Plant Syst. Evol. 221:89-105 (2000) Systematics and Evolution © Springer-Verlag 2000 Printed in Austria

Generic delimitation and phylogeny of the - complex () based on ITS sequences

R. Vilatersana, A. Susanna, N. Garcia-Jacas, and T. Garnatje

Institut Botfinic de Barcelona (CSIC), Barcelona,

Received August 18, 1999 Accepted October 21, 1999

Abstract. Within the Mediterranean complex 50 species (Dittrich 1977). These two genera Carduncellus-Carthamus, taxonomic classification share a native distribution ranging from the has proven problematic. Numerous attempts to Iberian Peninsula and to West , clarify the relative systematic boundaries have but the core of Carduncellus centers in the included splitting Carduncellus and Cartharnus western Mediterranean (Iberian Peninsula and into several genera, but none of these proposed , eastward to and Egypt), classifications have been generally accepted. For a comprehensive resolution of the relationships while Carthamus centers in the eastern Medi- within this group, we used sequences of the terranean Basin, except for a small section Internal Transcribed Spacers (ITS) of nuclear endemic to southern Spain and Morocco ribosomal DNA. The results indicate that the (Fig. 1). complex should be classified into four genera: In addition, some species such as Car- Carduncellus, Carthamus, Ferneniasia and Phonus. tharnus lanatus L., probably of hybrid origin, The relationship between the western group and Sibth. & Sm. are (Carduncellus, Femeniasia and Phonus) and the widely naturalized as noxious weeds in the eastern Carthamus are not resolved by western Mediterranean region, as well as in analysis of ITS sequences, but the two groups Mediterranean climatic regions of Argentina, are probably not close relatives. The ITS classifi- Australia, California, and South Africa (Ashri cations corresponded with biogeography and less and Knowles 1960, Hanelt 1963, Estilai and with morphological characters, which have also been the main source of confusion in traditional Knowles 1978). The wild origin is unknown for classifications. Most of the unusual morphological the cultivated species Carthamus tinctorius L. features in the Carduncellus-Carthamus complex (), although it is widely grown as an appear to be reversals to ancestral character important source of oil in subtropical coun- states. tries (Hanelt 1963, 1976) and as a substitute for saffron. Key words: Carduncellus, Carthamus, Femeniasia, The subtribe position of this complex Phonus, phylogeny, ITS sequences. within Cardueae is problematic and many genera are not yet clearly delineated. The In the Mediterranean basin, the complex of existing confusion was compounded by the two Asteraceae-Cardueae genera, Carduncellus recent incorporation of a new genus, Femeni- Adanson and Carthamus L., comprises about asia Susanna (Susanna et al. 1995, Susanna 90 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus

Fig. 1. Geographic distribution of the genera of the complex. Light grey: Carduncellus. Dark grey: Carthamus (excluded the cosmopolitan section Atractylis). Diamonds: Phonus (=Carthamus sect. Thamnacanthus). Inverted triangle: Femeniasia and Vilatersana 1996, Wagenitz and Hellwig Carduinae, but highly unusual in Centaurei- 1996), into the complex. nae. Secondly, the caducous single pappus Subtribal placement. Cassini (1819) includ- found in some species of Carduncellus (Cassini ed Carduncellus and part of Carthamus in the 1819, Dittrich 1969) and Femeniasia (Susanna subtribe Carduinae, while classifying Car- 1988) is more representative of the subtribe thamus sensu stricto in the Centaureinae. Later Carduinae than the tribe Centaureinae. How- classifications proposed a new subtribe Carth- ever, the spiny habit could be easily attributed aminae for Carduncellus and Carthamus, posi- to secondary adaptation against predators; and tioned between the two major subtribes of the the reduction from a double to a single pappus Cardueae: Carduinae and Centaureinae (De occurs frequently in the subtribe Centaureinae, Candolle 1838, Nyman 1878-1890). Subse- thus it provides minimal systematic value quently, Godron (1852) and Battandier (Dittrich 1968, 1969; Wagenitz and Hellwig (1890) reverted to the classification of Cassini 1996). For this reason, current general opinion (1819) and moved Carduncellus to Carduinae favors classifying the entire complex in Cen- and Carthamus to Centaureinae. After that, taureinae (Dittrich 1977, Bremer 1994, Susan- Bentham (1873) classified both genera in the na et al. 1995, Wagenitz and Hellwig 1996). subtribe Centaureinae, with concurrence by Generic delineation. At the generic level, Hoffmann (1894). Similarly, Femeniasia was the main problem in the complex is the described and classified in Carduinae (Susanna delineation of Carduncellus and Carthamus. 1988), and was later transferred to Centaur- Unfortunately, numerous attempts to solve einae (Bremer 1994, Susanna et al. 1995, this central problem by splitting both genera Susanna and Vilatersana 1996, Wagenitz and have created even greater quandary. For Hellwig 1996). example, the systematic position is not clear Two peculiar morphological traits are pri- for two genera split from Carthamus (Kent- marily responsible for this fluctuating subtribal rophyllum Necker and Phonus Hill) and placement. First, most species in the complex another genus (Larnottea Pomel) segregated have spiny leaves, a frequent characteristic in from Carduncellus. Also problematic are two R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 91

Table 1. Different generic classifications of the complex Cassini (1819) De Candolle (1838) Pomel (1874) Kentrophyllum Kentrophyllum Kentrophyllum sect. Atraxyle Onobroma sect. Odontognathia sect. Euonobroma sect. Thamnacantha sect. Cirsiastrum Car duncellus Carduncellus Carduncellus Onobroma Carthamus Carthamus Durandoa Bentham (1873) Battandier (1890) Hanelt (1963) Boissier (1875) Hoffmann (1894) Kentrophyllum Carthamus sect. Durandoa sect. Atractylis Carduncellus sect. Atraxyle sect. Carthamus sect. Lepidopappus Carthamus Carthamus sect. Odontognathius sect. Thamnacanthus Car duncellus sect. Cyanoidei Carduncellus sect. Cirsiastri sect. Phalolepides sect. Carduncellus L6pez Gonzfilez (1990) Carthamus sect. Atractylis sect. Carthamus sect. Odontognathius Phonus Lamottea CarduncelIus

