(American Journal of Botany. 2001;88:503-533.)
© 2001 Botanical Society of
America, Inc.
The dalbergioid legumes (Fabaceae): delimitation of a pantropical monophyletic clade1
Matt Lavin2,3,
R. Toby Pennington 4 ,
Bente B. Klitgaard 5 ,
Janet I. Sprent 6 ,
Haroldo Cavalcante de Lima 7 and
Peter E. Gasson 5
3Department of Plant Sciences, Montana State University, Bozeman, Montana 59717 USA;
4Tropical Biology Group, Royal Botanic Garden Edinburgh, 20a Inverleith Row, Edinburgh EH3 5LR, UK;
5Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3AB, UK;
6Department of Biological Sciences, University of Dundee, Dundee DD1 4HN, UK; and
7Jardim Botânico do Rio de Janeiro, Rua Pacheco Leão No. 915, Gavea 22.460 Rio de Janeiro—RJ, Brazil
Received for publication January 11, 2000.
Accepted for publication June 2, 2000.
 |
ABSTRACT
|
|---|
A monophyletic pantropical group of papilionoid legumes, here referred to as the "dalbergioid" legumes, is circumscribed to include all genera previously referred to the tribes Aeschynomeneae and Adesmieae, the subtribe Bryinae of the Desmodieae, and tribe Dalbergieae except Andira, Hymenolobium, Vatairea, and Vataireopsis. This previously undetected group was discovered with phylogenetic analysis of DNA sequences from the chloroplast trnK (including matK) and trnL introns, and the nuclear ribosomal 5.8S and flanking internal transcribed spacers 1 and 2. All dalbergioids belong to one of three well-supported subclades, the Adesmia, Dalbergia, and Pterocarpus clades. The dalbergioid clade and its three main subclades are cryptic in the sense that they are genetically distinct but poorly, if at all, distinguished by nonmolecular data. Traditionally important taxonomic characters, such as arborescent habit, free stamens, and lomented pods, do not provide support for the major clades identified by the molecular analysis. Short shoots, glandular-based trichomes, bilabiate calyces, and aeschynomenoid root nodules, in contrast, are better indicators of relationship at this hierarchical level. The discovery of the dalbergioid clade prompted a re-analysis of root nodule structure and the subsequent finding that the aeschynomenoid root nodule is synapomorphic for the dalbergioids.
Key Words: aeschynomenoid nodule • dalbergioid legumes • Fabaceae • papilionoid legumes • root nodule
|
INTRODUCTION
|
|---|
The "dalbergioid" legumes are a previously unrecognized monophyletic group of papilionoid legumes in spite of the extensive taxonomic history of its four constituents: tribes Adesmieae, Aeschynomeneae, Dalbergieae, and Desmodieae subtribe Bryinae. The formal recognition of this group represents a major rearrangement of papilionoid legumes. It combines elements conventionally considered disparate and characterized as either "primitive" or having varying levels of "advancement" (Fig. 1).
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 1. Putative relationships among tribes of the subfamily Papilionoideae according to Polhill (1981a)
. Tribes underscored include genera that are now known to be members of the dalbergioid clade (e.g., Desmodieae then included subtribe Bryinae, and Robinieae the genus Diphysa). Accumulation of nonprotein amino acids and fusion of floral parts occur frequently in Tephrosieae and all tribes positioned above it. The absence of such traits is traditionally viewed as primitive and is most frequent in tribes positioned below Tephrosieae
|
|
The Dalbergieae originally included tropical trees with fused floral parts and indehiscent pods (Bentham, 1860
). Three subtribes were recognized: Pterocarpeae with samaroid pods, Lonchocarpeae marked by pods having at most small marginal wings, and Geoffroyeae having drupaceous fruits. Polhill (1971, 1981d, 1994)
revised this classification by combining morphological evidence with that of seed chemistry and wood anatomy. This new Dalbergieae included 19 tropical woody genera mainly from Bentham's Pterocarpeae and Geoffroyeae. Lonchocarpinae were relegated to a closer relationship with other legumes that accumulated nonprotein amino acids in seed (e.g., Evans, Fellows, and Bell, 1985
). The revised Dalbergieae were diagnosed by supposedly plesiomorphic flower morphologies (i.e., free keel petals, staminal filaments partly fused and without basal fenestrae), pods with specialized seed chambers, and seeds that accumulated alkaloids or other than nonprotein amino acids. Geesink (1981, 1984)
accepted Polhill's circumscription with slight modification, whereas Sousa and de Sousa (1981)
proposed a classification similar to Bentham's because Dalbergieae (sensu Polhill, 1981d
) supposedly shared a determinate inflorescence with the Lonchocarpinae.
The Aeschynomeneae (Rudd, 1981a
) are one of five tribes traditionally characterized by lomented pods (Polhill, 1981a
). Although some Aeschynomeneae lack such pods (e.g., Arachis, Ormocarpopsis, Diphysa spp., Ormocarpum spp., Pictetia spp.), none of the members of this tribe have ever been confused or classified with the genera of Dalbergieae. Adesmieae (Polhill, 1981f
) have a notable history independent of the other dalbergioid legumes. This is because this tribe combines a presumed plesiomorphic trait, free staminal filaments, with a supposedly very derived one, lomented pods. This combination has suggested either a taxonomically isolated position or a relationship with other papilionoids also with free stamens (e.g., Burkart, 1952
). Bryinae, with lomented pods, possess other traits confirming its placement in the tribe Desmodieae (e.g., explosive secondary pollen presentation; Ohashi, Polhill, and Schubert, 1981
). However, Bryinae have seeds that do not accumulate nonprotein amino acids and lack a structural mutation in the chloroplast rpl2 locus (Bailey et al., 1997
). Both are atypical of the rest of Desmodieae.
