Molecular Phylogenetics and Evolution 44 (2007) 89–103www.elsevier.com/locate/ympev

A phylogeny of the “evil tribe” (Vernonieae: Compositae) reveals Old/New World long distance dispersal: Support from separate

and combined congruent datasets (trnL-F, ndhF, ITS)

Sterling C. Keeley a,¤, Zac H. Forsman b, Raymund Chan a

a Department of Botany, University of Hawaii at Manoa, 3190 Maile Way, Honolulu, HI 96822, USAb Department of Biology, University of Hawaii at Manoa, 2450 Campus Road, Honolulu, HI 96822, USA

Received 30 June 2006; revised 21 December 2006; accepted 28 December 2006Available online 8 January 2007

Abstract

The Vernonieae is one of the major tribes of the largest family of Xowering plants, the sunXower family (Compositae or Asteraceae),with ca. 25,000 species. While the family’s basal members (the Barnadesioideae) are found in South America, the tribe Vernonieae origi-nated in the area of southern Africa/Madagascar. Its sister tribe, the Liabeae, is New World, however. This is the only such New/OldWorld sister tribe pairing anywhere in the family. The Vernonieae is now found on islands and continents worldwide and includes morethan 1500 taxa. The Vernonieae has been called the “evil tribe” because overlapping character states make taxonomic delimitations diY-

cult at all levels from the species to the subtribe for the majority of taxa. Juxtaposed with these diYcult-to-separate entities are monotypicgenera with highly distinctive morphologies and no obvious aYnities to any other members of the tribe. The taxonomic frustration gener-ated by these contrary circumstances has resulted in a lack of any phylogeny for the tribe until now. A combined approach using DNAsequence data from two chloroplast regions, the ndhF gene and the noncoding spacer trnL-F, and from the nuclear rDNA ITS region for90 taxa from throughout the world was used to reconstruct the evolutionary history of the tribe. The data were analyzed separately and incombination using maximum parsimony (MP), minimum evolution neighbor-joining (NJ), and Bayesian analysis, the latter producingthe best resolved and most strongly supported tree. In general, the phylogeny shows Old World taxa to be basal and New World taxa tobe derived, but this is not always the case. Old and New World species are found together in two separate and only distantly relatedclades. This is best explained by long-distance dispersal with a minimum of two trans-oceanic exchanges. Meso/Central America has hadan important role in ancient dispersals between the Old and New World and more recent movements from South to North America in theNew World.© 2007 Elsevier Inc. All rights reserved.

Keywords: Vernonieae; Vernonia; Compositae; Old and New World; Phylogeny; ITS; trnL-F; ndhF; Congruence; Mesoamerica: Central America

1. Introduction

The Vernonieae is one of the most poorly understood ofthe >20 recognized tribes of the Compositae (Funk et al.,2005). A perplexing array of intergrading morphologiesand overlapping character states juxtaposed with highlyautapomorphic character state combinations (Keeley and

* Corresponding author. Fax: +1 808 956 3923.E-mail address: sterling@hawaii.edu (S.C. Keeley).

1055-7903/$ - see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2006.12.024

Turner, 1990; Robinson, 1999a,b) make the Vernonieaeamong the most refractory of tribes to elucidating system-atic relationships of any members of the family. The frus-tration felt by those who have attempted to understandrelationships within the Vernonieae has given rise to itsnickname as the “evil tribe” (Funk et al., 2005). No phylog-eny has ever been proposed for the tribe and only a fewrelationships have been suggested even among the bestknown species groups before the present study (Jones,1977; JeVrey, 1988; Robinson, 1999a,b, 2007). Despite diY-culties understanding relationships within the tribe, the

90 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

monophyly of the Vernonieae has never been in doubt. Theoverall circumscription of the tribe has changed little sinceits initial description by Cassini (1819, 1828) and Bentham(1873) and only minor changes have been made as a resultof recent molecular work (Keeley and Jansen, 1994; Kimet al., 1998). The tribe has been traditionally placed in thesubfamily Cichorioideae, a position reconWrmed in a recentstudy of the family (Funk et al., 2005).

The “Vernonia problem” (Bremer, 1994) is one of themajor reasons for the historical diYculties in establishingrelationships within the tribe. Until recently (Robinson,1999a,b), the vast majority of species (>1000) was foundwithin a single worldwide core genus, Vernonia. Aroundthis enormous core genus swirled a cloud of largely mono-typic genera with unusual and distinctive morphologiesthat made it diYcult to relate these taxa to those with themore common morphological ground plan of the tribe. Forexample, Stokesia, a monotypic endemic of the southeast-ern US, is the only member of the tribe with zygomorphicXorets. Similarly, the monotypic Pacourina edulis fromnorthern South America has an unusual head morphologyand is the only truly aquatic member of the family, whileHesperomannia from the Hawaiian Islands has become somodiWed by adaptation to bird pollination that it was untilrecently (Kim et al., 1998) thought to belong to the tribeMutisieae. Out of the 121 recognized genera in the Vern-onieae 48 are monotypic and another 30 have only two spe-cies, leaving most species even now in only a few genera(Robinson, 1999a,b, 2007). Studies by Robinson and Kahn(1986) and Robinson and Funk (1987) pointed out theparaphyly of Vernonia s.l., a Wnding supported by Keeleyand Jansen (1994) in a chloroplast DNA (cpDNA) restric-tion site study. Robinson (1999a,b, 2007) made sweepingchanges in the circumscription of Vernonia, limiting thegenus to a small group of eastern North American speciesthat includes the type species for the tribe (Vernonia noveb-oracensis (L.) Willd.). The remaining taxa were placed innewly created genera which were in turn placed into one ofapproximately 20 subtribes (Robinson, 2007). Few relation-ships were suggested among subtribes and genera, however,leaving relationships among tribal members unresolved ashas been the case since the tribe’s original description.

One of the few distinctions generally noted within thetribe has been that of two geographically separate lineages,one in the Old World and the other in the New World. Thissubtribal dichotomy, initially proposed by Gleason (1906),was extended by Jones (1977) in an overview of the tribe,followed by synoptic treatments of Vernonia in the NewWorld (Jones, 1979) and the Old World (Jones, 1981). TheOld World species of Vernonia were placed into the subge-nus Orbisvestus and the New World species into subgenusVernonia. In a treatment of African species, JeVrey (1988)noted features of morphology, chemistry and pollen thatalso suggested separate lineages for New and Old Worldspecies. Despite the inclusion of most species in the genusVernonia s.l., JeVrey (1988) proposed that the closest rela-tionships were among taxa within each hemisphere, again

suggesting two separate evolutionary lines within the tribe.Following this tradition, Robinson (1999a,b, 2007) erectedseparate subtribes for Old and New World Vernonieae inthe most complete taxonomic treatments of the tribe todate.

Despite emphasis on the diVerences between the Newand Old World subtribal lineages, cross-hemisphere rela-tionships have been proposed. Turner (1981) suggested apossible connection to the Old World for the CentralAmerican Leiboldia (Leboldiinae) group. Their morpholo-gies, chromosome numbers and chemistry did not entirelyWt with other New World species, but were similar to sometaxa in the Old World. Keeley and Turner (1990), in a cla-distic analysis of morphological characters, and Keeley andJansen (1994), using cpDNA restriction site data, foundclades containing both New and Old World species sug-gesting a connection between the hemispheres. In his recenttreatments Robinson (1999a,b, 2007) also pointed out caseswhere the New/Old World dichotomy did not seem to hold.For example, the Old World genus Manyonia was postu-lated to be close to the New World genus Heterocypsela,and conversely the New World genera Telmatophila, Acile-pidopsis and Mesanthophora were placed in the Old Worldsubtribe Erlangeinae.

The overall goal of this study was to provide a phylog-eny for the Vernonieae for the Wrst time and with it toclarify New and Old World subtribal relationships.Within this framework, additional goals were to gain abetter understanding of the potentially important role ofMeso/Central America in connecting the two hemispheres(Keeley and Jansen, 1994), to ascertain the derived posi-tion of the North American taxa, and to further explorethe role of Brazil in the origin of New World taxa, as sug-gested by Keeley and Jones (1979). Three phylogeneticmarkers (the chloroplast non-coding trnL-F and codingndhF) and the nuclear rDNA ITS region were chosen toresolve relationships within the Vernonieae. We alsoexamined congruence and resolution of these three mark-ers for the tribe worldwide.

2. Materials and methods

2.1. Choice of taxa and regions to be sequenced

Vernonieae taxa from 90 species and 35 genera weresampled across as wide a geographical range as possible inthe Old and New Worlds. These taxa are listed in Table 1with GenBank Accession Numbers. Nomenclature isaccording to Robinson (1999a,b, 2007). Revision of the OldWorld Vernonieae has not yet been completed (Robinson,pers. comm.), however. Consequently, a number of OldWorld species have yet to be formally transferred fromVernonia s.l. and thus must retain that genus name heredespite the fact that Vernonia s.s is entirely New World(Robinson, 1999a).

