387 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. WOJCIECHOWSKI, 2 PETER GASSON, 3 COLIN HUGHES, 4 and ELISABETH WHEELER 5 1 Department of Plant Sciences, Montana State University, Bozeman, Montana 59717 ([email protected]); 2 Department of Plant Biology, Arizona State University, Tempe, Arizona 85287; 3 Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3AB, UK; 4 Department of Plant Sciences, University of Oxford, South Parks Road, Oxford, OX1 3RB, UK; 5 Department of Wood and Paper Science, North Carolina State University, Raleigh, NC 27695; 6 Author 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, and the genus Poissonia is resurrected to accommodate four morphologically disparate species previously classified 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 continentalsister 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 10 29 substitutions/site/year, approximately an order of magnitude faster than that estimated for the matK locus. The tribe Robinieae comprises 12 genera concen- trated in the deserts and seasonally dry tropical forests of North America (Lavin and Sousa 1995). In this study, we distinguish the robinioid legumes from the tribe Robinieae so as to exclude the genus Sesbania. Phylogenetic evidence from chloroplast matK sequenc- es firmly place the robinioids in a monophyletic clade along with Sesbania and the genera of the tribes Loteae and Coronilleae. Collectively this trichotomous clade is sister to the very large legume clade marked by the loss of the chloroplast DNA inverted repeat, or the in- verted-repeat-lacking clade (IRLC; Wojciechowski, in press; Wojciechowski et al. 2000). Although three monographs detailing the relation- ships and circumscriptions of robinioid genera have been produced (Lavin 1988; Lavin 1993; Lavin and Sousa 1995), accumulating DNA sequence and mor- phological data suggest that some of these genera and their relationships are in need of reinvestigation. Al- though the problematic taxonomies are confined to two robinioid genera, Coursetia and Gliricidia, a comprehen- sive reanalysis has been undertaken for the first time at the species level for all robinioid legumes. Such ex- haustive sampling enabled the detection of a root for the robinioid phylogeny, a reevaluation of all generic circumscriptions and relationships, and a biogeo- graphic analysis that included an evolutionary rates analysis. Robinioids are represented in the fossil re- cord throughout North America, and to some extent Europe, by Late Eocene to Pliocene wood and leaf samples (Matten et al. 1977; Wheeler and Landon 1992; Page 1993; Wheeler 2001). The early continental distri- bution of these well characterized fossils, as well as recent advances in Bayesian likelihood and evolution- ary rates analysis, provides an opportunity to confi- dently estimate the ages of insular and continental ro- binioid diversifications. In particular, these include two Greater Antillean robinioid genera, Hebestigma and Po- itea, the ages of which were estimated previously (Lav- in et al. 2001b) but with parsimony methods and nom- inal consideration of apomorphic traits shared between the fossil and extant taxa. MATERIALS AND METHODS Taxon Sampling. All 12 genera and all constituent species tra- ditionally classified in the tribe Robinieae, excepting those of Ses- bania, have been sampled for morphological data (Appendix A and B). Such data have been derived in large part from a monographic treatment of the tribe (Lavin and Sousa 1995) and two of the larger constituent genera, Coursetia (Lavin 1988) and Poitea (Lavin 1993). All robinioid genera and most constituent species have been sam- pled for nuclear ribosomal internal transcribed spacers and inter- vening 5.8S sequences (the ITS region; Appendix C). This includes Lennea (all three species sampled), Gliricidia (all three), Hybosema
23
Embed
Phylogeny of Robinioid Legumes (Fabaceae) Revisited ...mfwojci/pdfs/LavinetalSYSBOT2003.pdf · Phylogeny of Robinioid Legumes (Fabaceae) Revisited: Coursetia and ... (S. vesicaria).
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
387
Systematic Botany (2003), 28(2): pp. 387–409q Copyright 2003 by the American Society of Plant Taxonomists
Phylogeny of Robinioid Legumes (Fabaceae) Revisited: Coursetia andGliricidia Recircumscribed, and a Biogeographical Appraisal of the
Caribbean Endemics
MATT LAVIN,1,6 MARTIN F. WOJCIECHOWSKI,2 PETER GASSON,3 COLIN HUGHES,4 andELISABETH 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 intronand matK locus were sampled from robinioid legumes to infer phylogenetic relationships. The monophyletic robinioid cladeincludes 11 genetically and often morphologically distinct subclades ranked as genera with the following well supportedhigher 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 withGliricidia, and the genus Poissonia is resurrected to accommodate four morphologically disparate species previously classifiedin 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 froma rate-smoothed Bayesian likelihood approach on sequences from the ITS region and the matK locus. Time constraints arederived from the Tertiary fossil wood species Robinia zirkelii, which shares apomorphic wood characters with the Robiniastem clade. The Cuban endemic Hebestigma is estimated to have diverged at least 38 Ma from its Mesoamerican sister genusLennea, 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 levelsof 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-trated in the deserts and seasonally dry tropical forestsof North America (Lavin and Sousa 1995). In thisstudy, we distinguish the robinioid legumes from thetribe Robinieae so as to exclude the genus Sesbania.Phylogenetic evidence from chloroplast matK sequenc-es firmly place the robinioids in a monophyletic cladealong with Sesbania and the genera of the tribes Loteaeand Coronilleae. Collectively this trichotomous clade issister to the very large legume clade marked by theloss of the chloroplast DNA inverted repeat, or the in-verted-repeat-lacking clade (IRLC; Wojciechowski, inpress; Wojciechowski et al. 2000).
Although three monographs detailing the relation-ships and circumscriptions of robinioid genera havebeen produced (Lavin 1988; Lavin 1993; Lavin andSousa 1995), accumulating DNA sequence and mor-phological data suggest that some of these genera andtheir relationships are in need of reinvestigation. Al-though the problematic taxonomies are confined to tworobinioid genera, Coursetia and Gliricidia, a comprehen-sive reanalysis has been undertaken for the first timeat the species level for all robinioid legumes. Such ex-haustive sampling enabled the detection of a root forthe robinioid phylogeny, a reevaluation of all genericcircumscriptions and relationships, and a biogeo-
graphic analysis that included an evolutionary ratesanalysis. Robinioids are represented in the fossil re-cord throughout North America, and to some extentEurope, by Late Eocene to Pliocene wood and leafsamples (Matten et al. 1977; Wheeler and Landon 1992;Page 1993; Wheeler 2001). The early continental distri-bution of these well characterized fossils, as well asrecent advances in Bayesian likelihood and evolution-ary rates analysis, provides an opportunity to confi-dently estimate the ages of insular and continental ro-binioid diversifications. In particular, these include twoGreater Antillean robinioid genera, Hebestigma and Po-itea, the ages of which were estimated previously (Lav-in et al. 2001b) but with parsimony methods and nom-inal consideration of apomorphic traits shared betweenthe fossil and extant taxa.
MATERIALS AND METHODS
Taxon Sampling. All 12 genera and all constituent species tra-ditionally classified in the tribe Robinieae, excepting those of Ses-bania, have been sampled for morphological data (Appendix A andB). Such data have been derived in large part from a monographictreatment of the tribe (Lavin and Sousa 1995) and two of the largerconstituent genera, Coursetia (Lavin 1988) and Poitea (Lavin 1993).All robinioid genera and most constituent species have been sam-pled for nuclear ribosomal internal transcribed spacers and inter-vening 5.8S sequences (the ITS region; Appendix C). This includesLennea (all three species sampled), Gliricidia (all three), Hybosema
388 [Volume 28SYSTEMATIC BOTANY
(all two), Poitea (12 of 13 taxa), Coursetia (35 of 40 species), Peteria(three of four), Robinia (all four), and the following monotypic gen-era: Hebestigma, Olneya, Genistidium, and Sphinctospermum. The ITSregion was sampled exhaustively because this provides a meansof identifying potentially problematic paralogs (e.g., Buckler et al.1997) even though such have never been detected among robinioidlegumes and close relatives. Also, these sequences contain muchinformative variation among genera and species, and they arereadily amplified and are not problematic to align for robinioidlegumes and close relatives (e.g., Lavin et al. 2001a; Lavin et al.2001b; Wojciechowski et al. 1993). In addition, exhaustive samplingincreases overall phylogenetic accuracy (Zwickl and Hillis 2002).The single exception to the sampling of the morphological dataand the ITS sequences is the genus Sesbania. Because this genushas been shown to be a distinct lineage from the robinioid le-gumes (Wojciechowski, in press; Wojciechowski et al. 2000), Ses-bania has been sampled as one of the designated outgroups in thisanalysis.
The chloroplast trnL intron and the matK loci were more dis-criminately sampled for selected species to verify certain of thefindings derived from the analysis of the morphological data andITS region. The matK locus was sampled to verify findings at high-er levels, particularly the root of the robinioid phylogeny. For thesechloroplast loci, sampling was performed for at least one speciesfrom each of the robinioid genera (Hebestigma, Lennea, Gliricidia,Hybosema, Poitea, Coursetia, Genistidium, Olneya, Peteria, Robinia, andSphinctospermum) and designated outgroups (Sesbania, Anthyllis,Lotus, Ornithopus, and Securigera).