In spite of a taxonomic history of Dalbergieae that has been separate from those of Aeschynomeneae, Adesmieae, and Bryinae, we present evidence that they collectively form a monophyletic group. The focus on these putatively disparate taxa was motivated by the taxonomic distribution of the distinctive aeschynomenoid root nodule (Corby, 1981
; Faria et al., 1994
) and four cladistic analyses: three involving nonmolecular data (Lavin, 1987
; Chappill, 1995
; Beyra-M. and Lavin, 1999
), and one with rbcL sequence data (Doyle et al., 1997
). We have expanded on these previous analyses by sampling exhaustively to reveal the exact constituents of the dalbergioid clade and enumerate the nonmolecular characters that have been used in the conventional tribal classification of these legumes. As such, we demonstrate where molecular and nonmolecular data are taxonomically concordant. We also show that many traditionally important taxonomic characters in this group are more homoplasious than previously considered. Because taxon sampling has focused on just the putative members of the dalbergioid clade, a point to be briefly addressed here but more thoroughly developed elsewhere is the higher level relationships of this newly recognized clade (Hu et al., 2000
; Pennington et al., in press
; M. Wojciechowski et al., unpublished data).
|
MATERIALS AND METHODS
|
|---|
DNA sequence data
DNA isolations, polymerase chain reaction (PCR) amplifications, and template purifications were performed with Qiagen Kits (i.e., DNeasy Plant Mini Kit, Taq PCR Core Kit, QIAquick PCR Purification Kit; Qiagen, Santa Clarita, California, USA). DNA sequences analyzed were the nuclear ribosomal 5.8S and flanking internal transcribed spacers (ITS1 and ITS2), the chloroplast trnK intron, including matK, and the trnL intron. PCR and sequencing primers for ITS and 5.8S sequences are described in Beyra-M. and Lavin (1999)
and Delgado-Salinas et al. (1999)
. Primers for matK and flanking trnK intron sequences are described in Lavin et al. (2000)
. Primers for the trnL intron are described by Taberlet et al. (1991)
. DNA sequencing was performed on an automated sequencer at the Iowa State University DNA Sequencing Facility (Ames, Iowa, USA) and Davis Sequencing (Davis, California, USA).
DNA sequences were aligned manually with Se-Al (Rambaut, 1996
). Bias introduced by the manual alignment was evaluated with a sensitivity analysis (cf. Whiting et al., 1997
; Beyra-M. and Lavin, 1999
; Delgado-Salinas et al., 1999
). Alignment-variable regions were variably aligned or excluded, a step matrix (cf. Cunningham, 1997
) was invoked or not, and gaps were treated as missing data, a fifth state, or as separate characters. Each of the different sensitivity analyses were subjected to the same heuristic search options. Missing data included 12.9% of the matK/trnK data set, 5.4% of the trnL data set, 1.5% of the ITS/5.8S data set, and 7.6% of the nonmolecular data set.
Maximum parsimony analyses were performed with PAUP* (Swofford, 2000
). Heuristic search options included 100 random-addition replicates, tree- bisection-reconnection branch swapping, and steepest descent. A maximum of 10 000 trees was allowed to accumulate, which is sufficient to capture all topological variation (cf. Sanderson and Doyle, 1993
). Clade stability tests involved bootstrap resampling (Felsenstein, 1985
; Sanderson, 1995
), where each of the 10 000 bootstrap replicates was subjected to heuristic search options that included one random-addition sequence per replicate, swapping with tree-bisection-reconnection, and invoking neither steepest descent nor mulpars.
Taxon sampling
Sampling of molecular and nonmolecular data was as exhaustive as possible at the generic level in order to determine membership in the dalbergioid clade, as well as the principal phylogenetic structure within this clade. Molecular and nonmolecular data were obtained for at least one species from every genus ever placed in the Dalbergieae (Burkart, 1952
; Polhill, 1981d
), Aeschynomeneae (Rudd, 1981a
), Adesmieae (Polhill, 1981f
), or Bryinae (Ohashi, Polhill, and Schubert, 1981
). The only exception is the presumably extinct genus Peltiera (Labat and Du Puy, 1997
), where no successful PCR amplifications were obtained from the few available DNAs. In addition to the advantages of being able to detail the taxonomic implications, exhaustive sampling for molecular data increases the probability of subdividing long branches (e.g., Hillis, 1998
).
Our original intent was to sample the same DNA accessions for each of the data sets. This proved impossible for DNA sequences because of inconsistencies in DNA quality and quantity and PCR amplification. We consequently had to resort to multiple methods of sampling. The DNA sequence data were sampled using the exemplar approach. Multiple species per terminal taxon were sampled where possible (Appendix A). Because nonmolecular data are generally open to visual inspection across all species of a particular terminal taxon, the "democratic" method of sampling (Bininda-Emonds, Bryant, and Russell, 1998
) was used for nonmolecular data. In this approach, we included all possible character states represented by any one terminal, which was usually a traditionally recognized genus (i.e., multistate terminal taxa were coded). The reasoning is that in the evaluation of traditionally important taxonomic characters, the degree of polymorphisms within terminals should be explicitly enumerated. For those few terminals in which species-level phylogenetic analysis has been completed (e.g., Andira and Pictetia), we employed the ancestral method of sampling nonmolecular data (Bininda-Emonds, Bryant, and Russell, 1998
). The justification for ultimately combining data that have been sampled differently is that a combined analysis should still allow us to best estimate where the traditionally important taxonomic characters lie on the continuum from strongly phylogenetically constrained to maximally homoplasious.
The genera Bergeronia, Dalbergiella, Lonchocarpus, and Muellera have been placed in the tribe Dalbergieae (e.g., Burkart, 1952
; Geesink, 1981
) and Pongamiopsis has been synonymized with the genus Aeschynomene (Hutchinson, 1964
). However, they were not included in this analysis because other phylogenetic analyses (Lavin et al., 1998
; Hu et al., 2000
) have shown these genera to be closely related to Millettia and relatives, all of which accumulate nonprotein amino acids in seed. Similarly, Poecilanthe and Cyclolobium should be allied with more basal Papilionoideae that accumulate alkaloids in seed (Greinwald et al., 1995
; Lavin et al., 1998
; Hu et al., 2000
). This is the reason that Poecilanthe is retained as a designated outgroup.
Outgroups were sampled extensively as part of large-scale molecular phylogenetic studies of the subfamily Papilionoideae (Hu et al., 2000
; Pennington et al., in press
; M. Wojciechowski et al., unpublished data). Sampling outgroups was guided by phylogenetic studies involving nonmolecular data (e.g., Chappill, 1995
; Herendeen, 1995
; Beyra-M. and Lavin, 1999
). For example, all outgroups chosen have leaves with punctate glands, a trait common to dalbergioids. In the end, the outgroups retained in this analysis included Acosmium and Myrospermum (tribe Sophoreae; Polhill, 1981b
), Dipteryx and Pterodon (Dipterygeae; Polhill, 1981c
), Poecilanthe (variously classified; see Lavin and Sousa, 1995
), and Apoplanesia, Amorpha, Eysenhardtia, and Marina (tribe Amorpheae; Barneby, 1977
; Polhill, 1981e
). This sampling was considered sufficient to demonstrate membership in the dalbergioid clade. The findings reported here did not change with a more extensive sampling of outgroups.