The chloroplast gene ndhF has been used in a variety ofphylogenetic studies at several taxonomic levels (Olmstead

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 91

(continued on next page)

Table 1Taxa sampled with species codes, geographic locations, and GenBank Accession Numbers

Genus (code) Species Author Geographic source

Collector, Number, Herbarium

Genbank Accession Number

ITS NDHF TRNL-F

Albertinia brasiliensis Spreng. Brazil Stannard, Ganev & Queiroz 51633-US

EF155744 EF155656 EF155832

Baccharoides (A) adoensis (Sch.Bip ex Walp.) H. Rob. Africa (cult.) Kew 453-68-45801-K EF155745 EF155657 EF155833Baccharoides (L) lasiopus (O.HoVm.)H.Rob. Africa Keeley cp-1 (930)-CONN EF155796 EF155708 EF155884Bothriocline(L) laxa H.Wild & G. Pope Rhodesia S. B. Jones 76-114 (Cult.)-G EF155746 EF155658 EF155834Bothriocline(U) ugandensis (S. Moore)M.G. Gilbert Burundi Reekmans 8820-MO EF155747 EF155659 EF155835Brachythrix brevipapposa H.Wild & G. Pope Tanzania Carter et al. 2527-K EF155748 EF155660 EF155836Cabobanthus polysphaera (S. Moore) H. Rob. Zambia Philcox et al. 10264-MO EF155749 EF155661 EF155837Centauropsis decaryi Humbert Madagascar J. L. Zarucchi et al. 7361-MO EF155750 EF155662 EF155838Centrapalus pauciXorus (Willd.)H. Rob. Zimbabwe Purdue USDA-US EF155751 EF155663 EF155839Centratherum1 punctatum Cass. Brazil (cult.) Funk & Keeley 12,443-US EF155753 EF155665 EF155841Centratherum2 punctatum Cass. Brazil J. Panero-TEX EF155754 EF155666 EF155842Chresta sphaerocephala DC. Brazil Azeviedo et al. 533-K EF155755 EF155667 EF155843Chrysolaena(F) Xexuosa (Sims)H.Rob. Paraguay E. Zardini & R. Velasquez

25832-MOEF155756 EF155668 EF155844

Chrysolaena(P) platensis (Spreng.)H.Rob. Paraguay E. Zardini & R. Velasquez 24552-MO

EF155757 EF155669 EF155845

Critoniopsis(H) huairacajana (Hieron.) H. Rob. Ecuador Keeley 4129-US EF155821 EF155733 EF155909Critoniopsis(So) sodiroi (Hieron. Ex Sodiro) H. Rob. Ecuador Keeley & Keeley s.n.-WHIT EF155760 EF155672 EF155848Cyanthillium patulum (Ait.) H.Rob. China Keeley cp-6 (1557)-CONN EF155812 EF155724 EF155900Distephanus(M) madagascariensis Less. Madagascar P. Phillip 1905-MO EF155761 EF155673 EF155849Distephanus(P) polygalaefolia (Less.) H.Rob.& B. Kahn Madagascar Kew 282-85-03275-K EF155762 EF155674 EF155850Eirmocephala brachiata (Benth.)H.Rob. Costa Rica W. Haber & W. Zuchowski

10536-MOEF155763 EF155675 EF155851

Elephantopus(C) carolinianus Willd. Georgia Urbatsch 6017-LSU EF155764 EF155676 EF155852Elephantopus(E) elatus Gleason Georgia Urbatsch 6051-LSU EF155765 EF155677 EF155853Elephantopus(M) mollis Domin. Hawaii Funk& Keeley-US EF155766 EF155678 EF155854Elephantopus(M) mollis Domin Singapore Lum s.n.-Singapore EF155767 EF155679 EF155855Elephantopus(N) nudatus A.Gray Georgia Urbatsch 6049-LSU EF155768 EF155680 EF155856Elephantopus(T) tomentosus L. Georgia Coile s.n.-GA EF155769 EF155681 EF155857Eremanthus erythropappus (DC.) MacLeish Brasil de Moray 661-MO EF155770 EF155682 EF155858Eremosis shannoni (Coult.) H.Rob. Guatemala S.C. Keeley & J. E. Keeley

3161-MOEF155758 EF155670 EF155846

Ethulia conyzoides L. Zaire Pauly 234-K EF155772 EF155684 EF155860Gorceixia decumbens Baker Brazil MGC 935-US EF155773 EF155685 EF155861Gymnanthemum

(Am)amygdalinum Sch.Bip. Ex Walp. Africa Kew 318-86-02802-K AY504695 AY504737 AY504777

Gymnanthemum(M)

mesipifolium (Less.) H.Rob. GH-Africa Jansen 995-MICH EF155775 EF155687 EF155863

Hesperomannia(Arb)

arbuscula Hillebr. Oahu,HI Ching 11a-DNA library UH EF155776 EF155688 EF155864

Hesperomannia(Arr)

arborescens H. Gray Oahu,HI Ching A20-DNA library UH AY504696 AY504738 AY504778

Hesperomannia(L)

lydgatei Forbes Kauai,HI Ching K60-DNA library UH EF155777 EF155689 EF155865

Hesperomannia(SN)

sp. nov. HI Ching H28-DNA library UH EF155778 EF155690 EF155866

Heterocypsela andersonii H.Rob. Brasil A. Salino 3043-US EF155779 EF155691 EF155867Hilliardiella(A) aristata (DC.) H.Rob. South Africa Funk 12,410-US EF155780 EF155692 EF155868Hilliardiella(L) leopoldii (Vatke) H.Rob. Ethiopia Tadesse 7551-MO EF155781 EF155693 EF155869Hilliardiella(O) oligocephala (FV.) H.Rob. South Africa Funk 12,429-US EF155782 EF155694 EF155870Leiboldia guerreroana (S.B. Jones) H.Rob. S. Mexico Spooner & Burgos 2625-

TEXEF155820 EF155732 EF155908

Lepidaploa(A) arborescens (L.) H.Rob. Costa Rica Keeley 4085-LAM EF155783 EF155695 EF155871Lepidaploa(Ba) balansae (Hieron.) H.Rob. Paraguay R. Degen 1606-MO EF155784 EF155696 EF155872Lepidaploa(Bo) borinquensis (Urb.) H.Rob. Costa Rica Keeley 1550-LAM EF155785 EF155697 EF155873Lepidaploa(C) canescens (Kunth) H.Rob. Costa Rica Keeley s.n.-CONN EF155786 EF155698 EF155874Lepidaploa(E) ehretifolia (Aristeg.) H. Rob. Venezuela Keeley & Keeley 4443F-US EF155829 EF155741 EF155917Lepidaploa(T) tortuosa (L.) H.Rob. Costa Rica S. Keeley 3252-LAM EF155787 EF155699 EF155875Lepidonia jonesii (B.L.Turner) H.Rob.& V.A.

FunkMexico Todzia 2835-TEX EF155788 EF155700 EF155876

92 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

and Sweere, 1994; Olmstead et al., 2000; Pfeil et al., 2002)including several in the Asteraceae (Kim and Jansen, 1995;Kim et al., 1998, 2002) and the Vernonieae (Kim et al.,

1998). Following the work of Kim et al. (1998, 2002) onlythe 3� half of the gene was used due to the lack of variationin the 5�end for the Vernonieae. The non-coding trnL-F

Table 1 (continued)

Genus (code) Species Author Geographic source

Collector, Number, Herbarium

Genbank Accession Number

ITS NDHF TRNL-F

Lessingianthus buddleiifolius Mart. Ex DC. Brasil R. Anderson 9706-MO EF155789 EF155701 EF155877Linzia(Ac) accommodata (Wild) H. Rob. Zimbabwe Bayliss 10387-MO EF177478 EF177479 EF177480Linzia(G) gerberiformis (Oliv. & Hiern.) H. Rob Rwanda Auquier 3263-MO EF155791 EF155703 EF155879Linzia(M) melleri (Oliv. & Hiern.) H. Rob Burundi Reekmans 10202-MO EF155792 EF155704 EF155880Linzia(G) gerberiformis (Oliv. & Hiern.) H. Rob Zimbabwe Keeley cp-4 (31)-CONN EF155752 EF155664 EF155840Linzia(M) melleri (Oliv. & Hiern.) H.Rob. Malawi Christenson & Liperde 1491-

USEF155790 EF155702 EF155878

Munnozia giganteum Rusby Peru Dillon s.n.-F AY504697 AY504739 AY504779Muschleria sp. Tanganika Milne-Redhead & Taylor

9039-KEF155793 EF155705 EF155881

Orbivestus cinarescens (Sch.Bip.) H. Rob South Africa Koekemoer 232-MO EF155795 EF155707 EF155883Orbivestus cinerascens (Sch.Bip.) H. Rob South Africa M. Koekemoer 232-PRE EF155794 EF155706 EF155882Parapolydora fastigiata (Oliv. & Hiern.) H. Rob South Africa Koekemoer 225-MO EF155817 EF155729 EF155905Polydora poskeana (Vatke & Hillebr.) H. Rob. South Africa Koekemoer 233-MO EF155797 EF155709 EF155885Sipolesia lanuginosa Glaz. Ex Oliv. Brasil Harley et al. 24885-K EF155798 EF155710 EF155886Stokesia laevis Greene SE .US Kew 068-62-06802 –K (cult.) EF155799 EF155711 EF155887Stokesia laevis Greene SE.USA Kew 068-62-06802 –K (cult.) EF155800 EF155712 EF155888Stramentopappus pooleae (B.L.Turner) H.Rob.& V.A.

FunkS. Mexico WoodruV & Todzia

199-TEXEF155801 EF155713 EF155889

Strobocalyx arborea (Buch.-Ham.) H.Rob. Singapore Lum s.n.-Singapore EF155774 EF155686 EF155862Tephrothamnus paradoxa (Sch.Bip.) H. Rob. Venezuela S. Keeley 4500-6-CONN EF155759 EF155671 EF155847Vernonanthura (A) alamanii DC. Texas Todzia 2869-TEX EF155802 EF155714 EF155890Vernonanthura(B) brasiliana (L.) H.Rob. Brazil unknown-MO EF155827 EF155739 EF155915Vernonanthura(P) patens (H.B.K. )H.Rob. Panama S. Keeley 4685-LAM EF155803 EF155715 EF155891Vernonia(Ab) abyssinica Sch.Bip. Ex Walp. Ethiopia Gilbert, Sebsebe, Vollensen

7408-MOEF155805 EF155717 EF155893

Vernonia(Al) altissima Gleason E.N. America Keeley cp-2-CONN EF155806 EF155718 EF155894Vernonia(An) angustifolia Michx. South Carolina Coile s.n.-GA EF155807 EF155719 EF155895Vernonia(Ba) baldwinii Torr. E. North

AmericaKew 643-52-64320-K (cult.) EF155808 EF155720 EF155896

Vernonia(Br) brachycalyx O.HoVm. Uganda Lock 88/16-K EF155809 EF155721 EF155897Vernonia(Bu) bullata Benth.ex Oerst. Zambia Philcox, Pope, Chisumpa,

Ngoma 10265-MOEF155810 EF155722 EF155898

Vernonia(C) capensis (Houtt.) Druce South Africa L. E. Watson & J. Panero 94-120-TEX

EF155811 EF155723 EF155899

Vernonia(E) elliptica DC. SE Asia (cult.) Funk & Keeley 12442-US EF155813 EF155725 EF155901Vernonia(Fc) fasciculata Michx. E North

America (cult.)U. Posnaniensis, Poland, 573 (cult.)

