Incongruent plastid and nuclear DNA phylogenies reveal ancient intergeneric hybridization in Pilosella hawkweeds (Hieracium, Cichorieae, Asteraceae)

Molecular Phylogenetics and Evolution 42 (2007) 347–361 www.elsevier.com/locate/ympev

Incongruent plastid and nuclear DNA phylogenies reveal ancient intergeneric hybridization in Pilosella hawkweeds (Hieracium, Cichorieae, Asteraceae) Judith Fehrer a

a,*

, Birgit Gemeinholzer

b,1

, Jindrˇich Chrtek Jr a, Siegfried Bra¨utigam

c

Institute of Botany, Academy of Sciences of the Czech Republic, 25243 Pru˚honice, Czech Republic b Institute of Plant Genetics and Crop Plant Research, 06466 Gatersleben, Germany c Staatliches Museum fu¨r Naturkunde, Postfach 300154, 02806 Go¨rlitz, Germany Received 21 October 2005; revised 28 June 2006; accepted 7 July 2006 Available online 20 July 2006

Abstract Phylogenetic relationships for Hieracium subgen. Pilosella were inferred from chloroplast (trnT–trnL, matK) and nuclear (ITS) sequence data. Chloroplast markers revealed the existence of two divergent haplotype groups within the subgenus that did not correspond to presumed relationships. Furthermore, chloroplast haplotypes of the genera Hispidella and Andryala nested each within one of these groups. In contrast, ITS data were generally in accord with morphology and other evidence and were therefore assumed to reflect the true phylogeny. They revealed a sister relationship between Pilosella and Hispidella and a joint clade of Hieracium subgenera Hieracium and Chionoracium (Stenotheca) while genus Andryala represented a third major lineage of the final ingroup cluster. Detailed analysis of trnT–trnL character state evolution along the ITS tree suggested two intergeneric hybridization events between ancestral lineages that resulted in cytoplasmic transfer (from Hieracium/Chionoracium to Pilosella, and from the introgressed Pilosella lineage to Andryala). These chloroplast capture events, the first of which involved a now extinct haplotype, are the most likely explanation for the observed incongruencies between plastid and nuclear DNA markers. Ó 2006 Elsevier Inc. All rights reserved. Keywords: Andryala; Apomixis; Asteraceae; Chloroplast capture; Hieracium; Incongruence; Intergeneric hybridization; Internal transcribed spacer; matK; Molecular evolution; Molecular phylogeny; Pilosella; Polyploidy; Speciation; trnT–trnL

1. Introduction Hybridization plays a major role in angiosperm speciation (Stebbins, 1971; Arnold, 1992; Rieseberg and Carney, 1998). This is especially true for groups with a high number of polyploids such as Hieracium L. s.l. Hawkweeds of the subgenus Pilosella on which we focus in this study show a mixture of sexual and facultatively apomictic taxa that exhibit a heavily reticulate pattern of relationships according to morphology and field observations. This led the *

Corresponding author. Fax: +420 267750031. E-mail address: [email protected] (J. Fehrer). 1 Present address: Botanic Garden and Botanical Museum BerlinDahlem (FU Berlin), Ko¨nigin-Luise-Str. 6–8, 14191 Berlin, Germany. 1055-7903/$ - see front matter Ó 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2006.07.004

authors of the monographs on Hieracium s.l. (including subgenera Hieracium, Chionoracium DUMORT. (syn. Stenotheca (MONN.) TORR. et GRAY), and Pilosella (HILL) S.F. GRAY) to develop a classification and species concept attempting to reflect this situation (Na¨geli and Peter, 1885; Zahn, 1921–23, 1922–38). They defined species in a broad sense and distinguished between ‘basic’ and ‘intermediate’ species. While basic species feature unique morphological characters, intermediates are defined by a combination of morphological characters of two or more basic species, and are supposed to be of hybrid origin. British authors (e.g., Sell and West, 1976) treat most intermediate species as hybrids. In this article, we will follow the classification of the monographs.

