PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON 115(4):909-922. 2002. Toward a phylogenetic subfamilial classification for the Compositae (Asteraceae) Jose L. Panero and Vicki A. Funk (JLP) Section of Integrative Biology, The University of Texas, Austin, Texas 78712 (VAF) U.S. National Herbarium, NMNH, Smithsonian Institution MRC 166, Washington, D.C. 20013-7012 Abstract. —Molecular studies of the flowering plant family Compositae (As- teraceae) based on comparative DNA sequence data of chloroplast genes pro- vide new insights into the evolution and radiation of the family. The results support the creation of new groups to maintain a classification that reflects evolutionary history. We are proposing the following new names: subfamilies Corymbioideae, Gochnatioideae, Gymnarrhenoideae, Hecastocleoideae, and Pertyoideae; tribes Athroismeae, Corymbieae, Dicomeae, Gochnatieae, Gym- narrheneae, Hecastocleideae, Polymnieae; subtribe Rojasianthinae. The totals now stand at 1 1 subfamilies and 35 tribes. Only one tribe, the Mutisieae, is non-monophyletic having two branches. Most of the new groups are derived from taxa included in tribe Mutisieae, long suspected to be a paraphyletic group. Molecular studies that support these changes are discussed elsewhere but a summary of their results is presented. Molecular phylogenetic studies of the of the Compositae using sequence data of family Compositae (Asteraceae) have been the chloroplast gene ndhF and provide ad- instrumental in providing important insight ditional evidence for the paraphyly of sub- into the evolution and radiation of the fam- family Cichorioideae. Their data provided ily. Assessments of phylogenetic relation- some support for the recognition of three ships in the family using molecular data be- main groups within subfamily Cichorioi- gan with restriction fragment length poly- deae s.l. splitting sequentially from above morphisms of the chloroplast genome. the Barnadesioideae: 1) Mutisieae, followed These studies identified a 22 kb inversion by 2) Cardueae, and finally, 3) the tribes present in all members of the Compositae Cichorieae, Liabeae, Vernonieae, and Arc- except the Mutisieae subtribe Barnadesieae toteae forming a monophyletic group sister (Jansen & Palmer 1987), which had a chlo- to Asteroideae. The latter group plus Aster- roplast organization similar to the rest of oideae was recognized by Bremer (1994, land plants. This result was used to elevate 1996) as the vernonioid group. Since the the Barnadesieae to subfamily status (Bre- publication of the molecular phylogeny of mer & Jansen 1992), and the traditional two the Compositae by Kim & Jansen (1995) subfamily system consisting of Cichorioi- only the study by Bayer & Starr (1998) has deae and Asteroideae was expanded to in- provided a competing hypothesis of tribal clude the Barnadesioideae. Kim & Jansen relationship based on noncoding regions of (1995) published what can arguably be the chloroplast genome, but this study did viewed as one of the most important papers not provide good support for any new pat- using DNA sequence data aimed at eluci- terns of relationship. Most studies using the dating the phylogeny of the Compositae. In ndhF data matrix to date have concentrated this paper they identified the major clades on elucidating the placement of a particular 910 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON taxon in the tree of life of daisies (Karis et argue that eventually they might be part of al. 2001). the Asteroideae and Cichorioideae respec- The goal of our study was to use several tively, but given the amount of data that molecular markers to produce a phylogeny support this cladogram and the fact that the that would clarify the phylogenetic relation- isolated position of these genera has long ships of members of the tribe Mutisieae s.l., been recognized, it is doubtful. Details of and the circumscription of the subfamily the analyses and a discussion of the result- Cichorioideae in general. Our data matrix ing clades will be published elsewhere includes the genes and markers ndhF, trnh- (Funk & Panero, in prep.; Panero et al., in trnF, matK, ndhD, rbcl^, rpoB, rpoCl prep.); we publish these names here so that exonl, 23S-trnl, and ndhl for an approxi- they are available for use in those papers. mate total data matrix of 13,380 bp. Com- The changes to the classification of the bining these chloroplast markers has pro- family proposed herein have been carefully duced a solid phylogeny of the Compositae. reviewed by the authors with the under- Included here is a strict consensus tree of standing that these changes represent the the relationships of Compositae based on largest overhaul to the classification of the our studies for the Cichorioideae (Panero & Compositae