EF155815 EF155727 EF155903

Vernonia(Fc) fasciculata Michx. E North America (cult.)

Kew 590-53-59006-K EF155816 EF155728 EF155904

Vernonia(Gi) gigantea (Walt.) Trel. E. North America (cult.)

Kew 611-66-61103-K EF155818 EF155730 EF155906

Vernonia(Hu) humbloti Drake Madagascar H.J. Beentje 4823-K EF155819 EF155731 EF155907Vernonia(L) lindheimerii A.Gray & Englem. Texas Lievens 4100-TEX EF155822 EF155734 EF155910Vernonia(L) lindheimerii A.Gray & Englem. Texas J. Kim 10573-TEX EF155823 EF155735 EF155911Vernonia(M) missurica RaWn. E. North

America (cult.)Urbatsch 5870-LSU EF155824 EF155736 EF155912

Vernonia(N) noveboracensis (L.) Michx. E. North America (cult.)

Kew 000-69-18590-K EF155825 EF155737 EF155913

Vernonia(P) profuga L. E. North America (cult.)

Keeley cp-5 (1561)-CONN EF155826 EF155738 EF155914

Vernonia(Sp) sp Brazil unknown-US EF155814 EF155726 EF155902Vernonia(SpS) sp S. Africa M. Koekemoer 1973-US EF155828 EF155740 EF155916Vernonia(Su) subplumosa O.HoVm. Zambia Philcox et al. 10296-MO EF155830 EF155742 EF155918Vernonia(T) texana (A.Gray) Small E. North

AmericaUrbatsch 5889-LSU EF155831 EF155743 EF155919

Vernoniastrum nestor (S.Moore) H.Rob. Malawi Banda, Mwyanyambo 3869-MO

EF155804 EF155716 EF155892

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 93

including the spacer region also has been used in numerousphylogenetic studies and for a wide range of plants (for asummary see Shaw et al., 2005). This region provides vary-ing degrees of resolution, depending on the group. Shawet al. (2005) found that even when trnL-F alone was notsuYcient to provide a resolved phylogeny it was eVectivewhen used in combination with other gene regions. Bayerand Starr (1998) and Bayer et al. (2000) also found trnL-Fuseful at the tribal level in the Asteraceae.

The ITS has been successfully used in studies of theAsteraceae at the species, tribal and family level (e.g. Bald-win, 1992; Marcos and Baldwin, 2001; Chan et al., 2001;Baldwin et al., 2002; Roberts and Urbatsch, 2003, 2004).Work by Kim and Jansen (1995) and Kim et al. (1998,2002) have shown ITS to resolve relationships at the tribaland generic levels within the family and within the Vern-onieae. These data were also combinable with dataobtained from the ndhF chloroplast region (Kim et al.,1998). In analyses of two tribes within Asteraceae subfam-ily Cichorideae, the same subfamily as the Vernonieae(Panero and Funk, 2002; Funk et al., 2005) Samuel et al.(2003) and Funk et al. (2004) also found that ITS providedgood resolution.

2.2. Sequence alignment and phylogenetic analysis

To avoid possible problems with ITS alignments(Álvarez and Wendel, 2003; Goertzen et al., 2003) NJ trees(implemented in PAUP* with an HKY85 substitutionmodel for computational speed) were evaluated for threealternative alignments; one including all data, and two thatexcluded columns containing lower than 50% or 20% pos-terior probability scores generated in ProAlign 0.5a(Loytynoja and Milinkovitch, 2003). Including all positionsresulted in higher resolution but no signiWcant change intopology, therefore all positions were included in the Wnalanalyses. The results of these preliminary studies indicatedthat the ITS, trnL-F and ndhF would provide good phylo-genetic signal for resolving clades within the Vernonieae.To determine combinability the incongruence length diVer-

ence (ILD) test (Farris et al., 1994), implemented as parti-tion homogeneity in PAUP*4.01b.10 (SwoVord, 2002), wasperformed. The settings used the nearest neighbor inter-change (NNI) method of branch swapping with 100 repli-cations, random addition for 3 replicates, nchuckD2,chuckscoreD1; scores are indicated in Table 2. Initial testswere also performed to determine the eVects of gaps; gapsexcluded, treated as missing data or as a Wfth character.Treating gaps as a Wfth character resulted in signiWcant het-erogeneity, but excluding gaps or treating them as a missingcharacter resulted in no signiWcant heterogeneity (Table 3).Including gaps as a missing character resulted in a better-resolved topology without signiWcant topological diVer-ences, therefore gaps were treated as missing data. Missingdata (no sequence) occurred very rarely for the tailing endof some sequences and were coded as ambiguous data (N).

Maximum parsimony (MP), neighbor joining (NJ) andBayesian methods were used to estimate phylogenies foreach separate data partition, and for all partitionscombined. Maximum parsimony trees were generated inPAUP*4.01b.10 (SwoVord, 2002) with the following condi-tions; gaps were treated as missing data in a full heuristicsearch with the tree-bisection-and-reconnection (TBR)branch swapping algorithm, with 10 random additions, and5 trees held each step, with 1000 non-parametric bootstrappseudo-replicates. NJ trees were also estimated in PAUP*

with 1000 non-parametric bootstrap pseudo-replicates,using the Minimum Evolution criteria. Evolutionary mod-els were selected using the AIC criterion (Akaike Informa-tion Criterion, Akaike, 1974) in ModelTest (Posada andCrandall, 1998). Models for separate and combined data

Table 3Primer sequences used for PCR and cycle sequencing

Name Sequence (5�–3�)

ITS5A GGA AGG AGA AGT CGT AAC AAG GITS4 TCC TCC GCT TAT TGA TAT GCtrnL C CGA AAT CGG TAG ACG CTA CGtrnL F ATT TGA ACT GGT GAC ACG AGndhF 1603 CCT YAT GAA TCG GAC AAT ACT ATG CndhF +607 ACC AAG TTC AAT GYT AGC GAG ATT AGT C

Table 2The size of each data partition and the number of parsimony informative characters (PIC) with gaps treated as missing and gaps excluded for each datapartition are given above the double line

ILD scores are given below the double line with alignment gaps excluded (upper matrix) and alignment gaps treated as missing data (lower matrix). Abbre-viations: sp D spacer; NUC D nuclear data partition; ORG, organellar data partition.

ITS-1 5.8S ITS-2 trnL trnL-sp ndhF ndhF-sp NUC ORG

Size (bp) 310 159 236 549 463 601 181 705 1794PIC gaps D missing 234 17 193 23 47 91 58 444 219PIC gaps D excluded 138 17 112 13 31 85 0 267 129ITS-1 » 0.01 0.34 0.55 0.28 0.15 0.99 » »5.8S 0.03 » 0.06 0.23 0.48 0.01 0.96 » »ITS-2 0.25 0.53 » 0.98 0.94 0.68 1 » »trnL 0.74 0.01 0.91 » 0.81 0.63 0.92 » »trnL-sp 0.39 0.19 0.83 0.44 » 0.14 1 » »ndhF 0.33 0.01 0.81 0.39 0.19 » 1 » »ndhF-sp 0.01 0.19 0.01 0.26 0.06 0.01 » »NUC » » » » » » » » »ORG » » » » » » » 0.16 »

94 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

sets were: ITS, GTR + I + G; trnL, TVM + G; ndhF,TVM + I + G; combined data set, TIM + I + G (�D0.5809,pinvD 0.5689). Bayesian analysis was performed in MrBa-yes 3.0b4 (Hulsenbeck and Ronquist, 2001) using theMarkov Chain Monte Carlo analyses (Geyer, 1991), usingfour chains sampled every 100 generations. The Wrst 2000trees were discarded as the burn-in period, this value wasdetermined empirically from plotting the likelihood valuesto determine convergence, and each analysis was run forone million generations. All trees saved from the indepen-dent runs (excluding burn-in) were used to construct 50%majority-rule consensus trees. Trees were drawn withPAUP*4.01b.10 (SwoVord, 2002) or with the programMEGA 3.0 (Kumar et al., 2004).

Outgroup taxa in initial analyses included Munnozia ofthe sister tribe, Liabeae, (sequences included in Table 1 forreference) as well as the yellow-Xowered Madagascan spe-cies of the genus Distephanus. The latter taxon was previ-ously shown to belong to the basal group within theVernonieae (Keeley and Jansen, 1994). As no diVerence intopologies or clade composition resulted from the use ofDistephanus alone it was used as the outgroup in the Wnalanalyses for economy of computation. Nucleotide diver-gence rates were assumed for ndhF and ITS based on esti-mates provided by Kim et al. (1998). Mean nucleotidedivergence between groups was calculated using the pro-gram MEGA 3.0 (Kumar et al., 2004) with 1000 bootstrapreplicates to estimate the standard error and 95% conW-dence intervals.

