53 (3) • August 2004: 637–655 Funk & al. • Evolution of the tribe Arctoteae A S T E R AC E A E Insights into the evolution of the tribe Arctoteae (Compositae: subfamily Cichorioideae s.s.) using trnL-F, ndhF, and ITS Vicki A. Funk1, Raymund Chan2 & Sterling C. Keeley2 1 U.S. National Herbarium, Smithsonian Institution MRC 166, P.O. Box 37012, Washington D.C. 20013 U.S.A. funk.vicki@nmnh.si.edu (author for correspondence) 2 Department of Botany, University of Hawaii at Manoa, Hawaii 96822, U.S.A. raymundc@hawaii.edu; sterling@hawaii.edu Compositae (Asteraceae) are the largest flowering plant family (23,000 to 30,000 species) and its members are found throughout the world in both temperate and tropical habitats. The subfamilies and tribes of Compositae remained relatively constant for many years; recent molecular studies, however, have identified new subfamilial groups and identified previously unknown relationships. Currently there are 35 tribes and 10 subfamilies (Baldwin & al., 2002; Panero & Funk, 2002). Some of the tribes and subfamilies have not been tested for monophyly and without a clear understanding of the major genera that form each tribe and subfamily, an accurate phylogeny for the family cannot be reconstructed. The tribe Arctoteae (African daisies) is a diverse and interesting group with a primarily southern African distribution (ca. 17 genera, 220 species). They are especially important in that most of the species are found in the Cape Floral Kingdom, the smallest floral kingdom and the subject of intense conservation interest. Arctoteae are part of the monophyletic subfamily Cichorioideae s.s. Other tribes in the subfamily include Eremothamneae, Gundelieae, Lactuceae, Liabeae, Moquineae, and Vernonieae, and these were all evaluated as potential outgroups. Ultimately 29 ingroup taxa and 16 outgroup taxa with a total of 130 sequences (125 newly reported), from three genetic regions, two from chloroplast DNA (trnL-F and ndhF) and one from the nuclear genome (ITS), were used to evaluate the tribe and its proposed outgroups. Each molecular region is examined separately, the chloroplast markers are examined together, and the data are combined. The data were analyzed with and without outgroups and problem taxa using parsimony and maximum likelihood methods. The analyses showed robust support for two outgroup clades, Liabeae-Vernonieae and Gundelieae-Lactuceae and two main subtribes within Arctoteae: Arctotineae and Gorteriinae. Support for monophyly of Arctoteae is weak. Within Arctoteae, some taxa of interest are easily placed: Didelta, Cuspidia and Heterorhachis are consistently part of subtribe Gorteriinae, Cymbonotus, the Australian genus, is nested within subtribe Arctotineae, and Haplocarpha is at the base of Arctotineae. Berkheya, Haplocarpha, and Hirpicium are probably paraphyletic. Furthermore, Platycarpha most likely does not belong in Arctoteae, and Heterolepis and the tribe Eremothamneae are within Arctoteae but not within either of the two main subtribes. After some rearrangements, the two main subtribes, Arctotineae and Gorteriinae, are monophyletic and the latter has three clades. The study shows that the unusual taxa are of critical importance, and they should be included in any molecular analysis. Adequate representation of the ingroup is also important as all previous studies of Arctoteae had involved only a few taxa from the core subtribes, and so did not reveal the problems. Multiple outgroups evaluated in an iterative manner had pronounced effects on the relationships within the ingroup, not only on the position of the root. Finally, unrooted consensus trees and unrooted phylograms were found to be very useful in analyzing the data, allowing for examination of placement of taxa without the bias of a rooted tree. KEYWORDS: Arctoteae, Asteraceae, Cape Floral Kingdom, Cichorioideae, Compositae, ITS, molecular phylogenetics, ndhF, outgroups, South Africa, trnL. INTRODUCTION The Compositae family has the largest number of species of any family of seed plants (23,000–30,000), and today its members can be found on every continent except Antarctica. The subfamilial classification has had minor revisions, but until recently it had not seriously changed since the 13 tribes of Bentham (Bentham, 1873a, b). The advent of DNA sequence data from the chloroplast and nuclear genomes has changed Compositae systematics in dramatic ways: first, they have identified basal clades that turned the ideas about 637 Funk & al. • Evolution of the tribe Arctoteae evolution within the family upside down (Jansen & al., 1991; Kim & Jansen, 1995) and second, more recently by revisions that have divided the family into 10 subfamilies and 35 tribes (Baldwin & al., 2002; Panero & Funk, 2002; Panero & al., unpubl.). Many of the traditional tribes had broad distributions covering hemispheres (i.e., Lactuceae, Heliantheae s.l.), or nearly the whole globe (e.g., Senecioneae). Three of the more modestly sized tribes, Arctoteae (southern Africa and Australia), Calenduleae (southern Africa), and Liabeae (Andean South America, Central America, Mexico, and the Caribbean), have rather restricted distributions and therefore lend themselves to detailed systematic and biogeographic investigations. One of these, Arctoteae, is the subject of our investigation. This tribe is particularly interesting not only because it is more or less confined to southern Africa, but also because most of its species are endemic to the Cape Floral Kingdom. The tribe Arctoteae belongs to the redefined subfamily Cichorioideae s.s. (Panero & Funk, 2002). The subfamily contains from four to seven tribes, depending on where the member taxa are found in the final analysis and on one’s personal philosophy with regard to Compositae classification. The four traditional and larger tribes are Arctoteae (African daisies), Lactuceae (or Cichorieae; dandelions), Liabeae (Andean sunflowers), and Vernonieae (ironweeds). The three small tribes (with one or two genera each) are Eremothamneae (Eremothamnus and Hoplophyllum; authors for all genera are found in Table 1), Gundelieae (Gundelia, crown 53 (3) • August 2004: 637–655 of thorns), and Moquineae (Moquinea and Pseudostifftia). None of the relationships among the tribes have been resolved, and some of the taxa are not universally accepted at the tribal level. For purposes of this study, Vernonieae, Liabeae, and Lactuceae were included as potential outgroups for Arctoteae. The tribe Moquineae was assumed to be close to, if not actually in, the Vernonieae clade and was not included. The genera of Eremothamneae were treated as part of Arctoteae and Gundelia was originally treated as part of Arctoteae, but it eventually fell outside the tribe. Data from a variety of sources (Cassini, 1816, 1821; Beauverd, 1915a, b; Robinson & Brettell, 1973a; Norlindh, 1977; Robinson, 1992, 1994; Bremer, 1994; Herman & al., 2000; Leistner, 2000) as well as personal observations show that the tribe Arctoteae does not have an abundance of characters to support its monophyly. Beginning with Cassini, the traditional definition of Arctoteae was based on the presence of a swollen area with a ring of hairs located just below the branch point on the style. While there are hairs in a ring just below the division in the style, this character is also found in some thistles, and the presence of a swollen area is found mostly in the subtribe Arctotinae. Members of the two main subtribes of Arctoteae have sagittate but not tailed anthers (except for Arctotis which has calcarate, short caudate anthers), and the heads usually have both ray and disc florets. The stigmatic papillae are spread over the inner surfaces of the style branches, but this character is found in many tribes in the bottom third of the family Table 1. Highlights of classification and nomenclature of the tribe Arctoteae (Compositae). Numbers in parentheses indicate the number of species in the genus. Traditional (modified Bentham, 1873b) Robinson & Brettell, 1973 Subtribe Arctotinae Less. Arctotheca Wendl. (4) Arctotis L. (~64) Cymbonotus Gaud. (2) Dymondia Compton (1) Haplocarpha Less. (10) Subtribe Gorteriinae Benth. & Hook. f. Berkheya Ehrh. (~75) Cullumia R. Br. (15) Cuspidia Gaertn. (1) Didelta L’Herit (2) Gazania Gaertn. (17) Gorteria L. (3) Heterolepis Cass. (3) Heterorhachis Sch. Bip. (1) Hirpicium Cass. (12) Subtribe Gundeliinae Bentham Gundelia L. (1) Tribe Gundelieae H. Rob. & Brettell Eremothamnus O. Hoffm. (1) Tribe Eremothamneae H. Rob. & Brettell Platycarpha Less. (3) Tribe Cynareae Hoplophyllum DC. (2) Ursinia Gaertn. 