The designated outgroups in this study include representativesof two monophyletic clades that are the closest relatives of therobinioid clade, the genus Sesbania and the Loteae-Coronilleae trib-al alliance. Five species of the Loteae-Coronilleae alliance forwhich at least matK sequences were available have been incorpo-rated into this analysis. These include Anthyllis vulneraria, Securi-gera varia, Lotus unifoliolatus, Lotus japonicus, and Ornithopus com-pressus. According to the phylogeny of Allan and Porter (2000),Anthyllis and Securigera represent the basal-most branching lineageof the Loteae-Coronilleae clade, whereas Ornithopus represents amore distally branching clade very closely related to species tra-ditionally placed in the genus Lotus. Five species were selected thatrepresent the major subgenera of Sesbania (Lavin and Sousa 1995):the most speciose subgenus Sesbania (Old World S. cannabina andS. tomentosa, and New World S. emerus), subgenus Daubentonia (S.drummondii), subgenus Agati (S. grandiflora), and subgenus Glotti-dium (S. vesicaria). These outgroup species were sampled especiallyfor sequences from the ITS region and the matK locus, but morediscriminately for morphological and trnL intron variation becausethese last two sources of data were used mostly for addressingquestions at phylogenetic levels within the robinioid legumes (Ap-pendix C).
Sequence Data and Analysis. DNA isolations, PCR amplifica-tions, and template purifications were performed with Qiagen Kits(i.e., DNeasy Plant Mini Kit, Taq PCR Core Kit, QIAquick PCRPurification Kit; Qiagen, Santa Clarita CA). PCR primers and am-plification conditions for the ITS region, the trnL intron, and thematK locus are described in Beyra-M. and Lavin (1999), Lavin etal. (2000), and Lavin et al. (2001a and 2001b). Sequencing productswere run on an ABI 377 and 3700 automated sequencer at North-woods DNA (Becida, Minnesota), Davis Sequencing (Davis, Cali-fornia), DNA Sequencing and Synthesis Facility (Ames, Iowa), andThe ASU-DNA Laboratory (Tempe, Arizona).
DNA sequences were aligned manually with the aid of Se-Al(Rambaut 1996). Multiple manual alignments were continuouslyanalyzed during the development of the various data sets to iden-tify those clades consistently resolved regardless of taxon com-position or sequence alignment of the data set. Individual datasets were developed for sequence data from the ITS region, thetrnL intron, and the matK locus, morphological characters. Com-bined data sets were developed for all possible combinations ofthese four individual data sets. Maximum likelihood and parsi-mony analyses were performed with MrBayes (Huelsenbeck andRonquist 2001) and PAUP* (Swofford 2001). For nucleotide se-
quence data, multiple Bayesian analyses each began with a ran-dom tree and the most general nucleotide substitution model, ageneral time reversible model plus a shape parameter for a gammadistribution and invariant sites parameter. Parsimony heuristicsearches on all data sets included 100 random addition replicates,tree bisection reconnection branch swapping, and retention ofmultiple parsimonious trees. Clade stability tests involved Bayes-ian posterior probabilities (Huelsenbeck et al. 2001) and parsi-mony bootstrap resampling (Felsenstein 1985). For the latter, eachof 10,000 non-parametric bootstrap replicates was subjected to aheuristic search as above, but with one random addition sequenceand only one tree saved per replicate. For the various data sets,the percentage of data matrix cells scored as missing data are re-ported in Table 1.
Evolutionary Rates Analysis. The program r8s (Sanderson2001) was used to assess variance in evolutionary substitutionrates for nucleotide sequences from the ITS region and the matKlocus, and incorporate such variance into the estimation of agesof lineages (Sanderson 1997, 1998, 2001, 2002). This program usesa rate smoothing approach, penalized likelihood (PL), to identifyan optimal rate smoothing parameter that renders evolutionarysubstitution rates and ages for each of the branches in a phylogeny.The optimal smoothing parameter is determined by a cross vali-dation approach whereby the value chosen best predicts the over-all terminal branch lengths in a saturated rate model. This predic-tive ability is then compared with that of an autocorrelated ratesmoothing approach (nonparametric rate smoothing; Sanderson1997) and rate constant model (Langley and Fitch 1974), whichpotentially define the extremes of the continuum from the satu-rated to the clock-like rates model. These latter two approachesare also implemented in r8s.
For the ITS and matK data sets, branch lengths were estimatedduring a maximum likelihood analysis that involved a search oftree parameter space using a Bayesian approach. This involves aMetropolis-coupled Markov Chain Monte Carlo permutation oftree parameters, an initial random tree, 1,000,000 permutations oftree parameters, and four chains (Huelsenbeck and Ronquist, 2001;Huelsenbeck et al. 2001). Parsimony searches in PAUP* (Swofford2001) were used to validate the branching order parameter esti-mated with MrBayes, whereas the AIC model selection approachwas used to validate the estimated nucleotide substitution param-eters (Posada and Crandall 1998). All estimated parameters(branching order, branch lengths, and nucleotide substitution)were then used to generate parametric bootstrap replicates usingSeq-Gen (Rambaut and Grassley 2001), each of which were sub-jected to analyses with r8s to obtain mean and standard deviationsof evolutionary substitution rates and ages of specified clades.
Relative substitution rates and ages estimated with r8s wereconverted to absolute rates and ages by enforcing two time con-straints. One time constraint was derived from an evolutionaryrates analysis of matK sequences for all Fabaceae (Wojciechowski,in press; Wojciechowski et al., in mss.). This large-scale analysisuses an age constraint for the legume crown clade of 59.9 Ma,which represents the age of the oldest unequivocal legume fossil(Herendeen et al. 1992 and personal communication). From this,the maximum age estimate for the robinioid crown clade (i.e., themost recent common ancestor of Hebestigma and Robinia) is 45 Ma.A second constraint was derived from the robinioid fossil record,where Tertiary North American fossil wood samples of Robiniazirkellii (Platen) Matten, Gastaldo, and Lee range in age from LateEocene to Pliocene (Matten et al. 1977; Wheeler and Landon 1992;Page 1993; Wheeler 2001). The apomorphic traits displayed bythese fossils place it clearly on the Robinia stem clade, which there-fore must have a minimum age of Late Eocene, or 33.7 Ma follow-ing Berggren et al. (1995). These wood characters are detailed inthe discussion.
RESULTS
ITS Data Set. The parsimony analysis of sequencesfrom the ITS region yielded 10,000 trees (Table 1). Thestrict consensus resolves clades that are well support-
2003] 389LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
TAB
LE
1.D
ata
set
and
tree
stat
istic
sfo
rth
ese
para
tean
dco
mbi
ned
anal
yses
.PI
5pa
rsim
ony
info
rmat
ive,
CI
5co
nsis
tenc
yin
dex,
RI
5re
tent
ion
inde
x,an
dPH
T5
part
ition
hom
ogen
eity
test
.Th
ep
valu
esfo
rth
ePH
Tw
ere
deri
ved
from
1000
rand
ompa
rtiti
ons,
each
anal
yzed
with
only
info
rmat
ive
char
acte
rs(a
ctua
lpar
titio
nsar
ein
dica
ted
inth
em
ostl
eft-
hand
colu
mn;
NA
5no
tapp
licab
le).
Boot
stra
pva
lues
are
give
nin
pare
nthe
ses
inla
stfo
urco
lum
ns.I
nal
lana
lyse
s,th
em
axim
umnu
mbe
rof
tree
sw
asse
tat
10,0
00.
Dat
ase
tTa
xa
Tota
lch
arac
ters
(%m
issi
ng)
PIch
arac
ters
Tre
esL
engt
hPH
T(p
)C
IR
I
Heb
esti
gma-
Lenn
eacl
ade
Glir
icid
ia-
Hyb
osem
acl
ade
Poite
a-G
liric
idia
clad
ePo
isso
nia
clad
e
nrD
NA
ITS1
,5.
8S,I
TS2
cpD
NA
trnL
intr
on&
inde
lscp
DN
Am
atK
178 53 37
737
(0.6
)68
2(1
.0)
1557
(2.8
)
435
111
291
10,0
00
10,0
00 180
2003 247
660
0.20
9
0.47
4
NA
0.44
7
0.79
1
0.81
1
0.90
1
0.91
6
0.89
8
basa
lmos
tcl
ade
(88%
)no
tre
solv
ed
not
reso
lved
mon
ophy
letic
(86%
)m
onop
hyle
tic(7
3%)
not
reso
lved
mon
ophy
letic
(100
%)
not
reso
lved
mon
ophy
letic
(100
%)
mon
ophy
letic
(94%
)m
onop
hyle
tic(6
0%)
mon
ophy
letic
(71%
)m
orph
olog
y
mor
phol
ogy
and
ITS
regi
onal
lda
tase
ts
82 82 26
40(6
.7)
777
(0.4
)30
17(0
.7)
40 400
554
10,0
00
3,36
0 2
104
1639
1743
NA
0.03
9
0.28
8
0.45
2
0.48
1
0.69
0
0.90
7
0.82
9
0.75
2
not
reso
lved
basa
lmos
tcl
ade
(88%
)ba
salm
ost
clad
e(9
4%)
not
reso
lved
mon
ophy
letic
(79%
)m
onop
hyle
tic(9
7%)
mon
ophy
letic
(62%
)m
onop
hyle
tic(1
00%
)m
onop
hyle
tic(1
00%
)
not
reso
lved
mon
ophy
letic
(100
%)
mon
ophy
letic
(100
%)
ed by bootstrap analysis (Figs. 1–3). Unambiguous andphylogenetically informative indels were few in the ITSregion and the few that were detected correspondedwith already well supported clades derived from nu-cleotide substitutions (e.g., the genus Lennea, or sub-clades within Poitea). The clade including Hebestigmaand Lennea is sister to the rest of the robinioids, andGliricidia and Hybosema, the latter depicted as G. robus-tum and G. ehrenbergii in Fig. 1, form the well-sup-ported sister clade of Poitea. The robinioid genera bear-ing a style with a pollen brush, Olneya, Robinia, Peteria,Genistidium, Sphinctospermum, and Coursetia, form afairly well supported monophyletic clade (Fig. 2). OnlyCoursetia in this group of six genera is not resolved asmonophyletic (Figs. 2–3). The Coursetia clade contain-ing most species of the genus (Fig. 3) is weakly re-solved as sister to a clade comprising Peteria, Sphinc-tospermum, and Genistidium (Figs. 2–3). The other Cour-setia clade comprises four species, Coursetia hypoleuca,C. orbicularis, C. heterantha, and C. weberbaueri, depictedas Poissonia in Fig. 2, that form the moderately sup-ported sister clade of Robinia.
