (Fig. 10). Morphological differences between the genera (Table 1) are striking as Hemiorchis (Fig. 6) has a narrow, creeping rhizome, a short stamen, a relatively large, well-developed labellum, and large lateral staminodes. Gagnepainia flowers (Fig. 5) are similar to those of Globba in having a long, arching stamen, small labellum, and lateral staminodes. The rhizome of Gagnepainia, like Globba, is highly compact, rounded, and has numerous long, tuberous roots radiating out from it. The short stamen of Hemiorchis (Fig. 6) has led some authors (Larsen et al., 1998⇓) to suggest the genus may be misplaced in the tribe. However, a morphological examination of Hemiorchis and Gagnepainia clearly supports a close relationship, as they share a common leaf shape, inflorescence structure, fruit type, pollen morphology, and most importantly, a raised central line along the center of the labellum that terminates as a pointed tip or even a third lobe (Figs. 5, 6; Table 1). This third lobe is believed to be a unique character to these genera within the family and is the strongest synapomorphy uniting the genera. In fact, these taxa were all originally described under a single genus (Hemiorchis) until Schumann (1904)⇓ separated them into two genera based on the differences mentioned above. While closely related, Gagnepainia and Hemiorchis are quite distinct and should be maintained as two genera. These genera are sympatric only along an extremely narrow band roughly following the Thailand–Myanmar border between Mae Hong Son and Mae Sot, Thailand.

Mantisia and Globba

Mantisia is clearly nested within Globba in the combined analysis (Fig. 10). This result, in particular the sister relationship of Mantisia to Globba sect. Haplanthera (excluding G. substrigosa), is strongly influenced by the matK data. The close relationship between Mantisia and Globba has been recognized since the first species of Mantisia (M. saltatoria Sims) was described. This species was described three times in the same year, twice as a Globba (Andrews, 1810⇓; Roxburgh, 1810⇓) and once as Mantisia (Sims, 1810⇓). For reasons of priority this species should properly be referred to as M. radicalis (Roxburgh) D. P. Dam & N. Dam, not M. saltatoria Sims. The complex nomenclatural history of Mantisia is summarized by Dam et al. (1997)⇓.

The main differences separating Globba and Mantisia have traditionally been inflorescence position and lateral staminode position (e.g., Sims, 1810⇓; Schumann, 1904⇓; Table 1). Even before the results of this study were known, three species brought the Globba-Mantisia distinction into question. Kurz (1870⇓: p. 84) described G. arracanensis, a species from far western Myanmar that “so much resembles at the first aspect Globba spathulata, Rxb. (Mantisia spathulata, Schult.), that it might easily be taken for it.” The lateral staminode position of this species is consistent with Globba however. More significantly, Burtt and Smith (1968)⇓ described M. wardii, also from western Myanmar. This species flowers terminally on a leafy stem. The position of M. wardii as sister to the rest of Mantisia (Fig. 10) suggests that the evolution of a lateral inflorescence occurred after the shift in lateral staminode position. Further eroding the utility of distinguishing these genera based upon inflorescence position is a recently discovered, as yet undescribed species of Globba sect. Haplanthera from the vicinity of Kanchanaburi, Thailand (Y. Paisooksantivatana, Kasetsart University, unpublished data). This species is the first nonmantisioid Globba known to produce leaves and inflorescences consistently on different shoots. Like Mantisia, flowering occurs only at the beginning of the rainy season. This leaves lateral staminode position, and a potentially distinct chromosome number (2n = 20) in Mantisia, as the distinguishing characteristics separating it from Globba. For these reasons, Mantisia is here transferred into Globba (Appendix 1), but retained as a distinct section.

Mantisia radicalis and M. spathulata (Roxb.) Schult. are not resolved as monophyletic in the large ITS analysis. These taxa form a species complex in need of much field and laboratory work.

By submerging Mantisia into Globba, two characters become synapomorphies for the genus within the Zingiberaceae. All species of Globba (as here defined) have a labellum that is fused to the floral tube for a significant portion (i.e., at least 20%) of its length. Likewise, the flowers in every species of Globba have a floral tube that is conspicuously reflexed to nearly 90° (Figs. 1, 2). In genera with similarly elongated flowers (e.g., Gagnepainia, Hedychium, and Pommereschia), the floral tube is essentially straight. While some genera, especially Boesenbergia, have noticeably deflexed flowers, no genus in the Zingiberaceae except Globba has flowers with a consistently and strongly reflexed floral tube.