638 Tribe Ursinieae H. Rob. & Brettell Heywood & al. (various chapters), 1977 Bremer, 1994 Arctotinae Arctotinae Arctotinae Arctotinae Arctotinae Arctotinae Arctotinae Arctotinae Arctotinae Arctotinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae by Norlindh, Mutisieae by Merxmüller Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae Gorteriinae unassigned to subtribe Subtribe Gundeliinae Subtribe Eremothamninae Lenis Tribe Cynareae (Norlindh) (rejected by Dittrich) Tribe Vernonieae Tribe Anthemideae Subtribe Gorteriinae unassigned to tribe unassigned to subtribe Gorteriinae Gorteriinae unassigned to tribe Tribe Anthemideae 53 (3) • August 2004: 637–655 phylogeny. The pappus is usually made up of scales (absent in Cymbonotus and Cullumia), but Heterolepis has what has been described as “bristle-like” scales. Most members of the tribe are annual or perennial herbs, but some are shrubs or subshrubs; most have yellow flowers, but there are some noticeable exceptions such as Arctotis and Gazania (Figs. 1B, C, E). All of the species in the Arctoteae, except those in Cymbonotus (2–3 species, Australia), are native to southern Africa. The tribe Arctoteae was first recognized by Cassini (1816) who described a total of 14 genera, three of which were later placed in synonymy. In 1832, Lessing reduced the group to the rank of subtribe within the tribe Cardueae (thistles); he recognized 18 genera, including five new ones. Most modern treatments are based on the work of Bentham (Bentham, 1873a, b) who reinstated Arctoteae at the tribal level with three subtribal groups: Euarctoteae (now Arctotinae), Gorterieae (Gorteriinae), and Gundelieae (Gundelinae). Hoffmann (1890) used Bentham’s subtribes, but he interpreted the genus Arctotis in the broad sense, encompassing most of the modern-day genera of the subtribe. Minor movements of problem genera ensued. Bentham (1873b) removed Heterolepis from Arctoteae and placed it in Inuleae. Hoffmann placed Eremothamnus in Senecioneae (1890). Beauverd (1915a) transferred Ursinia from Arctoteae to Anthemideae based on style morphology. Moore (1929) moved Eremothamnus into Inuleae, and Merxmüller (1967) put it back into Senecioneae. Stix (1960) suggested that Platycarpha should be in Mutisieae. Prior to 1970 Eremothamnus had been placed in the tribes Senecioneae, Liabeae, Astereae, Gnaphalieae, and Inuleae, but in 1970 Leins suggested that Eremothamnus was in Arctoteae, close to but not inside Gorteriinae. In 1973, Robinson and Brettell (1973a) took Ursinia out of Arctoteae and put it into its own tribe based on pollen characters. Also in the same publication, Robinson and Brettell put Platycarpha in Cardueae (thistles) and returned Heterolepis to Arctoteae. In that same year, Robinson & Brettell (1973b) moved both Eremothamnus and Gundelia into tribes of their own. In the book Biology and Chemistry of the Compositae (Heywood & al., 1977), each tribe was discussed in one or more chapters, and comments were made concerning the placement of problem taxa. Norlindh (1977) prepared the treatment of Arctoteae; he accepted the three subtribes of Bentham and added the subtribe Eremothamninae while excluding Platycarpha and Ursinia. Dittrich (1977) in his treatment of the thistles (Cardueae) rejected both Platycarpha and Gundelia and suggested that they should be placed in Arctoteae. Merxmüller (1977), in his treatment of Inuleae, agreed that Heterolepis did not belong in that tribe and suggested a placement in Mutisieae, and Jones (1977) included Funk & al. • Evolution of the tribe Arctoteae Hoplophyllum in Vernonieae. Heywood & Humphries (1977) placed Ursinia into Anthemideae but considered the placement dubious. More recently, Karis (1992) published a cladistic study based on morphology that suggested that Hoplophyllum was sister to Eremothamnus, and he had moderate support for the two genera being closely related to Arctoteae. Robinson (1994) used pollen data to support his hypotheses that Gundelia and Eremothamnus belonged in tribes of their own and not in Arctoteae. Bergqvist & al. (1995) used cpDNA restriction site data from 58 genera of the family to examine the position of Eremothamnus and found that it was sister to Arctoteae, but he used only one genus (Gazania) from Arctoteae. The most recent classification of Arctoteae was published by Bremer (1994); he recognized only the two main subtribes by sinking Gundelieae into Gorteriinae. He listed Eremothamnus and Hoplophyllum as belonging to the subfamily Cichorioideae s.l. but “unplaced” as to tribe. Karis & al. (2001) used sequence data from ndhF to suggest that Gundelia did not belong in Arctoteae but was the sister group of Cichorieae (Lactuceae). Based on these data, Gundelia could be included in Lactuceae or left in its own monotypic tribe, Gundelieae. Most recently the Saharan genus Warionia has been shown to be closely related to, and possibly the sister taxon of, Gundelia (Panero & Funk, 2002), and these two together now constitute Gundelieae and are sister to Lactuceae. Ursinia has been placed in Anthemideae based on morphological and molecular data (Heywood & Humphries, 1977; Watson & al., 2000). However, problems remain concerning the placement of several genera: Eremothamnus, Heterolepis, Hoplophyllum, and Platycarpha. In addition, Cuspidia, Didelta, Haplocarpha, and Heterorhachis are unusual taxa morphologically, and a confirmed placement of these genera would help with understanding the evolution of the morphology of the tribe. Heterorhachis and Cuspidia are monotypic genera that have unusual features, Haplocarpha has a variable chromosome number, and Didelta has an unusual condition where its receptacle breaks into parts enclosing the achenes. The outer parts are adnate to one of the outer involucral bracts that become thickened and lignified and new plants may germinate out of these parts (Fig. 1A). Finally, Cymbonotus, the only genus found outside of southern Africa, may be designated “endangered” in its native Australia and so its position in the tribe needs to be confirmed. Table 1 presents an historical summary of taxonomy of Arctoteae. The ultimate goal of our research effort is to produce a detailed morphological study and molecular analysis of Arctoteae and to use the results to investigate the evolution and biogeographical history within the tribe and among tribes of the subfamily Cichorioideae s.s. Obviously, this task will take some time and involve sev639 Funk & al. • Evolution of the tribe Arctoteae 53 (3) • August 2004: 637–655 Fig. 1. Variation in shape and color of heads in the mostly southern African tribe Arctoteae. A, Didelta; B, C, Arctotis; D, Berkheya; E, Gazania. All photos by V. Funk. 640 53 (3) • August 2004: 637–655 eral projects. The goals of this paper are to investigate the monophyly of Arctoteae, determine the sister group(s) and/or most appropriate outgroup(s), determine placement of problem and unusual genera, and establish a sound basis for a subtribal classification of the tribe. MATERIALS AND METHODS Taxon sampling. — The terms “ingroup” and “outgroup” are, of course, relative to one another and change based on the specific question being addressed. For this study we consider the ingroup to include all taxa that were previously thought to be in the tribe Arctoteae, with the exception of Ursinia (now in Anthemideae). These ingroup genera were sampled thoroughly by including several species of each genus so that all “groups” within the genera were sampled. At least two populations were sampled in the monotypic genera. The first set of outgroups for this study included all taxa outside of traditional Arctoteae that are found in the subfamily Cichorioideae s.s. (Appendix). A second set of outgroups from outside the subfamily were added later (Appendix). In general, all outgroup taxa were sampled less thoroughly than the ingroup taxa. Ingroup. Seventeen ingroup genera were used in this study, including all but one of the genera previously placed in Arctoteae (excluding Ursinia: Anthemideae). A total of 76 samples was analyzed and 28 of these ingroup samples (Appendix) were selected to be used in the pruned tree. There are 78 newly reported sequences, 26 ITS, 27 trnL-F, and 25 ndhF, and two taken from GenBank. Outgroups. Sixteen species in 15 outgroup genera were sampled for this study (Appendix). There are 44 newly reported sequences, 14 ITS, 15 trnL-F, and 15 ndhF, one sequence was taken from GenBank and two were sent by J. Panero (pers. comm.). The tribes of Cichorioideae s.s. (Arctoteae, Eremothamneae, Gundelieae, Lactuceae, Liabeae, Moquineae, and Vernonieae) form a monophyletic group with good support (Panero & Funk, 2002); the relationships among these tribes, however, is poorly understood. All of the tribes were considered as possible outgroups for Arctoteae, but only four were used. The Eremothamneae were considered by Norlindh (1977) to be part of Arctoteae, so it was included in the ingroup. Gundelieae (as Gundelia) was also included in Arctoteae by Norlindh (1977) and Bremer (1994), but, based on ndhF sequence data, Karis suggested that Lactuceae and Gundelia were sister taxa (Karis & al., 2001). Our initial analyses supported this, and so the tribe Gundelieae was moved to the outgroup. The Gundelieae were, until recently, believed to be monotypic, but they have now Funk & al. • Evolution of the tribe Arctoteae been shown to contain Warionia as well as Gundelia (Panero & Funk, 2002); both genera were used in this study. Lactuceae, morphologically distinct but poorly known molecularly, was represented by three common genera (Cichorium, Hypochaeris, and Lactuca). The taxa from Liabeae, a modest-sized tribe from the Americas, were selected because they represent the two main lineages of the tribe and one member of the paraphyletic base of the tribe (Kim & al., 2003). Vernonieae, a large and diverse tribe with ca. 2500 species, was included as an outgroup. The taxa from Vernonieae used in the study were selected because they were near the base of the preliminary trees produced in an ongoing molecular study of the tribe (Keeley & Chan, pers. comm.). Previously it has been suggested that Vernonieae and Liabeae might be closely related, in fact Cassini had Liabeae as a subunit of Vernonieae (Cassini, 1828, 1830) and other authors have also placed the two together (e.g., Robinson & Funk, 1987; Bergqvist & al., 1995). Moquineae were considered to be part of Vernonieae and were not included in this study. Once the data were combined, additional outgroups could be added from outside of subfamily Cichorioideae including Barnadesia and Dasyphyllum from subfamily Barnadesioideae, Gerbera (two species) from the tribe Mutisieae s.s., and Carthamus and Cirsium from the tribe Cardueae (thistles). The DNA sequences were obtained from several sources. Ongoing studies by Keeley and Chan (unpubl.) provided sequences for Vernonieae. Panero (unpubl.) provided the trnL-F and ndhF for Warionia. GenBank provided two ITS sequences (Hoplophyllum and Warionia) and one ndhF (Eremothamnus). Although we tried repeatedly to obtain additional sequences from