trnL Data Set. Analysis of the trnL locus did notshow much variation among robinioid legumes andwas thus not sampled as thoroughly as the ITS region.Parsimony analysis of the trnL data set (670 alignedsites plus 12 shared indel characters) yielded 10,000equally minimal length trees (Table 1). Branch supportfor the relatively few resolved clades was generallylow. Exceptions included the clade comprising Robinia,Olneya, Peteria, Genistidium, Sphinctospermum, and Cour-setia, which was resolved at 94% support. The Gliricidiaclade (including Hybosema) was resolved with 73%bootstrap support, and the Poissonia clade comprisingCoursetia hypoleuca, C. orbicularis, C. heterantha, and C.weberbaueri was resolved as monophyletic with 60%bootstrap support. Notably, there was no conflictamong any of the clades supported at 60% or moreand those that were detected during the analysis ofsequences from the ITS region.
matK Data Set. Parsimony analysis of the matK lo-cus yielded 180 equally minimal length trees (Table 1).The Poitea-Gliricidia clade and that comprising Robinia,Olneya, Poissonia, Peteria, Genistidium, Sphinctospermum,and Coursetia were very well supported by high boot-strap values of over 96% (Fig. 4). Hebestigma and Lenneawere resolved as a clade and as sister to the rest ofrobinioids in most of the minimal length trees, but thestrict consensus placed these two in an unresolved po-sition at the base of the robinioid clade. Relationshipsresolved during the analysis of matK sequences did notconflict with any of those hypothesized by the ITS re-gion or the trnL intron sequence data.
Morphological Data Set. Analysis of 40 morpho-logical traits did not resolve many well supported re-lationships among the robinioid legumes (Table 1). The
390 [Volume 28SYSTEMATIC BOTANY
FIG. 1. Strict consensus of 10,000 (5maxtrees) equally most parsimonious trees derived from the analysis of the ITS region.The 178 sequences comprise 737 aligned sites, 435 of which are parsimony informative. The maximum parsimony trees havea length of 2003, a consistency index of 0.447, and a retention index of 0.901 (Table 1). Values above the nodes are parsimonybootstrap values. The species classified as the genus Gliricidia in this study are shown inside the box. The genus Lennea andthe genera depicted in Figs. 2–3 are shown marked by states of character 15, which are described in Appendix A.
2003] 391LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
FIG. 2. Continuation of Fig. 1 showing the clade comprising the genera Olneya, Robinia, Poissonia, Peteria, Genistidium, Sphinc-tospermum, and Coursetia, genera traditionally grouped as the barbistyled genera because of the shared pollen brush on thestyle. Values above the nodes are parsimony bootstrap values. The species classified as the genus Poissonia in this study areshown inside the box. A subgroup of Poissonia is marked by a post-pollination floral resupination syndrome and a longpedunculate inflorescence (character #’s 20 and 22 in Appendix A).
392 [Volume 28SYSTEMATIC BOTANY
FIG. 3. Continuation of Figs. 1 and 2 showing the clade comprising the species of the genus Coursetia (sensu stricto—asdefined in this study). Values above the nodes are parsimony bootstrap values. Coursetia subclades possessing a post-pollinationfloral resupination syndrome and a long pedunculate inflorescence are indicated (character #‘s 20 and 22 in Appendix A).
robinioids are moderately supported as a monophylet-ic clade (76% bootstrap support) comprising four ma-jor subclades. The genera Hebestigma and Lennea eachrepresent a basal branching clade. Gliricidia, Hybosema,and Poitea comprise a third clade (62% bootstrap sup-
port), and Robinia, Olneya, Peteria, Genistidium, Sphinc-tospermum, and Coursetia a fourth (73% bootstrap sup-port). These are the same four groups that were re-solved in all analyses of the individual molecular datasets. The bulk of the missing entries for this particular
2003] 393LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
FIG. 4. Strict consensus of 180 equally most parsimonious trees derived from the analysis of the matK locus. The 37 sequencescomprise 1557 aligned sites, 291 of which are parsimony informative. This tree has a length of 660, a consistency index of0.811, and a retention index of 0.898. Values above the nodes are parsimony bootstrap values. The species classified as thegenera Gliricidia and Poissonia in this study are shown inside the boxes.
394 [Volume 28SYSTEMATIC BOTANY
data set came from the wood anatomical characters,because many species have yet to be studied for woodtraits.
Combined Data Sets. A partition homogeneity testrevealed no significant conflict among any of the pos-sible data partitions, except between the sequencesfrom the ITS region and the morphological data set(Table 1). Other than the Hebestigma-Lennea clade, noconflict was detected among the these two data setsduring individual data set analyses, if only because allthe well supported nodes detected in one analysiswith bootstrap values well over 60% were also re-solved with the other data set. Apparent conflict de-tected in the combined morphology and ITS data setwas the result of the many missing entries in the woodcharacters. Omitting wood characters during this testsuggested no conflict between the morphological andITS data sets.
Just two of the combined data sets will be presentedhere because they capture the essence of results for allpossible data combinations. The first is the combinedmorphological (40 characters) and ITS region (737aligned sites) data set. Because of the exhaustive sam-pling of these two data sets at the species level, a com-bined analysis allows for the most comprehensivecomparisons. Analysis of the combined ITS-morpho-logical data set yielded 3,360 equally parsimonioustrees (Table 1). Only the Sesbania species were used asoutgroups because of the difficulty of scoring homol-ogous morphological characters among the robinioidsand species of the Loteae-Coronilleae alliance.
The only substantive differences between the anal-ysis of this combined data set and that of the ITS re-gion alone was that branch support increased signifi-cantly in a few clades. Most notably, the genus Cour-setia (sensu stricto) showed an increase from 92% inthe ITS analysis to 98% in the combined analysis. Al-though, the combined analysis of the ITS region andthe morphological data sets yielded very similar re-sults to the individual analyses of constituent data sets,combined data allowed for mapping the evolution ofthe morphological characters that have most heavilyinfluenced previous taxonomies of the robinioid genera(see discussion).
The second combined analysis involved all of themolecular data sets in combination with the morpho-logical one. Because of limited sampling of the matKlocus, this analysis included only 23 ingroup terminaltaxa (Table 1; Fig. 5). One significant finding was thatHebestigma and Lennea form a well supported cladethat is resolved as one of three basal branches in therobinioid diversification. A second major subclade inthe robinioid diversification is the clade comprising Po-itea, Gliricidia, and Hybosema, and a third includes Ro-binia, Poissonia, Olneya, Peteria, Genistidium, Sphinctos-permum, and Coursetia. In both of the combined anal-
yses, samples of the two hybosemas (depicted as Glir-icidia robustum and G. ehrenbergii in Figs. 4 and 5) arenested within Gliricidia, and Coursetia includes twomajor sublineages (Coursetia sensu stricto and Poisson-ia). Coursetia is resolved with weak support as a closerelative of Genistidium, Peteria, and Sphinctospermum,whereas Poissonia is weakly resolved as the sister toRobinia. This latter clade comprises Coursetia hypoleuca,C. orbicularis, C. weberbaueri, and C. heterantha (depictedas Poissonia in Fig. 5), four South American species con-fined to the Andes in southern Peru, Bolivia, andnorthern Argentina.
Evolutionary Rates Analysis. A likelihood ratiotest rejected models with a uniform rate across allbranches for both the ITS and matK phylogenies (ITSregion: LR5182.87, df577, p50.000000; matK:LR5345.56, df527, p50.000000). With the root of eachof the ITS and matK robinioid phylogenies constrainedto a maximum of 45 Ma, and that of the Robinia stemclade set to 33.7 Ma, the penalized likelihood (PL) andrate constant (LF) methods of rate smoothing yieldedvery similar age and rate estimates (Tables 2–3). Au-tocorrelated rate smoothing (NPRS) generally pro-duced faster rates or older age estimates than PL andLF for sequences from both the ITS region and matKlocus. Sequences from the matK locus yielded slightlyolder age estimates than sequences from the ITS region(compare clades A–D in Tables 2–3; Figs. 6–7). Re-gardless, the estimated substitution rates using PL fallwithin a very close range, suggesting a high degree ofconstancy in substitution rate. The PL estimates areuniformly 3.9 x 10210 substitution/site/year for matKand 3.1–3.5 x 1029 substitutions/site/year for the ITSregion (Tables 2–3).
The two vicariant Caribbean clades, Hebestigma-Len-nea (crown clade A in Fig. 6, Table 2) and Poitea-Gliri-cidia (crown clade B in Figs. 6–7, Tables 2–3) are esti-mated to be at least 38 and 16 Ma in age, respectively.The Poitea diversification (crown clade B2, Table 2, Fig.6), estimated at least 9 Ma in age, is at least as old asits mainland sister diversification, Gliricidia (crownclade B1, Table 2, Fig. 6). Andean South Americanclades occur within the robinioid diversification, andinvolve only the species of Poissonia (crown clade C inFig. 6), Gliricidia brenningii (crown clade B1), and sub-clades within Coursetia (all within crown clade D1).The crown clade Poissonia (clade C in Figs. 6–7, Tables2–3) is at least 18 Ma in age, which is at least as old orolder than the other South American diversifications(crown clade D1 in Fig. 6, Table 2). The youngest cladewithin the robinioid diversification that contains bothNorth and South American species is that of involvingCoursetia caribaea and close relatives (crown clade D2in Fig. 6, Table 2), which is estimated at around 5 Main age.
