| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics |
2The School of Plant Sciences, The University of Reading, Whiteknights, Reading, RG6 6AS, UK; 3The Royal Botanic Gardens, Kew, Richmond, Surrey, TW 9 3AB, UK
Received for publication November 3, 2004. Accepted for publication April 22, 2005.
ABSTRACT
The monophyly of the Peltophorum group, one of nine informal groups recognized by Polhill in the Caesalpinieae, was tested using sequence data from the trnL-F, rbcL, and rps16 regions of the chloroplast genome. Exemplars were included from all 16 genera of the Peltophorum group, and from 15 genera representing seven of the other eight informal groups in the tribe. The data were analyzed separately and in combined analyses using parsimony and Bayesian methods. The analysis method had little effect on the topology of well-supported relationships. The molecular data recovered a generally well-supported phylogeny with many intergeneric relationships resolved. Results show that the Peltophorum group as currently delimited is polyphyletic, but that eight genera plus one undescribed genus form a core Peltophorum group, which is referred to here as the Peltophorum group sensu stricto. These genera are Bussea, Conzattia, Colvillea, Delonix, Heteroflorum (inedit.), Lemuropisum, Parkinsonia, Peltophorum, and Schizolobium. The remaining eight genera of the Peltophorum group s.l. are distributed across the Caesalpinieae. Morphological support for the redelimited Peltophorum group and the other recovered clades was assessed, and no unique synapomorphy was found for the Peltophorum group s.s. A proposal for the reclassification of the Peltophorum group s.l. is presented.
Key Words: Bayesian analysis • Caesalpinioideae • cladistic analysis • classification • Fabacaceae • molecular data • parsimony • phylogeny
The Leguminosae is among the most extensively studied of all angiosperm families and yet, despite extensive phylogenetic research, unexpected relationships are still being revealed. The tribe Caesalpinieae is one for which phylogenetic studies are discovering relationships not in accord with the most widely used classification of the legumes, that of Polhill (1994)
. The Peltophorum group is one of eight informal groups within the tribe Caesalpinieae described by Polhill and Vidal (1981)
and later redefined and augmented by Polhill (1994)
. Recent classifications of the genera (Hutchinson, 1964
; Heywood, 1971
; Pettigrew and Watson, 1977
; Polhill and Vidal, 1981
; Watson and Dallwitz, 1983
; Polhill, 1994
) have differed in their treatment of these 16 genera (Table 1) and generic groupings in the Caesalpinieae are still uncertain. Molecular analyses of the rbcL gene have suggested that the Peltophorum group as currently delimited is polyphyletic (Doyle et al., 2000
; Kajita et al., 2001
). This has been supported by data from the trnL intron (Bruneau et al., 2001
) and from the trnL-F region (Haston et al., 2003
; Simpson et al., 2003
). The Peltophorum group referred to here is that of Polhill (1994)
unless otherwise specified. The geographical distribution of the Peltophorum group was suggested by Polhill and Vidal (1981)
to be centered in Amazonia. Indeed, eight of the 16 genera are restricted to South America, and three of the other genera are also represented there. There are, however, other areas of diversity within the group. Two genera are endemic to Mexico, and four genera are found only in the palaeotropics, of which three are restricted to Africa and Madagascar and one extends into Asia. There are also two pantropical genera in the group, one of which occurs in South and Central America, Mexico, southern United States, and Africa, while the other occurs in South and Central America, Africa, southeast Asia, and northern Australia. The areas of diversity observed in the group will be discussed with reference to the phylogenetic hypotheses generated by our analyses.
|
; Kajita et al., 2001
; Haston et al., 2003
; Simpson et al., 2003
) are upheld when more data and more taxa are included in molecular phylogenetic analyses, (2) determine the relationships of the genera currently placed in the group, whether in one or more clades, (3) test the monophyly of the genera in the Peltophorum group when sufficient species have been studied, and (4) characterize the clades recovered, based on a morphological survey carried out to determine the utility of morphological characters for phylogeny reconstruction in the group (Haston, 2003
). MATERIALS AND METHODS
Taxon sampling
Accessions were selected to represent as many species as possible from the genera in the Peltophorum group (Appendix). Heteroflorum, a new genus first mentioned in Sousa and Delgado (1993)
but not yet validly published, is reportedly closely related to Conzattia (M. Sousa, Universidad Nacional Autónoma de México, unpublished data) and is therefore included within the ingroup. Complete sampling of all the species in the group for all gene regions has not been possible, due to a lack of material for some species and difficulties in sequencing. However, all genera have been represented in at least one analysis. Apart from Campsiandra, for genera containing more than a single species, at least two species have been included in at least one analysis. For the larger genera, including Delonix, Moldenhawera, Parkinsonia, and Peltophorum, at least three species were included. We were not able to obtain suitable leaf material of Campsiandra, and a single sequence for the trnL intron supplied by A. Bruneau and F. Forest (Université de Montréal) was therefore used.
To test the monophyly of the Peltophorum group as rigorously as possible, outgroup taxa were selected from seven of the other eight informal groups of the Caesalpinieae (sensu Polhill, 1994
). Where a group is not monogeneric, exemplars were included from at least two genera to help break any long branches (Smith, 1994
). Orphandodendron bernalii, the only member of the recently established Orphanodendron group (Polhill, 1994
), is not included due to the scarcity of collected material. Poeppigia procera, a monospecific genus from Central America, was selected to root the tree since higher level analyses have suggested that it is the most distantly related of the outgroup taxa (Bruneau et al., 2001
).
DNA sequence data
Silica-dried leaf material was available for 23 individuals: 14 collected from plants cultivated in the UK and nine collected in the wild. Three sequences were downloaded directly from GenBank (National Center for Biotechnology Information, website http://www.ncbi.nlm.nih.gov). Leaf material for all other species was collected from herbarium specimens (Appendix). DNA extracted from specimens in which the leaves retained some green coloring was found to amplify more successfully than DNA from browner leaves, irrespective of age. The leaves of one genus, Campsiandra, appeared to turn brown very quickly following collection and DNA was not successfully extracted from herbarium specimens of this genus.
DNA was extracted from leaf material using a "microprep" protocol (modified from Harris, 1995
). ß-Mercaptoethanol (2%) was included only for leaf material from herbarium specimens. When it was difficult to extract sufficient DNA the protocol was modified to include a longer precipitation with propan-2-ol for 2 weeks at –18°C. For tough, leathery leaves, autoclaved sand was used in grinding the leaves. Finally, DNA was successfully extracted from the leaves of Arapatiella species using a solution of 3% CTAB with the addition of approximately 1% polyvinyl polypyrrolidone (PVPP) and a 2-week precipitation.