Eremothamnus, we were unsuccessful; these two regions, therefore, are missing for this taxon. It does not appear to have created any problems, however; all ndhF and combined analyses were conducted with and without the genus, and it made no difference in the results. The remaining 125 sequences are newly reported; the Appendix contains a list of vouchers and the GenBank numbers for all sequences. All plant tissue was collected in the field and stored in silica gel, sampled from herbarium specimens, or freshly collected. DNA amplification, sequencing, and analysis. — DNA extractions were performed using a Qiagen DNeasy Plant Mini Kit following the instructions supplied but with an extended incubation period (up to 40 minutes) for herbarium material. Primer ITS5A (Downie & Katz-Downie, 1996), based on White & al.’s (1990) fungal primer ITS5 and corrected at two positions for angiosperms, was substituted for ITS5 in this study. All primer sequences are given in Table 2. Primers used to amplify and sequence the trnL-F and downstream trnL-F spacer region of cpDNA were designed by Taberlet & al. 641 Funk & al. • Evolution of the tribe Arctoteae (1991), and those used for the 3' end of the ndhF region were designed by Jansen (1992). For the PCR amplification reactions, each 25 µl PCR reaction cocktail contained 12.9 µl of sterile water, 2.5 µl of 10x PCR reaction buffer A (Promega), 2 µl of 20 mM dNTPs (Pharmacia) in an equimolar ratio, 2.5 µl of 25 mM magnesium chloride, 0.5 µl of 10 mg/µl Bovine Serum Albumin (Sigma), 1 µl of a 10 µM concentration each of the forward and reverse primer, 0.1 µl of Taq DNA polymerase enzyme (5 units/µl from Promega), and 2.5 µl of sample DNA. The amount of template DNA was adjusted when necessary to generate sufficient PCR products for DNA sequencing. The amplification reactions were conducted using thin-walled 0.2 ml PCR reaction tubes in a GeneAmp PCR System 9700 (Perkin Elmer). The PCR program consisted of an initial preheating at 94º C for 2 minutes. Then, the first reaction cycle proceeded as follows: 1 minute at 94º C to denature the template DNA, followed by 1 minute at 48º C (54º C for cpDNA) to allow primer annealing and 45 seconds at 72º C for primer extension. Primer extension time was increased by 4 seconds (7 seconds for cpDNA) for each subsequent reaction cycle. After a total of 40 reaction cycles, an additional 7 minute extension at 72º C was allowed for completion of unfinished DNA strands. All PCR products were quantified by agarose gel electrophoresis with comparison of an aliquot of products with a known quantity of a 100 bp DNA ladder (GeneChoice; visualized with ethidium bromide). The remainder was stored at 4º C until utilized. PCR products used for sequencing were first purified for sequencing using an enzymatic PCR product presequencing kit (USB). This procedure involved mixing 8 µl of the PCR product with 1 µl of each enzyme from the kit and then incubating the mixture first at 37º C for 30 min, to degrade excess primers and dNTPs, and then raising the temperature to 80º C for 15 minutes, to denature the enzymes themselves. This method of purification without loss of PCR products (no filtration, precipitation, or washes are necessary) is especially important for DNA extracted from herbarium vouchers, which is sometimes only weakly amplified and yields barely sufficient PCR product for sequencing. Table 2. Primer sequences used for PCR and cycle sequencing. Name Sequence (5' to 3') ITS5A ITS4 trnL C trnL F ndhF 1603 ndhF +607 GGA AGG AGA AGT CGT AAC AAG G TCC TCC GCT TAT TGA TAT GC CGA AAT CGG TAG ACG CTA CG ATT TGA ACT GGT GAC ACG AG CCT YAT GAA TCG GAC AAT ACT ATG C ACC AAG TTC AAT GYT AGC GAG ATT AGT C 642 53 (3) • August 2004: 637–655 The cycle sequencing reactions were done using 96 well microplates in a PTC-100 thermal cycler (MJ Research). Each one-eighth cycle sequencing reaction cocktail contains 50–150 ng of the purified PCR product, 2 µl of a 1 mM concentration of the sequencing primer, 0.6 µl of a 5x reaction buffer (400 mM Tris HCl, 10 mM magnesium chloride at pH 9.0), and 1 µl of the reagent pre-mix from the BigDye (Version 2/3) dye terminator cycle sequencing pre-mix kit (Applied Biosystems). The cycle sequencing program consisted of an initial preheating at 96º C for 30 seconds. Then, the first reaction cycle proceeded as follows: 10 seconds at 92º C to denature the template DNA, followed by 15 seconds at 55º C to allow primer annealing and 4 minutes at 60º C for primer extension. Unincorporated dye terminators were removed by Sephadex (Sigma) gel filtration using MultiScreen plates (Millipore). The purified cycle sequencing products were then resolved by electrophoresis on a 5% polyacrylamide (MJ Research Kilobasepack) gel using a BaseStation 51 automated DNA sequencer (MJ Research). Sequences from both strands of each PCR product were examined, compared, and corrected using Sequence Navigator software (Applied Biosystems). All trnL-F, ndhF, and ITS sequences were aligned visually, with the insertion of gaps where necessary. There were five unambiguous insertions and deletions, mainly in the trnL-F data set. Three deletions and one insertion supported all or part of the Gazania-GorteriaHirpicium clade. The remaining deletion (ndhF) was found in all of the ingroup and first set of outgroups (Cichorioideae s.s.) but not in the second set of outgroup taxa. The taxon groups that these indels supported occurred in the phylogenies without coding the indels (see Results), so it did not seem necessary to separately code the indels. Maximum parsimony analysis and parsimony bootstrap analysis (with 1000 replicate runs, each with 10 random taxon additions, TBR branch swapping, and MULPARS in effect) of the aligned trnL-F, ndhF, and ITS sequences were performed (with and without the outgroups) for each region and for the cpDNA data and the combined data sets via full heuristic searches with PAUP* (Swofford, 2002). No weighting was used. Maximum parsimony analysis using a branch-and-bound search was also performed with a reduced dataset of 17 ingroup and three outgroup taxa selected from previous results. The bootstrap runs employed 1000 replicates with branch-and-bound searches. The likelihood ratio tests were done using the “Tree Scores” function in PAUP* under the likelihood criterion, the HasegawaKishino-Yano model of sequence evolution (Hasegawa & al., 1985), and a gamma distribution of rate variation among sites (with the shape parameter estimated and with four rate categories). The likelihood and the boot- 53 (3) • August 2004: 637–655 Funk & al. • Evolution of the tribe Arctoteae ARCT strap analyses were used on each molecular region separately, the two chloroplast regions combined, and all three regions combined. Arctotis 14 Arctotis 08 Cymbonotus 45 Arctotheca 95 Haplocarpha 77 Dymondia 53 RESULTS A number of full heuristic searches were run using various outgroups and adding and subtracting members of the ingroup. These initial runs contained all taxa and all outgroups. These runs produced thousands of equally parsimonious trees, but they consistently showed several things. The core elements of two subtribes formed strongly supported clades, Arctotineae (ARCT) and Gorteriinae (GORT), and within Gorteriinae there are three groups, the Didelta-Berkheya clade (Did), Berkheya-Cullumia clade (Ber), and the GazaniaGorteria-Hirpicium clade (Gaz). Berkheya appears to be paraphyletic and Hirpicium and Haplocarpha are most likely paraphyletic as well. The 10 outgroup taxa from the first set of outgroups (Appendix 1) consistently divided into two clades, Liabeae and Vernonieae (LIAVER) and Gundelieae and Lactuceae (GUN-LAC). However, the placement of some of the problem taxa, as well as the relationships of the outgroups to the clades within the ingroup and to one another, was unresolved. An examination of many of the equally parsimonious trees showed that most of the alternative trees were the result of closely related taxa changing positions. In order to elucidate relationships among the problem taxa, outgroup taxa, and core subtribes, some taxa were pruned from the dataset. For instance, in the full dataset there were eight samples of Arctotis, 18 of Berkheya, nine of Gazania, and nine of Hirpicium and all of these were not pertinent to the goals of this study. In order to decide which taxa to remove from this portion of the study, we ran the data without outgroups. The resulting unrooted phylograms allowed us to select taxa for removal that had similar positions, and so there was no loss of relationships within or among genera. Taxa were selected so as to keep the basic structure of the unrooted phylogram, e.g., all potentially paraphyletic groups were maintained and members of the core subtribes were arranged in the same manner. Because all of the members of each genus indicated to be paraphyletic were contained within a single subtribe, the fact that they are paraphyletic does not affect the questions being asked in this study. After pruning the database there were 28 ingroup samples. The analyses were then re-run using the pruned data set to confirm to the overall branching structure. Figure 2 shows an unrooted phylogram with labels on the major groups plus abbreviations that will be used in most figures; the remaining problem taxa and outgroups are not included. The basic shape of Fig. 2 was taken from Gaz Gazania 62 Gazania 59 Gazania 64 Hirpicium 86 Gorteria 69 Gorteria 73 Didelta 51 Didelta 50 Did Berkheya 24 Hirpicium 25 Hirpicium 87 Cuspidia 104 Berkheya 18 Berkheya 99 Cullumia 37 Cullumia 39 Berkheya 31 GORT Ber Heterorhachis 84 Fig. 2. An unrooted phylogram of Arctoteae showing the location of the two main subtribes, Arctotinae (ARCT), Gorteriinae (GORT), and the three clades of the latter: Gazania-Gorteria-Hirpicium (Gaz), Berkheya-Cullumia (Ber), and