ELSEVIER Available online at www.sciencedirect.com %' ScienceDirect Molecular Phylogenetics and Evolution 47 (2008) 757-782 MOLECULAR PHYLOGENETICS AND EVOLUTION www.elsevier.com/locate/ympev The value of sampling anomalous taxa in phylogenetic studies: Major clades of the Asteraceae revealed Jose L. Paneroa'*, V.A. Funkb a Section of Integrative Biology, 1 University Station, A6700, 141 Patterson Building, University of Texas, Austin, TX 78712, USA b US National Herbarium, Department of Botany, P. O. Box 37012, Smithsonian Institution MRC 166, Washington, DC 20013-7012, USA Received 17 August 2007; revised 11 February 2008; accepted 12 February 2008 Available online 28 March 2008 Abstract The largest family of flowering plants Asteraceae (Compositae) is found to contain 12 major lineages rather than five as previously suggested. Five of these lineages heretofore had been circumscribed in tribe Mutisieae (Cichorioideae), a taxon shown by earlier molec- ular studies to be paraphyletic and to include some of the deepest divergences of the family. Combined analyses of 10 chloroplast DNA loci by different phylogenetic methods yielded highly congruent well-resolved trees with 95% of the branches receiving moderate to strong statistical support. Our strategy of sampling genera identified by morphological studies as anomalous, supported by broader character sampling than previous studies, resulted in identification of several novel clades. The generic compositions of subfamilies Carduoideae, Gochnatioideae, Hecastocleidoideae, Mutisioideae, Pertyoideae, Stifftioideae, and Wunderlichioideae are novel in Asteraceae systematics and the taxonomy of the family has been revised to reflect only monophyletic groups. Our results contradict earlier hypotheses that early divergences in the family took place on and spread from the Guayana Highlands (Pantepui Province of northern South America) and raise new hypotheses about how Asteraceae dispersed out of the continent of their origin. Several nodes of this new phylogeny illustrate the vast differential in success of sister lineages suggesting focal points for future study of species diversification. Our results also provide a backbone exemplar of Asteraceae for supertree construction. ? 2008 Elsevier Inc. All rights reserved. Keywords: Asteraceae; Compositae; Mutisieae; Phylogeny; Taxon sampling; Supermatrix; Supertree; Speciation; Sweepstakes dispersal; Guayana Hig- hlands 1. Introduction Well-resolved and statistically well-supported phytoge- nies are essential for many kinds of evolutionary studies, but such an estimate of relationships between genera and higher taxa is still needed for the Asteraceae, the largest family of vascular plants. With more than 23,600 species, this family constitutes approximately 8% of all flowering plants and is distributed on all continents of the world except Antarctica (Stevens, 2001 onwards). Asteraceae pro- vides an excellent opportunity to understand adaptation in the recent radiation of a plant group at a global scale. The Corresponding author. Fax: +1 512 471 3878. E-mail address: panero@mail.utexas.edu (J.L. Panero). secondary chemistry, inflorescence morphology, and habit plasticity of Asteraceae are characteristics routinely assumed to be responsible for the worldwide success of the family (Carlquist, 1976; Hendry 1996; Stuessy and Car- ver, 1996). Polyploidy has also been associated with an increase in speciation rates (Vamosi and Dickinson, 2006) and this phenomenon could be responsible for the large number of species in several clades of the Asteraceae (e.g. Heliantheae alliance). The flowers and stems of many spe- cies of Asteraceae are hosts to numerous insect species in parasitic or mutualistic relationships of which only a few have been documented or studied (Ronquist and Liljeblad, 2001; Craig et al., 2007). The specialized inflorescence of Asteraceae, or capitulum, and the secondary pollen presen- tation mechanism of its flowers have evolved in concert, 1055-7903/$ - see front matter ? 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2008.02.011 758 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 presumably in response to several conspicuous evolution- ary pressures including herbivory, pollination and/or fruit dispersal. These interactions have resulted in the evolution of several breeding systems (Ferrer and Good-Avila, 2007) and an astonishing array of morphologies used in classifi- cation but still largely unexplained in an evolutionary con- text. Testing assumptions that specific traits are key innovations (Schluter, 2000) or tracing the transition of character suites that lead to diversification (Donoghue, 2005) requires a phylogenetic model that includes at a min- imum all major lineages of the family. Our ability to inves- tigate macroevolutionary patterns and ultimately the processes and timing associated with the worldwide radia- tion of sunflowers cannot progress without phylogenies that sample sufficient characters and taxa to offer strongly supported hypotheses of relationship and include many more branches of the sunflower tree. In spite of a long history of taxonomic work and more recently of molecular phylogenetic studies, not all the major lineages of the Asteraceae have been identified, and their relationships are certainly not known with confi- dence. Both Carlquist (1976) and Wagenitz (1976) reviewed the morphology of the family and concluded that the tribes of the Asteraceae could be grouped into just two main branches, namely subfamilies Cichorioideae and Asteroi- deae. Subsequent molecular studies based on restriction fragment length polymorphisms of the chloroplast genome identified three genera of Cichorioideae (Mutisieae: Bar- nadesiinae) that lacked a 22-kb inversion common to all other sunflowers (Jansen and Palmer, 1987). Bremer and Jansen (1992) erected subfamily Barnadesioideae for these and six additional genera establishing a three-subfamily system of classification. Analysis of chloroplast ndhF gene sequences (Kim and Jansen, 1995) showed Barnadesioideae as sister to all other sunflowers comprising a monophyletic Asteroideae and a paraphyletic Cichorioideae. Their study recovered a phylogenetic tree placing members of tribe Mutisieae in a paraphyletic grade of three clades diverging early in the family and a paraphyletic tribe Cynareae sister to the rest of Cichorioideae and Asteroideae. These results were used by Bremer (1996) to resurrect subfamily Car- duoideae to accommodate the Cynareae lineage at a rank comparable to the Barnadesioideae, Cichorioideae, and Asteroideae. Recognition of Barnadesioideae and Carduoi- deae implied that the other lineages of the Mutisieae grade would need recognition at the subfamily level as well, once their relationships are clarified (Bremer, 1996). There have been surprisingly few family-wide molecular studies since Kim and Jansen (1995) and these (Bayer and Starr, 1998; Goertzen et al., 2003) have not improved significantly our understanding of relationships among the major lineages of the Asteraceae. Prior to the present study the family has been considered to have five primary branches formally recognized as subfamilies Asteroideae, Barnadesioideae, Carduoideae, Cichorioideae, and tribe Mutisieae (Bremer, 1996). The monophyly of at least one of these, the Muti- sieae, cannot be supported (Kim and Jansen, 1995) and the "Mutisieae problem" has been viewed as the key to resolving Asteraceae systematics at higher ranks. In fact, the monophyly of most tribes, not only of Muti- sieae, remains to be tested rigorously with molecular data. Attempts at natural classification of Asteraceae began with Cassini (1819a,b) who placed genera into tribes and associ- ated these tribes based on their collective morphological similarities. Lost in obscurity, then resurrected by Bentham (1873), Cassini's concepts of the tribal lineages of Astera- ceae were the focus of modern debate and refinement until the early 1990s (Poljakov, 1967; Carlquist, 1976; Wagenitz, 1976; Cronquist, 1977; Jeffrey, 1978; Robinson, 1983; Thome, 1983; Bremer, 1987, 1994) and probably the most important tool for understanding lineages before molecular systematics. DNA studies have only begun to test the 17 tribes recognized by Bremer (1994) but have already clari- fied the limits of the Anthemideae (Watson et al., 2000), Chaenactideae, Eupatorieae, Helenieae, and Madieae (Baldwin and Wessa, 2000). These five tribes appear to be monophyletic. Tribe Cynareae, monophyletic in previ- ous molecular and morphological studies based on broad taxon sampling (Susanna et al., 2006, other references therein), has recently been shown to include the anomalous taxon Dipterocome (Anderberg et al. 2007), left as unplaced in Asteroideae by Bremer (1994). A number of studies have identified misplaced genera and refined tribal circumscrip- tions (Feddea, Cariaga et al., 2008; Gymnarrhena, Zout- pansbergia, Callilepis, Anderberg et al., 2005), clarified the phylogenetic position of unplaced genera {Cratystylis, Bayer and Cross 2003; Hoplophyllum, Karis et al. 2001) or identified genera whose tribal placements are equivocal {Heterolepis, Platycarpha, Funk et al., 2004; Jaumea, Bald- win et al., 2002). In addition to placing anomalous genera, i.e. those without synapomorphies of their presumed tribe, molecular studies have identified genera that are transi- tional between tribes (Abrotanella, Wagstaff et al., 2006) or astonishingly, several transitional genera found to con- stitute a heterogeneous but monophyletic tribe (Athrois- meae, Panero, 2007a). Carlquist (1976) drew attention to both misplaced (anomalous) genera and transitional genera ("non-missing" links) as the major obstacles hindering con- struction of a new comprehensive classification of Astera- ceae. Most of the studies listed above have been unable for the most part to clarify the phylogenetic positions of anomalous or transitional genera. We initiated this study to construct a robust phyloge- netic hypothesis of Asteraceae at the subfamily and tribal levels that would aid in constructing a classification recog- nizing only monophyletic groups and facilitate future evo- lutionary studies. In particular, we sought to define a monophyletic Mutisieae, since previous studies had brought the traditional circumscription into question but had not provided the resolution or statistical support needed to make taxonomic changes. Since a number of important taxa had not been sampled, our study includes the key mutisioid genera Hecastocleis and Wunderlichia as well as representatives of the Guayana Highlands J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 759 Mutisieae, some hypothesized to be among the deepest divergences of the tribe (Karis et al., 1992). The Guayana Highlands1 have a high proportion of endemic taxa and are considered to be an autochthonous plant evolution cen- ter for South American plants and mutisioid Asteraceae in particular (Huber, 2005). We have also included Corymbi- um, Gymnarrhena, and Warionia, identified by Bremer (1994) as of uncertain position in subfamily Cichorioideae. We avoid the lack of resolution and statistical support of previous studies by implementing a multi-gene strategy sampling nucleotide sequences from 10 loci of the plastome (cpDNA), representing more than 13,000 characters, so that we might obtain a fully-resolved and statistically sup- ported phylogenetic tree for Asteraceae. Novel, well-sup- ported groups identified by our studies and lacking taxonomic recognition are formally named in Panero and Funk (2002) and Panero and Funk (2007). 2. Materials and methods 2.1. Taxon sampling Taxa sampled for this study were chosen to represent all 17 tribes of the family recognized by Bremer (1994) with denser sampling within tribe Mutisieae. Special emphasis was given to include anomalous genera of Cichorioideae unplaced to tribe by Bremer (1994) including Adenocaulon, Corymbium, Gymnarrhena, Hecastocleis and Warionia. A total of 108 taxa, 56 of these representing the Mutisieae sensu Cabrera (1977) were sampled. The Asteraceae are monophyletic and sister to family Calyceraceae (DeVore and Stuessy, 1995; Kim and Jansen, 1995; Lundberg and Bremer, 2003). We chose Acicarpha (Calyceraceae) and Scaevola (Goodeniaceae) to serve as the outgroup. Appen- dix A lists specimens with their collection localities and her- baria where vouchers are deposited with corresponding Genbank accession numbers for the sequences used in this study. 