Sampling for the molecular data was re-evaluated as aligned DNA sequences accumulated. It became obvious that the matK/trnK sequences were by far the most informative at higher taxonomic levels, as seen in increased resolution in the strict consensus and higher bootstrap values. The primary effort then changed to sample as exhaustively as possible matK/trnK sequences and, secondarily, the ITS/5.8S and trnL intron sequences. Thus, the data analysis of this study centers on the matK/trnK data set. Sampling of ITS/5.8S sequences was guided by species level analyses of certain dalbergioid genera (e.g., Beyra-M. and Lavin, 1999
; Lavin et al., 2000
). Sampling of the trnL intron data was guided by a phylogenetic analysis of putatively basal Papilionoideae (Pennington et al., in press
). Unevenness in sampling was exacerbated by inconsistencies in PCR amplifications (mentioned above). A combined molecular analysis was not attempted because unevenness in sampling would result in a combined data set not exhaustively sampled at the genus level. Thus, consensus among the data sets was evaluated by congruence of the major clades resolved with high bootstrap values (cf. Huelsenbeck, Bull, and Cunningham, 1996
).
Nonmolecular character analysis
A nonmolecular data set was developed from that in Beyra-M. and Lavin (1999)
and is presented in Appendix B. Characters that have been considered traditionally important in the taxonomy of Dalbergieae, Aeschynomeneae, Adesmieae, and Bryinae (e.g., Burkart, 1952
; Ohashi, Polhill, and Schubert, 1981
; Polhill, 1981d
; Rudd, 1981a
; Sousa and de Sousa, 1981
) were targeted for analysis. As discussed above, multistate taxa were coded as polymorphic (cf. Weins, 1995
; Weins and Servedio, 1997
), in spite of the recommendation of Nixon and Davis (1991)
. Although this can underestimate the degree of homoplasy (see individual character discussions in Appendix B), splitting polymorphic terminals into two or more monomorphic ones does not change our findings (e.g., as evaluated in the fashion of a sensitivity analysis). This is because the focus is strictly at wide-scale relationships of groups of genera, and the potentially problematic polymorphisms are at a different level, within genera. Polymorphisms are discussed in the presentation of characters or ingroup terminal taxa (Appendices B and C). Inapplicable character states in certain terminals (e.g., leaf traits of Ramorinoa, a genus that doesn't produce leaves) were variously treated as a missing state, an uncertain state, or an extra state (as in a sensitivity analysis). The nonmolecular data were gathered primarily from field observations or herbarium specimens. Literature reports were usually verified by observations of the plants.
|
RESULTS
|
|---|
Parsimony analysis of the 1266 informative sites from the 95 taxa by 2966 sites matK/trnK data set produced 10 000 trees (the set maximum) each with a minimal length of 4352, a consistency index of 0.570 and a retention index of 0.830. The monophyly of the dalbergioid clade, including all genera of Aeschynomeneae, Adesmieae, Bryinae, and most Dalbergieae, was very well supported by bootstrap analysis (Fig. 2). Four members of tribe Dalbergieae (Andira, Hymenolobium, Vatairea, and Vataireopsis) and two sampled genera of Dipterygeae (Dipteryx and Pterodon) were not included. Indeed, the sister group to the dalbergioid clade includes genera sampled from the tribe Amorpheae (Apoplanesia and Amorpha). Within the dalbergioid clade, there are three well-supported subclades marked as the Adesmia, Dalbergia, and Pterocarpus clades (Fig. 2). The earliest branching Adesmia clade includes the genus Adesmia (sole member of the tribe Adesmieae) and mostly herbaceous to subshrubby genera of the tribe Aeschynomeneae (Poiretia, Amicia, Zornia, Chaetocalyx, and Nissolia). The remaining two subclades each include members of the Aeschynomeneae and Dalbergieae. The Pterocarpus clade additionally includes two genera, Brya and Cranocarpus, of Desmodieae (subtribe Bryinae).
View larger version (53K):
[in this window]
[in a new window]
|
Fig. 2. Bootstrap majority rule (50%) consensus from the analysis of matK/trnK sequences. The dalbergioid clade and its three constituent subclades are indicated
|
|
For the 481 informative sites from the 118 taxa by 719 sites ITS/5.8S data set, 120 trees were generated each with a minimal length of 5009, a consistency index of 0.259, and a retention index of 0.714. The same higher level relationships described for the matK/trnK analysis were resolved in this analysis, though with less bootstrap support (Fig. 3). Although the Pterocarpus clade was resolved in the strict consensus of the parsimony analysis, it was resolved in less than 50% of the analyses of the bootstrap replicates. In no case (majority- rule bootstrap consensus or strict consensus of minimal length trees) was the sister-group relationship of the Amorpheae samples resolved.
View larger version (60K):
[in this window]
[in a new window]
|
Fig. 3. Bootstrap majority rule (50%) consensus from the analysis of ITS/5.8S sequences. The dalbergioid clade and two of its three constituent subclades are indicated. The clade marked by a closed circle was also detected in the analysis of matK/trnK and trnL intron sequences
|
|
Analysis of the 293 informative sites from the 93 taxa by 737 sites trnL intron data set generated 10 000 trees each with a minimal length of 1102, a consistency index of 0.603, and a retention index of 0.804. Although the dalbergioid clade is well resolved by bootstrap analysis, only the Adesmia clade is further resolved (Fig. 4). Not in any case was the Dalbergia or Pterocarpus clades resolved as monophyletic. Regardless, the relationships resolved by majority-rule bootstrap consensus did not conflict with those similarly resolved in either the matK/trnK and ITS/5.8S analyses.
View larger version (51K):
[in this window]
[in a new window]
|
Fig. 4. Bootstrap majority rule (50%) consensus from the analysis of trnL intron sequences. The dalbergioid clade and the Adesmia subclade are indicated. Clades marked by a closed circle were also detected in the analysis of matK/trnK and ITS/5.8S sequences
|
|
Analysis of the 55 nonmolecular characters (Appendix B) yielded poorly resolved and supported relationships, such that the majority-rule bootstrap consensus was largely unresolved above the genus level. Resolved intergeneric relationships include a clade with Aeschynomene, Cyclocarpa, Bryaspis, Geissaspis, Humularia, Kotschya, Smithia, and Soemeringia (60% bootstrap support), one with Chapmannia, Arachis, and Stylosanthes (65%), Brya and Cranocarpus (67%), Chaetocalyx and Nissolia (100%), Amicia, Poiretia, and Zornia (67%), and Ormocarpopsis and Peltiera (93%). Because Peltiera is not represented by DNA sequence data, this nonmolecular data provide the only evidence for its relationships (the relationships of Peltiera are a focus of another study; M. Thulin and M. Lavin, unpublished data). The only well-supported clade that was resolved during this analysis and that was not seen during the previous molecular analyses was one with Etaballia and Inocarpus (80%), apomorphically diagnosed as having nearly regular flowers (characters 22–23 in Appendix B).