Didelta (Did). No outgroups or problem taxa were included. The basic shape was taken from the results of the combined analysis. results of the combined analysis. In all analyses conducted for this study, some of the taxa in which we were interested are clearly placed (Fig. 2; no matter what outgroup is used to root this diagram it always attaches somewhere along the long branch between the two subtribes, thus allowing use of the monophyletic subtribes in an unrooted diagram). Heterorhachis was consistently part of the BerkheyaCullumia clade (Ber) of the subtribe Gorteriinae, Haplocarpha was at the base of the subtribe Arctotineae (ARCT), and Cymbonotus, the Australian genus, was always part of Arctotineae. In addition, in the analyses containing outgroups, Gundelia was always the sister group of Lactuceae and was, therefore, maintained as part of the outgroup for the rest of the study. The morphologically interesting taxa, Cuspidia and Heterorhachis, are always found in the BerkheyaCullumia clade and Didelta is in the Didelta-Berkheya clade; both clades are in the subtribe Gorteriinae. Because they were stable in their placements, they are considered part of one of the main subtribes and their location is not discussed further. However, the positions of the three other taxa of interest, the tribe Eremothamneae (Eremothamnus and Hoplophyllum), and the genera Heterolepis and Platycarpha, were not clear nor was their relationship to the two core subtribes or to the outgroups consistent. The position of these three problem taxa was, therefore, a concern for the remainder 643 Funk & al. • Evolution of the tribe Arctoteae 53 (3) • August 2004: 637–655 of the analysis. trnL-F. — Data were available for all taxa except Eremothamnus (Appendix), resulting in the use of 27 ingroup taxa and 16 outgroup taxa. In an effort to determine the basic ingroup patterns, the pruned dataset was analyzed without outgroups and excluding the three remaining problem taxa (Eremothamneae, Heterolepis, and Platycarpha). This resulted in an analysis with 35 “parsimony informative” characters (pi) which produced 72 trees (L = 84, CI = 0.893, RI = 0.936). The three problem taxa were then added to the same analysis and one of the unrooted phylograms is shown in Fig. 3 (547 trees, L = 109, pi = 38, CI = 0.881, RI = 0.914). All three problem taxa were on long branches and very nearly in the same location. When one of the outgroups, the Liabeae-Vernonieae clade (LIA-VER), was added to the analysis, it formed a ARCT 1 change Pla Het 2 1 GUN-LAC Ere * Did GORT Ber Gaz LIA-VER LIAVER Ber Gaz 66 Het 100 Ere ARCT 100 61 53 GUNLAC 57 59 Pla 75 GORT Did Fig. 3. Unrooted phylogram from trnL-F data analyzed with problem taxa but without outgroups. The arrows indicate placement of the two outgroup clades, Gundelieae-Lactuceae (GUN-LAC) and LiabeaeVernonieae (LIA-VER) when they are added to the analysis singularly. Note that the placement of Platycarpha changes from position “1” when the outgroup is LIA-VER to position “2” when the outgroup is GUN-LAC. The * indicates a branch that collapsed in the unrooted strict consensus tree (547 trees, L = 109, pi = 38, CI = 0.881, RI = 0.914). The lower diagram is a bootstrap consensus tree reduced to show the main groups. 644 polytomy with Eremothamneae, Platycarpha and Heterolepis. When the other outgroup clade, the Gundelieae-Lactuceae clade (GUN-LAC), was added it also attached at the same node, however Platycarpha changed its position and became the sister taxon of Heterolepis. When any single problem taxon or outgroup clade was added to the analysis, it attached at or near the same location; when the taxa were combined, it resulted in hundreds of trees with varying placement of the taxa in question. The maximum likelihood tree for the trnL-F data (score of single best tree found = 2612.57292) supported monophyly of the two main subtribes and of the two outgroup clades, but the position of the two outgroup clades and the problem taxa remained unresolved, and Arctoteae were not indicated to be monophyletic. The bootstrap consensus tree, shown in a reduced form at the bottom of Fig. 3, is for an analysis with all problem taxa and both outgroup clades. Note the weak support for the placement of the problem taxa and outgroups. ndhF. — There were no ndhF data for Heterorhachis and for one of the Gorteria species (Gorteria 69) so they were not included in these analyses. There were ndhF data available for Eremothamnus, so the taxon was added for a total of 26 ingroup samples; all 10 outgroup samples were included (Appendix). When examined without outgroups or problem taxa, the ndhF data set yielded 996 trees (pi = 42, L = 95, CI = 0.821, RI = 0.906). It had the same basic structure as the unrooted phylogram from the trnL-F sequence data. When all three problem taxa were analyzed together (6029 trees, L = 122, pi = 47, CI = 0.803, RI = 0.879) the resulting placement of the taxa put Platycarpha closer to Gorteriinae, Eremothamneae closer to Arctotineae and Heterolepis between the two (Fig. 4). The two outgroup clades were added separately and their placement is indicated by the arrows on Fig. 4. The LiabeaeVernonieae clade is placed closer to Platycarpha and Gorteriinae, and the Gundelieae-Lactuceae clade is closer to Arctotineae; Heterolepis is between the two. The maximum likelihood tree for ndhF (score of single best tree found = 2447.8924) supported the monophyly of the two main subtribes and of the two outgroup clades, but the position of the outgroups and the problem taxa remained unresolved. The Arctoteae could be monophyletic if they included Eremothamneae but not Platycarpha or Heterolepis. The bootstrap consensus tree, shown at the bottom of Fig. 4, is for an analysis with all problem taxa and both outgroup clades; it shows weak support for the placement of Eremothamneae, and it has an unresolved cluster at the center of the diagram. Combined chloroplast dataset. — Neither of the chloroplast datasets had sufficient information to fully resolve the trees. However, the two datasets can be combined to produce a larger number of characters that 53 (3) • August 2004: 637–655 Funk & al. • Evolution of the tribe Arctoteae ARCT 1 change * GUN-LAC Ere Het LIA-VER Pla * * * Did Ber Gaz GORT Ere Ber Gaz 100 Het 73 Pla ARCT 100 64 60 GUNLAC LIAVER 75 GORT Eremothamneae was sister to Arctotineae. In either explanation, the tribe Arctoteae could not be made monophyletic. However, when the data were analyzed with outgroups but without the three problem taxa, the tribe Arctoteae was monophyletic (36 trees, L = 379, pi = 128, CI = 0.815, RI = 0.862). The bootstrap consensus diagram for the chloroplast data with all problem taxa and the two outgroups is shown at the base of Fig. 5. ITS. — Nuclear sequence data were gathered for all ingroup taxa except Eremothamnus (28 taxa) and all 10 outgroup taxa. The initial analysis, without outgroups or problem taxa, produced four trees (4 trees, L = 546, pi = 179, CI = 0.663, RI = 0.809), and in all four the three problem taxa attach, when analyzed separately, between the two subtribes as in the chloroplast sequences. Figure 7 illustrates one of the three unrooted phylograms produced when all three problem taxa are included without the outgroup clades (3 trees, L = 729, pi = 196, CI = 0.595, RI = 0.741). Note that while Platycarpha remains between the two subtribes, Heterolepis and Did Fig. 4. The upper diagram is an unrooted phylogram from ndhF data with problem taxa and without outgroups. The arrows indicate placement of the outgroups when they are added to the analysis singularly. The * indicates branches that collapse in the unrooted strict consensus tree (6029 trees, L = 122, pi = 47, CI = 0.803, RI = 0.879). The lower diagram is a bootstrap consensus tree reduced to show the main groups. GUN-LAC Ere ARCT Het LIA-VER Pla Ber Did GORT Gaz Pla LIAVER Ber Gaz 100 Ere 83 ARCT 100 90 50 GORT 78 80 may shed light on the unresolved and conflicting areas of the trees as well as the outgroup placement. As with both trnL-F and ndhF data, all three problem taxa attached between the two subtribes (Figs. 3, 4). When the outgroup taxa were added separately, the LiabeaeVernonieae clade connected at the base of the Platycarpha and Heterolepis clade, and the GundelieaeLactuceae grouped with Eremothamneae becoming the sistergroup of Arctotinae. Overall, the GundelieaeLactuceae clade was closer to Eremothamneae and Arctotinae and Liabeae-Vernonieae to Heterolepis and Gorteriinae. When both outgroup clades were added to the analysis at the same time, two very different basal topologies resulted depending on which outgroup was used to root the tree (Figs. 5, 6; 108 trees, L = 429, pi = 135, CI = .804, RI = .844). When the tree was rooted on Gundelieae-Lactuceae, the Liabeae-Vernonieae clade was nested in the ingroup, sister to Gorteriinae along with Heterolepis and Platycarpha (Fig. 5). When the Liabeae-Vernonieae clade was used as the root, Heterolepis formed a trichotomy at the base of the tree and Gundelieae-Lactuceae was sister to the ArctotineaeEremothamneae clade (Fig. 6). In both figures, Platycarpha was the sister group of Gorteriinae and Het GUNLAC Did Fig. 5. The upper diagram is a strict consensus tree from the chloroplast data (trnL-F and ndhF) with both outgroup clades and all three problem taxa rooted on the Gundelieae-Lactuceae clade (108 trees, L = 429, pi = 135, CI = 0.804, RI = 0.844). The lower diagram is a bootstrap consensus tree reduced to show the main groups. 