2.2. DNA sampling and sequencing Total genomic DNA was isolated from field-collected leaves preserved in liquid nitrogen, CTAB solution, or sil- ica using the CTAB method (Doyle and Doyle, 1987) mod- ified to include a 1 volume phenol-chloroform-isoamyl alcohol extraction, resuspending DNA in water-7 M sodium acetate (10:1), precipitating the DNA with 1 vol- 1 Also known as the Pantepui Province (Huber, 1987), this distinctive phytogeographic region comprises more than 25 Roraima sandstone table mountains (tepuis) on the Guayana Shield (Maguire, 1956) between 1500 and 3000 m (Rull, 2004). The Guayana Highlands lie mostly within the Venezuelan Guayana and adjacent northern Brazil and western Guyana, north of the Amazon and south of the Orinoco River (Maguire, 1956). Isolated from surrounding lowlands by steep base slopes and vertical walls, and mostly inaccessible except by helicopter, the tabular summits covered with a diverse and highly endemic flora were described as a fictional "Lost World" by Arthur Conan Doyle in 1912. ume of ethanol, followed by two 70% ethanol washes. Her- barium samples were isolated using the DNeasy? Plant Mini Kit from Qiagen, Inc. (Qiagen, Valencia, California, USA). Chloroplast loci sequenced in this study include the genes matK, ndhD, ndhF, ndhl, rbcL, rpoB, and exon 1 of rpoC, as well as the trnF-trnF and 23S-trnA intergenic spacer regions, and the 5' portion of the trnK split intron. The rbcL gene was the first and most exploited marker for plant phylogenetic studies and continues to be employed above the genus level (Ritland and Clegg, 1987; Saarela et al., 2007). The ndhF gene has been utilized in earlier studies aimed at elucidating the major branches of the Asteraceae (Kim and Jansen, 1995; Jansen and Kim, 1996) and more recently the utility of matK has been shown in numerous intergeneric and infrageneric studies within the family (Susanna et al. 2006; Wagstaff et al., 2006; Watanabe et al., 2006). The genes ndhD and ndhl are part of the same gene family of ndhF, therefore they were explored in our study because of the known phyloge- netic utility of the ndhF subunit in previous sunflower DNA phylogenetic reconstructions. Preliminary studies of the Heliantheae alliance (Panero et al., unpublished) found that the co-transcribed polymerase genes rpoB and rpoC also provide many informative characters with low levels of homoplasy. We favor protein coding sequence over non- coding markers for their reduced ambiguity of alignment. However, noncoding data provide informative characters and have been used extensively in Asteraceae systematics, so approximately one-fifth of the characters we analyzed come from noncoding regions. DNA fragments were amplified by Polymerase Chain Reaction (PCR) using primers described in Panero and Crozier (2003) or as other- wise noted. The ndhF gene was amplified in two segments using primers 52 and 1212 (52: 5-AGG TAA GAT CCG GTG AAT CGG AAA-3', Jansen, 1992; 1212R: 5-GGT GGA ATA CCA CAA AGA-3') and primers 972F and 607 (972F: 5-GTC TCA ATT GGG TTA TAT GAT G-3'; 607: 5-ACC AAG TTC AAT GTT AGC GAG ATT AGT C-3', Jansen, 1992). All primers except 607 were used as sequencing primers also. Primer 1587F (5'-CCA ACC CTT TCT TTC TAT TCC G-3') was used to sequence the last segment of the gene. The genus Inula contains an extra codon in the primer region so primer 1587Inula (5'- CCA ACC CTT TCT TTC TAT TCC TCC G-3') was used instead. The matK gene was sequenced in three segments using primers 3914F and 884R (3914F: 5'-TGG GTT GCT AAC TCA ATG G-3', Johnson and Soltis, 1994; 884R: 5-TGT CAT AAC CTG CAT TTT CC-3'), primers 816F and 1857R (816F: 5-ATC TTT CAG GAG TAT ATT TAT G-3'; 1857R: 5-CCA GAG GCA TAA TTG GAA C-3'), and primers 1755F and trnK2R (1755F: 5'- TCC TAT TTT TAC CTG TGG TCT CA-3'; trnK2R: 5-AAC TAG TCG GAT GGA GTA G-3', Johnson and Soltis, 1994). These primers produce a complete sequence 760 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 of the gene. A partial sequence of the 5'-trnK intron as read by primer 884R was also used in the phylogenetic analysis. Primers trnK2R and 3914F were not used in sequencing. The rbch gene was amplified in two segments using primers rbcLl and rbcL911R (rbcLl: 5-ATG TCA CCA CAA ACA GAG ACT AAA GC-3' Hillis et al., 1996; rbcL911R: 5-TTT CTT CGC ATG TAG CCG C-3') and primers rbcL876F and rbcL2 (rbcL876F: 5-CAG GTG AAA TCA AAG GGC-3'; 5-CTT TTA GTA AAA GAT TGG GCC GAG-3'; Olmstead et al., 1992). All PCR primers were used as sequencing primers. The ndKD gene was amplified in two segments using primers ndhDF and 732R (ndhDF: 5-TTC GAC CTT GTC AAC TGC-3'; 732R: 5-TTG CCG ATT CTA CCC CTA C-3') and primers ndhDR and 672F (ndhDR: 5'- GAA CTC CTT CTA ACG ACT TAT GC-3'; 672F: 5'- CGG CTA GAA GCA TAC AAG-3'). Primers 732R and 672F were used as sequencing primers. The ndhl gene was amplified using primers ndhGF (5'- CCG ACC CTA GAA AGA CTA AAA G-3') and primer ndhAexon2R (5-CGT CCC AAC TTC TTT CAC TG-3'). Primer ndhAexon2R was used as the sequencing primer. The trriL intron and trnL-trnF spacer were amplified using primers C-Aster (5-CGA AAT TGG TAG ACG CTA CG-3') and primer F (5'-ATT TGA ACT GGT GAC ACG AG-3') (Taberlet et al., 1991). Primer C-Aster was used as the sequencing primer. The rpoB gene was amplified in three sections using primers rpoCF and rpoB1394R (rpoCF: 5-GAA ACT GAT CCA ATT CGG AG-3'; rpoB1394R: 5-TGG GGA TAC TCT AAG GAT TCC-3'), rpoB1270F and rpob2503R (rpoB1270F: 5-TTC GCC ACC AAC TGT AGC AG-3'; rpoB2503R: 5-TTG TGT AGA GGG AGA TCC G-3') and rpoB2426and either rpoBRl or rpoBR2 (rpoB2426: 5-AAT TGG GAG GGA TTG GTC G-3'; rpoBRl, 5-CAA GGT TTG ACG GAA GAA C-3'; rpoBR2: 5'-GAT CAA GGT TTG ACG GAA G-3'). Exon 1 of the rpoCl gene was amplified and sequenced for this study. Primers rpoC952F and either rpoBSRl or rpoBSR2 were used as PCR and sequencing primers (rpoC952F: 5-CCC TCT TTG CCT TCA ATT AC-3'; rpoBSRl: 5-CGG TTG TTC GTT CGA GAA C-3'; rpoBSR2: 5-CGA TCT TTA GCT CTG GAA CTG-3). The 23S-trnA spacer and the trnA intron were sequenced using primers 23SF (5-ATC CAC CGT AAG CCT TTC-3') and trnIR (5-ATT GGT TGG ACC GTA GGT GC-3'). Primer 23SF was used to sequence. The PCR products were cleaned with QIAquick PCR Purification Kit (Qiagen, Inc.) according to manufacturer's protocols. Sequencing was performed at the DNA sequenc- ing facility of the University of Texas using Big Dye termi- nator chemistry (version 3.0, Applied Biosystems, Foster city, California, USA). Sequence chromatograms were proofread and nucleotides aligned visually using Sequen- cher (version 4.5, Gene Codes Corp., Ann Arbor, Michi- gan, USA). Trimmed fragment matrices (contigs) were concatenated in NEXUS format for each data partition. Coding sequence partitions were translated and codon alignment checked, especially for uniformity and homology of stop codons using MacClade 4.05 (Maddison and Maddison, 2002). All character partitions are expected to have the same underlying history because chloroplast genes are linked and non-recombining. 2.3. Phylogenetic analyses Maximum parsimony analyses of the combined data were performed using PAUP* 4.0M0 (Swofford, 2003). The most parsimonious trees were found by a heuristic search using tree bisection reconnection (TBR), MUL- PARS, and simple taxon addition. The same analysis was performed with 100 random additions replicates of the data also. All characters in the data matrix were unordered and equally weighted and gaps were treated as missing data. Support for monophyletic groups was assessed using 1000 non-parametric bootstrap replicates (Felsenstein, 1985) using the same settings as the parsimony analysis. Bootstrap proportions >70% are considered well supported (Hillis and Bull, 1993). Tree statistics including consistency index and the retention index were calculated using PAUP*. Bayesian analysis of the combined data set was accom- plished using MrBayes 3.0b4 (Huelsenbeck and Ronquist, 2001). A best model of nucleotide evolution was chosen from the 24 models implemented in MrModeltest 2.1 (Nylander, 2004) using Akaike information criterion (Akaike, 1974). Four replicate Metropolis-coupled Markov chain approximations were run for 10 million generations, each starting from random trees. For each replicate run four Markov chains were run, one cold and three heated (temp ? 0.5) to facilitate mixing, sampling likelihood val- ues and trees from the cold chain every 100 generations for a total of 100,000 trees. Stationarity of the Markov chain was ascertained by plotting likelihood values against number of generations and the four plots compared for apparent stationarity. Trees sampled before stationarity were discarded as burnin. Following Alfaro et al. (2003) we consider posterior probabilities >0.95 as significant probability for a clade. 2.4. Bayesian hypothesis tests We used our data to test two a priori hypotheses of Mutisieae. To test the monophyly of the Guayana High- lands genera, including Stifftia and Wunderlichia as pro- posed by Jimenez Rodriguez et al. (2004), we first constructed a tree consistent with that hypothesis contain- ing a single resolved node shared by all the Guayana High- lands genera of classical Mutisieae that we sampled including Stenopadus, Stomatochaeta, Chimantaea, Dui- daea, Gongylolepis, Wunderlichia and Stifftia. We then used this constraint tree to filter all post-burnin trees sampled in one of our Bayesian analyses. The number of trees found to J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 761 be consistent with the constraint divided by the total num- ber of trees, yielded a proportion that was converted to a percentage and taken as the probability of monophyly of such a clade given the data and the GTR +1+ T model used. A similar procedure was used to test the monophyly of Mutisieae including Stifftia as found by Kim and Jansen (1995) using a tree constraining the nine genera sampled by Kim and Jansen (Stifftia, Onoseris, Trixis, Acourtia, Pere- zia, Nassauvia, Mutisia, Adenocaulon, and Gerbera), along with 17 genera not sampled by Kim and Jansen but shown here to be included in our Mutisia and Stifftia clades namely Lycoseris, Plazia, Aphyllocladus, Lophopappus, Proustia, Leucheria, Jungia, Dolichlasium, Pachylaena, Trichocline, Brachyclados, Chaetanthera, and Chaptalia (Mutisia clade) and Gongylolepis, Duidaea, Hyaloseris and Dinoseris (Stifftia clade). 3. Results 3.1. Phylogenetic analyses Concatenation of all 10 markers resulted in a data matrix containing 13,299 nucleotides for 108 taxa. The data matrix contained 1080 sequences of which more than 1060 were new and contributed to the GenBank database (Appendix A). Nine sequences included in the analyses were obtained from Genbank and correspond to the gene ndhF for the genera Athroisma, Barnadesia, Blepharisper- mum, Helianthus and Tagetes and the gene rbcL for the genera Barnadesia, Dasyphyllum, Scaevola, and Stokesia (Appendix A). Missing data amounted to approximately 3.9% of the total. A summary of the Maximum Parsimony statistics for each data partition is presented in Table. 1. Approximately 20% of the 13,299 characters in the data matrix for our taxon sampling were parsimony informa- tive. Among the 10 data partitions, matK had the highest percentage of informative characters with 28% whereas the 23S-trnl region had the lowest with only 4% (Table 1). Maximum parsimony analysis of the concatenated matrix by simple taxon addition produced 72 most parsimonious trees each 11,043 steps in length. The strict Table 1 Properties of istics resultin; data partitions used in this study and I from MP analyses statistical character- Partition Aligned Informative characters Tree CI RI length (percentage of total) length matTL 1610 446 (28%) 1848 0.47 0.62 ndKD 1408 264(19%) 1070 0.45 0.62 ndh? 2328 572(25%) 2586 0.42 0.63 ndh\ 501 75(15%) 270 0.49 0.73 rbcL 1412 251(18%) 1217 033 0.60 rpoB 3151 573(18%) 2021 0.52 0.53 rpoClexonl 518 71 (14%) 222 0.64 086 23S-95%) posterior probabilities (PP) in the Bayesian analyses and moderate to high (>70%) bootstrap proportions (BS) in parsimony analysis with the exception of the Wunderlichioideae clade (52% BS, 91% PP). Relationships among ten of the 12 lineages were also congruent between methods, and supported by signif- icant posterior probabilities and strong bootstrap propor- tions except for Gochnatioideae (65% BS). However, placement of the Wunderlichioideae and the Stifftioideae lineages was equivocal in Bayesian and parsimony 762 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 9 bp deletion ndhF (Kim and Jansen, 1995) Vernoniod Group (Bremer, 1 9 bp deletion rpoB 18 bp insertion rpoB 15 bp deletion rpoB Achillea Oncosiphon Ursinia Baccharis Erigeron Gamochaeta Syncarpha Relhania C h rysa nthem o i des Osteospermurn Dimorphateca Phaneroglossa Senecio Psacalium Heliarthus Oyedaea Perityle Helenium Athroisma Blepharispermum Pluchea Corymbium Gorteria Hoplophyllurn Berkheya Arctotis Hete role pis Centratherum Eremanthus Stokes ta Hesperomannia Sinclairia Sonchus Young ia Scolymus Gundelia Warronia Gymnarrhena Ainsliaea apiculata Ainsliaea macrocephala Pertya Carthamus Centaurea Atractylis Echinops Dicoma sp. Dicoma capensis Pasaccardoa Macledium Brachylaena Tarchonantrius Oldenburgia Hecastocleis Richterago angustifolia Richterago amplexifolia Cnicathamnjs Gochnatia hypoleuca Gochnatia hjriartjana Cyclolepis Stenopadus Stomatochaeta Chimantaea Wunderlichia lanthopappus Leucomeris Nouelia Dinoseris Hyaloseris Duidaea Gongyldepis Stifflia Gerbera piloselloides Gerbera serrata Chaptalia Brachyclados Trichocline Adenocaulon chilense Ada n oca u I on bicolor Mutisia Pachylaena Chaotanthera Nassauvia Acourtia Dolichlassurn Jungia Leucheria Lophopappus Proustia Aphyllocladus Plazia Lycos eris Onosens Barnadesia Dasyphyllum Chuquiraga Doniophyton Acicarpha Scaevnla Asteroideae Anthemideae I Astereae Gnaphalieae Calenduleae Senecioneae Heiiantheae alliance Athroismeae Inuleae Corymbieae i Corymbioideae Arctotideae Vernonieae Liabeae Cichorieae Gundelieae Gymnarrheneae Pertyeae Cynareae Dicomeae Cichorioideae Gymnarrhenoideae Pertyoideae Carduoideae Tarchonantheae Hecastocleideae I Hecastocieidoideae Gochnatieae Wunderlichieae Hyalideae Stiffiieae Mutisieae Nassauvieae Gochnatioideae Wunderlichioideae Stifftioideae Mutisioideae Onoserideae Barnadesieae Fig. 1. Strict consensus of 72 most parsimonious trees resulting from Maximum Parsimony analysis of combined data (10 chloroplast loci). Bootstrap proportions shown above the branches. Black bars map major indels discussed in Section 3.2. J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757?782 763 Achillea Oncosiphan Baccharis Erigeron Felicia ? Gamochaeta J ' Syncarpha I Relh 'uui Chrysantr J I Osteospei ' Dimorpho' l ania hemoides rmum teca Phanerogtossa Senecio Psacalium Helianthus Stevia Perityle Tagetes Helenium Athroisma Blepharispermum Pluchea Corymb ium Gorier j a Hoplophyllum Berkheya Arctotis Heterolepts GentratherLim Eremanthus Stokes ia Hesperomannia EJnclaJna Sonchus Youngia Scolymus Gund elia Warionia Gymnarrhena AJnsliaea apiculata Ainsliaea macrocephala Dicoma sp. Dicoma ca pens is Pasaccardoa Macledknn Brachylaena Tarchonarithus Oldenburgia Carthamus Centaurea Echinops Atractylis Hecastocleis Richterago angustifolia Richterago amplexifolia Cnicothamnua Gochnatia hypoleuca Gochnatia hiriartiana Cyclolepis Stenopadus Stomatochaeta Chimantaea Wunderlichia Hyalis lanthopappus Leucomeris Nouelia Dinoseris Hyaloseris Duidaea Gongylolepis Stifftia Gerbera piloselloides Gerbera serrata Chaptalia Brachyclados Trichocline Adenocaulan chilense Adenocaulan bicotor Mutisia Pachylaeria Chaetanthera Nassauvia Perezia Acourtia Dolichlasium Jungia Leucheria Lophopappus Proustia Aphyllocladus Plazia Lycoseris Onoseris Barnadesia Dasyphyllum Ghuquiraga Doniophyton Acicarpha Scaevola Anthemideae Astereae Gnaphalieae Calenduleae Senecioneae Heliantheae alliance Athroismeae Inuleae Corymbieae Arctotideae Vernonieae Liabeae Cichorieae Gundelieae Gymnarrheneae ' Pertyeae Dicomeae Tarchonantheae Cynareae Hecastocleideae l Gochnatieae Wunderlichieae Hyalideae Stiffiieae Mutisieae Nassauvieae Onoserideae Barnadesieae Asteroideae i Corymbioideae Cichorioideae Gymnarrhenoideae Pertyoideae Carduoideae Hecastocleidoideae Gochnatioideae Wunderlichioideae Stifftioideae Mutisioideae Barnadesioideae Fig. 2. Majority Rule consensus of 98,000 post-burnin trees obtained from Bayesian analyses of the combined data (10 chloroplast loci) using GTR + 74- r. Numbers above branches represent posterior probabilities. 