2.3. Extraction and ampliWcation

DNA extractions were performed on herbarium or silicadried material using Qiagen DNeasy. Plant Mini Kits fol-lowing the instructions supplied but with an extended incu-bation period (up to 40 min) for herbarium material. Whennecessary, an additional clean up and concentration stepwas done using an UltraClean 15 DNA PuriWcation Kit(Mo Bio Laboratories). Primer ITS5A (Downie and Katz-Downie, 1996), based on White et al.’s (1990) fungal primerITS5 and corrected at two positions for angiosperms, wassubstituted for ITS5 in this study. Primers used to amplifyand sequence the trnL region of cpDNA were designed byTaberlet et al. (1991) and those used for the 3� end of thendhF region were designed by Jansen (1992). All primersequences are given in Table 3.

For the PCR ampliWcation reactions, each 25 �l PCRreaction cocktail contained 12.9�l of sterile water, 2.5 �l of10£ PCR reaction buVer A or B (Promega), 2 �l of 20 mMdNTPs (Pharmacia) in an equimolar ratio, 2.5 �l of 25-mMmagnesium chloride, 0.5�l of 10 mg/ml Bovine SerumAlbumin (Sigma), 1 �l of a 10 �M concentration of the for-ward primer, 1 �l of a 10�M concentration of the reverseprimer, 0.1�l of Taq DNA polymerase enzyme (5 U/�l fromPromega), and 2.5 �l of template DNA. The amount of tem-plate DNA was adjusted when necessary to generate suY-cient PCR products for DNA sequencing.

The ampliWcation reactions were conducted in a Gene-Amp PCR System 9700 (Perkin-Elmer). The PCR proWleconsisted of an initial preheating at 94 °C for 2 min fol-lowed by 40 cycles of: 1 min at 94 °C, followed by 1 min at48 °C (54 °C for cpDNA) and 45 s at 72 °C. Primer exten-sion time was increased by 4 s (7 s for cpDNA) for each sub-sequent reaction cycle. An additional 7 min extension at72 °C was added for completion of unWnished DNAstrands. All PCR products were quantiWed by agarose gelelectrophoresis with comparison of an aliquot of productswith a known quantity of a 100-bp DNA ladder (Gene-Choice). The remainder was stored at 4 °C until utilized.

PCR products were puriWed for sequencing using anenzymatic PCR product pre-sequencing kit (USB) follow-ing recommendations from the manufacturer. This methodof puriWcation without loss of PCR products (no Wltration,precipitation, or washes are necessary) is especially impor-tant for DNA extracted from herbarium vouchers, which issometimes only weakly ampliWed and yields barely suY-cient PCR product for sequencing. The cycle sequencingreactions were done using 96-well microplates in a PTC-100thermal cycler (MJ Research). Each one-eighth cyclesequencing reaction cocktail contains 50–150 ng of thepuriWed PCR product, 2 �l of a 1-mM concentration of thesequencing primer, 0.6 �l of a 5£ reaction buVer (400 mMTris HCl, 10 mM magnesium chloride at pH 9.0), and 1�lof the reagent pre-mix from the BigDye (Version 2/3) dyeterminator cycle sequencing pre-mix kit (Applied Biosys-tems). The cycle sequencing program consisted of an initialpreheating at 96 °C for 30 s followed by 25 cycles of: 10 s at92 °C, followed by 15 s at 55 °C and 4 min at 60 °C. Unin-corporated dye terminators were removed by Sephadex(Sigma) gel Wltration using MultiScreen plates (Millipore).The puriWed cycle sequencing products were then resolvedby electrophoresis on a 5% polyacrylamide (MJ ResearchKiloBasePack) gel using a BaseStation 51 automated DNAsequencer (MJ Research). Sequences from both strands ofeach PCR product were examined, compared, and cor-rected using Sequence Navigator software (Applied Biosys-tems). Sequence alignments were generated by ProAlign0.5a0, and adjusted manually (Loytynoja and Milinkovitch,2003).

3. Results

3.1. Relationship of New and Old World Vernonieae

New and Old World Vernonieae are not entirely distinct.Taxa from both hemispheres are found together in twowidely separated clades (clade B� and clade 3, Figs. 2 and 3).Outside of these clades, however, New World and OldWorld taxa are in separate monophyletic lineages.

The most signiWcant bi-hemispheric clade includes taxafrom Mesoamerica and southeast Asia (clade B�, Figs. 2and 3). This is not the Wrst time that Mesoamerican taxahave been found together with Old World species. Keeleyand Jansen (1994) found members of Vernonia subtribe

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 95

Leiboldiinae (Lepidonia jonesii and Stramenopappus poo-leae), plus the North American monotypic genus Stokesia,at the intersection of the Old and New World lineages, asthey are here (Fig. 2). In Keeley and Jansen (1994) thesethree New World taxa were derived above a group of OldWorld species (clades 5 and 6 here) and together formed thebase from which all other New World species arose. Simi-larly, in Kim et al. (1998) Lepidonia jonesii and Stokesiawere placed between Old and New World species in treesconstructed using ndhF and ITS sequence data.

The aYnities of the Leiboldiinae to both New and OldWorld taxa is also shown by the the somewhat labile aYni-ties of the individual taxa of this group (clade B�). The NewWorld Leiboldiinae and Stokesia, and the Old World taxaStrobocalyx arborea and Vernonia elliptica, are the onlytaxa to switch positions between New and Old World poly-tomies in the individual gene trees (Fig. 1). Clade B� is alsothe only one to change position in the combined analyses(Fig. 2 versus Fig. 3). Regardless of the position of thisclade, the taxa are more closely related to each other thanthey are to other species in their respective hemispheres.

A diVerent connection between New and Old Worldtaxa is seen in the monophyletic genus Elephantopus (clade3, Figs. 2 and 3). In this case both Old and New World spe-

cies apparently arose from a New World Andean-basedlineage, (using the species available here). Once the elephan-topoid ancestral line formed in the New World there wasdispersal to the Old World along with continued radiationwithin the New World.

3.2. Individual gene regions

No single gene region alone was suYcient to provide afully resolved phylogeny. Six clades, labeled 1–6, werestrongly supported in all individual gene trees (Fig. 1). Ofthe four other clades, labeled A-D, all four were found inonly in the ITS tree, with three and one clades found in thendhF and trnL-F trees, respectively. All 10 clades werepresent in trees from the combined data (Figs. 2 and 3). Theconsistency of these clades suggests that the phylogeneticsignal is similar in both nuclear and organellar DNA (e.g.,Chen et al., 2003). Taxa in recurring clades are fullyenumerated in Figs. 2 and 3, and are summarized in Fig. 1to save space. Bayesian trees consistently had higher resolu-tion and support values for each data partition (Fig. 1) andfor the combined analyses (Fig. 2) than MP (Fig. 3) and NJtrees. Only Bayesian results are shown for the individualregions.

Fig. 1. Phylograms generated by Bayesian analyses for ITS, ndhF and trnL-F regions. Posterior probability scores higher than 70 are indicated by an aster-isk. Analyses were run for 1 million generations, log-likelihood scores: ITS ¡14495.97, ndhF ¡2623.95, trnL-F ¡3057.80 (see Section 2 for individual data

set models). Numbered and lettered brackets indicate recurring clades. Darkened triangles are collapsed clades; a full listing of taxa in these clades is givenin Fig. 2. Asterisks next to the collapsed clades indicate the number of groupings within the clade with posterior probability scores higher than 70.

96 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

3.2.1. ITSThe ITS tree was the most resolved and best supported

of the individual trees (Fig. 1). Overall the tree was dividedinto two general sections, one New World (clades 1, 2, 3, A,B) and the other Old World (clades C, D, 4, 5, 6). While noNew World taxa were found in the Old World portion,some Old World taxa were found intermixed in the primar-ily New World portion of the tree. Two geographically

anomalous species, Vernonia elliptica and Strobocalyxarborea from southeast Asia were found among the NewWorld species, and clade 3 consisting of the mostly NorthAmerican species in the genus Elephantopus, also includedone widespread Old World taxon. E. mollis.

Several distinct clades (clades 1, A, B) contained taxathat occupy a similar geographic region within the NewWorld. Clade 1, for example, was primarily composed of

Fig. 2. Bayesian analysis of the combined data set (ITS, ndhF, trnL-F) run for 1 million generations, with model TIM + I + G, log-likelihood score¡20656.5. Posterior probability values are given above the line, bootstrap values (1000 replicates) from a majority rule consensus NJ tree (minimumevolution) constructed using the same model are given below the line.

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 97

North American Vernonia species (Robinson, 1999a), witha small number of South and Central American taxa (seeFig. 2 for the expanded clade). Clade A was composed ofBrazilian taxa and clade B of Central American species.The majority of Old World taxa were found in a large poly-tomy below the New World groups. This polytomyincluded clades C and D from south and east Africa and agroup formed from clade 4 (South Africa) paired withCentauropsis (Madagascar). Clades 5 and 6 formed a sister

group to the other Old World clades. Clade 5 included taxafrom Africa, Madagascar and Hawaii, supporting therelationship of Hawaiian and African species reported inKim et al. (1998). Clade 6 included east African species inthe genus Linzia, recently described by Robinson (1999b).

3.2.2. ndhFThe ndhF region included nine of the 10 clades present

in the ITS tree (Fig. 1). New World clades 1, 2, 3 and A were

Fig. 3. Majority rule consensus tree of 1000 bootstrap pseudoreplicates. Bold lines indicate the topology of the strict consensus of 672 most parsimonioustrees for the combined data set (ITS, ndhF, trnL-F). Bootstrap values are provided above the nodes, values below 50% are not shown. Tree length 3127,CI D 0.395, RI D 0.783.