348

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

Hybridization between species of Pilosella is common (e.g., Houliston and Chapman, 2001; Morgan-Richards et al., 2004), and apomixis (asexual seed production) in this group is a facultative trait, i.e., even apomicts can introgress into other taxa (e.g., Chapman and Bicknell, 2000; Krahulcova´ and Krahulec, 2000). The whole subgenus can be considered as a large, heavily reticulate complex that comprises about 200 (macro)species. In order to study species relationships in such an intricate group, we focused on basic species (ca 25) as they are supposed to be the precursors of the multitude of intermediate taxa. Some basic species comprise diploids, but different ploidy levels within single species are common (Schuhwerk, 1996; Krahulcova´ et al., 2000). Interspecific hybridization is one of the major factors leading to phylogenetic incongruence among loci. This is especially true for plastid and nuclear genomes with a maternal or biparental mode of transmission. The most extreme case is chloroplast capture where the cytoplasm of one species is replaced by that of another species through hybridization/introgression, a process that can occur at a variety of taxonomic levels (Rieseberg and Soltis, 1991). As a result, clustering taxa on the basis of chloroplast DNA often does not correspond to taxonomic units, neither to groups supported by analysis of morphological characters, nor to groups indicated by nuclear markers (e.g., Soltis and Kuzoff, 1995; Smissen et al., 2004). In order to avoid erroneous phylogenetic conclusions based on cpDNA data, comparisons with phylogenetic hypotheses based on nuclear gene sequences as well as comprehensive sampling are required (Rieseberg and Soltis, 1991). In a previous molecular study on a selection of Pilosella species co-occurring in a rather small area (Fehrer et al., 2005), we discovered rather high intraspecific chloroplast haplotype uniformity and a general lack of geographic pattern. Actually, species usually belonged to one of two divergent haplotype groups that showed no correlation to species morphology. In order to investigate whether this result was an artifact of species sampling or due to the selected chloroplast marker (trnT–trnL intergenic spacer), for the present study, we sampled all Pilosella basic species, added a second cpDNA marker (matK), and compared the results with a molecular phylogeny of the internal transcribed spacer (ITS) region of nuclear ribosomal DNA. For outgroup comparison, the other Hieracium subgenera as well as closely and more distantly related genera of the Cichorieae (Asteraceae) were employed. We present here what we think is the simplest model of the complex evolutionary history for Pilosella hawkweeds that explains all currently available data. It provides an example of how detailed analysis of molecular evolution in differently inherited gene regions can be applied to elucidate speciation processes. The pattern of relationships revealed here shall serve as a framework for further studies.

2. Materials and methods 2.1. Plant material and ploidy level determination All basic species of subgenus Pilosella (sexual as well as apomictic) were sampled. Species with different degrees of presumed relationships were used as preliminary outgroups: (1) the subgenera Hieracium and Chionoracium, represented by a selection of species covering a reasonable range of sections and geographic areas, (2) all accessible species from the most closely related genus Andryala (Stebbins, 1953), (3) the monotypic west-Mediterranean Hispidella hispanica, which Jeffrey (1966) associated to Hieracium, (4) Tolpis virgata and Tolpis proustii as representatives of the Hieraciinae sensu Bremer (1994), (5) Koelpinia turanica and Cichorium intybus due to their placement in preliminary ITS trees of the Cichorieae (Gemeinholzer in preparation), and (6) Mycelis muralis due to its similarity in the alignment of areas of structural mutations in the chloroplast trnT–trnL region. The majority of the material originated from living plants collected in 1999–2003. Most were cultivated in the experimental garden at Pru˚honice for at least some time. Specimens of either the individual plant (whenever possible) or from the same locality were deposited in herbaria (see Table 1). Species were complemented by up to 30year-old herbarium specimens from several sources. All determinations were done or confirmed by at least one Hieracium specialist. A detailed list of species is given in Table 1. In a group as rich in polyploids like Hieracium—especially if frequently different cytotypes are known to occur in same species—knowledge about the ploidy level is important in molecular studies for proper marker application and result interpretation. We used chromosome counts as well as flow cytometry, either alternatively or combined. Chromosome counts were performed from root tip metaphases as described previously (Mra´z et al., 2005); flow cytometry analyses were done according to Bra¨utigam and Bra¨utigam (1996) and Krahulcova´ et al. (2004) using diploid Hieracium lactucella as an internal standard. Where only herbarium material was available, ploidy information was inferred from Schuhwerk (1996; and unpublished data). 2.2. Molecular methods DNA was isolated from fresh, silica-gel dried, CTABpreserved or herbarium material either by sorbitol extraction (Sˇtorchova´ et al., 2000) or by use of the DNeasy kit (Qiagen, Hilden, Germany). Amplifications of the trnT–trnL intergenic spacer were done in 50 ll-reactions containing 2.5 mM MgCl2, 200 lM of each dNTP, 1 lM of primers a and b (Taberlet et al., 1991), a few nanograms of genomic DNA, and 1 unit of Taq-DNA polymerase (Fermentas, Ontario, Canada). After 2 min of predenaturation at 94 °C, 35–40 cycles of