764 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 methods. In parsimony analysis these two clades were unresolved together with Mutisioideae, whereas in all Bayesian analysis the Wunderlichioideae was placed sister to the Gochnatioideae-Asteroideae clade. The most parsi- monious trees placed Wunderlichioideae variously, either as sister to the Gochnatioideae-Asteroideae clade as in Bayesian analyses, or sister to the Mutisioideae, or sister to Stifftioideae and collectively sister to the Mutisioi- deae. The Stifftioideae were either placed as the next line- age to split after Barnadesioideae or sister to the Mutisioideae. Not only are the phylogenetic positions of Stifftioideae and Wunderlichioideae novel but their compositions are as well. The taxa endemic to the Guayana Highlands were found not to be monophyletic contrary to previous studies (Maguire, 1956; Pruski 1991; Bremer, 1994). Both the Stiff- tioideae and Wunderlichioideae contain tropical Brazilian genera sister to clades of Andean, eastern temperate South America and/or Guayana Highlands genera. The Stifftioi- deae was found to comprise the mostly Brazilian genus Stifftia sister to two clades namely, Hyaloseris and Dinose- ris of the central Andes sister to Gongylolepis and Duidaea of the Guayana Highlands. The Wunderlichioideae was found to contain two clades: Wunderlichieae and Hyali- deae both of whom share a 6 bp deletion in rpoB unique among Asteraceae. The former holds the Brazilian Planalto endemic genus Wunderlichia sister to Guayana Highlands genera Chimantaea, Stenopadus, and Stomatochaeta. The latter clade holds the southern South American genera Hyalis and Ianthopappus, and the Asian genera Nouelia and Leucomeris. The Mutisioideae was found to contain three lineages with Onoserideae sister to the Mutisieae-Nassauvieae clade. Nassauvieae and Mutisieae were found to be mono- phyletic and Adenocaulon was strongly supported as a member of the Mutisieae. Phylogenetic relationships among members of the three tribes were fully resolved except for those within Mutisieae. The scapose inflores- cence group of Mutisieae including Brachyclados, Tricho- cline, Chaptalia, and Gerbera was strongly supported as a monophyletic group (100% BS, 100% PP). Mutisia is sister to Pachylaena. Lophopappus and Proustia are genera with species having actinomorphic and bilabiate corollas sister to all other genera of Nassauvieae that are characterized by bilabiate corollas. The yellow-corolla Nassauvieae exemplified by Trixis were found to be sister to Nassauvia, Perezia, and Acourtia. The Onoserideae comprised two main clades with Plazia and Aphyllocladus sister to Onose- ris and Lycoseris (Figs. 1 and 2). The phylogenetic position of the Gochnatioideae was only weakly supported in parsimony analyses (65% BS), but strongly supported in Bayesian analyses (100% PP). Cnicothamnus and Richterago, the only genera of the sub- family with marginal corollas containing a limb, formed a clade. There is support for the recognition of Richterago as distinct from Gochnatia although no South American representative of Gochnatia was sampled. Cyclolepis was found to be the sister taxon to all other genera of the sub- family sampled. The Carduoideae branch was strongly supported in both analyses (100% BS, 100% PP). The subfamily Carduoideae was found to include not only members of tribe Cynareae but also Tarchonanthus, Brachylaena, Dicoma, Macledium, Pasaccardoa, and Oldenburgia, genera traditionally placed in tribe Mutisieae. The Carduoideae contained three major lineages: tribes Cynareae, Tarchonantheae, and Dicomeae. Resolution among these three tribes was equivocal. The strict consensus of most parsimonious trees collapses these branches to a trichotomy. Bayesian analysis placed Dico- meae as sister to Tarchonantheae with only 74% posterior probability. Oldenburgia is supported as a member of Tarchonantheae (70% BS, 100% PP). The monophyly of the Cichorioideae was strongly sup- ported in both Maximum Parsimony and Bayesian analy- ses (96% BS, 100% PP). Our studies showed there are three main lineages in the subfamily, Arctotideae, Vern- onieae plus Liabeae, and Cichorieae plus Gundelieae. Par- simony analysis failed to resolve the relationships among these three lineages, whereas Bayesian analyses placed the Vernonieae-Liabeae clade as sister to Arctotideae without strong support (93% PP) and collectively sister to the Gun- delieae-Cichorieae clade with strong support (100% PP). Warionia was strongly placed (86% BS, 100% PP) as sister to Gundelia of the Gundelieae (Figs. 1 and 2). Parsimony and Bayesian analyses strongly support sub- family Asteroideae (80% BS, 100% PP). Both methods also strongly supported a clade consisting of tribe Calenduleae as sister to tribes Gnaphalieae, Anthemideae, and Astereae (95% BS, 100% PP) and another containing tribes Inuleae, Athroismeae, and Heliantheae alliance (100% BS, 100% PP). Parsimony analysis placed Senecioneae sister to the Calenduleae clade but without support (Fig. 1), whereas Bayesian analysis failed to resolve the relationship of the Senecioneae to other tribes of Asteroideae (Fig. 1). The relationships among tribes of the Heliantheae alliance were found to be in agreement with those reported in Panero (2007b). 3.4. Bayesian tests of monophyly The hypothesis that classical Mutisieae genera of the Guayana Highlands are monophyletic has a 3% posterior probability and is rejected. The monophyly of classical Mutisieae including Stifftia (here Mutisioideae including Stifftioideae) has a 26.6% posterior probability and there- fore could not be rejected at a 5% significance level. Results from these Bayesian analyses are conditional probabilities, conditioned on the data, the GTR+I+F model of nucleo- tide evolution, and the prior probabilities used. 4. Discussion Our phylogenetic analyses identified 12 major clades in Asteraceae. Eleven of these lineages are statistically sup- J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 765 ported; the exception being Wunderlichioideae with only 91% posterior probability. Only Barnadesioideae, Cicho- rioideae, and Asteroideae had been identified with strong bootstrap support in previous studies. We recognize all 12 lineages at the subfamily level. Phylogenetic relation- ships among 10 of the 12 subfamilies are strongly sup- ported by both bootstrap proportions and posterior probabilities and this represents a significant advancement in our understanding of the evolutionary history of Aster- aceae. Discussion of the systematic and taxonomic implica- tions of each major clade follows the branching order represented in Figs. 1 and 2. Members of Barnadesioideae are strongly supported as sister to a clade encompassing the 11 other major lineages of the family. The classical circumscription of tribe Mutisieae (Cabre- ra, 1977) is herein amended. The lineage containing the type genus Mutisia is here recognized as Mutisioideae. The subfamily is much reduced from its classical circum- scription with several of its members referred herein to the Stifftioideae, Wunderlichioideae, Gochnatioideae, Hec- astocleidoideae, Carduoideae-Dicomeae, Carduoideae- Tarchonantheae, and Pertyoideae. 4.1. Mutisioideae Mutisioideae was found to contain three main branches: Mutisieae, Nassauvieae, and Onoserideae. Mutisioideae as recognized here contains approximately 44 genera and 630 species. The subfamily is primarily South American with the exception of three derived genera [Chaptalia, Gerbera, Trichocline) and Adenocaulon that have attained cosmopol- itan distributions except Europe. The subfamily can be characterized as having disc corollas with deeply dissected lobes, some of its members having bilabiate corollas, capit- ula with imbricate phyllaries, anthers calcarate and caudate with strongly sclerified anther appendages, and styles usu- ally well-exserted from the floret and essentially glabrous. Most species are annual or perennial herbs, although shrubs, small trees and vines are also present. Hybrids of species of the genus Gerbera are widely cultivated for their large capitula and vibrant colors. Many studies have been aimed at clarifying the relation- ships of classical Mutisieae using detailed morphological features ranging from ligule micromorphology to pollen ultrastructure (Hansen, 1991 and references therein; Zhao et al., 2006). Revisionary or cladistic studies of the group have invariably concluded that Mutisieae are paraphyletic and require dismemberment (Cabrera, 1977; Hansen, 1991; Karis et al., 1992; Bremer, 1994, 1996). However, among these studies there has been no consensus for con- struction of a stable classification for the group and conclu- sions about clade circumscription and/or monophyly have often been contradictory. 4.1.1. Mutisieae Our studies support a reduced tribe Mutisieae to include approximately 14 genera and 200 species with the great majority endemic to South America. The composition of the tribe as supported by our molecular studies roughly corresponds to the circumscription of subtribes Gerberinae and Mutisiinae of Hind (2007) excluding the genera of the Onoserideae (see below). The genera sampled include: Adenocaulon, Brachyclados, Chaetanthera, Chaptalia, Ger- bera, Mutisia, Pachylaena, and Trichocline. The southern Andean, mostly herbaceous genus Chae- tanthera has been traditionally allied to Brachyclados, Pachylaena and Trichocline (Cabrera, 1937). Palynological studies by Tellerfa and Katinas (2004) support this view. Most species of the genus share a translucent wing on the phyllaries and trichomes on the anthers (Hansen, 1991). Chloroplast ndhF studies by Kim et al. (2002) revealed Chaetanthera as sister to Duidaea, a genus endemic to the Guayana Highlands. In contrast, our results show Chae- tanthera is not closely related to Duidaea of the Stifftioideae but firmly nested within Mutisieae, although its relation- ships within the tribe are equivocal. Our study shows that Gerbera is closely related to Chaptalia and collectively sister to Brachyclados and Trichocline (^Gerberinae). This result is consistent with morphological studies (Hind, 2007) and the molecular study of Kim et al. (2002). These four genera for the most part can be characterized by their acaulescent habit and solitary monocephalous inflorescences on long scapes. Their primary distribution is in South America with the genus Chaptalia and segregates also present in North America and Asia. Kim et al. (2002) showed Gerbera derived from Chaptalia making the latter not monophy- letic. Gerbera is considered to be an Old World endemic (Hansen, 1991). However, other authors believe Gerbera to have one or two species in America (Gerbera hieracio- ides (Kunth) Zardini, Zardini, 1974; Gerbera hintonii (Bullock) Katinas, Katinas, 1998). Speculation about the affinities of Adenocaulon has been extensive, and historic taxonomic views are summa- rized in Katinas (2000). Adenocaulon contains five species with a disjunct distribution in North America, Mesoamer- ica, South America, and the Himalayas. It is perhaps the most widespread genus of Mutisioideae. Adenocaulon is an anomalous genus that has lost several of the obvious morphological synapomorphies of Mutisieae and pos- sesses characters that point to a relationship to subfamily Asteroideae. These include basally constricted anther appendages, small anthers, and disciform capitula (Kati- nas, 2000). Katinas concluded that Adenocaulon along with Eriachaenium should be placed in their own tribe within Cichorioideae. Previous molecular studies have placed the genus in tribe Nassauvieae (Kim et al., 2002), although this position was not supported (24% BS), or in the Mutisieae (Jansen and Kim, 1996). We sampled two American species, Adenocaulon bicolor and Adenoca- ulon chilense, and found them to be strongly supported as a lineage within tribe Mutisieae. However, the relation- ship of Adenocaulon to other Mutisieae genera was not resolved (Figs. 1 and 2). 