Generic boundaries and evolution of characters in the Arctium group : a nuclear and chloroplast DNA analysis

Definir los límites genéricos en el grupo Arc/;/l1/I (Composi/(ll!. Car(/ul!al!-C{/l'dllilwe). formado por los géneros ArctiulI1, COlIsillitl. Hypac(IIl/!liulI1 y Schlllal1wlIsell;a, ha resultado ser una tarea muy complicada. Concretamente, la separación precisu de Are/il/m y COl/sillia es muy difrcil de establecer. En consecuencia. hemos :lOaliz.ado las secuencias del DNA de dos regiones. el gen cloropltistico marK y los espaciadores ITS 1 y 2 del ONA ribos6mico-nuc1cllT. de una amplia representación de todos los géneros del grupo: en el caso de CO/l.~il1ia, centrándonos en las especies más obviamente relacionadas con Arctilflll, ¡'Iemos cncontrtldo una correlaci6n exacta entre Iilogenia molecular y dos caracteres fundamentales. el tipo de polen y el número cromos6mico: todas las especies estudiadas que lienen el tipo de polen Arctiastmll1 y el número x= 18, característicos de Arctilllll sensu stricto. forman un ciado monofilético. hermano de ouo ciado monofilético formado por las especies de COlIsillia sensu suicto. Sin embargo. el ciado "Arclioide" no se puede definir con caracteres morfol6gicos macrosc6picos: el carácter que separa Arctil/lI1 y COIlSillia son las hojas espinoso·pinnatifidas o pinnatisectas de COlIsiltia, que es adaptativo y tiene poca relevancia taxonómica. Según nuestros resultados, las espinas habrían aparecido en dos linajes diferelllcs: en COl/sillia. por una parte, y en H)'fJacanr/¡irllll y Sdll/rtlllral/sel1ia. que son espinosos y morfológicamente más pr6ximos a COlIsinia, pero que están sin duda más relacionados con Arctil/lIl, inerme. Evaluamos también las implicaciones de esta incongruencia entre morfologfa, por un lado, y datos moleculares, palinol6gicos y cariológicos, por otro. Proponemos algunas soluciones, pero ninguna es totalmeme satisfactoria: se necesitan más estudios que incluyan más especies de COl/sillia subg. H)'fJacanrltotles.

Having acknowledged the tribal and subtribal limits, the next step was lo e1arify the delimitation of sorne of the major genera in the tri be.Now, among the most interesting and vexing problems of genus rank that persist in the Cardueae is the marking of lhe generic boundaries in the Arelillm group.
Three important eharacters never found in combination elsewhere in the tribe characterize basically theAre/illm group.First, the receptaele has strongly twisted scales.Second, the achenes are always tigrine (with a panero of wavy fringes), very often winged, and without a neclary.Third, the pappus is formed by free deciduous bristles (SUSANNA & GARCIA-JACAS, 2003).80th on morphological and molecular grounds, the genera Arclium and COIIsit¡ill belong to the subtribe Carduill11e.More precisely, the Aretium group is part of a large e1ade ineluding also the subtribe Celllaureillae and the genera Saussl/rea Oc. and JI/rillea Cass.(GARCIA-JACAS el aL, 2002).
These entangling cases are illustralive enough, but Lhe besL example 01' lhe problems of generic range in the group is the case of Arcri///1/ and COllsi"ia.
We shall shortly explain lhe maio traits thal have been used in the delimitation of the genera of the Are/j//m group: morphology, karyoiogy, pollen and geography: Morphological characters: leaves, bracls and heads Arcril/m leaves are always unanned, often very big (lO 80 cm) and usally cardate, bOlh characlers being shared by COl/sinia subg.Cy"aroides and Hypacanthodes.In contrast, COl/sinia subg.COl/sinia leaves are small, usually lanceolate and with few exceptions very spiny.
Dispersal is also differenl.80th genera share a [eature infrequent in lhe tri be: achenes are dispersed logelher with the heads.In Arctillm and most of the "Arctioid" group, propagules are usually the whole heads, dispersed by exozoochory by means of the hooks of the involucral bracts.On the contrary, mosl species of COllsillia are turnble weeds, like rnany plants fram the steppe: the whole plant is dispersed by the wind.

