| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, UK; and 3Instituto de Ecología, Universidad Nacional Autónoma de México, Apartado Postal 70-275, 04510 México, D.F. México
Received for publication December 14, 2000. Accepted for publication March 15, 2001.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
Key Words: ITS • matK • Orchidaceae • Pleurothallidinae • rDNA • trnL
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
30 genera (Luer, 1986a
), accounting for 15–20% of the species in the entire family. The vast majority are dipteran-, deceit-pollinated epiphytes with sympodial growth, unifoliate nonpseudobulbous stems or "ramicauls," conduplicate leaves, velamentous roots, and an articulation between the pedicel and ovary. Genera have been circumscribed on the basis of number of pollinia—eight, six, four, or two—although there can be either eight or six in Brachionidium Lindl. (Luer, 1986a
) and two or four (one large pair and one small pair) in Myoxanthus Poepp. & Endl. and Lepanthes Sw. (Stenzel, in press). Other floral characters used to distinguish genera include number of stigma lobes, sepal connation, resupination, and similarity of perianth parts (Luer, 1986a
). Luer (1986a, 1987, 2000b)
also placed great weight on the evolution of lip mobility and segregated species having any one of the various mechanisms in which this trait has independently evolved. This feature presumably represents a more efficient method of pollen transfer to vectors not known for their efficiency (e.g., flies), although pollination has never been observed for any of the species with these mechanisms.
From a broadly described Pleurothallis R.Br., several genera have been segregated since 1813: Barbosella Schltr., Lepanthopsis (Cogn.) Ames, Restrepiella Garay & Dunst., Dresslerella Luer, and Frondaria Luer. In addition, the type species of Trichosalpinx Luer and Zootrophion Luer were originally described as species of Specklinia Lindl., currently treated as a subgenus of Pleurothallis (Luer, 1986c
). Even now, Pleurothallis includes some 2000 species grouped artificially into 32 subgenera with numerous sections, subsections, and series (Luer, 1986c, 1989, 1994, 1998a, b, 1999
). Lindley (1859)
was loath to split Pleurothallis further in the absence of distinguishing characters and preferred to maintain the admittedly artificial assemblage for ease of study and identification. Luer (1986c)
agreed with Lindley and followed his broad-based approach, saying that "a Pleurothallis might be described as any pleurothallid that does not fit into any of the other genera," but he divided the genus artificially into numerous subgenera, sections, and subsections using morphological characters and approached the intrageneric classifications of Masdevallia (Luer, 1986b
) and Dracula (Luer, 1993
) similarly.
Identification of morphological and anatomical synapomorphies in the subtribe is complicated by the homoplasy rife in vegetative and floral features (Pridgeon, 1982
), as shown in the cladistic study by Neyland, Urbatsch, and Pridgeon (1995)
. Morphological features such as fleshy or terete leaves, variously connate sepals, attenuated petals with apical osmophores, actively mobile labella, and ornamented ovaries occur in clearly unrelated species (Luer, 1986a
). The same is true for anatomical features such as thickenings in the foliar hypodermis, differentiation of foliar chlorenchyma, and spirally thickened idioblasts (Pridgeon, 1982
; Neyland, Urbatsch, and Pridgeon, 1995
). Most of the above are either xeromorphic adaptations or phenotypic responses to selection pressures imposed by pollinators with similar behaviors. Thus, in the absence of reliably homologous morphological and anatomical characters to interpret as synapomorphies, no satisfactory phylogenetic treatment of this large group has been published to date.
To evaluate the monophyly of the subtribe and constitutent genera and reveal evolutionary relationships and trends, we sequenced the nuclear ribosomal DNA internal transcribed spacers (ITS1 and ITS2) and the 5.8S gene (hereafter simply ITS) for 185 taxa of Pleurothallidinae, including two accessions each of Masdevallia venezuelana, Brachionidium valerioi, and Ophidion pleurothallopsis, as well as the outgroup taxa Dilomilis montana, Neocogniauxia hexaptera, Arpophyllum giganteum, and Isochilus amparoanus (mostly Laeliinae sensu Dressler, 1981, 1993
). All but seven of the 32 subgenera of the megagenus Pleurothallis are represented here by one or more taxa; those subgenera not represented are monospecific or comprise only a few species. As a result, the overall morphological diversity is sampled to minimize spurious attractions; such a strategy is recommended for large study groups in particular (Hillis, 1998
). Internal transcribed spacer sequence variation has been previously used in phylogenetic studies of orchids to identify monophyletic groups at the genus level and below and to provide a molecular basis for taxonomic restructuring, particularly in Cypripedioideae (Cox et al., 1997
), Orchidinae (Pridgeon et al., 1997
; Bateman, Pridgeon, and Chase, 1997
), Catasetinae (Pridgeon and Chase, 1998
), Diseae (Douzery et al., 1999
), Pogoniinae (Cameron and Chase, 1999
), Lycastinae (Ryan et al., 2000
), Laeliinae (van den Berg et al., 2000
), and Maxillarieae (Whitten, Williams, and Chase, 2000
).
To resolve internal nodes in the ITS topology and offer additional evidence from another genome, we also sequenced the plastid gene matK and the trnL intron with the trnL-F intergenic spacer (hereafter simply trnL-F) for a representative subset of the taxa in the ITS study. Sequences of rbcL (Chase et al., 1994
; Kores et al., 1997
; Cameron et al., 1999
; van den Berg, 2000
), matK (Ryan et al., 2000
; van den Berg, 2000
; Whitten, Williams, and Chase, 2000
; Kores et al., in press
), and trnL-F (van den Berg, 2000
; Whitten, Williams, and Chase, 2000
; Kores et al., in press
) have been useful in evaluating higher-level relationships in Orchidaceae by virtue of the relatively conservative evolution of the plastid genome.