766 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757?782 4.1.2. Nassauvieae Our studies show Nassauvieae to be sister to Mutisieae, with the genera Lophopappus and Proustia as sister to the other genera of the tribe. Nassauvieae contain 24 genera and approximately 370 species (Panero, 2007c) distributed mainly in South America with the genera Acourtia and Berylsimpsonia, and some species of Trixis endemic to North America. The tribe has been traditionally viewed as a natural group because most of its genera have capitula with only bilabiate corollas and truncate (rarely round) style branch apices, a combination of characteristics dis- tinctive in the family. Crisci (1974) modified the traditional concept of the tribe by adding Lophopappus and Proustia, which he considered sister genera. These two genera have species with actinomorphic corollas, and rounded styles but share similar pollen morphology with Nassauvieae. Our results are consistent with his circumscription of the tribe based on pollen and floral morphology. Ours is the first molecular study to provide a robust hypothesis of generic relationships in Nassauvieae. We sampled nine of the 24 genera of the tribe along with Aden- ocaulon. Kim et al. (2002) included seven of the same gen- era we sampled but did not include Lophopappus or Dolichasium. They hypothesized relationships based on one of 5056 most parsimonious trees found. In their strict consensus tree (bootstrap values not provided) major clades of Nassauvia collapse to a polytomy, though the relationships of Proustia with Trixis, Perezia with Nassau- via, and Triptilion embedded within Nassauvia were main- tained, as was the association of Jungia with Leucheria (shown with 34% bootstrap support on one of the most parsimonious trees). Although we also found Perezia and Nassauvia to form a clade, our results differ significantly from those reported by Kim et al. (2002). Based on cladistic analyses of morphological characters, Crisci (1980) posited three scenarios for relationships within Nassauvieae depending on whether Dolichlasium, Trixis, or an hypo- thetical ancestor was used to polarize characters. Our results do not agree with any of these three scenarios, although we did find Lophopappus and Proustia to be sister, and these to be sister to the other genera of the tribe. These two genera have floral characteristics seen in the sister tribe Mutisieae, but rare in Nassauvieae. Leucheria was found to be sister to a clade containing the two largest genera of the tribe, Acourtia and Trixis. The Acourtia clade also contains Nassauvia, and Perezia. Acourtia was historically included in Perezia but later recognized to be distinctive and shown to share features with Lophopappus and Proustia (Crisci, 1980). The close relationship between Perezia and Nassau- via found here corroborates that of Kim et al. (2002). The Trixis clade also contains Dolichlasium and Jungia. These three genera have for the most part yellow corollas and tapered cypselae (Bremer, 1994). Since the basal lineages of Nassauvieae, Lophopappus, Proustia, and Leucheria, and most members of the sister tribe Mutisieae, are ende- mic to the south central Andes, it is reasonable to assume that Nassauvieae originated in dry areas of this region and subsequently expanded to the southern Andes, North America, and Brazil. 4.1.3. Onoserideae Our study identified a novel clade containing Aphyllo- cladus, Lycoseris, Onoseris, and Plazia. Recent morpholog- ical studies of selected members of classical Mutisieae (Telleria and Katinas, 2004) have identified a combination of morphological characteristics that are shared by these four genera and Gypothamnium and Urmenetea, namely similar tubular corollas and dimorphic pappi of narrowly paleaceous bristles. Several taxonomic and morphological studies of classical Mutisieae are in agreement that Aphyllocladus, Gypothamnium, and Plazia form a natural group based on the shared character of red anther append- ages that are distinctive in Mutisioideae (Hansen, 1991; Bremer, 1994; Telleria and Katinas, 2004). We sampled four of the six genera of Onoserideae (Panero and Funk, 2007) and our results show the tribe is biphyletic, with Onoseris sister to Lycoseris and Plazia sister to Aphyllocla- dus. Morphological studies have not allied Lycoseris to Onoseris but Hansen (1991) noted a similarity (potentially a synapomorphy) between Lycoseris and Onoseris in their short corolla lobes as compared to other members of Muti- sieae. There is general agreement among most workers that Urmenetea is closely related to Onoseris (Sancho, 2004). Onoserideae as circumscribed here contains the genera Aphyllocladus, Gypothamnium, Lycoseris, Onoseris, Plazia and Urmenetea, comprising 53 species distributed mostly in the dry Andes of northern Chile, Argentina, Bolivia, and southern Peru, with Onoseris and Lycoseris having a sizeable number of species in mesic or seasonally dry low- land forests of Mesoamerica and South America. 4.2. Stifftioideae Stifftia, Gongylolepis, Duidaea, Hyaloseris and Dinoseris form a clade recognized as subfamily Stifftioideae. This clade is strongly supported in both Parsimony and Bayes- ian analyses (99% BS, 100 PP) though the relationship of this clade to other subfamilies is equivocal (Figs. 1 and 2). We were unable to statistically reject the hypothesis of Stifftia and relatives within Mutisioideae at the 5% level of significance, although the conditional probability that this hypothesis is correct is low (26.6% PP). The common denominator in the taxonomic history of Stifftia has been its inclusion, by virtue of its morphology, as a member of the early radiation of the Asteraceae. Stifftia has been con- sidered a "primitive genus" in the family because several of its species have large imbricate involucres, long actinomor- phic corollas with strongly coiled lobes, and an arborescent habit (Maguire, 1956). Most of its eight species are found in the tropical forest of eastern Brazil with two additional species found in the rainforests of northern Brazil and French Guyana (Hind, 1996; Robinson, 1991). Nonethe- less, Maguire (1956) considered Stifftia one of the Guayana Highlands genera placed in classical Mutisieae and an early J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 767 offshoot of the progenitor of those taxa. He allied Stifftia to Stenopadus, another genus that figures prominently in dis- cussions about the origins and characteristics of the early sunflowers (Maguire, 1956; Pruski, 1991; Karis et al. 1992; Bremer 1994). The four genera here found to be strongly supported as sister to Stifftia have bilabiate/ligulate corollas. Our results show that Stifftia is sister to two clades, one containing Gongylolepis and Duidaea of the Guayana Highlands and the other Hyaloseris and Dinoseris of Bolivia and Argen- tina. Because Gongylolepis and Duidaea share a similar cor- olla morphology and other floral characteristics with the Guayana Highlands genera Achnopogon, Glossarion, Eurydochus, Neblinaea, and Quelchia (Jimenez Rodriguez et al., 2004 and references therein) we consider these genera to belong in subfamily Stifftioideae as well. A potential syn- apomorphy for the Guayana Highlands genera of Stifftioi- deae is the presence of laticifers observed by Carlquist (1958) in the genera Gongylolepis, Duidaea, Neblinaea, and Quelchia. The inclusion of Hyaloseris and Dinoseris in the Stifftioideae is strongly supported in our analyses. Cladistic analyses of morphological data by Karis et al. (1992) placed Hyaloseris as sister to Stifftia and Gongylol- epis. Hind (2007) included Hyaloseris (with Dinoseris in synonymy) in his Stifftia group along with Stifftia and Wunderlichia. The seven species of Hyaloseris are endemic to the mountains of northern Argentina and southern Boli- via. The ligulate corollas of Hyaloseris, Dinoseris, and Glossarion (Hansen, 1991) are of interest as this unusual corolla morphology may represent a synapomorphy for the group, subsequently lost in the other genera of the Guayana Highlands, or acquired in parallel. 4.3. Wunderlichioideae Results from Bayesian analyses provide statistical sup- port for the placement of the lineage containing Wunderli- chia as the next branch to split from Mutisioideae- Stifftioideae, this sister to the rest of the Asteraceae. The subfamily contains two main assemblages whose composi- tions are novel. One clade contains Wunderlichia sister to the Guayana Highlands genera Chimantaea, Stenopadus, and Stomatochaeta hereafter referred to as tribe Wunderli- chieae. The sister lineage contains the genera Hyalis and Ianthopappus sister to Nouelia and Leucomeris hereafter as tribe Hyalideae. The geographic distribution of members of subfamily Wunderlichioideae parallels that of Stifftioideae with three allopatric areas of endemism across South America: one in the Andes or temperate eastern South America, one in cen- tral Brazil, and one in the Guayana Highlands. Wunderli- chia is a genus of five or six species of the Planalto of central Brazil having large, homogamous capitula with actinomorphic corollas produced at the end of the dry sea- son on leafless stems. The gross morphology of Wunderli- chia (tree-like habit, coriaceous, caducous, densely pubescent leaves) is distinctive among genera included in classical Mutisieae, and it appears to be the result of adap- tation to the seasonally dry conditions of the Campo Rupestre of central Brazil. In spite of different life form, comparative studies of floral features have suggested a close relationship of Wunderlichia to members of the Gua- yana Highlands genera with actinomorphic corollas repre- sented by Stenopadus (Barroso and Maguire, 1973; Carlquist, 1957). However, the close relationship of this group to members of Hyalideae has never been hypothesized. Relationships within tribe Hyalideae are interesting because of the wide geographical separation of sister clades. Hyalis and Ianthopappus are genera endemic to sub- tropical, eastern South America, whereas Nouelia and Leu- comeris are endemic to the mountainous regions of southeast Asia and the foothills of the Himalayas. It is remarkable that sister clades are so distant in their geo- graphic distribution. This amphi-pacific pattern has never been reported for any South American sunflower except for the ChaptalialLeibnitzia and the Adenocaulon lineages but these also have species present in North America. Nou- elia and Leucomeris can be added to the list of Asteraceae genera with astonishing sister-taxon disjunctions ranging thousands of kilometers that includes Abrotanella (Chile, New Zealand-Tasmania, Wagstaff et al., 2006) and Hespe- romannia (Africa-Hawaii, Kim et al., 1998) among others. The presence of these genera in Asia either results from vacariance and extinction in North America or very long distance dispersal. Hyalis, Ianthopappus, and Leucomeris share with some Gochnatioideae rather short imbricate involucres with corollas and pappi conspicuously exserted beyond the involucre. This characteristic is not present in Nouelia, the latter having strongly imbricate involucres, but Nouelia shares with Ianthopappus and Hyalis radiate capitula with bilabiate (3 + 2) corollas. All four genera have leaves with white, pubescent abaxial surfaces. Our studies clearly show that Leucomeris does not belong to Gochnatia sensu Freire et al. (2002). Ianthopappus was recently recognized by Roque and Hind (2001) as a distinctive genus and not clo- sely related to Richterago (Gochnatioideae), its original placement. Among classical Mutisieae, only Ianthopappus and Lulia are restricted to semi-aquatic habitats. Characters that have been traditionally used to recog- nize natural groups or maintain certain groups within clas- sical Mutisieae appear labile. Most authors are in agreement that actinomorphic corollas are the ancestral condition in the family (Koch, 1930; Bremer 1994). This assertion is supported by our study and it can now be assured that bilabiate corollas are not ancestral as Jeffrey (1977) asserted, but rather are interpreted as having evolved in parallel in different groups including Mutisioi- deae, Stifftioideae, Wunderlichioideae, Gochnatioideae, Dicomeae, and Tarchonantheae. Equally important in the circumscription of groups in classical Mutisieae has been the use of anther appendage morphology. For example, the eight genera of the Wunderlichioideae share with some 768 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 Gochnatioideae, some Dicomeae, and most members of the Guayana Highlands genera of Stifftioideae apiculate anther appendages, a characteristic shared by these nested lineages and lost in a few species. Understanding how the Asteraceae first diversified beyond the Andes and Patagonia hinges on reconstructing the historical relationships among the lineages constituting classical Mutisieae. Although our study was unable to ascertain the precise relationship of Stifftioideae to the Mutisioideae, the statistically supported relationships of Stifftioideae and Wunderlichioideae enabled us to falsify an almost universally accepted historical hypothesis con- cerning the early diversification of sunflowers on the Gua- yana Highlands, and shed light on the evolution of one aspect of early floral evolution in Asteraceae. The considerable age of the Guayana Highlands together with the distinctive and apparently primitive mor- phology of genera traditionally assigned to Mutisieae that inhabit that area have been used to conclude that these genera are monophyletic and the probably most "ances- tral" lineage of the family (Maguire, 1956; Maguire and Wurdack, 1957; Huber, 2005). The 10 mutisioid genera of the Guayana Highlands are Achnopogon, Chimantaea, Duidaea, Eurydochus, Glossarion, Gongylolepis, Neblinaea, Quelchia, Stenopadus, and Stomatochaeta. After the discov- ery of Barnadesioideae as the lineage sister to the rest of the family, a monophyletic Guayana Highlands classical Muti- sieae (Pruski, 1991; Bremer, 1994) has remained the best hypothesis of the next lineage of the family to diverge (Pru- ski, 1991; Karis et al., 1992). Implicit in the writings of Maguire (1956) and Maguire and Wurdack (1957) is that they considered all the classical Mutisieae genera of the Guayana Highlands to be monophyletic, arising from a Guayanan progenitor. Following the taxonomy of the time they conveniently positioned genera with bilabiate and actinomorphic corollas into two taxonomic groups, namely Mutisiinae and Gochnatiinae respectively. Maguire's (1956) evolutionary scenario placed the largest mutisioid genera endemic to the Guayana Highlands, Gongylolepis and Stenopadus, as a bridge between his Mutisiinae and Gochnatiinae. He considered these two genera to be the primitive elements of the group and Stifftia and Moquinia (now all but the type species transferred to Gochnatia) to be extra-Guayana members early diversifying beyond the region. Hansen (1991) provided a potential synapomorphy for the Guayana Highlands Mutisioideae in the blackish color of the stems. He also commented that the black