rare North African endemics, Carduncellus Carduncellus and Carthamus. The line mareoticus (Del.) Hanelt and Carduncellus separating Carduncellus from Carthamus has fruticosus (Maire) Hanelt, which have been fluctuated widely depending on the characters classified in four different genera: Carduncellus, chosen. L6pez Gonzfilez (1990) lists three Carthamus, Femeniasia and Phonus. These and potential criteria for differentiating between other conflicting generic classifications are Carthamus and Carduncellus: structures of the reviewed in Table 1. Finally, there is the pappus and the pericarp and morphology of question of where the monotypic genus Fe- the middle bracts of the capitulum. meniasia belongs in the complex. By the time it Two main types of pappus are found in the was first described on the basis of a single rare Carduncellus-Carthamus complex. The more endemic species (Susanna 1988), the relation- common type is a double pappus, composed of ship of this genus to the Carduncellus-Car- two whorls, a very short convergent inner thamus group was unsuspected and it was first pappus and an outer pluriseriate pappus with classified within the Carduinae. longer, unequal, linear or paleaceous setae To minimize confusion, this paper follows (Dittrich 1968, 1969). This type of pappus can the nomenclature of Hanelt (1963). be persistent or deciduous and is frequently 92 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus absent in the peripheral cypselas. This pappus (1874). He described it as Durandoa Pomel, type is present in Carthamus and in a group of but this illegitimate name has been corrected to Carduncellus species that has been segregated Phonus (cf. Ldpez Gonz/tlez 1990). Species of to a different genus (Lamottea) on the basis of this section of Cartharnus possess some con- this character. The second type of pappus is flicting characters that deviate from the rest of characterized by simple, more or less equal the genus. They are subshrubs, whereas the linear setae, deciduous as a single unit by rest of the species of Carthamus are annual. means of a basal ring. This pappus is typical of The cypselas resemble those of Carduncellus Carduncellus sensu stricto (excluding Lamo- by having a semideciduous pappus, and the ttea). Pappus characteristics are the basis of pericarp is undifferentiated (Dittrich 1969). classifications proposed by De Candolle (1838), Their geographic distribution is western Medi- Boissier (1873), Bentham (1873), Battandier terranean with two narrow endemics centered (1890), and Ldpez Gonzfilez (1990) (Table 1). in southern Spain and Morocco (Fig. 1). It was primarily on the basis of pericarp However, species of Phonus lack the append- structure that some species were segregated ages of the middle bracts typical of Card- from the complex into a new genus. Members uncellus, a fact that led Hanelt (1963, 1976) to of this genus (Phonus) are characterized by an retain them in Carthamus, while recognizing undifferentiated pericarp (Dittrich 1969). that this classification was a conservative Ldpez Gonzfilez (1990) concurred that this compromise. character strongly supported segregation of Kentrophyllum. On the basis of some dif- Phonus as a new genus. ferences in vegetative characters and in the The third important character is the pres- cypselas, Cassini (1819) and De Candolle ence of semicircular, cucullate, brown scariose (1838) suggested that Car thamus should include appendages in the middle involucral bracts, a only the Carthamus tinctorius group, and that character typical of Carduncellus in the present the remaining species be classified in a separate circumscription (Hanelt 1963, 1976). Ldpez genus Kentrophyllum. Chromosome numbers Gonzfilez (1990) claims that this character is support this segregation, i.e., species of Car- irrelevant for the classification because these thamus sect. Carthamus have only the base appendages are missing in some species usually chromosome number x = 12, whereas species of included in Carduncellus. the rest of the sections show a complex dysploid Lamottea and Phonus. Pomel (1860) series from x = 12 to x = 10 (Estilai and Know- segregated some species of Carduncellus les 1976, 1978). Experimental evidence demon- (C. caeruleus (L.) C. Presl., C. calvus Boiss. & strates that although hybrids are frequent and Reut., C. dianius Webb and C. helenioides fertile within Kentrophyllum, they are sterile (Desf.) Hanelt) into a new genus Lamottea, and difficult to obtain between Carthamus and mostly on the basis of carpologic characters. Kentrophyllurn (Ashri and Knowles 1960; Although subsequent authors (Boissier 1873, Schank and Knowles 1964; Estilai and Knowles Battandier 1890, Jahandiez and Maire 1934, 1976, 1978; Estilai 1977). Qu~zel and Santa 1963) moved the species of Femeniasia. This monotypic genus was Lamottea to Carthamus on the basis of pappus described on the basis of balearica characteristics, L6pez Gonz~.lez (1990) trans- Rodriguez Femenias, a narrow endemic from ferred them back to Lamottea. Because most the island of Minorca (Balearic Islands, Fig. 1). species in this genus have bracts with cochle- Susanna (1988) pointed out that the cypselas of ariform appendages, Hanelt (1963) and Rivas the species were aberrant in Centaurea and Goday and Rivas Martinez (1967) moved suggested that it could belong to the subtribe them back to Carduncellus. Carduinae. Subsequently, Bremer (1994), On the other hand, Phonus was referred to Susanna et al. (1995), Susanna and Vilatersana Carthamus sect. Thamnacanthus by Pomel (1996), and Wagenitz and Hellwig (1996) R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 93 suggested that it should be classified in the 1995). This DNA region has already been used Centaureinae, close to the Carduncellus-Car- in the phylogenetic studies of the Centaureinae thamus complex. Thus, uncertainties remain in (Susanna et al. 1995, 1999). its generic status and relationships. Our research analyzed the sequences of a Carduncellus fruticosus and Carduncellus comprehensive representation of genera within mareoticus. These two North African species the complex and several more distantly related are problematical. Both were described as species, with the following three objectives: 1) Carthamus, but have rudimentary appendicu- to examine the adequacy of molecular evidence late middle bracts and Hanelt (1963) trans- for differentiating the six genera described in ferred them to Carduncellus. But because they the complex (CardunceIlus, Carthamus, Fe- also have a semideciduous pappus and an meniasia, Kentrophyllum, Lamottea and Pho- undifferentiated pericarp, L6pez GonzAlez nus), 2) infer their phylogenetic relationships (1990) classified them as Phonus. On the other based on DNA sequences, and 3) to explore hand, Carduncellus fruticosus was recently the correspondence between morphological changed to Femeniasia fruticosa (Maire) Petit characters commonly used in classification of (cf. Petit 1997). the group and the new molecular evidence. Two opposite approaches have been pro- posed as comprehensive solutions to the pre- ceding taxonomic confusion. Dittrich (1969) Material and methods suggested returning to the old classification by Both previously published and new sequences were Linn6 (1753), with all the species of both used in the analysis. Sequences of Carduncellus genera grouped in a single genus Carthamus. mitissimus (L.) DC., L. and Alternatively, L6pez Gonzfilez (1990) pro- Femeniasia balearica Susanna (from the ingroup) posed splitting Carduncellus and Carthamus and Acroptilon repens (L.) DC., Cheirolophus ar- into four genera: Carthamus, Carduncellus, but(folius (Svent.) Kunkel, Centaurea africana Lamottea and Phonus (Table 1). As demon- Lain., C. americana Nutt., C. clementei Boiss., C. strated by the history of taxonomic classifica- dealbata Willd., and C. toletana Boiss. & Reuter (from the outgroup) were obtained from previous tions and reclassifications, it is obvious that studies (Susanna et al. 1995). Twenty-eight new neither the characters of cypselas nor those of species were examined, including representatives of bract appendages, alone or in combination, Carthamus, Carduncellus, Lamottea, and Phonus. adequately provide a means of cleanly sepa- The origin of these samples, the herbaria where rating the two genera, an observation also vouchers were deposited, and the GenBank acces- made by Hanelt (1963), Dittrich (1969) and sion numbers are detailed in Table 2. For represen- L@ez Gonzfilez (1990). tatives of Carthamus, we followed the sectional The molecular approach may provide a classification of Hanelt (1963) (Table 1), including new option for clarifying systematic relation- representatives of each section. The outgroup was ships within the Carduncellus-Carthamus com- chosen among Centaureinae from our previous ITS plex because structural or morphological sequence analysis (Susanna et al. 1995). The out- characters have repeteadly proven to be con- group included two species of the more advanced fusing. Comparison of sequences of the Inter- clade according to previous analysis (Susanna et al. 1995), Centaurea clementei and C. toletana, and five nal Transcribed Spacers (ITS-1 and 1TS-2) of representatives of more primitive groups (Acropti- nuclear ribosomal DNA has demonstrated its lon repens, Centaurea africana, C. americana, C. suitability for unraveling systematic problems dealbata and Cheirolophus arbutifolius). in many genera of Asteraceae (Antennaria, cf. In a preliminary analysis, we included the ITS Bayer et al. 1996; Calycadenia, cf. Baldwin sequences of three species of suspected hybrid 1993; Dendroseris, cf. Sang et al. 1995; Krigia, origin (Hanelt 1963, Ldpez Gonzfilez 1990): Car- cf. Kim and Jansen 1994; Robinsonia, cf. Sang thamus creticus L., C. lanatus L. and C. turk- et al. 1995; and Senecio, cf. Bain and Jansen estanicus Popov. This analysis resulted in trees with 94 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus

Table 2. Origin of the studied material (vouchers in the herbarium BC). GenBank accession numbers are bracketed Species Voucher and GenBank accession # Carduncellus araneosus SPAIN, TOLEDO: between Huerta de Vatdecarfibanos and Cabafias Boiss. & Reut. de Yepes, Garcia-Jacas, Susanna 1603 and Vilatersana, 22-06-1996 [AF140464, AF140465]. Carduncellus calvus MOROCCO, : S side of Djebel Bou Messoud, AbdeIkader, Molero, J. Montserrat 3642, Pallds, Vicens and Veny, 13-06-1994 [AF140438, AF140439]. Carduncellus caeruleus SPAIN, M~_LAGA: road C-344 between Coin and Tolox, Garcia-Jacas, Susanna 1610 and Vilatersana, 22-06-1996 [AF140442, AF140443]. Carduncellus cuatrecasasii SPAIN, JAI~N: Sierra de Mfigina, between Mancha Real and Torres, Garcia-Jacas, Susanna 1608 and Vilatersana, 22-06-1996 [AF140430, AF140431]. Carduncellus dianius SPAIN, VALENCIA: Cape San Antonio near Jfivea, Garcia-Jacas, Susanna 1479 and Vilatersana, 17-06-1995 [AF140440, AF140441]. Carduncellus duvauxii MOROCCO, AL HOCEIMA: 8 km S of Tafraoute, G6miz, 23-04-1994 Batt. & Trab. [AF140434, AF140435]. Carduncellus eriocephalus MOROCCO, BOUARFA: 100 km from Bouanane to Bouarfa, Garnatje, Boiss. Susanna 1785 and Vilatersana, 16-06-1997[AF140470, AF140471]. Carduncellus fruticosus MOROCCO, OUARZAZATE: gorges of the river Todgha, Benedi, G. Montserrat and J. M. Montserrat 2407, 5-06-1989 [AF140450, AF140451]. Carduncellus helenioides MOROCCO, FES: Oued Zloul valley near , Garnatje, Susanna 1801 and Vilatersana, 16-06-1997 [AF140472, AF140473]. Carduncellus hispanicus SPAIN, ALMERIA: Sierra de Gitdor, near the TV tower on the road Boiss. ex DC. F~lix-Canj/tyar, Garcia-Jacas, Susanna 1614 and Vilatersana, 22-06-1996 [AF140462, AF140463]. Carduncellus mareoticus EGYPT: road Alexandria-Marsah Matruh km 106, Susanna 1860 and Vilatersana, 8-06-1998 [AF140480, AF140481]. Carduncellus SPAIN, TARRAGONA: Montsant range, Grau del Carrasclet, Vilatersana monspelliensium All. 18, 18-10-1997 [AF140468, AF140469]. Carduncellus pinnatus MOROCCO, CHEFCHAOUEN: Between Bab Berred and Ketama, (Desf.) DC. Arista, Garcla Murillo, Gibbs, Lughadha and Talavera 4154/94, 23-06-1994 [AF140466, AF140467]. Carduncellus pomelianus MOROCCO, AL HOCEIMA: , Jbel Akechar, Boratynski Batt. and Romo 8792, 16-06-1995 [AF140436, AF140437]. Carduncellus rhaponticoides Cultivated in the Royal Botanic Garden of Madrid, Spain (234/87) Coss. & Dur. [AF140476, AF140477]. Carthamus arborescens L. SPAIN, ALMERIA: Sierra de Gfidor near F61ix, J. M. Montserrat, 27-08-1990 [AF140444, AF140445]. Carthamus boissieri Halficsy GREECE, KRITI, RETHIMNON: road N-77 between Rethimnon and Armeni, Vilatersana 26, 7-07-1996 [AF140456, AF140457]. MOROCCO, AL HOCEIMA: 38 km S of A1 Hoceima on the road to Nador, Garnatje, Susanna 1772 and Vilatersana, 15-06-1997 [AF140474, AF140475]. Carthamus dentatus GREECE, KRITI, RETHIMNON: road between Fouforas and Kourotes, (Forssk.) vahl subsp, ruber Vilatersana 33, 8-07-1996 [AF140428, AF140429]. (Link) Hanelt Carthamus glaueus M. Bieb. ARMENIA, EKHEGNADZOR: near Agarakadzor, Fajvush, Gabrielyan, Garcia-Jacas, Guara, Hovannisyan, Susanna 1551, Tamanyan and Vallds, 20-08-1995. [AF140454, AF140455]. R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 95

Table 2 (continued) Carthamus gypsicola Iljin ARMENIA, ARARAT: Vedi, Fajvush, Gabrielyan, Garcia-Jacas, Guara, Hovannisyan, Susanna 1579, Tamanyan and Vallds, 25-08-1995 [AF140446, AF140447]. Carthamus leucocaulos GREECE, KRITI, HANIA: base of the Mount Hrissokalitissas, Vilatersana 40, 11-07-1996 [AF140460, AF140461]. Carthamus nitidus Boiss. ISRAEL: Negev Desert, Dead Sea, R. Levy, 09-1997 [AF140482, AF140483]. Carthamus oxyacantha IRAN, TEHRAN: Sorkhehesar near Tehran, Garcia-Jacas, Mozaffarian, M. Bieb. Susanna 1626 and Vall~s, 2-08-1996 [AF140452, AF140453]. Carthamus riphaeus MOROCCO, AL HOCEIMA: Tleta Oued Laou between Tarerha and Font Quer & Pau Azenti, J. M. Montserrat 4360, Pallhs & Veny, 23-06-1993 [AF140448, AF140449]. Carthamus tenuis ISRAEL: Negev Desert, Dead Sea, R. Levy, 09-1997 (Boiss. & Blanche) Bornm. [AF140478, AF140479]. Carthamus tinctorius Seeds from the Botanic Garden of Nancy, (607/96) [AF140458, AF140459]. Carthamus turkestanicus ARMENIA, ARARAT: near Surenavan following a water conduction 1 km from the road, Fajvush, Gabrielyan, Garcia-Jacas, Guara, Hovannisyan, Susanna 1532, Tamanyan and Vallds, 19-08-1995 [AF140432, AF140433].