645 Funk & al. • Evolution of the tribe Arctoteae 53 (3) • August 2004: 637–655 LIA-VER Het GUN-LAC Ere ARCT Pla Ber Did GORT 8009.8773) supported the monophyly of the two main subtribes and of the two outgroup clades, but the position of the outgroups remained problematic. Platycarpha is sister to Liabeae-Vernonieae, and a weak clade of Heterolepis and Hoplophyllum is sister to Gorteriinae. Arctoteae are not monophyletic, with GundelieaeLactuceae closer to Arctotinae and Liabeae-Vernonieae closer to Gorteriinae, depending on the rooting. Combined data analysis of all sequences. — All 28 ingroup taxa and 10 outgroup taxa were used for the combined data analysis (trnL-F, ndhF, ITS). Figure 2 is an unrooted phylogram for the combined dataset (without problem taxa and outgroups), and the two main subtribes and the three clades within Gorteriinae are distinct. When the problem taxa were added (2 trees, L = 976, pi = 281, CI = 0.644, RI = 0.767), they were spread out along the branch that separates the two subtribes. The position of each outgroup clade was examined individually and the Liabeae-Vernonieae clade was Gaz ARCT 646 Pla GUN-LAC LIA-VER Het Gaz Ere Did * * Ber GORT LIAVER Ber 0 99 Gaz 100 94 Ere Het GUNLAC GORT 100 Pla ARCT 78 Eremothamneae are sister to one another and are nested within Gorteriinae. The position of the outgroups (if added separately) is indicated with arrows (Fig. 7). The rooted trees that were run with both outgroups included at the same time showed the outcome of rooting with one outgroup or the other (Figs. 8, 9; 2 trees, L = 1443, pi = 314, CI = 0.463, RI = 0.628). In Fig. 8, when the tree was rooted with the Liabeae-Vernonieae clade, Platycarpha was basal followed by monophyletic Arctotineae, while the Gundelieae-Lactuceae clade was between Arctotineae and Gorteriinae. When the same unrooted phylogram was rooted on the GundelieaeLactuceae clade (Fig. 9), Platycarpha became sister to Liabeae-Vernonieae and the two together were then sister to Arctotineae. The Gorteriinae clade (including the Eremothamneae and Heterolepis) was the same with either rooting of the tree. Note that in the analysis of the nuclear DNA data, Platycarpha was most closely related to either the Liabeae-Vernonieae clade or Arctotineae, while in the chloroplast DNA it was closer to Gorteriinae. The results of the bootstrap consensus analysis involving all problem taxa and both outgroup clades is shown at the base of Fig. 7; note the position of the Gazania-Gorteria-Hirpicium clade in the unresolved center of the diagram. The maximum likelihood tree for the ITS data (score of single best tree found = 10 changes 10 Fig. 6. A strict consensus tree from the chloroplast data (trnL-F and ndhF) with both outgroup clades and all three problem taxa rooted on the Liabeae-Vernonieae clade. [108 trees, L = 429, pi = 135, CI = 0.804, RI = 0.844] Did Fig. 7. The upper diagram is an unrooted phylogram from ITS data with problem taxa but without outgroups. The arrows indicate placement of the outgroups when they are added to the analysis singularly. The * indicates branches that collapse in the strict consensus tree (3 trees, L = 729, pi = 196, CI = 0.595, RI = 0.741). The lower diagram is a bootstrap consensus tree reduced to show the main groups. 53 (3) • August 2004: 637–655 Funk & al. • Evolution of the tribe Arctoteae LIA-VER GUN-LAC Pla Pla LIA-VER ARCT ARCT GUN-LAC Gaz Gaz Het Ere Het Ere Did Did Ber Ber 10 changes 10 changes Fig. 8. A phylogram from the ITS data with problem taxa and outgroups, rooted with Liabeae-Vernonieae (LIAVER). The other outgroup (GUN-LAC) is sister to the subtribe Gorteriinae. Fig. 9. A phylogram from ITS data with problem taxa and outgroups, rooted with Gundelieae-Lactuceae. The other outgroup (LIA-VER), along with Platycarpha, is sister to the subtribe Arctotinae. always the sister group of Platycarpha; however the Gundelieae-Lactuceae clade attached in five different places. When the two outgroups were analyzed together, a four-branched polytomy resulted with the two outgroup clades, the subtribe Arctotineae, and on a short branch the remainder of the ingroup (Fig. 10; 8 trees, L = 1850, pi = 441, CI = 0.530, RI = 0.665). The bootstrap consensus analysis of the same dataset produced the diagram shown at the bottom of Fig. 10. This analysis was, as usual, less resolved than the parsimony analysis so it lacks the branch between the Ere-Het-GORT clade and the rest of the taxa. Since the position of the LiabeaeVernonieae clade was unambiguous in the parsimony analysis, one could argue that it would make a “better” outgroup and root the tree using that clade. In such an analysis, Arctoteae (including Heterolepis and Eremothamneae but excluding Platycarpha) are monophyletic. Another alternative, and the one used here, was to gather sequence data from additional taxa that were outside the subfamily Cichorioideae. Prior attempts to use these taxa in this study had failed because with trnL-F and ndhF data there were too few characters to resolve the tree, and with the ITS some of the new outgroups were so different that the results were ambiguous. However, with the combined dataset there was a sufficiently large number of “parsimony informative” characters, shorter branches, and more reliable results making the inclusion of these outside taxa possible. The additional outgroups (Appendix 1) included two taxa from the subfamily Barnadesioideae (Barnadesia and Dasyphyllum), the basal branch in the family, along with two species of African Mutisieae (Gerbera), and two species of Cardueae (thistles; Carthamus and Cirsium). Data from ndhF and trnL-F were collected for all six taxa and ITS data from five (minus Dasyphyllum) for a total of 17 additional sequences. All of these taxa are located relatively close to the base of the family. The trnL-F and the ndhF of these six additional taxa aligned relatively easily, but the ITS sequences were aligned with less confidence. Fortunately, the ambiguous areas were confined to two places in the data matrix, and the data could be analyzed with and without these two areas. Figure 11 shows one of two trees produced by the combined dataset (trnL-F, ndhF and ITS) of all taxa 647 Funk & al. • Evolution of the tribe Arctoteae 53 (3) • August 2004: 637–655 Hesperomannia 1 LIA-VER 10 changes Gymanthemum K11 Philoglossa 120 Munnozia K39 Cymbonotus 45 Oligactis 121 ARCT Arctotis 14 Arctotis 08 Haplocarpha 77 Arctotheca 95 Dymondia 53 Ere Pla Didelta 51 Did Didelta 50 Berkheya 24 * * Gundelia 96 Cichorium 136 Gaz Het GUN-LAC Warionia Berkheya 18 Cullumia 37 Berkheya 99 Cullumia 39 Heterorhachis 84 Berkheya 31 Ber Cuspidia 104 Hirpicium 87 Hirpicium 25 Gorteria 73 Gazania 64 Gorteria 69 Gazania 62 Gazania 59 Hirpicium 86 GORT Hypochaeris 33 Lactuca 133 Fig. 10. The upper diagram is an unrooted phylogram from the combined data set (trnL-F, ndhF, ITS), with problem taxa and outgroup clades. The * indicates branches that collapse in the strict consensus tree (8 trees, L = 1850, pi = 441, CI = 0.530, RI = 0.665). The lower diagram is a bootstrap consensus tree reduced to show the main groups. including the second set of outgroups (L = 2345, pi = 516, CI = 498, RI = 0.638). The two trees differed only in a minor change in the position of one species of Cullumia. The Gundelieae-Lactuceae clade is more distant from Arctoteae than the Liabeae-Vernonieae clade. Platycarpha is basal to the Liabeae-Vernonieae clade and cannot be considered part of Arctoteae. Within monophyletic Arctoteae there are several clades: the subtribe Arctotineae, the genus Heterolepis, the tribe Eremothamneae, and the subtribe Gorteriinae. Furthermore, Gorteriinae can be broken up into three clades: the Gazania-Gorteria-Hirpicium clade, the Didelta-Berkheya clade and the Berkheya-Cullumia clade. The same result was also obtained using just Mutisieae and Cardueae as outgroups or Cardueae or Mutisieae alone. The bootstrap values are listed on the branches; note that although there is good support for nearly all the branches (all but two critical nodes have support of 72% or higher), the support for Arctoteae is less than 50%. Likewise there is little support for the grouping of Eremothamneae and Heterolepis or for these two taxa grouping with the subtribe Gorteriinae. If the 648 data are analyzed without the three problem taxa, the support for monophyly of Arctoteae is 72%. The maximum likelihood tree of the ingroup rooted on Platycarpha has monophyletic Arctotineae and monophyletic Gorteriinae with Eremothamneae and Heterolepis between the two subtribes (score 8781.5000). The maximum likelihood tree of the ingroup with the Liabeae-Vernonieae clade as the outgroup gave the same tree except that, because the tree was rooted with the Liabeae-Vernonieae clade, Platycarpha was basal to this clade; Fig. 11 (score 11513.36510). Table 3 gives the size of the aligned matrices, the number of gaps added in order to align the sequences, the number of informative characters, and the percent of sequence divergence (calculated using PAUP 4.0*) for all three regions. When applicable, the figures were calculated for both the ingroup alone and for the ingroup plus the outgroup. This table shows that the percent sequence divergence is low in the cpDNA (2.9–4.5% for trnL-F and 5.6–7.7% for ndhF) and higher in the ITS sequences (19.9–28.3%). However, none of the divergence percentages were alarmingly high. Of more con- 53 (3) • August 2004: 637–655 Funk & al. • Evolution of the tribe Arctoteae Dasyphyllum 465 BAR Barnadesia K79 Gerbera 75 100 MUT Gerbera 134 Carthamus 117 CAR Cirsium 118 100 Gundelia 96 GUN Warionia 100 Cichorium 136 93 100 Hypochaeris 33 LAC 87 Lactuca 133 Platycarpha 91 72 Gymanthemum K11 100 91 VER Hesperomannia 1 92 Oligactis 121 100 Munnozia K39 LIA 86 Philoglossa 120 Haplocarpha 77 72 100 Dymondia 53 67 Cymbonotus 45 ARCT Arctotheca 95 54 Arctotis 08 100 Arctotis 14 Heterolepis 82 96 Hoplophyllum 109 Eremothamnus GB Hirpicium 87 Gaz Gazania 64 100 99 56 Gazania 59 Gazania 62 99 Hirpicium 25 57 Hirpicium 86 ARC 95 90 93 Gorteria 69 Gorteria 73 Berkheya 24 100 Didelta 50 Did GORT 66 Didelta 51 Cuspidia 104 100 Berkheya 18 Berkheya 31 