98 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

found in the New World polytomy, as they were in the ITStree, but their relationships were slightly diVerent. Clade A(Brazilian) and clade 1 (primarily North American) werederived from the same branch with Tephrothamnus(Venezuelan) and Eirmocephala (Central American) at thebase of these clades. This changed the relationshipsbetween these clades from independent lines, as was thecase for ITS, to ancestral and derived positions. Chresta andHeterocypsela (both Brazilian) were on separate branchesin the polytomy, again suggesting a more distant relation-ship to taxa in clade A, similar to the Wndings for ITS.Clade 3 was composed of New and Old World Vernonieaeand remained with the New World taxa.

The major diVerence between the ndhF and ITS treeswas the position of Central American taxa of clade B in thesubtribe Leiboldiinae (Leiboldia, Lepidonia, Stramentopap-pus) and the North American Stokesia in the otherwise OldWorld polytomy that included clades 4, C and D. Addition-ally, Strobocalyx arborea and Vernonia elliptica, found withNew World species in the ITS tree, were this time placedwith other Old World taxa. Old World clades 5 and 6,remained in the same basal position as in the ITS tree.

3.2.3. trnL-FThe trnL-F region had the fewest informative characters

of the three regions (Table 2) and correspondingly pro-duced the least resolved tree (Fig. 1). There was no statisti-cal support for a distinction between Old and New Worldtaxa, other than for the basalmost clades 5 and 6. Smallerclades identiWed in the other trees were present, i.e., clades 1,2, 3, 4 and B, but their relationship to other taxa was unre-solved.

3.3. Combined analyses

Results of ILD tests (partition homogeneity in PAUP*)showed that the partitions were congruent except for thendhF spacer region, and the 5.8S ribosomal gene (Table 3).The ndhF spacer could not be reliably aligned and wasexcluded from further analyses. The 5.8S ribosomal genecontributed a small number of informative characters(Table 2), and excluding it did not alter the tree topology(data not shown). It was left in the Wnal analysis in an eVortto obtain as much phylogenetic information as possible.Combined analyses using Bayesian (Fig. 2), NJ (not shown)and MP (Fig. 3) methods produced similar topologies. TheBayesian and NJ methods produced trees with higher reso-lution and statistical support than MP, which only uses asubset of the data (parsimony informative characters).Bayesian posterior probability values were considerablyhigher than NJ (values given in Fig. 2) or MP bootstrapvalues (Fig. 3). There was no conXict between the Bayesiananalysis and clades supported by the other methods. Theinterpretation of Bayesian posterior probability values iscontroversial, and may overestimate statistical support (e.g.Simmons et al., 2004; Suzuki et al., 2002). We consider theNJ and MP bootstrap values as potentially more conserva-

tive estimates of clade support compared to the Bayesianposterior probability values. Posterior probability andbootstrap values are both provided, as they may provideestimates of upper and lower limits of statistical support.

As in the individual analyses the overall pattern was thatof basal groups consisting of primarily Old World taxa and aderived group containing most New World taxa. All 10clades were present and an additional one was recognized(clade B�, Figs. 2 and 3). The formation of this clade providedresolution for the somewhat ambiguous positions of the NewWorld Stokesia and clade B taxa (subtribe Leiboldiinae taxa,Leiboldia, Lepidonia,Stramentopappus) and the Old Worldpair Strobocalyx/Vernonia elliptica. The taxa in clade B� aremost closely related to each other rather than to other generain their respective geographical areas. Clade 3, with both Oldand New World Elephantopus species, remained with NewWorld taxa as in individual analysis. The major topologicaldiVerence between MP and Bayesian analyses was in theposition of clade B� versus the grade including clades 2 and 3at the intersection of the large Old World group and the pre-dominately New World group (Figs. 2 and 3). There was nodiVerence in relationships of the Old World clades to eachother in any of the analyses. The Old World basal portion ofthe tree consisted of sister clades 5 and 6 and a large groupincluding clades C, D, and clade 4 (plus Centauropsis) (Figs. 2and 3). Bootstrap support and posterior probability valueswere similar for these groups as well, underscoring the stabil-ity of the clades in this portion of the tree.

4. Discussion

Turner (1981) was the Wrst to suggest a connectionbetween Old and New World Vernonieae involving Leibol-diinae taxa. Based on morphology Turner (1981, p. 403)said that section Lepidonia (now a genus within the Leibol-diinae Robinson (1999a)) “ƒ is as closely related to someof the Old World sections (e.g. Cyanopsis) as to those of theNew World sections.” He also noted that the chromosomenumber, nD19 for Stramentopappus pooleae supported anAfrican relationship. Old World Vernonieae typically havea chromosome number of nD 9 or 10 while New WorldVernonieae are typically nD 17 (Jones, 1977; Turner, 1981).Chemical data were somewhat equivocal, however. Ger-shenzon et al. (1984) found the Old World type of non-glaucolide germacranolides in Lepidonia jonesii while theNew World type of glaucolide was found in Stramentopap-pus pooleae. Robinson and Funk (1987) disagreed withTurner (1981) based on a morphological cladistic analysissaying the Leiboldiinae were autochthonous New Worldelements not related to Old World taxa. If Turner’s reason-ing is followed then Stokesia, found with the Leiboldiinaetaxa here as well as in these other studies, may similarlyhave ties to the Old World. With a chromosome number ofnD7 Stokesia is anomalous among New World taxa. Acount of nD 7 is otherwise known only from the Madaga-scan species, Vernonia appendiculata (Rabakonandrianinaand Carr, 1987). It is also tempting to consider

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 99

nD7 + nD 10 as a possible source of the nD17 common inNew World Vernonieae today.

4.1. Long distance dispersal and the role of Mesoamerica

Zavada and de Villiers (2000) reported the earliestunequivocal record of the Compositae (Asteraceae) at ca.60 MYA from Paleocene-Eocene sediments of SouthAfrica. This new date nearly doubles the most widelyaccepted earlier estimates of 30–40 MYA (Raven and Axel-rod, 1974) and 42–47 MYA recently proposed by Kim(2005). The family is undoubtedly Gondwanan (Bremerand Gustafsson, 1997; Zavada and de Villiers, 2000) andSouth American in origin (Jansen and Palmer, 1987;Bremer, 1994). However, several near-basal tribes, forexample the Vernonieae, Arctoteae and Cichorieae in sub-family Cichorioideae, are Old World in origin. Given therelatively young age of the family with respect to continen-tal separations (90–120 MYA (Scotese, 2002)) and itsworldwide distribution, long distance dispersal must haveplayed an important role in creating current distributionpatterns.

For the Vernonieae long distance dispersal is likely tohave been especially important. The Vernonieae originatedin the region of southern Africa/Madagascar. The sistertribe to the Vernonieae, the Liabeae (Keeley and Jansen,1994; Robinson, 1999a), is entirely New World. This is theonly such New/Old World sister relationship found amongtribes anywhere within the Compositae (Funk et al., 2005).Vernonieae taxa are now found from Argentina to Canadain the New World and throughout Africa, south and south-east Asia, and Australia in the Old World, and on islandchains in both hemispheres. Using the rates of 0.0007 nucle-otide changes/MY for ndhF and 0.0078 nucleotide changes/MY for ITS established by Kim et al. (1998) for the Vern-onieae, we estimate a date of ca. 14–20 MY for the majorradiation of Old to New World for the Vernonieae (ITS18.9§1.4 MY, ndhF 17.1§4.2 MY), values that are inagreement with those of Tremetsberger et al. (2005) for theage of the tribe. South America and Africa were already intheir current positions by this time (Scotese, 2002).

Long distance dispersal has been shown for severaltribes in the Compositae. Undoubtedly the best-docu-mented case is that of the Hawaiian silversword alliance(tribe Heliantheae) which originated 5–7 MYA from a Cal-ifornian ancestor (Baldwin, 1992; Baldwin and Sanderson,1998; Baldwin, 2003). In addition to the original coloniza-tion from the mainland, there has been spectacular adap-tive radiation (3 genera, 27 species) involving multipleinter-island dispersals (Carlquist et al., 2003). The Hawai-ian islands are volcanic, originating from an undersea hotspot and have never been connected to land (Clague andDalrymple, 1987). They are also among the most distantland masses in the world today (ca. 4000 km from the edgeof western North America the nearest land). Other recentlyreported examples of long distance dispersal in the Com-positae include that of Hypochaeris (tribe Cichorieae) from

Africa to South America (Tremetsberger et al., 2005) andSenecio (tribe Senecionieae) from Mediterranean Europe towestern North America (Coleman et al., 2003).

The Hawaiian endemic Hesperomannia (Vernonieae),like the silverswords, came from a distant mainland ances-tor, in this case from east Africa/Madagascar (Kim et al.,1998) and this paper (Figs. 2 and 3), a distance of almost12000 km. Since initial colonization there has also been dis-persal between the islands. Additionally, there are >30endemic Vernonieae in the West Indies (Keeley, 1978). TheCompositae is not the only group where long distance dis-persal is important. There are numerous other cases of longdistance dispersal for both plant and animal groups whoseyoung age precludes vicariance as an explanation for pres-ent day distributions. (e.g., Melastomataceae, Renner et al.,2001; Renner, 2004a; and other trans-Atlantic disjunctplant genera, Renner, 2004b; monkeys, Xightless insects,frogs, baobabs etc. summarized in de Queiroz, 2005; landsnails, Cowie and Holland, 2006).