Because of the poorly resolved relationships obtained from analysis of the nonmolecular data set, it was combined with the matK/trnK data set in order to explore the evolution of the traditionally important taxonomic characters. Integration with just the matK/trnK is justified by how well this data set can resolve relationships (discussed in MATERIALS AND METHODS) and because of noncompatibility of molecular data sets with respect to sampling. Parsimony analysis of the 1319 informative characters of the combined matK/trnK and nonmolecular data set (95 taxa by 3021 characters) produced 2340 trees with a minimal length of 4664, each with a consistency index of 0.551 and a retention index of 0.821. The resulting relationships are essentially those described previously for the analysis of just the matK/trnK data set (Fig. 5).
View larger version (53K):
[in this window]
[in a new window]
|
Fig. 5. Bootstrap majority rule (50%) consensus from the analysis of combined nonmolecular and matK/trnK sequence data. The dalbergioid clade and its three constituent subclades are indicated
|
|
Sensitivity analysis
Making different assumptions about the molecular data sets, deleting characters with many missing entries (e.g., nonmolecular characters 50–54), splitting polymorphic terminals into two or more monomorphic ones, or recoding inapplicable nonmolecular characters to uncertain states, missing data, or as an extra state, did not affect the results described above (Figs. 2–5). The monophyly of the dalbergioid legumes was consistently resolved, as generally was the monophyly of the three constituent subclades. There were no cases of clades with bootstrap values over 70% that conflicted among the molecular data sets. Also, clades with high bootstrap values (i.e., >90%) in individual analyses of the matK/trnK, ITS/5.8S, trnL intron, or combined nonmolecular and matK/trnK data sets were consistently resolved regardless of the assumptions made about any one of the particular data sets. This is exemplified by analysis of just the matK coding region (i.e., excluding the flanking noncoding portion of the trnK intron), where some accessions in the data matrix were missing either the 5' or 3' half of this locus (for a total of 12.1% missing entries). The strict consensus of the parsimony analysis of the matK locus was essentially identical to that of the analysis of the matK/trnK data set. Bootstrap analysis resulted in values that were sometimes lower than in the analysis of the entire matK/trnK data set: 80% for the Amorpheae + dalbergioid clade, 100% for the dalbergioid clade, 100% for the Adesmia clade, 94% for the Dalbergia clade, and 71% for the Pterocarpus clade.
|
DISCUSSION
|
|---|
As now circumscribed, the dalbergioids comprise 44 genera (Appendix C) and 
1100 species of trees, shrubs, and perennial to annual herbs. Included are economically important hardwoods (e.g., Dalbergia and Pterocarpus spp.), forage legumes (Stylosanthes spp.), and crops (e.g., Arachis spp.). Like most pantropical legume taxa, the dalbergioids are concentrated in the neotropics and subSaharan Africa. Although the position of the dalbergioid clade within the Fabaceae is not fully developed here, its sister group is the tribe Amorpheae, which contains eight New World genera confined mostly to warm temperate and tropical North America. What is generally certain of higher level relationships is that the dalbergioids are distantly related to papilionoids that accumulate nonprotein amino acids in seed. This most notably includes Lonchocarpus, Derris, Millettia, and Hologalegina (e.g., tribes Robinieae, Galegeae, etc.; Wojciechowski, Sanderson, and Hu, 1999
), which at times have been taxonomically confused with various elements now included in the dalbergioid clade.
Implications for traditional classifications
The classification of certain genera into tribes and subtribes of Papilionoideae (e.g., Rudd, 1981a
; Ohashi, Polhill, and Schubert, 1981
; Ohashi, 1999
; Polhill, 1981a, d
) needs to be greatly modified in light of the evidence presented here. The genera Brya and Cranocarpus (subtribe Bryinae of tribe Desmodieae) share many unusual synapomorphies, such as periporate pollen and glochidiate trichomes, that have served to obscure higher level relationships. The explosive pollen presentation mechanism that Brya shares in common with Desmodieae is shown to have evolved independently. So have the lomented pods that Brya and Cranocarpus share with Desmodieae.
Four of the five subtribes of Aeschynomeneae are either monotypic (e.g., Discolobiinae) or are polyphyletic. Aeschynomeneae subtribe Ormocarpinae includes three different elements: Diphysa, Ormocarpum, Ormocarpopsis (and Peltiera), and Pictetia form one lineage in the Dalbergia clade, Fiebrigiella is in the Pterocarpus clade, and Chaetocalyx and Nissolia are part of the Adesmia clade. The pod valves with distinctive parallel venation that previously allied all of these genera now are considered to have evolved on three separate occasions. Indeed, this derived pod trait is homologous among Fiebrigiella, Chapmannia, Arachis, and Stylosanthes.
Aeschynomeneae subtribe Poiretiinae includes two different elements. Amicia, Poiretia, and Zornia form a monophyletic group within the Adesmia clade, and Weberbauerella is phylogenetically isolated within the Dalbergia clade. The marked pustular glands of Weberbauerella are no longer considered homologous to those of Amicia, Poiretia, and Zornia. In the recent classification of Japanese legumes (Ohashi, 1999
), Poiretia and Zornia are classified as the sole members of the tribe Poiretieae, a taxonomy that finds no support in this analysis.
Aeschynomeneae subtribe Aeschynomeninae includes eight genera (Aeschynomene, Cyclocarpa, Soemmeringia, Kotschya, Smithia, Humularia, Bryaspis, and Geissaspis) that form a very well-supported monophyletic group. A nonmolecular character supporting this relationship is the medifixed stipule, although it is not universal in this clade and has evolved independently in Zornia. An extrapolation from our small sample, however, suggests that species of Aeschynomene having basifixed stipules (e.g., A. fascicularis and A. purpusii) are more closely related to Machaerium and Dalbergia than they are to the species of Aeschynomene with medifixed stipules. Thus, the subtribe Aeschynomeninae includes two disparate elements.
Only Aeschynomeneae subtribe Stylosanthinae, with Arachis, Stylosanthes, and Chapmannia (and the segregates Pachecoa and Arthrocarpum), has been long recognized as a distinct taxonomic group and is also revealed as monophyletic in this analysis. The well-known nonmolecular character supporting the monophyly of this clade is a sessile papilionoid flower with a long hypanthium. However, these three genera are very closely related to Fiebrigiella and Fissicalyx and together all of these genera are set apart from other members of the Pterocarpus clade by large genetic distances. Notably, nonmolecular characters do not support most of the relationships in this clade that are so well supported by independent molecular data. For example, there are no known nonmolecular data that support the monophyly of the genus Chapmannia (Thulin, 2000
) or the relationship of Fissicalyx and Fiebrigiella.