Heterorhachis 84 66 Berkheya 99 58 Cullumia 37 Ber 79 Cullumia 39 10 changes Fig. 11. Phylogram with all taxa from the combined data set (trnL-F, ndhF, ITS). The tribe Arctoteae is monophyletic, the subtribes Arctotineae and Gorteriinae are monophyletic; two problem taxa, the genus Heterolepis and the tribe Eremothamneae, are weakly grouped together and are sister to Gorteriinae. The closest outgroup is the LiabeaeVernonieae clade and the problem genus Platycarpha. The bootstrap values are listed on the branches; note that although there is good support for nearly all the branches, the support for Arctoteae is less than 50%. cern is the number of informative characters, which reaches 50% for ITS data in the combined analysis (with all outgroups). The cpDNA scores reach as high as 8.6% for the trnL-F and as high as 15% for the ndhF (combined analysis using all outgroups). The alignment problems in the ITS were confined to two areas, positions 39–165 (ITS 1) and positions 457–484 (ITS 2). As an experiment, these two areas were replaced with question marks for all of the outgroups and the three problem taxa. All of the analyses were rerun, including the addition and subtraction of all outgroups and problem taxa. The only real difference between the final figure (combined analyses) for this altered dataset and Fig. 11 is a slight change in the bootstrap values for some nodes and the fact that 649 Funk & al. • Evolution of the tribe Arctoteae 53 (3) • August 2004: 637–655 Table 3. Size of aligned matrices, number of gaps added, number of informative characters, and percent sequence divergence for trnL-F, ndhF, and ITS. Figures usually given for both “only the ingroup” and for “ingroup plus all of the outgroups”; single outliers have been removed. Regions trnL-F ndhF ITS Total Size of aligned matrices 0897 0699 0679 2275 # of gaps added (ingroup/outgroup) 47–49 /50–65 22–26/27–42 39–42/34–64 ca. 113/ ca. 141 Heterolepis moves to a position as the sister taxon to subtribe Arctotineae. Since the position of Heterolepis in Fig. 11 or in the new analyses is not supported by either high bootstrap values or the maximum likelihood analysis, this shifting of positions is not surprising. DISCUSSION Molecular data. — The final phylogeny (Fig. 11) supports three out of four of the previously described subtribes of Arctoteae: Arctotinae, Gorteriinae, and Eremothamninae; the remaining subtribe, Gundeliinae, was previously elevated to tribal level and is here confirmed as the sister group of Lactuceae. It is clear that Haplocarpha and Cymbonotus belong in Arctotineae, and Heterorhachis, Cuspidia, and Didelta in Gorteriinae. However, to achieve monophyly in Arctoteae, Platycarpha must be removed from the tribe. Over the years this genus has been in the tribes Mutisieae and Cardueae as well as Arctoteae, and in the most recent classification it was “unassigned to tribe” (Bremer, 1994). Of any existing tribes it may be the closest to the LiabeaeVernonieae clade. However, final placement of this genus must await the addition of several taxa which may be near the base of the Liabeae-Vernonieae clade. The phylogeny from the combined analysis (Fig. 11) is a major step forward in understanding relationships among taxa within Arctoteae, specifically the composition of the subtribes, and within subfamily Cichorioideae s.s., particularly the sistergroup of Arctoteae, and placement of Eremothamneae. The extensive analyses have also provided insights into affinities of taxa of Arctoteae that are not expressed in the final figure. For instance, when any of the three DNA regions were examined alone or in combination, but without any of the problem taxa, the results were similar: the two main subtribes were monophyletic, the two outgroup clades were monophyletic, and either Arctoteae was monophyletic or the relationship among the four clades (two outgroups and two subtribes) was unresolved. The clades always have the same composition, however, and there is little if any conflict; it is just a problem of resolution. However, when the problem taxa were added, changes occurred. In 650 # of informative characters (ingroup/outgroup) 049/77 047/103 196/346 292/526 % sequence divergence (ingroup/outgroup) 0.02965/0.04505 0.05578/0.07716 0.19897/0.28353 n/a results of the analysis using chloroplast data, Platycarpha was the sister group of Gorteriinae, Eremothamneae was the sister group of Arctotinae and Heterolepis was near the Liabeae-Vernonieae clade, but its exact position was unresolved (Figs. 5, 6). In ITS results, Platycarpha was closer to the LiabeaeVernonieae clade, which in turn was closer to Arctotineae while Heterolepis and Eremothamneae were inside Gorteriinae (Figs. 8, 9). This presents a nearly complete reversal of their previous positions. The combined analysis produced results that were somewhere in the middle, with Platycarpha sister to Liabeae-Vernonieae (as in the ITS tree), Arctoteae and its two main subtribes (Arctotinae and Gorteriinae) monophyletic; Heterolepis and Eremothamneae were between the two subtribes and the sister group of Gorteriinae. It is not unusual for there to be some discordance of chloroplast and nuclear DNA (e.g., Yoo & al., 2002), and it is usually attributed to the possibility of hybridization. Following that reasoning, the problem taxa could have such a history. Chromosome numbers might be useful in searching for hybrids but they are not available for most taxa in this analysis. For seed plants it has been shown that inadequate taxon sampling can be responsible for conflicting results (Rydin & Källersjö, 2002), but that is most likely not the problem with this study because the taxa removed from the analysis were very similar in their molecular characters to ones that were retained in the study. For now it seems best to keep Heterolepis and Eremothamneae in Arctoteae but out of either of the main subtribes. Taxa within each of the two subtribes are closely related, as evidenced by the short branch lengths that are usually found within them. However, the problem taxa and outgroups have much longer branches and a few of the branches have weak or no support from bootstrap analysis (Fig. 11). It is possible that there has been a relatively recent radiation within the two subtribes and that the outgroups and problem taxa represent old lineages that have been separated for a long time. Morphology. — Based on the literature (Cassini, 1816, 1828; Robinson & Brettell, 1973a; Norlindh, 1977; Robinson, 1992, 1994; Bremer, 1994; Herman & al., 2000) and personal observation, the morphology can be examined in light of the phylogeny (Fig. 11). 53 (3) • August 2004: 637–655 Funk & al. • Evolution of the tribe Arctoteae Fig. 12. Photographs of several heads and habits of Arctoteae. A, Didelta; B, Arctotis; C, Cullumia; D, Hirpicium; E, Gazania; F, Berkheya; G, Didelta; H, Arctotis; I, Gorteria; J, Berkheya. All photos by V. Funk. Photographs of several members of the tribe are found in Figs. 1 and 12. The members of the subtribe Arctotinae (Arctotis, Arctotheca, Cymbonotus, Dymondia and Haplocarpha; Figs. 1B, C, 12B, H) can be characterized as follows: they do not produce latex; their involucral bracts are not fused (one species of Arctotheca and two of Haplocarpha have a tendency toward fusion), the outermost bracts are foliaceous and have a scarious apical lamina and the receptacle is smooth or shallowly areolate-alveolate; the ligulate florets have 4 veins and 3lobes when lobes are present (similar to the tribe Liabeae and the subfamily Asteroideae), and they are fertile (except for Arctotheca); the central florets have shallow651 Funk & al. • Evolution of the tribe Arctoteae ly lobed corollas, the anthers are not tailed (except for Arctotis the largest genus of the subtribe), and the style has a swollen portion below the branch point sometimes with a ring of hairs; the achenes have 3–5 dorsal, welldeveloped ribs or wings. Many of the members of this subtribe are rosette-forming, perennial herbs (Fig. 12H). Most of these characters would be considered plesiomorphic except for the swollen portion of the style with the ring of sweeping hairs (also found in some thistles) and the 3–5 well-developed ribs or wings on the achenes. The lack of latex is most likely apomorphic, but it comes and goes so often in the subfamily that it is difficult to use at this level. The members of the subtribe Gorteriinae (Berkheya, Cullumia, Cuspidia, Didelta, Gazania, Gorteria, Heterorhachis, and Hirpicium; Figs. 1A, D, E, 12A, C–G, I, J) are better defined as follows: they have laticifers (the cells that normally contain latex), and latex is present in Gazania, some Berkheya and possibly others; involucral bracts are connate at least at the base (Fig. 12E), or in various degrees upwards, forming a cup; receptacles are more or less deeply alveolate and enclosing at least the base of the achenes, the apex of the ray florets often have 5 veins and 4 lobes when lobes are present, but it can vary from 2–5 lobes in some taxa; the rays are sterile, the central florets are deeply lobed (Figs. 1A and D), the anthers are sagittate but not tailed, and the previously mentioned conspicuous swollen portion of styles is either not present or not well-developed; the achenes are without well-developed ribs or wings. Figure 11 shows the three clades of this subtribe: the GazaniaHirpicium-Gorteria clade comprises annual or perennial herbs with strongly fused involucral bracts (Fig. 12E), the Berkheya-Cullumia clade comprises a mixture of subshrubs, shrubs, and annual and perennial herbs that are usually spiny in some manner (Figs. 1D, 12C, F, J), and the Didelta-Berkheya clade consists of subshrubs or shrubs with the Didelta members having heads that break apart (Figs. 1A, 12A, G). Members of both the BerkheyaCullumia and the Didelta-Berkheya clades have partially fused