since Bentham (1873). Our Funk, in prep.) and those of Panero et al. confidence in proposing these changes is (in prep.) for tribe Heliantheae and relatives based on two things, first, an analysis of (Fig. 1). This phylogeny includes a majority one of the largest data matrices for any of genera of the classical Mutisieae and rep- group of flowering plants produced to resentatives of most other tribes of Com- date—a combination of multiple genetic positae as well as genera deemed, based on markers and a broad sampling of taxa—and their morphology, to be of difficult place- second, an analysis of these data by con- ment (see Bremer 1994). The resulting phy- servative methods (i.e., maximum parsi- logeny has clarified substantially the rela- mony) without the coding of indels or char- tionships of the Compositae and resulted in acter weighting of any type. Bootstrap sup- the need to publish new names to maintain port for nearly all of the clades being rec- a formal taxonomic nomenclature for ognized as new taxa is 100%. Our monophyletic clades. All groups are now experience with this large data matrix monophyletic and reflect the evolutionary shows that additions of new data partitions history of the family except for two branch- only solidifies the phylogenetic signal and es near the base of the tree that constitute branch support. It is clear that past a certain a paraphyletic Mutisioideae. It is most like- threshold, unique to each dataset and deter- ly that additional data will provide enough mined by the number of characters and taxa support to recognize the "Stifftia group" as sampled, the phylogenetic signal stabilizes. a new subfamily and tribe, however the It is our view that the addition of new data data are missing several of the genera that partitions in the future will only solidify the should fall into the "Stifftia group" clade branch support reported here. and final resolution of the problem will have to wait until those genera have been Subfamilies collected and analyzed. Two of the mono- Corymbioideae Panero & V. A. Funk, sub- phyletic subfamilies proposed, Corymbioi- fam. no v., Corymbieae Panero & V. A. deae and Gymnarrhenoideae, contain gen- Funk, trib. nov. era whose placement has long been prob- Type: Corymbium L., Coroll. Gen. 14 lematic. Their isolated placement on the (1737). cladogram combined with their unusual Monotypic (7 species). morphology supports their being named at the subfamily (and tribal) level. One might Plantae in foliis sessilibus et in laminis VOLUME 115, NUMBER 4 911 Tribes Subfamilies Heliantheae | 100 58 Madieae Perityleae I 68 51 Eupatorieae 72 Millerieae 95 Polymnieae 981 Bahieae Chaenactideae 98 Tageteae 100 Neurolaeneae Asteroideae Coreopsideae 100 Helenieae 100 Athroismeae Plucheeae 100 1 Inuleae 54 Anthemideae Astereae 60 Gnaphalieae 100 ^ Calenduleae Senecioneae Corymbieae Corymbioideae 79 Arctoteae Vernonieae 90 Liabeae Cichorioideae 100 Cichorieae 100 Gundelieae Gymnarrheneae Gymnarrhenoideae 100 Pertyeae Pertyoideae Cardueae 100 100 Tarchonantheae Carduoideae Dicomeae 53 E Hecastocleideae Hecastocieioideae Gochnatieae Gochnatioideae 100 Mutisieae (SA) Mutisioideae iqoj Mutisieae (SA, A) "The Stifftia Group" Bamadesieae Bamadesioideae t ^^^^^—^^^^—^^-^^^^^^ Outgroups (Calyceraceae, Goodeniaceae) Fig. 1. Strict consensus tree of 12 most parsimonious trees for 122 representative Compositae taxa based on chloroplast DNA sequence data. Bootstrap values indicated above nodes, SA = South America, A = Asia. Tree is simplified to show only tribe and subfamily placement and arrangement. 912 PROCEEDINGS OF THE BIOLOGICAL SOCIETY OF WASHINGTON conduplicatis. Capitula uniflora; bracteis in- obtuse or acute, with short sweeping hairs volucri biseriatis, dimorphis; bracteis inter- on abaxial side, sweeping hairs restricted to ioribus 2, involutis; limbis corollarums sal- style and scarcely present below bifurcation verformibus. point on the style shaft. Achenes linear to Scapose, perennial herbs, tufted and elliptic, hirsute or villous, silvery white, forming rosettes. Leaves alternate, sessile, pappus of short, crown-like scales variously sometimes conduplicate, blades lanceolate fused or divided into lobes and bristles. to linear, leaf bases attenuate, leaf apices Chromosome number: 2N = 16 (Weitz acuminate, terete, involute or flat in cross 1989). section, margins entire, venation parallel, The phylogenetic position of the South leaf surfaces glabrous to villose or hirsute, African genus Corymbium has been histor- sometimes with glandular trichomes. Inflo- ically controversial (Bentham 1873) and in rescence of open to congested corymbiform the latest revision of the family at the ge- cymes at the end of scapes that are as tall neric level, the genus is placed in subfamily