The tribe Dalbergieae also is not monophyletic. Excluded from the dalbergioid clade are Andira, with 30 species largely confined to the neotropics and with one species distributed in the neotropics and tropical Africa (Lima, 1990
; Pennington, 1996
; Pennington, Aymard, and Cuello, 1997
), Hymenolobium with 10–15 species in tropical South America and one species in Central America (Polhill, 1981d
; Lima, 1982a, 1990
), Vatairea with seven species from Mexico to Brazil (Lima, 1982b, 1990
), and Vataireopsis with three species in Brazil and the Guianas (Polhill, 1981d
; Lima, 1990
). The distinction of these four genera from others traditionally included in the tribe Dalbergieae has been noted with wood anatomy (Baretta-Kuipers, 1981
) and estimates of overall similarity (Lima, 1990
). For example, the wood of Andira, Hymenolobium, Vatairea, and Vataireopsis lacks the storied structure and uniseriate rays that are characteristic of dalbergioid wood and is generally of less commercial value.
The remaining genera of the tribe Dalbergieae belong to either the Dalbergia or Pterocarpus clades. Only Dalbergia and Machaerium are part of the Dalbergia clade, where they are most closely related to Aeschynomene species that have basifixed stipules. The rest of the genera previously classified in the tribe Dalbergieae form the bulk of the Pterocarpus clade along with some genera previously classified in the tribe Aeschynomeneae (e.g., Fiebrigiella, Chapmannia, Arachis, Stylosanthes, and Discolobium) and subtribe Bryinae of Desmodieae.
The genera of Dipterygeae (Taralea, Dipteryx, and Pterodon; Polhill, 1981c
) are not part of the dalbergioid clade. Burkart (1952)
originally included Dipteryx (then Coumarouna) in the tribe Dalbergieae, and a phylogenetic analysis of nonmolecular data by Beyra-M. and Lavin (1999)
suggested Dipterygeae was part of the dalbergioid clade. Even the combination of paripinnate leaves bearing glandular punctae is known only from Dipterygeae and the dalbergioid legumes. However, this analysis strongly suggests that the punctate glands are plesiomorphic because they are found in all genera included in this analysis. Paripinnate leaves evolved independently among Dipterygeae and various elements in the dalbergioid clade.
Phylogenetic information among the various nonmolecular characters
While the matK/trnK phylogeny was not greatly influenced by the addition of the 55 nonmolecular characters (compare Figs. 2 and 5), there is some phylogenetic information in the nonmolecular characters, as evinced by high retention indices (Table 1). The consistency (CI) and retention (RI) indices for each of the 55 nonmolecular characters (Appendix B) in the combined analysis were compared to the same values obtained when each of the nonmolecular characters was mapped onto the matK/trnK phylogeny. In the combined analysis, the average CI and RI were 0.427 and 0.672, respectively. When mapped onto the matK/trnK trees, the average CI and RI were 0.390 and 0.627, respectively. Regardless of the small but significant differences (for RI, two-tailed t test, t = 2.94, P = 0.005, df = 52), no character had a higher consistency or retention index when mapped onto the matK/trnK phylogeny as when combined with the matK/trnK sequence data during parsimony analysis. This suggests that mapping a few selected nonmolecular characters onto a molecular phylogeny may involve a bias of excess levels of homoplasy.
View this table:
[in this window]
[in a new window]
|
Table 1. Average lengths (L) and consistency (CI) and retention (RI) indices for each of the 55 nonmolecular characters. These are compared for the combined analysis and when each of the 55 is mapped onto the matK/trnK phylogeny. An "=" indicates that the CI and RI of the combined and mapped character are equal. A ">" signifies a higher CI and RI value for a character in the combined analysis compared to when mapped. The reverse situation did not occur
|
|
Different classes of characters (e.g., vegetative, floral, and fruiting) were equally as prone to having homoplasy overestimated when mapped onto a molecular phylogeny. These include, for example, an asymmetric leaflet base (character 9 in Appendix B), persistent floral bracts (character 16), and a long pod stipe (character 33). The states of the leaflet base had an average retention index of 1.000 in the combined analysis and 0.500 when mapped to the matK/trnK trees (Table 1). The corresponding values were 0.667 and 0.167 for the states of the floral bracts, and 0.647 and 0.559 for the pod stipe (Table 1). Also, no particular class of characters (e.g., vegetative, floral, and fruiting) was more informative than another. For vegetative characters (1–13, 44–45, 50–55), the average retention index is 0.705. For floral characters (14–30, 46–49), it is 0.604. For fruiting characters (31–43), the average retention index is 0.726. These differences are not significant (single- factor ANOVA, F = 1.174, P = 0.317, df = 52). The lack of a difference in behavior among the various classes of characters, as also generally found by Bateman and Simpson (1998)
for vascular plants, weakens the suggestion of Tucker and Douglas (1994)
that floral characters necessarily provide the best taxonomic information in Leguminosae. These findings also weaken the implication that pod morphology is prone to higher rates of convergent evolution than other types of characters (e.g., Geesink, 1984
; Hu et al., 2000
).
Conventional taxonomic evidence
Some traditionally important taxonomic characters are determined in this analysis to be more homoplasious than previously considered. This is especially true of the character states of growth habit, staminal fusion, and pod segmentation. Herbaceous and woody relatives generally are separated into different taxonomic groups when a temperate vs. tropical distinction correlates with habit (Judd, Sanders, and Donoghue, 1994
). This is especially true of papilionoid legumes where tribes have been categorized by habit (e.g., temperate herbaceous vs. tropical woody tribal division in Polhill, 1981a, 1994
). An herbaceous habit (number 1 in Appendix B) has evolved at least three times in monomorphic condition but more times than this in polymorphic condition (Table 1). The Adesmia clade contains mostly herbaceous species, although some species of Adesmia and Poiretia are shrubs. That an herbaceous growth form maps as the ancestral state in the Adesmia clade stands in contrast to the conventional wisdom that woody taxa form basal clades in tropical Papilionoideae (e.g., Polhill, 1981a
; Tucker and Douglas, 1994
).
Genera containing both woody and herbaceous species also occur in the clade containing Aeschynomene sect. Aeschynomene, Kotschya, Humularia, and Geissaspis. The same is true for the clade including Fiebrigiella, Chapmannia, Stylosanthes, and Arachis. Fissicalyx and some species of Chapmannia are woody in a clade dominated by herbaceous to subshrubby species. Representing yet two other clades, species of Machaerium, Dalbergia, Brya, and Cranocarpus vary from trees or shrubs to weak subshrubs. Clearly, there is no evidence from this analysis that the ability to produce a strongly woody growth habit is a good indicator of relationship.