Following extraction, DNA was amplified using the polymerase chain reaction (Saiki et al., 1988
). Previous analyses have found the trnL intron region and the trnL-F spacer region to be phylogenetically informative at a tribal level in the Leguminosae (Bruneau et al., 2000
, 2001
; Ireland et al., 2000
; Luckow et al., 2000
; Pennington et al., 2001
). The intron was sequenced here and, when the DNA was not too degraded or weak, the trnL-F spacer was also included. The rbcL gene has been used extensively for higher level phylogenetic analyses, including several legume studies (Kass and Wink, 1996
; Doyle et al., 1997
; Kajita et al., 2001
). A faster evolving region was desirable to resolve relationships within genera, and the rps16 region was selected, earlier studies having shown that the region may have a suitable rate of evolution (Oxelman et al., 1997
; Andersson and Andersson, 2000
; Lee and Hymowitz, 2001
).
The following primers were used: (1) trnL-F: primers c, d, e, and f (Taberlet et al., 1991
). Internal primers d and e were used only for degraded DNA. (2) rbcL: primers N, 674, 724/SS722, and R (Kass and Wink, 1996
; Sulaiman, 1997
). In taxa for which the forward primer N failed to produce a readable sequence a reverse primer 402R (5' ARTCGCAAATCCTCCAGACG 3') was designed from successful sequences and was used to sequence the first 402 bp. (3) rps16: primers F and R2 (Oxelman et al., 1997
).
PCR reactions of 50 µl contained 2 µl of undiluted DNA template with a final concentration of 1x reaction buffer, 2.5 mM MgCl2, 1 mM dNTPs, 1.5 µM of each primer, 2.5 units Taq DNA polymerase, made up to 50 µL with nanopure H20. A higher concentration of MgCl2 (3 mM) was used to amplify the rps16 region. The following PCR profiles and modifications were used: (1) trnL-F: 95°C for 2 min; 95°C for 50 s, 50°C for 50 s, 72°C for 110 s, for 35 cycles; 72°C for 7 min; (2) rbcL: 94°C for 2 min; 94°C for 50 s, 49°C for 50 s, 72°C for 110 s, for 35 cycles; 72°C for 7 min; (3) rps16: 95°C for 2 min; 95°C for 30 s, 58°C for 1 min, 72°C for 2 min, for 30 cycles; 72°C for 10 min. The annealing temperature was increased to 60°C for template DNA, which produced a weak band at 58°C and a low molecular mass smear, which interfered with the cycle sequencing reaction. A lower concentration of MgCl2 (2 mM) was found to produce cleaner bands with less smearing.
The PCR products were cleaned (Qiagen PCR Purification kit, Qiagen Ltd., Crawley, West Sussex, UK, or NucleoSpin Extract kit, Macherey-Nagel GmbH & Co., Düren, Germany). Cycle sequencing was carried out using a standard profile: 25 cycles of 97°C for 15 s, 50°C for 5 s, 60°C for 4 min. The products of the cycle-sequencing reaction were cleaned using Sephadex columns (Amersham Pharmacia Biotech, Piscataway, New Jersey, USA) and dried at 60°C in a rotary vacuum evaporator. The sequences were generated on the ABI 373 and the ABI prism 3100 capillary sequencer (Applied Biosystems, Foster City, California, USA).
Sequence alignment, indel coding, and data set summary
Sequences were assembled in Seqman (DNAStar Inc., Madison, Wisconsin, USA) and aligned using ClustalV as implemented by Megalign 4.05 (DNAStar Inc., Madison, WI, USA), prior to final editing by eye. The three alignments were combined in MacClade 4.0 (Sinauer, Sunderland, Massachusetts, USA) and indels were coded as additional binary characters and appended to the data set. The indels were coded using the "simple indel coding" of Simmons and Ochoterena (2000). Indels of uniform length for each taxon were scored as present or absent. Where a shorter gap is nested within a longer gap, the two gaps were scored as separate characters, and the shorter gap was scored as missing data for the taxa with the longer gap in the alignment. The ends and all ambiguously aligned sequence (highly variable homopolymer and indel regions) were excluded in all analyses. The data sets are summarized in Table 2.
|
), with a heuristic search algorithm of 1000 random replicates holding 25 trees at each step, and tree-bisection-reconnection (TBR) branch swapping (MulTrees on; Steepest Descent off). Separate analyses were performed including and excluding the indels to determine the effect of these extra characters on topology and support. The trnL intron data set comprised the intron and 3' exon. The trnL-F data set comprised the trnL intron, the 3' exon and the trnL-F intergenic spacer. For each analysis, the strict consensus was calculated and bootstrap support values were generated using a heuristic search algorithm of 1000 bootstrap replicates holding five trees at each step, and nearest-neighbor-interchange (NNI) branch swapping (MulTrees on; Steepest Descent off).
Combined analyses
A partition homogeneity test was carried out to assess incongruent length differences between each pairing of the three gene regions: trnL-F and rbcL, trnL-F and rps16, and rbcL and rps16. However, previous studies have questioned the validity of the incongruence length test (Reeves et al., 2001
; Yoder et al., 2001
), and it may be more reasonable to identify strongly supported topological conflict as a measure of incongruence (Wiens, 1998
). The partition homogeneity test found the rbcL data set to be significantly incongruent (0.002) with the other two data sets. However, any topological conflict between the rbcL and the trnL-F and rps16 analyses was not strongly supported in each of the conflicting data sets, and therefore combined analyses of all three regions were carried out. In a small combined matrix, only taxa which had at least partial sequence data for all three gene regions were included to reduce the problem of missing data in parsimony analysis. In a large combined matrix, all taxa were included, even those for which only one partial sequence was available.
Bayesian analysis
All Bayesian analyses were performed using MrBayes 3.0b3 (Huelsenbeck and Ronquist, 2001
). Model parameters were determined by the Akaike information criterion (AIC) in ModelTest 3.5 (Posada and Crandall, 1998
). Based on the AIC test, the transversion model (TVM) was selected for the trnL-F and rps16 data sets, the transition model (TIM) was selected for the rbcL data set and the general time reversible (GTR) model was selected for the trnL intron data set. The small combined and the large combined data sets were each analyzed using heterogeneous Bayesian analysis in MrBayes with a separate submodel, as specified earlier, assigned to each of the three gene regions represented in the matrix. The analyses were run for 5 000 000 generations, with four chains, sampling every 100th generation. The burn-in stage needed to reach stationarity was determined by plotting the likelihood scores in Microsoft (Redmond, Washington, USA) Excel 2002. The trees sampled from within the burn-in stage were excluded, and the remaining trees were assumed to be representative of the posterior probability distribution. The majority rule consensus tree was calculated in PAUP*, and the resulting branch values represent the posterior probabilities.
Morphological survey
The 85 morphological characters investigated by Haston (2003)
for a subset of taxa included in the molecular anlayses were reviewed with respect to the clades recovered using molecular data. Morphological support for the clades was sought, and levels of morphological variation within the clades assessed.