Meso/Central America is an important region connect-ing Old and New World taxa in the Vernonieae (Figs. 2 and3). Over the past 15–16 MY a chain of volcanic islandsformed proto-Central America. When the islands joinedtogether to form a solid landmass and the Panamanianisthmus closed 3–3.5 MYA (Keigwin, 1982; Coates et al.,2003; Morley, 2003; Sanmartin and Ronquist, 2004) a land-based corridor for dispersal became established betweenNorth and South America. Dispersal overland once Northand South America were connected is well known (e.g., theGreat American Exchange (Marshall et al., 1979; Morley,2003)). Not all species would have moved at the same rateor necessarily in the same direction, however, and somemay not have moved at all. Changes in climate and Xuctua-tions in sea level likely created disjunct populations espe-cially in some of the higher elevation areas of Mesoamerica(particularly in southern Mexico and northern Guatemala).In turn, these isolated areas may have become refugia.Rzedowski (1993) pointed out that many of the lineagesnow found in the Neotropics might have arrived in Mexicoand the Neotropics from other parts of the world, withsome subsequently becoming extinct in important parts oftheir original area of distribution. Taxa whose historiesshow this pattern include, for example, northernhemisphere Liriodendron, Nyssa, and Tilia and southernhemisphere Ayenia, Coccoloba and Enterolobium (Graham,1993; Rzedowski, 1993).

Not all relicts leave clear fossils, but their past presencein an area can be reasonably inferred on the basis of theircurrent distributions and phylogenetic relationships. TheMesoamerican Vernonieae of subtribe Leiboldiinae (cladeB�, Figs. 2 and 3) are rare and restricted to small, widelyseparated populations in cloud forests and on wet moun-tain slopes of volcanoes from 800 to 2400 m (in southernMexico above the Isthmus of Tehuantepec (Oaxaca, Guer-erro) and northern Guatemala (Alta Vera Paz)) and in themountains of central Costa Rica (Cartago) (Turner, 1981).These areas were likely high enough to have remained

100 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

above water as parts of the volcanic island chain that madeup proto-Central America were variously submerged anduplifted. It is these relict taxa that show the greatest aYnityto Old World species (clade B�) and if the topology of Fig. 2is correct, established the basal lineage of Vernonieae in theNew World from Old World ancestors.

Movement of Vernonieae taxa in the New Worldappears to be from South to North America, with Meso/Central America once again playing an important role. Forexample, the Costa Rican Vernonieae species, Eirmocephalabrachiata, groups with the Venezuelan genus Tephrotham-nus and in a position between Brazilian taxa (clade A) andthe Mesoamerican/Brazilian/North American taxa of clade1. The genus Eirmocephala includes two other species (notsampled) that range from Panama, to Colombia and Vene-zuela and into northern Ecuador. Given the derived posi-tion of the North American clade it is not hard to imaginethat South American species migrated northward throughCentral America. A somewhat more complex history is sug-gested by the Mesoamerican taxon, Vernonanthura alamaniiin clade 1. This species is found at the base of the clade andabove which are derived Brazilian (and northern SouthAmerican) species, on the one hand, and North Americanspecies on the other. The Brazilian derivatives may reXect aback dispersal or perhaps V. alamanii is a relict taxon nowisolated from its southern relatives. In any case, the historyof Mesoamerican taxa is important in understanding thecurrent distribution of the Vernonieae. Rzedowski (1993)noted the importance of southern elements in the Xora ofMexico (and mega-Mexico including Mesoamerica) con-curring with Raven and Axelrod (1974) that an importantpart of the Mexican Xora must have originated in Centraland South America.

Understanding the history of the tribe will likely requirelooking at the interplay of land-based radiations and longdistance dispersals in Meso/Central America. This is sug-gested, for example, by the contrasting geographical distri-butions of the Mesoamerican taxa in the relictual subtribeLeiboldiinae, previously discussed, and the species of thewidespread genus Lepidaploa. In the latter case, Lepidaploatortuosa and L. borinquesis are both restricted to montaneCosta Rica, but are not each other’s closest relatives. Lep-idaploa tortuosa is basal to its congeners (at least those usedhere), including Lepidaploa ehretifolia from Venezuela. Thespecimen of the latter species was collected by the Wrstauthor from the top of Auyantepui, one of several isolatedremnants of the Guyana shield. Given its position in thetree, L. ehretifolia is more recently derived than a numberof other New World taxa and is apparently not a relicttaxon which might otherwise be suggested by its geographi-cal location. Other Lepidaploa species are derived from aBrazilian/Paraguayan group with roots in the Andes. Asmentioned earlier, there are also endemic species of Lepida-ploa in the West Indies, particularly in the older islands ofCuba and Jamaica, and one species, Lepidaploa arbores-cens, is present on several islands in the West Indies and inboth South and Central America (Keeley, 1978, 1982). Like

Hesperomannia, it is clear Lepidaploa species can disperseover signiWcant distances.

Given the multiple possible pathways for movementbetween North and South America one route, not yet men-tioned, does not seem to apply to the Vernonieae. This isthe Boreotropic pathway (TiVney, 1985). In this scenario aneast Asian (Laurasian) ancestral lineage is postulated tohave crossed into North America via one of the landbridges formed during the Tertiary with subsequent south-ward movement to tropical South America. Climatic eventsafter these initial radiations may have resulted in the elimi-nation of North American taxa thus leaving behind theextant members in South America. For taxa with African/South American connections formed in this way (i.e. dahl-bergioid legumes (Lavin et al., 2000)) the most derived spe-cies would then be South American. The distribution andphylogenetic histories of the Vernonieae precludes the tribefrom having this kind of history. The tribe is not found ineast Asia and is also lacking in Europe making a Laurasianancestry unlikely. Vernoniae are also found in south andsoutheast Asia, i.e., India, Thailand, Viet Nam, and parts ofMalayasia and adjacent southern China, but not north ofthe Himalayas. In addition, the most derived New Worldtaxa are in eastern North America, contrary to the predic-tions of the Boretropic hypothesis.

4.2. New World relationships

Within the New World lineages (clades A, 1, 2, 3) severaldiVerent patterns can be seen that create a complex net-work of relationships involving South, Central and NorthAmerican taxa. The relative positions of the clades remainthe same with the exception of clade B� discussed above(Figs. 2 and 3). Brazilian taxa of clade A form a sister groupto other Brazilian species and the grade of American taxathat culminates in clade 1. Within clade 1 itself, Braziliantaxa are also in the sister group to the monophyletic NorthAmerica Vernonia s.s (Robinson, 1999a). The basal mosttaxon of clade 1 is the morphologically distinctive V. ala-manii from southern Mexico (Mesoamerica) whose charac-ters are such that it may deserve separate genericrecognition (Robinson,pers. comm.).

The sister group to the combination of clade A and thegrade including clade 1 (discussed above) is composed ofAndean species (Critoniopsis) plus clades 2 and 3 (Fig. 2).This relationship suggests two separate major New Worldlineages despite the fact that both contain Brazilian taxa. Inthe Wrst case with dispersal from Brazil through Mesoamer-ica to North America, with back radiation to Brazil, and inthe second case radiation of taxa from western SouthAmerica with dispersal east to Brazil, and up to Centraland North America. Secondary radiations and exchangesmay also have occurred. For example, Lepidaploa speciesare found widely distributed in South America (Argentiana,Bolivia, Brazil, Colombia, Ecuador, Peru, Venezuela), Cen-tral America and southern Mexico, and across the WestIndies (Keeley, 1978, 1982; Robinson, 1999a). If these

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 101

originated in the Andes then there could have been longdistance dispersal directly to the West Indies and overlandspread to Brazil and nearby regions or overland spreadmay have occurred Wrst followed by dispersal to the islands.The events surrounding the uplift of the Andes 5–15 MYA(Funk et al., 2005) are also likely to have aVected the Vern-onieae by separating and perhaps recombining species pop-ulations and providing new pathways for dispersal. Inaddition, the geological history of the Antilles and proto-Central America is complex (Keigwin, 1982; Coates et al.,2003) providing opportunities for dispersal away fromSouth America at one time, and in the reverse direction atothers.

It is apparent from the other long distance dispersalevents discussed above, and the relationship among Ele-phantopus species that an exchange of taxa occurredbetween the New and Old Worlds on at least two occasionsand in two directions. One dispersal was from southeastAsia to the Americas (clade B�) and the other from theAmericas to the Old World (clade 3).

4.3. Old World relationships

The Vernonieae originated in the Old World in theregion of Madagascar/southern Africa (Keeley andTurner, 1990; Keeley and Jansen, 1994; Kim et al., 1998).The several African lineages sampled here provide betterresolution of the relationships among them. For example,Keeley and Jansen (1994) found Distephanus (as Vernoniapopulifolia) in a basal clade that also contained clade 5 spe-cies, Baccharoides adoensis, Gymnanthemum amygdalinum,G. mesipifolium (all considered species of Vernonia at thattime) and one species of clade 6, Linzia melleri (as Verno-nia glabra). Distephanus is now a separate genus (Robin-son, 1999b) and has been shown to be basal in the tribe, asWrst suggested in a morphologically-based study by Keeleyand Turner (1990). The position of taxa in clades C, D and4 (Figs. 2 and 3) is also better resolved as a sister group tothe combination of newly circumscribed genera Linzia,Gymnanthemum,and Hesperomannia (Robinson, 1999a,2007). The taxa in clades C and D are found broadly ineastern Africa (Burundi, Botswana, Ethiopia, Malawi,Rwanda, Zimbabwe) while those in clade 4 are found onlyin South Africa. Additionally, taxa shared by the presentstudy and that of Keeley and Jansen (1994) remain inmuch the same relationship to each other. For example,Centrapalus galamensis and Parapolydora fastigiata (cladeD) remain sister taxa and the three species, Orbivestus cin-arescens, Baccharoides lasiopus and V. brachycalyx (previ-ously treated as species of Vernonia) are members of thesame clade as in Keeley and Jansen (1994). None of theHilliardiella (clade 4) species were included in Keeley andJansen (1994) so their position was not previouslydescribed. Robinson (1999b) placed taxa from these clades(C, D, 4) within a broadly constructed subtribe, the Erlan-geinae. Relationships within the Old World taxa sampledto date are stable.