The staminal character number 26 (Appendix B) includes five states that provide an average length of 15.0 to the most parsimonious trees. The consistency index of 0.267 and the retention index of 0.633 demonstrate that this character is homoplasious. Even state zero, free staminal filaments, added a length of two because this state occurs ancestrally in some of the outgroup genera and represents a reversion in the genus Adesmia. That a legume group with free stamens can evolve this condition secondarily from a fused condition (e.g., 9 + 1 diadelphous) is not surprising. Four species of Pictetia have nearly free staminal filaments in a clade otherwise represented by species with fused filaments (Beyra-M. and Lavin, 1999
). Also, Käss and Wink (1995, 1997)
have implicitly shown in an unrelated papilionoid group that the evolution of staminal morphology does not necessarily involve a unique transformation from free filaments into the fused condition. Perhaps related to this issue, Klitgaard (1999a)
showed that order of initiation and loss of stamens are more variable among the dalbergioids than previously appreciated. No doubt, the a priori view that free staminal filaments represent necessarily a plesiomorphic condition among papilionoid legumes will have to be abandoned.
All papilionoid legumes with lomented pods were at one time classified together, although more recently five tribes (Adesmieae, Aeschynomeneae, Coronilleae, Desmodieae, and Hedysareae) were thought to have gained this pod type independently (Polhill, 1981a
). We scored three states pertaining to articulation of pod segments (number 31 in Appendix B), which added an average length of 9.0 to the most parsimonious trees. The consistency index of 0.250 and a retention index of 0.825 suggest that, although homoplasious, this character provided phylogenetic resolution towards the tips of the tree. The Adesmia clade is uniform for lomented pods, but the Dalbergia and Pterocarpus clades are variable, with a minimum of three separate origins of this pod type in each of these clades. What was thought to be two separate origins of lomented pods in Adesmieae and Aeschynomeneae is now considered at least six origins combined with at least two reversals, and not counting polymorphisms.
New taxonomic evidence
In contrast to the above, a few previously overlooked characters are shown by analysis of combined molecular and nonmolecular data to be taxonomically informative. Short shoots (character 2 in Appendix B) evolved only once in the clade containing Pictetia, Ormocarpum, and Ormocarpopsis (also Peltiera). However, the support for this clade is moderate (Fig. 5), both in this analysis, and in those of Beyra-M. and Lavin (1999)
and Lavin et al. (2000)
. Bilabiate calyx lobes (state 2 of character 19 in Appendix B) mark the monophyly of the clade containing Aeschynomene sect. Aeschynomene, Smithia, Kotschya, Humularia, Cyclocarpa, Soemmeringia, Bryaspis, and Geissaspis. In contrast to short shoots, this calyx morphology marks a very well-supported clade (Fig. 5). The other nonmolecular characters with a high retention index (Table 1), however, either mark small clades (e.g., characters 13 and 46 and the clade with Brya and Cranocarpus), or have homoplasy that was underestimated because of scoring polymorphic taxa (e.g., see characters 39 and 40 in Appendix B).
The aeschynomenoid root nodule (Fig. 6, character 55 in Appendix B) is the most notable nonmolecular character in that it is inferred to be a synapomorphy for the dalbergioid clade. The idea that nodule morphology could be a useful character in legume taxonomy was pioneered by Corby (1981)
. He described a number of shapes, named according to the genus from which he had most observations. The aeschynomenoid type has as its main feature a small oblate nodule (transverse diameter greater than axial) with determinate growth. Corby noted that aeschynomenoid nodules are often associated with fine rootlets, but his otherwise excellent drawings omitted these "for clarity." Such nodules were found primarily in the tribes Adesmieae, and Aeschynomeneae, but also in some members of the Abreae, Dalbergieae, Phaseoleae, and Robinieae (Corby, 1988
). On his retirement, Corby kindly gave the Sprent laboratory his collection of preserved nodules. These were used, together with new material, for more detailed structural studies. As a result, the definition of an aeschynomenoid nodule has been adapted to include additional features. In particular, this nodule is always associated with a lateral or (in the case of stem nodules) adventitious root. The central infected tissue contains few or no uninfected cells. Differentiated infection threads are not involved in the process of infection, which (where studied in detail) takes place at the lateral root junction (Sprent, Sutherland, and Faria, 1989
). All nodules of the tribe Aeschynomeneae that have been examined conform to this description, together with ten genera of the Dalbergieae: Centrolobium, Dalbergia, Etaballia, Geoffroea, Machaerium, Platymiscium, Platypodium, Riedeliella, Tipuana, and Pterocarpus (two Brazilian species, P. rohrii and P. santalinoides are not known to nodulate). The evidence for Adesmia, Brya, and Cranocarpus, although slightly less detailed, is entirely consistent with the revised description of aeschynomenoid nodules.
View larger version (68K):
[in this window]
[in a new window]
|
Figs. 6–8. Selected nonmolecular characters (scale bar = 1 cm for all figures). 6. Aeschynomenoid root nodule associated with lateral root (character number 55, Appendix B). 7. Short shoots of Ormocarpum (character number 2). 8. Pseudopetiole of Arachis (character number 4)
|
|
Members of the Dalbergieae that have been omitted from the revised clade on morphological and molecular grounds would also be omitted on grounds of nodule structure (Andira and Hymenolobium) or absence of nodules (Vatairea and Vataireopsis; Sprent, Sutherland, and Faria, 1989
). Two genera of the dalbergioid clade that do not nodulate are Chaetocalyx and Nissolia (Faria and Lima, 1998
). Both of these are lianes. Notably, a group of species in Acacia with a semiscandent habitat cannot nodulate (Harrier et al., 1997
). These acacias have retained some of the characters associated with nodulation, such as some of the nod genes, and the ability to stimulate rhizobial attachment to roots. It was thus suggested that they may have lost the ability to nodulate because, living on the forest margins, they were not nitrogen limited (Harrier, 1995
). It would be interesting to carry out similar tests on Chaetocalyx and Nissolia as one of their principal habitats is forest margins.
It is now generally agreed that nodulation in legumes may have evolved more than once (Sprent, 1994
; Soltis et al., 1995
). One of these nodulation events involved an infection process through a wound, such as where a lateral or an adventitious root emerges. Compared with the more familiar root hair infection pathway (see Sprent and Sprent, 1990
for details), this pathway is simpler, involving less complex recognition systems. Apart from some species of the mimosoid genus Neptunia (James et al., 1992
), this wound infection pathway is associated with only aeschynomenoid nodules. In Neptunia, however, nodule processes subsequent to infection involve production of infection threads and development of an indeterminate nodule.
Our phylogenetic results are in agreement with molecular and biochemical evidence that nodule structure and infection site are largely plant determined (e.g., Gualtieri and Bisseling, 2000
). Given a phylogenetic lineage, nodule morphology and infection processes are generally the same regardless of which species or genus of rhizobia is involved (six genera of bacteria nodulating legumes are now recognized, and they are collectively known as rhizobia). Another general inference is derived from the observation that all species of the genus Aeschynomene that have stem nodules are nodulated by photosynthetic rhizobia (Molouba et al., 1999
). Given that the aeschynomenoid root nodule has an unelaborated morphology and infection mode, the ancestral rhizobial form could have been photosynthetic. As legumes moved into drier areas, nodules developed on roots and lost photosynthetic ability (Sprent, 1994
).
