Molecular Phylogenetics and Evolution Vol. 22, No. 1, January, pp. 51– 64, 2002 doi:10.1006/mpev.2001.1038, available online at http://www.idealibrary.com on Tribal and Subtribal Delimitation and Phylogeny of the Cardueae (Asteraceae): A Combined Nuclear and Chloroplast DNA Analysis Núria Garcia-Jacas, Teresa Garnat je, Alf onso Susanna, 1 and Roser Vilat ersana Botanical Institute of Barcelona (C.S.I.C.-Ajuntament de Barcelona), Av. M untanyans, s.n., E-08038 Barcelona, Spain Received August 28, 2000; revised M arch 15, 2001; published online December 5, 2001 this matter (Susanna et al., 1995) revised exhaustively the different classifications. Table 1 summarizes the different classifications. The earliest classification by Cassini (1819) recognized three tribes, Echinopeae, Carlineae, and Cardueae, the latter with two subtribes, Carduinae and Centaureinae. Bentham (1873) and Hoffmann (1894), in contrast, suggested a broadly defined Cardueae comprising four subtribes: Echinopsidinae, Carlininae, Carduinae, and Centaureinae. This treatment has been generally accepted, but the problem surfaced again when Wagenitz (1976) suggested the segregation of the Echinopeae as a separate tribe. Soon after, Dittrich (1977) returned to Cassini’s views and segregated the Echinopeae and the Carlineae. Petit (1988) followed Wagenitz (1976) and kept a separate tribe Echinopeae. In view of the contradicting views, the last general survey of the Asteraceae (Bremer, 1994) adopted the more conservative, broad tribal concept and recognized only one tribe. In a previous molecular analysis based on internal transcribed spacer (ITS) sequences (Susanna et al., 1995), we concluded that the Cardueae were monophyletic, including Carlininae and Echinopsidinae. The same conclusion had been reached by Jansen et al. (1990, 1991), on the basis of chloroplast DNA restriction site data, and by Kim et al. (1992) on the basis of rbcL sequences. Cladistic analyses of morphological data in the Compositae (Bremer, 1987, 1994; Karis et al., 1992) also supported the monophyly of a broadly defined Cardueae, including Echinopsidinae and Carlininae in addition to Carduinae and Centaureinae. However, these results did not settle the discussion: in a cladistic analysis of morphologic characters, Petit (1997) concluded again that Echinopeae should not be included in the Cardueae. Most of the difficulties have originated in the Echinopsidinae, characterized by uniflowered capitula grouped in second-order heads. This structure has been the main reason for the segragation of Echinopsidinae as a different tribe by Wagenitz (1976), Dittrich (1977), and Petit (1988, 1997). Homologies of the synflorescence of Echinops L. and related genera are ex- Tribal delimitation of Cardueae is controversial, and the traditional classification in four subtribes (Echinopsidinae, Carlininae, Carduinae, and Centaureinae) has fluctuated widely. Most of the problems are centered in subtribes Echinopsidinae and Carlininae, often segregated with tribal rank. We therefore analyzed DNA sequences of the internal transcribed spaces (ITS) of the nuclear ribosomal DNA genes and the matK gene of the chloroplast DNA of a broad representation of the tribe to examine (1) the phylogeny of the tribe, (2) the position of Echinopsidinae and Carlininae, (3) the circumscription of the subtribes and the position of some conflicting genera, and (4) the delimitation of some generic complexes in the Carduinae. Phylogenetic analysis of ITS and matK sequence variation, both separate and combined, strongly support the monophyly of Cardueae including Carlininae and Echinopsidinae. The combination of both genomes suggest that Xeranthemum and its allies should be included among the Echinopsidinae rather than the Carlininae, which implies that the capitulum of Xeranthemum could be interpreted as a syncephaly. The subtribe Centaureinae forms a well-supported clade, and their sister clades contain the genera Arctium, Cousinia, Jurinea, and Saussurea from the Carduinae. However, some problems persist: Carduinae are a paraphyletic assemblage, and the subtribal placement of Berardia, Cardopatium, Cousiniopsis, and Staehelina remains unresolved. Our results also indicate that present classification in four subtribes is unsatisfactory, but it is still the only practical approach. © 2001 Elsevier Science Key Words: Asteraceae; Cardueae; ITS; matK; phylogeny; tribal delimitation. INTRODUCTION The tribe Cardueae (Asteraceae) is usually divided into four taxonomic entities, but the rank and delimitation of these units are highly problematic. A paper on 1 To whom correspondence should be addressed. Fax: 34 93 4269321. E-mail: asusanna@ibb.csic.es. 51 1055-7903/01 $35.00 © 2001 Elsevier Science All rights reserved. 52 GARCIA-JACAS ET AL. TABLE 1 Different Subtribal Classifications of the Cardueae Cassini (1819), Dittrich (1977) Bentham (1873), Hoffmann (1894), Bremer (1994) Wagenitz (1976), Petit (1988, 1997) Tribe Echinopeae Tribe Carlineae Tribe Cardueae Subtribe Carduinae Subtribe Centaureinae Tribe Cardueae Subtribe Echinopsidinae Subtribe Carlininae Subtribe Carduinae Subtribe Centaureinae Tribe Echinopeae Tribe Cardueae Subtribe Carlininae Subtribe Carduinae–Centaureinae tremely difficult to establish (Petit, 1988), as is usually the case with highly derived syncephalies (Stuessy and Spooner, 1988). A similar problem is posed by the complicated involucral structures of the Carlininae, which have been interpreted even as a third-order syncephalia in the genus Carlina L. (Meusel and Kohler, 1960; Meusel and Kästner, 1994), an interpretation rejected by Petit (1988). In fact, contradictory results obtained by different authors in cladistic analyses of morphologic data (e.g., Karis et al., 1992 versus Petit, 1997) could be partly attributed to different points of view in addressing the problem of homologies of the receptacular structures of Echinops, Carlina, and related genera. Not only the limits of the tribe are problematic. The boundaries between the classic four subtribes are very difficult to establish. By example, most of the authors (Bentham, 1873; Hoffmann, 1894; Dittrich, 1977; Bremer, 1994) included the genera Amphoricarpos Vis., Cardopatium Juss., Chardinia Desf., Cousiniopsis Nevski, Siebera J. Gay, Staehelina L., and Xeranthemum L. among the Carlininae, whereas Petit (1997) moved Cardopatium and Cousiniopsis to the Echinopeae (to which he assigned tribal rank) and the rest of genera to the Carduinae–Centaureinae. There are some other genera that are very difficult to classify, such as Berardia Vill.: it was included among the Cardueae by Bremer (1994) and moved to the Mutisieae by Dittrich (1996a). Finally, there are the problems of generic delimitation posed by some large genera of the tribe: Carduus L. (90 species), Cirsium Mill. (250 species), Centaurea L. (400 species), Cousinia Cass. (800 species), Jurinea Cass. (100 species), and Saussurea DC. (more than 300 species). The natural delimitation of Centaurea (400 species) was recently cleared up by Garcia-Jacas et al. (2000, 2001), and some apportionments on the limits of Cirsium and Carduus were made recently by Häffner and Hellwig (1999), but the rest of the problems persist. Many small genera from central and west Asia have been described on the basis of splits of Cousinia and Jurinea (Susanna and Garcia-Jacas, 2001). By example, the genera Hyalochaete Dittrich & Rech. f., Jurinella Jaub. & Spach, and Outreya Jaub. & Spach are doubful segregates from Jurinea. Also, according to Duistermaat (1996, 1997) and Petit (1997), the limits between Arctium L. and Cousinia are unclear and controversial. Clearly, a molecular approach seemed most applicable for this conflicting group, as the results of cladistic morphological analysis leads to contradictory results. Problems encompassed a broad taxonomic range, as we intended to address questions from the generic to the tribal level. Thus, we gathered