RESULTS
The Peltophorum group is polyphyletic in all the analyses carried out. In 12 of the 14 analyses, a clade is recovered containing the core Peltophorum group described in Haston et al. (2003)
with Heteroflorum (inedit.) nested within. Thus, the core Peltophorum group, referred to here as the Peltophorum group s.s., now comprises nine genera, Peltophorum, Bussea, Schizolobium, Parkinsonia, Conzattia, Colvillea, Delonix, Lemuropisum, and Heteroflorum (inedit.). The remaining genera of the Peltophorum group s.l. are dispersed throughout the tree (Figs. 1, 2).
|
|
|
The parsimony analysis of the large combined matrix exceeded the maximum number of trees that could be held by the computer, and the strict consensus of the trees saved was highly unresolved. When the number of taxa was limited to those with at least partial sequence data for all three gene regions in the small combined matrix and indels included, only 60 most parsimonious trees were discovered. Of the parsimony analyses of the combined data sets, only the results of the small combined analysis are presented and discussed here (Fig. 1).
Bayesian analysis
All clades with posterior probabilities higher than 95% in the trnL intron and trnL-F analysis are also found in the corresponding parsimony analyses. In the Bayesian analysis of the rps16 data, two clades within the Caesalpinia group, the Cordeauxia edulis/Caesalpinia velutina clade and the Caesalpinia coriaria/C. calycina/C. eriostachys clade, have over 96% posterior probability but are unresolved in the parsimony analysis. In the Bayesian analysis of the small combined data set, the Batesia/Recordoxylon/Melanoxylon clade finds a posterior probability of 99% (Fig. 1), but in the parsimony analysis Batesia floribunda is sister to Pterogyne nitens, although unsupported.
Within Parkinsonia, there is conflict in some well-supported clades between the rbcL data set and the other data sets. In the Bayesian analyses of the trnL-F and the rps16 data, the African species have a posterior probability of 99% and 93%, respectively, for their position as nested within a clade which includes the South American species and one representative of P. florida. In the rbcL analysis, however, the African species form a clade with P. microphyllya, which is sister to all the American species. In the Bayesian analyses of the large combined data set, this conflict results in lack of support for several clades within Parkinsonia and support for the African/South American clade falls to 75% (Fig. 2). This clade is not recovered in the analyses of the small combined data set (Fig. 1).
In our analyses, the method of analyzing the data generally has little effect on the topology of well-supported relationships. However, there can be large shifts in the position of genera or entire clades that are equivocally placed in the parsimony analyses. The higher support values seen in the Bayesian analyses here are consistent with the findings of others (Sanderson and Wojciechowski, 2000
; Alfaro et al., 2001
; Wilcox et al., 2002
).
Morphological survey
Although many clades found good morphological characters in support of their recognition, these clades mostly comprised single genera. The higher level clades found less unambiguous morphological support and several of the clades recovered here were cryptic (as in Wojciechowski et al., 1993
; Sanderson and Wojciechowski, 1996
; Pennington and Gemeinholzer, 2000
; Lavin et al., 2001
; Pennington et al., 2001
).
DISCUSSION
The Peltophorum group, when originally defined by Polhill and Vidal in 1981
, was an association of genera showing "a remarkable diversity in their flowers, fruits, seeds and physical defences" (Polhill and Vidal, 1981
, p. 84). This diversity was seen to comprise a series of increasing complexity with the simplest members, Batesia and Vouacapoua, rather different from the most elaborate, Colvillea, Delonix, and Jacqueshuberia. However, the results of our analyses show a different scheme of relationships. Six of the other groups of the Caesalpinieae are now nested within this series, and the genera of the Peltophorum group are no longer tenable as a natural group. The polyphyly of the group suggested by several higher level analyses (Bruneau et al., 2001
; Kajita et al., 2001
), the analysis of the Caesalpinia group (Simpson et al., 2003
) and an earlier analysis of the trnL-F data (Haston et al., 2003
) is demonstrated here.
The use of informal groups in the Leguminosae has provided a means of communicating preliminary hypotheses of generic groupings without creating extra layers of formal ranks within the family. They have allowed workers to present hypotheses that may then be tested as more data become available. The alternative classifications of Hutchinson (1964)
, Pettigrew and Watson (1977)
, Polhill and Vidal (1981)
, Watson and Dallwitz (1983)
, and Polhill (1994)
all use informal groups in their classifications of the Caesalpinieae. These working hypotheses have been useful in suggesting sampling strategies, but the results of our analyses conflict to a certain extent with all of these classifications. Choosing to redelimit the informal groups based on our analyses would affect several of the informal groups of Polhill (1994)
. Taxonomic recommendations are made here with regard to the genera of the Peltophorum group s.l., which minimize taxonomic changes, but are based on a monophyletic system of classification.
To retain a monophyletic Peltophorum group, it would be necessary to include the whole of the Caesalpinieae, as represented by the genera in our analyses. In rejecting this solution, two alternatives could be considered: (1) to include all genera within clade A (Fig. 2), which would include 11 genera of the Peltophorum group, Heteroflorum, the Sclerolobium group, and possibly the Dimorphandra group and the Mimosoideae if the analyses of Bruneau et al. (2001)
, Kajita et al. (2001)
and Luckow et al. (2000)
are supported with further data; (2) to include all genera within the core Peltophorum group described in Haston et al. (2003)
including Heteroflorum (inedit.) to form the Peltophorum group s.s. This option would allow the reorganization of the Peltophorum group with the least number of changes to the remaining groups in the tribe and would therefore appear to be the best option based on the data here (Fig. 2).
Monophyly of genera and relationships within the Peltophorum group s.s
Eight genera of the Peltophorum group of Polhill (1994)
and Heteroflorum (inedit.) form a clade which is regarded here as the Peltophorum group s.s. (Figs. 1, 2). In terms of the morphological survey, there is no unambiguous morphological character defining the Peltophorum group s.s., rather the clade is characterized by a combination of features. Whereas the Peltophorum group s.l. held a mixture of bipinnate- and pinnate-leaved trees, most of the trees with pinnate leaves are excluded from the Peltophorum group s.s., which is now essentially bipinnate. Pinnate leaves are now present only in Lemuropisum edule and, in association with bipinnate and partially bipinnate leaves, in Heteroflorum (inedit.). In the Peltophorum group s.s. the petals are generally yellow, but may be white or red in Delonix, and rarely white or pink in Peltophorum. Seeds are generally narrow, less than 8 mm wide, and between 2.5 mm and 10 mm in thickness.
Five suprageneric clades within the Peltophorum group s.s. are identified in the analyses of the combined data (Figs. 1, 2). Support for these clades and the relationship between them are discussed next. Where more than one species or subspecies represents a genus in at least one analysis, the monophyly of the genus and the relationships within the genus are discussed.
The Bussea/Peltophorum clade
In all the analyses, Bussea and Peltophorum form a clade, sister to the rest of the core Peltophorum group. This close relationship is implied by their nomenclatural history. The first Bussea species was described as P. massaiense by Taubert (1895)
, 7 years later becoming the type species for a new genus described by Harms (1902)
. The clade is morphologically distinct, sharing the presence of a peltate stigma from which Peltophorum takes its name (pelta = shield). The endocarp of Bussea and P. dubium is pubescent, a feature not found in any other species examined in our analyses. The endocarp of the two other Peltophorum species, P. africanum and P. pterocarpum, was not observed. The stamens tend to be shorter than the petals, although they are equal in length in P. africanum. The two genera are morphologically distinct, Bussea having thick, dehiscent fruit without wings, Peltophorum having thin, indehiscent fruit with wings on both margins. The anther connective in Peltophorum is split from the base of the anther to near the union with the filament forming a sagittate anther, and the sepals are connate for a short length at the base, forming a ring of tissue that can detach from the hypanthium and lie loosely round the fruit stalk. None of these characteristics are found in Bussea.