Finally, we combined the plastid data with the corresponding ITS sequences for a separate analysis of 58 representative taxa to assess congruence among the separate and combined data sets. In this way we were able to compare topologies of DNA regions with different functional constraints (e.g., coding vs. noncoding, nuclear vs. plastid, concerted evolution in ribosomal ITS sequences) before combining them to limit spurious results in the separate analyses (Johnson and Soltis, 1998
; Wiens, 1998
; Soltis, Soltis, and Chase, 1999
).
Following the definitions proposed by Seelanan, Schnabel, and Wendel (1997), we considered bootstrap consensus trees to be incongruent only if they showed "hard" incongruence (high bootstrap support), which possibly reflects biological processes such as hybridization. "Soft" incongruence (weakly supported conflicts) on the other hand, probably represent stochastic error from undersampling of taxa or characters. The effects of differing functional constraints are best corrected by directly combining data; common patterns in each partition, presumably the historical ones, will be strengthened and overcome the unique patterns created by different constraints (Qiu et al., 1999
).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
are listed in Table 2. All genera of Pleurothallidinae except Chamelophyton Garay (one species) and Teagueia (Luer) Luer (six species) were included in the study. Designation of outgroup was based on ITS nrDNA studies of Laeliinae and other Epidendreae (van den Berg et al., 2000
), a four-loci study of the same groups (van den Berg, 2000
), and rbcL sequences of Orchidaceae (Cameron et al., 1999
): Arpophyllum giganteum. Authority abbreviations for taxa follow those of Brummitt and Powell (1992)
.
|
|
. DNA was then precipitated with 100% ethanol or isopropanol, chilled for at least 24 h at 4°C, pelleted and purified by centrifugation through CsCl2-ethidium bromide (1.55 g/mL) and subsequent dialysis in sterile double-distilled H2O and TE buffer, pH 8.0. For some taxa, DNA was instead purified using QIAquick (Qiagen, Crawley, UK) or CONCERT (Life Technologies, Paisley, UK) silica columns following the manufacturer's protocols. Purified DNAs were then stored at –80°C in the Royal Botanic Gardens, Kew, DNA bank.
Amplification and sequencing
ITS was amplified using the methods and primers described by Baldwin (1992)
or Sun et al. (1994)
. The thermal cycling protocol of the polymerase chain reaction comprised 25–28 cycles, each with 1 min denaturation at 97°C, 1 min annealing at 48–50°C, and an extension of 3 min at 72°C, concluding with an extension of 7 min at 72°C. Amplification was improved with the addition of 1 mol/L betaine (final concentration; Sigma-Aldrich, Poole, Dorset, UK, product no. B0300) to the polymerase chain reaction (PCR) mixture, which is reported to reduce the effects of base-pair composition on DNA strand melting, thus improving PCR primer binding.
Amplification of the matK gene and spacer was achieved in two parts using primers designed for Epidendroideae:—19F (CGTTCTCATATTGCACTATG; Whitten, Williams, and Chase, 2000
), 881R (TMTTCATCAGAATAAGAGT; new for this study), 731F (TCTGGAGTCTTTCTTGAGCGA; new for this study), and trnK-2R (AACTAGTCGGATGGAGTAG; Johnson and Soltis, 1995
). The PCR protocol comprised 27–30 cycles, each with 1 min denaturation at 94°C, 30 sec annealing at 48–52°C, and an extension of 1 min at 72°C, ending with an extension of 7 min at 72°C. Amplification of trnL-F used the primers c and f described by Taberlet et al. (1991)
. The thermal cycling program was the same as that for matK.
Amplified products were cleaned with QIAquick or CONCERT silica columns following manufacturer's protocols, including the addition of guanidinium chloride 35% to remove primer dimers (if present). Cleaned products were then sequenced using the BigDyeTM Terminator Cycle Sequencing Ready Reaction kit with AmpliTaq® DNA Polymerase (Applied Biosystems, Warrington, Cheshire, UK). Unincorporated dye terminators were removed by precipitating with 1 : 25 3 mol/L NaOAc : 100% ethanol and then washing twice with 70% ethanol. Pelleted samples were sequenced on an Applied Biosystems 377 automated sequencer. Sequence editing and assembly of complementary strands were accomplished using the "Sequence Navigator" and "AutoAssembler" programs, respectively, of Applied Biosystems. Each base position was checked for ambiguity and agreement with the complimentary strand. GenBank accession numbers for all sequences used in this study are listed in Table 1. Matrices are also available from AMP (a.pridgeon@rbgkew.org.uk) and MWC (m.chase@rbgkew.org.uk).
Phylogenetic analyses
All sequences were initially aligned with CLUSTAL W (Thompson, Higgins, and Gibson, 1994) and adjusted by eye. Each data set was analyzed with PAUP* version 4.0b4a (Swofford, 2000
) with Arpophyllum giganteum designated as the single outgroup. Gaps were coded as missing values. Sequencing artefacts at the ends of all matrices and two regions of ambiguous alignment in the trnL matrix (totaling 529 bases and 40 bases, respectively) were excluded prior to analysis. Additionally, unambiguous indels of four or more base pairs were coded for all sequences, totaling an additional seven characters for the ITS matrices, three for matK, and 26 for trnL.
For each heuristic search, 1000 replicates of random sequence additions were run using subtree-pruning-regrafting (SPR) branch-swapping with MulTrees under the Fitch criterion (unordered states and equal weights; Fitch, 1971
), but limiting the number of trees saved per replicate to ten to reduce time spent in swapping on large islands of trees. The shortest trees identified from this search were then used as starting trees, swapping on them up to 10 000 trees using tree bisection-reconnection (TBR). Relative support for trees from each of the equally weighted, indel-coded data sets was evaluated with 1000 bootstrap replicates (Felsenstein, 1985
), saving no more than ten trees per replicate. We describe bootstrap values of 85–100% as strong, 75–84% as moderate, and 50–74% as weak.
For comparison with results of Fitch parsimony, successive approximation weighting (SW; Farris, 1969
) using the rescaled consistency index (RC) was applied to the combined data set only (using the Fitch trees to calculate the initial weights) until tree length remained the same in two successive rounds. Tree and character manipulations were accomplished using MacClade version 4.0b9 (Maddison and Maddison, 1997
).