2003] 395LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
FIG. 5. The strict consensus of two equally most parsimonious trees derived from the combined data set comprising 26species represented by sequences from the combined ITS region (737 aligned sites), the trnL intron (670 aligned sites plus 12indel characters), and the matK locus (1557 aligned sites plus one indel character), as well as 40 morphological characters. Ofthese 3017 characters, 554 were parsimony informative. This tree has a length of 1743, a consistency index of 0.690, and aretention index of 0.752 (Table 1). Values above the nodes are parsimony bootstrap values. Species classified in the generaGliricidia and Poissonia in this study are shown inside the boxes.
396 [Volume 28SYSTEMATIC BOTANY
TABLE 2. Estimated ages (Ma) and rates (per site per Ma—below the age estimate) of selected clades from the rate smoothed Bayesianlikelihood trees derived from sequences from the ITS region. The root of the robinioid crown clade was calibrated at a maximum of 45Ma and the Robinia stem clade calibrated at 33.7 Ma (Fig. 6). The rate smoothing methods are PL 5 penalized likelihood, LF 5 rateconstant, and NPRS 5 nonparametric rate smoothing. Means and standard deviations were estimated with a rate smoothing analysisof 100 Bayesian likelihood trees, each of which was derived from a parametric bootstrapped data set.
Crown clade PL (45) LF (45) NPRS (45)
A
B
B1
38.3 6 3.80.0031 6 0.0012
16.3 6 4.30.0034 6 0.0004
8.2 6 1.60.0035 6 0.0002
38.4 6 3.80.0035 6 0.0002
16.3 6 4.30.0035 6 0.0002
8.2 6 1.60.0035 6 0.0002
37.9 6 4.10.0055 6 0.0022
26.4 6 3.90.0060 6 0.0011
20.8 6 4.10.0029 6 0.0007
B2
C
D
9.3 6 1.20.0035 6 0.0002
18.1 6 2.10.0034 6 0.0002
19.8 6 2.50.0034 6 0.0002
9.3 6 1.20.0035 6 0.0002
17.7 6 2.10.0035 6 0.0002
19.6 6 2.40.0035 6 0.0002
18.4 6 2.80.0044 6 0.0006
25.8 6 2.20.0036 6 0.0007
28.2 6 3.00.0046 6 0.0008
D1
D2
15.7 6 1.80.0034 6 0.0002
4.8 6 0.90.0035 6 0.0002
15.6 6 1.80.0035 6 0.0002
4.8 6 0.90.0035 6 0.0002
23.8 6 2.90.0037 6 0.0006
8.1 6 1.90.0026 6 0.0006
TABLE 3. Estimated ages (Ma) and rates (per site per Ma—below the age estimate) of selected clades from the rate smoothed Bayesianlikelihood trees derived from sequences from the matK locus. The root of the robinioid crown clade was calibrated at a maximum of45 Ma and the Robinia stem clade calibrated at 33.7 Ma (Fig. 7). The rate smoothing methods are PL 5 penalized likelihood, LF 5 rateconstant, and NPRS 5 nonparametric rate smoothing. The estimated age of the crown clade A (the Hebestigma-Lennea clade; Fig. 6) didnot differ significantly from the fixed age of the robinioid crown clade (45 Ma). Means and standard deviations were estimated with arate smoothing of 100 Bayesian likelihood trees, each of which was derived from a parametric bootstrapped data set.
Crownclade PL LF NPRS
B
C
D
18.0 6 2.40.00039 6 0.00003
22.3 6 3.10.00039 6 0.00003
23.2 6 3.30.00039 6 0.00003
17.9 6 2.40.00039 6 0.00003
22.3 6 3.00.00039 6 0.00003
23.2 6 3.30.00039 6 0.00003
17.0 6 3.20.00059 6 0.00011
22.1 6 3.50.00048 6 0.00008
25.2 6 3.10.00042 6 0.00007
DISCUSSION
The results of this analysis bear on two general ar-eas. The first involves the taxonomies of the generaGliricidia and Coursetia. The evolution of certain dis-tinctive morphological traits that have been used todiagnose these genera (e.g., Lavin 1988; Lavin and Sou-sa 1995) are now at odds with the findings derivedfrom molecular data. The second involves the Carib-bean endemic genera, Hebestigma and Poitea. With theroot of the robinioid phylogeny firmly established forthe first time with multiple data sets, it is clear thatthis primarily North American tropical radiation in-cludes two independent cases of Caribbean and con-tinental vicariant sister clades: Hebestigma (Cuba) andLennea (Mesoamerica), and Poitea (Greater Antilles andDominica) and Gliricidia (Mesoamerica but with onespecies from Ecuador and adjacent Peru).
The Genus Gliricidia. The Gliricidia and Poitea phy-logenies derived from the sequence analysis of the ITSregion are highly congruent with previously published
phylogenies of these genera, which were derived fromcpDNA restriction site data (e.g., Lavin et al. 1991; Lav-in 1993). As such, the genus Gliricidia is herein circum-scribed to include five species: Gliricidia brenningii, G.ehrenbergii, G. maculata, G. robustum, and G. sepium. Thegenus is not diagnosed by morphological characters,but the taxonomic keys to genera and species providedby Lavin and Sousa (1995) remain valid as long as Hy-bosema species are placed within Gliricidia. That Hy-bosema is nested within a paraphyletic Gliricidia sensustricto is revealed not only by molecular data, but alsoby wood anatomy. For example, wood with septate fi-bers (#35; Table 4, Appendix A and B) and storied rays(character #36) is derived at the base of the clade con-taining Gliricidia robustum, G. ehrenbergii, G. maculata,and G. sepium (storied rays are secondarily lost in G.ehrenbergii). Wood and molecular data are concordantin resolving the South American Gliricidia brenningii assister to the rest of the species of Gliricidia. As nowcircumscribed, Gliricidia is confined to Mexico and ad-
2003] 397LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
FIG. 6. Chronogram derived from penalized likelihood rate smoothing of a consensus Bayesian likelihood tree, which wasestimated with sequences from the ITS region. The 45 Ma maximum age constraint at the basal node is derived from a largescale rates analysis of all legumes (Wojciechowski, in press; Wojciechowski et al., in mss.). The 33.7 Ma minimum age constraintis derived from the fossil wood record (see methods and material, and discussion). The average nucleotide substitution param-eters for 10,000 Bayesian trees at stationarity are r(GT) 5 1.00, r(CT) 5 6.379, r(CG) 5 0.980, r(AT) 5 1.758, R(AG) 5 3.018,r(AC) 5 1.155, pi(A) 5 0.204, pi(C) 0.2715, pi(G) 5 0.293, pi(T) 5 0.233, alpha 5 1.418, iP 5 0.216. See Table 2 for the estimatedages and rates of substitution for the clades marked A, B, C, and D. Values above nodes are Bayesian posterior probabilitiesfor selected clades.
398 [Volume 28SYSTEMATIC BOTANY
FIG. 7. Chronogram derived from penalized likelihood rate smoothing of a consensus Bayesian likelihood tree, which was esti-mated with sequences from the matK locus. The 45 Ma maximum age constraint at the basal node is derived from a large scale ratesanalysis of all legumes (Wojciechowski, in press; Wojciechowski et al., in mss.). The 33.7 Ma minimum age constraint is derived fromthe fossil wood record (see methods and material, as well as discussion). The average nucleotide substitution parameters for 10,000Bayesian likelihood trees at stationarity are r(GT) 5 1.00, r(CT) 5 2.201, r(CG) 5 2.059, r(AT) 5 0.274, R(AG) 5 1.925, r(AC) 5 1.508,pi(A) 5 0.324, pi(C) 5 0.145, pi(G) 5 0.147, pi(T) 5 0.384, SS(1) 5 0.849, SS(2) 5 0.705, SS(3) 5 1.446. See Table 3 for the estimatedages and rates of substitution for the clades marked B, C, and D (which correspond with those marked in Fig. 6). Values above nodesare Bayesian posterior probabilities.
2003] 399LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
TABLE 4. Parsimony tree scores for the 40 morphological characters listed in Appendix A and taken from the analysis of the combinedmorphological and nrDNA ITS/5.8S data set. The range of the length, consistency index, and retention index is over the 3,360 minimallength trees. These morphological characters added a length of 120 to the total 1639 (Table 1). The CI for all 40 morphological charactersis 0.392, and the RI is 0.882, suggesting that the morphological traits are comparable to ITS in resolving relationships (compare to valuesin Table 1).
Character
Length
Minimum Maximum
Consistency index
Best Worst
Retention index
Best Worst
12345
73612
83612
0.1430.3330.3331.0001.000
0.1250.3330.3331.0001.000
0.7690.8950.6361.0001.000
0.7310.8950.6361.0001.000
67890
21134
21134
0.5001.0001.0000.3330.500
0.5001.0001.0000.3330.500
0.8891.0001.0000.9440.923
0.8891.0001.0000.9440.923
1112131415
21214
21214
0.5001.0000.5001.0000.750
0.5001.0000.5001.0000.750
0.5001.0000.7501.0000.971
0.5001.0000.7501.0000.971
161718192021
112343
112343
1.0001.0000.5000.3330.2500.333
1.0001.0000.5000.3330.2500.333
1.0001.0000.9550.9290.8640.926
1.0001.0000.9550.9290.8640.926
2223242526
48523
48524
0.2500.1250.2000.5000.667
0.2500.1250.2000.5000.500
0.8640.8060.6000.9410.969
0.8640.8060.6000.9410.938
2728293031
36611
36611
0.3330.1670.1671.0001.000
0.3330.1670.1671.0001.000
0.9460.8080.7221.0001.000
0.9460.8080.7221.0001.000
3233343536
12333
12333
1.0001.0000.3330.3330.333
1.0001.0000.3330.3330.333
1.0001.0000.8570.6670.500
1.0001.0000.8570.6670.500
37383940
1382
1382
1.0000.3330.1250.500
1.0000.3330.1250.500
1.0000.5000.5000.667
1.0000.5000.5000.667
jacent parts of Central America, with the exception ofG. brenningii, which is from Ecuador and Peru (Lavinand Sousa 1995). The following nomenclature is mod-ified from Lavin and Sousa (1995) to adjust for thisrecircumscription.
