Phylogeny of Robinioid Legumes (Fabaceae) Revisited: Coursetia and Gliricidia Recircumscribed, and a Biogeographical Appraisal of the Caribbean Endemics
Total Page:16
File Type:pdf, Size:1020Kb
Systematic Botany (2003), 28(2): pp. 387±409 q Copyright 2003 by the American Society of Plant Taxonomists Phylogeny of Robinioid Legumes (Fabaceae) Revisited: Coursetia and Gliricidia Recircumscribed, and a Biogeographical Appraisal of the Caribbean Endemics MATT LAVIN,1,6 MARTIN F. W OJCIECHOWSKI,2 PETER GASSON,3 COLIN HUGHES,4 and ELISABETH WHEELER5 1Department of Plant Sciences, Montana State University, Bozeman, Montana 59717 ([email protected]); 2Department of Plant Biology, Arizona State University, Tempe, Arizona 85287; 3Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AB, UK; 4Department of Plant Sciences, University of Oxford, South Parks Road, Oxford, OX1 3RB, UK; 5Department of Wood and Paper Science, North Carolina State University, Raleigh, NC 27695; 6Author for correspondence Communicating Editor: Thomas G. Lammers ABSTRACT. Morphological data and sequences from the nuclear ribosomal ITS region, and the chloroplast trnL intron and matK locus were sampled from robinioid legumes to infer phylogenetic relationships. The monophyletic robinioid clade includes 11 genetically and often morphologically distinct subclades ranked as genera with the following well supported higher level relationships: ((Hebestigma, Lennea), ((Gliricidia, Poitea), (Olneya, Robinia, Poissonia, Coursetia, Peteria, Genistidium, and Sphinctospermum))). In order to render all 11 robinioid genera monophyletic, the genus Hybosema is synonymized with Gliricidia,andthegenusPoissonia is resurrected to accommodate four morphologically disparate species previously classi®ed in Coursetia. Three new combinations are required to accommodate these two generic recircumscriptions: Gliricidia robus- tum, Poissonia heterantha, and Poissonia weberbaueri. Ages of clades and evolutionary substitution rates are derived from a rate-smoothed Bayesian likelihood approach on sequences from the ITS region and the matK locus. Time constraints are derived from the Tertiary fossil wood species Robinia zirkelii, which shares apomorphic wood characters with the Robinia stem clade. The Cuban endemic Hebestigma is estimated to have diverged at least 38 Ma from its Mesoamerican sister genus Lennea, whereas the Greater Antillean Poitea is estimated to have diverged at least 16 Ma from its continental sister Gliricidia. This study reveals that sequences from the ITS region are amenable to exhaustive taxon sampling because of the high levels of variation at and below the species level. The evolutionary substitution rate for the ITS region is estimated at 3.1±3.5 x 1029 substitutions/site/year, approximately an order of magnitude faster than that estimated for the matK locus. The tribe Robinieae comprises 12 genera concen- graphic analysis that included an evolutionary rates trated in the deserts and seasonally dry tropical forests analysis. Robinioids are represented in the fossil re- of North America (Lavin and Sousa 1995). In this cord throughout North America, and to some extent study, we distinguish the robinioid legumes from the Europe, by Late Eocene to Pliocene wood and leaf tribe Robinieae so as to exclude the genus Sesbania. samples (Matten et al. 1977; Wheeler and Landon 1992; Phylogenetic evidence from chloroplast matK sequenc- Page 1993; Wheeler 2001). The early continental distri- es ®rmly place the robinioids in a monophyletic clade bution of these well characterized fossils, as well as along with Sesbania and the genera of the tribes Loteae recent advances in Bayesian likelihood and evolution- and Coronilleae. Collectively this trichotomous clade is ary rates analysis, provides an opportunity to con®- sister to the very large legume clade marked by the dently estimate the ages of insular and continental ro- loss of the chloroplast DNA inverted repeat, or the in- binioid diversi®cations. In particular, these include two verted-repeat-lacking clade (IRLC; Wojciechowski, in Greater Antillean robinioid genera, Hebestigma and Po- press; Wojciechowski et al. 2000). itea, the ages of which were estimated previously (Lav- Although three monographs detailing the relation- in et al. 2001b) but with parsimony methods and nom- inal consideration of apomorphic traits shared between ships and circumscriptions of robinioid genera have the fossil and extant taxa. been produced (Lavin 1988; Lavin 1993; Lavin and Sousa 1995), accumulating DNA sequence and mor- phological data suggest that some of these genera and MATERIALS AND METHODS their relationships are in need of reinvestigation. Al- Taxon Sampling. All 12 genera and all constituent species tra- though the problematic taxonomies are con®ned to two ditionally classi®ed in the tribe Robinieae, excepting those of