We analyzed patterns of sequence evolution using MacClade and PAUP* (Swofford, 2000
) with the same abridged matrices used in parsimony analysis based on the one successively weighted tree from the combined matrix, which had the highest level of bootstrap support. To determine number of steps, CI (consistency index), and RI (retention index) for tranversions in each of the gene regions, we used a stepmatrix to weight the transitions to zero. Using the transversion data, we then calculated the number of transitions and their CI and RI.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
|
) received strong support (98% and 97%, respectively), but there was no support for the genus s.l. (sensu lato). Interrelationships of the supported groups collapsed in the strict consensus tree (arrowheads in Fig. 1). The two accessions of Brachionidium valerioi (six pollinia) have identical sequences.
|
|
The next successively branching clade (D) is the strongly supported (91%) Lepanthes clade, including Frondaria, Lepanthes, Lepanthopsis (Cogn.) Ames, Pleurothallis subgenus Acuminatia Luer (P. angustilabia, P. linearifolia), Pleurothallis subgenus Specklinia sect. Muscosae Lindl. (P. microgemma, P. corticicola, P. minutalis, P. sertularioides), Trichosalpinx, and Zootrophion (Fig. 1). The Pleurothallis subclade receives 97% support, with strong support as well for the monophyly of Zootrophion (96%), Lepanthes (100%), and Trichosalpinx subgenus Trichosalpinx (100%) but not Trichosalpinx s.l. Trichosalpinx berlineri, a pendent rather than caespitose species in subgenus Trichosalpinx, and T. arbuscula, in subgenus Tubella Luer, are isolated from the others in all shortest trees in positions that receive >50% bootstrap support or collapse in the strict consensus tree.
The Pleurothallis s.s.–Stelis clade (clade E; Fig. 2) encompasses the majority of subgenera of Pleurothallis, most of which are not monophyletic. Pleurothallis mentosa and P. tripterantha (representing two sections of subgenus Specklinia) are moderately supported (78%) as sister to P. mirabilis of subgenus Mirandia Luer. In the Stelis Sw. subclade, Stelis is clearly monophyletic (100%) but is embedded in a grade of Pleurothallis subgenera Crocodeilanthe (Rchb.f. & Warsc.) Luer (P. velaticaulis), Physothallis (Garay) Luer (P. neoharlingii), Physosiphon (Lindl.) Luer (P. tubata, P. tacanensis [in ed.]), Effusia Luer (P. resupinata, P. amparoana, P. immersa), Mystax (P. mystax), Elongatia Luer (P. guttata), Dracontia Luer (P. powellii, P. tuerckheimii, P. cobanensis), and Salpistele lutea (subgenus Salpistele Dressler). Several branches collapsed in the strict consensus, blurring the distinction between the large genus Stelis and a miscellany of Pleurothallis subgenera. Condylago rodrigoi and Pleurothallis segoviensis (subgenus Unciferia Luer) are only weakly supported (59%) as belonging here.
|
Sister to the Pleurothallis s.s.–Stelis clade is the Scaphosepalum Pfitz. clade (clade F; Fig. 2), which comprises a monophyletic Dryadella Luer (100%), several sections of Pleurothallis subgenus Specklinia, including the type of the subgenus (P. lanceola), Pleurothallis subgenera Empusella Luer (P. endotrachys) and Pseudoctomeria (Kränzl.) Luer (P. lentiginosa), Acostaea Schltr., Scaphosepalum, and Platystele Schltr. There was only weak support for the monophyly of the latter two. The two accessions of Pleurothallis costaricensis are sister to each other with strong support (99%), although the sequences are more different than those between other paired accessions of the same species.
The Luerella-Ophidion-Pleurothallis peperomioides group (clade G; Fig. 3) lacks bootstrap support >50% in the ITS analysis but received strong support in the combined treatment (see below). The two accessions of Ophidion pleurothallopsis are sister to each other with 100% support.
|
Finally, sister to Masdevallia-Porroglossum is moderately supported Dracula. Subgenus Sodiroa (Luer) Luer (D. sodiroi) is embedded in species of subgenus Dracula, and subgenus Xenosia Luer (D. xenos) is instead sister to Masdevallia picturata (79%).
Small ITS matrix
This data set comprised 58 taxa, representing each of the major clades and outgroups identified in the larger analysis and also those taxa in the plastid and combined analyses. The same seven indels were included as before. Of the 394 variable sites (51%), 296 (75%) were potentially parsimony informative. Twelve most parsimonious trees of Fitch length 1749 were produced with a CI of 0.39 and RI of 0.52 (Table 3).
The ITS bootstrap consensus tree (Fig. 4) shows that the level of support for monophyly of Pleurothallidinae increased slightly to 94%, whereas that for inclusion of Dilomilis and Neocogniauxia was still weak. Except for clades F and H, resolution is poor. Support for the monophyly of Masdevallia increased to 80%, and support remained high for the entire Dracula–Masdevallia–Porroglossum–Trisetella clade. Sister relationships were somewhat better supported than in the complete ITS study for Scaphosepalum–Platystele, Pleurothallis lentiginosa–P. endotrachys, P. cardiantha–P. ruscifolia, P. mentosa–P. tripterantha, Myoxanthus uncinatus–M. aspasicensis, and for clades D and E.
|
|
|
The bootstrap consensus tree (Fig. 6) is slightly more resolved than either the matK or small ITS tree and supports the monophyly of Pleurothallidinae, more strongly with the inclusion of Dilomilis and Neocogniauxia. There is a clear relationship (100%) between Luerella Braas and P. peperomioides (clade G) not present in the other topologies. Compared to the matK analysis, there is stronger support for clades D and H, although as before there is still strong support for clade C, P. mentosa–P. tripterantha and P. ruscifolia–P. cardiantha in clade E, and Myoxanthus uncinatus–M. aspasicensis (excluding M. punctatus) in clade B.
|
A heuristic search of the combined matrices resulted in four trees of 4180 (CI = 0.51, RI = 0.54). After three rounds of successive weighting, one of the Fitch trees was identified as the optimal SW tree, SW length = 1382.888 steps, SW CI = 0.84, and SW RI = 0.79 (Fitch length = 4180; Table 3). Figure 7 shows the single successively weighted tree with Fitch lengths above the branches and equally weighted bootstrap percentages below.