Gliricidia H. B. K., Nov. gen. Sp. 6: 393. 1823.—Type:Gliricidia sepium (Jacq.) Steud.
1. Gliricidia robustum (M. Sousa & Lavin) Lavin, comb.nov. Hybosema robustum M. Sousa & Lavin, AnalesInst. Biol. Univ. Nac. Mexico, Ser. Bot. 63: 143.1992.—Type: MEXICO. Chiapas: Canon del Sumi-dero, Martınez S. & Reyes G. 22047 (holotype:MEXU; isotypes: BM, MEXU, MO).
2. Gliricidia ehrenbergii (Schltdl.) Rydb., N. Amer. Fl. 24:239. 1924. Robinia ehrenbergii Schltdl., Linnaea 12:303. 1838. Hybosema ehrenbergii (Schltdl.) Harms,Repert. Spec. Nov. Regni Veg. 19: 66. 1923.—Type:MEXICO. In solo calcareo boream versus ab aquiscalidis pr. Grande, Ehrenberg 645 (holotype: HAL,photo: MONT; isotypes: B, destroyed, HAL, frag-ment of B isotype: F, photos of B isotype: F, G,
400 [Volume 28SYSTEMATIC BOTANY
GH, MEXU, MICH, MO, NY, TEX). See Lavin andSousa (1995) for taxonomic synonyms.
3. Gliricidia brenningii (Harms) Lavin, Syst. Bot. Mon-ogr. 45: 83. 1995. Sesbania brenningii Harms, Re-pert. Spec. Nov. Regni Veg. 19: 68. 1923. Yucara-tonia brenningii (Harms) Burkart, Darwiniana 15:526. 1969.—Type: ECUADOR. Guayaquil, an We-grandern in der Nahe des Rıo Salado, Brenning238 (holotype: B, destroyed, fragment: F, photos:F, NY).
4. Gliricidia maculata (H. B. K.) Steud., Nom. Bot., ed.2, 1: 688. 1840. Robinia maculata H. B. K., Nov. gen.Sp. 6: 393. 1823. Lonchocarpus maculatus (H. B. K.)DC., Prodr. 2: 260. 1825.—Type: MEXICO. Cam-peche: crecit prope Campeche, Humboldt & Bonp-land s.n. (holotype: P-HBK, microfiche IDC 6209.161:II.7).
5. Gliricidia sepium (Jacq.) Steud., Nom. Bot., ed. 2, 1:688. 1840. Robinia sepium Jacq., Enum. Syst. Pl. 28.1760 [Select. stirp. amer. hist. 211, t. 179, f. 101.1763]. Lonchocarpus sepium (Jacq.) DC., Prodr. 2:260. 1825.—Type: COLOMBIA. Cartegena, N. J. Jac-quin s.n. (holotype: not located). See Lavin andSousa (1995) for taxonomic synonyms.
The present circumscription of Gliricidia representsonly a minor modification from previous taxonomies.Indeed, Hybosema ehrenbergii was formally treated as aspecies of Gliricidia until Lavin and Sousa (1995) de-termined that no apomorphic characters diagnosedGliricidia unless G. ehrenbergii (and the then newly de-scribed H. robustum) were removed to the genus Hy-bosema. Phylogenetic analysis of morphological datasuggested that Gliricidia, Hybosema, and Poitea formeda distinct clade but with a trichotomous or unresolvedrelationship (Lavin and Sousa 1995). Indeed, the twohybosemas (Gliricidia ehrenbergii and G. robustum)shared certain morphological apomorphies with otherrobinioid genera. For example, the pseudomonadel-phous staminal column (character 13 in Table 4 andAppendix A and B) was derived among robinioidsonly in Lennea and Hybosema, and the bilabiate calyxwas determined to be derived only in Lennea, Hybose-ma, and Poitea (character 10; Table 4, Appendix A andB). Clearly, molecular data reveal that the hybosemasare derived from within the Gliricidia radiation. It isnotable that in spite of the high degree of floral simi-larity, including the unique banner callus that is cen-trally located on the nectar guide (character 4; Table 4,Appendix A and B), the two hybosemas are separatedfrom each other by long terminal branches and areonly occasionally suggested to be sister species butthen with weak support (compare Figs. 1, 4–7).
The Genus Poissonia. The circumscription of Pois-sonia is revised such that the genus includes four spe-cies, Coursetia hypoleuca, C. orbicularis, C. heterantha, and
C. weberbaueri. Two of these species, Coursetia heteranthaand C. weberbaueri, share a very unusual post-pollina-tion floral resupination syndrome, which results in amature resupinate pod, with certain subclades of Cour-setia sensu stricto (character 20; Figs. 2–3, Table 4, Ap-pendix A and B). Indeed, Lavin (1988) and Lavin andSousa (1995) considered such a trait to be a synapo-morphy of the clade containing Coursetia sections Neo-cracca (monotypic, including just Coursetia heterantha)and section Craccoides. Co-occurring precisely with re-supinate pods is an inflorescence with a long peduncle(character 22; Figs. 2–3, Table 4, Appendix A and B).Both of these traits, previously thought to have evolvedonce at the base of a clade containing sections Neo-cracca and Craccoides, are now considered to have in-dependently evolved four separate times (Table 4, Figs.2–3).
Another remarkable parallelism occurs with the la-trorse pollen brush (state 3 of character 15; see Figs.1–3, Table 4, Appendix A and B). Previously thoughtto have evolved once at the base of the clade containingall the species of Coursetia (Lavin 1988), this characterstate must now be hypothesized to have evolved twice,once each at the bases of the Poissonia and Coursetiaclades. In spite of these extraordinary shared morpho-logical similarities, Coursetia and Poissonia are never re-solved as even weakly supported sister clades in anyother individual or combined analyses.
As now characterized, Poissonia is apomorphicallydiagnosed by leaves bearing orbicular leaflets (char-acter 31; Table 4, Appendix A and B) and seedlingsthat bear two eophylls (character 4; Table 4, AppendixA and B), although the latter has been observed fromonly two of the four species, Poissonia hypoleuca and P.heterantha. Also, petal pigments that are bluish at an-thesis are unique (in the context of robinioids, Sesbania,and Loteae-Coronilleae) to Poissonia hypoleuca, P. orbi-cularis, and P. heterantha. The brick-red petals of Pois-sonia weberbaueri could represent an evolutionary trans-formation away from an originally blue-color state. Al-though these shared vegetative and floral similaritieswere noted for Poissonia hypoleuca, P. orbicularis, and P.heterantha by Lavin (1988) and Lavin and Sousa (1995),blue petal pigments and orbicular leaflets were consid-ered too trivial to be scored as cladistic characters, andtheir occurrence in P. heterantha was thought to be dueto a hybrid origin of this species. With both nuclearand chloroplast sequences showing a close relationshipof these species, such interpretations are no longer con-sidered valid. Regarding Poissonia weberbaueri from Ar-equipa, Peru, no one had previously suspected a closerelationship of this species with the Argentine-BolivianP. heterantha. Lavin (1988) and Lavin and Sousa (1995)had suggested this species to be closely related toCoursetia dubia, C. grandiflora, and C. tumbezensis. All
2003] 401LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
four of these species have a similar distribution includ-ing southern Ecuador and Peru.
The following nomenclature is modified from Lavinand Sousa (1995) to adjust for the new circumscriptionof Poissonia.
The genus Poissonia comprises two well-supportedclades, one with P. hypoleuca and P. orbicularis, and theother with P. heterantha and P. weberbaueri. Notably, asimilar geographic pattern occurs within each of thesetwo clades: an Argentine-Bolivian species (P. hypoleucaor P. heterantha) as sister to a southern Peruvian species(P. orbicularis or P. weberbaueri). The difference betweenthe two clades is that P. hypoleuca and P. orbicularis oc-cur in seasonally dry tropical forests, whereas P. het-erantha and P. weberbaueri occur in the deserts (Montein northern Argentina and Arequipa in Peru). The evo-
lutionary rates analysis (see below and Fig. 6) suggeststhat the potential vicariant event involving the two de-sert species is about three times older (i.e., 3 Ma) thanthat involving the two species from the seasonally dryforest.
Other Taxonomic Implications. The other potentialtaxonomic implications of this analysis mainly involvespecies of Coursetia (sensu stricto), which were classi-fied into sections by Lavin (1988) and Lavin and Sousa(1995). The ITS and combined morphological and ITSdata suggest that of the species retained in Coursetia,only sect. Madrenses is monophyletic (e.g., Fig. 3). Incontrast, species of Coursetia sections Coursetia andCraccoides form disparate groups of small clades. Forexample, the species in the clade including the mostrecent common ancestor of C. brachyrhachis and C. mar-aniona (Fig. 3) were classified into sect. Coursetia, aswere the distantly related C. ferruginea, C. rostrata, andC. glandulosa. Similarly, the species in the clade delim-ited by the most recent common ancestor of C. hassleriand C. pumila, and that by the most recent commonancestor of C. gracilis and C. grandiflora, as well as thedistantly related C. hintonii were classified into sect.Craccoides (Lavin 1988). Generally, the species of bothof these sections are interdigitated on the ITS and com-bined phylogeny (Fig. 3). Because of this, the formalsectional classification of Coursetia (Lavin 1988) isabandoned.