herb- age trichomes of these species, once believed to be a syna- pomorphy for the group, is the result of a fungal infection. Maguire (1956) considered the genus Stenopadus with its large capitula and conspicuously imbricate involucres to represent the ancestral lineage of Asteraceae. Karis et al. (1992) identified Stenopadus as the first lineage to split from Barnadesioideae, this based on cladistic analysis of mor- phological data of the Cichorioideae. Bremer (1994) included all Guayana Highlands Mutisioideae in his Stenopadus group. Cladistic analyses of the Guayana High- lands genera by Jimenez Rodriguez et al. (2004) support the traditional view that these taxa are monophyletic, but only with the inclusion of Stifftia and Wunderlichia. Our results contradict traditional views that the genera of the Guayana Highlands classical Mutisieae are mono- phyletic. Even though our sampling of the Guayana High- lands taxa was limited to one exemplar each of five of the 10 genera recognized to exist in this area (Hind, 2007), our analyses support two independent introductions into these mountains (Figs. 1 and 2). These two lineages correspond to the two corolla types observed in the group, bilabiate and actinomorphic. The bilabiate corolla genera Gongylol- epis and Duidaea are members of the Stifftioideae whereas the actinomorphic corolla genera Chimantaea, Stenopadus and Stomatochaeta are members of the Wunderlichioideae. Our results show that the progenitors of these taxa arrived in the Guayana Highlands region from the eastern Andes or central/northern Brazil early in the radiation of the fam- ily and have diversified there in isolation probably for mil- lions of years. Guayana Highland Asteraceae (Berry and Riina, 2005) belonging to other subfamilies are more recent arrivals. Except for Chaptalia, no genera of Mutisioideae occur in the Guayana Highlands area. 4.4. Gochnatioideae The well-supported clade comprising Cyclolepis sister to Gochnatia, Cnicothamnus, and Richterago recovered by our study does not correspond to any previous concept of a lineage containing Gochnatia. Cabrera (1977) recognized a subtribe Gochnatiinae with 36 genera distributed world- wide in his treatment of tribe Mutisieae. He believed this subtribe to be the most primitive group in tribe Mutisieae. He characterized most members of the subtribe by having actinomorphic corollas and never truly bilabiate corollas. Hansen (1991) narrowed the concept of the group by removing the African members of the tribe (Dicoma and relatives). In the same publication, Hansen considered monophyletic a subset of Gochnatiinae informally labeled as the Gochnatia group that also included the genera Acti- noseris, Cyclolepis, Gochnatia, Hyalis, Leucomeris, and Nouelia. A cladistic analysis of tribe Mutisieae by Karis et al. (1992) revealed Gochnatiinae as a paraphyletic assemblage. Bremer (1994) believed Gochnatiinae to be polyphyletic and placed all genera of the subtribe Goch- natiinae in subtribe Mutisiinae recognizing this and Nas- sauviinae as the two main lineages of Mutisioideae (classical Mutisieae). Molecular studies by Kim and Jansen (1995), and Kim et al. (2002) provide evidence for the poly- phyletic circumscription of the subtribe as conceived by Cabrera (1977). The Gochnatia complex of Freire et al. (2002) included Gochnatia and the genera Ianthopappus, Actinoseris, Cyclolepis, Cnicothamnus, Hyalis, and Nouelia. Their detailed morphological studies characterized this complex by the combination of smooth style branches and apiculate anther appendages and allowed them to exclude Chucoa, Pleiotaxis, and Wunderlichia from it. They J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 769 placed Leucomeris, Pentaphorus and Richterago within Gochnatia. Hind (2007) recognizes only the genera Gochna- tia, Richterago, and Pentaphorus in this lineage. Our studies support the recognition of the Brazilian endemic genus Richterago as circumscribed by Roque and Pirani (2001) including Actinoseris as distinctive from Gochnatia and sister to Cnicothamnus. Most species of Richterago form small rosettes with long, scapose inflores- cences. The genus is a distinctive element of seepages and edges of intermittent rivers primarily in the sandy moun- tains of the Serra do Cipo and Diamantina plateau in Min- as Gerais, central Brazil, whereas the two species of Cnicothamnus are treelets endemic to the moist forests of northern Argentina and southern Bolivia. The genera Hyalis, Ianthopappus, and Nouelia of the Gochnatia com- plex sensu Freire et al. (2002) are herein referred to the Wunderlichioideae. 4.5. Hecastocleidoideae Hecastocleis, a shrub endemic to the mountains that sur- round the Mojave Desert of California and Nevada, was found here to form a lineage distinct from Mutisioideae and sister to the Carduoideae-Asteroideae clade. The genus was named by Gray (1882), who believed its closest relative to be the Asian genus Ainsliaea. Gray commented on the distinctiveness of the genus in Mutisieae because of its spiny leaves, involucre of aciculate phyllaries, single- flowered capitula, and spiny inflorescence bracts. The corollas are white, and like those of Pertyoideae have five, long, broadly expanded, narrowly triangular lobes. Most corollas are actinomorphic but a few are zygomorphic with equivalent lobe lengths (Jose L. Panero, personal observa- tion). The inflorescence of Hecastocleis is distinctive in Asteraceae as its capitula are loosely surrounded by mostly five spiny bracts that enclose up to nine capitula that sit at the end of the shoot axis in an expanded, receptacle-like structure. Accessory flowering branches (paracladia) below these bracts expand after anthesis of the terminal capitula and have a similar arrangement to that of the terminal group. The single-flowered capitula with aciculate phyl- laries aggregated in globose secondary heads are reminis- cent of some Cynareae. Bremer (1994), echoing the views of Hansen (1991), considered the genus an isolated member of tribe Mutisieae. Hansen (1991) considered the style of Hecastocleis similar to that of Carlina, the latter a basal lineage of the Cynareae (Susanna et al., 2006). In addition, Hecastocleis has several other characteristics rare in Aster- aceae, including ring porous wood and imperforate trac- heids with prominent bordered piths probably the result of adaptation to seasonally dry habitats (Carlquist, 1957). A similar wood morphology is also present in Proustia, and to a lesser degree in Nouelia, Dasyphyllum (as Floto- via), and Trixis (Carlquist, 1957). According to Tellen'a and Katinas (2005; but see also Woodhouse, 1929), Hecas- tocleis along with selected species of Ainsliaea share tricol- pate pollen as yet unknown in other species of Asteraceae. Hecastocleis shares with Pertyoideae distinctive trichomes (Hansen, 1991). 4.6. Carduoideae With the inclusion of Dicomeae and Tarchonantheae along with the core thistles Cynareae we have expanded the Carduoideae (100% BS, 100% PP). This subfamily was recognized by Bremer (1996) to include only members of tribe Cynareae. Bayesian analyses (Fig. 2) placed Tarchonantheae and Dicomeae as sister and collectively sister to Cynareae but without significant support. Parsi- mony analysis (Fig. 1) was unable to resolve relationships among these three well-supported lineages. We cannot cite any morphological synapomorphy that defines Carduoi- deae but nearly all members have papillose styles with papillae mostly confined to a ring below the stigmatic branches, as do some members of the Arctotideae, or as a tuft of trichomes on the abaxial surface of the style branches. Oldenburgia and some Dicomeae have apiculate anther appendages similar to those of some Gochnatioi- deae and Wunderlichieae. The subfamily is dominated by tribe Cynareae which accounts for more than 90% of the species diversity of the group. Some members of Cynareae contain latex, a characteristic found almost exclusively in the Cichorioideae. The highest species and generic diversity of Cynareae is found in Europe and central Asia with very sparse representation in America and Australia. The distinctiveness of Dicomeae within classical Muti- sieae was pointed out by Jeffrey (1967) but it was Hansen (1991) who supported their removal from Mutisieae. The latter worker speculated that the inclusion of Dicomeae in Mutisieae by other workers was based upon the bilabiate corolla of some of its members and the style trichome dis- tribution dissimilar from Cynareae. Morphological studies of 'Mutisieae' endemic to eastern Africa convinced Hansen that the genera Dicoma, Erythrocephalum, Pasaccardoa, and Pleiotaxis are not closely related to Mutisieae but rather to Cynareae. Hansen (1991) believed the corolla epi- dermal trichomes, the conspicuous difference between the narrow tube and broad limb, and style branches with sub- apical trichomes are not present in any other Mutisieae and are similar to Cynareae. Bremer (1994) followed conclu- sions from cladistic studies by Karis et al. (1992) that Dico- ma, Erythrocephalum, and Pleiotaxis were members of Mutisieae, and he maintained the genera as such under the informal Dicoma group. Cladistic studies of the Dicoma group by Ortiz (2000) using Oldenburgia and Gochnatia as outgroups, showed Erythrocephalum and Pleiotaxis to be sister to a paraphyletic Dicoma with the species of Pasa- ccardoa derived from within it. Subsequent studies by Ortiz (2001, 2006) refined the concept of Dicoma and the Dicoma group to contain also Cloiselia, Dicoma, Erythrocephalum, Gladiopappus, Macledium, Pleiotaxis, and Pasaccardoa. Our studies support the conclusions of Hansen (1991) that the Dicoma group is a member of subfamily Carduoideae and that the genera we sampled, Dicoma, Macledium, and 770 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 Pasaccardoa form a monophyletic group (Figs. 1 and 2). The relationships of the Dicoma clade to other members of the Carduoideae are equivocal based on our data. Our inclusion of Tarchonantheae in Carduoideae has no historical precedent. The suprageneric taxonomic history of Tarchonanthus and Brachylaena has been one of contro- versy. The two genera were initially assigned to the Inuleae then transferred to Mutisieae after pollen studies by Leins (1971) and Skvarla et al. (1977) showed the genera to have anthemoid type pollen. Cabrera (1977) did not include these taxa in his treatment of the tribe, and Hansen (1991) considered the genera not to be Mutisieae. Restric- tion fragment analyses of the chloroplast DNA by Keeley and Jansen (1991) showed these constitute a clade and one of the early lineages of the family. Their discovery was used to recognize the lineage formally as tribe Tarcho- nantheae (a later homonym of Tarchonantheae Kostel.). Our studies recovered the Tarchonanthus-Brachylaena- Oldenburgia clade with significant statistical support by both phylogenetic methods (100% BS, 100% PP). Oldenbur- gia was revised by Bond (1987) to include four species of shrubs and trees of the Cape region of South Africa. The genus is characterized by its large, radiate capitula and the dense tomentum of their herbage. Bond (1987) believed the genus not to be closely related to the Dicomeae but rather to certain New World Mutisieae including Chiman- taea, Cnicothamnus, and Wunderlichia. She believed Cnico- thamnus in the Gochnatioideae to be the closest relative of Oldenburgia. Hansen (1991) maintained Oldenburgia in Mutisieae but did not provide any insights about its rela- tionships. Cladistic analysis of Cichorioideae by Karis et al. (1992) placed Oldenburgia in a relatively basal posi- tion within Mutisieae and Bremer (1994) used these results to conclude that the genus represented an isolated member of the tribe, a survivor from an early diversification of the family. Our results support the close relationship of Olden- burgia to the parapatric Tarchonanthus and Brachylaena. The latter two genera are different from Oldenburgia in gross morphology but share with it the arborescent habit and the dense herbage pubescence. Here we expand Tarchonantheae to include Oldenburgia and recognize the need for a comprehensive synthetic study of the tribe. 4.7. Pertyoideae The Pertyoideae represent the most highly nested lineage of classical Mutisieae. The subfamily is strongly supported as sister to Gymnarrhenoideae and the Cichorioideae- Asteroideae clade, and it contains the genera Ainsliaea, Macroclinidium, Myripnois, and Pertya (sensu Hind 2007). The only molecular study to include a majority of the species of the genus Ainsliaea showed that the mono- typic genus Diaspananthus, traditionally included in Ains- liaea (Ainsliaea uniflora Sch. Bip.), is the sister taxon to all the other species sampled (Mitsui et al., 2008). Cladistic analysis of morphological features support as well A. unifl- ora as sister to all other species of the genus (Freire, 2007). According to Jeffrey (2007) the monotypic Himalayan genus Catamixis is a member of this group. The Pertyoideae have approximately 70 species restricted to temperate eastern Asia and the Himalayas {Pertya group, Hind, 2007). Most species are hygrophytes that inhabit the understory of temperate forests. According to Hansen (1991) the Pertyoideae (Ainsliaea group) has one synapomorphy in the laterally arranged capitula. The capitula are discoid, cylindric or campanulate, and few- flowered. The corollas are distinctive as they are deeply lobed and the lobes in many species are perpendicular to the axis of the capitulum. Some corollas are zygomorphic with one deeper sinus resulting in a corolla that resembles a slightly concave hand (Pertya, Koyama, 1975). The genus Ainsliaea shares with Chaptalia and Leibnitzia the forma- tion of cleistogamous and chasmogamous capitula (Freire, 2007). The Pertyoideae share a few morphological charac- teristics with Hecastocleis (see above). 