very low support within the Carthamus clade. primer and 307-R (D. Nickrent, personal commu- Removal of the three species did not modify the nication) located in the 26S gene as the reverse topology of the tree, but resulted in a significant primer. Sometimes for Carduncellus we used ITS-4 increase of the bootstrap support for the clades (White et al. 1990), located also in the 26S gene, as within Carthamus. Thereafter, none of the analyses the reverse primer. presented here included the three purportedly Double-stranded ITS-1 DNA was purified with hybrid species. The occasional problems produced QIA PCR purification minicolumns (Qiagen Inc., by hybrids in cladistic analysis are discussed in Valencia, California) and eluted in 30 gl of elution McDade (1992). buffer. Double-stranded ITS-2 DNA was purified ITS amplification and sequencing strategies. Fresh by precipitation with 3M sodium acetate and or dry leaves were taken from one or more absolute ethanol. After 15 rain of centrifugation, individuals, either seedlings or adult , but the supernatant was discarded and the pellet was always from the same population. In some cases vacuum dried and resuspended in water. The leaves were sampled from herbarium specimens. purification method was modified when there was DNA was isolated following the miniprep proce- more than one band in the amplification products, dure of Doyle and Doyle (1987) as modified in originating from infection of some cultivated Soltis et al. (1991). specimens with a mildew-type fungus. In these The ITS-1 and ITS-2 regions were amplified cases, double-stranded ITS-1 and ITS-2 DNA were separately using the polymerase chain reaction purified by electrophoresis through 1% low melting (PCR) protocol described in Soltis and Kuzoff point agarose. After staining with ethidium bro- (1993). For amplification of ITS-l, we used the mide, the band was excised and the agarose 1406-F primer (D. Nickrent, Southern Illinois fragment was digested using the Gelase Agarose University, Carbondale, IL, personal communica- Gel-Digesting Preparation, Epicentre Technolo- tion), located near the 3' end of the 18S gene; the gies. DNA was then precipitated with 5M ammo- reverse primer was ITS-2 (White et al. 1990), nium acetate and ethanol, and vacuum dried. The located in the 5' end of the 5.8S gene. The ITS-2 sequencing primers used for the ITS-1 and ITS-2 region was amplified using the ITS-3 primer (White regions were ITS-1 and ITS-4 (White et al. 1990), et al. 1990) located in the 5.8S gene as the forward respectively. Both were labeled with digoxigenine, 96 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus

Boehringer Mannheim. After direct sequencing of Parsimony analysis involved heuristic searches the double stranded DNA using the cycle sequenc- conducted with PAUP version 3.1.1 (Swofford ing Fmol kit, Promega Corp., sequences were 1991) using TBR branch swapping with character resolved in 6% acrylamide gels. After transferring states specified as unordered and unweighted. All the DNA fragments to a positively charged nylon most-parsimonious trees were saved. To locate membrane, sequences were stained with the DIG other potential "islands" of most-parsimonious detection kit, Boehringer Mannheim. trees (Maddison 1991), we performed 100 replica- Carduncellus eriocephalus and C. helenioides tions with random taxon addition, also with TBR presented a sequencing problem, either a polymor- branch swapping. Bootstrap analysis (Felsenstein phism or a PCR artifact, in positions 42 and 191, 1985), and decay analysis (Bremer 1988, Donoghue respectively, of the ITS-I region, resulting in et al. 1992) were carried out for an estimate of unreadable sequences from the artifact towards support of the branches. the end of the ITS-1 region. The problem was Because the bootstrap could not be done in the solved by sequencing the complementary strand, usual way with PAUP 3.1.1 (the limit of trees was which yielded a perfect reading from the base always hit in the first ten replicates regardless of change to the 255 position. The primer used for starting seed used), we used the approach by Lid6n sequencing the complementary strand was 5.811 et al. (1997) using 1000 replicates, random taxon (Sun et al. 1994). The ITS-2 region presented a addition with 20 replicates per replicate and no similar problem in Carduncellus araneosus, C. swapping. In the decay analyses, the memory limit dianius, C. monspelliensium, C. pinnatus, C. pom- of PAUP (32700 trees) was reached while swapping elianus and Carthamus nitidus, between the posi- on trees only one step longer than the shortest tions 23 and 41, a zone whose alignment required trees. Therefore, successive decay analyses were the insertion of some gaps (notably, an indel of 22 conducted using the clade constraint approach as bp in Carduncellus cuatrecasasii). The complemen- discussed in Morgan (1997). tary strand was sequenced using the primer 5.812 (Sun et al. 1994). This kind of problem is not frequent in the Centaureinae, but there are other Results cases in ITS amplification (cf. Sun et al. 1994). Size and composition of ITS. The length of the Data analysis. DNA sequences were aligned ITS-1 region in species of Carduncellus and visually by sequential pairwise comparison (Swof- Carthamus ranged from 250 bp in Car- ford and Olsen 1990). The alignment of ITS-1 sequences for all taxa required interpretation of six duncellus fruticosus to 253 bp in Carduncellus indels of (1-2bp) at 2.34% of the sites. Alignment caeruleus. The ITS-2 region varied between of ITS-2 sequences required inference of five indels 196 bp in Carduncellus cuatrecasasii to 217 bp of (1-5 bp) at 2.26% of atl sites, except for in Carthamus arboreseens and Carthamus Carduncellus cuatrecasasii, which had a 22bp riphaeus. deletion located in the position 26-48bp (Table 3). Sequence divergence was calculated sepa- The aligned matrix is available from the first rately for the ITS-1 and ITS-2 regions using author. the Distance Matrix option available in Because it is now generally accepted to con- PAUP. ITS-1 sequence divergence varied from serve the phylogenetic signal of the gaps, which 0% within species of Carduncellus (C. duvauxii, may be lost if they are coded as missing values, we C. pomelianus and C. rhaponticoides) and carried out two different analyses. In the first one, Carthamus ( C. dentatus subsp, ruber, following Wojciechowski et al. (1993), we coded gaps as missing values. In a second analysis, C. glaucus, C. boissieri and C. leucocaulos; following Sun et al. (1994), Bain and Jansen C. tinctorius and C. oxyacantha) to 6.5% (1995), and Samuel et al. (1997), gaps were coded (pairwise distance between Femeniasia baleari- as a fifth base. Other different ways of coding indels ca and Carduncellus dianius). Divergence be- (e. g. recoding them as additional characters tween ingroup and outgroup genera ranged appended to the sequence matrix) did not modify from 5.4% between Centaurea dealbata and the topology of the trees. Cartharnus boissieri to 11.2% between R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 97