information from different sources by sequencing two regions of the genome: the nuclear ribosomal DNA (nrDNA) internal transcribed spacers (ITS1 and ITS2) and the chloroplast DNA (cpDNA) gene matK. The ITS region was extensively demonstrated to be a source of phylogenetic information in the Cardueae, in the tribal (Susanna et al., 1995), subtribal (Häffner and Hellwig, 1999; Garcia-Jacas et al., 2001), and generic (Susanna et al., 1999; Vilatersana et al., 2000) levels. In regard to matK, if ITS has its major weakness in the resolution of remote subtribal groups (Susanna et al., 1995), the matK gene was more applicable to the suprageneric level. It had been used already in the Asteraceae (Denda et al., 1999) and in similar taxonomic levels in other groups such as the Apiaceae (Plunkett et al., 1996), Cornaceae (Xiang et al., 1998), Myrtaceae (Gadek et al., 1996), Polemoniaceae (Johnson and Soltis, 1995), and Saxifragaceae (Johnson and Soltis, 1994; Soltis et al., 1996; Brochmann et al., 1998). On the other hand, the combination of different genomes is at the present moment one of the best tools for phylogenetic reconstruction (Qiu et al., 1999). The combined approach solves one of the most problematic issues of molecular phylogenies based on one genome: despite offering a great number of characters, at the end they conform to a one-character-based taxonomy (Doyle, 1992). Our goals in this study were (1) to evaluate the monophyly of the tribe Cardueae; (2) to compare the molecular phylogeny with the traditional classification in four subtribes (Echinopsidinae, Carlininae, Carduinae, and Centaureinae); (3) to examine the boundaries between the four subtribes and the position of conflicting genera such as Berardia, Cardopatium, Cousiniopsis, Staehelina, and the Xeranthemum group; and (4) to verify the delimitation of some of the large genera of COMBINED NUCLEAR–CHLOROPLAST DNA PHYLOGENY OF TRIBE CARDUEAE the tribe (the Arctium–Cousinia group and the Jurinea complex). MATERIALS AND METHODS Plant Material Sampling was based on the subtribal classification of the Cardueae in four subtribes by Bentham (1873), Hoffmann (1894), and Bremer (1994) with the modifications suggested by Susanna and Garcia-Jacas (2001), which recognize 74 genera. With regard to Centaureinae, only 3 genera were included, because this subtribe is a very well-defined monophyletic group and had been the object recently of a deep survey, with 25 of 33 genera studied (GarciaJacas et al., 2001). As for the rest of the tribe, we included 31 of 41 genera (according to Susanna and Garcia-Jacas, 2001) of the subtribes Carduinae, Carlininae, and Echinopsidinae. Two outgroup species were chosen in the tribe Mutisieae according to previous sequence analysis (Susanna et al., 1995). Voucher data, source, and GenBank sequence accession numbers for the 62 studied species are given in Table 2. The analysis of ITS sequences used published sequences along with new sequences. Sequences of Centaurea involucrata Desf., Cirsium arvense (L.) Scop., Cynara humilis L., Jurinea humilis (Desf.) DC., Ptilostemon hispanicus (Lam.) Greuter (wrongly published as Galactites durieui Spach), and Zoegea mianensis Boiss. were from previous studies (Susanna et al., 1995; Garcia-Jacas et al., 2000). All the matK sequences analyzed are new. DNA Extraction, Amplification, and Sequencing Total genomic DNA was extracted, following the CTAB method of Doyle and Doyle (1987) as modified by Soltis et al. (1991), from silica-gel-dried leaves collected in the field or fresh leaves of plants cultivated in the Botanic Institute of Barcelona. In some cases, herbarium material was used. cpDNA matK gene strategies. Double-stranded DNAs of matK were amplified by PCR with trnK-710F and matK-1848R (Johnson and Soltis, 1995) as PCR primers. An additional primer, AST-1R, was designed for this study, as a substitute for primer matK-1848R. The sequence is as follows: 59 CCGCACACTTGAACG/ CATAACCCAG 39. This primer combination failed to generate double-stranded products for some taxa. Therefore, trnK-3914F and trnK-2R (Johnson and Soltis, 1995) were used in place of those primers. The profile used for amplification included a warm start at 94°C for 2 min, followed by 80°C for 5 min, during which the polymerase (Ecotaq; Ecogen S. R. L., Barcelona, Spain) was added. Forty cycles of amplification were carried out under the following conditions: 94°C 53 for 1 min 30 s, 48°C for 2 min, and 72°C for 3 min, with an additional extension step of 15 min at 72°C. The double-stranded PCR products were cleaned with a QIAquick PCR Purification Kit (Qiagen Inc.) and sequenced. In general, four sequencing primers, trnK-710F, matK-1168R, matK-1848R (Johnson and Soltis, 1995), and AST-1R were used. Direct sequencing of the amplified DNA segments was performed with a Thermo Sequenase II Dye Terminator Cycle sequencing kit (Amersham), following the protocol recommended by the manufacturer. The nucleotide sequencing was performed at the Serveis Cientı́fico-Tècnis of the University of Barcelona on an ABI 377 Automated DNA Sequencer (Perkin–Elmer). Nucleotide sequences of matK were edited with Chromas 1.56 (Technelysium Pty. Ltd.) and easily aligned by hand. nrDNA ITS region strategies. Double-stranded DNA of the ITS region was amplified with the 1406F primer (Nickrent et al., 1994) and ITS4 (White et al., 1990). In some cases, we used ITS1 (White et al., 1990) and 17SE (Sun et al., 1994) as forward primers and 26SE (Sun et al., 1994) as reverse primer. The profile used for amplification included a warm start at 94°C for 2 min, followed by 80°C for 5 min, during which the polymerase (Ecotaq; Ecogen S. R. L., Barcelona, Spain) was added. Thirty cycles of amplification were carried out under the following conditions: 94°C for 1 min 30 s, 55°C for 2 min, and 72°C for 3 min, with an additional extension step of 15 min at 72°C. The PCR products were purified with the QIAquick PCR Purification Kit (Qiagen Inc.). Both strands were sequenced with the sequencing primers 1406F, ITS1, and 17SE as forward primers and ITS 4 and 26SE as reverse primers. Direct sequencing of the amplified DNA segments was performed as for the matK region. Phylogenetic Analysis DNA sequences were aligned visually by sequential pairwise comparison (Swofford and Olsen, 1990). Alignment was checked with ClustalX (Thompson et al., 1997). Data matrices are available on request from the corresponding author. Parsimony analysis involved heuristic searches conducted with PAUP version 4.0b4a (Swofford, 1999) with TBR branch swapping and character states specified as unordered and unweighted. The indels were coded as fifth base, following the now generally accepted trend of saving the potential phylogenetic information of shared indels (Sun et al., 1994; Bain and Jansen, 1995; Samuel et al., 1998). All most parsimonious trees (MPTs) were saved. To locate other potential islands of most parsimonious trees (Maddison, 1991), we performed 100 replications with random taxon addition and TBR branch swapping. Bootstrap analyses (BS) were performed (Felsenstein, 1985), and decay indices (DI) were calculated (Bremer, 1988; Donoghue et al., 1992) to obtain esti- 54 GARCIA-JACAS ET AL. TABLE 2 Origin of the Materials and Herbaria Where the Vouchers are Deposited Acantholepis orientalis Less. Alfredia cernua (L.) Cass. Arctium lappa L. Arctium minus Bernh. Atractylis cancellata L. Atractylis carduus (Forssk.) Christ. Atractylis humilis L. Atractylodes japonica Koidz. ex Kitam. Berardia subacaulis Vill. Cardopatium corymbosum (L.) Pers. Carduus carlinoides Gouan Carduus pycnocephalus L. Carlina acanthifolia All. Carlina falcata Svent. Carlina gummifera (L.) Less. Carlina lanata L. Carlina macrophylla (Desf.) DC. Carlina vulgaris L. Centaurea hajastana Tzvel. Centaurea involucrata Desf. Centaurea lingulata Lag. Chardinia orientalis (L.) O. Kuntze Cirsium arvense (L.) Scop. Cirsium echinus (M. Bieb.) Hand.Mazz. Cirsium