Table 1 List of species Species

Hieracium subgen. Pilosella H. alpicola SCHLEICHER ex STEUDEL & HOCHST. ssp. ullepitschii (BŁOCKI) ZAHN H. angustifolium HOPPE [H. glaciale REYN. ex FR.] H. argyrocomum (FR.) ZAHN

H. castellanum BOISS & REUTER H. caucasicum NA¨GELI & PETER H. cymosum L. ssp. cymosum H. cymosum L. ssp. cymigerum (RCHB.) PETER H. echioides LUMN. H. hoppeanum SCHULT. H. lactucella WALLR. [H. auricula LAM. & DC.] H. macranthum (TEN.) TEN. H. onegense (NORRL.) NORRL. [H. caespitosum DUMORT. ssp. brevipilum (NA¨GELI & PETER) P.D. SELL] H. pavichii HEUFF. H. peleterianum ME´RAT

H. pilosella L. H. piloselloides VILL. ssp. obscurum (RCHB.) ZAHN H. piloselloides VILL. ssp. praealtum (VILL.) ZAHN H. procerum FR. H. pseudopilosella TEN. ssp. tenuicaule NA¨GELI & PETER

Ploidy

Identifiers

GenBank Accession Nos.i trnT–trnL

matK

ITS

AJ633181

AJ633401

Vysoke´ Tatry Mts/Slovakia, herbarium P. Mra´z

2xa

pic1141, GAT-bg166

AY573309

Col de Valbelle, De´p. Hautes-Alpes/France, herbarium GLM 155905 Prov. Granada/Spain, Bot. Garden Munich, Merxmu¨ller & Gleisner, culture H11 Krkonosˇe/Czech Republic, herbarium PRA Oberlausitz/Germany, herbarium GLM38443 Lac de Bouillouses/France, herbarium M, H. Merxmu¨ller 26985 & B. Zollitsch Krkonosˇe/Czech Republic, herbarium GLM 157788

2xe

ang.Fra, GAT-bg265

AY573315

AJ633407

2xa

agy.Gra, GAT-bg242

AY573320

AJ633395

4xb 5xd 2xe

aur.Spb.1 bau.AM.1 brc.Bou, GAT-bg188

AY192646g AY192647g AY573317

4xc

glo.Gau.1

AY192650g

Erzgebirge/Germany, herbarium GLM157765 Sierra Nevada/Spain, herbarium GLM 156412 Kayseri/Turkey, herbarium M, Buttler 13880 Louny/Czech Republic, herbarium PRA Krusˇne´ Hory/Czech Republic, herbarium GLM 157800