Based on the single successively weighted tree from the combined analysis (Fig. 7; Table 5), the number of transversions in ITS1 was 336 (CI = 0.41; RI = 0.46) and the number of transitions 600 (CI = 0.35; RI = 0.52). For ITS2, 274 of the 783 steps were transversions (CI = 0.46; RI = 0.57) and 509 were transitions (CI = 0.37; RI = 0.51). Finally, of the 32 changes in 5.8S, nine were transversions (CI = 0.56; RI = 0.20) and 23 transitions (CI = 0.42; RI = 0.73). The transition/transversion (ts/tv) ratio for the coding region (5.8S) was predictably higher (2.56) than for either ITS1 (1.79) or ITS2 (1.86). Again based on the single successively weighted tree, 682 (60%) of the changes in the matK tree were transversions (CI = 0.51; RI = 0.48) and 448 (40%) transitions (CI = 0.69; RI = 0.56) with a ts/tv ratio of 0.66 (Table 5). As for trnL, transversions for all three regions—intron, exon, and spacer—contributed more than transitions did to the number of steps (62, 63, and 70%, respectively; Table 5). The intron had the highest ts/tv value (0.61) and the spacer the lowest (0.43). The CI of transitions in all three regions (0.74, 0.76, 0.84, respectively) was higher than that of tranversions, as was the RI of transitions except for the exon (0.79, 0.23, 0.67, respectively; Table 5).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONLITERATURE CITED |
|---|
), there is a substantial excess of transversions over transitions in matK with an equivalent ts/tv ratio (0.66). Most changes occur in the third-position codon in both studies, although there are slightly more in Pleurothallidinae (41.7%) than in Maxillarieae (39.2%). First-codon changes in both studies account for 32% of the total. The proportion of change at third positions for matK is substantially less than for other plastid protein-coding genes, such as rbcL and atpB (Savolainen et al., 2000
). The rate of change at variable positions in matK is similar to that of trnL-F (i.e., the average changes per variable site). The situation is reversed in ITS nrDNA, in which an excess of transitions was observed (Table 5), which Bakker et al. (2000)
attributed to different patterns of evolution between nuclear and plastid DNA.
Sampling effects
Comparisons of the statistics for the large and small data sets of ITS (Table 3) show that adding taxa greatly increases the estimated number of changes at variable sites and therefore lowers the CI slightly but produces a much higher RI and perhaps more accurate result by better revealing the structure of homoplasy. If the object of phylogenetic studies is to determine accurately the number of times each character changes, then increased sampling is likely to produce a better result due to the drastically greater number of changes detected in variable base positions. If the improved RI is an accurate measure of character performance, then increased taxonomic sampling improves the information content of each variable position because the RI is higher with the larger matrix than it is with the smaller matrix. Ironically, in this case, the analysis with the higher RI receives lower overall levels of bootstrap support, but that information alone does not tell us which result is more accurate.
To us, it would be interesting to know which topology is more accurate, the one with higher bootstrap support or the one with a higher RI. If we can assume that the combined tree is more accurate than any of the trees from the component matrices (because it has higher levels of overall bootstrap support than any of the individual analyses), then by standardizing the numbers of taxa and characters to those held in common by all analyses, we can ask which version of the ITS matrix (large or small) produced a tree more similar to the combined tree (i.e., the tree with a length more similar to the combined tree is therefore the more accurate one). Compared to the combined tree topology, the better-sampled (larger) ITS analysis is more similar in length; the length of the small ITS data set was 1749 steps, whereas that of the large ITS matrix stripped to just the taxa in the combined analysis was 1763 steps; the ITS contribution to the combined tree was 1761 steps. Thus, the length of the trees produced with more taxa was more like that produced with more characters (matK and trnL-F plus ITS). Which of the three is the most accurate is still a question we cannot answer, but it seems clear to us in this case that RI is a better measure of data performance than overall levels of bootstrap support. However, the use of overall measures for a tree (such as RI, which tells us how well character changes collectively fit the resulting trees) cannot tell us anything about the accuracy of individual groups, so ultimately bootstrap support for a given clade is the only measure of internal support available.
Many previous phylogenetic studies of DNA sequence data have used frequency of change as the basis for weighting under the assumption that characters that change more frequently are less reliable. However, based on the combined analysis, frequency of change and performance (as measured by the RI) are not correlated. Therefore, if weighting of any sort is to be employed, it must not be based on whole-category weights, which is why we employed successive approximations weighting. A similar conclusion about whole-category weighting was reached by Olmstead, Reeves, and Yen (1998)
, although they favored simply eliminating characters that were changing excessively (i.e., weighting them to zero), whereas SW uses a graded scale.
Previous cladistic studies
In a cladistic study using 45 morphological and anatomical characters, Neyland, Urbatsch, and Pridgeon (1995)
also designated Arpophyllum giganteum as outgroup along with Brassavola nodosa (L.) Lindl. and Epidendrum ciliare L. of Laeliinae. Some results were similar to those reported here, e.g., Porroglossum was sister to Masdevallia, and Trisetella was sister to both of them (but they fell in the same clade as Scaphosepalum–Platystele–Dryadella). Furthermore, Lepanthes was sister to sect. Hymenodanthe of Pleurothallis subgenus Specklinia instead of Lepanthopsis, which was part of a polytomy with Pleurothallis s.s. and Restrepia. Brachionidium was sister to Dracula, a relationship based in large part on the absence of a leaf hypodermis. In light of these results, the anatomical similarity between those two genera represents a reversal in Dracula, and perhaps also in Brachionidium, rather than a synapomorphy. Although the morphological analysis likewise clearly showed the polyphyly of Pleurothallis, the distribution of its various components differed substantially from the highly bootstrap supported topology shown here.