palustre (L.) Scop. Cousinia canescens DC. Cousinia esfandiarii Rech. F. & Aellen Cousinia onopordioides Lbd. Cousiniopsis atractyloides (C. Winkl.) Nevski Cynara cornigera Lind. Cynara humilis L. Echinops ritro L. Echinops sp. Echinops spinosissimus Turra Echinops viscosus DC. Galactites tomentosa Moench Gerbera kunzeana A. Br. & Aschers. Hyalochaete modesta Dittrich & Rech. f. Jurinea humilis (Desf.) DC. Jurinea macrocephala DC. Jurinea sp. Jurinella moschus (Habl.) Bobrov Uzbekistan: Kyzylkum, 30 Km N of Bukhara, Khassanov IX-1999 (BC) [AF319046, AF319100, AY013518]. Denmark: Copenhagen Botanical Garden (BC) [AF319047, AF319101, AY013519]. Belgium: Lovaina Botanical Garden (BC) [AF319048, AF319102, AY013520]. Belgium: Lovaina Botanical Garden (BC) [AF319049, AF319103, AY013521]. Spain, Madrid: Torrelaguna, Garcia-Jacas & Susanna 1458 (BC) [AF319050, AF319104, AY013522]. Egypt: near El Amiriya, Susanna 1856 & Vilatersana (BC) [AF319051, AF319105, AY013523]. Spain, Toledo: Huerta de Valdecarábanos, Susanna 1883 (BC) [AF319052, AF319106]. Japan: Tokyo Botanical Garden (BC) [AF319053, AF319107, AY013524]. France, Alpes Maritimes: Col de la Cayolle, Garnatje 27 & Luque (BC) [AF319054, AF319108, AY013525]. Greece, Macedonia: between Thermi and Thessaloniki, Roché & Susanna 1951 (BC) [AF319055, AF319109, AY013526]. Spain, Girona: Abella, Garnatje 11 & Luque (BC) [AF319056, AF319110, AY013527]. Spain, Barcelona: Montjuı̈c, Garnatje & Susanna 1827 (BC) [AF319057, AF319111, AY013528]. Spain, Girona: Abella, Garnatje 8 & Luque (BC) [AF319058, AF319112, AY013529]. Spain, Canary Islands, La Palma: Gallegos, Garnatje 3 & Luque (BC) [AF319059, AF319113, AY013530]. Switzerland: Genève Botanical Garden (BC) [AF319060, AF319114, AY013531]. Creta, Hania: Phalasarna, Vilatersana 39 (BC) [AF319061, AF319115, AY013532]. Morocco: between Xauen and Oued Laou, Garnatje, Susanna 1893 & Vilatersana (BC) [AF319062, AF319116]. Switzerland: Zürich, Botanical Garden (BC) [AF319063, AF319117, AY013533]. Armenia, Talin: between villages Pokr Artik and Bagravan, Fajvush, Gabrielyan, Garcia-Jacas, Guara, Hovannisyan, Susanna 1587, Tamanyan & Vallès (BC) [AF319064, AF319118, AY013502]. Susanna et al. (1995) for the ITS data [matK: AY013503]. Spain, Madrid: Puerto de Navafrı́a, Garcia-Jacas & Susanna 1462 (BC) [AF058851, AF058876, AY013505]. Iran, Kordestan: 35 km W of Divandarreh, Garcia-Jacas, Mozaffarian, Susanna 1715 & Vallès (BC) [AF319065, AF319119, AY013534]. Susanna et al. (1995). Iran, Azarbayjan-e-Sharghi: 25 Km from Ahad on the road to Kaleibar, Garcia-Jacas, Mozaffarian, Susanna 1667 & Vallès (BC) [AF319066, AF319120, AY013535]. Spain, Girona: Abella, Garnatje 17 & Luque (BC) [AF319067, AF319121, AY013536]. Iran, Azarbaijan-e-Sharghi: 25 Km from Ahad on the road to Kaleibar, Garcia-Jacas, Mozaffarian, Susanna 1608 & Vallès (BC) [AF319068, AF319122]. Iran, Mazandaran: between Gachsar and Valiabad, Garcia-Jacas, Mozaffarian, Susanna 1618 & Vallès (BC) [AF319069, AF319123, AY013537]. Iran, Tehran: between Firuzkuh and Semnan, Garcia-Jacas, Mozaffarian, Susanna 1637 & Vallès (BC) [AF319070, AF319124]. Uzbekistan: Babatak mountains, 40 Km E from Lalmikar, Sukervanik, 18-V-1979 (BC) [AF319071, AF319125]. Egypt, Alexandria: 2 Km N from the city of Alexandria, Susanna 1840 & Vilatersana (BC) [AF319072, AF319126, AY013538]. Susanna et al. (1995). Spain, Tarragona: Xerta, Garnatje & Susanna 1870 (BC) [AF319074, AF319128]. Iran, Hamadan: 5 Km E of Hamadan on the road to Teheran, Garcia-Jacas, Mozaffarian, Susanna 1721 & Vallès (BC) [AF319073, AF319127]. Germany: Berlin Botanical Garden (BC) [AF319075, AF319129, AY013539]. Germany: Berlin Botanical Garden (BC) [AF319076, AF319130, AY013540]. Spain, Barcelona: Montjuı̈c, Garnatje & Susanna 1867 (BC) [AF319077, AF319131, AY013541]. Germany: Stuttgart Botanical Garden (BC) [AF319078, AF319132, AY013542]. Germany: Berlin Botanical Garden (BC) [AF319080, AF319134]. Susanna et al. (1995). Iran, Azarbaijan-e-Sharghi: 10 Km northern Qarabchaman, Garcia-Jacas, Mozaffarian, Susanna 1650 & Vallès (BC) [AF319081, AF319135]. Armenia, Krasnoselsk: Artanish Peninsula, Fajvush, Gabrielyan, Garcia-Jacas, Guara, Hovannisyan, Susanna 1528, Tamanyan & Vallès (BC) [AF319082, AF319136, AY013543]. Iran, Azarbaijan-e-Sharghi: Sabalan mt., 2800 m, Garcia-Jacas, Mozaffarian, Susanna 1662 & Vallès (BC) [AF319083, AF319137]. COMBINED NUCLEAR–CHLOROPLAST DNA PHYLOGENY OF TRIBE CARDUEAE 55 TABLE 2—Continued Mutisia spinosa R. & P. Notobasis syriaca (L.) Cass. Onopordon algeriense Pomel Onopordon leptolepis DC. Outreya carduiformis Jaub. & Spach Picnomon acarna (L.) Cass. Ptilostemon afer (Jacq.) Greuter Ptilostemon chamaepeuce (L.) Less. Ptilostemon hispanicus (Lam.) Greuter Saussurea alpina (L.) DC. Saussurea discolor (Willd.) DC. Serratula coronata L. Silybum marianum (L.) Gaertner Staehelina baetica DC. Synurus palmatopinnatifidus Kitam. Thevenotia scabra (Boiss.) Boiss. Tyrimnus leucographus (L.) Cass. Xeranthemum cylindraceum Sibth. & Sm. Xeranthemum inapertum (L). Miller. Zoegea mianensis Boiss. Spain: Barcelona Botanical Garden (BC) [AF319079, AF319133, AY013544]. Egypt, Alexandria: Burg el Arab, Susanna 1844 & Vilatersana (BC) [AF319085, AF319139, AY013545]. Switzerland: Zürich Botanical Garden (BC) [AY013546]. Iran, Teheran: Sorkhehesar, Garcia-Jacas, Mozaffarian, Susanna 1632 & Vallès (BC) [AF319086, AF319140, AY013547]. Iran, Teheran: Sorkhehesar, Garcia-Jacas, Mozaffarian, Susanna 1631 & Vallès (BC) [AF319087, AF319141, AY013548]. Iran, Azarbaijan-e-Gharbi: Orumiyeh, Garcia-Jacas, Mozaffarian, Susanna 1692 & Vallès (BC) [AF319088, AF319142, AY013549]. Germany: Freiburg Botanical Garden (BC) [AF319088, AF319142, AY013550]. France: Lyon Botanical Garden (BC) [AF319090, AF319144]. Susanna et al. (1995, as Galactites durieui Coss.). Italia: Cogne Botanical Garden (BC) [AF319091, AF319145]. Switzerland: Meyrin Botanical Garden (BC) [AF319092, AF319146]. Austria: Vienna Botanical Garden (BC) [AF319093, AF319147, AY013513]. Spain, Madrid: Torrelaguna, Garcia-Jacas & Susanna 1461 (BC) [AF319094, AF319148, AY013551]. Spain, Málaga: Sierra Bermeja, Garcia-Jacas & Susanna 1826 (BC) [AF319095, AF319149]. Japan: the Nippon Shinyaku Institute for Botanical Research (BC) [AF319096, AF319150, AY013551]. Uzbekistan: Kyzylkum, 50 km northern town by Zhdavan, Khassanov, IX-1999 (BC) [AY013553]. Spain, Tarragona: La Vilella Baixa, R. Pascual, VI-1998 (BC) [AF319097, AF319151, AY013554]. Switzerland: Genève Botanical Garden (BC) [AF319098, AF319152, AY013556]. Spain, Madrid: Torrelaguna, Garcia-Jacas & Susanna 1456 (BC) [AF319099, AF319153, AY013555]. Garcia-Jacas et al. (2000) for the ITS data [matK: AY013517]. Note. ITS 1 and 2 (beginning with AF) and matK (beginning with AY) GenBank Accession Nos. between brackets. Previous publications refer to ITS sequences. mates of support for each monophyletic group. Three parsimony analyses were performed, with three different data sets: the ITS data, the matK data, and the combined ITS and matK data. Due to the unpracticability of performing bootstrap analysis in the usual way with PAUP 4.0b4a in the cpDNA (matK) matrix, we used the approach by Lidén et al. (1997) using 1000 replicates, random taxon addition with 20 replicates per replicate, and no branch swapping. Results obtained with this method are very similar to those obtained with other approaches (Mort et al., 2000). With regard to decay analysis, the search for trees three steps longer than the shortest trees was unpractical in all cases. Therefore, the next decay analyses were conducted with the clade-constraint approach as discussed in Morgan (1997). The nrDNA ITS and cpDNA matK data sets were tested for congruence with the partition homogeneity test (Farris et al., 1995) as implemented in PAUP 4.0b4a, before the data sets were combined. The partition homogeneity test was conducted with 1000 replicates, heuristic search option with simple addition sequence, TBR, and