Anther appendage evolution, infrageneric relationships, and a new classification of Globba

The combined analysis of ITS and matK data yields a highly resolved strict consensus parsimony tree and 95% majority-rule Bayesian tree (Fig. 10). Unfortunately, the backbone of this tree (i.e., the relationship of the four primary clades) lacks statistical support. While the traditional classification of Globba is not strictly supported by this phylogeny, anther appendage number and shape remain the most important characters for recognizing sections of the genus and ultimately for understanding the evolution of Globba itself. The synthesis of molecular and morphological data necessitates a new classification of Globba that submerges Mantisia into Globba and recognizes three subgenera, seven sections, and two subsections (Table 3, Appendix 2). A topological comparison of the old (i.e., Schumann, 1904⇓) and new classification is shown in Fig. 10, while a diagrammatic tree of the major clades in Globba and the Globbeae is presented in Fig. 11.

View this table:

In this window

Table 3. Comparison of the sections and subsections used in the new classification of Globba presented in this study. The defining morphological characteristics of each subgenus are included in Appendix 2. Medio. is an abbreviation of Mediocalcaratae. Type 1 anther appendages are narrow, nearly linear (at least near tips), and attached to a small area near the base of the anther. Type 2 anther appendages are broad, never linear, and attached across most of the lateral side of the anther

View larger version:

Download as PowerPoint Slide

Fig. 11. Phylogeny of the Globbeae including the subgenera, sections, and subsections of Globba and genera of the Globbeae. Classification corresponds to the new classification presented in the text. Triangles represent the major lineages (i.e., genera, sections, and subsections), as shown in Fig. 10. Numbers inside of triangles are the approximate number of species in each lineage. The topology of the three subgenera relative to each other has been reduced to a basal polytomy. Figures to the right of the clades are the generalized anther appendage shape found in the lineage. Probable gains (+) and losses (−) of appendages are indicated above the branch in which the gain or loss is believed to have occurred. All known chromosome numbers for a clade are listed below the corresponding branch, with three exceptions: Globba nuda (2n = 34), G. marantina (2n = 24, 44, 48, 64, 80, and 96), and G. schomburgkii (2n = 48 and 64), all of which are members of subgenus Globba

In a small number of cases, species sampled more than once in the large ITS phylogeny do not resolve as monophyletic. All cases appear to be lack of resolution due to low molecular divergence between species, except for taxa belonging to subgenusCeratanthera. Globba pendula Roxb. and G. leucantha Miq. are distinct species morphologically but overlap for most of their range, and these results (Fig. 8) could be explained by ancient cases of introgression. Although further study is warranted, these cases do not affect the overall relationships within the genus.

While modifications of the anther are common in the Zingiberaceae, lateral anther appendages of the type seen in Globba are unique to this genus within the family (Larsen et al., 1998⇓). Therefore, anther appendages in Globba apparently have evolved no more distantly than the most recent common ancestor of the genus. The most parsimonious reconstruction of anther appendage number based upon molecular data (Fig. 11) suggests either the evolution of two basal appendages in the ancestor to all modern Globba species with a secondary loss in subgenus Mantisia or two major independent derivations of anther appendages (excluding sect. Mantisia and G. geoffrayi Gagnep., see below). The latter possibility suggests that appendage evolution occurred once in subgenus Ceratanthera (zero to two appendages) and once in subgenus Globba (zero to four appendages). However, support for this topology is weak (<50% bootstrap support, <95% posterior probability value, Bremer support of one; Fig. 10). Another possibility, three steps longer than the MPTs but consistent with morphological characters and not significantly different in topology from the MPTs, suggests that the first branch within Globba is subgenus Mantisia and that the species with linear to triangular appendages (i.e., subgenus Ceratanthera + Globba) are monophyletic. This reconstruction would also place all 2n = 32 (48) taxa (Table 3) into a single clade (along with the 2n = 34 G. nuda K. Larsen, which may have an aneuploid gain of one chromosome). The ambiguity of relationships between the three subgenera of Globba on morphological grounds, and the weak molecular support for this arrangement (Fig. 10), is reflected in the conservative topology (Fig. 11).

The anther appendages found in sect. Mantisia (Fig. 4) are of two types (mucronate in G. wardii, quadrate in the remaining species), neither of which resembles the appendages found in any other Globba species (Figs. 7, 11). The phylogenetic position of sect. Mantisia as nested within a clade of nonappendaged Globba species (Fig. 11) further suggests that the anther appendages of sect. Mantisia are independently derived.