5xc 2xe 4xe 2xa 4xc

glo.Fuer.4 cas.Nev, GAT-bg187 cau.Kay, GAT-bg189 cym.12/4, GAT-bg106 cymi.Wbs.A

Go¨hrisch near Meißen/Germany, herbarium GLM 101286 Upper Egger-Alm, Alps/Austria, herbarium PRA Erzgebirge/Germany, culture Go¨rlitz, herbarium GLM 140613 Jonsdorf, Oberlausitz/Germany, herbarium GLM 140619 Nitrica/Slovakia, herbarium PRA Krkonosˇe/Czech Republic, herbarium GLM157784

2xc

AJ633174

AJ633393

AY192649g AY573316 AY573310 AY192652g AY192653g

AJ633173 AJ633183 AJ633178

AJ633392 AJ633403 AJ633398

ech.Goer.1, GAT-bg105

AY192655g

AJ633177

AJ633397

2xa 2xe

34hop, GAT-bg239 la.GR

AY192665g AY192670g

AJ633184

AJ633405

2xe 2xa 2xc

lac.Jon.1, GAT-bg102 137mac, GAT-bg240 caeb.Jbo.2, GAT-bg104

AY192669g AY192671g AY192651g

AJ633170 AJ633185

AJ633389 AJ633406 AJ633396

2xe

pav.Oly, GAT-bg168

AY573312

Mt. Olympos/Greece, herbarium M, A. Strid & S.O. Hansen 9638 Bavarian Forest/Germany, herbarium M, F. Schuhwerk 96/1 Kanton Wallis/Switzerland, herbarium GLM 155337 Krkonosˇe/Czech Republic, herbarium GLM156942

2xa,e

pel.Bay

AY573321

2xe 4xd

Oberlausitz/Germany, herbarium GLM38421

4xd

pel.Wal, GAT-bg249 pla.Jbo.1 pla.Jbo.2 poio.AM.1

AY573322 AY192674g AY192673g AY192679g

Go¨rlitz/Germany, herbarium GLM140620

5xd

poip.GRG.5

AY192680g

¨ ren/Turkey, herbarium M, M. Nydegger 46839 SW of O Prov. Civdad Real/Spain, herbarium GLM 153084

4xe 2xe

pro.Oere, GAT-bg164 pse.Civ, GAT-bg169

AY573313 AY573318

AJ633400

AJ633186

AJ633504

AJ633182 AJ633171

AJ633402 AJ633390

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

H. aurantiacum L. H. bauhini SCHULT. H. breviscapum DC. [H. candollei MONN.] H. caespitosum DUMORT. [H. pratense TAUSCH]

Source

(continued on next page) 349

Species

H. saussureoides (ARV.-TOUV.) ST.-LAG. [H. tardans PETER, H. niveum (MUELLER ARG.) ZAHN] H. vahlii FROEL.

H. eriophorum ST.AMANS H. fritzei F.W. SCHULTZ H. hryniawiense WOŁ. [H. crocatum FR. s.l.] (H. conicum group) H. humile JACQ. H. intybaceum ALL.

H. lachenalii C.C. GMEL H. magocsyanum JA´V. (H. sparsum group) H. nigrescens WILLD. ssp. decipiens (TAUSCH) ZAHN H. pannosum BOISS. H. piliferum (HOPPE) s.str. H. pojoritense WOŁ H. prenanthoides VILL. H. sabaudum L. H. schmidtii TAUSCH ssp. schmidtii H. sparsum s.str. FRIV. H. umbellatum L. H. transsilvanicum HEUFF. H. villosum JACQ. Hieracium subgen. Chionoracium H. albiflorum HOOK H. antarcticum D’URV. H. aff. asplundii SLEUMER H. carneum GREENE H. frigidum WEDD.