Ses- robinioid genera, Coursetia and Gliricidia, a comprehen- bania, have been sampled for morphological data (Appendix A and B). Such data have been derived in large part from a monographic sive reanalysis has been undertaken for the ®rst time treatment of the tribe (Lavin and Sousa 1995) and two of the larger at the species level for all robinioid legumes. Such ex- constituent genera, Coursetia (Lavin 1988) and Poitea (Lavin 1993). haustive sampling enabled the detection of a root for All robinioid genera and most constituent species have been sam- pled for nuclear ribosomal internal transcribed spacers and inter- the robinioid phylogeny, a reevaluation of all generic vening 5.8S sequences (the ITS region; Appendix C). This includes circumscriptions and relationships, and a biogeo- Lennea (all three species sampled), Gliricidia (all three), Hybosema 387 388 SYSTEMATIC BOTANY [Volume 28 (all two), Poitea (12 of 13 taxa), Coursetia (35 of 40 species), Peteria quence data, multiple Bayesian analyses each began with a ran- (three of four), Robinia (all four), and the following monotypic gen- dom tree and the most general nucleotide substitution model, a era: Hebestigma, Olneya, Genistidium,andSphinctospermum. The ITS general time reversible model plus a shape parameter for a gamma region was sampled exhaustively because this provides a means distribution and invariant sites parameter. Parsimony heuristic of identifying potentially problematic paralogs (e.g., Buckler et al. searches on all data sets included 100 random addition replicates, 1997) even though such have never been detected among robinioid tree bisection reconnection branch swapping, and retention of legumes and close relatives. Also, these sequences contain much multiple parsimonious trees. Clade stability tests involved Bayes- informative variation among genera and species, and they are ian posterior probabilities (Huelsenbeck et al. 2001) and parsi- readily ampli®ed and are not problematic to align for robinioid mony bootstrap resampling (Felsenstein 1985). For the latter, each legumes and close relatives (e.g., Lavin et al. 2001a; Lavin et al. of 10,000 non-parametric bootstrap replicates was subjected to a 2001b; Wojciechowski et al. 1993). In addition, exhaustive sampling heuristic search as above, but with one random addition sequence increases overall phylogenetic accuracy (Zwickl and Hillis 2002). and only one tree saved per replicate. For the various data sets, The single exception to the sampling of the morphological data the percentage of data matrix cells scored as missing data are re- and the ITS sequences is the genus Sesbania. Because this genus ported in Table 1. has been shown to be a distinct lineage from the robinioid le- Evolutionary Rates Analysis. The program r8s (Sanderson gumes (Wojciechowski, in press; Wojciechowski et al. 2000), Ses- 2001) was used to assess variance in evolutionary substitution bania has been sampled as one of the designated outgroups in this rates for nucleotide sequences from the ITS region and the matK analysis. locus, and incorporate such variance into the estimation of ages The chloroplast trnL intron and the matK loci were more dis- of lineages (Sanderson 1997, 1998, 2001, 2002). This program uses criminately sampled for selected species to verify certain of the a rate smoothing approach, penalized likelihood (PL), to identify ®ndings derived from the analysis of the morphological data and an optimal rate smoothing parameter that renders evolutionary ITS region. The matK locus was sampled to verify ®ndings at high- substitution rates and ages for each of the branches in a phylogeny. er levels, particularly the root of the robinioid phylogeny. For these The optimal smoothing parameter is determined by a cross vali- chloroplast loci, sampling was performed for at least one species dation approach whereby the value chosen best predicts the over- from each of the robinioid genera (Hebestigma, Lennea, Gliricidia, all terminal branch lengths in a saturated rate model. This predic- Hybosema, Poitea, Coursetia, Genistidium, Olneya, Peteria, Robinia,and tive ability is then compared with that of an autocorrelated rate Sphinctospermum) and designated outgroups (Sesbania, Anthyllis, smoothing approach (nonparametric rate smoothing; Sanderson Lotus, Ornithopus,andSecurigera). 1997) and rate constant model (Langley and Fitch 1974), which The designated outgroups in this study include representatives potentially de®ne the extremes of the continuum from the satu- of two monophyletic clades that are the closest relatives of the rated to the clock-like rates model. These latter two approaches robinioid clade, the genus Sesbania and the Loteae-Coronilleae trib- are also implemented in r8s. al alliance. Five species of the Loteae-Coronilleae alliance for For the ITS and matK data sets, branch lengths were estimated which