MULPARS. RESULTS Nuclear Ribosomal DNA ITS The ITS1 and ITS2 alignment of 60 taxa consisted of 515 positions and contained 199 phylogenetically informative substitutions and 156 phylogenetically informative indels. Mean pairwise distances (as calculated by PAUP) within ingroup varied from 0% (between Echinops viscosus DC. and Echinops spinosissimus Turra) to 31.2% (between Xeranthemum inapertum (L.) Miller and Carlina vulgaris L.). Pairwise distance between ingroup and outgroup varied from 25.7% (between Gerbera kunzeana A. Br. & Aschers. and Carlina lanata L.) to 36.2% (between Xeranthemum inapertum and Gerbera kunzeana). The parsimony analysis yielded eight MPTs of 2073 steps in one island. The strict consensus of all the trees is shown in Fig. 1; the consistency index (CI) excluding uninformative characters was 0.3726, the retention index (RI) was 0.6378, and the homoplasy index (HI) was 0.6274 (Table 3). The strict consensus of the eight MPTs produced from the ITS analysis (Fig. 1) support monophyly of the Cardueae (BS 5 100%, DI 5 20). The subtribe Carli- 56 GARCIA-JACAS ET AL. FIG. 1. Strict consensus tree of the eight most parsimonious trees generated by the ITS matrix. Length: 2073 steps; consistency index excluding uninformative characters: 0.3726; retention index: 0.6378; homoplasy index: 0.6274. Subtribal assignments follow Bremer (1994). CARD, Carduinae; CARL, Carlininae; CENT, Centaureinae; ECHI, Echinopsidinae; OUTG, outgroup. Berardia was unasigned to any subtribe. Ono; Onopordon. ninae was sister to the remaining Cardueae with weak support (BS 5 58%, DI 5 2). This clade had strong support (BS 5 92%, DI 5 3) including the genera Atractylis L., Atractylodes DC., and Carlina. The next clade was formed by Cardopatium corymbosum (L.) Pers. and Cousiniopsis atractyloides (C. Winkl.) Nevski, with weak support (BS 5 57%, DI 5 1). The next branches had no support from the bootstrap or the decay index. However, there were some groups with strong support, such as the clade formed by Alfredia Cass., Synurus Iljin and Onopordon L. (BS 5 92%, DI 5 5). Another clade was formed by the genera Chardinia and Xeranthemum (BS 5 100%, DI 5 15). Next, there were the genera Cynara L. (BS 5 100%, DI 5 15) and Galactites Moench and Ptilostemon Cass. (BS 5 100%, DI 5 36). The clade formed by Echinops and Acantholepis Less. was the next group with very high support (BS 5 100%, DI 5 30). Next was the clade formed by Cirsium, Notobasis Cass., Picnomon Adans., Silybum Adans., Carduus, and Tyrimnus Cass. (BS 5 COMBINED NUCLEAR–CHLOROPLAST DNA PHYLOGENY OF TRIBE CARDUEAE TABLE 3 Comparison of Results from the ITS, matK, and Combined Data Sets Data set ITS matK Combined Total characters Informative characters Number of MPTs Number of steps Consistency index Retention index Homoplasy index Range of divergence, ingroup (%) 515 355 8 2073 0.3726 0.6378 0.6274 0–31.2 1011 126 1887 271 0.5679 0.7384 0.4321 0–6.8 1542 443 4 1915 0.4389 0.6119 0.5611 0.8–13.5 93%, DC 5 8). Finally, there was the clade of the Centaureinae and their sister clades, with very low support (BS 5 64%, DI 5 3). Within this clade, there were the Centaureinae with strong support (BS 5 100%, DI 5 18), and three more clades: the genus Saussurea (BS 5 100%, DI 5 15), the clade with Arctium and Cousinia with weak support (BS 5 62%, DI 5 2), and finally the clade formed by Hyalochaete, Jurinea, Jurinella, and Outreya (BS 5 100%, DI 5 18). Chloroplast DNA matK The matK alignment of 44 taxa consisted of 1011 positions and contained 104 phylogenetically informative substitutions and 22 phylogenetically informative indels. Mean pairwise distances (as calculated by PAUP) within ingroup varied from 0% (between Carlina vulgaris and Carlina lanata) to 6.8% (between Xeranthemum inapertum and Echinops spinosissimus). Pairwise distance between ingroup and outgroup varied from 6.4% (between Outreya carduiformis Jaub. & Spach and Mutisia spinosa R. & P.) to 25.7% (between Gerbera kunzeana and Carlina lanata). The parsimony analysis yielded 1887 MPTs of 271 steps in one island. The strict consensus of all the trees is shown in Fig. 2; the CI excluding uninformative characters was 0.5679, the RI was 0.7384, and the HI was 0.4321 (Table 3). The strict consensus of the 1887 MPTs produced from the matK analysis (Fig. 2) supports monophyly of the Cardueae (BS 5 96%, DI 5 6). There was a polytomy, basal to the weakly supported Carduinae (BS 5 61%, DI 5 2). This grade was formed by the Carlininae (BS 5 82%, DI 5 2) with Atractylis, Atractylodes, Carlina, and Thevenotia DC., the Echinopsidinae with weak support (BS 5 63%, DI 5 2), the genera Berardia and Cardopatium, and, finally, the group formed by Chardinia and Xeranthemum (BS 5 92%, DI 5 4). The rest of the analysis had some groups with strong support. Within the Carduinae, there were the clade formed by Carduus, Notobasis, Picnomon, Tyrimnus, Cirsium, and Silybum (BS 5 87%, DI 5 3), the clade that included Jurinea and Outreya (BS 5 93%, DI 5 4), 57 and the monotypic clade formed by Arctium (BS 5 99%, DI 5 7). However, there was one clade with low support: the clade formed by the genera Alfredia and Synurus (BS 5 61%, DI 5 1). Finally, the clade of the Centaureinae had very low support (BS 5 58%, DI 5 1). Combined nrDNA ITS and cpDNA matK The P value resulting from the partition homogeneity test (P 5 0.245) indicates that data partitions are random, and there is good congruence between nrDNA ITS and cpDNA matK data sets, at a significance threeshold of P 5 0.05 (Farris et al., 1995). The combined ITS–matK alignment of 42 taxa consisted of 1542 positions with 286 phylogenetically informative substitutions and 157 phylogenetically informative indels. Mean pairwise distances (as calculated by PAUP) within ingroup varied from 0.8% (between Carlina lanata and Carlina falcata Svent.) to 13.5% (between Xeranthemum inapertum and Carlina vulgaris). Pairwise distance between ingroup and outgroup varied from 10.1% (between Gerbera kunzeana and Carlina lanata) to 15.2% (between Xeranthemum inapertum and Gerbera kunzeana). The parsimony analysis yielded 4 MPTs of 1915 steps in one island. The strict consensus of all the trees is shown in Fig. 3; the CI excluding uninformative characters was 0.4389, the RI was 0.6119, and the HI was 0.5611 (Table 3). The strict consensus of the four MPTs produced from the combined analysis (Fig. 3) support the monophyly of the Cardueae (BS 5 100%, DI 5 31) as in the ITS and matK single analyses. The subtribe Carlininae was sister to the remaining Cardueae with weak support (BS 5 58%, DI 5 2), as in the ITS analysis. The Carlininae had strong support (BS 5 97%, DI 5 5) including Atractylis, Atractylodes, and Carlina. The next monotypic clade was Cardopatium corymbosum, sister to the rest of the weakly supported Echinopsidinae 1 Carduinae (BS 5 66%, DI 5 2). The clade formed by the Echinopsidinae was the next sister group to the rest of the Cardueae with weak support (BS 5 55%, DI 5 1). The Echinopsidinae was well supported (BS 5 83%, DI 5 4); within this clade there were two branches, one with the genera Echinops and Acantholepis (BS 5 100%, DI 5 34) and the other with the genera Chardinia and Xeranthemum (BS 5 100%, DI 5 16). The next branches had no support from the bootstrap or the decay index, and this part of the tree can be considered a polytomy. The only groups with support were the clade formed by Alfredia, Synurus, and Onopordon (BS 5 97%, DI 5 4) and the clade formed by Carduus, Tyrimnus, Silybum, Cirsium, Notobasis, and Picnomon (BS 5 100%, DI 5 12). Finally, there was the clade of the Centaureinae and related clades, with very low support (BS 5 64%, DI 5 2). Within this clade, 58 GARCIA-JACAS ET AL. FIG. 2. Strict consensus tree of the 1887 most parsimonious trees generated by the matK matrix. Length: 271 steps; consistency index excluding uninformative characters: 0.5679; retention index: 0.7384; homoplasy index: 0.4321. Subtribal