A phylogenetic classification
The dalbergioid legumes are similar to a group of Papilionoideae that includes also Amorpheae and Dipterygieae. They share a distinctive combination of a base chromosome number of x = 10 (Goldblatt, 1981
), wood with uniseriate stored rays, vegetative growth with glandular punctae, flowers with fused keel petals or staminal filaments, and seeds that do not accumulate nonprotein amino acids (derived from Beyra-M. and Lavin, 1999
). The dalbergioids differ and are apomorphically defined (sensu de Queiroz and Gauthier, 1994
) as having glandular-based trichomes on vegetative or floral organs, a well-developed abaxial calyx lobe, and the "aeschynomenoid" root nodule. All of these traits have been secondarily transformed in some constituents of the dalbergioid clade (see characters 11, 19, and 55 in Appendix B; also Table 1).
The dalbergioid clade is distinguished more by molecular than nonmolecular data. It is another legume example of a cryptic clade, like "Neo-Astragalus" (Wojciechowski et al., 1993
) and the "temperate herbaceous clade" (Sanderson and Wojciechowski, 1996
). Regardless, it is informally recognized here as a distinctive taxonomic group. Furthermore, the three major constituent subclades are informally recognized and Appendix C enumerates the 44 current dalbergioid genera accordingly. The three subclades of dalbergioids are:
The Adesmia clade
This includes the genera Adesmia (of tribe Adesmieae; Polhill, 1981f
) and Poiretia, Amicia, Zornia, Chaetocalyx, and Nissolia of the tribe Aeschynomeneae. This clade is apomorphically defined as having an herbaceous growth habit (modified in some descendants—character 1), leaves with few opposite leaflets (evolved in parallel in Arachis and close relatives—character 8), and pedicels confluent with the calyx (modified only in a few species of Nissolia— character 17). A node-based definition (sensu de Queiroz and Gauthier, 1994
) includes all descendants from the common ancestor of Adesmia and Amicia.
The Dalbergia clade
This includes Dalbergia and Machaerium (of tribe Dalbergieae; de Candolle, 1825
; Polhill, 1981d
), and the following genera of Aeschynomeneae (sensu Rudd, 1981a
): Aeschynomene (all infrageneric taxa), Soemmeringia, Cyclocarpa, Kotschya, Smithia, Humularia, Bryaspis, Geissaspis, Weberbauerella, Diphysa, Pictetia, Ormocarpum, Ormocarpopsis, and Peltiera. This clade is apomorphically defined as having diadelphous staminal filaments splitting readily or tardily into two flanges, usually in a 5 + 5 arrangement (polymorphic with a 9 + 1 diadelphous condition in many species and occasionally monodelphous in Machaerium—character 26), and a persistent staminal flange that in some cases reflexes upward above the developing fruit (character 28). A node-based definition includes all descendants from the common ancestor of Dalbergia and Cyclocarpa.
The Pterocarpus clade
This includes Pterocarpus, Tipuana, Platypodium, Reideliella, Centrolobium, Grazielodendron, Paramachaerium, Ramorinoa, Inocarpus, Etaballia, Platymiscium, Cascaronia, Fissicalyx, Geoffroea from Dalbergieae; Brya and Cranocarpus from Desmodieae; and Fiebrigiella, Chapmannia, Stylosanthes, Arachis, and Discolobium from Aeschynomeneae. This clade is apomorphically defined as having commonly caducous bracteoles (character 18) and seedlings producing a simplified eophyll (secondarily transformed in Arachis and close relatives—character 45). A node-based definition includes all descendants from the common ancestor of Pterocarpus and Riedeliella.
Although data from matK/trnK, trnL, and ITS/5.8S were not combined in a single analysis, results from individual analyses showed significant consensus combined with no significant conflict. The combined matK/trnK and nonmolecular analysis yielded very robust results to support the conclusions outlined above. This study demonstrates that matK/trnK sequences provide excellent resolution at the broadest phylogenetic levels dealt with in this study. This same locus, along with ITS/5.8S, gives excellent resolution to within and among closely related genera. In contrast, trnL provides the least resolution. Ultimately, this study provides a framework for future studies that deal taxonomically with individual dalbergioid genera. There is now sufficient data from which to guide the choice of potential sister groups or outgroups in such studies.
View larger version (182K):
[in this window]
[in a new window]
|
Figs. 9–12. Selected nonmolecular characters. 9. Glandular-based trichome of dalbergioid legumes (character number 11; scale bar = 200 µm). 10. Base of trichome where glandular exudate is secreted (scale bar = 20 µm). 11. Pustular glands on leaflet of Centrolobium (character number 12; scale bar = 200 µm). 12. Glochidiate trichomes on leaf of Brya (character number 13; scale bar = 20 µm)
|
|
View larger version (21K):
[in this window]
[in a new window]
|
Figs. 13–16. Petal characters (scale bar = 5 mm for all figures). Figs. 13–15
. Petals differentiated into blade and claw in Geoffroea (character number 23). 13. Standard. 14. Wing. 15. Keel. 16. Petals not differentiated into a blade and claw in Inocarpus
|
|
View larger version (41K):
[in this window]
[in a new window]
|
Figs. 17–28. Representative species of the Adesmia clade (scale bar = 1 cm for all figures). Figs. 17–23
. Chaetocalyx brasiliensis. 17. Habit. 18. Calyx. 19. Gynoecium. 20. Androecium. 21. Keel petal. 22. Wing petal. 23. Standard. Figs. 24–27
. Nissolia wislizenii. 24. Habit. 25. Cauline leaf. 26. Flower. 27. Fruits. 28. Nissolia microptera, leafy stem with fruits. Reproduced from Volume 5 of Flora Novo-Galiciana by Rogers McVaugh
|
|
View larger version (60K):
[in this window]
[in a new window]
|
Figs. 29–47. Representative species of the Pterocarpus clade (scale bar = 1 cm for all figures). Figs. 29–36
. Platymiscium trifoliolatum. 29. Flowering branch. 30. Branch of fruiting inflorescences with wall of one fruit cut away to show seed. 31. Calyx. 32. Androecium. 33–34. Wing petals. 35. Keel petals. 36. Standard. Figs. 37–47
. Pterocarpus orbiculatus. 37. Detached leaf. 38. Inflorescence. 39. Mature fruits. 40. Immature fruits. 41. Calyx. 42. Androecium. 43. Gynoecium. 44. Keel petals. 45–46. Wing petals. 47. Standard. Reproduced from Volume 5 of Flora Novo-Galiciana by Rogers McVaugh
|
|
View larger version (61K):
[in this window]
[in a new window]
|
Figs. 48–69. Representative species of the Dalbergia clade (scale bar = 1 cm for all figures except where noted). Figs. 48–58
. Dalbergia congestiflora. 48. Leafy branch. 49. Flowering branch. 50. Fruits. 51. Seed. 52. Androecium. 53. Anther (scale bar = 1 mm). 54. Calyx. 55. Keel petals. 56–57. Wing petals. 58. Standard. Figs. 59–69. Machaerium kegelii. 59. Flowering branch. 60–61. Nodes with stipular spines. 62. Androecium. 63. Gynoecium. 64. Calyx. 65. Keel petals. 66. Wing petal. 67. Standard. 68. Flower. 69. Fruit. Reproduced from Volume 5 of Flora Novo-Galiciana by Rogers McVaugh
|
|
|
APPENDIX
|
|---|
Nonmolecular characters and character states. All references to clades are those derived from the combined matK/trnK and nonmolecular analysis (Fig. 5). References to ancestral states were inferred with the reconstruct tree option in PAUP (Swofford, 2000
) and the trace option in MacClade (Maddison and Maddison, 1999
).