The genus Peltophorum was represented by three (of between four and six) species in all our analyses and is strongly supported as monophyletic in all but the separate parsimony analysis of the rbcL data in which it finds moderate support (85% bootstrap). The species included here broadly represent the geographical distribution of the genus: P. dubium from tropical America, P. africanum from tropical Africa, P. pterocarpum from Asia. The species that have not been included are found mostly within the range of P. pterocarpum.
The African and Malagasy genus Bussea was represented by three (of seven) species in the analyses of the trnL intron and in the Bayesian analysis of the large combined matrix, B. gossweileri, B. sakalava, and B. xylocarpa. The genus was found to be monophyletic in all these analyses, with moderate bootstrap support in the parsimony analyses and 100% posterior probability in the Bayesian analysis. The species representing the genus here are from both the eastern and western part of its range, including Madagascar.
The Schizolobium clade
The position of the monospecific tropical American genus Schizolobium, linking the Bussea/Peltophorum clade and the three other clades within the core Peltophorum group only finds strong support in the Bayesian analysis of the small and large combined data sets (Figs. 1, 2). Schizolobium is distinguished from the other members of the Peltophorum group s.s. by the lack of stipules and by the fruit morphology and dehiscence type. The single-seeded fruit is flattened and oblanceolate or spoon-shaped. During dehiscence, the endocarp is released intact as a thin, papery wing-like layer, with the seed held inside.
The two varieties of Schizolobium parahyba have a disjunct Central American–Amazonian distribution and are further distinguished by the presence or absence of an articulated joint on the pedicels (Barneby, 1996
). Individuals from both geographical areas were represented in these analyses and they were always found to form a clade.
The Parkinsonia clade
All species of Parkinsonia and Cercidium have been included in at least one analysis here. Previous delimitations of the genus Parkinsonia have differed in the recognition of a distinct genus Cercidium (Sargent, 1889
; Johnston, 1924
; Carter, 1974
) or in treating the two genera as congeneric under the prior name, Parkinsonia (Watson, 1876
; Brenan, 1963
, 1980
; Polhill and Vidal, 1981
; Hawkins, 1996
). The results of our molecular analyses discover a phylogeny in which there is no support for Parkinsonia and Cercidium as distinct genera. Together, Parkinsonia and Cercidium form a strongly supported monophyletic group, well separated from Conzattia (Figs. 1, 2). The majority of the nomenclatural synonymy resulting from the sinking of Cercidium into Parkinsonia has been carried out, but one species, Cercidium andicola, and several doubtful subspecific names are yet to be published as new combinations in the genus Parkinsonia. The genera are regarded as congeneric here and will be referred to as Parkinsonia.
Within Parkinsonia, the African species form a well-supported clade that is recovered in every analysis, with the East African species, P. anacantha, P. raimondoi, and P. scioana, being more closely related to each other than to the southern African species, P. africana. The results of the morphological analyses of Hawkins (1996)
suggested that two East African species, P. scioana and P. raimondoi, were closely related but that the South African species, P. africana, was more closely related to the American species. There are morphological similarities between P. africana and the South American C. andicola, including the presence of a reduced primary rachis, an asymmetrical hypanthium, and leaflets that are glabrous on the abaxial surface and less than 2 mm in length. However, the molecular data strongly support an African clade in which P. africana is sister to a clade of the three East African species, although the position of the African clade within Parkinsonia is not yet unambiguously resolved. Within the large American clade, a smaller North American clade is present in all the analyses and strongly supported in the combined analysis, comprising P. texana and two representatives of P. florida. The third representative of P. florida is placed sister to P. praecox in all the Bayesian analyses except that of the rps16 data. The positions of two South American species, C. andicola and the newly discovered P. peruviana, are unsupported relative to the larger clade of mainly North American species. The dependence in these analyses on chloroplast data may cause some difficulties in confirming the relationships in the American species of Parkinsonia that have shown marked hybridization (Hawkins, 1996
; Hawkins and Harris, 1998
).
The Conzattia/Heteroflorum clade
The position of the Conzattia/Heteroflorum (inedit.) clade, sister to the Delonix/ Colvillea/Lemuropisum clade, contradicts the suggestion that Conzattia itself is more closely related to Parkinsonia (including Cercidium) and Peltophorum (Rose, 1909
; Miranda, 1955
). Heteroflorum (inedit.) is well supported as sister to Conzattia (Figs. 1, 2). Both genera are endemic to Mexico where they grow in semideciduous scrub and dry forest. Both genera also show cryptic dioecy, a trait also observed in Parkinsonia anacantha and the Caesalpinia group (Lewis et al., 2000
).
The Delonix/Colvillea/Lemuropisum clade
The relationship between Delonix, Colvillea, and Lemuropisum has been uncertain. Lemuropisum, a monospecific genus endemic to Madagascar, was described by Perrier in 1938 who suggested a close affinity between it, Colvillea, and Poinciana (Delonix) based on floral morphology. However, in some later classifications it has been separated from Colvillea and Delonix (Hutchinson, 1964
; Pettigrew and Watson, 1977
; Polhill and Vidal, 1981
). The more recent classification of Polhill (1994)
placed Lemuropisum in the Peltophorum group with Delonix and Colvillea following the reappraisal of the Caesalpinia group by Lewis and Schrire (1995)
, which was published the following year. A revision of the Malagasy species of Delonix by Du Puy et al. (1995)
examined the generic relationships of Delonix, Colvillea, and Lemuropisum and also recommended that Lemuropisum be classified with Delonix and Colvillea. In our analyses, the clade comprising Delonix, Colvillea, and Lemuropisum is well supported as monophyletic (Figs. 1, 2), characterized principally by the presence of a valvate calyx (also seen in Cordeauxia and Moldenhawera), the sepals of which do not reflex at anthesis. The general size of the flower parts is also a feature of the clade. The sepals are long, 11.5– 28 mm. The stamens are usually longer than the petals, and are associated with a long style. The seeds lie transversely across the fruit in Delonix, Colvillea, and Lemuropisum. The seeds of D. pumila were not observed in the fruit, and thus the orientation of the seeds could not be assessed.