Femeniasia balearica and Cheirolophus Carthamus sect. Thamnacanthus; and third, all arbutifolius. the included species of the genus Carduncellus. ITS-2 sequence divergence values were The branch formed by Carthamus arborescens slightly higher than those for ITS-l, which and C. riphaeus (the genus Phonus) had a high was consistent with our results in Cheirolophus support with a bootstrap of 100% and decay (Susanna et al. 1999), but not with the results of 7. The next clade, CardunceIIus, had a weak for some other Asteraceae (Baldwin 1992, bootstrap of 79% and decay index of 1. Within 1993; Susanna et al. 1995; Bayer et al. 1996). the Carduncellus branch, Carduncellus frutico- It could be a consequence of the relative age of sus was sister to the rest of species. Next was the group: both Cheirolophus and Carduncellus CarduncelIus mareoticus with a high bootstrap are groups of recent speciation (Susanna et al. of 90% and decay index of 3. The relationships 1999). among the remaining species were not well We found a minimal divergence of 0% in resolved; bootstrap support was generally low, Carduncellus between C. araneosus, C. hispa- except for a bootstrap of 72% and a decay nicus, and C. monspelliensium; C. eriocephalus index of 2 for the clade that included Card- and C. duvauxii; C. heleniodes, C. duvauxii, and uncellus dianius, C. duvauxii, C. calvus, C. C. caeruleus. The minimal divergence within caeruleus, C. eriocephalus, C. heleniodes, C. Carthamus was 0.9% between C. arborescens pomelianus and C. rhaponticoides. and C. riphaeus; C. gypsicola and C. tinctorius; The clade formed by Carthamus (excluding also between C. tenuis, C. dentatus subsp, ruber, sect. Thamnacanthus) had very low support. and C. leucocaulos. Maximum divergence Within this clade there were two groups. In the within genera was 10.8% between Carthamus first one, Carthamus nitidus was sister to the tinctorius and Carduncellus cuatrecasasii. Di- rest of species with a bootstrap of 77% and vergence between ingroup and outgroup genera decay of 2. Next, Carthamus dentatus subsp. ranged from 5.2% between Carthamus dentatus ruber, C. glaucus, C. boissieri, C. leucocaulos subsp, ruber, Centaurea toletana, and Cheirolo- and C. tenuis were grouped with a bootstrap of phus arbutifolius; Carthamus leucocaulos, C. 82% and decay of 2. In the second group, tenuis and CheiroIophus arbutifoIius; Carthamus Carthamus oxyacantha was sister to the rest of tenuis, C. nitidus, and Centaurea toletana to the species with a high bootstrap of 100% and 13.7% between Carduncellus cuatrecasasii and a decay index of 11. The next clade was formed Acroptilon repens. by Carthamus gypsicola and C. tinctorius with Phyiogenetic analyses. Aligned ITS-1 and a bootstrap of 75% and decay index of 1. ITS-2 sequences formed a matrix of 477 Second analysis: The strict consensus tree when characters, of which 100 were informative gaps were coded as a fifth base is shown in (excluding indels): 50 informative characters Fig. 3. The topology of this second analysis was in the ITS-1 and 50 in the ITS-2. Two different almost the same as in the first analysis except for analysis were conducted, with different treat- the position of Femeniasia as sister of the species ment of the gaps. Although significant, differ- of Carthamus sect. Thamnacanthus with a ences between the two consensus trees were bootstrap of 89% and decay index of 3. minor and are discussed later in the paper. First Analysis: The strict consensus of all the Discussion and conclusions trees when gaps were coded as missing data is shown in Fig. 2. The first analysis grouped Generic delineation. The delineation of genera Carduncellus and Carthamus, but with essen- suggested by the analysis of the ITS sequences tially no support: the bootstrap value was 33% supports the core proposal of Ldpez Gonzfilez and the decay index was 1. There was a (1990): Phonus (= Carthamus sect. Thamnacan- monophyletic clade with three branches: first, thus) should constitute a different genus. If the monospecific genus Femeniasia; second, Phonus is removed from Carthamus as a 98 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus

Carthams arborescens [Phonus] | Carthamus riphaeus [Phonus] Femeniasia balearlca Carduncellus mitissimus Carduncellus hiepanicus 1 ~ Carduneellus araneosus _m j Carduncellus cuatrecaeasil Carduncellus duvauxii

26 Carduncellus pomelianus 1 41 ~ Carduncellus calvus [Lamottea] 27 1 ~ Carduncellus caeruleus [Lamottea] 79 1 :~:i- 37 Carduncellus eriocephalus W ~:i 1 Carduncellus helenioides[Lamottea] 90 31 72 ~--- Carduncellus dianius [Lamottea] 3 1 53 2 n i Carduncelhls rhaponthicoides 1 67 m_ i Carduncellus pinnatus

1 i Carduncellus menspelllensium _n ~ Carduncellus mareoticus [Phonus?] l~ Carduncellus frnticosus [Phonus?]

33 ~ Carthamus dentatu~ subep, ruber i Car thamus gl aucus 82[: Carthamus boissierl 77 21: Carthamus leucocaulos F-2 Carthamus tennis 46 ~-- Carthamus nitidus :-:- • ~. ~ I 1 75 ~ Carthamus g~l~.icola i00 I 1 ~ Carthamus tinctorlus ii - ~I Carthamus oxyacantha i00 ..... Centaurea clementei I 12 [" Centaurea toletana Centaurea americana

. 26 Centaurea afrlcana 37 Cheirolophus arbutifoliua ~croptilon repens Centaurea dealbata

SYMBOLS

:.)iiAnnual iiilPerennial ~ Single pappus I Double pappus m Appendages of the bracts cucullate R Appendages of the bracts not cucullate E eastern distribution W western distribution

Fig. 2. Strict consensus tree of the 792 most parsimonious trees (gaps coded as missing values). Length of the trees: 189 steps. Consistency index excluding uninformative characters (CI)=0.596; retention index (RI) : 0.777. Numbers above branches are bootstrap values. Numbers below branches are decay indices R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 99

Femeniasia balearica

89 i I00 I Carthamus arborescens [Phonus]

8 Carthamus riphaeus [Phonus] ©| Carduncellus mitissimus :| Carduncellus cuatrecasasii i Carduncellus duvauxii

~ Carduncellus pomelianus

74 Carduncellus calvus [Lamottea]

w i----i- 1 67 Carduncellus dianius [Lamottea] 1 ~ -- Carduncellus caeruleus [Lamottea] 88 --~---Carduncellus eriocephalus

Carduncellus helenioldes [Lamottea]

--~---Carduncellus rhaponticoides

i Carduncellus hispanicus 73 Carduncellus araneosus ~ --Carduncellus pinnatus 31 ~ -----Carduncellus monspelllensium ~ --Carduncellus mareoticus [Phonus?] ~ ---Carduncellus fruticosus [Phonus?] Carthamus dentatus subsp, ruber

Carthamus glaucus

m Carthamus boissieri

76 I Carthamus leucocaulos I-" Carthamus tenuis 43 Carthamus nitidus x- I e ii:i I i [] Carthamus glrpsicola 100[ Carthamus Orlracantha 11 [] Carthamus tinctorius 100 Centaurea clementei