involucral bracts. The apomorphic characters for this subtribe include the connate involucres, deeply alveolate receptacles, and 4-lobed and 5-veined sterile rays. All three problem taxa have sets of characters that make it difficult to place them in either of the subtribes. The genus Platycarpha (three species) is a perennial herb confined to southern Africa. The heads are discoid and have only a few florets, which are purple, and these small heads are aggregated into a secondary head that is large and sessile, lying mostly flat on the ground with the leaves spread out like the spokes of a wheel. The plant is either acaulescent, without branches, or with subterranean branches. The receptacle has pales (bracts), the 652 53 (3) • August 2004: 637–655 anthers either have tails (or are sagittate), and the style has a swelling and a ring of sweeping hairs. The achenes are only faintly ribbed and the pappus is made up of scales. While Platycarpha has the swollen style and the ring of sweeping hairs of Arctotineae (and some Cardueae), it does not have the other potential apomorphies of either subtribe, and it has purple flowers and pales which they do not. Heterolepis (three species) is found only in the Western Cape of South Africa. It is a small shrub with involucral bracts somewhat connate in 1–3 rows with the outer ones sometimes herbaceous and the inner with membraneous margins. They have tailed anthers and the heads have fertile rays each with 3 lobes and 4 veins. The styles are slender with short branches, the apical portion either is not or is only slightly thickened, and it has a ring of sweeping hairs. There are no ribs on the achenes and the pappus is made up of slender “bristle-like” scales. Heterolepis has somewhat connate involucral bracts but none of the other apomorphies of Gorteriinae. The margins of its involucral bracts and style are somewhat similar to Arctotineae. The tribe Eremothamneae (Eremothamnus, one species from Namibia; Hoplophyllum, two species from Northwest Cape Province, South Africa) contains shrubs that are radiate or discoid (respectively). When present, the rays are female and fertile with three apical lobes. The anthers have short tails and the styles are long and slender (similar to Vernonieae and some Liabeae), they have unusual bi- and tri-cellular sweeping hairs but not the ring of sweeping hairs found in Arctotinae. There are no distinct ribs on the achenes, and the pappus is capillary. It has none of the apomorphies of either subtribe although its members have spines or thorns on the leaves as do many members of Gorteriinae. It is heartening that all three of the problem taxa (based on molecular data) have morphological characters that have traditionally made it difficult to place them in either of the two main subtribes of Arctoteae. Platycarpha has a swollen area on the shaft of the style, but it also has discoid heads with purple florets, a paleaceous receptacle, and tailed anthers, characters that are unlike either of the subtribes. Heterolepis has tailed anthers, unlike the two subtribes, connate involucral bracts as in Gorteriinae but without the deeply alveolate receptacle, the common type of ray floret as in Arctotinae, and an unusual pappus. Finally, members of Eremothamneae have unique hairs on long, slender styles, and, unlike members of the two subtribes, they have a capillary pappus and tailed anthers, and Hoplophyllum has a discoid head. None of them have the connate involucral bracts with the deeply alveolate receptacle and the 4-lobed sterile rays of Gorteriinae, nor do they have the prominently winged or ribbed achenes. Platycarpha and 53 (3) • August 2004: 637–655 Eremothamneae do not have the scarious-margined bracts of Arctotinae. Given the morphological and molecular data, it seems that we can be certain that the two main subtribes, Arctotinae and Gorteriinae, are monophyletic (as we have defined them) with robust support. The three problem taxa do not belong in either of these two subtribes nor are they closely related to one another. Our data indicate that Eremothamneae and possibly Heterolepis most likely belong in Arctoteae but that Platycarpha does not. Of course, with the inclusion of both Heterolepis and the tribe Eremothamneae (as a subtribe it would be Eremothamninae) the few morphological characters that existed for the tribe Arctoteae have been lost; the scaly pappus, lack of tailed anthers, usually radiate heads, and typical styles are not found in one or both of these groups. A separate morphological analysis is underway (Karis & Funk, pers. comm.) that will hopefully shed light on this problem. The weak support for the tribe is supported by repeated bootstrap analyses, with and without problem taxa, which have failed to find strong support for monophyletic Arctoteae. For now, our conclusion must be that there are little data (either from molecules or morphology) to support the monophyly of Arctoteae, and molecular work currently underway in both Lactuceae and Vernonieae may help in making this determination. There are several important lessons learned in this study. First, the problem taxa are of critical importance. It is clear that in the subfamily Cichorioideae s.s. unusual and morphologically hard to place taxa must be included in any molecular analysis. For instance, in Fig. 2 (all problem taxa and all outgroups removed), the resulting tree shows that the two subtribes could easily be monophyletic, and if only one of the two outgroup clades were used it would present a perfect picture of a monophyletic Arctoteae. If any of the problem taxa (Heterolepis, Eremothamneae, or Platycarpha) or any of the outgroups are added, they attach between the two subtribes, and when used one at a time or in certain combinations of two, they produce a clear result. Combinations of two or more problem taxa in conjunction with both main outgroups generates problems with monophyly of the tribe. All previous studies involving a few taxa from the core subtribes did not reveal the problems in the tribe. Second, multiple outgroups must be evaluated in an exhaustive process to determine how the outgroups are behaving and what effect they are having on the ingroup. This is probably one of the most underrated steps in cladistic analyses, and it has great impact on the outcome. One or two token representatives will not suffice unless a thorough analysis has been done previous to the study in question. Finally, unrooted phylograms and unrooted strict consensus trees give very use- Funk & al. • Evolution of the tribe Arctoteae ful results for analyzing data by freeing the investigator from preconceived notions about relationship. ACKNOWLEDGEMENTS We thank the many colleagues who have sent us material and provided assistance in the field, particularly Marinda Koekemoer (PRE) and Terry Trinder-Smith (BOL), and the various herbaria who have either provided material or loaned specimens (F, K, MO, NY, PRE, US). We also thank MJ Research, especially Mark Norman and Jonathan Schimmel, for allowing us to use their facilities, Jose Panero for sending two unpublished sequences, and K. Peyton, K. Redden, and C. Kelloff for technical assistance. We send a big thank you to Tom Hollowell for his help with the graphics. We especially appreciate the funding that was provided by the National Science Foundation (DEB-0075095 to SK), the office of the Dean of the College of Natural Sciences, University of Hawaii (to SK), and the Mellon Foundation and Scholarly Studies programs of the Smithsonian Institution’s Office of Fellowships and Grants (to VAF). We appreciate the help of Linda Watson who gave us advice on the analysis and Per Ola Karis and an anonymous reviewer who provided comments on the manuscript and, of course, the editors for their helpful suggestions. LITERATURE CITED Baldwin, B. G., Wessa, B. L. & Panero, J. L. 2002. Nuclear rDNA evidence for major lineages of Helenioid Heliantheae (Compositae). Syst. Bot. 27: 161–198. Beauverd, G. 1915a. Contributions à l’étude des Composées X. A. La place des Ursinia parmi les ComposéesTubuliflores. B. Les representants australiens des Arctotidées et leurs rapports avec les Arctotidenées africaines. Bull. Soc. Bot. Genève (2 sér.) 7: 21–40. Beauverd, G. 1915b. Conspectus generum specierumque Arctotidinearum. Bull. Soc. Bot. Genève (2 sér.) 7: 41–56. Bentham, G. 1873a. Notes on the classification, history and geographical distribution of Compositae. J. Linn. Soc., Bot. 13: 335–577. Bentham, G. 1873b. Compositae. Pp. 163–533, 536–537 in: Bentham, G. & Hooker, J. D. (eds.), Genera Plantarum, vol. 2(1). Reeve & Co., London. Bergqvist, G., Bremer, B. & Bremer, K. 1995. Chloroplast DNA variation and the tribal position of Eremothamnus (Asteraceae). Taxon 44: 341–350. Bremer, K. 1994. Asteraceae: Cladistics and Classification. Timber Press, Portland, Oregon. Cassini, H. 1816. Arctotidées. Pp. 118–119 in: Cuvier, G. L. (ed.), Dictionnaire des Sciences Naturelles, vol. 2 (Suppl.). Le Normant, Paris. Cassini, H. 1821. Les Arctotidées (Arctotideae). Pp. 364–366 in: Cuvier, G. L. (ed.), Dictionnaire des Sciences Naturelles, vol. 20. Le Normant, Paris. Cassini, H. 1828. Vernoniées. Pp. 338–347 in: Cuvier, G. L. (ed.), Dictionnaire des Sciences Naturelles, vol. 57. Le 653 Funk & al. • Evolution of the tribe Arctoteae Normant, Paris. Cassini, H. 1830. Tableau synoptique des Synanthérées. Pp. 566–587 in: Cuvier, G. L. (ed.), Dictionnaire des Sciences Naturelles, vol. 60. Le Normant, Paris. Dittrich, M. 1977. Cynareae-systematic review. Pp. 999–1015 in: Heywood, V. H., Harborne, J. B. & Turner, B. L. (eds.), The Biology and Chemistry of the Compositae, vol. 2. Academic Press, London. Downie S. R. & Katz-Downie, D. S. 1996. A molecular phylogeny of Apiaceae subfamily Apioideae: evidence from nuclear ribosomal DNA internal transcribed spacer sequences. Amer. J. Bot. 83: 234–251. Hasegawa, M., Kishino, H. & Yano T. 1985. Dating the human-ape split by a molecular clock of mitochondrial DNA. J. Molec. Evol. 22: 160–174. Herman, P. P. J., Retief, E., Koekemoer, M. & Welman, W. G. 2000. Asteraceae (Compositae). In Leistner, O. A. (ed.), Seed plants of Southern Africa. Strelitzia 10: 101–170. Heywood, V. H., Harborne, J. B. & Turner, B. L. (eds.). 