4.8. Gymnarrhenoideae Gymnarrhena is an excellent example of a "non-missing" link genus, linking Pertyoideae Carduoideae and the Ver- nonioid group (Cichorioideae, Corymbioideae and Aster- oideae, Fig. 1) of Bremer (1996). The genus had been placed in tribe Inuleae because of its habit and near the Cynareae because of its pollen morphology (Leins, 1973; Skvarla et al., 1977) and Bremer (1994) considered the genus as of uncertain position in Cichorioideae. Our anal- yses do not support its placement in any existing tribe or subfamily. Instead, Gymnarrhena was identified as an inde- pendent lineage with significant bootstrap support and pos- terior probability. Furthermore, Gymnarrhena lacks the 9- bp deletion in the ndhF gene identified by Kim and Jansen (1995) and subsequently used by Bremer (1996) as a molec- ular characteristic in support of the recognition of the Ver- nonioid group (Fig. 1). Gymnarrhena is an amphicarpic herb of the Mediterranean biome of North Africa and the Middle East. Research on the interesting floral dimor- phism associated with the reproductive biology of this unu- sual plant is summarized by Koller and Roth (1964). The subfamily contains a single genus and species, Gymnarrhe- na micrantha Desf. Our studies support the recognition of Gymnarrhena at the subfamily level as sister to the Cicho- rioideae-Corymbioideae-Asteroideae clade. 4.9. Cichorioideae Our results and taxonomic concept of Cichorioideae dif- fer slightly from those of Kim and Jansen (1995) and their taxonomic interpretation by Bremer (1996) because we included samples of Gundelia and Warionia, two previously unplaced or misplaced anomalous genera. Our analyses show that a monophyletic Cichorioideae must include these two genera along with tribes Arctotideae, Cichorieae, Lia- beae, and Vernonieae. Recent molecular studies of the sub- family by Karis et al. (2001) identified the genus Gundelia J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 771 as a member of tribe Cichorieae and placed the anomalous genera Eremothamnus and Hoplophyllum as members of Arctotideae. Here we confirm the placement of Hoplophyl- lum, and also place Heterolepis in Arctotideae. However, we find Gundelia together with Warionia to be a separate lineage sister to Cichorieae and morphologically different from Cichorieae by its actinomorphic corollas. Therefore, we accept Robinson's (1994) tribe Gundelieae but expand it to include Warionia, a large shrub endemic to North Africa. Our study places Gundelieae sister to Cichorieae with strong statistical support but otherwise found tribal relationships within Cichorioideae congruent with the ndhF gene phylogeny of Kim and Jansen (1995). Like Kim and Jansen we also do not have support for any relationship of the Cichorieae-Gundelieae clade with other tribes in the subfamily (<50% BS, 93% PP). Bremer (1994) provided a concise history of the classifi- cation and morphology of the group. The subfamily is characterized by the presence of latex as a plesiomorphic condition subsequently lost in most Vernonieae, Arctoti- deae, and some Liabeae. Members of Cichorieae, Vern- onieae, and some Liabeae have tangentially oriented style branches (Robinson, 1984) different from the radially ori- ented style branches present in most of the other lineages of the family. Cichorieae have ligulate corollas, whereas Arctotideae, Liabeae, and Vernonieae with few exceptions have discoid or radiate capitula. The subfamily contains approximately 2900 species of cosmopolitan distribution (Jeffrey, 2007). 4.10. Corymbioideae The anomalous genus Corymbium was found to be a dis- tinct lineage linking Cichorioideae to the tribes of Asteroi- deae. A small genus of only nine species of perennial herbs from the Cape region of South Africa (Weitz, 1989; Nor- denstam, 2007), Corymbium has been traditionally allied to the Vernonieae though several morphological and chem- ical features suggest otherwise (see Bremer, 1994). Its com- bination of morphological characteristics distinctive in Asteraceae namely parallel-veined leaves, single flowered capitula, and broad, spreading corolla lobes have made it difficult to place. The style morphology is similar to Vern- onieae with long, slender style branches with papillae cov- ering the abaxial surface and distal half of style. The combination of morphological features presented by Corymbium precluded Bremer from assigning it to any par- ticular tribe of his Cichorioideae. Results from preliminary molecular studies of Corymbium (Jansen and Kim, 1996) suggested placement in Senecioneae. However, our results provide strong support for this lineage outside of both Asteroideae and Cichorioideae. 4.11. Asteroideae Our study provides strong statistical support for most tribal relationships of the Asteroideae, except the relation- ship of Senecioneae. The previously unknown positions of Calenduleae and Gnaphalieae were resolved with high sta- tistical support (95% and 93% BS, respectively, 100% PP). Calenduleae was found sister to a clade containing Gna- phalieae as sister to Anthemideae and Astereae. Other tri- bal relationships found here are congruent with the ndhF phylogeny of Kim and Jansen (1995) including the clade containing Inuleae sister to Athroismeae and the tribes of the Heliantheae alliance and a third clade containing mem- bers of tribe Senecioneae. The tribal relationships within the Heliantheae alliance are in agreement with recent mor- phological studies of the group (Panero, 2007b) and molec- ular phylogenetic results based on denser sampling (Panero, unpublished). As in earlier studies the position of Senecioneae was equivocal in our analyses. Senecioneae was either unresolved in Bayesian analyses or placed as sis- ter to the Calenduleae clade without significant support (52% BS) in Maximum Parsimony analyses. The three main lineages of the Asteroideae identified by previous molecular studies have been recently recognized as super- tribes of the subfamily: Asterodae, Helianthodae, and Senecionodae (Robinson 2005). The monophyly of Aster- oideae was strongly supported (80% BS, 100% PP) as expected based on earlier morphological and molecular studies. The Asteroideae are the largest subfamily of the Astera- ceae. It contains 1210 genera and approximately 17,000 species or 72% of the diversity of the family (Jeffrey, 2007). According to Bremer (1994), a majority of Asteroi- deae are characterized by the presence of true ray florets, disc corollas with short lobes, caveate pollen, and style branches with two marginal stigmatic surfaces. 5. Early dispersal of Asteraceae out of South America Our chloroplast phylogeny confirms again the South American origin of Asteraceae. Of the basal lineages of Asteraceae, Barnadesioideae and Stifftioideae are endemic to South America and Mutisioideae, Wunderlichioideae, and Gochnatioideae are primarily South American. Together these five lineages represent only about 4% of the species diversity of the family whereas the tribal divergences giving rise to approximately 96% of the spe- cies of the family occurred outside of South America. The great success of Asteraceae in terms of species diver- sification appears to have been contingent upon the dis- persal of diaspores out of South America and subsequent worldwide expansion. Recently the origin of Asteraceae has been estimated to postdate the breakup of Gondw- ana in the mid-Eocene to late Paleocene-Selandian (42- 47 Ma, Kim et al., 2005; 60 Ma, McKenzie et al., 2006). Thus the early evolution of Asteraceae would have occurred when South America was essentially an isolated land mass connected with North America and Africa only by island chains exposed by fluctuating sea levels (Sclater et al., 1977; Iturralde-Vinent and MacPhee, 1999). 772 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 The only speculation as to the route of sunflowers out of the continent of their origin has been that of Bremer (1994) who hypothesized a Pacific-Asian route from South Amer- ica. His proposal of dispersal to Asia via Hawaii was pred- icated upon erroneous inferences that the Asian genus Ainsliaea (Pertyeae, Pertyoideae) and Hawaiian genus Hes- peromannia (Vernonieae, Cichorioideae) were early diverg- ing members of the "Mutisieae clade". In light of our phylogeny the North American endemic genus Hecastocleis could be interpreted as a North American link to Pertyoi- deae in Asia; similar floral and pollen morphology may support this. Fossil pollen records confirm the wide distri- bution of Asteraceae in the Northern Hemisphere on both sides of the Pacific as early as the Eocene (North Amer- ica?eastern Texas, Elsik and Yancey (2000); northwestern China, Song et al. (1999)). However, to fully explain the radiation of the family out of South America by a North American-Asian route alone (Fig. 3, red arrows) would require that the mostly African Dicomeae-Tarchonantheae lineage be derived from an Asian ancestor (Fig. 3, open cir- cle; most recent common ancestor (MRCA) of Carduoi- deae and Perty oideae-Asteroideae). The morphological similarity of Oldenburgia (Tarchonantheae) and some Dicomeae to South American Mutisieae suggests a more direct South America-Africa connection. The presence of Pertyoideae in Asia could be postulated alternatively as derived from an African or Eurasian ancestor. Direct dispersal easterly from South America to Africa or Eurasia (Fig. 3, solid gray arrow) is plausible because the earliest branching clades outside of South America are the monotypic Hecastocleidoideae sister to a clade con- taining all other Asteraceae whose basal branches are mostly African (Tarchonantheae/Dicomeae, Carduoideae), mostly African/Eurasian (Cichorioideae, Gymnarrhenoi- deae), and Mediterranean/Central Asian (Cynareae, Car- duoideae) as well as the eastern Asian/Himalayan Pertyoideae. Of these lineages the greatest diversity of spe- cies by far was attained in Africa, the Mediterranean and Central Asia. The basal lineages of Carduoideae, Cicho- rioideae and Gymnarrhenoideae are African or Mediterra- South American grade *$%/ J^f j> ^ #' Earnadesioideae Wlutisicideae Stifftioideae Wunderlichioideae Gochnalioideae OutolSo^ Fig. 3. Alternative "Out of South America" hypotheses. Depending on where the ancestral area of the "Out of South America" clade is reconstructed, three scenarios are postulated. (1) If the ancestral area is African or Eurasian (gray star) then transatlantic dispersal to Africa or Eurasia gave rise to a global expansion of Asteraceae including the Hecastocleidoideae in North America (gray arrows). (2) If the ancestral area is North American (red star) then long distance dispersal or "island-hopping" gave rise to global expansion via northern hemisphere routes (red arrows). (3) If the ancestral area is South American (white star) then two original dispersal events are hypothesized: one that founded a lineage in Africa or Eurasia giving rise to most of Asteraceae diversity, and another to North America that has been far less successful. Stars indicate hypothesized ancestral areas for the MRCA of Hecastocleis and the Carduoideae-Asteroideae clade. Solid lines signify first steps of Asteraceae leaving South America and dashed lines migration after initial dispersal event(s). Open circle indicates MRCA of Carduoideae and Pertyoideae-Asteroideae. Base map of middle Eocene continents (~40 Ma) inferred from tectonic plate reconstruction of Lawver et al. (2002). J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 773 nean. Samples from Paleocene-Eocene pollen deposits in southwestern Africa have been attributed to classical Muti- sieae (Zavada and De Villiers, 2000; De Villiers and Cad- man, 2001) or more specifically to a Dicoma-likQ taxon and recently dated to the mid-Eocene, approximately 38 Ma (Scott et al., 2006). With slightly later pollen reported from Egypt (~34-36 Ma; Kedves, 1971) we assume the family was probably widespread on the African continent by the Late Eocene. Sweepstakes dispersal across the Atlantic has been concluded to explain taxonomic sim- ilarities between South America and Africa in studies of several angiosperm groups using molecular phylogenies calibrated by fossil evidence (Pennington and Dick, 2004; Lavin et al., 2004), or to explain Eocene to Miocene pollen of South American angiosperms in Africa (Morley, 2003). Of 110 extant genera with species on both sides of the Atlantic, wind dispersal specifically has been invoked to explain the presence of a few South American taxa in Africa (Renner, 2004). Sweepstakes dispersal has been the mode commonly invoked to explain the presence of Asteraceae outside South America (Raven and Axelrod, 1974; Stuessy et al., 1996). The earliest successful colonizations of areas outside South America by Asteraceae may well have been the result of two long distance dispersal events, or possibly stepping- stone migration, across oceanic barriers to North America (Fig. 3, solid red arrow) and Africa or Eurasia, most prob- ably Africa (Fig. 3, solid gray arrow). If true, the African dispersal gave rise to the explosive radiation of Asteraceae across the world resulting in the largest family of flowering plants, whereas the dispersal to North America was much less successful and today is represented only by a single spe- cies, Hecastocleis shockleyi. Hecastocleis illustrates that all sweepstakes winners may not be equally successful in terms of speciation and adaptation after colonization depending on where and when they arrive, and reinforces that long distance dispersal is a continuous process only rarely rewarded. After the first steps out or South America several per- mutations of Northern hemisphere dispersal could be pos- tulated. Hecastocleis could represent an important "non- missing link" lineage that gave rise to worldwide diversity via the Bering land bridge or the North American land bridge, or both. Alternatively, Hecastocleis could be the only relict of an Asteraceae lineage that reached North America from the Old World by either route. The impor- tance of the North American land bridge to explain North American-African disjunctions has gained recent attention (Davis et al., 2002 and other references therein). However, we note that despite abundant sampling the earliest fossil evidence of Asteraceae in Europe dates only from the late Oligocene or early Miocene (Graham, 1996), later than in either North America or Africa. The high number of Asteraceae species of different lin- eages sympatric in most biomes of the world is indicative of a complex biogeographical history that began with first steps out of South America. Rigorous testing of alternative hypotheses, including those proposed here of dual dispers- als out of south America and more complex scenarios that invoke land bridge migrations and longer routes between the New and Old Worlds, depends upon reconstructing ancestral areas (sensu Bremer, 1993) and dating the major divergences of the family described here, and will appear in future studies. 