Molecules vs. morphology
The results of the Neyland, Urbatsch, and Pridgeon (1995)
attempt to classify Pleurothallidinae based on morphological/anatomical data differ radically from the results here in part because of the homoplasy of most of the characters used. For example, wall thickenings of leaf hypodermal cells and mesophyll idioblasts (specializations for water storage) have arisen or been lost several times in the phylogeny of the group (Fig. 7). The only unequivocal anatomical synapomorphy of which we are aware is that of elevated, cyclocytic stomata in Dresslerella and Myoxanthus subgenera Silenia and Satyria. The morphological analysis also suffered simply because not enough reliable characters relative to the number of taxa were analyzed.
Some of the morphological characters adduced by Luer (1986a)
, in assorted combinations, to classify Pleurothallidinae (Fig. 7) include presence of an annulus on the stem or ramicaul, variously connate sepals, presence of actively motile lips, anther position (apical, subapical, or ventral), number of pollinia (two, four, six, or eight), bilobed stigmas, and a number of autapomorphies. Actively motile lips have evolved independently in Acostaea, Condylago, Porroglossum, and Masdevallia teaguei, and bilobed stigmas occur in such unrelated genera as Lepanthes, Pleurothallis s.s./Stelis, and Platystele. The anther is apical, subapical, or ventral within every one of the major clades identified here. Number of pollinia is not necessarily evidence of relationship, as illustrated above by Restrepiopsis and Pleurothallopsis.
Such homoplasy in floral characters and their lack of congruence with molecular data are attributable to the fact that flowers of most Pleurothallidinae are deceit-pollinated and exhibit various combinations of the same features: (1) attract dipterans by simulation of a food source or breeding site, (2) literally toss them against the column for deposition and extraction of pollinia (though this has never observed in nature), or (3) limit pollinator size by the size of openings in connate sepals. In orchid groups such as Catasetinae and Stanhopeinae that offer rewards in the form of floral fragrances, however, floral characters more closely reflect phylogenetic relationships based on molecular data (Chase and Hills, 1992
; Pridgeon and Chase, 1998
; Whitten, Williams, and Chase, 2000
).
On the other hand, the stem annulus, which is not subject to pollination selection pressures, is consistently absent (Fig. 7) in the more ancestral genera represented by clades A, B, and C but (except for P. peperomioides) consistently present in all derived genera, which have two pollinia (Stern, Pridgeon, and Luer, 1985
).
The anatomy and morphology of Pleurothallidinae are well studied compared to many orchids, but these features are relatively homogenous, leaving us with the distinct impression that more study of nonmolecular characters is unlikely to reveal the vastly greater numbers of characters required for a more accurate assessment of intergeneric relationships. This is, of course, why we turned to DNA sequence analyses for additional characters. Compared to the extensive number of species that have to be sampled to characterize accurately the distribution of any newly found potentially useful character, DNA sequence studies are a vastly more efficient, if less elegant, method of finding useful information. We believe that both types of studies are required to understand the evolutionary history of Pleurothallidinae, but it is also clear to us that if we are going to make rapid strides in understanding these often bizarre and fascinating plants before they become extinct, then molecular studies are the only option that is likely to succeed in the limited time remaining. Furthermore, studies of DNA sequences have been demonstrated to produce results that are well corroborated by other characters, both at the level of angiosperm families (Nandi, Chase, and Endress, 1998
) as well as at the level of genus (Rudall et al., 1998, 2000
) and species (Cameron and Chase, 1999
). The results presented here demonstrate that the morphological characters of Neyland, Urbatsch, and Pridgeon (1995)
change much more frequently when mapped onto the molecular topology than they did in Neyland, Urbatsch, and Pridgeon (1995)
, but not to the point that their change is random. It is clear that their patterns of change are still highly structured when optimized on the DNA-based topology, just less so than on the morphological topology.
Subtribe Pleurothallidinae and outgroups
Dressler (1993)
suggested that Pleurothallidinae could be sister to or derived from something similar to Dilomilis, which has eight pollinia and reed stems with persistent leaf sheaths (Ackerman, 1995
). Its sister genus, Neocogniauxia, has sheathed stems terminated by a single leaf. The leaf anatomy of both (Baker, 1972
) is similar in many respects to that of most Pleurothallidinae: adaxial and abaxial hypodermis, helically thickened mesophyll cells, and absence of extravascular fibers (Pridgeon, 1982
). In the matK, trnL, and combined analyses here, Dilomilis montana and Neocogniauxia hexaptera received strong support (91, 85, and 98%, respectively) as the sister taxa to Pleurothallidinae. The comprehensive ITS study of Laeliinae (van den Berg et al., 2000
), the rbcL study of Orchidaceae (Cameron et al., 1999
), the four-region study of Epidendreae and Laeliinae (van den Berg, 2000
), and the mitochondrial DNA study by Freudenstein, Senyo, and Chase (2000)
offered even stronger support for inclusion of Dilomilis and Neocogniauxia in Pleurothallidinae. There is only one morphological synapomorphy uniting the members of Pleurothallidinae as presently understood—an articulation between the ovary and pedicel—that Dilomilis and Neocogniauxia lack. However, taking into account the highly supported molecular evidence from multiple DNA regions, the shared number of pollinia in some taxa (eight) and leaf anatomy, the ancestral reed-stem condition in other clades of Epidendroideae (van den Berg, 2000
), and evolutionary remnants thereof in present-day Pleurothallidinae (see below), Pleurothallidinae should be expanded to include Dilomilis, Neocogniauxia, and presumably the monospecific Tomzanonia Nir (segregated from Dilomilis by Nir, 1997
), thereby forming a more natural unit. Furthermore, recognition of a new subtribe comprising only three genera that are collectively sister to Pleurothallidinae seems an unnecessary case of taxonomic inflation.