Source

Ploidy

Identifiers

GenBank Accession Nos.i trnT–trnL

matK

ITS

niv.Lda, GAT-bg165

AY573311

AJ633172

AJ633391

2xa

vah.Sor, GAT-bg241

AY573319

AJ633175

AJ633394

4xe

vcl.Erz, GAT-bg170

AY573314

AJ633179

AJ633399

Hruby´ Jesenı´k Mts./Czech Republic, herbarium PRA Chornohora Mts., Ukrainian Carpathians/Ukraine, herbarium PRA Unie˛cice (Meffersdorf)/Poland (neophyte), herbarium GLM 44502 Biele Karparty/Slovakia, culture MS B 11/6, herbarium PRC B& T World Seed, http://www.b-and-t-world-seeds.com/index.html, herbarium GLM 157756 near Arcachon/France, Bot. Garden Munich, culture 94-102 Krkonosˇe/Czech Republic, herbarium PRA Mt. Pikuj/Ukraine, P. Mra´z culture 1430

3xa 2xa 3x or 4x 3xa 3xa,c

alp 9/14 alp.Ukr, GAT-bg197 amp.Mef bup.Bik canad

AY573344 AY512556h AY573331 AY573338 AY573341

AJ633201

AJ633429

2xa 3xa 2xa

bg49, GAT-bg49 schn 1/1 hry.Ukr, GAT-bg310

AY573337 AY573342 AY573340

AJ633196

AJ633424

Steiermark/Austria, herbarium PR Oberengadin, SSE St. Moritz/Switzerland, herbarium GLM 152405 Ka¨rnten/Austria, herbarium M, S. Jagalski 4 B & T World Seed, http://www.b-and-t-world-seeds.com/index.html, herbarium PRA Podyjı´, South Moravia/Czech Republic, herbarium PRA Retezat Mts/Romania, herbarium P. Mra´z

3xe 2x, 3x or 4x 2x, 3x or 4x 2xa,c

hum.Ste inb.Eng, GAT-bg266 inb.Kaer, GAT-bg192 H.intybac, GAT-bg311

AY573329 AY573323 AY573324

3x? 3xa

lai.Luk spa1058, GAT-bg194

AY573336 AY573332

Krkonosˇe/Czech Republic, herbarium PRA

4xa

dec 1/1

AY573343

Nevsehir/Turkey, herbarium PR Vysoke´ Tatry Mts/Slovakia, P. Mra´z, culture 1146 Pojorita, Cimpalung Moldovec/Romania, herbarium P. Mra´z near Scho¨nblick, Alps/Austria, herbarium GLM 147987 SE Scho¨nau-Berzdorf/Germany, herbarium GLM 46890 ˇ eske´ Strˇedohorˇ´ı/Czech Republic, herbarium PRA Hill Borecˇ, C

3x or 4x 4xa 2xb 4xc 4xb 4xa

pan.Nhi pif1146 poj.Rom.1, GAT-bg308 pre.Alp, GAT-bg262 sa.AM.1 smi.Ces.2

AY573325 AY573339 AY573328 AY573326 AY573334 AY573327

Pirin Mts, Vihren Mt./Bulgaria, garden culture Z. Szela˛g, herbarium PRA SE Scho¨nau-Berzdorf/Germany, herbarium GLM 46889 Borsßa/Rumania, herbarium P. Mra´z 1066 Mala´ Fatra/Slovakia, culture MS V 12/3, herbarium PRC

2xe 2xe 2xb 2xb,d 3xa

spa.sst.1, GAT-bg267 spa.sst.2 um.AM.1 tra.Boa, GAT-bg193 vil.Fat

AY573333 AY573335 AY512557h AY573330

B&T World Seed, http://www.b-and-t-world-seeds.com/index.html Patagonia National Park, Torres del Paine/ Chile, herbarium GLM 155579 Prov. Murillo, Dpto La Paz/Bolivia, herbarium M, Beck 14779 Tucson, Arizona/USA, herbarium PRA Prov. Tungurahua, P.N. Llangantes/Ecuador, det. Schuhwerk, herbarium PRA