assignments follow Bremer (1994). CARD, Carduinae; CARL, Carlininae; CENT, Centaureinae; ECHI, Echinopsidinae; OUTG, outgroup. Berardia was unasigned to any subtribe. there are two groups: the Centaureinae with strong support (BS 5 100%, DI 5 17) and two clades, one with Arctium and Cousinia (BS 5 73%, DI 5 3) and another one with Jurinea and Outreya (BS 5 100%, DI 5 28). DISCUSSION Relative Utility of ITS and matK Regions in the Cardueae Neither the ITS nor the matK alone have fully completed the objectives of our study. ITS analysis (Fig. 1) shows excellent resolution at the generic level; however, it leaves unresolved the relationships between most of the major groups. The consistency index is mediocre (CI 5 0.3726), and the homplasy index is high (HI 5 0.6274). The high homoplasy in the ITS data is confirmed by the high difference between the consistency index and the retention index (Table 3) (Plunkett and Downie, 1999). However, one of the reasons for the high homoplasy could be the size of the matrix (60 taxa), according to Archie (1989). In any case, it is obvious that ITS alone, as expected, cannot fully re- COMBINED NUCLEAR–CHLOROPLAST DNA PHYLOGENY OF TRIBE CARDUEAE 59 FIG. 3. Strict consensus tree of the four most parsimonious trees generated by the combined matK–ITS matrix. Length: 1915 steps; consistency index excluding uninformative characters: 0.4389; retention index: 0.6119; homoplasy index: 0.5611. Boxes with subtribal assignments are from Susanna and Garcia-Jacas (2001). Ono; Onopordon; Jur; Jurinea. solve the relationships between the widely diverging basal groups of the tribe Cardueae. On the other hand, matK analysis shows good resolution at the generic level in many cases, but does not support some indisputable genera, e.g., Echinops (Fig. 2), and the support for another indisputable clade such as the Centaureinae (Fig. 2) is practically null (BS 5 58%, DI 5 1). The homoplasy index and the consistency index are better than the ITS values (CI 5 0.5679; HI 5 0.4321), but the number of informative characters is very low (Table 3). It also leaves the basal groups in an unresolved grade, a rather unexpected result, as matK was supposed to be more adequate for higher taxonomic levels. Instead, the combination of both genomes shows better resolution, both at the generic and the subtribal level. This result is a confirmation of the superiority of the combined approach for phylogenetic studies (Qiu et al., 1999). The Delimitation of the Cardueae Sequence analysis of chloroplast and nuclear DNA, both individually and combined, strongly support the monophyly of the Cardueae [BS 5 100%, DI 5 20 (Fig. 60 GARCIA-JACAS ET AL. 1); BS 5 96%, DI 5 6 (Fig. 2); BS 5 100%, DI 5 31 (Fig. 3)]. The broad concept favored by Bentham (1873), Hoffmann (1894), and Bremer (1994) looks accurate, according to the new molecular evidence. The combined analysis classifies the tribe in, first, a basal grade formed successively by the subtribe Carlininae, the genus Cardopatium, and the subtribe Echinopsidinae (including the Xeranthemum group, usually placed among the Carlininae), and second, the polytomy of the subtribe Carduinae, which includes the subtribe Centaureinae (Fig. 3). Phylogeny suggested by the ITS alone (Fig. 1) coincides in the position of the Carlininae and the genus Cardopatium (with Cousiniopsis), but places Echinopsidinae (without the Xeranthemum group) as nested among the Carduinae. With regard to matK alone, the basal part of the tree is an unresolved grade with Carlininae, Echinopsidinae, Cardopatium, Berardia, and the Xeranthemum group (Fig. 2). We shall discuss each clade separately. The Carlininae. The Carlininae are consistently placed as sister to the rest of the Cardueae by the ITS alone and the combined analysis, with the same low support (BS 5 58%, DI 5 2; Figs. 1 and 3) as that in our previous analysis (Susanna et al., 1995). The analysis of the matK region leaves this part of the tree unresolved (Fig. 2), but this result is compatible with the others. Probably, the tribe Carlininae constitutes the oldest stock of the tribe, which, according with its distribution, would indicate an east Mediterranean origin for the Cardueae (Meusel and Kästner, 1990, 1994; Jäger, 1987). An interesting character that reinforces the basal position of the Carlininae is the presence of true ligules in Atractylis (Dittrich, 1977; even doubting—“almost ligulate”—whether they are true ligules), a character shared with many members of the tribe Mutisieae. According to our results, circumscription of Carlininae must be broadly redefined. The subtribe must include only the genera Carlina, Atractylis, Atractylodes [BS 5 92%, DI 5 3 in the ITS analysis (Fig. 1); BS 5 97%, DI 5 5 in the combined analysis (Fig. 3)], and Thevenotia (BS 5 82%, DI 5 2 in the matK analysis, Fig. 2). With regard to Tugarinovia Iljin, this genus was not included in our analysis, but there are strong doubts on its relationship to the rest of the group, contrary to Dittrich et al. (1987). The exclusion from the Carlininae of the Xeranthemum group and the genera Cardopatium and Cousiniopsis was already pointed out by Petit (1988, 1997), who sent Cardopatium and Cousiniopsis to the Echinopeae (which he treated as a separate tribe) and Xeranthemum to the Carduinae. Homologies of the complex involucral pattern of bracts of Carlina and the receptacular scales of the Xeranthemum group were obviously misinterpreted by previous authors. Our results also suggest interesting considerations on the evolution of some characters in the group. With regard to life forms, Carlina falcata, from the Canary Islands, is an unarmed and shrubby plant, two very unusual features in a group formed mainly by spiny perennial herbs. This led Meusel and Kästner (1994) to suggest that the Canarian species of Carlina were the oldest stock of the genus. However, all the analyses group in a very robust clade Carlina falcata and the widespread European spiny herbs Carlina lanata and C. vulgaris [BS 5 73%, DI 5 2 in the ITS analysis (Fig. 1); BS 5 100%, DI 5 6 in the matK analysis (Fig. 2); BS 5 93%, DI 5 6 in the combined analysis (Fig. 3)]. We must conclude that the shrubby habit is a secondary adaptation in insular species of Carlina, as suggested by Carlquist (1976) for island species of the Asteraceae and recently verified in another Canarian genus of the Cardueae (Cheirolophus Cass.; cf. Susanna et al., 1999). The robust clade of morphologically very different Carlina falcata, C. lanata, and C. vulgaris also indicates that, at least in Canarian plants, extensive changes in habit and morphology, such as the secondary loss of the spines, are not correlated to DNA divergence, a confirmation of the recent origin of some Macaronesian plants (Susanna et al., 1999). It also illustrates the unsuitability of adaptative characters in the systematics of this group. With regard to involucral morphology, two species of Carlina from the west Mediterranean and North Africa, Carlina gummifera (L.) Less. and C. macrophylla (Desf.) DC., were either classified in the genus Atractylis (De Candolle, 1838) or segregated to a different genus, Chamaeleon Cass. The main difference with Carlina is the absence of the inner row of radiant, showy involucral bracts usually associated to Carlina (Petit, 1987; Meusel and Kästner, 1990, 1994). ITS analysis groups both species in a very strong clade (BS 5 100%, DI 5 14), but this clade is nested within Carlina with even higher support (BS 5 100% and DI 5 44; Fig. 1). On the other hand, both matK and the combined analyses include Carlina gummifera in the Carlina clade with the highest support (BS 5 81%, DI 5 2; BS 5 100%, DI 5 49; Figs. 2 and 3, respectively). These results suggest that both species must remain in Carlina, a view strongly supported by the structure of the involucre: Carlina macrophylla and C. gummifera share with the rest of species of Carlina the peculiar and exclusive structure