Clade A
Octomeria, a genus of 150 species distributed throughout the Neotropics but most diverse in Brazil, is sister to the rest of Pleurothallidinae and has features that could be considered unspecialized, e.g., eight pollinia and a stem without an annulus at the insertion of the inflorescence (Fig. 7). A relationship with Brachionidium, which likewise has subsimilar sepals and petals, eight pollinia (some species), and also lacks an annulus (Luer, 1986a
; Stenzel, in press), received moderate support in the matK and combined analyses but was at best weakly supported in the remaining analyses. Sampling additional species of Brachionidium and Octomeria as well as Chamelophyton might alter the position of Brachionidium.
Clade B
This clade, which received <50% support (Fig. 7), comprises clearly monophyletic genera with four or eight pollinia (Restrepia, Restrepiella, Barbosella including Barbrodia, and Restrepiopsis–Pleurothallopsis Porto & Brade). All lack an annulus (the ancestral condition), but there are specializations to attract pollinators. Species of Restrepia have well-developed osmophores at apices of the dorsal sepal and petals (Pridgeon and Stern, 1983
), and more generalized osmophores occur over the sepals of Restrepiella (A. M. Pridgeon and W. L. Stern, unpublished data).
Luer (2000b)
maintained that on the basis of the different form of attachment of the lip to the column (a simple hinge instead of the ball-and-socket articulation that characterizes Barbosella species), Barbrodia should be maintained as a monospecific genus. The same ball-and-socket joint also occurs in Pleurothallis subgenus Crocodeilanthe and elsewhere. The same holds true for the column character (apical anther) that Luer (2000b)
used to distinguish Barbrodia from Barbosella (with a ventral anther). An apical anther also occurs in such unrelated taxa as Porroglossum, Lepanthes, Andinia Luer, Acostaea, and Pleurothallis s.s. Contrary to Luer's segregation of Barbrodia, it is embedded within Barbosella in all analyses (with 100% support) and cannot be maintained if Barbosella is to remain monophyletic.
There is strong support (88–100%) in all analyses for a sister relationship between Restrepiopsis (four pollinia) and Pleurothallopsis (eight pollinia), with relatively few steps between them to justify treating them as separate genera. Pleurothallopsis has been treated as a subgenus of Octomeria (Luer, 1991
), and this relationship is strongly refuted. This disparity casts doubt on the conventional wisdom (Brieger, 1977
; Freudenstein and Rasmussen, 1999
; and others) of using pollinium number to circumscribe genera, which has also been shown to vary in Broughtonia R.Br. (van den Berg, 2000
), Myoxanthus (Stenzel, in press), and Maxillarieae in general. Reduction in number is a multiple parallelism in Laeliinae (van den Berg et al., 2000
) and probably also in Pleurothallidinae. However, from these data it is impossible to determine with certainty how many times this has occurred in Pleurothallidinae.
Dresslerella and Myoxanthus s.l. form a polytomy with the remaining genera in the bootstrap consensus trees of most analyses but comprise a clade with <50% support in the combined analysis. The Pleurothallis subgenus Acianthera clade (C) is next to clade B in the grade in the successively weighted tree, but the relationship between it and clade B received <50% bootstrap support. The Myoxanthus uncinatus–M. aspasicensis group, treated by Luer as subgenus Silenia (1992), is highly supported (92–100%) and may include the M. punctatus group (subgenus Myoxanthus) based on these results. Myoxanthus subgenus Silenia and subgenus Satyria Luer share cyclocytic, elevated foliar stomata with Dresslerella (Pridgeon and Williams, 1979
; Pridgeon, 1982
; Pridgeon and Stern, 1982
), a significant synapomorphy not yet found elsewhere in Orchidaceae, including subgenus Myoxanthus. On the other hand, subgenera Silenia and Satyria both lack the autapomorphic coralloid raphide clusters that characterize the foliar epidermis of species in subgenus Myoxanthus (Pridgeon, 1982
; Pridgeon and Stern, 1982
). Both Dresslerella and Myoxanthus s.l. also lack an annulus on the stem or ramicaul. Further, the discovery of an additional, smaller pair of pollinia in some species of subgenus Myoxanthus (Stenzel, in press) strengthens the link to Dresslerella, which also has one pair of large and one pair of small pollinia.
Clade C
In the plastid and combined analyses there is strong support (90–100%) for a monophyletic P. subgenus Acianthera, far removed from the type clade of Pleurothallis, although the ITS trees provide strong support for only three of the terminal clades (Fig. 1). From Luer's (1986c)
treatment, sections Brachystachyae Lindl. (P. johnsonii, P. leptotifolia, P. ochreata, P. saurocephala, P. strupifolia), Cryptophoranthae Luer (P. fenestrata), Phloeophilae Luer (P. raduliglossa), Sicariae Lindl. (P. circumplexa, P. luteola, P. pectinata, P. prolifera, P. sicaria), and Tricarinatae Luer (P. glumacea) are all represented, though not necessarily in subclades reflecting these same relationships. In addition, subgenera Arthrosia (P. auriculata) and a monophyletic Sarracenella (P. sarracenia, P. asaroides) are embedded in P. subgenus Acianthera (Fig. 1). Like other species in P. subgenus Acianthera, species in these two subgenera lack an annulus on the stem. Pleurothallis peperomioides of subgenus Phloeophilae is placed not with subgenus Acianthera at all but with Luerella (see below). Pleurothallis melanocthoda, which Luer (1996)
assigned to subgenus Specklinia (Lindl.) Garay sect. Muscosae Lindl., is instead a member of subgenus Acianthera according to the complete ITS study. All species in clade C and those that follow below have only two pollinia.