Although there is no support for the relationship between Colvillea and Delonix, with Colvillea being placed alternatively within Delonix or as sister to a Delonix/Lemuropisum clade, the monospecific Malagasy genus Lemuropisum forms a strongly supported clade with three Malagasy species of Delonix, D. floribunda, D. brachycarpa, and D. pumila, in the analysis of the small combined matrix. This corroborates the close affinity of Lemuropisum to the genus Delonix that was noted by Du Puy et al. (1995)
. The inclusion of Lemuropisum edule within this grouping, as sister to D. pumila, is supported by several morphological characters, including (1) reduced size—both species are shrubby with slightly contorted branches (Du Puy et al., 1995
); (2) reduced leaves—D. pumila has one or occasionally two pairs of pinnae, each with 2–4 pairs of leaflets and L. edule has pinnate leaves, which may be considered to be reduced, each bearing 2–3 pairs of leaflets, (3) in the absence of articulation on the pedicel, which is present in all other species sampled in the core Peltophorum group; and (4) in the spiraling of the pods. Delonix brachycarpa shares the fewer pairs of pinnae and leaflets with the other species in this grouping, usually having two pairs of pinnae and 3–6 pairs of leaflets per pinna. There is also a tendency in this group to have shorter stipules, glabrous adaxial leaflet surfaces and stamen filaments, as well as pods that are constricted along the margins. In the analysis of the large combined data set, the clade is no longer strongly supported, and additional data are needed to test this relationship. The position of this clade with reference to the remaining Delonix species and Colvillea is equivocal.
The genus Delonix comprises 11 species (Du Puy et al., 1995
), eight of which were sampled here. Nine species are endemic to Madagascar and two are non-Malagasy, one restricted to East Africa and the other found through East and northeastern Africa to the Middle East and India. In their revision of the genus in Madagascar, Du Puy et al. (1995)
discussed the morphological variation and relationships both within the genus and between Delonix, Colvillea, and Lemuropisum. They found that the earlier grouping of two Delonix species, D. floribunda and D. velutina, as a separate genus, Aprevalia, (Baillon, 1884
; Capuron, 1968
) based solely on the reduction of the petals in these two species was not supported by the morphology of the leaves and pods. They suggested that D. floribunda bears a stronger resemblance to D. decaryi, D. leucantha, and D. pumila, with leaves having few pinnae and leaflets and pods that are linear-oblong, and that D. velutina is more similar to D. regia and D. tomentosa with numerous pinnae and leaflets and long strap-shaped pods. Du Puy et al. (1995)
suggested that D. brachycarpa and D. boiviniana may be conspecific and that their position in the genus was unclear. They found that the two non-Malagasy species, D. elata and D. baccal, were morphologically distinct and closely related to one another. The implication is therefore that there are three distinct subgroups within the genus, two within Madagascar and one outside.
In our analyses, a clade comprising the two non-Malagasy species, D. baccal and D. elata, finds strong support from the small combined data set (Fig. 1). Both species have filiform rather than foliaceous stipules and flat fruit, under 7 mm thick. When the sampling of D. elata is increased in the large combined data set, one of the three exemplars does not fall within the D. baccal/D. elata clade and its position is unresolved (Fig. 2). Although Delonix regia forms a clade with D. baccal and D. elata in some analyses (Fig. 1), this clade collapses when D. velutina and further representatives of D. regia and D. elata are included (Fig. 2), and the position of D. regia remains equivocal. The second Malagasy group, comprising Delonix velutina, D. regia, and D. tomentosa, was represented in our analyses by D. velutina and D. regia, although D. velutina could not be included in all molecular analyses due to missing data. Morphological data suggest that Colvillea racemosa forms part of this grouping, sharing bipinnate leaves that bear many pinnae and leaflets and longer strap-shaped pods (Du Puy et al., 1995
) as well as the pubescent adaxial leaflet surfaces that appear to characterize this clade. Colvillea racemosa and Delonix regia both bear stipules that are highly branched, almost fernlike. They also share longer bracts and petals than the other Delonix species, and pubescence on the banner petal. Unfortunately, no flowers of D. velutina were observed in our study, and the stipules of Delonix velutina and D. tomentosa have not been observed here nor described in the literature. Delonix regia has been introduced throughout the tropics as an ornamental street tree, but is considered to be originally endemic to Madagascar where its distribution is threatened by logging for charcoal (Du Puy et al., 1995
). The monospecific genus Colvillea is also endemic to Madagascar, but is also widely cultivated as an ornamental street tree (Du Puy et al., 1995
).
The results of our analyses therefore support the relationships within Delonix suggested by Du Puy et al. (1995)
but include the monospecific Lemuropisum and possibly also Colvillea, within the genus. The nomenclatural implications of our results would suggest that both Lemuropisum edule and possibly Colvillea racemosa be sunk within a broader circumscription of Delonix. The alternative would require the description of at least two new genera. The earlier name of Aprevalia could be resurrected for the Delonix pumila clade, which contains the type species of Aprevalia, D. floribunda, but the exclusion of D. velutina may cause some confusion in the use of this name. The other name that has been associated with Delonix is Poinciana, but the type species of this name is P. pulcherrima which is now placed within Caesalpinia. These nomenclatural complications argue in favor of a single genus, but the new combinations are not made here. We suggest that more data should be brought to bear on the problem, particularly because Lemuropisum edule is an economically important food source for the arid regions of Madagascar and Africa, and Colvillea racemosa is an important street tree in many parts of the tropics.
Monophyly and relationships of genera outwith the Peltophorum group s.s. and proposal for a reclassification of the genera of the Peltophorum group s.l
The eight genera currently placed in the Peltophorum group but not included in the Peltophorum group s.s. in our analyses are all restricted to South America. The relationships of these eight genera are reviewed here.
Campsiandra and Vouacapoua
Campsiandra comosa was only included in the parsimony analysis of the trnL intron and the Bayesian analysis of the large combined matrix. In both analyses, it was sister to Vouacapoua americana (Fig. 2). This relationship is supported by their geographical distribution in the tropical northeast of South America, and by several morphological features. Both have once-pinnate, imparipinnate leaves, and small flowers in terminal panicles. Bracteoles are present but caducous, and the calyx is imbricate in both genera (Watson and Dallwitz, 1983
). Both genera have also been recorded as nodulating (Sprent, 1994
, 2000
). The Campsiandra/ Vouacapoua americana clade is moderately supported as sister to the large clade of the remaining species apart from Poeppigia procera. In the small combined analysis in which Campsiandra was not included, Vouacapoua americana finds strong support in this position (Fig. 1).
The taxonomy of the tropical South American genus Vouacapoua has been historically closely linked with that of the papilionoid genus Andira (Pennington, 2003). When Larmarck (1783) described Andira, he included the earlier genus, Vouacapoua, within his circumscription, citing V. americana in synonymy with A. racemosa, making Andira illegitimate. Vouacapoua is now accepted as a distinct caesalpinioid genus, and Andira Lam. has now been conserved with a conserved type, A. inermis (Pennington, 2002). Vouacapoua is now thought to comprise three species. Of the two species represented in our analyses, one is the type species of the genus, V. americana, and the other, V. macropetala, was described by Sandwith (1937)
as a new species after the main elements of taxonomic and nomenclatural confusion had been resolved. The genus is not monophyletic in our analyses. Vouacapoua macropetala, in the Bayesian analysis of the large combined data, is strongly supported as phylogenetically separated from V. americana, but with no clear alternative sister relationship to assess. More data are required to adequately test the monophyly of the genus.