The anther appendages in subgenus Ceratanthera are essentially identical to the lower appendages in the four-appendaged species, while the upper appendages of four-appendaged globbas are extremely similar to those of sect. Nudae subsect. Mediocalcaratae (Fig. 7; Table 3). The upper, bump-like secondary appendages in G. geoffrayi (subgenus Ceratanthera) are considered an autapomorphy (K. J. Williams, personal observation). Morphologically, the homology is quite clear between the upper appendages in four-appendaged Globba species and the solitary pair of appendages in sect. Nudae subsect. Mediocalcaratae, as is the homology of the lower appendages and those of subgenus Ceratanthera (Fig. 7).

It is therefore likely that four independent evolutions of anther appendages and one or two losses occurred in Globba (Fig. 11). These innovations include the evolution of sect. Mantisia appendages, basal appendages, upper appendages, and the bump-like upper appendages of G. geoffrayi. A loss of lower appendages occurred in sect. Nudae subsect. Mediocalcaratae and possibly in subgen. Mantisia (if subgen. Ceratanthera is sister to the rest of the genus as in Fig. 10). Studying anther appendage homology through the use of developmental morphology techniques is difficult as anther appendages are nonvascularized and form very late in the floral development process (Endress, 1994⇓). Advances in our understanding of the genetic basis of floral development may eventually lead to a better understanding of anther appendage development in Globba. The function of anther appendages is unknown, but they may act as a lever, increasing the likelihood of the anther making full contact with a pollinator (Endress, 1994⇓; K. J. Williams, personal observation).

The most unexpected result from the combined analysis (Fig. 10) is the placement of G. substrigosa (= G. aphanantha K. Larsen). This species lacks anther appendages, placing it in sect. Haplanthera in Schumann's (1904)⇓ classification of the genus. However, G. substrigosa is one of the most distinctive species of Globba because of its small stature (rarely exceeding 30 cm), white to pale yellow flowers and fruit, and especially in the combination of zero anther appendages and persistent, colorful, involucrate bracts (Table 3). Persistent involucrate bracts are common in sections Globba, Sempervirens, and Mantisia but are not found in the rest of the genus outside G. substrigosa and two rare, poorly known species of sect. Haplanthera, G. andersonii Baker and G. arracanensis. Morphologically, G. andersonii and G. arracanensis are similar enough to G. substrigosa to unite them into a single, putatively monophyletic, section, at least until samples can be obtained and sequenced to provide definitive placement.

Although the position of the G. substrigosa lineage is not well supported, its placement within subgenus Mantisia and sister to the Haplanthera-Mantisia clade (Fig. 11) is most feasible on morphological grounds. If the G. substrigosa lineage is similar to the ancestor of the Haplanthera-Mantisia lineage, then differential gains and losses of relatively few characters could lead to the evolution of both lineages. In particular, a shift in lateral staminode position onto the filament would create a mantisia-type flower. Conversely, a reduction in bract size and a shift from persistent bracts to caducous ones would result in a taxon assignable to sect. Haplanthera.

The placement of the Mediocalcaratae group with the four-appendaged Globba species (Fig. 11) initially seems odd, but the results of this study suggest that an evolutionary loss of the lower two anther appendages has occurred. Equally interesting is that these species form a clade sister to section Nudae subsect. Nudae. The characteristics of each clade are nearly identical, with anther appendage number being the only significant difference between the taxa (Table 3). The strongest synapomorphy uniting these clades is a fruit type that is unique within the genus (Table 3). Andromonoecy in Globba has been recently discovered (K. J. Williams, unpublished data) and appears to be an additional synapomorphy for sect. Nudae subsect. Nudae.

Globba sects. Globba and Sempervirens form two well-supported clades of approximately equal size and are the largest sections in the genus. One clade consists of taxa that remain evergreen throughout the year under normal conditions, while the other contains taxa that enter dormancy during the dry season (Table 3).