2xf 2xf

albifl, GAT-bg111 ant.2, GAT-bg109

2xf 2xa,c,f 2xc,f

asp.Bol.2, GAT-bg243 H. carneum.1, GAT-bg309 Chi.Ecu

Prov. Leida, Pyrenees/Spain, herbarium GLM 153085

e

4x

Prov. Soria/Spain, Bot. Garden Munich, H. Merxmu¨ller & W. Lippert, culture H43 Prov. Erzurum/NE Turkey, herbarium GLM 154133

AJ633415

AJ633198

AJ633430 AJ633426 AJ633414

AJ633428

AJ633202

AJ633412 AJ633417

AJ633203

AJ633431

AJ633199

AJ633427

AY573352 AY573350

AJ633189 AJ633188

AJ633418 AJ633409

AY573348 AY573351 AY573346

AJ633191

AJ633420 AJ633413

AJ633194

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

H. verruculatum LINK Hieracium subgen. Hieracium H. alpinum L. ssp. alpinum H. alpinum L. H. amplexicaule L. H. bupleuroides C.C. GMEL. H. canadense MICHX.

350

Table 1 (continued)

H. glaucifolium POEPP. ex FROELICH H. cf. patagonicum HOOK H. scabrum MICHX. H. stachyoideum ARV.-TOUV. H. trichodontum (SCHULZ-BIP.) ARV.-TOUV. H. venosum (L.) WILLD.

Andryala glandulosa LAM. s.str. Andryala integrifolia L.

Andryala levitomentosa (NYA´R.) P.D. SELL Andryala pinnatifida AIT. Andryala pinnatifida AIT. Andryala ragusina L. Andryala varia DC. Hispidella hispanica BARNADES ex LAM. Tolpis proustii PIT. Tolpis virgata BERTOL. Cichorium intybus L. Koelpinia turanica VASS. Mycelis muralis (L.) DUMORT.

2xf

bg53, GAT-bg53

Refugio Redeto/Chile, Bot. Garden Munich, culture no. 94-55 Cobequid Hills, Nova Scotia/Canada, herbarium GLM 152506 Prov. Murillo, Dpto Cochabamba/Bolivia, herbarium M, Beck 14565 & Seidel Prov. Camacho, Dpto La Paz/Bolivia, herbarium M, leg. Beck 4233 Michigan/USA, herbarium MICH, coll. Reznicek 11266, herbarium GLM 141161

2xa,f

pat.Chi, GAT-bg263

AY573349

2xf

sca.can

AY573353

2xf

sty.Bol, GAT-bg244

2xf 2xf

Sierra de la Sagra, Prov. Granada/Spain, herbarium B, col.: Valdes et al. 1916/1988 Sierra Nevada/Spain, herbarium PRA Ponta do Pargo/Madeira, herbarium GLM 148659 Malaga/Andalusia/Spain, herbarium GLM 141138 Botanic Garden Potsdam, 66-H02 St. Segal, Finistere/France, herbarium P, 96-184-63/02 Porquerolles Var/France, herbarium P, 98-62-63/02 Pietrosul Bogolin/Romania, herbarium GLM 156367 El Cercado/La Gomera/Spain, herbarium GLM 158131 Elpino/La Gomera/Spain, herbarium GLM 158080 San Jose´/SE-Andalusia/Spain, herbarium GLM 141140 Levada nova, Prazeres/Madeira, herbarium GLM 148639 Sierra de Guadarrama/Spain, herbarium M, leg. J. Pizarro et C. Navarro no CN 2460 La Gomera/Spain, herbarium GLM 158013 Bivio di Lavi/Aurelia (Grosseto)/Italy Bot. Garden of Masaryk Univ. Brno 92-25/02 Aschrabat Karakum, between Anai and Giaursom/ Turkmenistan Go¨rlitz/Germany, herbarium GLM 46891 Botanic Garden of Friulano, Udine/Italy, 144/02