Clade D
In many respects this clade, which received strong support in every analysis except matK and 100% support in the combined analysis, is the most interesting and unanticipated from a phylogenetic perspective. It includes Lepanthes and Zootrophion, genera with highly divergent floral morphologies, as well as Pleurothallis subgenus Acuminatia, P. subgenus Specklinia sect. Muscosae, Frondaria, and Trichosalpinx. What unites this florally disparate group is a many-noded stem with infundibular sheaths, either imbricating as in Zootrophion and P. subgenus Specklinia sect. Muscosae or sclerotic as in Lepanthes, Lepanthopsis, Trichosalpinx, and some members of P. subgenus Acuminatia. Reduced and sclerified sheaths would reduce water loss at lower elevations or during the alternating wet and dry periods of the Pleistocene described by Gentry (1982)
. It is most parsimonious to explain the expanded leaf sheaths on the stem of Frondaria as a reversal toward the reed-stem condition; indeed, with the leaf sheaths subtending a true apical leaf it approaches Neocogniauxia in vegetative morphology.
All genera in this clade are monophyletic with the possible exception of Lepanthopsis (only one species was studied) and Trichosalpinx. There is at best a weakly supported relationship between the pendent species T. berlineri and the erect species-pair T. orbicularis–T. blaisdellii in subgenus Trichosalpinx (Figs. 1, 4–7). Subgenus Tubella, represented by T. arbuscula here, has a different, proliferating habit such that new plants arise from nodes on the stem or ramicaul (Luer, 1997
); Trichosalpinx arbuscula has an isolated position (Fig. 1) with no apparent relationship to T. berlineri or T. orbicularis–T. blaisdellii.
Clade E
Three subclades comprise this strongly supported clade (Fig. 7). The first unites sections Mentosae Luer and monospecific Tripteranthae Luer of Pleurothallis subgenus Specklinia with monospecific subgenus Mirabilia Luer (Figs. 2, 4–7). Vegetatively all three groups are similar, the latter differing florally from the others in having a much longer column foot.
Sister to this subclade is the type group of Pleurothallis, including P. ruscifolia and P. cardiantha. It, too, is a highly supported group in all analyses (Figs. 2, 4–7). However, the monophyly of the subclade can be established only if several other subgenera, separated by only a few steps (Fig. 2), are sunk into it: Scopula (P. penicillata), Ancipitia (P. viduata, P. niveoglobula), Mirandia (P. miranda), Restrepioidia (P. hemirhoda), Rhynchopera (P. loranthophylla), and Talpinaria (P. talpinaria). Furthermore, some sections and subsections of P. subgenus Pleurothallis are not supported as monophyletic units. For example, P. truncata (sect. Truncatae Luer) forms a polytomy with P. teaguei, P. cardiantha, and P. cardiothallis (sect. Pleurothallis subsect. Macrophyllae–Fasciculatae Lindl.). Pleurothallis rowleei and P. allenii (sect. Pleurothallis subsect. Acroniae Luer) form a polytomy with other subgenera (Fig. 2). Criteria used to distinguish these subgeneric taxa are almost exclusively floral: shape and margins of the petals, shape of the lip, etc. (Luer, 1986a
). It is clear from the above that reliance on trivial characters that often are subjectively emphasized over others has led to gross taxonomic inflation, which obscures and even contradicts much closer relationships in Pleurothallis s.s.
The polyphyly of Pleurothallis is further reflected in the Stelis subclade of clade E, in which seven subgenera of Pleurothallis (Dracontia, Elongatia, Mystax Luer, Effusia, Physosiphon (Lindl.) Luer, Physothallis (Garay) Luer, Crocodeilanthe) and Salpistele lutea form a grade in the complete ITS tree to a monophyletic Stelis s.s. (Fig. 2). Sister to the subclade are P. segoviensis (subgenus Unciferia) and the monospecific genus Condylago Luer. In the combined tree (Fig. 7), P. segoviensis is still sister to that subclade in a weakly supported alliance with P. amparoana.
Stelis s.s. is distinguished florally by (1) equal to subequal, variously connate or almost free sepals; (2) small, transversely elongated, fleshy petals more or less thickened on the margins; (3) a concave, fleshy, three-sided lip; and (4) a short, broad column with or without a column foot but with an apical anther and often bilobed stigma (Garay, 1979
; Luer, 1986a
). Stelis ciliaris, which Garay (1979)
segregated with 32 other species as the genus Apatostelis Garay on the basis of having only one stigma lobe instead of two, is here embedded among other Stelis species. There is thus no support for recognition of Apatostelis. Perhaps more important, this result diminishes the reliability of number of stigma lobes as a taxonomic character at the genus level.
Sister to Stelis s.s. (Figs. 2, 4–7) is P. velaticaulis of P. subgenus Crocodeilanthe. Species in this subgenus as well as subgenus Pseudostelis, removed from Crocodeilanthe by Luer (1999)
, are vegetatively similar to Stelis and also have an apical anther, connate lateral sepals, and concave lip. Pleurothallis neoharlingii (subgenus Physothallis), sister to Stelis–P. velaticaulis, also bears a resemblance to sect. Nexipous (Garay) Luer of Stelis with its lateral sepals more connate to the dorsal sepal than to each other. Next in the grade (Fig. 2) are P. tubata and P. tacanensis of subgenus Physosiphon, characterized by a tubular calyx and a vegetative morphology indistinguishable from most species of Stelis, the former species originally described as Stelis tubata Lodd. The remaining members of the grade in Fig. 2, representing four other subgenera of Pleurothallis s.l., exhibit low levels of divergence. It seems likely that in the evolution of this clade, vegetative morphology remained essentially unchanged, but there was progressive connation of sepals (often with trichomes) and thickening of all floral parts, culminating with the highly reduced but fleshy petals and lip of Stelis s.s. Condylago, a monospecific genus that Luer (1986a, 1987)
likened to P. flexuosa of subgenus Effusia (Luer, 2000b
), differs only from that subgenus by having a sensitive lip, which arose independently in other clades (see below). In light of the low levels of divergence, the vegetative similarities and floral homoplasy, and moderate (81%) support in the combined tree for an expanded, phylogenetic concept of Stelis, there is little justification for continuing to recognize Salpistele, Condylago, and the several subgenera of Pleurothallis as anything but species of Stelis, the sister genus to Pleurothallis s.s. These taxonomic transfers will be made elsewhere.