The members of Globba sects. Globba and Sempervirens have similar and distinctive characteristics (Table 3) that were likely gained from a common ancestor. Large, persistent (at least to initial flowering), usually colorful bracts are found without exception in every member of sects. Globba and Sempervirens, but not in sect. Nudae. Persistent bracts are also seen in sects. Mantisia and Substrigosa. Inflorescences in sects. Globba and Sempervirens are always arching to pendulous while in all other species of Globba they are erect (at least at the tips). A unique three base-pair deletion in ITS2 can be found in all members of sects. Globba and Sempervirens sampled, but never in sect. Nudae (or any other species of Globba; Table 3). If Fig. 11 is reflective of the true evolutionary history of the group, then these characters were most probably lost in sect. Nudae. However, support in the combined analysis (Fig. 10) is low (except for Bayesian) and a monophyletic clade of sects. Globba + Sempervirens is a possibility. The innovation of colorful bracts and variable inflorescence morphologies in sects. Globba and Sempervirens is consistent with a hypothesis of pollinator-driven selection for inflorescence traits. However, more work is necessary to verify or refute this possibility.

Biogeography

The overlapping distributions of the genera of the Globbeae and infrageneric groups of Globba (Fig. 12) make unequivocal statements about biogeography difficult; however, at least one clear pattern is evident. Peninsular Thailand, centered at the Isthmus of Kra (Fig. 12), forms a clear biogeographic barrier within the genus. Members of sections Ceratanthera and Sempervirens are distributed south of the Isthmus of Kra, while sections Globba, Haplanthera, Mantisia, Substrigosa, and most of sect. Nudae are distributed north of that point. This distribution pattern suggests that climate adaptation (i.e., strongly seasonal vs. largely aseasonal) was fixed early in the evolution of these groups and is not evolutionarily plastic. Also of note is that no species of Globba, except G. marantina, is distributed east of Wallace's line.

View larger version:

Download as PowerPoint Slide

Fig. 12. Distribution of Globba. Map represents most of South and Southeast Asia plus southern China. Classification corresponds to the new classification presented in the text. Distribution of sections falls within outlines shown (see key). The distribution of G. marantina (sect. Globba) is not included (see Discussion: Biogeography)

Three exceptions to this distribution pattern are known, of which only sect. Nudae subsect. Mediocalcaratae has multiple taxa on both sides of the isthmus. This distribution may be explained by the subsection's specialization for growing on rocks and boulders, a specialized, xeric habitat found in both seasonal and aseasonal climates. Globba marantina (sect. Globba) is distributed exclusively, but widely, south of the Isthmus of Kra (and extreme southern India) from Sri Lanka to the Solomon Islands. Despite this distribution, G. marantina is highly nested within (Figs. 8–10) and has all of the characteristics of sect. Globba, including entering a yearly dormancy period. Therefore, the distribution of G. marantina is not included in Fig. 12, as this weedy apomict appears to have originated in the monsoonal region and spread into the aseasonal tropics, thus obscuring an otherwise clear case of geographic disjunction within the genus. The other exception, G. geoffrayi (sect. Ceratanthera), is noteworthy as it is endemic to western Cambodia and far southeastern Thailand (Chantaburi Province). This region has a low degree of seasonality, and the flora is very similar to that of southern Thailand and northern Malaysia (Whitmore, 1984⇓) and may be a relictual pocket of a once more widespread Malaysian flora.

Future directions

Further exploration of variable DNA regions are necessary to elucidate the relationships of Globba to Gagnepainia and Hemiorchis as well as to the rest of the Zingiberoideae, as increased taxon sampling is unlikely to resolve the issue. Adding variable DNA regions should help to clarify the topology at the base of the Globba tree, thereby allowing for better interpretations of biogeography and anther appendage evolution. Taxon sampling is largely complete in terms of higher-level relationships within the genus. However, sampling of G. andersonii, G. arracanensis, and the undescribed (nonmantisioid) basal flowering Globba species is being pursued, as they are critical to fully understanding the relationships within subgenus Mantisia. Increased field collections of Globba and its relatives are critical to fully understanding this complex group, as numerous new species are certain to be discovered. Areas most in need of exploration are Myanmar, Laos, and Cambodia as well as northeastern India. These areas represent the most poorly explored regions in Asia for plant diversity of all types (Frodin, 2001⇓).

APPENDIX

Formal transfer of Mantisia into Globba, with synonomy.