We propose a new Campsiandra group (Fig. 2). This group may also include the genus Vouacapoua if the relationship between them is supported with denser sampling, but the position of V. macrocarpa outwith the Campsiandra/Vouacapoua clade suggests that the monophyly and the relative position of Vouacapoua is not yet clear.
Batesia, Melanoxylon and Recordoxylon
Three South American genera, Batesia, Melanoxylon, and Recordoxylon form a clade in the analysis of the rps16 data and in both the small and large combined analyses (Figs. 1, 2). The affinity between Melanoxylon and Recordoxylon is apparent from their nomenclatural history. The genus Recordoxylon was described in 1934 by Ducke based on Melanoxylon amazonicum, defined by wood anatomy and fruit morphology, and named after Professor Record who had compared the anatomy of Melanoxylon brauna and M. amazonicum (Ducke, 1934
).
The small clade comprising Batesia, Recordoxylon, and Melanoxylon, may best be considered as a new Batesia group (Fig. 2). If the Peltophorum group s.s. is accepted, then this is the most inclusive option for this clade which would not render it paraphyletic.
Moldenhawera
The position of Moldenhawera within the Caesalpinieae has been uncertain, although the preliminary analyses of L. P. de Queiroz (Universidade Estadual de Feira de Santana, Bahia, Brazil, unpublished data) suggested two genera as putative sister groups, Melanoxylon and Recordoxylon. Given the amount of geographic structure we are finding in the Peltophorum group, the Brazilian distribution suggests ties with the Sclerolobium group or the Batesia/Melanoxylon/ Recordoxylon clade discovered here. The valvate calyx suggests an affinity with the Peltophorum group s.s. In the Bayesian analysis of the combined data presented here, Moldenhawera is supported as sister to a large clade comprising the Sclerolobium/Tachigali/Arapatiella/Jacqueshuberia clade and the Peltophorum group s.s. (Fig. 1). However, in the analysis of the large combined matrix, the inclusion of Erythrophleum ivorense, currently in the Dimorphandra group of the Caesalpinieae, adds a degree of uncertainty to the relationships in this part of the tree (Fig. 2).
The genus Moldenhawera comprises nine species, all of which are endemic to eastern Brazil. The three species representing Moldenhawera in our analyses form a strongly supported clade with a bootstrap value and posterior probability of 100%. This agrees with the views of Queiroz et al. (1999)
, which were based on a preliminary morphological analysis (L. P. de Queiroz, Universidade Estadual de Feira de Santana, Bahia, Brazil, unpublished data) and on the presence of several observed synapomorphies (Queiroz et al., 1999
). The plasticity of the leaf morphology in Moldenhawera was observed by Queiroz et al. (1999)
and is marked particularly by the presence of partially bipinnate leaves in M. papillanthera and M. polysperma. Moldenhawera also has dimorphic anthers, which are not shared by any other taxa in this analysis. All three species included here are from within the section Brasilianae of Queiroz et al. (1999)
. The other two sections, section Acuminatae and section Moldenhawera, are not represented here.
Moldenhawera appears to be unclearly associated with any other genus or clade and could therefore form a monogeneric group, the Moldenhawera group (Fig. 2), although the affinities of this genus may become clearer when more taxa are sampled from the Dimorphandra group of the Caesalpinioideae and from the subfamily Mimosoideae in future analyses.
Jacqueshuberia and Arapatiella
In the combined analyses, Jacqueshuberia and Arapatiella, small genera from the tropical north and northeast of South America, are placed in a clade with Tachigali and Sclerolobium, large South American genera mainly found in Amazonia. Tachigali and Sclerolobium have been considered congeneric under the prior name Tachigali (Brako and Zarucchi, 1993
; Barneby, 1996
; Kirkbride et al., 2000
), but are treated as separate genera in our sampling. Arapatiella and Tachigali have been associated in their nomenclatural history. Harms (1915)
first described Tachigali psilophylla. Rizzini and Mattos (1972)
described the genus Arapatiella based on A. trepocarpa. In 1973, Cowan decided that these two species were conspecific and placed A. trepocarpa in synonymy with Tachigali psilophylla under the new combination of A. psilophylla (Cowan, 1973
). The close relationship between Arapatiella and Jacqueshuberia is supported by pollen morphology. The pollen of Jacqueshuberia has long been known to have viscin threads to which the pollen adheres (Arroyo, 1981
). Arapatiella has been found to have similar viscin threads present in the anthers (Banks et al., 2003
).
Of the seven species of Jacqueshuberia currently described, two were included in our analyses. All species are found in northern South America, with each species having a narrow distribution, usually within a single country. The two species included in our analyses, J. loretensis and J. purpurea, are found in Peru and Brazil respectively. Both species formed a strongly supported clade in the small combined analysis (Fig. 1), the trnL-F analyses and the Bayesian analysis of the rps16 data. The clade is less well supported in the large combined analysis (Fig. 2).
The genus Arapatiella comprises two species, both of which are restricted to Brazil. Both species were included in the trnL intron analyses and the Bayesian analysis of the large combined matrix. Parsimony fails to resolve a polytomy comprising species of Arapatiella and Jacqueshuberia. Bayesian analysis results in Arapatiella being polyphyletic but with low posterior probablities. Arapatiella psilophylla has only trnL intron sequence data at present, and further data may influence the phylogenetic placement of the two species.
The Sclerolobium group currently comprises three genera, Sclerolobium, Tachigali, and Diptychandra. If Sclerolobium and Tachigali are accepted as congeneric under the prior name, Tachigali, then the name of the informal group would be based on a name in synonymy. The results of our analyses suggest that the delimitation of the group could be broadened to include Arapatiella and Jacqueshuberia to form a more appropriately named Tachigali group (Fig. 2).
The informal groups proposed here are another step towards a revised phylogenetic classification that represents the Caesalpinieae. These hypotheses are put forward to be tested with denser sampling of the Dimorphandra and Caesalpinia groups, the Cassieae, and the basal Mimosoideae. As the phylogeny becomes more robust, more cryptic characters may be found to validate the generic groupings discovered here. In their phylogenetic analysis of the Caesalpinioideae, Bruneau et al. (2001)
remarked that the five genera that are confirmed as nodulating in the Caesalpinieae were members of the Dimorphandra and Peltophorum groups. Interestingly, when these five genera and two further records (Sprent, 2000
) are reexamined in light of these analyses, all nodulating species fall outwith the core Peltophorum group s.s. apart from a single unconfirmed report of nodulation in Colvillea (Allen and Allen, 1981
), which Sprent has suggested should be treated with caution (Sprent, 2001
). Areas of the molecular phylogeny that are at present equivocal may yet correspond with data from other sources, including nodulation data.
When Polhill and Vidal proposed the informal groups in the Caesalpinieae, they referred to centers of diversity within the groups, and they suggested that the Peltophorum group was centered in tropical South America (Polhill and Vidal, 1981
). However, the exclusion of eight genera, all of which are South American, has resulted in the Peltophorum group s.s. now being centered in Central America and Africa with distinct geographical clades within the group. It remains to be seen if the two other geographically diverse groups in the tribe, the Caesalpinia and Dimorphandra groups, are also geographically delimited. Phylogenetic analyses have been carried out on the genera of the Caesalpinia group (Simpson et al., 2003
), but relationships within the Dimorphandra group remain uncertain.