6. Phylogenetic reconstruction of closely spaced cladogenesis The historical difficulty in solving the relationships of the deep divergences of Asteraceae represented by members of classical Mutisieae is probably due to closely spaced cladogenic events. Informative characters supporting these branches are relatively few. By increasing the number of characters sampled approximately sixfold over earlier sin- gle marker studies and extending taxon sampling to include a good representation of the deep divergences of the family, we were able to amplify the phylogenetic signal to resolve most branches with strong statistical support. As data par- titions were successively concatenated, preliminary analy- ses yielded increasing resolution and support for branches as we had expected. Future expansion of this data matrix will likely allow the placements of Stifftioideae and Sene- cioneae to be resolved also. Although systematic bias can also be amplified as more data are added as in the case of the long-branch attraction problem (Felsenstein, 1978), our phylogeny appears robust to different phylogenetic methods. Both maximum parsimony and the general time reversible modeled Bayesian method yielded highly congru- ent topologies. 7. Comparison with phylogenetic hypothesis from the nuclear compartment We would like to compare our plastome phylogeny with an estimate based on data obtained from the nuclear com- partment. Consensus among independent studies would corroborate findings, however we expect that the evolu- tionary histories of the two compartments differ somewhat since modes of inheritance differ and the nuclear genome is subject to significant recombination in contrast to the plas- tid genome (Clegg and Zurawski, 1992). Incongruence between phylogenetic estimates could also be useful to dis- cover reticulation, incomplete lineage sorting, and evidence of horizontal gene transfer. Asteraceae provides well- known and important examples of hybridization in the evolution of plants (Rieseberg et al., 2003; Comes and Abbott, 2001; Francisco-Ortega et al., 1996; Schilling and Panero, 1996) and we like to investigate the extent to which reticulation may have played a role in the evolution of the major lineages of sunflowers. Only one study of Asteraceae has sampled the nuclear compartment across the entire family, a combined analyses of the internal transcribed spacers of the 18S-5.8S-26S nuclear ribosomal cistron (ITS1 and ITS2) by Goertzen et al. (2003). Maximum parsimony analysis of 288 taxa 774 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 resulted in more than 34,000 trees, the strict consensus of which was summarized as a tribal phylogeny (17 termi- nals). The generic composition of tribal clades (one MP tree shown) is highly congruent past chloroplast studies and our results. However, relationships between tribal clades are highly incongruent with both. Goertzen et al. (2003) nonetheless characterize their ITS results as having considerable topological congruence of major lineages of the family with chloroplast studies when bootstrap values are considered. To support this assertion they compare a mostly unresolved 50% bootstrap consensus ITS tree with a mostly resolved ndhF tree for comparable analyses of a reduced taxon set (82-taxon matrices). Since meaningful discussion of tribal relationships based on the bootstrap consensus is ineffectual, the authors shift their comparison of ITS with chloroplast results to a strict consensus over- view (Fig. 2, Goertzen et al., 2003) that shows resolved tri- bal relationships, and we will compare our results to that tree also. Goertzen et al. cite the monophyly of Asteroi- deae, paraphyly of Cichorioideae (defined to include Muti- sieae) and the position of classical Mutisieae as congruent with chloroplast studies. They also cite examples of incon- gruence with past studies including the phylogenetic posi- tions of tribes Arctotideae, Cichorieae, Cynareae, Liabeae, and Vernonieae. Comparing their results to those of Kim and Jansen (1995) further incongruences include the position of Anthemideae, the sister taxon relationship of Calenduleae to Senecioneae, the sister taxon relationship of the previous to Inuleae and Plucheeae, and the sister taxon relationship of Astereae to Gnaphalieae. Our results are congruent with the ITS strict consensus topology in only three respects: (1) placing the Mutisioi- deae as the next lineage to split above Barnadesioideae, (2) placing Heliantheae Alliance sister to Athroismeae, and (3) placing Inuleae sister to Plucheeae (now Inuleae, Anderberg et al., 2005). The ITS topology showing Cic- horieae as sister to the other tribes of Asteraceae except Barnadesioideae and Mutisieae is incongruent with our results and other studies based on sequence data of chloro- plast markers (Kim and Jansen, 1995; Karis et al., 2001). Cynareae placed as sister to Arctotideae and Liabeae in the ITS study is also incongruent with our results as well as those of past sequence studies based on the ndhF gene. The tribal relationships within Asteroideae reconstructed in the ITS study are vastly in disagreement with our and other published results based on chloroplast sequence data (Kim and Jansen, 1995; Kim et al., 2005). Goertzen et al. (2003) trace the topological incongruence between their ITS and earlier chloroplast results to unspec- ified analytical or biological phenomena. Comparing the 82-taxon ITS and chloroplast ndhF trees in that study, the ndhF alone had far more power to resolve relationships among the tribal lineages than did the ITS; the relatively lower number of informative characters ITS provides for resolving these relationships (380 ITS vs. 465 ndhF) and higher homoplasy (homoplasy index: 0.88 ITS vs. 0.61 ndhF) resulted in low bootstrap values supporting tribal splits in the ITS trees. Their study highlights that when ITS is the only marker used, it appears to be more appro- priate for phylogenetic comparisons at lower taxonomic levels rather than addressing deep divergences in Astera- ceae. This conclusion resonates with that of Bailey et al. (2006) who found ITS provided only limited resolution of deeper nodes of Brassicaceae. If the evolutionary histories of the nuclear and chloro- plast compartments in Asteraceae truly differ so extensively at the tribal relationships it would be remarkable, and could suggest an even greater role of hybridization in gen- erating Asteraceae diversity. However, this interpretation of the incongruence between these nuclear and plastome topologies is confounded by other factors that can lead to conflicting phylogenetic signal, namely sampling error and homoplasy arising from the assessment of sequence orthology as well as nucleotide substitution saturation. Our chloroplast and the ITS study differ significantly in both dimensions of the data matrices. Overall more than twice as many taxa were sampled in the ITS study, while sampling of the basal lineages of Mutisioideae, Stifftioi- deae, Wunderlichioideae, Gochnatioideae, and Hecasto- cleidoideae is nearly four times greater in our study (43 in the chloroplast study compared to 11 in the ITS study). Increased taxon sampling is thought to reduce phylogenetic error (Wheeler, 1992; Graybeal, 1998; Zwickl and Hillis, 2002; Pollock et al., 2002) primarily through the subdivi- sion of long branches (Lyons-Weiler and Hoelzer, 1997; Purvis and Quicke, 1997; Poe and Swofford, 1999). When long-branch attraction (Felsenstein, 1978) is a problem, accuracy is apparently facilitated when taxa are added at deeper nodes near the base of long branches rather than at the tips of long branches (Geuten et al., 2007; Graybeal, 1998; Poe, 2003), but the existence of long-branch prob- lems in either of these Asteraceae data sets has not been demonstrated. Taxon sampling differences must account for some different placements but may not fully explain incongruence in tribal relationships in so many branches across the tree. Hypothesizing positional homology of nucleotides from hypervariable spacer data does present a challenge for alignment of more divergent taxa (Baldwin et al., 1995; Kim and Jansen, 1995). In the ITS study, sam- ples were first aligned manually by roughly tribal groups, then tribal groups aligned with the aid of 80% consensus sequences based on each group. The high correspondence of generic composition of each tribal lineage contrasts starkly with the low correspondence of tribal relationships between the ITS and chloroplast studies; it is not clear what effect the primary assessment of homology may have had on these ITS results. Also unexplored is the possibility of unidentified divergent paralogues among the ITS sequences assembled from various sources that could mislead phylo- genetic inference (Sanderson and Doyle, 1992; Buckler et al., 1997; Bailey et al., 2006). Alvarez and Wendel (2003) recently reviewed some characteristics of nrlTS and advocate the alternative use of low copy nuclear genes for plant phylogenetic studies. A well-resolved statistically J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 775 supported nuclear phylogeny based on balanced taxon sampling is still needed to compare with chloroplast results to understand the organismal phylogeny of Asteraceae. 8. Conclusions With this study a few more branches of the sunflower tree of life have been identified and a new paradigm has been established in Asteraceae systematics. We now know with confidence that the family comprises many more major lineages than previously thought. The 'Mutisieae problem' has been mostly solved as the component lin- eages have been identified and recognized as new subfam- ilies of the Asteraceae. However, the position of Stifftioideae is still equivocal and the phylogenetic rela- tionships among the three main lineages of the Cichorioi- deae are still problematic. Our results support the close relationship of Liabeae and Vernonieae, but failed to identify the relationship of this lineage to either Arctoti- deae or the Cichorieae-Gundelieae clades. Equivocal too are the phylogenetic position of Senecioneae within Aster- oideae and some of the tribal relationships of the Helian- theae alliance. Our phylogeny clearly shows that from its South American origin the great diversity of the family was obtained in the Old World with the subsequent reintro- ductions of these lineages into the New World and Aus- tralia. Although the Stifftioideae and Wunderlichioideae are considered to be the most characteristic members of the Guayana flora (Berry and Riina, 2005) the Guayana Highlands did not give rise to any major lineages of the Asteraceae. The phylogenetic position of the North Amer- ican endemic Hecastocleis sister to the Carduoideae- Asteroideae clade precipitates new hypotheses about how sunflowers may have first expanded beyond the con- tinent of their origin, and we suggest that dual long dis- tance dispersals to North America and to Africa met with dual fates. The capacity of Asteraceae to disperse long distances and establish successfully in other habitats is demonstrated again with one new example revealed here: Nouelia and Leucomeris are Asian genera whose sis- ter taxa are endemic to South America. Several lineages of the Asteraceae have experienced sig- nificant cladogenesis resulting in the formation of thou- sands of species. The first step in recognizing a pattern attributable to adaptive radiation is the differential diversi- fication among lineages that results from differences in extinction and speciation rates (Guyer and Slowinski, 1993). We have identified several lineages whose sister taxa contain significantly larger numbers of species and includ- ing: Cyclolepis/Gochnatioideae (1:74 spp.); Gundelieae/ Cichorieae (2:1500 spp.); Liabeae/Vernonieae (190:1000 spp.); Hecastocleis/Carduoideae-Asteroideae clade (1:22,000 spp.); Gymnarrhena/Cichorioideae, Asteroideae, and Corymbioideae (1:20,000 spp.); Corymbium/Asteroi- deae (9:17,000 spp.); Calenduleae/Anthemideae, Astereae, and Anthemideae (120:6200 spp.). The relatively recent ori- gin of the family and the extraordinary cladogenesis of some of its more derived lineages suggest the family may contain groups (e.g. Astereae) with some of the fastest diversification rates in the flowering plants. With a strong phylogenetic framework and denser taxon sampling future studies aimed at comparing salient features of the morphol- ogy, chemistry, breeding systems, pollination, and herbiv- ory of the taxa contained in these lineages may document traits responsible for the extraordinary global diversifica- tion of Asteraceae. Even denser taxon sampling is still needed to address important macroevolutionary questions in Asteraceae. Our study demonstrates that despite the large size of the family more inclusive taxonomic and character sampling is profitable. We have demonstrated the utility of five chloroplast markers new to Asteraceae studies, with pri- mer sequences and protocols for these markers tested across all tribes of the family. With the increasing ease of gathering sequence data these sequences provide an infrastructure on which even larger supermatrices can be built in the near future. Our hypothesis of phylogenetic relationships among the major clades of Asteraceae, based on preliminary results of this study, has already been uti- lized as the backbone of a metatree (or informal supertree sensu Bininda-Emonds, 2004) that summarizes phyloge- netic hypotheses of many independent studies at various taxonomic levels across the family (Funk et al., 2005). Source trees based on molecular studies of Asteraceae have typically sampled within existing taxonomic catego- ries with little or no taxon overlap among studies. As the present study points out, existing suprageneric taxa in Asteraceae may or may not be monophyletic. Formal supertree methods using source trees available prior to our study could not have identified the lineages found here. Furthermore, there is no reason to expect that our study has uncovered all the major lineages of the family. A continually updated metatree for the family can be obtained at http://www.tolweb.org/asteraceae. The surprising number of new major lineages found in this study results from sampling genera identified by mor- phological studies as anomalous, in the context of denser taxon and broader character sampling than in previous studies. Our strategy could work well for the design of supermatrix studies aimed at building a backbone for big- ger trees at the family and ordinal levels. Empirical studies have demonstrated the value of a multigene approach to building backbone phylogenies for large flowering plant families (Bailey et al., 2006; Potter et al., 2007) and have stressed the importance of broad taxon sampling. Sampling anomalous and transitional genera in the context of dense and balanced sampling should also be considered a prior- ity, as some of these underrepresented taxa may constitute novel lineages. Although the explosive radiation of Astera- ceae is not as recent as previously thought (Kim et al., 2005), Carlquist's (1976) expectation that transitional gen- era should be extant has been validated by this molecular study. 