Globba radicalis Roxb., Asiat. Res. 11: 359 (Jan.–Jun. 1810).
Synonyms: Globba purpurea Andrews, Bot. Repos. 10: t. 615 (Aug. 1810), Mantisia saltatoria Sims, Bot. Mag. t.1320 (Sept. 1810), Globba subulata Roxb., Fl. Ind. 1: 78 (1820). Globba saltatoria (Sims) Roscoe, Monandr. t. 112 (1828). Mantisia radicalis (Roxb.) D. P. Dam & N. Dam, Bull. Bot. Surv. India 34: 190 (1997 “1992”).
Globba spathulata Roxb., Fl. Ind. 1: 80 (1820).
Synonym: Mantisia spathulata (Roxb.) Schult., Mant. 1: 49 (1822).
Globba wardii (B. L. Burtt & R. M. Sm.) K. J. Williams, comb. nov.
Basionym: Mantisia wardii B. L. Burtt & R. M. Sm., Notes Roy. Bot. Gard. Edinburgh. 28: 288 (1968).
Globba wengeri (C. E. C. Fisch.) K. J. Williams, comb. nov.
Basionym: Mantisia wengeri C. E. C. Fisch., Kew Bull. 1931: 283 (1931).

APPENDIX

Typification of infrageneric groupings of Globba recognized in this study. Formal descriptions are provided for subgenera. Latin descriptions are provided for the new taxa Globba sect. Sempervirens and Globba sect. Substrigosa.

Globba L. subgenus Ceratanthera (Horan.) K. J. Williams stat. nov. Anthers with two linear (at least near the tips) appendages attached near the base of anther. Upper appendages lacking, or if present, then minute processes resembling small bumps. Type: G. pendula Roxb., Asiat. Res. 11: 359 (1810). Basionym: Ceratanthera T. Lestib. Ann. Sc. Nat. 15: 335 (1841).
Included taxa:
Globba sect. Ceratanthera (Horan.) Petersen. Nat. Pflanzenfam. 2 (6): 29 (1889). Type: G. pendula Roxb., Asiat. Res. 11: 359 (1810). Basionym:Ceratanthera T. Lestib. Ann. Sc. Nat. 15: 335 (1841).
Globba L. subgenus Globba. Anthers appendages two or four. Upper pair (or only pair) triangular and broadly attached along the sides of the anther. Type: G. marantina L., Mant. Pl. ed. 2: 170 (1771).
Included taxa:
Globba L. sect. Globba. Type: G. marantina L., Mant. Pl. ed. 2: 170 (1771);
Globba sect. Nudae K. Larsen subsect. Mediocalcaratae (K. Schum.) K. J. Williams comb. et stat. nov. Lectotype here designated: G. gracilis K. Schum., Pflanzenr. 4: 145 (1904). Basionym: Globba sect. Ceratanthera ser. Mediocalcaratae K. Schum. Pflanzenr. 4/46: 133 (1904).
Globba sect. Nudae K. Larsen subsect. Nudae. Notes Roy. Bot. Gard. Edinburgh 31: 235 (1972). Type: G. nuda K. Larsen, Notes Roy. Bot. Gard. Edinburgh 31: 235 (1972).
Globba sect. Sempervirens K. J. Williams sectio nov. (A sectionibus ceteris appendicibus quatuor antherum, bracteis persistentibus et decurrentibus in pedunculis ad folio summo, foliis sempervirentibus differt.). Type here designated: G. atrosanguinea Teijsm. & Binn., Natuurw. Tijdschr. Ned.-Indie. 27: 22 (1864).
Globba L. subgenus Mantisia (Sims) K. J. Williams stat. nov. Anther appendages lacking, or if present then two and quadrate or mucronate. Type: Mantisia saltatoria Sims, Bot. Mag. 32: t. 1320 (1810). Basionym:Mantisia Sims, Bot. Mag. 32: t. 1320 (1810).
Included taxa:
Globba sect. Haplanthera (Horan.) Petersen. Nat. Pflanzenfam. 2 (6): 29 (1889). Lectotype here designated: G. orixensis Roxb. Asiat. Res. 11: 358 (1810). Basionym: Globba (unranked) Haplanthera Horan. Prodr. Monogr. Scitam. 19 (1862).
Globba sect. Mantisia. Type: Mantisia saltatoria Sims, Bot. Mag. 32: t. 1320 (1810); Basionym: Mantisia Sims, Bot. Mag. 32: t. 1320 (1810).
Globba sect. Substrigosa K. J. Williams sectio nov. (A sectionibus ceteris antherum sine appendicibus, bracteis et bracteolis grandibus, persistentibus et involucratis differt). Type here designated: G. substrigosaKing ex Baker, Fl. Brit. India. 6: 202 (1890).