The abandonment of the sectional classification ofCoursetia is motivated by the realization that morpho-logical traits that formed the basis of this classificationare prone to a higher level of independent evolutionthan previously considered. This is exemplified by theresupinate pod and long pedunculate inflorescence(state 1 for each of characters 20 and 22 in AppendixA). These two traits, in part, readily distinguished sect.Craccoides from sect. Coursetia (Lavin 1988). The com-bined morphology and ITS data set suggests that eachof these traits has evolved independently four separatetimes among robinioid legumes (Figs. 2–3; Table 4).The high retention indexes of most of the 40 morpho-logical traits used in this analysis suggest that mor-phological data are phylogenetic informative (comparevalues in Tables 1 and 4). However, the consistencyindexes are low enough to suggest that morphologyshould be used in combination with molecular data forphylogenetic inference.
Evolutionary Rates Analysis. The best evidencefor the age of robinioid legumes (i.e., without Sesbania)comes from Tertiary fossil wood samples. In contrastto wood, the fossil leaves and fruits reported for Ro-binia are doubtful (Herendeen et al. 1992). For example,Robinia californica Axelrod (Axelrod 1987) is describedfrom fossil leaves that are definitively not those of Ro-binia (Wolfe and Schorn 1990; personal observation ofpublished photos) and fruits that have been deter-
402 [Volume 28SYSTEMATIC BOTANY
mined to be fossilized crane-fly larvae (Jack Wolfe, inlitt.).
From thorough comparisons of extant and Tertiaryfossil woods, Matten et al. (1977), Wheeler and Landon(1992), Page (1993), and Wheeler (2001) have recog-nized all fossil Robinia woods as Robinia zirkelii (Platen)Matten, Gastaldo, and Lee. As a fossil wood species,Robinia zirkelii existed from the Late Eocene throughthe Pliocene and was widespread across North Amer-ica, as well as in western Europe. Although spanninga great range in time and geography, these wood sam-ples have been assigned to this single fossil speciesbecause they display very uniform qualitatively diag-nostic traits found only in the modern genus Robinia.A set of diagnostic wood traits displayed by Robiniazirkelii that is found as a combination otherwise onlyin extant robinioid legumes are 1) vestured intervesselpits, 2) storied axial parenchyma and vessel elements,and 3) numerous thin-walled tyloses. In particular, thethin-walled tyloses are very uncommon in wood ofpapilionoid legumes and definitely have never beenobserved in numerous wood samples of Sesbania or(when wood is produced) the Loteae-Coronilleae tribalalliance. More importantly, extant species of Robiniaand Robinia zirkelii share apomorphic wood characters,including 1) ring-porous wood (approached otherwiseto some extent only in Sesbania punicea—character #37in Appendix A), 2) homocellular rays (otherwise ob-served only in some samples of Lennea, Hybosema, Glir-icidia, and Coursetia—character #38 in Appendix A),and 3) spiral sculpturing in narrow vessels (otherwiseobserved only in Genistidium—character #40 in Appen-dix A). The three apomorphies listed immediatelyabove map to the Robinia stem clade in the robinioidphylogeny using the character reconstruction option inPAUP on the combined data sets. Although ring po-rosity (#37) may be ecologically determined (Wheelerand Landon 1992), the other two apomorphies (#’s 38and 40), and the unique combination of the three ad-ditional wood traits shared between extant Robinia andR. zirkellii compel us to date the Robinia stem cladefrom the Late Eocene, or 33.7 Ma in accordance withBerggren et al. (1995).
The analysis of rates of evolution in sequences fromthe ITS region and the matK locus revealed similar es-timates of ages of clades within the robinioid phylog-eny. The ITS region gave better resolution amongclosely related species, as is typical for this locus inlegumes (e.g., Lavin et al. 2001a; Lavin et al. 2001b).The substitution rate of 3.1–3.4 x 1029 substitutions/site/year (Table 2) is typical for the ITS region of le-gumes (e.g., Lavin et al. 2001b; Richardson et al. 2001;Wojciechowski et al. 1999). The substitution rate for thematK locus, estimated in this study to be 3.9 x 10210
substitutions/site/year (Table 3), is an order of mag-nitude slower, and this is in conformity with the use
of this locus for resolving higher level relationshipsthat can be achieved with sequences from the ITS re-gion (e.g., Hu et al. 2000; Lavin et al. 2001a).
The three main robinioid lineages, Hebestigma-Len-nea, Poitea-Gliricidia, and Robinia and close relatives, alldiverged from each other sometime between the Mid-dle and Late Eocene (Figs. 6–7). That the Cuban He-bestigma diverged from the Mesoamerican Lennea dur-ing this same time frame strongly implicates a Carib-bean vicariance event in explaining the distribution ofHebestigma on the island of Cuba (e.g., Rosen 1976;Iturralde-Vinent and MacPhee 1999). The other GreaterAntillean endemic, Poitea, diverged from its sistermainland clade, Gliricidia, much later (Figs. 6–7), atabout 16 Ma. This suggests a different historical eventas causing the island distribution of this clade. How-ever, Poitea and Gliricidia show the pattern of reciprocalmonophyly (Cunningham and Collins 1998; Riddle1996). Notably, the pattern of reciprocal monophyly israre if non-existent among studies of oceanic islandradiations (Lavin, in mss.). The significance of recip-rocal monophyly is that any attempt at ancestral areaestimation at the base of lineages such as that contain-ing Gliricidia and Poitea would render equivocal ances-tral states. That is, both a continental and island dis-tribution would be optimized at the base of the lineageespecially because of the relatively long branches lead-ing to each sister clade. This implies that the islandlineage could just as well serve as the source for thecontinental lineage, rather than reverse. The pattern ofreciprocal monophyly is not associated with an endem-ic diversification following chance dispersal, but ratherwith a historical vicariant event (Cunningham andCollins 1998; Lavin et al. 2000). With the Greater An-tilles represented in two of the three basal robinioidclades, the pattern of reciprocal monophyly involvedin one of these (i.e., Poitea-Gliricidia), and an estimatedrange of Caribbean lineages all well into the Tertiary,a Greater Antillean representation in the ancestral areaof robinioids is strongly suggested.
ACKNOWLEDGEMENTS. Alfonso Delgado-Salinas, Bente Klit-gaard, Gwilym Lewis, Tom Ranker, and Toby Pennington suppliedleaf material of species that were critical to this study. Toby Pen-nington provided the wood sample of Poissonia weberbaueri. Thecurators of the following herbaria made loans available: ASU, F,GH, K, MO, MONT, NY, TEX, and US.
LITERATURE CITED
ANDERSON, J. L. and J. M. PORTER. 1994. Astragalus tortipes (Faba-ceae): a new species from desert badlands in southwesternColorado and its phylogenetic relationships within Astraga-lus. Systematic Botany 19: 116–125.
ALLAN, G. J., and J. M. PORTER. 2000. Tribal delimitation and phy-logenetic relationships of Loteae and Coronilleae (Faboideae:Fabaceae) with special reference to Lotus: evidence from nu-clear ribosomal ITS sequences. American Journal of Botany87: 1871–1881.
AXELROD, D. I. 1987. The Late Oligocene Creede Flora, Colorado.
2003] 403LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
University of California Publications in Geological Sciences,volume 130. Berkeley, California: University of CaliforniaPress.
BERGGREN, W. A., D. V. KENT, C. C. SWISHER III, and M. P. AUBRY.1995. A revised Cenozoic geochronology and chronostratig-raphy. Pp. 129–212 in Geochronology, Time Scales and GlobalStratigraphic Correlation, eds. W. A. Berggren, D. V. Kent, M.P. Aubry, and J. Hardenbol. Society for Sedimentary GeologySpecial Publication No. 54. Tulsa, Oklahoma: SEPM.
BEYRA-M., A., and M. LAVIN. 1999. Monograph of Pictetia (Papi-lionoideae; Leguminosae) and review of tribe Aeschynome-neae. Systematic Botany Monographs 56: 1–93.
BUCKLER, E. S., A. IPPOLITO, and T. P. HOLTSFORD. 1997. The evo-lution of ribosomal DNA: divergent paralogues and phylo-genetic implications. Genetics 145: 821–832.
CUNNINGHAM, C. W., and T. COLLINS. 1998. Beyond area relation-ships: extinction and recolonization in molecular marine bio-geography. Pp. 297–321 in Molecular approaches to ecologyand evolution, eds. R. DeSalle, B. Schierwater. Birkhauser, Ba-sel.
FELSENSTEIN, J. 1985. Confidence limits on phylogeny: an approachusing the bootstrap. Evolution 39: 783–791.
GUTIERREZ, D. G., L. KATINAS, and S. S. TORRES-ROBLES. in press.Type material of Carlos L. Spegazzini in the Museo de LaPlata Herbarium (LP), Argentina: II. Fabaceae. Darwiniana.
HERENDEEN, P. S., W. L. CREPET, and D. L. DILCHER. 1992. Thefossil history of the Leguminosae: phylogenetic and biogeo-graphic implications. Pp. 303–316 in Advances in legume sys-tematics, part 4, the fossil record, eds. P.S. Herendeen, D.L.Dilcher. Kew: Royal Botanic Gardens.