Geographic distribution
The genus ArClium has a peculiar dislribulion, with lwo groups wilh marked differences in this respeee In lhe definition by DUISTERMAAT (1996), the species of ArCI;1l1ll sensu striclo (sect.Arc!illm) are subeosmopolitan.The resl of the species (Le., alllhe species previously e1assified in the genus COllsillia) have a cenlral Asian distribution, as lhe gcnus Co//sinia.

Our objectives
The combined analysis of the nrONA region ITS and the chloroplast ONA gene marK has pro ven an excellent tool for unravelling problems at this level in lhe Cardueae (GARCIA-JACAS et aL, 2001, 2002).Thereafter, \Ve have carried out a molecular survey of bolh regions in a sample of aH lhe genera of lhe ArctiulII group \Vith the following goals: 1) To verify lhe monophyly of the ArctiuIIl group 2) To establish the generic boundaries of Arcl;ulII and CO/lsi/l;a 3) To compare nuelear ribosomal and chloroplasl DNA phylogeny and lhe evolulion of lhe palien types, basic chromosome numbers and lTlorphology in lhe group 4) To contribute with a preliminary study lO lhe infrageneric classificalion of the genus COllS;ll;a.

Plant material
Sampling was based on lhe seclional c!assificatioll of lhe genus CO/lsillio by TSCHERNEVA (l988b, 1988c).We have inc1uded 27 taxa of the genus COl/sillia from 19 seclions, lhree species of the genus Arcliul1I, one species oC lhe genus Hypacalllllium and the only species of the monospecific genus Sc!lmalllOlIsellio.The five outgroup species were chosen in the genera Juri/leo and Saussureo according 10 previous sequence analysis (SUSANNA et al., 1995;OARCIA-JACAS el al., 2002) SOLTIS, 1995) primer, as a subslitUle of primer AST-lR.40 eycles of amplification were carried OUl under lhe following conditions: 94°C for 1 minute 30 seconds, 480C for 2 minules and 720C for 3 minules, with an addilional eXlension step of 15 minutes at 72°C.
The double-stranded PCR produets were c1eaned using QlAquiek PCR Purifieation Kit (Qiagen lne.) and sequeneed.Sequeneing primers tmK-7IOF, matK-1848R and AST-lR were used.Oirecl sequencing of lhe amplified ONA segments was performed using a BigDye Terminator Cycle Sequeneing v2.0 (PE Biosystems), following the protoeol recommended by the manufacturero The nucleotide sequencing was performed al the "Serveis Cienlífieo-Téenics" of the University of Barcelona on an AB] PRlSM 3700 DNA Analyzed (PE Biosystems).Nucleolide sequences of /l/a/K were edited using Chromas 1.56 (Teehnelysium Pty Ltd) and easily aligned by hand.
nrONA ITS region slralegies Double-slranded DNA of lhe lTS region was amplified using the l7SE as forward primer and the 26SE as reverse primer (SUN et al., 1994).The profile used for amplifiealion included a warm start at 94°C for 2 minules, followed by 800C for 5 minutes, during which the polymerase {Ecolaq, Eeogen S. R. L.} was added.30 eyeles of amplifieation were carried out under the following condilions: 94°C for 1 minute 30 seconds, 57°C for 2 minutes and 72°C for 3 minutes, with an additional extension slep of 15 minutes at 72°c' The PCR produets were purified wilh the QlAquiek PCR Purification Kil (Qiagen lne.).OfuRa BOlanieal Garden (Japan) AY373751 AY373682 Herder AY373718 (SWOFFORD, 1999) using TBR branch swapping wilh character states specified as unordered and unweighted.The indels were coded as fiflh base.AH most-parsimonious trces (MPTs) were saved.To locate other potential islands of most-parsimonious trces (MADDISON, 1991), we performed 1000 replications with random taxon addition, also with TBR branch swapping.Bootstrap analyses (BS) were performed (FELSENSTE1N, 1985) wilh 1000 replicates, and decay indices (DI) were calculated (BREMER, 1988;DONOGHUE el al., 1992) to obtain estimates of support for each monophyletic group.AH the decay analyses were conducted using the elade-constraint approach as discussed in MORGAN (1997).Three parsimony analyses were performed, with three different data sets: the ITS dala, the mQtK data and Ihe combined ITS and mal K data.
The nrDNA ITS and cpDNA matK data sets were tesled for congruence using the partition homogeneity test (FARRIS et aL, 1995) as implemented in PAUP 4.0b4a, before combining the data sets.The partition homogeneity test was conducted with 1000 replicates, heuristic search option with simple addilion sequence, TBR, and MULPARS.
The parsimony analysis yielded 91 MPTs of346 steps in one island.Thc strict conscnsus of aH the trees is shown in Fig. 1; the consistency index (el) exeluding uninformative eharaeters was 0.4780; the retenlion index (RI) was 0.7570; and the homoplasy index (HI) was 0.5220.