13 Centaurea toletana Centaurea americana

26 Centaurea africana 37 20 Cheirolophus arbutifolius

1 Acroptilon repens

Centaurea dealbata

SYMBOLS

ii~iRnnual ~ii Perennial ~ Single pappus I Double pappus

m Appendages of the bracts cucullate m ~%ppendages of the bracts not cucullate

E eastern distribution W western distribution

Fig. 3. Strict consensus tree of the 32700 most parsimonious trees (gaps codified as fifth base). Length of the trees: 232 steps. Consistency index excluding uninformative characters (CI)= 0.558; retention index (RI)= 0.739. Numbers above branches are bootstrap values. Numbers below branches are decay indices 100 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus distinct genus, most of the delimitation prob- Africa, supports the idea of recent divergence lems would disappear. The result is a discrete (Fig. 1). Failure to resolve relationships within segregation between annual, eastern Car- the genus Carduncellus reiterates the extremely thamus and the wide complex of perennial, low level of variation of the ITS region in western genera (CardunceIIus, Phonus and many genera of recent origin (Nepokroeff and Femeniasia), without intermediates. That Pho- Sytsma 1996, Susanna et al. 1999). Other nus is closer to Carduncellus, both being regions of the genome might be more effective. perennial and western, than to Carthamus For example, Francisco-Ortega et al. (1996) (annual and eastern), would also be an obvious obtained a well resolved phylogeny of the prediction based on biogeography alone young insular genus Argyranthemum based on (Fig. 1). As Hanelt (1963) pointed out, the cp-DNA restriction site analysis. species of Phonus grow in natural plant Carduncellus fruticosus and C. mareoticus. associations, as do all the species of Card- Our results indicated that Carduncellus fruti- unceIlus. In contrast, species of Carthamus cosus and C. mareoticus were successive sisters grow only in human-disturbed habitats. to the rest of Carduncellus (Fig. 2), in agree- Although it effectively differentiated be- ment with Hanelt (1963). They share unusual tween Cartharnus and the other genera within (for Carduncellus) features: Carduncellus fruti- the complex (Carduncellus, Phonus and cosus and C. mareoticus are spiny shrublets, Femeniasia), the ITS phylogeny did not resolve while the rest of the species of the genus are the relationship between the two major groups hemicryptophytes. On the other hand, they of the complex. Even though the Carduncellus share with Phonus and Femeniasia the unique group and Carthamus were united in a clade, feature of having cypselas with an undifferen- there was essentially no support for this branch tiated pericarp. On this basis, Ldpez Gonz~tlez (Fig. 2, Fig. 3). (1990) included both of them in Phonus. Carduncellus. Our delineation of Car- However, the ITS sequences indicate that their duncellus coincided with the conclusions of systematic position is within Carduncellus, as Hanelt (1963, 1976). The ITS sequence analysis the most basally divergent species of that did not support segregation of Lamottea as a genus. According to the ITS phylogeny, their different genus. Carduncellus caeruIeus, C. cal- inclusion in Phonus would make this genus vus, C. dianius, and C. helenioides, all included paraphyletic (Fig. 2 and 3). Petit (1997) sub- in Larnottea by Ldpez Gonzfilez (1990), did not ordinated C. fruticosus to Femeniasia on the form a group in our analysis (Fig. 2, Fig. 3). basis of vegetative similarities that may have Morphological differences in the cypselas indi- been overemphasized. For example, spines of cated by Ldpez GonzAlez (1990) among species Femeniasia are true leaves that retain their of CardunceIlus are an example of the abrupt basal sheath (Susanna 1988), whereas C. fruti- evolution of cypselas that can occur in the cosus bears fully lignified cauline spines. Most Centaureinae (Wagenitz and Hellwig 1996). of the characters that purportedly connect Our analysis also suggests that speciation Femeniasia balearica with Carduncellus fruti- in Carduncellus is very recent relative to other cosus are adaptations to habitats that are xeric Centaureinae, except for the more archaic (C. fruticosus) or extremely windy (Femeni- C. fruticosus and C. mareoticus. The ITS asia). divergence values are very low, as in other Femeniasia. Unfortunately, our results for young genera of the Centaureinae (e. g. Femeniasia were somewhat ambiguous. ITS Cheirolophus, cf. Susanna et al. 1999). Even if sequence analysis indicated that it belonged to species of the group are perennials, thus having the western complex, along with Carduncellus lower rates of ITS evolution, the narrow and Phonus. However, its final position within geographic distribution of Carduncellus, limit- the group depended on the interpretation of ed mainly to the Iberian Peninsula and North gaps in the phylogenetic analysis. If gaps were R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 101

Table 3. Bases 8-36 of the ITS-2 region. The insertion (1 bp long) which groups Femeniasia and Phonus is boldfaced

CAT-CTCCC-ATGGGGAC-ACGTGTTTG Femeniasia balearica CATG-CTCCC-CATGGGGACGACGTGTTTG Carthamus arborescens CATg-CTCCC-CATGGGGACGACGTGTTTG Carthamus riphaeus CAT?-CTCCC-CATGGGGAC--GTTTG Carduncellus araneosus CATG-CTCCC-CATGGGGRC--GTGTTTG Carduncellus calvus CATG-CTCCC-CATGGGGAC--GTGTTTG Car duncellus caeruleus CATG- -TTTG Carduncellus cuatrecasasii CATG-CTCCC-CATGGGGAC--GTGTTTG Car duncellus dianius CATG-CTCCC-CATGGGGAC--GTGTTTG Carduncellus duvauxii CATG-CTCCC-CATGGGGAC--GTGTTTG Carduncellus eriocephalus CATG- C T C C C - C ATGGGGAC--GTGTTTG Carduncellus fruticosus CATG-CTCCC-CATGGGGAC--GTGTTTG Carduncellus helenioides CATG-CTCCC-CATGGGGAC--GTTTG Carduncellus hispanicus CATG-CTCCCCCATGGGGAC--GTGTTTG Carduncellus mareoticus CATG-CTCCC-CATGGGGA?--??GTTTG Carduncellus mitissimus CATG-CTCCC-CATGGGGAC--GTGTTTG Carduncellus monspelliensium CATG-CTCCC-CTTGGGGAC--GTGTTTG Carduncellus pinnatus CATG-CTCCC-CATGGGGAC--GTGTTTG Carduncellus pomelianus CATG-CTCCC-CATGGGGAC--GTGTTTG Carduncellus rhaponticoides CACG-CTCCC-CATGGGGAC--GTGTTTG Carthamus boissieri CATG-CTCCC-GATGGGRAC--GTGTTTG Carthamus dentatus subsp, ruber CATG-CTCKC-CATGGGGAC--GTGTTTG CATGGCTACC-CATGGGGAA--GTGTTTG Carthamus gypsicola CACG-CTCCC-CATGGGGAC--GTGTTTG Carthamus leucocaulos CATG- CTC C C C CATAGGGAC--GTTTG Carthamus nitidus CATG-CTCCC-CATGGGGAA--GTGTTTG Carthamus oxyacantha CAYG-CTCCC-CATGGGGAC--GTGTTTG Carthamus tenuis CATG-CTACC-CATGGGGAA--GTGTTTG Carthamus tinctorius CACG-CTCCC-CATGGGGAT--GTGTTTT Acroptilon repens CACG-CTCCC-CATGGGGAC--GTTTTGC Centaurea africana CATG-CTCCC-CATGGGGAT--GTGTTTT Centaurea americana CATG-CTCCC-CATGGGGAC--GTTTTTG Centaurea clementei CACG-YTCCCCTATGGGGAC--GTGTTTC Centaurea dealbata CATG" CTC C C- CATGGGGAC-AT-TTTTTG Centaurea toletana CATG-CTCCCTCATGGGGAC-T-TGTTTT Cheirolophus arbutifolius

coded as missing, Femeniasia stood as an belongs to the same lineage from which Phonus isolated genus. When we coded gaps as a fifth also evolved, but it is a distinct genus. We base, Femeniasia was sister to Phonus, with a propose to keep it at the generic rank until bootstrap of 89; however, the indel which more unambiguous evidence clarifies its rela- forced Femeniasia into Phonus was formed by tionship to Phonus. only one bp (Table 3). There are obvious An interesting sidelight of our analysis was morphologic differences with Phonus, in habit, the relationship of Femeniasia placement rela- appendages of the bracts, and pappus (Vila- tive to its biogeography. Femeniasia balearica tersana, unpublished data). However, Femeni- is a narrow endemic in the north of the island asia shares with Phonus an undifferentiated of Minorca, where numerous endemic ele- pericarp. Our conclusion is that Femeniasia ments are believed to have affinities with the 102 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus

Tyrrhenic flora (Cardona and Contandriopou- Middle bract appendages represent another los 1979). The ITS sequences indicated instead prominent example of classification difficulties that Femeniasia belonged to an Ibero-North generated by the frequent reversals and sec- African stock. ondary reductions that characterize the sub- Carthamus. For Carthamus, our results tribe Centaureinae. These appendages are supported the idea of two rather different missing in Carthamus, Phonus and some spe- groups: Carthamus sensu stricto, which in- cies of Carduncellus (L6pez GonzAlez 1990; cludes only sect. Carthamus, and the rest of the Vilatersana, unpublished data). However, the sections of Hanelt (1963). De Candolle (1838) presence of cucullate-lacerate appendages in and Cassini (1819) classified these groups as the middle bracts is one of the core characters separate genera: Carthamus and Kent- that segregate the subtribes Centaureinae and rophyllum. Differences in morphology and Carduinae (Cassini 1819, Bentham 1873, karyology are evident (Hanelt 1963) and the Hoffmann 1894, Dittrich 1977, Bremer 1994, splitting of Carthamus should be reconsidered. Wagenitz and Hellwig 1996). However, our results are provisional because The presence of so many primitive charac- our sampling of sect. Carthamus was not ters (for the subtribe Centaureinae!) in some comprehensive, including only three out of species of the Carduncellus-Carthamus com- nine species of sect. Carthamus. According to plex is remarkable because this complex be- Hanelt (1963), these three are very closely longs to the modern clade of Centaureinae as related. A wider representation of Carthamus defined by Susanna et al. (1995) and Wagenitz sect. Carthamus would likely show more and Hellwig (1996). The presence of an undif- clearly the relationship of the two groups. ferentiated pericarp and a deciduous single Evolution of characters within the complex. pappus and the absence of cucullate append- A comparison of morphological characters ages could represent the persistence of primi- with the ITS phylogeny supports the opinion tive states, rather than reversals of advanced of Dittrich (1969) that features of cypselas and characters to a more primitive-appearing state. appendages of the bracts cannot be used as But because these characters (especially the exclusive diagnostic characters for the com- lack of cucullate appendages) occur only in plex. In the case of cypselas, a double pappus scattered species within genera of the group is present in Carthamus and in some species of (Figs. 2 and 3), and are not clustered in Carduncellus. Most species of Centaureinae, taxonomic lines, we find the hypothesis of including the most primitive groups of the reversals more credible than random survival subtribe, have a double pappus, thus forming a of truly primitive characters. critical character for the separation of Card- As shown in a previous publication on the uinae and Centaureinae (Dittrich 1968, 1969, phylogeny of the Centaureinae (Susanna et al. 1977; Susanna etal. 1995; Wagenitz and 1995), clarification of systematic and phyloge- Hellwig 1996). We believe that the single netic relationships in this subtribe requires not pappus of many species of Carduncellus is a only a very careful analysis of morphological secondary reduction. We also interpret the characters, but also the interpretation of other undifferentiated pericarp present in the basally evidence (biogeographic, karyological and mo- diverging species of Carduncellus, C. fruticosus lecular). While this rule generally applies in all and C. mareoticus, in Phonus (Dittrich 1969) systematics, an integrative approach using all and in Femeniasia (Vilatersana, unpublished characters becomes an undeniable necessity data) as a reversal. A differentiated pericarp is when interpreting the numerous complex rela- present in the rest of the species of the complex tionships in Centaureinae. and in all the other genera of the subtribe Centaureinae, including the more primitive Financial support by the CICyT (project PB93- ones as well (Dittrich 1968, 1969). 0032) is gratefully acknowledged. The authors R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 103 thank G. Ldpez Gonzfilez, for his invaluable Boissier E. (1873) Flora orientalis, sive enumeratio contribution to the design of this work and his plantarum in Oriente a Graecia et Aegypto ad help in the determination of the materials; E. Indiae fines hucusque observatarum. Gen~ve. Gabrielyan, M. Hovhannisyan, K. Tamanyan and Bremer K. (1988) The limits of amino acid G. Fajvush, from the Botanic Institute of the sequence data in angiosperm phylogenetic re- Academy of Sciences of Armenia, for their warm construction. Evolution 42: 795-803. hospitality in Erevan and their help in the field Bremer K. (1994) Asteraceae. Cladistics and clas- work; M. Jaffari, A. Jalili, Z. Jamzad, A. Ma- sification. Timber Press, Portland. assoumi and V. Mozaffarian, from the Research Cardona M. A., Contandriopoulos J. (1979). En- Institute of Forest and Rangelands, Iran, for their demism and evolution in the islands of the help in the field collections in Iran; R. Levy, from Western Mediterranean. In: Bramwell D. (ed.) Beniamina, Israel, who got his hands pricked Plants and Islands. Academic Press, London. collecting Carthamus nitidus and C. tenuis for us; Cassini H. (1819) [Different articles]. Dictionnaire the botanic gardens of Madrid, Spain, and Nancy, de Sciences Naturelles. Paris. Cited by King R., France; and M. Torrell, for her help in the lab Dawson H. W. (1975). Cassini on Compositae. work. Our grateful thanks to C. Roch6, who Oriole Editions, New York. converted the manuscript into readable English. De Candolle A. P. (1838) Prodromus systematis naturalis regni vegetabilis 6. Paris. Dittrich M. (1968) Morphologische Untersuchun- References gen an den Frfichten der Subtribus Cardueae- Ashri A., Knowles P. F. (1960) Cytogenetics of Centaureinae (Compositae). Willdenowia 5: safflower (Carthamus L.) species and their hy- 67-107. brids. Agron. Jour. 52:11 17. Dittrich M. (1969) Anatomische Untersuchungen an Bain J. F., Jansen R. K. (1995) A phylogenetic den Frfichten yon Carthamus L. und Carduncellus analysis of the auroid Senecio (Asteraceae) Adans. (Compositae). Candollea 24(2): 263-277. complex based on ITS sequence data. Plant Dittrich M. (1977) - systematic review. Syst. Evol. 195: 209-219. In: Heywood V. H., Harbone J. B., Turner B. L. Baldwin B. G. (1992) Phylogenetic utility of the (eds.) The Biology and Chemistry of Comp- Internal Transcribed Spacers of nuclear ribo- ositae. Academic Press, London. somal DNA in plants: an example from the Donoghue M. J., Olmstead R. G., Smith J. F., Compositae. Molec. Phylogenet. Evol. 1: 3-16. Palmer J. D. (1992) Phylogenetic relationships of Baldwin B. G. (1993) Molecular phylogenetics of Dipsacales based on rbcL sequences. Ann. Mis- Calycadenia (Compositae) based on ITS se- souri Bot. Garden 79: 333-345. quences of nuclear ribosomal DNA: chromo- Doyle J. J., Doyle J. L. (1987) A rapid DNA somal and morphological evolution reexamined. isolation procedure for small quantities of fresh Amer. J. Bot. 80: 222-238. leaf tissue. Phytochem. Bull. 19: 11-15. Battandier J. A. (1890) Dycotyledon6es. In: Bat- Estilai A. (1977) Interspecific hybrid between tandier J. A., Trabut A. (eds.) Flore de l'Alg~rie, Carthamus tinctorius and C. alexandrinus. Crop contenant la description de toutes les plantes Sci. 17: 800-802. signal6es jusqu'fi ce jour comme spontan~es en Estilai A., Knowles P. F. (1976) Cytogenetic studies Alg~rie et catalogue des plantes du Maroc. of Carthamus divaricatus with eleven pairs of Alger. chromosomes and its relationship to other Bayer R. J., Soltis D. E., Soltis P. S. (1996) Carthamus species (Compositae). Amer. J. Bot. Phylogenetic inferences in Antennaria (Aster- 63: 771-782. aceae: Gnaphalieae: Cassiniinae) based on se- Estilai A., Knowles P. F. (1978) Relationship of quences from nuclear ribosomal DNA internal Carthamus leucocaulos to other Carthamus spe- transcribed spacers (ITS). Amer. J. Bot. 83: 516- cies (Compositae). Canad. Jour. Genet. Cytol. 527. 20: 221-233. Bentham G. (1873) Compositae. In: Bentham G., Felsenstein J. (1985) Confidence limits on phylo- Hooker J. D. (eds.) Genera plantarum 2(1). genies: an approach using the bootstrap. Evolu- London. tion 39: 783-791. 104 R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus

Francisco-Ortega J., Jansen R. K., Santos- Petit D. (1997) Generic interrelationship of the Guerra, A. (1996) Chloroplast DNA evidence Cardueae (Compositae): a cladistic analysis of of colonization, adaptive radiation and hybrid- morphological data. Plant Syst. Evol. 207: 173- ization in the evolution of the Macaronesian 203. flora. Proc. Natl. Acad. Sci. USA 93: 4085M090. Pomel A. (1860) Mat6riaux pour la Flore Atlan- Godron D. A. (1852) Synanther6es. In: Grenier J. tique. Caen. C., Godron D. A. (eds.) Flore de la France 2. F. Pomel A. (1874) Nouveaux mat6riaux pour la Savy, Paris, pp. 81-392. Flore Atlantique. Paris. Hanelt P. (1963) Monographische Ubersicht der Qu6zel P., Santa S. (1963) Nouvelle flore de Gattung Carthamus L. (Compositae). Feddes l'Alg~rie et des r~gions d~sertiques m~ridionales. Repert. 67: 41-180. Editions du Centre National de la R6cherche Hanelt P. (1976) Carthamus L. and Carduncellus Scientifique, Paris. Adans. In: Tutin T. G., Heywood V. H., Burges Rivas Goday S., Rivas Martinez S. (1967) Acerca N. A., Valentine D. H., Walters S. M, Webb D. de los Carthamo-Carduncellus de la Ononido- A. (eds.) Flora Europaea 4. Cambridge Univer- Rosmarinetea peninsular. Anales Inst. Bot. Ca- sity Press, Cambridge, pp. 302-304. vanilles 25: 188-197. Hoffmann O. (1894) Compositae. In: Engler A., Samuel R., Bachmair A., Jobst J., Ehrendorfer F. Prantl K. (eds.) Die nattirlichen Pflanzenfamilien (1997) ITS sequences from nuclear rDNA sug- 4(5). Leipzig, pp. 87-387. gest unexpected phylogenetic relationships be- Jahandiez E., Maire R. (1934) Catalogue des tween Euro-Mediterranean, East Asiatic and plantes du Maroc. Alger. North American taxa of Quercus (Fagaceae). Kim K. J., Jansen R. K. (1994) Comparisons of Plant Syst. Evol. 211: 129-139. phylogenetic hypothesis among different data Sang T., Crawford D. J., Stuessy T. F., Silva M. sets in dwarf dandelions (Krigia, Asteraceae): (1995) ITS sequences and the phylogeny of the additional information from internal transcribed genus Robinsonia (Asteraceae). Syst. Bot. 20: 55- spacers sequences of nuclear ribosomal DNA. 64. Plant Syst. Evol. 190: 157-185. Schank S. C., Knowles P. F. (1964) Cytogenetics of Lid~n M., Fukuhara T., Rylander J., Oxelman B. hybrids of Carthamus species (Compositae) with (1997) Phylogeny and classification of Fu- ten pairs of chromosomes. Amer. J. Bot. 51 (10): mariaceae, with emphasis on Dicentra s. 1. based 1093-1102. on the plastid gene rpsl6 intron. Plant Syst. Soltis D. E., Kuzoff R. K. (1993) ITS sequence Evol. 206: 411-420. homogeneity within and among populations of Linn6 C. (1753) Species Plantarum. Holmiae. Lomatium grayi and L. laevigatum (Umbellife- L6pez Gonz/tlez G. (1990) Acerca de la clasifica- rae). Molec. Phylogenet. Evol. 2: 166-170. ci6n natural del g6nero Carthamus L. s. 1. Anales Soltis D. E., Kuzoff R. K., Soltis P. S., Collier T. Jard. Bot. Madrid 47(1): 11-34. G., Edgerton M. L. (1991) The Heuchera group Maddison D. R. (1991) The discovery and impor- (Saxifragaceae): evidence for chloroplast transfer tance of multiple islands of most parsimonious and paraphyly. Amer. J. Bot. 78: 1091-1112. trees. Syst. Zool. 40: 315-328. Sun Y., Skinner D. Z., Liang G. H., Hulbert S. H. McDade L. A. (1992) Hybrids and phylogenetic (1994) Phylogenetic analysis of Sorghum and systematics II: the impact of hybrids on cladistic related taxa using internal transcribed spacers of analysis. Evolution 46: 1329-1346. nuclear ribosomal DNA. Theor. Appl. Genet. Morgan D. R. (1997) Decay analysis of large sets of 89: 26-32. phylogenetic data. Taxon 46: 509-517. Susanna A. (1988) Femeniasia, novus genus Car- Nepokroeff M., Sytsma K. J. (1996) Systematics duearum. Collect. Bot. Barcelona 17: 83-88. and patterns of speciation and colonization in Susanna A., Garcia-Jacas N., Soltis D. E., Soltis P. Hawaiian Psychotria and relatives based on S. (1995) Phylogenetic relationships in tribe phylogenetic analysis of ITS sequence data. Cardueae (Asteraceae) based on ITS sequences. Amer. J. Bot. 83, Suppl.: 181-182. Amer. J. Bot. 82: 1056-1068. Nyman C. F. (1878-1890) Conspectus Florae Susanna A., Garnatje T., Garcia-Jacas N. (1999) Europaeae. Orebro. Molecular phylogeny of Cheirolophus (Aster- R. Vilatersana et al.: Phylogeny of Carduncellus-Carthamus 105

aceae: Cardueae-Centaureinae) based on ITS White T. J., Bruns T., Lee S., Taylor J. (1990) sequences of nuclear ribosomal DNA. Plant Amplification and direct sequencing of fungal Syst. Evol. 214: 147-160. ribosomal RNA genes for phylogenetics. In: Susanna A., Vilatersana R. (1996) Las afinidades de Innis M., Gelfand D., Sninsky J., White T. (eds.) Ferneniasia Susanna (Compositae), o rectificar es PCR protocols: a guide to methods and appli- de sabios. Anales Jard. Bot. Madrid 544: 355-357. cations. Academic Press, San Diego, pp. 315- Swofford D. L. (1991) PAUP: Phylogenetic anal- 322. ysis using parsimony, version 3.1. Illinois Natu- Wojciechowski M. F., Sanderson M. J., Baldwin B. ral History Survey, Champaign. G., Donoghue M. J. (1993) Monophyly of Swofford D. L., Olsen G. J. (1990) Phylogeny aneuploid Astragalus (Fabaceae): evidence from reconstruction. In: Hillis D., Moritz C. (eds.) nuclear ribosomal DNA Internal Transcribed Molecular systematics. Sinauer, Sunderland, pp. Sequences. Amer. J. Bot. 80: 711-722. 411-501. Wagenitz G., Hellwig F. H. (1996) Evolution of characters and phylogeny of the Centaureinae. Addresses of the authors: R. Vilatersana, A. In: Hind D. J. N., Beentje H. G. (eds.) Comp- Susanna, N. Garcia-Jacas and T. Garnatje (e-mail: ositae: Systematics. Proceedings of the Interna- [email protected]) Institut Botanic de Barce- tional Compositae Conference, Kew, 1994. lona (CSIC). Av. Muntanyans, s. n., E-08038 Royal Botanical Gardens, Kew, pp. 491-510. Barcelona, Spain.