of receptacular scales (Petit, 1987). The lack of the showy bracts could be a primitive character, as C. gummifera is basal to the rest of the genus in the matK and combined analyses (Figs. 2 and 3). Alternatively, the loss of the bracts could be interpreted as a secondary loss as small, unconspicuous bracts are present in Carlina gummifera (Susanna and Garcia-Jacas, 2001). We favor this second hypothesis because secondary reductions are very frequent in the tribe (and they have often contributed COMBINED NUCLEAR–CHLOROPLAST DNA PHYLOGENY OF TRIBE CARDUEAE to the difficulty to the classification in many genera; cf. Vilatersana et al., 2000). The genera Cardopatium and Cousiniopsis. The placement of Cardopatium and Cousiniopsis is a problem that our analysis has failed to solve. These two monotypic genera have been classified usually among the Carlininae (Bentham, 1873; Hoffmann, 1894; Dittrich, 1977; Bremer, 1994). Petit (1997) moved them to the Echinopeae, which he considered a separate tribe, a view not supported by our results. The analysis of the ITS region (Fig. 1) suggests a weak relationship between Cardopatium and Cousiniopsis, as was already pointed out by Bremer (1994). The Echinopsidinae. In view of the strong support of both genomes for the position of Echinops within the tribe Cardueae, the secondary heads of Echinops must be regarded from a rather different point of view. Wagenitz (1976), Dittrich (1977), and Petit (1988, 1997) have considered syncephaly a trait without parallel in the rest of the groups of the Cardueae sensu lato. However, we should rather conclude that the compound inflorescence of Echinops is an extreme adaptation without high systematic relevance. In fact, the trend toward the grouping of small, few-flowered heads in compound inflorescences is common across all the subtribes of the Cardueae: in Carlininae (Atractylis polycephala Coss. has only three flowers per head), in Carduinae (Cousinia congesta Bunge, Cousinia triflora Schrenk, Cirsium congestum Fisch. & C. A. Mey.), and even in the Centaureinae (Centaurea aggregata Fisch. & C. A. Mey.). The epithets of all these taxa are very explicit. The inclusion of Xeranthemum and Chardinia—traditionally classified among the Carlininae—in a robust clade (BS 5 83%, DI 5 4) within the Echinopsidinae, even if only in the combined analysis (Fig. 3), is a rather surprising result: no one had suspected a sister relationship between both groups. Furthermore, this inclusion is not supported by ITS and matK alone. A priori, it is very difficult to connect usually perennial, robust, spiny species of Echinops with predominantly annual, delicate, unarmed plants of the Xeranthemum complex. However, some characters could have given a hint. The only species of the monotypic genus Acantholepis, A. orientalis Less., unanimously considered very closely related to Echinops by its compound second-order inflorescence, is an unarmed, frail annual. Vegetative similarities between Acantholepis and the Xeranthemum group are obvious, especially in the unarmed, entire, linear-lanceolate leaves with velvety indument, but resemblances between the Echinops and Xeranthemum groups are not only vegetative. As is obvious from Dittrich (1977, Figs. 7a and 7b; 1996b), both groups share a similar achene anatomy: they have a parenchymatic pericarp densely covered with multicellular hairs (both characters shared by the Carlini- 61 nae), and the scaly pappus is directly attached to the pericarp. Finally, the most important shared trait was hitherto neglected. The receptable of Xeranthemum and related genera has long tapering scales, longer than florets, which have been interpreted usually as homologous to the receptacular bracts of Carlina (true bracts sensu Petit, 1988; Stuessy and Spooner, 1988) by most authors (Kruse and Meusel, 1972). This was the main reason for the inclusion of the Xeranthemum complex in the Carlininae (Bentham, 1873; Hoffmann, 1894; Dittrich, 1977; Bremer, 1994). However, they can be interpreted in a very different way: the bracts of Xeranthemum could be the invoucral bracts of a primary head, a view supported by the disposition of the scales (see Kruse and Meusel, 1972; Petit, 1987). According to this interpretation, the involucre of Xeranthemum and related genera must be considered a second-order sincephaly, the same as that in Acantholepis and Echinops. With regard to chemistry, Echinops and Xeranthemum are connected by the presence of tiophenes that are unique in all the Cardueae (Wagner, 1977; Bremer, 1994). However, this chemical character has been only partially explored, and no strong conclusions can be drawn yet. The problem of Berardia. The position of the alpine genus Berardia supports its inclusion among the Cardueae. Dittrich (1996a), on the basis of achene anatomy, moved Berardia to the tribe Mutisieae, where it would be aberrant from a biogeographic standpoint: there are no European representatives of this mainly tropical tribe. However, Berardia remains isolated in the Cardueae: it is a monotypic genus from the Alps without any morphologic affinities to any other genus of the tribe (Dittrich, 1977, 1996a; Bremer, 1994). Some striking parallelisms between Berardia and species of Saussurea, Jurinea, and related genera from the mountains of central Asia are only the typical adaptations of plants to high mountains: acaulescence, dense wooly indument, and stout woody rootstock. The subtribe Carduinae. The problem posed by the paraphyletic (if we exclude the Centaureinae) assemblage of the Carduinae (Susanna et al., 1995; Häffner and Hellwig, 1999) remains unresolved. However, some results are relevant for the systematics of the subtribe. The first interesting group is the “Onopordon” group (Susanna and Garcia-Jacas, 2001), formed by Alfredia, Synurus, and Onopordon [BS 5 92%, DI 5 5 in the ITS analysis (Fig. 1); BS 5 97%, DI 5 4 in the combined analysis (Fig. 3)]. Synurus was included among the Centaureinae by Dittrich (1977) and Bremer (1994). Recently, Häffner and Hellwig (1999), on the basis of ITS sequence analysis, moved it to the Carduinae. However, the strongly supported association of Alfre- 62 GARCIA-JACAS ET AL. dia and Synurus (BS 5 100%, DI 5 15 both in the ITS and in the combined analyses; Figs. 1 and 3) has a deeper basis: they share the habit, the form and indument of the leaves, the floral morphology, and the structure of the achenes. Synurus must be considered a mere synonym of Alfredia (Susanna and Garcia-Jacas, 2001). Onopordon and Alfredia share a very uncommon feature in the tribe: the receptacle is naked, honey-combed, and without scales or bristles. The next important natural group within the Carduinae is the “Carduus” group (Susanna and GarciaJacas, 2001), formed by the mainly rose-flowered spiny plants classically grouped under the common name “thistles.” This group includes two large genera (Carduus and Cirsium) and many other smaller genera represented here by Cynara, Galactites Moench, Notobasis, Picnomon, Ptilostemon, Silybum, and Tyrimnus. The group formed by Carduus, Cirsium, Notobasis, Picnomon, Silybum, and Tyrimnus was especially strongly supported by all the analyses [BS 5 93%, DI 5 8 in the ITS analysis (Fig. 1); BS 5 87%, DI 5 3 in the matK analysis (Fig. 2); BS 5 100%, DI 5 12 in the combined analysis (Fig. 3)]. Similarities in morphology, habit, and biogeography among all these genera are great, and the group can be considered a natural one. Häffner and Hellwig (1999), whose ITS analysis focused on this complex, arrived at the same conclusion. However, our analysis includes only a few representatives of the two largest genera, Carduus and Cirsium, and we cannot draw further conclusions on their delimitation, which is problematic according to Häffner and Hellwig (1999). However, there is a significant difference between the results of our analysis in Cirsium and the results obtained in another very large genus, Cousinia. Species