Conclusions
The recovery of a monophyletic core Peltophorum group, which is referred to here as the Peltophorum group s.s., and the distribution across the Caesalpinieae of four other components of the Peltophorum group s.l., allow us to propose some changes to the classification of the tribe. The Dimorphandra group has been under-represented in phylogenetic analyses to date and may prove to be critical in the understanding of relationships among Pterogyne, Moldenhawera, the Tachigali group, and the Mimosoideae. The inclusion of genera from the Dimorphandra group, as well as from the Cassieae and basal Mimosoideae in further phylogenetic analyses, will enable a more complete picture of the tribe to be discovered. The results of our molecular analyses suggest that faster evolving gene regions will be needed to fully resolve relationships between these genera and to recover a robust phylogeny on which to base future classifications.
The novel insights from the phylogeny presented here provide better explanations of both morphology and geographical distribution. In the future, it will be desirable to communicate these insights in the form of a revised classification. Such a reclassification can only take place as part of a more extensive project to reconstruct relationships in the Caesalpinioideae. Until this project is realized, we have retained the informal group rank, which has proved to be useful in allowing communication of results and concepts without major changes in formal nomenclatural ranks.
FOOTNOTES
1 We thank Anne Bruneau and Felix Forest for providing sequence data, the Regional Plant Introduction Station, Iowa and the Oxford Forestry Institute for providing seed, and the curators at the following herbaria for allowing access to collections: BM, E, FHO, K and NY. We thank Colin Hughes for constructive comments on an early version of this paper, Mario Sousa S. for information on Heteroflorum, Hannah Banks for providing unpublished information on the pollen of Arapatiella, Sue Mott, George Gibbings and Alastair Culham for technical assistance, and Dylan de Silva and Rolando Barcenas Luna for helpful discussion throughout the study. Finally, we thank Anne Bruneau and one anonymous reviewer for their constructive comments, which led to a more concise paper. This work was funded by the Natural Environment Research Council, UK (Grant: GT04/1999/TS/0235). 
4 Author for correspondence (e-mail: e.haston{at}rbge.org.uk
) present address: Tropical Diversity, The Royal Botanic Garden, Edinburgh, 20a Inverleith Row, Edinburgh, EH3 5LR, UK
LITERATURE CITED
Alfaro M. E. S. Zoller F. Lutzoni 2001 Comparative performance of the bootstrap and Bayesian MCMC sampling in assessing phylogenetic confidence: a simulation study. American Zoologist 41: 1379-1379
Allen O. N. E. K. Allen 1981 The Leguminosae: a source book of characteristics, uses, and nodulation. University of Wisconsin Press, Madison, Wisconsin, USA
Andersson L. S. Andersson 2000 A molecular phylogeny of Tropaeolaceae and its systematic implications. Taxon 49: 721-736[CrossRef][ISI]
Arroyo M. T. K. 1981 Breeding systems and pollination biology in Leguminosae. In R. M. Polhill and P. H. Raven [eds.], Advances in legume systematics, part 2, 723–769. Royal Botanic Gardens, Kew, Richmond, UK
Baillon, 1884 Liste des plantes de Madagascar: un nouveau type de Caesalpiniees monopetales. Bulletin Mensuel de la Societe Linneenne de Paris 1: 428-429
Banks H. B. B. Klitgaard G. P. Lewis P. R. Crane A. Bruneau 2003 Pollen and the systematics of tribes Caesalpinieae and Cassieae (Caesalpinioideae: Leguminosae). In A. Bruneau and B. B. Klitgaard [eds.], Advances in legume systematics, part 10, 95–122. Royal Botanic Gardens, Kew, Richmond, UK
Barneby R. C. 1996 Neotropical Fabales at NY: asides and oversights. Brittonia 48: 174-187[CrossRef][ISI]
Brako L. J. L. Zarucchi 1993 Catalogue of the flowering plants and gymnosperms of Peru. Monographs in systematic botany, vol. 45. Missouri Botanical Garden, St. Louis, Missouri, USA
Brenan J. P. M. 1963 Notes on African Caesalpinioideae. Kew Bulletin 17: 197-214[CrossRef]
Brenan J. P. M. 1980 A new species of Parkinsonia (Leguminosae) from Somalia. Kew Bulletin 35: 563-565[CrossRef]
Bruneau A. F. J. Breteler J. J. Wieringa G. Y. F. Gervais F. Forest 2000 Phylogenetic relationships in tribes Macrolobieae and Detarieae as inferred from chloroplast trnL intron sequences. In P. S. Herendeen and A. Bruneau [eds.], Advances in legume systematics, part 9, 121–150. Royal Botanic Gardens, Kew, Richmond, UK
Bruneau A. F. Forest P. S. Herendeen B. B. Klitgaard G. P. Lewis 2001 Phylogenetic relationships in the Caesalpinioideae (Leguminosae) as inferred from chloroplast trnL intron sequences. Systematic Botany 26: 487-514[ISI]
Capuron R. 1968 Contribution a l'etude de la flore forestiere de Madagascar. Reduction du genre Aprevalia Baillon au rang de section du genre Delonix Raf. et description d'une espece nouvelle (Leguminosae, Caesalp). Adansonia, series 2 8: 11-16
Carter A. M. 1974 The genus Cercidium (Leguminosae: Caesalpinioideae) in the Sonoran Desert of Mexico and the United States. Proceedings of the California Academy of Sciences 40: 17-57
Cowan R. S. 1973 Studies of tropical American Leguminosae. VII. Proceedings of the Biological Society of Washington 86: 447-460
Doyle J. J. J. A. Chappill C. D. Bailey T. Kajita 2000 Towards a comprehensive phylogeny of legumes: evidence from rbcL sequences and non-molecular data. In P. S. Herendeen and A. Bruneau [eds.], Advances in legume systematics, part 9, 1–20. Royal Botanic Gardens, Kew, Richmond, UK
Doyle J. J. J. L. Doyle J. A. Ballenger E. E. Dickson T. Kajita H. Ohashi 1997 A phylogeny of the chloroplast gene rbcL in the Leguminosae: taxonomic correlations and insights into the evolution of nodulation. American Journal of Botany 84: 541-554[Abstract]
Du Puy D. J. P. B. Phillipson R. Rabevohitra 1995 The genus Delonix (Leguminosae: Caesalpinioideae: Caesalpinieae) in Madagascar. Kew Bulletin 50: 445-475[CrossRef]