Combine<! nrONA ITS and cpONA lI/olK
The P-value resulting fram the partition homogeneity test (P = 0.0570) ¡ndieates (hat data partitions are random, and there is congruence between nrONA ITS and cpDNA IIIatK data seIS, al a significance lhreeshold of P = 0.05 (FARRIS el al., 1995).
The parsimony analysis yielded 8 MPTs of 329 steps in one island.The strict consensus of all the trees is shown in Fig. 3; lhe eonsistency index (el) exeluding uninforrnative charaeters was 0.5650; Ihe relenlion index (RI) was 0.7895; and Ihe homoplasy index (HI) was 0.4350.

D1SCUSSION
Monophyly of lhe Arctium group The use of a combinalion of /1JalK and lTS data has proven again a good choice for our study: the resulting tree from the analysis of the combined dala shows excel1enl resolulion at all taxonomical levels (Fig. 3).The matK alone is not suitable for our purposes (Fig. 2)  as il even fails in supporting monophyly of lhe group (SS; 78% and DI ; 2 only).However, it is worlh remarking thallhe l1IatK analysis reflects lhe cul between Ihe "Arctioid" and lhe Colts;llia e1ades.even with poor support (SS:;; 71%, DI:;; 1).This demonstrates thar lhe separalion of both groups is very old, as could be expected from lhe different pollen types and chromosome numbers.
Character evolution in the Areliu", group

Habil aud morphology
Our results suggest that the spiny habil is a derived character in the group.We can recall lhal OARC1A-JACAS el al. ( 2002) plaeed lhe An:lilllll group in a clade fonned by lhe subtribe Centallreinae and ils polential sisler groups, the genera Jurinea and Sallss/lrea.Wilhin Ihis group, unarmed habit is overwhelmingly dominant: aH the species of Jurinea and Saussl/rea are unarmed, and spines are a very rare fealure among the Centallre;nae: the only spiny genera in lhe subtribe are those of the Carthamlls complex, which are a derived group (GARCIA-JACA5 el al., 200 1), a fact thal reinforces the hypothesis of the derived character of spilles.Anceslors of the Arctillm group were probably unarmed plants lhat developed spines, like the genera of lhe CartlulI1J11s complex, as a secondary adaptation lO the defence againsl herbivores.In lhe case of COllsinia, spines have appeared twice in different lineages (Figs.1,3): one in COllsin;a subgenus COllsinia and a second one in Hypaeallthillm and Schmalhallse"ia.

Pollen and chromosomes
We shall discuss together both characters as they appear to be correlaled.As has been repeatedly demonslraled in the e10sely related Cemallreinae, pollen and karyology are the mosl reliable characters for phylogenetic reconstruction and usually show also a narrow eorrelalion lo molecular phylogenies (SUSANNA el al., 1995;OARCtA-JACAS el al., 2ool).In the Arcti",lI group.Ihe correlation is a1so very narrow and Ihey are the only ones that define truly monophylelie groups (Figs. 1, 3).
As lo chromosomal evolution, Ihe "Arctioid" e1ade is also ancestral to the Cousinia e1ade in this character.It is generally accepted lhat higher chromosome numbers are ancestral as compared lO lower ones, and this Irend has been demonSlrated in groups e10sely related to the Arclillfll group like subtribe Cenlaureinae (GARCIA-JACA5 et al., 2001).In our case, the base chromosome number x :;;; 18 of the "Arclioid" e1ade is ancestral as compared to lhe • ; ll, 12 or 13 of lhe COIISillio clade (Figs. 1, 3).