———, and R. B. MILLER. 2000. Utility of wood anatomical char-acters in cladistic analyses. International Association of WoodAnatomists Journal 21: 247–276.
HU, J.-M., M. LAVIN, M. F. WOJCIECHOWSKI, and M. J. SANDERSON.2000. Phylogenetic systematics of the tribe Millettieae (Le-guminosae) based on chloroplast trnK/matK sequences andits implications for evolutionary patterns in the Papiliono-ideae. American Journal of Botany 87: 418–430.
HUELSENBECK, J. P., and F. R. RONQUIST. 2001. MrBayes: Bayesianinference of phylogeny. Bioinformatics 17: 754.
———, R. RONQUIST, R. NIELSON, and J. P. BOLLBACK. 2001. Bayes-ian inference of phylogeny and its impact on evolutionarybiology. Science 294: 2310–2314.
ITURRALDE-VINENT, M. A., and R. D. E. MACPHEE. 1999. Paleoge-ography of he Caribbean region: implications for Cenozoicbiogeography. Bulletin of the American Museum of NaturalHistory 238: 1–95.
LANGLEY, C. H., and W. FITCH. 1974. An estimation of the con-stancy of the rate of molecular evolution. Journal of MolecularEvolution 3: 161–177.
LAVIN, M. 1988. Systematics of Coursetia (Leguminosae-Papilion-oideae). Systematic Botany Monographs 21: 1–167.
———. 1993. Systematics of the genus Poitea (Leguminosae): in-ferences from morphological and molecular data. SystematicBotany Monographs 37: 1–87.
———, S. MATHEWS, and C. HUGHES. 1991. Chloroplast DNA var-iation in Gliricidia sepium (Leguminosae): intraspecific phy-logeny and tokogeny. American Journal of Botany 78(11):1576–1585.
———, and M. SOUSA S. 1995. Phylogenetic systematics and bio-geography of the tribe Robinieae. Systematic Botany Mono-graphs 45: 1–165.
———, M. THULIN, J.-N. LABAT, and R. T. PENNINGTON. 2000. Af-rica the odd man out: molecular biogeography of dalbergioidlegumes (Fabaceae) suggests otherwise. Systematic Botany25: 449–467.
———, R. T. PENNINGTON, B. KLITGAARD, J. SPRENT, H. C. DE
LIMA, and P. GASSON. 2001a. The dalbergioid legumes (Fa-
baceae): delimitation of a monophyletic pantropical clade.American Journal of Botany 88: 503–533.
———, M. F. WOJCIECHOWSKI, A. RICHMAN, J. ROTELLA, M. J. SAN-DERSON, and A. BEYRA-M. 2001b. Identifying Tertiary radia-tions of Fabaceae in the Greater Antilles: alternatives to cla-distic vicariance analysis. International Journal of Plant Sci-ences 162(6 supplement): S53–S76.
MATTEN, L. C., R. A. GASTALDO, and M. R. LEE. 1977. Fossil Ro-binia wood from the western United States. Review of Paleo-botany and Palynology 24: 195–208.
PAGE, V. M. 1993. Anatomical variation in the wood of Robiniapseudoacacia L. and the identity of Miocene fossil woods fromsouthwestern United States. International Association ofWood Anatomists Journal 14: 299–314.
POSADA, D., and K. A. CRANDALL. 1998. ModelTest: testing themodel of DNA substitution. Bioinformatics 14: 817–818.
RAMBAUT, A. 1996. Se-Al ver.1.0a1, sequence alignment editor. Ox-ford: University of Oxford (http://evolve.zoo.ox.ac.uk/Se-Al/Se-Al.html)
———, and GRASSLY, N. C. 1997. Seq-Gen: An application for theMonte Carlo simulation of DNA sequence evolution alongphylogenetic trees. Computer and Applied Biosciences 13:235–238.
RICHARDSON, J. E., R. T. PENNINGTON, T. D. PENNINGTON, and P.M. HOLLINGSWORTH. 2001. Rapid diversification of a species-rich genus of neotropical rain forest trees. Science 293: 2242–2245.
RIDDLE, B. R. 1996. The molecular phylogenetic bridge betweendeep and shallow history in continental biotas. Trends inEcology and Evolution 11: 207–211.
ROSEN, D. 1976. A vicariance model of Caribbean biogeography.Systematic Zoology 24: 431–464.
SANDERSON, M. J. 1997. A nonparametric approach to estimatingdivergence times in the absence of rate constancy. MolecularBiology and Evolution 14: 1218–1231.
———. 1998. Estimating rate and time in molecular phylogenies:beyond the molecular clock. Pp. 242–264 in Molecular Sys-tematics of Plants, eds. D. Soltis, P. Soltis, and J. J. Doyle. NewYork: Chapman and Hall.
———. 2001. r8s, version 1.0(beta), User’s Manual (June 2001). Dis-tributed by the author (http://ginger.ucdavis.edu/r8s/). Da-vis: University of California.
———. 2002. Estimating absolute rates of molecular evolution anddivergence times: a penalized likelihood approach. MolecularBiology and Evolution 19: 101–109.
SWOFFORD, D. 2001. PAUP*. Phylogenetic Analysis Using Parsi-mony (*and Other Methods), version 4.0beta8. Sunderland:Sinauer Associates, Inc.
WHEELER, E. A. 2001. Fossil dicotyledonous woods from the Flor-issant Fossil Beds National Monument, Colorado. Pp. 197–213in Proceedings of the Denver Museum of Science and Nature.Series 4, eds. E. Evanoff, K.M. Gregory-Wodziki, K.R. John-son, Proceedings of the Denver Museum of Nature and Sci-ence. Series 4. Denver: Denver Museum of Nature and Sci-ence.
WHEELER, E. A., and J. LANDON. 1992. Late Eocene (Chadronian)dicotyledonous woods from Nebraska: evolutionary and eco-logical significance. Review of Paleobotany and Palynology74: 267–282.
WOJCIECHOWSKI, M. F. in press. Reconstructing the phylogeny oflegumes (Fabaceae): an early 21st century perspective. Pp.00–00 in A. Bruneau and B. Klitgaard, eds. Advances in le-gume systematics, part 10. Kew: Royal Botanic Gardens.
———, M. J. SANDERSON, B. G. BALDWIN, and M. J. DONOGHUE.1993. Monophyly of aneuploid Astragalus (Fabaceae). Evi-dence from nuclear ribosomal DNA internal transcribedspacer sequences. American Journal of Botany 80: 711–722.
———, ———, and J-M. HU. 1999. Evidence on the monophyly
404 [Volume 28SYSTEMATIC BOTANY
of Astragalus (Fabaceae) and its major subgroups based onnuclear ribosomal DNA ITS and chloroplast DNA trnL introndata. Systematic Botany 24: 409–437.
———, ———, K. P. STEELE, and A. LISTON. 2000. Molecular phy-logeny of the ‘‘Temperate Herbaceous Tribes’’ of papilionoidlegumes: a supertree approach. Pp. 277–298 in Advances inLegume Systematics, part 9, eds. P.S. Herendeen and A. Bru-neau. Kew: Royal Botanic Gardens.
WOLFE, J. A., and H. E. SCHORN. 1990. Taxonomic revision of theSpermatopsida of the Oligocene Creed Flora, southern Col-orado. U. S. Geological Survey Bulletin 1923: 1–40.
ZWICKL, D. J., and D. M. HILLIS. 2002. Increased taxon samplinggreatly reduces phylogenetic error. Systematic Biology 51:588–598.
APPENDIX AEnumeration of the morphological characters derived from
the monographs of Coursetia (Lavin 1988), Poitea (Lavin 1993),and the tribe Robinieae (Lavin and Sousa 1995).
1. Floral pedicles articulated with calyx50 (as in Gliricidia),confluent with the calyx51 (as in Hebestigma).
2. Standard claw gradually tapered from the blade of pet-al50 (as in Poitea and Sphinctospermum), abruptly contractedfrom the blade51 (most robinioids).
3. Petal pigments with reddish to yellowish pigments50,predominantly whitish51 (as in Coursetia brachyrhachis), pre-dominantly bluish52 (Coursetia heterantha, C. hypoleuca, C. or-bicularis). Bluish petal pigments are confined to three of thefour species of the newly circumscribed Poissonia.
4. Callus of the nectar guide on standard petal: as a pair,one on either side of the midrib50 (nearly all robinioids), sin-gle, centered along the midrib51 (Hybosema). The single cen-tral nectar guide is unique to the two hybosemas, which arenested within the Gliricidia clade.
5. Wing petals lateral to the keel petals50 (most robinioids),assuming the position of the standard51 (as in Poitea galego-ides), highly reduced and covered by the base of standard52(P. glyciphylla and P. multiflora).
6. Keel petals fused along the abaxial side to near the distaltip50 (most robinioids), fused for a short distance along theabaxial side equidistant between basal and distal ends51 (asin Poitea galegoides).
7. Keel petals markedly shorter than wing and standardpetals50 (most robinioids), keel petals longer than the otherpetals51 (Poitea).
8. Wing and keel petals free from each other50 (most ro-binioids), wing and keel connate via a boss and socket joint51(as in Poitea florida).
9. Keel petals blunt50 (most robinioids), distal keel tipcurved upward to a sharp point, rostrate51 (Coursetia andPoissonia). The rostrate keel tip has evolved several times, in-cluding independently in the newly circumscribed generaPoissonia and Coursetia.
10. Calyx lobes much shorter than and evenly spacedaround calyx tube50 (Hebestigma and Gliricidia), lobes as longor longer than the tube51 (as in Coursetia), calyx lobes shortand unevenly spaced to render a bilabiate calyx52 (Lennea,Hybosema, and Poitea). The last character condition was usedby Lavin and Sousa (1995) as evidence for the segregation ofHybosema from Gliricidia.