Ducke A. 1934 Recordoxylon: a new genus of Leguminosae-Caesalpinioideae. Tropical Woods 39: 16-18
Harms H. 1902 Beitrage zur flora von Afrika, 24. Botanische Jahrbucher fur Systematik, Pflanzengeschichte und Pflanzengeographie 33: 159-160
Harms H. 1915 Leguminosae-Caesalpinioideae. Notizblatt des Koniglichen botanischen Gartens und Museums du Berlin, Leipzig 6: 304-310[CrossRef]
Harris S. A. 1995 Systematics and randomly amplified polymorphic DNA in the genus Leucaena (Leguminosae, Mimosoideae). Plant Systematics and Evolution 197: 195-208[CrossRef][ISI]
Haston E. M. 2003 A phylogenetic investigation of the Peltophorum group (Caesalpinieae, Leguminosae) and taxonomic revision of the genus Conzattia Rose. Ph.D. dissertation, University of Reading, Reading, UK
Haston E. M. G. P. Lewis J. A. Hawkins 2003 A phylogenetic investigation of the Peltophorum group (Caesalpinieae: Leguminosae). In A. Bruneau and B. B. Klitgaard [eds.], Advances in legume systematics, part 10, 149–159. Royal Botanic Gardens, Kew, Richmond, UK
Hawkins J. A. 1996 Systematics of Parkinsonia L. and Cercidium Tul. (Leguminosae: Caesalpinioideae). Ph.D. dissertation, University of Oxford, Oxford, UK
Hawkins J. A. S. A. Harris 1998 RAPD characterisation of two neotropical hybrid legumes. Plant Systematics and Evolution 213: 43-55[CrossRef][ISI]
Heywood V. H. 1971 The Leguminosae: a systematic purview. In J. B. Harborne, D. Boulter, and B. L. Turner [eds.], Chemotaxonomy of the Leguminosae, 1–30. Academic Press, London, UK
Huelsenbeck J. P. F. Ronquist 2001 MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics 17: 754-755
Hutchinson J. 1964 The genera of flowering plants. Clarendon Press, Oxford, UK
Ireland H. R. T. Pennington J. Preston 2000 Molecular systematics of the Swartzieae. In P. S. Herendeen and A. Bruneau [eds.], Advances in legume systematics, part 9, 217–231. Royal Botanic Gardens, Kew, Richmond, UK
Johnston I. M. 1924 Taxonomic records concerning American spermatophytes. Contributions from the Gray Herbarium of Harvard University 70: 61-68
Kajita T. H. Ohashi Y. Tateishi C. D. Bailey J. J. Doyle 2001 rbcL and legume phylogeny, with particular reference to Phaseoleae, Millettieae, and allies. Systematic Botany 26: 515-536[ISI]
Kass E. M. Wink 1996 Molecular evolution of the Leguminosae: phylogeny of the three subfamilies based on rbcL sequences. Biochemical Systematics and Ecology 24: 365-378[CrossRef]
Kirkbride J. C. R. Gunn A. L. Weitzmann M. J. Dallwitz 2000 Legume (Fabaceae) fruits and seeds. Parkway Publishers, Boone, North Carolina, USA
Lamarck J. B. A. P. M. 1783 Encyclopédia méthodique, botanique, vol. 1, 171. Panckoucke, Paris, France
Lavin M. R. T. Pennington B. B. Klitgaard J. I. Sprent H. C. de Lima P. E. Gasson 2001 The dalbergioid legumes (Fabaceae): delimitation of a pantropical monophyletic clade. American Journal of Botany 88: 503-533
Lee J. T. Hymowitz 2001 A molecular phylogenetic study of the subtribe Glycininae (Leguminosae) derived from the chloroplast DNA rps16 intron sequences. American Journal of Botany 88: 2064-2073
Lewis G. P. B. D. Schrire 1995 A reappraisal of the Caesalpinia group (Caesalpinioideae: Caesalpinieae) using phylogenetic analysis. In M. D. Crisp and J. J. Doyle [eds.], Advances in legume systematics, part 7, 41–52. Royal Botanic Gardens, Kew, Richmond, UK
Lewis G. P. B. B. Simpson J. L. Neff 2000 Progress in understanding the reproductive biology of the Caesalpinioideae (Leguminosae). In P. S. Herendeen and A. Bruneau [eds.], Advances in legume systematics, part 9, 65–78. Royal Botanic Gardens, Kew, Richmond, UK
Luckow M. P. J. White A. Bruneau 2000 Relationships among the basal genera of Mimosoid legumes. In P. S. Herendeen and A. Bruneau [eds.], Advances in legume systematics, part 9, 165–180. Royal Botanic Gardens, Kew, Richmond, UK
Miranda F. 1955 Ensayo de evaluación de las relaciones entre los generos Conzattia, Peltophorum y Cercidium. Boletín de la Sociedad Botanica de México 18: 7-10
Oxelman B. M. Liden D. Berglund 1997 Chloroplast rps16 intron phylogeny of the tribe Sileneae (Caryophyllaceae). Plant Systematics and Evolution 206: 393-410[CrossRef][ISI]
Pennington R. T. 2002 Proposal to change the authorship of Andira nom. cons. (Leguminosae-Papilionoideae) and to conserve it with a conserved type. Taxon 51: 385-386[CrossRef]
Pennington R. T. 2003 A monograph of Andira (Leguminosae-Papilionoideae). Systematic Botany Monographs, vol. 64. American Society of Plant Taxonomists, Ann Arbor, Michigan, USA
Pennington R. T. B. Gemeinholzer 2000 Cryptic clades, fruit wall morphology and biology of Andira (Leguminosae: Papilionoideae). Botanical Journal of the Linnean Society 134: 267-286[CrossRef]
Pennington R. T. M. Lavin H. Ireland B. B. Klitgaard J. Preston J. M. Hu 2001 Phylogenetic relationships of basal papilionoid legumes based upon sequences of the chloroplast trnL intron. Systematic Botany 26: 537-556[ISI]
Pettigrew C. J. L. Watson 1977 On the classification of Caesalpinioideae. Taxon 26: 57-64
Polhill R. M. 1994 Complete synopsis of legume genera. In F. A. Bisby, J. Buckingham, and J. B. Harborne [eds.], Phytochemical dictionary of the Leguminosae, xlix–liv. Chapman & Hall, London, UK
Polhill R. M. J. E. Vidal 1981 Caesalpinieae. In R. M. Polhill and P. H. Raven [eds.], Advances in legume systematics, part 1, 81–95. Royal Botanic Gardens, Kew, Richmond, UK
Posada D. K. A. Crandall 1998 Modeltest: testing the model of DNA substitution. Bioinformatics 14: 817-818
Queiroz L. P. de. G. P. Lewis R. Allkin 1999 A revision of the genus Moldenhawera Schrad. (Leguminosae—Caesalpinioideae). Kew Bulletin 54: 817-852[CrossRef]
Reeves G. M. Chase P. Goldblatt P. Rudall M. Fay A. Cox B. Lejeune T. de Chies 2001 Molecular systematics of Iridaceae: evidence from four plastid DNA regions. American Journal of Botany 88: 2074-2087
Rizzini C. T. A. D. Mattos Filho 1972 Sobre Arapatiella trepocarpa n.g. & sp. (Leguminosae, Caesalpinioideae). Revista Brasileira de Biologia 32: 323-333