Generic implications
The delineation of Aretium and Cousinia Our results pose a very inleresling but vexing problem on the generic delineation of Aretillm and COlIsinia: how to conjugate conflicting morphological, on the one hand.and molecular, pollen and karyological dala, on lhe olher hand.
Molecular analyses suggest a clear separation (and a very ancient one, as renected by the matK data) between the "Arctioid" clade and COlIsinia sensu slriclo, strongly supported by pollen and karyological dala (Figs. 1, 2. 3).The problem is, lhen, whal lO do wilh Hypacamhium and Schmalhausenia.On strict molecular, pollen and karyological grounds, they belong without any doubt 10 {he "Arctioid" dade... whose homogeneity they destroy on morphological grounds!As stated in the introduClion of this papero ¡he only useful characlers for segregating ArClium and COltsinia are (he Icaves and, partly.(he appendages of the bracls.Pollen types and karyology are excellent markers, hUl lhey are nol macroscopic and henee of lillle praelieal utilily.Morphology of Ihe stigmas eould be useful (il should be rully explored in COlls;llia s. sIr.), hUI il is very difficult lo observe too.OUT own observations suggesl that there are sorne differences in length and distribution of (he collector hairs: lhey are long and sparsely distributed along the whole lenglh of the stigma in COllsi/lia s. 8lT., and short and concentrated on the basal thickening in Arctium.Schmalhausen;a is intermediate belween bOlh models, but c10ser to Aretium.However, from our point of view, differences are nol so substantive as c1aimed by DUlSTERMAAT (1996).
We have to resort to lhe leaves and, if we follow this morphological definition, Hypaca1lth;1l111 and ScllIllalhal/sell;a (probably a single genus) wauld be c1assified in COIIS;Il;a sensu stricto, conlradicting molecular, karyological and pollen evidence.There are three possible solutions to this incongruence.
The first ane should be to keep present c1assification in four genera with a different delimitation: Aret;ulII (ineluding COIls;llia subgenera Cynaro;des and Hypacallthodes), COlls;1lia, Hypaca1lthillm and Schmalhal/sen;a, disregarding the fact of the polyphyly of the resulting genus Aret;um (Figs. 1,3).This solution does nal seem adequate, if our goal is lO achieve a natural c1assification.
A second altemative would be to elevate the subgenera of COlIs;n;a wilh Arct;aSTrlllll palien type to the genus rank.The resulting Arct;um group would encompass six genera: Arct;ut1l sensu str., COlls¡'lia s. str., Cy"aro;des, Hypaca1lth;um, Hypacallthodes and Schmalhallsenia.The possible genus CY1laroides stands as a very well defined group on molecular grounds (SS = 100%, DI = II in Ihe ITS lree, Fig. 1; SS = 100%, DI = 12 in Ihe combined tree, Fig. 3), and our sampling is complete enough as to draw conelusions (five oul of seven seelions of Ihe subgenus, according lO TSCHERNEVA'S (l988b, 1988e) c1assification).In contrast, our sampling of sect.Hypacallthodes is too short (one out of four seclions) and the only included species forms a well•supported c1ade with Hypacalllh;lIm and Sehllla/hal/senia (SS = 93%, DI = 5 in the ITS Iree, Fig. 1; SS = 90%, DI = 5 in Ihe combined Iree [withoul Hypacallth;lIm], Fig. 3), a grouping lhal needs verificmion.However, we can anticipate that the problem of making up these six genera ¡s, again, morpho• logical.The boundaries between Hypacalllh;u11l and Schmalhallsen;a, on the one hand, and Cous;/I;a sensu str., on the other hand, are unc1ear: pollen type and karyology are the only reliable characters thal separale lhem.The stigma characters could be of help, bUl further studies are needed in COIIS;ll;a.Besides, the limits between Arct;u11I and a new genus Cy"aroides would also be uncertain on slrict morphological grounds because subg.Cy"aroides ¡neludes the arctioid species of Cous;nia.
Finally, Ihere is Ihe solulion suggesled by KUNTZE (1891): lO lump logelher alllhe genera described in lhe Arcliu11I group in a single genus Arctillm (as this genus would be nomcnelaturally prioritary).We are conscious of the nomenclatural consequences of this solution: sorne 420 species of Cousinia will have to be changed to Areti/l'" -KUNTZE (1891) already proposed ca.180 new combinations•.Nevertheless, if new studies in the group (especially with the ¡nclusion of mOfe species of CO/ls;/lia subgenus Hypacallthodes) confirm our results, it will probably be the only solution.

in the field, or fresh ¡caves of plants cultivated in the Botanic Institute of Barcelona. In sorne cases, herbarium material was used. cpDNA malK gene strategies
Double-stranded DNAs of matK were amplified by PCR with tmK-71OF (JOHNSON & SOLTIS, 1995) and AST-I R {OARCJA-JACAS el al., 2002} as PCR primers.Sometimes this combinalion