Clade F
The disparate membership of this strongly supported clade (excluding Andinia) in the combined analysis (Fig. 7) underscores the problems associated with excessive reliance on floral features for delimitation of taxonomic categories. Scaphosepalum shares several synapomorphies and indels with Platystele, as unexpected a pairing as Lepanthes–Zootrophion. Scaphosepalum flowers are nonresupinate with an undivided stigma and prominent osmophores on lateral and/or dorsal sepals (Pridgeon and Stern, 1985
; Luer, 1988
), whereas Platystele flowers are resupinate with a bilobed stigma and generalized osmophores. Apart from general differences in size of the plants (species of the latter are among the smallest Neotropical orchids), the two genera are almost identical vegetatively: (1) stem with an annulus shorter than the leaf and enclosed by two or three imbricating sheaths and (2) leaf thinly to thickly coriaceous, obovate, the apex notched with a mucro or apiculum in the sinus (Luer, 1988, 1990
). Sister to the Scaphosepalum–Platystele subclade with 94% support (Fig. 7) is a subclade comprising species of Pleurothallis subgenus Specklinia (sects. Hymenodanthe Barb.Rodr., Tribuloides Luer, Muscariae Luer), P. subgenus Empusella, P. subgenus Pseudoctomeria, and Acostaea. Once again the low levels of sequence divergence (Fig. 2) indicate that many of the current intrageneric concepts of Pleurothallis are trivial, and all taxa in this subclade could be accommodated in the resurrected genus Specklinia Lindl. (lectotype Epidendrum lanceola Sw., included here). That Acostaea, which like Condylago has a sensitive lip (but a different mechanism), is embedded among species of subgenus Specklinia sister to Pleurothallis costaricensis once again reveals problems with a priori, subjective weighting of the sensitive lip in generic circumscription. Dryadella is sister to both of these subclades (Fig. 7), and Andinia, segregated on the basis of these studies from Salpistele Dressler (Luer, 2000b
), is supported (84%) as sister to all the rest in clade F.
Clade G
Sister to the Clade H is the small clade comprising Luerella and Pleurothallis peperomioides. Although resolved in the large ITS tree (Fig. 3), there is <50% bootstrap support for a relationship between them. However, there is 100% support for a sister relationship between them in the trnL-F analysis (Fig. 6) and 92% in the combined tree (Fig. 7). Vegetatively, P. peperomioides differs from Luerella in having a creeping habit and small, round leaves. There are differences in lip and petal morphology and in the degree of sepal connation. Furthermore, the stem of Luerella has an annulus (Stern, Pridgeon, and Luer, 1985
), whereas that of P. peperomioides is said to lack one (Luer, 1986c
). The latter is one of nine species attributed to sect. Phloeophilae Luer of subgenus Acianthera (Luer, 1986c
). Another in this section is P. raduliglossa, which in this study fell within Acianthera (Fig. 1) rather than with P. peperomioides, indicating that the section is not monophyletic. Ophidion could be a member of this group, but the combined tree resolves it as a member of clade E without bootstrap support >50%. Its floral and vegetative morphology compare well with the other members of clade G, so perhaps the ITS result (Fig. 3) that places it with these will end up being accurate. Overall none of the data collected resolve its position with any confidence.
Clade H
This clade comprises four monophyletic genera—Dracula, Masdevallia, Porroglossum, Trisetella—and Masdevallia erinacea. Dracula xenos, which is sister to M. picturata in the ITS trees, is potentially a natural hybrid between M. picturata and an unspecified Dracula, as it has a Masdevallia habit but the distinctive lip of Dracula composed of a hypochile and an epichile with radiating lamellae. Artificial hybrids between the two genera inherit the Dracula lip and fit this interpretation (C. Head, J & L Orchids, Easton, Connecticut, USA, personal communication). The lack of ITS sequence heterogeneity for D. xenos may mean that through gene conversion the maternal ITS pattern has been retained, which would explain why the same position for D. xenos appears in plastid DNA results. The other interpretation is that it indeed is a species of Masdevallia that independently developed the Dracula-type lip; given the high homoplasy observed in orchid floral morphology, especially in this subtribe, this is a viable hypothesis.
Although there is a low level of molecular divergence among the species of Masdevallia, we are not proposing further changes to the finely split subgeneric classification of Luer (1986b, 2000b)
. Erection of such a complicated subgeneric classification appears to us as unnecessary and unlikely to reflect evolutionary patterns, leading us to question as well its accuracy and therefore the utility of such a complex infrageneric scheme. However, there is no justification for the erection of the monospecific genus Jostia (Luer, 2000b
) to accommodate M. teaguei solely on the basis of its sensitive lip, a feature that has arisen independently as many as four times in clades E, F, and H (Fig. 7). Much the same holds true for the infrageneric scheme of Dracula, well represented in this study.
Masdevallia erinacea is one of five species in subgenus Masdevallia sect. Pygmaeae (Luer, 1986b
) distinguished from other species of Masdevallia by possessing carinate and echinate or papillose ovaries. In the large ITS study (Fig. 3) here, M. erinacea is sister to Dracula–Masdevallia–Porroglossum; it and its relatives warrant generic status if Masdevallia is to remain monophyletic.
Conclusions
This is the first extensive molecular phylogenetic study of subtribe Pleurothallidinae. In this analysis, we have been able to assess generic circumscriptions with hundreds of characters and produce trees for which internal support can be assessed rather than relying on a few characters selected a priori for reasons that are subjective and inconsistent. Just as important, we can now reorganize the highly polyphyletic taxa under the umbrella of Pleurothallis and bring order to an artificial megagenus and a subtribe that have confounded taxonomists since the time of Lindley. The minimal divergence among the various subgenera, sections, and even series of Pleurothallis, Masdevallia, and Dracula blurs the