776 J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757?782 Acknowledgments We thank M. Bonifacino, M. Koekemoer, P. Simon, and T. Trinder-Smith for collecting plant material in South America and the Republic of South Africa while collabo- rating with V.A.F. We thank S. Keeley for providing DNA of Hesperomannia, R.K Jansen for providing DNA of Inula, E.E. Schilling for providing DNA of Stevia and Helianthus, and Bruce Baldwin for providing DNA of Layia heterotricha. We also thank Tetsukazo Yahara and Aiko Ohtsuka of Kyushu University, Japan, for providing DNA of Pertya and Ainsliaea. We are grateful to the Hes- ler Fund, Herbarium, University of Tennessee, Knoxville, for providing funds to cover costs associated with sequenc- ing of the rpoB gene. We are grateful to B. Crozier for assistance in the laboratory, Bayesian analyses, and critical reading of the manuscript. We thank Billie L. Turner, Ja- vier Francisco-Ortega, and an anonymous reviewer for reading the manuscript and providing helpful comments. Funds from the National Science Foundation DEB 03- 44116 (to J.L.P.) and the Mellon Foundation (to V.A.F.) supported the molecular studies and fieldwork in North America; a Smithsonian Scholarly Studies (to V.A.F.) grant funded additional fieldwork in South America. Appendix A. Appendix Voucher information and GenBank Accession numbers for sequences used in this study. Voucher information listed in the following order: taxon name, collection, coun- try of origin, herbarium. Genbank numbers listed in the following order: trnK intron and matK, ndhD, ndhl, ndhF, rbch, rpoB, rpoCl exonl, 23S-trnA spacer, trnL intron- trnL-F spacer (in some taxa 2 Genbank numbers comprise this region). ND, missing sequence. Achillea millefolium L., Panero 2002-55, USA, TEX. EU385315, EU385219, EU243242, EU385124, EU384938, EU385410, EU385506, EU243147, EU385030. /Wcmp&z spathulata R. Br., Salgado 7660, Brazil, TEX. EU385316, EU385220, EU243243, EU385125, EU384939, EU385411, EU385507, EU243148, EU385031. v4cowrfia W,mafa (La Llave & Lex.) DC, Panero 2891, Mexico, TEX. EU385317, EU385221, EU243244, EU385126, EU384940, EU385412, EU385508, EU243149, EU385032. /f&nocaw- lon chilense Less., Simon 382, Argentina, US. EU385319, EU383223, EU243246, EU385128, EU384942, EU385414, EU385510, EU243151, EU385034. /WeMocWon 6:Ww Hook., Twisselmann 7661, USA, TEX. EU385320, EU385224, EU243247, EU385129, EU384943, EU385415, EU385511, EU243152, EU385035. Ainsliaea apiculata Sch. Bip ex. Zoll., Ohtsuka s.n., Japan, no voucher. EU385321, EU385225, EU243248, EU385130, EU384944, EU385416, EU385512, EU243153, EU385036. Ainsliaea macrocephala (Mattf.) Y.Q. Tseng, Bartholomew and Buf- ford 6167, Taiwan, US. EU385322, EU385226, EU243249, EU385131, EU384945, EU385417, ND, EU243154, EU385037. Aphyllocladus spartioides Wedd., Simon 508, Argentina, US. EU385323, EU385227, EU243250, EU385132, EU384946, EU385418, EU385513, EU243155, EU385038. Arctotis hirsuta (Harv.) P. Beauv., Panero 2002-61, cultivated, seed source: Kirstenboch Botanical Garden, South Africa, TEX. EU385224, EU385228, EU243251, EU385133, EU384947, EU385419, EU385514, EU243156, EU385039. Athroisma gracile (Oliv.) Mattf. ssp. psyllioides (Oliv.) T. Eriksson, Eriksson, Kalema, and Leliyo 559, Tanzania, TEX. AY215765, AF384437, AF383757, L39455, AY215085, AY213763, EU385515, AY216277, AY216019/AY216144. Xfrncfy/zj cwicef/ofa L, Panero 7098, Spain, TEX. EU385325, EU385229, EU243252, EU385134, EU384948, EU385420, EU385516, EU243157, EU385040. Baccharis neglecta Britton ex Brit- ton and A. Br., Panero 2002-31, USA, TEX. EU385326, EU385230, EU243253, EU385135, EU384949, EU385421, EU385517, EU243158, EU385041. Barnadesia spinosa L. f., Panero and Crozier 8492, Argentina, TEX. EU385327, EU385231, EU243254, L39394 {Barnadesia caryophylla (Veil.) S.F. Blake), AY874427 {Barnadesia caryophylla), EU385422, EU385518, EU243159, EU385042. aerk/wryo purpurea (DC.) Mast., Panero 2002-49, cultivated, seed source: Kirstenboch Botanical Garden, South Africa, TEX. EU385328, EU385232, EU243255, EU385136, EU384950, EU385423, EU385519, EU243160, EU385043. Blepharispermum zanguebaricum Oliv. and Hiern., T. Eriks- son 604, Kenya, TEX. AY215768, AF384440, AF383760, L39456, AY215088, AY213766, ND, AY216280, AY216022/AY216147. Brachyclados caespitosus (Phil.) Speg., Bonifacino 459, Argentina, US. EU385329, EU385233, EU243256, EU385137, EU384951, EU385424, EU385520, EU243161, EU385044. Brachylaena elliptica (Thunb.) DC, Koekemoer and Funk 1971, South Africa, US. EU385330, EU385234, EU243257, EU385138, EU384952, EU385425, EU385521, EU243162, EU385045. Carthamus tinctorius L., al-Hosseini s.n., Iran, US. EU385331, EU385235, EU243258, EU385139, EU384953, EU385426, EU385522, EU243163, EU385046. Ce/zmurea melitensis L, Panero 2002-48, USA, TEX. EU385332, EU385236, EU243259, EU385140, EU384954, EU385427, EU385523, EU243164, EU385047. Centratherum puncta- tum Cass., Panero 2002-53, USA cultivated, TEX. EU385333, EU385237, EU243260, EU385141, EU384955, EU385428, EU385524, EU243165, EU384048. C/zaefaM- thera pentacaenoides (Phil.) Hauman, Bonifacino 293, Argentina, US. EU385334, EU385238, EU243261, EU385142, EU384956, EU385429, EU385525, EU243166, EU384049. Chaptalia nutans (L.) Pol., Panero 2002-19, USA, TEX. EU385335, EU385239, EU243262, EU385143, EU384957, EU385430, EU385526, EU243167, EU385050. Chimantaea humilis Maguire, Steyermark and Wurdack, Weitzman et al. 412, Venezuela, US. EU385336, EU385240, EU243263, EU385144, EU384958, EU385431, EU385527, EU243168, EU385051. Chrysanthe- moides monilifera (L.) Norl., Panero 2002-5, cultivated, seed source: Kirstenbosch Botanical Garden, South Africa, TEX. EU385337, EU385241, EU243264, EU385145, J.L. Panero, V.A. Funk I Molecular Phylogenetics and Evolution 47 (2008) 757-782 777 EU384959, EU385432, EU385528, EU243169, EU385052. Chuquiraga spinosa Less., Simon 522, Argentina, US. EU385338, EU385242, EU243265, EU385146, EU384960, EU385433, EU385529, EU243170, EU385053. Cmcof/wzm- nus lorentzii Griseb., Panero 1934, Argentina, TENN. EU385339, EU385243, EU243266, EU385147, EU384961, EU385434, EU385530, EU243171, EU385054. Corymbium glabrum L., Moffett 8764, South Africa, TEX. EU385340, EU385244, EU243267, EU385148, EU384962, EU385435, EU385531, EU243172, EU385055. Cyc/o/epK genzjfoz&j D. Don, Bonifacino 3, Argentina, US. EU385341, EU385245, EU243268, EU385149, EU384963, EU385436, EU385532, EU243173, EU385056. fkMyp/zy//wm refzcw&z- tum (DC.) Cabrera, Roque, Funk & Kim 485, Brazil, US. EU385342, EU385246, EU243269, EU385150, AY874428 {Dasyphyllum argenteum Kunth. in H.B.K.), EU385437, EU385533, EU243174, EU385057. Dicoma capensis Less., Trinder-Smith 349, South Africa, US. EU385344, EU385247, ND, EU385152, EU384965, EU385439, EU385534, EU243176, EU385059. Dicoma sp., Funk 1960, South Africa, US. EU385343, EU385248, EU243270, EU385151, EU384964, EU385438, ND, EU243175, EU385058. Dimorphoteca sinuata DC, Panero 2002-3, cultivated, seed source: Kirstenbosch Botanical Garden, South Africa, TEX. EU385345, EU385249, EU243271, EU385153, EU384966, EU385440, EU385535, EU243177, EU385060. Dinoseris salicifolia Griseb., Simon 330, Argentina, US. EU385346, EU385250, EU243272, EU385154, EU384967, EU385441, EU385536, EU243178, EU385061. Dolichlasium lagascae D. Don, Simon 811, Argentina, US. EU385347, EU385251, EU243273, EU385155, EU384968, EU385442, EU385537, EU243179, EU385062. Doniophyton anomalum (D. Don) Kurtz, Bonifacino 96, Argentina, US. EU385348, EU385252, EU243274, EU385156, EU384969, EU385443, EU385538, EU243180, EU385063. DwK&zea fmz/b&z S. F. Blake, V. A. Funk 8010, Venezuela, US. EU385349, EU385253, EU243275, EU385157, EU384970, EU385444, EU385539, EU243181, EU385064. Echinops ritro L., Panero 2002-71. cultivated, TEX. EU385350, EU385254, EU243276, EU385158, EU384971, EU385445, EU385540, EU243182, EU385065. Eremanthus erythropappus (DC.) MacLeish, Acosta 1661, Brazil, TEX. EU385351, EU385255, EU243277, EU385159, EU384972, EU385446, EU385541, EU243183, EU385066. Erigeron tenuis Torr. and Gray, Panero 2002-25, USA, TEX. EU385352, EU385256, EU243278, EU385160, EU384973, EU385447, EU385542, EU243184, EU385067. fWzcza Wen#y//a (Cass.) Grau, Panero 2002-1, cultivated, seed source: Kirstenbosch Botanical Garden, South Africa, TEX. EU385353, EU385257, EU243279, EU385161, EU384974, EU385448, EU385543, EU243185, EU385068. Gamochaeta pensylva- nica (Willd.) Cabr., Panero 2003-27, USA, TEX. EU385354, EU385260, EU243282, EU385162, EU384977, EU385449, EU385544, EU243188, EU385070. Gerbera ser- rata (Thunb.) Druce, Koekemoer 2001, South Africa, US. EU385356, EU385258, EU243281, EU385164, EU384976, EU385451, ND, EU243187, EU385069. GerWoffAWb- ides (L.) Cass., Koekemoer and Funk, 1972, South Africa, US. EU385355, EU384259, EU243280, EU385163, EU384975, EU385450, EU385545, EU243186, ND. Gochnatia hiriartiana Medrano, Villasenor and Medina, Panero MEX-2, Mexico, TEX. EU385358, EU385262, EU243284, EU385166, EU384979, EU385453, ND, EU243190, EU385072. Goc/znafza /r%Wewca (DC.) A. Gray, Panero MEX-1, Mexico, TEX. EU385357, EU385261, EU243283, EU385165, EU384978, EU385452, EU385546, EU243189, EU385071. GwzgyMepzj 6. Stuessy, T.F., Garver, D., 1996. The defensive role of pappus in heads of Compositae. In: Caligari, P.D.S., Hind, D.J.N. (Eds.), Compositae: Biology & Utilization. Proceedings of the International Compositae Conference, Kew, 1994. Royal Botanic Gardens, Kew, vol. 2, pp. 81- 91. Stuessy, T.F., Sang, T., DeVore, M.L., 1996. Phylogeny and biogeography of the subfamily Barnadesioideae with implications for early evolution of the Compositae. In: Hind, D.J.N., Beentje, H.J. (Eds.), Compositae: Systematics. Proceedings of the International Compositae Conference, Kew, 1994. Royal Botanic Gardens, Kew, vol. 1, pp 463^190. Swofford, D.L., 2003. PAUP*. Phylogenetic Analysis Using Parsimony ('and Other Methods), version 4.0M0. Sinauer Associates. Taberlet, P., Gielly, I., Patou, G, Bouvet, J., 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Mol. Biol. 17, 1105-1109. Tellerfa, M.C., Katinas, L., 2004. A comparative palynological study of Chaetanthera (Asteraceae, Mutisieae) and allied genera. Syst. Bot. 29, 752-773. Telleria, M.C., Katinas, L., 2005. The unusual occurrence of tricolpate pollen within Mutisieae (Asteraceae). Grana 44, 91-97. Thome, R., 1983. Proposed new realignments in the angiosperms. Nord. J. Bot. 3, 85-117. Vamosi, J.C., Dickinson, T.A., 2006. Polyploidy and diversification: a phylogenetic investigation in Rosaceae. Int. J. Plant Sci. 167, 349-358. Wagenitz, G, 1976. Systematics and phylogeny of the Compositae (Asteraceae). Plant Syst. Evol. 125, 29-46. Wagstaff, S.J., Breitwieser, I., Swenson, U, 2006. Origin and relationships of the austral genus Abrotanella (Asteraceae) inferred from DNA sequences. Taxon 55, 95-106. Watanabe, K, Kosuge, K, Simamura, R., Konishi, N., Taniguchi, K., 2006. Molecular systematics of Australian Calotis (Asteraceae: Aste- reae). Austral. Syst. Bot. 19, 155-168. Watson, L.E., Evans, T.E., Boluarte, T., 2000. Molecular phylogeny of tribe Anthemideae (Asteraceae), based on chloroplast gene ndh?. Mol. Phylogenet. Evol. 15, 59-69. Weitz, F., 1989. A revision of the genus Corymbium (Asteraceae). S. Afr. J. Bot. 55, 598-629. Wheeler, W.C., 1992. Extinction, sampling and molecular phylogenetics. In: Novacek, M.J., Wheeler, Q.D. (Eds.), Extinction and Phylogeny. Columbia University Press, New York, pp. 205-215. Woodhouse, R.P., 1929. Pollen grains in the identification and classifica- tion of plants. IV. The Mutisieae. Am. J. Bot. 16, 297-313. Zardini, E.M., 1974. Sobre la presencia del genero Gerbera en America. Bol. Soc. Argent. Bot. 16, 103-108. Zavada, M.C., De Villiers, S.E., 2000. Pollen of Asteraceae from the Paleocene-Eocene of South Africa. Grana 39, 39^45. Zhao, Z., Skvarla, J.J., Jansen, R.K., 2006. Mutisieae (Asteraceae) pollen ultrastructure atlas. Lundellia 9, 51-76. Zwickl, D.J., Hillis, DM., 2002. Increased taxon sampling greatly reduces phylogenetic error. Syst. Biol. 51, 588-594.