of Cirsium are grouped in a clade with low support, whereas species of Cousinia form a very strong clade (Fig. 1). Sequence divergence among species of Cousinia is lower than that in Cirsium, which indicates that Cousinia is a more recent genus with an explosive diversification (800 species!). This is concordant with the geographic distribution: Cirsium has an Eurasian, African, and North American distribution, whereas Cousinia has a much more restricted distribution centered in the Irano-Turanian region (Kazmi, 1963; Dittrich, 1977; Häffner and Hellwig, 1999; Susanna and Garcia-Jacas, 2001). Another clade is formed by the Centaureinae and some related groups, among which the sister group to the Centaureinae should be looked for, as was already pointed out by Susanna et al. (1995), Petit (1997), and Häffner and Hellwig (1999). As the Centaureinae have been the object of a deeper analysis elsewhere (GarciaJacas et al., 2001), we shall comment only that, as in previous results, they form a robust monophyletic clade. The groups that are more closely related to the Cen- taureinae are the clade which includes the genera Arctium and Cousinia, the clade formed by Jurinea and Outreya (Fig. 3), plus Hyalochaete, Jurinella, and Saussurea in the ITS tree (Fig. 1). However, the relationships of these genera to the subtribe Centaureinae are only very weakly supported by the ITS and the combined analysis and not supported at all by the matK analysis. Other conclusions can be drawn from this clade, such as the association of Arctium and Cousinia in a clade with only moderate support [BS 5 62%, DI 5 2 in the ITS analysis (Fig. 1); BS 5 73%, DI 5 3 in the combined analysis (Fig. 3)] and not supported by the matK sequences (Fig. 2). Relationships between both genera have been signaled from old (see Duistermaat, 1996, 1997; Petit, 1997). However, the problem of Arctium “getting entangled to Cousinia” (Duistermaat, 1997) is easy to solve: we only have to move the “Arctioid” Cousinia species to Arctium. The last clade from the combined analysis that we shall comment on is formed by Outreya and Jurinea (BS 5 100%, DI 5 28; Fig. 3). In the ITS tree, this clade includes also Hyalochaete and Jurinella with very high support (BS 5 100%, DI 5 18; Fig. 1). The strong support for this group confirms the suggestions by Susanna and Garcia-Jacas (2001): on strict morphological grounds, Hyalochaete, Jurinella, and Outreya cannot be segregated from Jurinea (significantly, all three were originally described within Jurinea). The Subtribal Classification: A Recapitulation Some final conclusions can be drawn on the subtribal classification. Support for the basal branches is very low or nonexistent in all the analyses (Figs. 1–3). Moreover, two genera, Cardopatium and Cousiniopsis, cannot be included in any subtribe, and the positions of Staehelina (Fig. 1) and Berardia remain uncertain (Figs. 1–3). These results have forced us to be very cautious when suggesting a subtribal classification in a recent synthesis of the tribe (Susanna and GarciaJacas, 2001). Another problem is that the segregation of the Centaureinae with subtribal rank leaves the subtribe Carduinae paraphyletic, as has been repeatedly pointed out (Bremer, 1994; Susanna et al., 1995; Petit, 1997; Häffner and Hellwig, 1999). However, the cladistically correct solution has a major drawback: if we made a natural Carduinae 1 Centaureinae, the resulting subtribe would include the genera Centaurea (400 spp.), Cousinia (800 spp.), Saussurea (more than 300 spp.), Jurinea (100 spp.), Carduus (90 spp.), and Cirsium (250 spp.). This totals nearly 2000 species of the ca. 2500 species of the Cardueae, 80% of the total. Probably, the Centaureinae 1 Carduinae are a natural group (even if unsupported by molecular data), but to keep them united in a subtribe would be extremely unpractical. COMBINED NUCLEAR–CHLOROPLAST DNA PHYLOGENY OF TRIBE CARDUEAE Finally, at the present state of knowledge, we think that the most conservative approach is the only practical solution: keep the four classical subtribes, with the broad modifications of delimitation suggested by Susanna and Garcia-Jacas (2001) and reflected in Fig. 3. ACKNOWLEDGMENTS Financial support from the Dirección General de Enseñanza Superior, Spain, Projects PB 93/0032 and PB 97/1134, and grant 1999SGR0032, generalitat de Catalonia, are gratefully acknowledged. We thank N. Gabrielyan, L. Kapustina, F. Khassanov, and V. Mozaffarian for their assistance in the field work and help in the identification of some difficult materials. The collaboration of the botanical gardens listed in Table 2 is also acknowledged. We also thank Vicki A. Funk, Smithsonian Institution and an anonymous reviewer for many helpful commentaries for improving the text. REFERENCES Archie, J. W. (1989). A randomization test for phylogenetic information in systematic data. Syst. Zool. 38: 239 –252. Bain, J. F., and Jansen, R. K. (1995). A phylogenetic analysis of the aureoid Senecio (Asteraceae) complex based on ITS sequence data. Pl. Syst. Evol. 195: 209 –219. Bentham, G. (1873). Compositae. In “Genera Plantarum” (G. Bentham and J. D. Hooker, Eds.), pp. 162–533. Lovell Reeve & Co., London. Bremer, K. (1987). Tribal interrelationships of the Asteraceae. Cladistics 3: 210 –253. Bremer, K. (1988). The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 795– 803. Bremer, K. (1994). “Asteraceae: Cladistics and Classification,” Timber Press, Portland. Brochmann, C., Xiang, Q.-Y., Brunsfeld, S. J., Soltis, D. E., and Soltis, P. S. (1998). Molecular evidence for polyploid origins in Saxifraga (Saxifragaceae): The narrow arctic endemic S. svalbardensis and its widespread allies. Am. J. Bot. 85: 135–143. Carlquist, S. (1976). Tribal interrelationships and phylogeny of the Asteraceae. Aliso 8: 465– 492. Cassini, H. (1819). [Different articles]. Dictionnaire de Sciences Naturelles, Paris. In “Cassini on Compositae” (R. King and H. W. Dawson, Eds., 1975). Oriole, New York. De Candolle, A. P. (1838). “Prodromus Systematis Naturalis Regni Vegetabilis 6,” Treuttel et Würtz, Paris. Denda, T., Watanabe, K., Kosuge, K., Yahara, T., and Ito, M. (1999). Molecular phylogeny of Brachycome (Asteraceae). Pl. Syst. Evol. 217: 299 –311. Dittrich, M. (1977). Cynareae—Systematic review. In “The Biology and Chemistry of the Compositae” (V. H. Heywood, J. B. Harborne, and B. L. Turner, Eds.), pp. 999 –1015. Academic Press, London. Dittrich, M. (1996a). Bemerkungen zur tribuszugehörigkeit von Berardia. Ann. Naturhist. Mus. Wien 99(b): 329 –342. Dittrich, M. (1996b). Die bedeutung morphologischer und anatomischer Achänen-Merkmale für die Systematik der tribus Echinopeae Cass. und Carlineae Cass. Boissiera 51: 9 –102. Dittrich, M., Kästner, A., and Meusel, H. (1987). Tugarinovia—Eine mongolische Carlineen-Sippe? Bot. Jahrb. Syst. 108: 167–186. Donoghue, M. J., Olmstead, R. G., Smith, J. F., and Palmer, J. D. (1992). Phylogenetic relationships of Dipsacales based on rbcL sequences. Ann. Missouri Bot. Gard. 79: 333–345. 