11. Margins of calyx lobes without long whitish hairs50(nearly all robinioids), with long whitish hairs51 (Coursetiaferruginea, Poissonia hypoleuca, and P. orbicularis).
12. Calyx tube not persisting with maturing fruit50 (mostrobinioids), persisting with mature fruit51 (Hybosema, Poitea,and Gliricidia).
13. Staminal tube diadelphous50 (most robinioids), pseu-domonodelphous51 (as in Lennea and Hybosema). The narrow
distribution of the pseudomonadelphous staminal tubeamong robinioid legumes was used as evidence for distin-guishing Hybosema from Gliricidia by Lavin and Sousa (1995).
14. Stamens enveloped by keel tip at anthesis50 (most ro-binioids), protruding beyond the tip of keep petals51 (as inPoitea galegoides).
15. Style brush absent50 (as in Hebestigma), comprising loosewavy hairs51 (Lennea), comprising bunched straight hairs sur-rounding the distal end of the style52 (as in Robinia and Ol-neya), comprising bunched straight hairs along the side of thestyle53 (Coursetia and Poissonia). The pollen brush confinedto the side of the style (latrorse) has arisen independently in Pois-sonia and Coursetia, according to the molecular data.
16. Style base not differentiated from the rest of the style50(as in Hebestigma and Gliricidia), bulbous or inflated51 (allgenera with a pollen brush; as in Coursetia and Robinia).
17. Stigma apical50 (most robinioids), introrse51 (Poiteapaucifolia and P. dubia).
18. Ovary stipe shorter than calyx tube50 (most robinioids),as long or longer than calyx tube and remaining distinct inthe mature fruit51 (Hybosema, Gliricidia, and Poitea).
19. Mature pods with a continuous chamber housing theseeds50 (as in Hebestigma and Robinia), forming individualseed chambers51 (as in Coursetia, Peteria, and Sphinctosper-mum).
20. Orientation of the mature legume such that the placentaruns along the upper margin50 (most robinioids), such thatthe placenta runs along the lower margin51 (Coursetia sec-tions Neocracca and Craccoides sensu Lavin and Sousa 1995). Re-supinate legumes are rare among papilionoid legumes andsurprisingly have independently evolved several times amongrobinioids. This includes once in the two species of the newlycircumscribed Poissonia (P. heterantha and P. weberbaueri) andthree times in a large subclade nested within Coursetia (i.e., thatdefined by the most recent common ancestor of C. pumila andC. dubia). Post-pollination resupination of flower and fruithave been reported for other legumes, including Astragalustortipes (Anderson and Porter 1994) and A. miser (M. Lavin,personal observation). However, in these Astragalus species, re-supination results from hyperflexion of the floral pedicel andnot from the twisting action characteristic of the certain spe-cies of Coursetia and Poissonia.
21. Testa of seed uniform in color50 (as in Hebestigma andSphinctospermum), mottled with purple patches51 (as in Ro-binia and Coursetia).
22. Inflorescence with a peduncle no longer than the basalinternode length50 (most robinioids), with a very long pe-duncle much longer than lower internode length51 (Coursetiasections Neocracca and Craccoides).
23. Stipitate glands lacking especially on reproductive or-gans50 (as in Hebestigma and Gliricidia), present on at leastthe reproductive organs, especially the ovary51 (as in Robiniaand Peteria).
24. Spinescent stipules lacking50 (as in Hebestigma and Glir-icidia), present51 (as in Robinia, Olneya, and Peteria).
25. Stipule pairs free50 (as in Hebestigma and Robinia), ad-nate along inner surface51 (Sesbania and Poitea).
26. Leaflet nyctinasty with a downward movement50 (mostrobinioids), with a backward movement51 (as in Coursetiasection Coursetia), with a forward movement52 (Sesbania).
27. Dried leaflets lacking tanniniferous patches50 (most ro-binioids), with patches of tannins forming distinct patterns51(Hybosema, Gliricidia, Poitea, and various species of Coursetia).
28. Leaflets imparipinnate50 (most robinioids), paripinna-te51 (Sesbania and various species of Poitea and Coursetia).
29. Leaflets uniform in size from base to distal tip50 (mostrobinioids), distally accresent51 (as in various species of Poiteaand Coursetia).
2003] 405LAVIN ET AL.: ROBINIOID PHYLOGENY AND BIOGEOGRAPHY
30. Leaflets with a thin texture such that secondary veinsare readily visible50 (most robinioids), with a thick texturethat generally obscures the secondary venation51 (Poitea grac-ilis, P. paucifolia, and P. dubia).
31. Leaflets generally elliptical, at least longer than wide50(most robinioids), orbiculate, as wide as long51 (the newlycircumscribe Poissonia: P. hypoleuca, P. orbicularis, P. heterantha,and P. weberbaueri). Leaflets as wide as long are uniquely char-acteristic of the newly circumscribed Poissonia.
32. Short shoots occasionally produced but not covered bydistichous persistent stipules50 (most robinioids), commonlyproduced and densely covered by persistent distichous stip-ules51 (Poitea).
33. Seedlings producing only one eophyll50 (most robi-nioids), producing two eophylls51 (as in Poissonia hypoleucaand P. heterantha), producing no eophylls52 (Olneya). Seed-lings with two eophylls might be a distinction of the newlycircumscribed Poissonia, although this condition is unknownfor P. orbicularis and P. weberbaueri.
34. Secondary roots slender and branching from a centraltaproot50 (most robinioids), fusiform and fascicled51 (as inCoursetia caribaea).
35. Wood with non-septate fibers50 (most robinioids), withseptate fibers51 (Gliricidia robustum, G. ehrenbergii, G. maculata,G. sepium, Olneya, and Genistidium). Septate fibers occur in thetwo hybosemas (G. ehrenbergii and G. robustum) and two spe-cies traditionally recognized as Gliricidia (G. maculata and G.sepium). Notably, they are absent in G. brenningii, which is thebasal branching species in the Gliricidia clade. Although woodcharacteristics have been well sampled from the tribe Robi-nieae, only 14 of about 75 species of Sesbania and four of about180 species of Coronilleae-Loteae have been studied for woodanatomical variation (most Coronilleae-Loteae are herbaceous,
however). Regardless, wood traits have been sampled for allthe major subgroups for each of Coursetia, Poitea, and Sesbania,including Poissonia, the new segregate of Coursetia. This char-acter and the two other wood traits listed below generallyfollow the suggestions provided by Herendeen and Miller(2000).
36. Wood with unstoried rays50 (most robinioids), with allrays storied51 (Hebestigma, Hybosema robustum, Gliricidia ma-culata, and G. sepium). Wood with storied rays is notably con-centrated among the robinioid legumes in the species of thenewly circumscribed Gliricidia.
37. Wood diffuse porous50 (most robinioids), ring po-rous51 (Robinia and Sesbania punicea). Ring porosity, one ofthe critical wood characters used in the assignment of fossilwood to the genus Robinia, may be more ecologically thanphylogenetically determined because it is often used as anindicator of temperate climates, paleo or extant (e.g., Wheelerand Landon 1992; Herendeen and Miller 2000).
38. Cellular composition of rays heterocellular50 (Sesbania,Hebestigma and most other robinioids), homocellular51 (Len-nea viridiflora, Hybosema robusta, Gliricidia sepium, Robinia, Cour-setia glandulosa, and C. rostrata).
39. Tyloses absent50 (Loteae-Coronilleae, Sesbania), present51 (most robinioids). The tyloses in robinioid wood, whenpresent, are always distinctively thin-walled and commonlycontain crystals. Although tyloses are abundant in fossil andextant Robinia wood, crystals in tyloses have yet to be ob-served for this genus.
40. Spiral sculpturing in vessel elements absent50 (mostrobinioids), present51 (Robinia and Genistidium). Page (1993)suggests this could be an adaptation to xeric conditions, whichcould well be the case for Genistidium, an inhabitant of theChihuahuan Desert.
406 [Volume 28SYSTEMATIC BOTANY
APPENDIX B. Data matrix representing the 40 morphological characters scored for 82 terminal taxa. L 5 multistate taxon (01); J 5uncertain state taxon {01}. This morphological data set combined with the sequence data from the ITS data set is deposited with TreeBasestudy accession number S813 (http://www.treebase.org/treebase/index.html) and from http://gemini.oscs.montana.edu/;mlavin/data/robin.htm.
APPENDIX CSpecies, locality, voucher specimen, and GenBank accession
number (1ITS/5.8S, 2trnL, 3matK) are provided. Data sets andconsensus tree descriptions in nexus format are available fromTreeBase study accession number S813 (http://www.treebase.org/treebase/index.html) and from http://gemini.oscs.montana.edu/;mlavin/data/robin.htm.
Loteae: Anthyllis vulneraria L.; Allan and Porter (2000);1AF218499. Anthyllis vulneraria spp. lapponica (Hylander) Jalas;Sweden. Vasterbotten. R. Lampinen 10057 (UC 1586976);3AF543845. Securigera varia L., Allan and Porter (2000);1AF218537; USA. Arizona. Huachuca Mtns. McLaughlin 6823(ARIZ); 3AF543846. Lotus unifoliolatus (Hook.) Benth. USA.California. Wojciechowski 707 (DAV); 1AF467067. Hu et al.(2000); 3AF142729 [reported as Lotus purshianus (Benth.) F. Cle-ments & E. Clements]. Ornithopus compressus L.; Old World;Allan and Porter (2000); 1AF218533; USDA seed source: Spain.Hu 1074; 3AF142727.