5. Conclusions

The data in this study reveal several important featuresof Vernonieae phylogeny and biogeography.

(1) New and Old World lineages are not separate. Herethey occur together in two diVerent and distantlyrelated clades.

(2) Long distance dispersal has been important in New/Old World exchange and has occurred at least twiceand at least once in each direction.

(3) Mesoamerica has played an important role in bothancient and modern dispersals, and is a refuge forancient taxa.

(4) The direction of movement in the New World has beenpredominately from South America to North America.

(5) North American species originated from at least twowidely separated lines of New World taxa, and bothare recently derived within the tribe.

(6) The Vernonieae are Gondwanan and dispersal didnot follow a Boretropic route.

The Old World tribe Vernonieae is the only member ofthe Compositae to have a New World sister tribe (the Lia-beae (Funk et al., 2005)). Therefore, it is reasonable to sus-pect a more visible connection between New and OldWorld members of the Vernonieae than may exist for mostother tribes in the family. The phylogeny of the tribe pointsto the family wide need to focus on speciWc biogeographi-cally important regions where taxa from Old and NewWorlds meet. Long distance dispersal is important in thefamily and there are many possible source areas, times andpathways to be further investigated. The Vernonieae oVeran unusually good system to understand the history of thefamily; no mean feat for an evil tribe.

Acknowledgments

We thank Vicki Funk, Harold Robinson, Elizabeth Zim-mer, Kimberley Peyton and the reviewers for helpful com-ments and technical support, and the individual collectorsand the curators of K, MO and US for materials. Supportwas provided by National Science Foundation Grant DEB-0075095 to S.C.K., the Laboratory of Molecular Systemat-ics and the Department of Botany, National Museum ofNatural History, and by the College of Natural Sciences,University Research Council and the Department of Bot-any, University of Hawaii.

References

Akaike, H., 1974. A new look at statistical model identiWcation. IEEETrans. Auto. Control 19, 716–723.

Álvarez, I., Wendel, J.F., 2003. Ribosomal ITS sequences and plant phylo-genetic inference. Mol. Phylogenet. Evol. 29, 417–434.

Baldwin, B.G., 1992. Phylogenetic utility of the internal transcribed spacerof nuclear ribosomal DNA in plants: an example from the Compositae.Mol. Phylogenet. Evol. 1, 3–16.

102 S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103

Baldwin, B.G., 2003. A phylogenetic perspective on the origin and evolu-tion of Madiinae. In: Carlquist, S., Baldwin, B.G., Carr, G.D. (Eds.),Tarweeds and Silverswords. Evolution of the Madiinae (Asteraceae).Missouri Botanical Garden Press, St. Louis, pp. 193–228.

Baldwin, B.G., Sanderson, M.J., 1998. Age and rate of diversiWcation of theHawaiian silversword alliance (Compositae). Proc. Natl . Acad. Sci .USA. 95, 9402–9406.

Baldwin, B.G., Wessa, B.L., Panero, J.L., 2002. Nuclear rDNA evidence formajor lineages of Helenioid Heliantheae (Compositae). Syst. Bot. 27,161–198.

Bayer, R.J., Starr, J.R., 1998. Tribal phylogeny of the Asteraceae based ontwo non-coding chloroplast sequences, the trnL intron and trnL/trnFintergenic spacer. Ann. Missouri Bot. Gard. 85, 259–272.

Bayer, R.J., Puttock, C.F., Kelchner, S.A., 2000. Phylogeny of South Afri-can Gnaphalieae (Asteraceae) based on two non-coding chloroplastsequences. Amer. J. Bot. 87, 259–272.

Bentham, G., 1873. Compositae. In: Bentham, G., Hooker, J.D. (Eds.),Genera Plantarum 2. Reeve and Company, London, pp. 163–533.

Bremer, K., 1994. Asteraceae: Cladistics and ClassiWcation. Timber Press,Portland, OR. pp. 202–233.

Bremer, K., Gustafsson, M.H.G., 1997. East Gondwana ancestry of the sun-Xower alliance of families. Proc. Nat. Acad. Sci. USA 94, 9188–9190.

Carlquist, S., Baldwin, B.G., Carr, G.D. (Eds.), 2003. and Silverswords.Evolution of the Madiinae (Asteraceae). Springer, St. Louis, p. 293.

Cassini, H., 1819. Sur la famille des Synanthérées contenant les caractèresdes tribus. J. Phys. Chim. Hist. Nat. Arts. 88, 190–204.

Cassini, H., 1828. Vernoniées. In: Cuvier, G.L. (Ed.), Dictionnaire des Sci-ences Naturelles, Vol. 57. Le Normant, Paris, pp. 338–347.

Chan, R., Baldwin, G., OrnduV, R., 2001. GoldWelds revisited: a molecularphylogenetic perspective on the evolution of Lasthenia (Compositae:Heliantheae sensu lato). Inter. J. Plant Sci. 152, 1347–1360.

Chen, W.-J., Bonillo, C., Lecointre, G., 2003. Repeatability of clades as acriterion of reliability: a case study for molecular phylogeny of Acanth-omorpha (Teleostei) with larger number of taxa. Mol. Phylogenet.Evol. 26, 262–288.

Clague, D.A., Dalrymple, G.B., 1987. The Hawaiian-Emperor volcanicchain. In: Decker, R.W., Wright, T.L., StauVer, P.H. (Eds.), Volcanismin Hawaii. US Geol. Surv. Prof. Paper 1350. US. Government PrintingOYce, Washington, D.C, pp. 1–54.

Coates, A.G., Aubry, M.-B., Berggre, W.A., Collins, L.S., Kunk, M., 2003.Early Neogene history of the Central American arc from Bocas delToro, western Panama. Geol. Soc. Amer. Bull. 115, 271–287.

Coleman, M., Liston, A., Kadereit, J.W., Abbott, R.J., 2003. Repeat inter-continental dispersal and Pleistocene speciation in disjunct Mediterra-nean and desert Senecio (Asteraceae). Am. J. Bot. 90, 1446–1454.

Cowie, R.H., Holland, B.S., 2006. Dispersal is fundamental to biogeogra-phy and the evolution of biodiversity on oceanic islands. J. Biogeogr.33, 193–198.

de Queiroz, A., 2005. The resurrection of oceanic dispersal in historicalbiogeography. TREE 20, 68–73.

Downie, S.R., Katz-Downie, D.S., 1996. A molecular phylogeny of Apia-ceae subfamily Apioideae: evidence from nuclear ribosomal DNAinternal transcribed spacer sequences. Amer. J. Bot. 83, 234–251.

Farris, J.S., Kallersjo, M., Kluge, A.G., Bult, C., 1994. Testing signiWcanceof incongruence. Cladistics 10, 315–319.

Funk, V.A., Chan, R., Keeley, S.C., 2004. Insights into the evolution of thetribe Arctoteae (Compositae:subfamily Cichorioideae s. s. ) using trnL-F , ndhF, and ITS. Taxon 53, 637–655.

Funk, V.A., Bayer, R.J., Keeley, S., Chan, R., Watson, L., Gemeinholzer,B., Schilling, E., Panero, J.L., Baldwin, B.G., Garcia-Jacas, N., Susanna,A., Jansen, R.K., 2005. Everywhere but Antarctica: using a supertree tounderstand the diversity and distribution of the Compositae. Biol. Skr.55, 343–374.

Gershenzon, J., Pfeil, R.M., Liu, Y.L., Mabry, T.J., Turner, B.L., 1984. Ses-quiterpene lactones from two newly-described species of Vernonia: V.jonesii and V. pooleae. Phytochemistry 22, 777–780.

Geyer, C.J., 1991. Markov chain Monte Carlo maximum likelihood. In:Keramidas, E.M. (Ed.), Computing science and statistics: Proceedings

of the 23rd Symposium on the Interface. Interface Foundation, FairfaxStation, VA, pp. 156–163.

Gleason, H.A., 1906. A revision of the North American Vernonieae. Bull.New York Bot. Gard. 4, 144–243.

Goertzen, L.R., Cannone, J.J., Gutell, R.R., Jansen, R.K., 2003. ITS sec-ondary structure derived from comparative analysis: implications forsequence alignment and phylogeny of the Asteraceae. Mol. Phylogenet.Evol. 29, 216–234.

Graham, A., 1993. Historical factors and biological diversity in Mexico.In: Ramamoorthy, T.P., Bye, R., Lot, A., Fa, J. (Eds.), Biological diver-sity of Mexico; origins and distribution. Oxford University Press, Lon-don, pp. 109–123.

Hulsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference ofphylogenetic trees. Bioinformatics 17, 754–755.

Jansen, R.K., 1992. Current research. Pl. Molec. Evol. Newsl. 2, 13–14.Jansen, R.K., Palmer, J.D., 1987. A chloroplast inversion marks an ancient

evolutionary split in the sunXower family (Asteraceae). Proc. Natl.Acad. Sci. USA 84, 5818–5822.

JeVrey, C., 1988. The Vernonieae of east tropical Africa. Notes on theCompositae: V. Kew Bull. 43, 195–277.

Jones, S.B., 1977. Vernonieae—Systematic Review. In: Heywood, V.H.,Harborne, J.B., Turner, B.L. (Eds.), The Biology and Chemistry of theCompositae. Academic Press, London, pp. 503–521.

Jones, S.B., 1979. Synopsis and pollen morphology of Vernonia (Composi-tae: Vernonieae) in the New World. Rhodora 81, 425–447.

Jones, S.B., 1981. Synoptic classiWcation and pollen morphology of Verno-nia (Compositae: Vernonieae) in the Old World. Rhodora 83, 59–75.

Keeley, S.C., 1978. A revision of the West Indian vernonias (Compositae).Journal of the Arnold Arboretum 59, 360–413.