AJ633408 AJ633193

AJ633416

AY573347

AJ633192

AJ633421

tri.Bol, GAT-bg167

AY573345

AJ633190

AJ633419

venosum GAT-bg307 GAT-bg110

AY573354 AJ633411 AJ633410

2xf

GAT-bg198

2xc 2xc,f 2xd,f 2xb,f 2xb,f 2xb,f 2xb,d,f

A.agaJF A.glan.Mad.1 And.int.2/2 GAT-bg17 GAT-bg35 GAT-bg36 A.lev.maj.1, GAT-bg107

AY573363 AY573356 AY573360

AY573362

AJ633167

AJ633382 AJ633383 AJ633384 AJ633385

2xd,f 2xd,f 2xf 2xf

And.pin.Cer, GAT-bg108 And.pin.Elp ArSJ, GAT-bg246 A.var

AY573358 AY573359 AY573361 AY573357

AJ633168

AJ633386

AJ633169

AJ633388

2xf

His.his.2, GAT-bg245

AY573355

AJ633205

AJ633433

2xf 2xf 2xf 2xf

Tol.prs, GAT-bg196 bg64, GAT-bg64 cich166, GAT-cich166 bg142, GAT-bg142

AY573307 AY573308 AY573305 AY573304

AJ633210 AJ633207 AJ633132 AJ633264

AJ633439 AJ633435 AJ633451 AJ633491

2xf 2xb

Myc.mur, GAT-bg247 GAT-bg101

AY573306

AJ633236

AJ633387

AJ633165

AJ633339

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

other genera Andryala agardhii DC.

Chile, leg. J. Grau, herbarium M

a

Chromosome count for the individual plant. Chromosome count on a plant from the same locality. c Flow cytometry for the individual plant. d Flow cytometry on a plant from the same locality. e Only this ploidy level described for the (sub)species. f Only this ploidy level described for the taxon. g Sequence from Fehrer et al. (2005). h Sequence from Mra´z et al. (2005). i Whenever two identifiers are given for the same species and locality, the same plant was used for the respective sequences indicated by GenBank accession numbers. ?, not tested, probaly triploid. b

351

352

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

1 min each at 94, 46 and 72 °C were performed, followed by a final extension at 72 °C for 10 min. PCR products were either purified directly or after excision from a preparative gel using Qiagen Gel extraction and PCR purification kits. Generally, both strands were sequenced (GATC Biotech, Konstanz, Germany) using the PCR primers and, in case of complications due to secondary structure and polynucleotide stretches, additional sequencing primers a+ (50 -aagagagacagatgtatagc-30 ) and b+ (50 -tatacatctgtctctcttcc-30 ). In a few cases of unambiguous sequence of one strand and in absence of unique mutations, sequencing of the second strand was omitted. All positions were unambiguously resolved. Chromatograms were edited manually. For matK-fragment amplification, primers trnK 710F, trnK2R (Johnson and Soltis, 1995) and AST-1R (GarciaJacas et al., 2002) were utilized, and Cichorieae-specific primers were designed (matK-RL 50 -gacygcgtmccattgaag0 3, matK-iF 50 -taccttacccagcccatctg-0 3, matK-iR 50 -aaatgcaaagaggaagcatct-0 3). PCR conditions and sequencing were the same as for ITS (see below). For ITS amplification, primers ITS-A and ITS-B (Blattner, 1999) were used; if necessary, primers ITS-C and ITSD (Blattner, 1999) or ITS-SF and ITS-SR (Blattner et al., 2001) were applied. The PCR mix contained 0.25–0.5 lg of DNA, 0.2 U Taq DNA polymerase, 5 ll 10x buffer, 2 ll dNTP, 10 ll Q-solution (all Qiagen) and 0.5 ll of each primer (50 pmol/ll), added up with H2O to a total of 50 ll. An initial denaturation step at 95 °C for 3 min was followed by 32 cycles of annealing at 52 °C for 30 s, extension at 70 °C for 45 s and denaturation at 95 °C for 30 s, and a final extension at 70 °C for 8 min. PCR products were purified with the QIAquick PCR purification kit and served as templates for cycle sequencing with the ABI RhodamineMix Kit. An ABI 377 sequencer was used to generate the data. All GenBank accession numbers are listed in Table 1. 2.3. Data analyses All phylogenetic analyses were calculated with PAUP* 4b10 (Swofford, 2002). Several approaches were used, employing different reasonable indel/gap treatments, optimality criteria, outgroup combinations, and models of molecular evolution. Rough alignment of the trnT–trnL region was performed with CLUSTAL_X (Thompson et al., 1997) and further processed in BioEdit (Hall, 1999). Manual adjustments of indel positions were arranged according to repeat structure, i.e., gaps were opened so that the inserted parts formed duplications of adjacent motifs. For a hypervariable poly(A) region, only one gap of variable length was allowed. It was always opened at the variable 50 -end before the first A in order to ensure consistent treatment. A matrix coding the indels as single mutation events was constructed and appended to the sequences in order to avoid giving longer indels more weight (or none, if gaps were treated