63 Doyle, J. J. (1992). Gene trees and species trees: Molecular systematics as one-character taxonomy. Syst. Bot. 17: 144 –163. Doyle, J. J., and Doyle, J. L. (1987). A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochem. Bull. 19: 11–15. Doyle, J. J., and Doyle, J. L. (1988). Natural interspecific hybridization in eastern North American Claytonia. Am. J. Bot. 75: 1238 – 1246. Duistermaat, H. (1996). Monograph of Arctium L. (Asteraceae): Generic delimitation (including Cousinia Cass. p.p.), revision of the species, pollen morphology and hybrids. Gorteria 3: 1–143. Duistermaat, H. (1997). Arctium getting entangled to Cousinia (Asteraceae: Cardueae). Bocconea 5: 685– 689. Farris, J. S., Källersjö, M., Kluge, A. G., and Bult, C. (1995). Testing significance of incongruence. Cladistics 10: 315–319. Felsenstein, J. (1985). Confidence limits on phylogenies: An approach using the bootstrap. Evolution 39: 783–791. Gadek, P. A., Wilson, P. G., and Quinn, C. J. (1996). Phylogenetic reconstruction in Myrtaceae using matK, with particular reference to the position of Psiloxylon and Heteropyxis. Aust. Syst. Bot. 9: 283–290. Garcia-Jacas, N., Susanna, A., Garnatje, T., and Vilatersana, R. (2001). Generic delimitation and phylogeny of the subtribe Centaureinae (Asteraceae): A combined nuclear and chloroplast DNA analysis. Ann. Bot. 87: 503–515. Garcia-Jacas, N., Susanna, A., Mozaffarian, V., and Ilarslan, R. (2000). The natural delimitation of Centaurea (Asteraceae: Cardueae): ITS sequence analysis of the Centaurea jacea group. Pl. Syst. Evol. 223: 185–199. Häffner, E., and Hellwig, F. H. (1999). Phylogeny of the tribe Cardueae (Compositae) with emphasis on the subtribe Carduinae: An analysis based on ITS sequence data. Willdenowia 29: 27–39. Hoffmann, O. (1894). Compositae. In “Die Natürlichen Pflanzenfamilien” (A. Engler and K. Prantl, Eds.), pp. 324 –333. Engelmann, Leipzig. Jäger, E. J. (1987). Arealkarten der Asteraceen-tribus als grunlage der örogeographischen Sippencharakteristik. Bot. Jahrb. Syst. 108: 481– 497. Jansen, R. K., Holsinger, K. E., Michaels, H. J., and Palmer, J. D. (1990). The phylogenetic analysis of chloroplast DNA restriction site data at higher taxonomic levels: An example from the Asteraceae. Evolution 44: 2089 –2105. Jansen, R. K., Holsinger, K. E., and Palmer, J. D. (1991). Phylogeny and character evolution in the Asteraceae based on chloroplast DNA restriction site mapping. Syst. Bot. 16: 98 –115. Johnson, L. A., and Soltis, D. E. (1994). matK DNA sequences and phylogenetic reconstruction in Saxifragaceae s. str. Syst. Bot. 19: 143–156. Johnson, L. A., and Soltis, D. E. (1995). Phylogenetic inference in Saxifragaceae sensu stricto and Gilia (Polemoniaceae) using matK sequences. Ann. Missouri Bot. Gard. 82: 149 –175. Karis, P. O., Källersjö, M., and Bremer, K. (1992). Phylogenetic analysis of the Cichorioideae (Asteraceae), with emphasis on the Multisieae. Ann. Missouri Bot. Gard. 79: 416 – 427. Kazmi, S. M. A. (1963). Revision der Gattung Carduus (Compositae). Mitt. Bot. Staatssamml. München 5: 139 –198. Kim, K.-J., Jansen, R. K., Wallace, R. S., Michaels, H. J., and Palmer, J. D. (1992). Phylogenetic implications of rbcL sequence variation in the Asteraceae. Ann. Missouri Bot. Gard. 79: 428 – 445. Kruse, J., and Meusel, H. (1972). Zur Blattfolge und Blattbildung einiger Cynareen-Gattungen. 2. Acantholepis, Xeranthemum, Chardinia, Siebera, Amphoricarpos. Feddes Repert. Spec. Nov. Regni Veg. 83: 289 –308. Lidén, M., Fukuhara, T., Rylander, J., and Oxelman, B. (1997). 64 GARCIA-JACAS ET AL. Phylogeny and classification of Fumariaceae, with emphasis on Dicentra s. I. based on the plastid gene rps16 intron. Pl. Syst. Evol. 206: 411– 420. Maddison, D. R. (1991). The discovery and importance of multiple islands of most parsimonious trees. Syst. Zool. 40: 315–328. Meusel, H., and Kästner, A. (1990). Lebengeschichte der Gold- and Silberdisteln: Monographie der mediterran–mitteleuropäischen Compositen-Gattung Carlina. Band I. Österr. Akad. Wiss. Math.Naturwiss. Kl. Denkschr. 127. Meusel, H., and Kästner, A. (1994). Lebengeschichte der Gold- and Silberdisteln: Monographie der mediterran–mitteleuropäischen Compositen-Gattung Carlina. Band II. Österr. Akad. Wiss. Math.Naturwiss. Kl. Denkschr. 128. Meusel, H., and Kohler, E. (1960). Die Blattbidung von Carlina vulgaris L. Bot. Jahrb. Syst. 79: 192–207. Morgan, D. R. (1997). Decay analysis of large sets of phylogenetic data. Taxon 46: 509 –517. Mort, M. E., Soltis, P. S., Soltis, D. E., and Mabry, M. L. (2000). Comparison of three methods for estimating internal support on phylogenetic trees. Syst. Biol. 49: 160 –171. Nickrent, D. L., Schuette, K. P., and Starr, E. M. (1994). A molecular phylogeny of Arceuthobium (Viscaceae) based on nuclear ribosomal DNA internal transcribed spacer sequences. Am. J. Bot. 81: 1149 –1160. Petit, D. P. (1987). Révision des genres Atractylis, Carlina et Chamaeleon (Compositae, Cardueae) au Maroc. Bull. Mus. Natl. Hist. Nat. B Adansonia 4: 407– 440. Petit, D. P. (1988). Le genre Echinops L. (Compositae, Cardueae). 1. Position phylétique et interprétation de l’incapitulescence. Candollea 43: 467– 481. Petit, D. P. (1997). Generic interrelationships of the Cardueae (Compositae): A cladistic analysis of morphological data. Pl. Syst. Evol. 207: 173–203. Plunkett, G. M., and Downie, S. R. (1999). Major lineages within Apiaceae subfamily Apioideae: A comparison of chloroplast restriction site and DNA sequence data. Am. J. Bot. 86: 1014 –1026. Plunkett, G. M., Soltis, D. E., and Soltis, P. S. (1996). Evolutionary patterns in Apiaceae: Inferences based on matK sequence data. Syst. Bot. 21: 477– 495. Qiu, Y.-L., Lee, J., Bernasconi-Quadroni, F., Soltis, D. E., Soltis, P. S., Zanis, M., Zimmer, E. A., Chen, Z., Savolainen, V., and Chase, M. W. (1999). The earliest angiosperms: Evidence from mitochondrial, plastid and nuclear genomes. Nature 402: 404 – 407. Samuel, R. A., Bachmair, A., Jobst, J., and Ehrendorfer, F. (1998). ITS sequences from nuclear rDNA suggest unexpected phylogenetic relationships between Euro-Mediterranean, East Asiatic and North American taxa of Quercus (Fagaceae). Pl. Syst. Evol. 211: 129 –139. Soltis, D. E., Kuzoff, R. K., Conti, E., Gornall, R., and Ferguson, K. (1996). matK and rbcL gene sequence data indicate that Saxifraga (Saxifragaceae) is polyphyletic. Am. J. Bot. 83: 371–382. Soltis, D. E., Soltis, P. S., Collier, T. G., and Edgerton, M. L. (1991). The Heuchera group (Saxifragaceae): Evidence for chloroplast transfer and paraphyly. Am. J. Bot. 78: 1091–1112. Stuessy, T. F., and Spooner, D. M. (1988). The adaptative and phylogenetic significance of receptacular bracts in the Compositae. Taxon 37: 114 –126. Sun, Y., Skinner, D. Z., Liang, G. H., and Hulbert, S. H. (1994). Phylogenetic analysis of Sorghum and related taxa using internal transcribed spacers of nuclear ribosomal DNA. Theor. Appl. Genet. 89: 26 –32. Susanna, A., and Garcia-Jacas, N. (2001). Tribus Cardueae. In “The Families and Genera of Vascular Plants: Compositae” (K. Kubitzki, Ed.), Springer-Verlag, Heidelberg, in press. Susanna, A., Garcia-Jacas, N., Soltis, D. E., and Soltis, P. S. (1995). Phylogenetic relationships in tribe Cardueae (Asteraceae) based on ITS sequences. Am. J. Bot. 82: 1056 –1068. Susanna, A., Garnatje, T., and Garcia-Jacas, N. (1999). Molecular phylogeny of Cheirolophus (Asteraceae: Cardueae–Centaureinae) based on ITS sequences of nuclear ribosomal DNA. Pl. Syst. Evol. 214: 147–160. Swofford, D. L. (1999). “PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4”. Sinauer, Sunderland, MA. Swofford, D. L., and Olsen, G. J. (1990). Phylogeny reconstruction. In “Molecular Systematics” (D. Hillis and C. Moritz, Eds.), pp. 411– 501. Sinauer, Sunderland, MA. Thompson, J. D., Gibson, T. J., Plewniak, F., Jeanmougin, F., and Higgins, D. G. (1997). The ClustalX windows interface: Flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 24: 4876 – 4882. Vilatersana, R., Susanna, A., Garcia-Jacas, N., and Garnatje, T. (2000). Generic delimitation and phylogeny of the Carduncellus– Carthamus complex (Asteraceae) based on ITS sequences. Pl. Syst. Evol. 221: 89 –105. Wagenitz, G. (1976). Systematics and phylogeny of the Compositae (Asteraceae). Pl. Syst. Evol. 125: 29 – 46. Wagner, H. (1977). Cynareae-chemical review. In “The Biology and Chemistry of the Compositae” (V. H. Heywood, J. B. Harborne, and B. L. Turner, Eds.), pp. 1017–1038. Academic Press, London. White, T. J., Bruns, T., Lee, S., and Taylor, J. (1990). Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In “PCR Protocols: A Guide to Methods and Applications” (M. Innis, D. Gelfand, J. Sninsky, and T. White, Eds.), pp. 315– 322. Academic Press, San Diego. Xiang, Q.-Y., Soltis, D. E., and Soltis, P. S. (1998). Phylogenetic relationships of Cornaceae and close relatives inferred from matK and rbcL sequences. Am. J. Bot. 85: 285–297.