1977. The Biology and Chemistry of the Compositae, vols. 1–2. Academic Press, London. Heywood, V. H. & Humphries, C. 1977. Anthemideae-systematic review. Pp. 851–898 in: Heywood, V. H., Harborne, J. B. & Turner, B. L. (eds.), The Biology and Chemistry of the Compositae, vol. 2. Academic Press, London. Hoffmann, O. 1890. Compositae. Pp. 87–391 in: Engler, A. & Prantl, K. (eds.), Die natürlichen Pflanzenfamilien, vol. 4(5). Wilhelm Engelmann, Leipzig. Jansen, R. K. 1992. Current Research. Pl. Molec. Evol. Newsl. 2: 13–14. Jansen, R. K., Michaels, H. J. & Palmer, J. D. 1991. Phylogeny and character evolution in the Asteraceae based on chloroplast DNA restriction site mapping. Syst. Bot. 16: 98–115. Jones, S. 1977. Vernonieae-systematic review. Pp. 503–521 in: Heywood, V. H., Harborne, J. B. & Turner, B. L. (eds.), The Biology and Chemistry of the Compositae, vol. 1. Academic Press, London. Karis, P. O. 1992. Hoplophyllum DC., the sister group to Eremothamnus O. Hoffm. (Asteraceae)? Taxon 41: 193–198. Karis, K., Eldenäs, P. & Källersjö, M. 2001. New evidence for the systematic position of Gundelia L. with notes on delimitation of Arctoteae (Asteraceae). Taxon 50: 105–114. Kim, H.-G., Funk, V. A., Vlasek, A. & Zimmer, E. A. 2003. A Phylogeny of the Munnoziinae (Compositae, Liabeae): circumscription of Munnozia and a new placement of M. perfoliata. Pl. Syst. Evol. 239: 171–186. Kim, K. J. & Jansen, R. K. 1995. ndhF sequence evolution and the major clades in the sunflower family. Proc. Natl. Acad. Sci. U.S.A. 92: 10379–10383. Leins, P. 1970. Die Pollenkörner und Verwandtschaftsbeziehungen der Gattung Eremothamnus (Asteraceae). Mitt. Bot. Staatssamml. München 7: 369–376. Leistner, O. A. (ed.). 2000. Seed Plants of Southern Africa: families and genera. Strelitzia 10: 1–775. Lessing, C. F. 1832. Synopsis Generum Compositarum. Duncker & Humblot, Berlin. Merxmüller, H. 1967. Asteraceae. In: Merxmüller, H. (ed.), Prodromus einer Flora von Flora von Südwestafrika, vol. 654 53 (3) • August 2004: 637–655 139. Cramer, Vaduz. Merxmüller, H. 1977. Inuleae-systematic review. Pp. 577–602 in: Heywood, V. H., Harborne, J. B. & Turner, B. L. (eds.), The Biology and Chemistry of the Compositae, vol. 1. Academic Press, London. Moore, S. 1929. Alabastra diversa. XXXVI(2). Notes on African Compositae. J. Bot. 67: 273–274. Norlindh, T. 1977. Arctoteae-systematic review. Pp. 943–959 in: Heywood, V. H., Harborne, J. B. & Turner, B. L. (eds.), The Biology and Chemistry of the Compositae, vol. 2. Academic Press, London. Panero, J. L. & Funk, V. A. 2002. Toward a phylogenetic subfamilial classification for the Compositae (Asteraceae). Proc. Biol. Soc. Washington 115: 909–922. Robinson, H. 1992. Observations on the unique form of sweeping hairs on the styles of the Eremothamneae (Asteraceae). Taxon 41: 199–200. Robinson, H. 1994. Notes on the tribes Eremothamneae, Gundelieae, and Moquinieae, with comparisons of their pollen. Taxon 43: 33–44. Robinson, H. & Brettell, R. D. 1973a. Tribal revisions in the Asteraceae. VIII. Arctotideae. Phytologia 26: 78–80. Robinson, H. & Brettell, R. D. 1973b. Tribal revisions in the Asteraceae. XI. Eremothamneae. Phytologia 26: 163–166. Robinson, H. & Funk, V. A. 1987. A phylogenetic analysis of Leiboldia, Lepidonia, and a new genus Stramentopappus (Vernonieae: Asteraceae). Bot. Jahrb. Syst. 108: 213–228. Rydin, C. & Källersjö, M. 2002. Taxon sampling and seed plant phylogeny. Cladistics 18: 485–513. Stix, E. 1960. Pollenmorphologische Untersuchungen an Compositen. Grana Palynol. 2: 41–114. Swofford, D. L. 2002. PAUP*. Phylogenetic Analysis Using Parsimony (* and other methods). Version 4.0b10. Sinauer Associates, Sunderland, Massachusetts. Taberlet, P., Gielly, L., Pautou, G. & Bouvet, J. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Pl. Molec. Biol. 17: 1105–1109. Watson, L. E., Evans, T. M. & Boluarte, T. 2000. Molecular phylogeny and biogeography of the tribe Anthemideae (Asteraceae), based on chloroplast gene ndhF. Molec. Phyl. Evol. 15: 59–69. White, T. J., Bruns, T., Lee, S. & Taylor, J. 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. Pp. 315–322 in: Innis, M., Gelfand, D., Sninsky, J. & White, T. (eds.), PCR Protocols: A Guide to Methods and Applications. Academic Press, San Diego. Yoo, K.-O., Lowry P. P. II & Wen, J. 2002. Discordance of chloroplast and nuclear ribosomal DNA data in Osmorhiza (Apiaceae). Amer. J. Bot. 89: 966–971. Appendix. Sources of sequence data and GenBank numbers. Barnadesia K79 Carthamus 117 Cirsium 118 Dasyphyllum 465 Gerbera 75 Gerbera 134 Genus Species Authority Collector & No. Locality Herb. ITS trnL ndhF Arctoteae Arctotheca Arctoteae Arctotis Arctoteae Arctotis Arctoteae Berkheya Arctoteae Berkheya Arctoteae Berkheya Arctoteae Berkheya Arctoteae Cullumia Arctoteae Cullumia Arctoteae Cuspidia Arctoteae Cymbonotus Arctoteae Didelta Arctoteae Didelta Arctoteae Dymondia Eremothamneae Eremothamnus Arctoteae Gazania Arctoteae Gazania Arctoteae Gazania Arctoteae Gorteria Arctoteae Gorteria Arctoteae Haplocarpha Arctoteae Heterolepis Arctoteae Heterorhachis Arctoteae Hirpicium Arctoteae Hirpicium Arctoteae Hirpicium Eremothamneae Hoplophyllum Arctoteae Platycarpha calendula bellidifolia fastuosa carlinopsis cruciata cuneata spinosissima bisulca rigida cernua lawsonianus carnosa spinosa margaretae marlothianus sp. krebsiana tenuifolia diffusa personata scaposa aliena aculeata echinus gazanioides integrifolium spinosum carlinoides (L.) Levyns Berg. Jacq. Welw. ex O. Hoffm. Willd. Willd. Willd. (Thunb.) Less. DC. (L.f.) B.L. Burtt Gaudich. Ait. (L.f.) Ait. Compton O.Hoffm. Funk 12266 Koekemoer & Funk 1926 Trinder-Smith 238 Bourell et al. 2689 Koekemoer 2002 Funk 12275 Koekemoer & Funk 1962 Koekemoer & Funk 1935 Trinder-Smith 182 Koekemoer 1986 Holland & Fechner 1336 Koekemoer & Funk 1943 Trinder-Smith 142 Trinder-Smith 197 South Africa South Africa South Africa South Africa South Africa South Africa South Africa South Africa South Africa South Africa Australia South Africa South Africa South Africa Namibia South Africa South Africa South Africa South Africa South Africa South Africa South Africa South Africa South Africa Namibia South Africa South Africa South Africa US PRE US MO PRE PRE PRE PRE US PRE US PRE US US PRE PRE US US US US US PRE PRE MO PRE PRE PRE AY504703 AY504704 AY504705 AY504709 AY504712 AY504711 AY504710 AY504713 AY504714 AY504715 AY504706 AY504716 AY504717 AY504707 AY504718 AY504719 AY504720 AY504722 AY504721 AY504708 AY504700 AY504723 AY504724 AY504726 AY504725 AY190607 AY504701 AY504785 AY504786 AY504787 AY504791 AY504794 AY504793 AY504792 AY504795 AY504796 AY504797 AY504788 AY504798 AY504799 AY504789 AY504800 AY504801 AY504802 AY504804 AY504803 AY504790 AY504782 AY504805 AY504806 AY504808 AY504807 AY504784 AY504783 AY504745 AY504746 AY504747 AY504751 AY504754 AY504753 AY504752 AY504755 AY504756 AY504757 AY504748 AY504758 AY504759 AY504749 L39424 AY504760 AY504761 AY504762 AY504763 AY504750 AY504742 AY504764 AY504766 AY504765 AY504744 AY504743 Lactuceae Gundelieae Vernonieae Vernonieae Lactuceae Lactuceae Liabeae Liabeae Liabeae Gundelieae Cichorium Gundelia Gymnanthemum Hesperomannia Hypochaeris Lactuca Munnozia Oligactis Philoglossa Warionia intybus tournefortii amygdalina arborescens glabra sativa gigantea volubilis mimuloides saharae L. Kelloff & Funk 1415 L. al-Hosseini s.n. Delile Kew 318-86-02802 s.n. A. Gray Ching A20 L. Funk 12217 L. Funk s.n. (Rusby) Rusby Dillon s.n. (Kunth) Cass. Funk 12042 (Hieron.) H.Rob. & Cuatr. Funk 11453 Benth. & Cross Lippert 25346 USA Iran Africa (cultivated) Hawaii (cultivated) Australia cultivated Peru Colombia Ecuador Morocco US US K PTBG US none F US US US AY504694 AY504691 AY504695 AY504696 AY504692 AY504693 AY504697 AY504698 AY504699 AY190608 AY504736 AY504733 AY504737 AY504738 AY504734 AY504735 AY504739 AY504740 AY504741 AY702088 Barnadesieae Cardueae Cardueae Barnadesieae Mutisieae Mutisieae Barnadesia Carthamus Cirsium Dasyphyllum Gerbera Gerbera caryophylla oxycantha drummundii reticulatum crocea sp. S.F. Blake Bieb. Torr. & Gray (DC.) Cabrera Kuntze Peru (cultivated) Iran U.S.A. Brazil South Africa Africa (cultivated) MICH US US US PRE none AY504686 AY504689 AY504690 AY504687 AY504688 AY504776 AY504773 AY504777 AY504778 AY504774 AY504775 AY504779 AY504780 AY504781 AY702089/ AY702090 AY504768 AY504771 AY504772 AY504767 AY504769 AY504770 Less. Less. Thunb. L. Harv. (L.f.) Druce (Burm.f.) Roessler Less. (Harv.) Roessler Less. DC. Oliver & Hiern. Koekemoer & Funk 1929 Koekemoer & Funk 1969 Trinder-Smith 64 Trinder-Smith 103 Dodd 289 Trinder-Smith 191 Trinder-Smith 188 Oliver 3867 Koekemoer & Funk 1966 Long & Rae 734 Koekemoer & Funk 1956 Koekemoer 2045 Germishuizen 2613 Matthei Bot. Gard s.n. al-Hosseini s.n. Funk 12302 Roque, Funk & Kim 485 Koekemoer & Funk 1924 Funk s.n. AY504728 AY504731 AY504732 AY504727 AY504729 AY504730 Funk & al. • Evolution of the tribe Arctoteae Tribe 53 (3) • August 2004: 637–655 655 Label Ingroup Arctotheca 95 Arctotis 08 Arctotis 14 Berkheya 18 Berkheya 99 Berkheya 31 Berkheya 24 Cullumia 37 Cullumia 39 Cuspidia 104 Cymbonotus l45 Didelta 50 Didelta 51 Dymondia 53 Eremothamnus GB Gazania 59 Gazania 62 Gazania 64 Gorteria 73 Gorteria 69 Haplocarpha 77 Heterolepis 82 Heterorhachis 84 Hirpicium 25 Hirpicium 87 Hirpicium 86 Hoplophyllum 109 Platycarpha 91 Outgroup Cichorium 36 Gundelia 96 Gymanthemum K11 Hesperomannia 1 Hypochaeris 33 Lactuca 133 Munnozia K39 Oligactis 121 Philoglossa 120 Warionia s.n.