Keeley, S.C., 1982. Morphological variation and species recognition inthe neotropical taxon Vernonia arborescens (Compositae). Syst. Bot,7:, 71–84.

Keeley, S.C., Jones, S.B., 1979. Distribution of pollen types in Vernonia(Vernonieae:Compositae). Syst. Bot. 4, 195–202.

Keeley, S.C., Turner, B.L., 1990. A preliminary cladistic analysis of thegenus Vernonia (Vernonieae:Asteraceae). Plant Syst. Evol. 4 (Supp l),45–66.

Keeley, S.C., Jansen, R.K., 1994. Chloroplast restriction site variation inthe Vernonieae (Asteraceae), an initial appraisal of the relationship ofNew and Old World taxa and the monophyly of Vernonia. Plant Syst.Evol. 193, 249–265.

Keigwin, L.D., 1982. Isotopic paleoceanography of the Caribbean andeast PaciWc: role of Panama uplift late Neogene time. Science 217,350–352.

Kim, H.-G., Keeley, S.C., Vroom, P.S., Jansen, R.K., 1998. Molecular evi-dence for an African origin of the Hawaiian endemic Hesperomannia(Asteraceae). Proc. Natl. Acad. Sci. USA 95, 15440–15445.

Kim, H.-G., Loockerman, D.J., Jansen, R.K., 2002. Systematic implicationsof ndhF sequence variation in the Mutisieae (Asteraceae). Syst. Bot. 27,598–609.

Kim, K.-J., Choi, K.-S., Jansen, R.K., 2005. Two chloroplast DNA inver-sions originated simultaneously during the early evolution of the sun-Xower family (Asteraceae).

Kim, K.-J., Jansen, R.K., 1995. ndhF sequence evolution and themajor clades in the sunXower family. Proc. Natl. Acad. Sci. USA 92,10379–10383.

Kumar, S., Tamura, K., Nei, M., 2004. MEGA3: an integrated software formolecular evolutionary genetics analysis and sequence alignment.Brief. Bioinform. 5, 150–163.

Lavin, M., Thulin, M., Labat, J.-N., Pennington, R.T., 2000. Africa, the oddman out: molecular biogeographic studies of dalbergioid legumes(Fabaceae) suggest otherwise. Syst. Bot. 25, 449–467.

Loytynoja, A., Milinkovitch, M.C., 2003. A hidden Markov model for pro-gressive multiple alignment. Bioinformatics 19, 1505–1513.

Marcos, S., Baldwin, B.G., 2001. Higher-level relationships and major lin-eages of Lessingia (Compositae, Astereae) based on nuclear rDNAinternal and external transcribed spacer (ITS and ETS) sequences. Syst.Bot. 26, 168–183.

S.C. Keeley et al. / Molecular Phylogenetics and Evolution 44 (2007) 89–103 103

Marshall, L.G., Butler, R.F., Drake, R.E., Curtis, G.H., Tedfpord, R.H., 1979.Calibration of the great American interchange. Science 204, 272–279.

Morley, R.J., 2003. Interplate dispersal paths for megathermal angio-sperms. Perspect Plant Ecol. Evol. System. 6, 5–20.

Olmstead, R.G., Sweere, J.A., 1994. Combining data in phylogenetic sys-tematics: an empirical approach using three molecular data sets in theSolanaceae. Syst. Biol. 43, 467–481.

Olmstead, R.G., Kim, K.-J., Jansen, R.K., WagstaV, S.J., 2000. The phylog-eny of the Asteridae sensu lato based on chloroplast ndhF genesequences. Mol. Phylogenet. Evol. 16, 96–112.

Panero, J., Funk, V.A., 2002. Toward a phylogenetic subfamilial classiW-cation for the Compositae (Asteraceae). Proc. Biol. Soc. Washington115, 909–922.

Pfeil, B.E., Brukabker, C.L., Craven, A., Crisp, M.D., 2002. Phylogeny ofHibiscus and the Hibisceae (Malvaceae) using chloroplast DNAsequences of ndhF and rpl16 intron. Syst. Bot. 27, 333–350.

Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNAsubstitution. Bioinformatics 14, 817–818.

Rabakonandrianina, E., Carr, G.D., 1987. Chromosome numbers of Mad-agascar plants. Ann. Missouri Bot. Gard. 74, 123–125.

Raven, P.H., Axelrod, D.I., 1974. Angiosperm biogeography and past con-tinental movements. Ann. Missouri Bot. Gard. 61, 539–673.

Renner, S., 2004a. Bayesian analysis of combined chloroplast loci, usingmultiple calibrations, supports the recent arrival of Melastomataceaein Africa and Madagascar. Amer. J. Bot. 91, 1427–1435.

Renner, S., 2004b. Plant dispersal across the tropical Atlantic by wind andsea currents. Int. J. Plant. Sci. 165 (4 Suppl), S23–S33.

Renner, S.S., Clausing, G., Meyer, K., 2001. Historical biogeography ofMelastomataceae: the roles of tertiary migration and long-distance dis-persal. Amer. J. Bot. 88, 1290–1300.

Roberts, R.P., Urbatsch, L.E., 2003. Molecular phylogeny of Ericameria(Asteraceae, Astereae) based on nuclear ribosomal 3 ETS and ITSsequence data. Taxon 52, 209–228.

Roberts, R.P., Urbatsch, L.E., 2004. Molecular phylogeny of Chrysotham-nus and related Genera (Asteraceae, Astereae) based on nuclear ribo-somal 3� ETS and ITS sequence data. Syst. Bot. 29, 199–215.

Robinson, H., 1999a. Generic and subribal classiWcation of AmericanVernonieae. Smithsonian Contrib. Bot. 89, 1–116.

Robinson, H., 1999b. Revision of paleotropical Vernonieae (Asteraceae).Proc.Biol. Soc. Washington (USA) 112, 220–247.

Robinson, H., 2007. Vernonieae. In: Kadereit, J., JeVrey, C. (Eds.), Vol. 8:Asterales. The families and genera of vascular plants (K. Kubitzki,Series Ed.). Springer, Berlin, Heidelberg, New York, pp. 149–174.

Robinson, H., Kahn, B., 1986. Trinervate leaves, yellow Xowers, tailedanthers, and pollen variation in Distephanus Cassini (Vern-onieae:Asteraceae). Proc. Biol. Soc. Washington 99, 493–495.

Robinson, H., Funk, V.A., 1987. A phylogenetic analysis of Leiboldia,Lepidonia, and a new genus Stramentopappus (Vernonieae: Astera-ceae). Bot. Jahrb. Syst. 108, 213–228.

Rzedowski, J., 1993. Diversity and origins of the phanerogamic Xora ofMexico. In: Ramamoorthy, T.P., Bye, R., Lot, A., Fa, J. (Eds.), Biologi-cal diversity of Mexico; origins and distribution. Oxford UniversityPress, London, pp. 129–144.

Samuel, R., Stuessy, T.D., Tremetsberger, L., Baeza, C.M., Siljak-Yakov-lev, S., 2003. Phylogenetic relationshps among species of Hypochaeris(Asteraceae:Cichorideae) based on ITS, plastid trnL intron, trnL-Fspacer and matK sequences. Amer. J. Bot. 90, 496–507.

Sanmartin, I., Ronquist, F., 2004. Southern hemisphere biogeographyinferred by event-based models: plant versus animal patterns. Syst.Biol. 54, 216–243.

Scotese, C.R. 2002. PALEOMAP website. (http://www.scotese.com).Shaw, J., Lickey, E.B., Beck, J.T., Farmer, S.B., Liu, W., Miller, J., Siripun,

K.C., Winder, C.T., Schilling, E.E., Small, R.L., 2005. The tortoise andthe hare II: relative utility of 21 noncoding chloroplast DNA sequencesfor phylogenetic analysis. Amer. J. Bot. 92, 142–166.

Simmons, M.P., Pickett, K.M., Miya, M., 2004. How Meaningful areBayesian support values? Mol. Biol. Evol.s 21 (1), 188–199.

Suzuki, Y., Glasko, G.V., Nei, M., 2002. Overcredibility of molecular phy-logenies obtained by Bayesian phylogenetics. Proc. Natl. Acad. Sci.USA 99, 16138–16143.

SwoVord, D.L. 2002. PAUP*: Phylogenetic analysis using parsimony (* andother methods), version 4.0b10. Sinauer Press, Sunderland, MA.

Taberlet, P., Gielly, L., Pautou, G., Bouvet, J., 1991. Universal primers forampliWcation of three non-coding regions of chloroplast DNA. PlantMolec. Bio. 17, 1105–1109.

TiVney, B.H., 1985. The Eocene North Atlantic land bridge: its importancein tertiary and modern phytogeography of the Northern Hemisphere.J. Arnold Arb. 66, 243–273.

Tremetsberger, K., Weiss-Schneeweiss, H., Stuessy, T., Samuel, R.,Kadlec, G., Ortiz, M.A., Talavera, S., 2005. Nuclear ribosomalDNA and kayotypes indicate a NW African origin of South Ameri-can Hypochaeris (Asteraceae, Cichorieae). Mol. Phylogenet. Evol.35, 102–116.

Turner, B.L., 1981. New species and combinations in Vernonia sectionsLeiboldia and Lepidonia (Asteraceae), with a revisional conspectus ofthe groups. Brittonia 33, 401–412.

White, T.J., Bruns, T., Lee, S., Taylor, J., 1990. AmpliWcation and directsequencing of fungal ribosomal RNA genes for phylogenetics. In: Innics,M., Glfand, D., Sninsky, J., White, T.J. (Eds.), PCR protocols: A guide tomethods and applications. Academic Press, San Diego, pp. 315–322.

Zavada, M.S., de Villiers, S.E., 2000. Pollen of the Asteraceae from thePaleocene–Eocene of South Africa. Grana 39, 39–45.