Vegetative characters
1. Habit: 0) woody (trees to shrubs), 1) herbaceous (subshrubs to herbs), 2) twining and herbaceous, 3) twining and woody. Predominantly herbaceous genera sometimes include subshrubby species, whereas woody genera usually do not, thus explaining the coding for state number 1. A herbaceous habit arose independently in the following clades: one represented by Fiebrigiella and Arachis, another by Chaetocalyx and Poiretia, and one by Weberbauerella and Kotschya. The twining herbaceous habit is restricted to the Adesmia clade where it is known from some species of Poiretia (Rudd, 1972c
) and all Chaetocalyx (Rudd, 1958
) and Nissolia (Rudd, 1956
). A twining woody habit occurs in polymorphic condition in the clade with Dalbergia and Machaerium.
2. Short shoots: 0) absent or not regularly present and then not covered by persistent stipules, 1) regularly present and covered by distichously arranged persistent stipules from the axils of which are born the inflorescence (Fig. 7). The short shoot condition is restricted to the clade including all descendants of the most recent common ancestor of Pictetia and Ormocarpopsis. Very similar short shoots were described for Poitea (tribe Robinieae; Lavin, 1993
), which is also from the Greater Antilles.
3. Stipule modifications: 0) attached to stem at base (basifixed) and foliaceous, 1) attached to stem in the middle and foliaceous (peltate or medifixed), 2) basifixed and lignescent. Medifixed stipules are referred to as appendiculate (e.g., Rudd, 1981a
) and are evolved independently in a clade including Aeschynomene sect. Aeschynomene, Cyclocarpa, Humularia, Geissaspis, Smithia, and another including just Zornia. Lignescent stipules evolved independently in polymorphic condition in the liana-forming species of Machaerium, in most species of Brya, and in all species of Pictetia. In Brya, the leaves of the long shoot are entirely transformed into a single spine.
4. Pseudopetiole: 0) absent, 1) present (Fig. 8). A pseudopetiole is traditionally defined as a petiole with stipules attached. It is here described as a pulvinus (leaf base) that is projected away from the main axis of the stem. The stipules are attached to this projected portion of the stem, and they superficially appear as if they are adnate to the petiole. The pseudopetiole evolved independently in a clade including just Adesmia, and another including Arachis and Stylosanthes.
5. Leaf rachis in cross section: 0) terete, 1) with a single continuous groove (canaliculate). A terete leaf rachis is recorded from Discolobium, Dalbergia, Machaerium, and Ormocarpopsis, Peltiera, Platymiscium, Centrolobium, Grazielodendron, Etaballia, Fissicalyx, Peltiera, and Pterocarpus, and in polymorphic condition from Ormocarpum, Aeschynomene (all subgroups) and closely related genera (Cyclocarpa, etc.). Grooved leaf rachises occur in the rest of the genera, except where the leaves are uniformly sessile, as in Brya and Inocarpus, and this trait is then scored as inapplicable. Otherwise, leaf rachises vary continuously between narrowly grooved and distinctly canaliculate. The motivation for using this trait is that terete leaf rachises are shown to be derived (but in polymorphic condition) in two clades: that including all descendents but Pictetia of the most recent common ancestor of Dalbergia and Ormocarpopsis, and that including most descendants of the recent common ancestor of Platymiscium and Pterocarpus.
6. Distal end of leaf rachis: 0) terminated by a leaflet, 1) not terminated by a leaflet (a mucro is often present). A leaf rachis not terminated by a leaflet is found in the large clade including Aeschynomene sect. Aeschynomene, Cyclocarpa, Humularia, Soemmeringia, Kotschya, Smithia, Geissaspis, and Bryaspis. This type of leaf also has evolved independently in the outgroup samples of Dipterygeae (Dipteryx and Pterodon), the clade including Amicia, Zornia, Adesmia, Arachis, and Poiretia, the clade including just Aeschynomene sect. Ochopodium, and the clade including Stylosanthes and Arachis.
7. Number of leaflets per leaf: 0) leaves unifoliolate/simple, 1) leaves tri- to 20-foliolate, 2) leaves more than 20-foliolate. State zero occurs uniformly in Etaballia, Inocarpus, and Brya, and in polymorphic condition in Cranocarpus. State two is restricted to just the Dalbergia clade where it occurs uniformly in Weberbauerella, and predominantly so (i.e., polymorphic) in Machaerium, Dalbergia, and all the sections and series of Aeschynomene (Aeschynomene, Viscidulae, Pleuronerviae, and Scopariae). This state is capturing "fern-like" leaves where the leaflets abut laterally, are narrowly elliptic, and have parallel lateral margins. Simple leaves are scattered throughout but with most occurrences (usually in polymorphic condition) in the Pterocarpus clade (Discolobium, Etaballia, Inocarpus, Platypodium, Byra, and Cranocarpus).
8. Leaflet arrangement: 0) alternate, 1) opposite. Two large clades have evolved opposite leaflets independently. One includes Adesmia, Chaetocalyx, Nissolia, Poiretia, Amicia, Zornia, and the other includes Fissicalyx, Fiebrigiella, Chapmannia, Stylosanthes, and Arachis. Opposite leaflets have evolved sporadically mostly within the Pterocarpus clade (Grazielodendron, Riedeliella, Cranocarpus, Paramachaerium), and rarely in the Dalbergia clade (Smithia). The genera with uniformly simple or unifoliolate leaves (e.g., Etaballia, Inocarpus, Brya, and Ramorinoa) were marked inapplicable. The species of Cran
TOP
ABSTRACT
INTRODUCTION