as missing data) and to account for intricate features like e.g., nested insertions. For character coding, the four nucleotides were used as multiple character states. Indels of identical position and length were considered homologous (Lloyd and Calder, 1991). Alignment and matrix are available upon request. Phylogenetic analyses of the trnT–trnL were performed (1) including all characters (gaps treated as missing data), (2) using the coded matrix instead of indels, and (3) using the matrix, but excluding the poly(A) region. All analyses were done by random addition of taxa. Identical sequences were represented only once in order to reduce computing time. Maximum parsimony (MP) analyses were performed as heuristic searches with 100 random addition replicates and tree bisection-reconnection (TBR) branch swapping, keeping no more than 100 trees of length greater than or equal to 1 in each replicate. For distance analyses, HKY85 or uncorrected P-distance measures were used in combination with either the minimum evolution (ME) or the least squares objective in heuristic searches. For maximum likelihood (ML) analyses, the model of sequence evolution fitting the data best was determined with Modeltest, version 3.06 (Posada and Crandall, 1998). A TVM + G model best fit the data (based on the alignment without matrix). It was applied along with the Modeltest-estimated parameters in heuristic searches with one addition sequence replicate and TBR branch swapping. Bootstrapping was performed for the three different character treatments in MP analyses, for one of the distance analyses, and for ML analysis with the coded character matrix without the poly(A) part, running 1000 replicates each with one random addition sequence without branch swapping. Trees were unrooted. For matK phylogenetic analyses, a representative subset of taxa was used for comparison, based upon the results of the trnT–trnL analyses. Alignment of the matK sequences was straightforward. A few taxa showed an insert of 9 bp (three triplets), which was treated as a single mutation event. Preliminary analyses revealed a similar topology as for the trnT–trnL. Therefore, and due to low matK overall variability, the two chloroplast DNA datasets were combined. Analyses were done as described above, using the matrix without poly(A) part for MP analyses of the trnT–trnL partition. For ML analysis, a F81 + G model corresponded best to the matK data. For the combined dataset, the same model was suggested by hierarchical Likelihood Ratio Tests (hLRTs) while the Akaike Information Criterion (AIC) favored a TVM + I model, comparable to the findings for the trnT–trnL region. Trees were calculated under both models, and bootstrapping was done like above. For the nuclear ITS region, only diploid taxa were included in the final analyses. Sequence alignment was performed manually in BioEdit. Indels were placed so as to keep the number of mutations within a region at a minimum while indels and substitutions were considered as events of equal probability. All analyses were calculated with different combinations of outgroup taxa (see results).

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

MP analyses were performed as heuristic searches with 100 random sequence additions and no more than 100 trees of length greater than or equal to 1 saved per replicate, TBR branch swapping and the MulTrees option in effect. Gaps were treated as 5th character state as most of them were one base long. For distance analyses, heuristic ME searches were performed using the p-distance measure, and ML analyses used the Modeltest-derived parameters corresponding to a SYM + G model with one random addition sequence replicate and TBR branch swapping. The most distant outgroup taxa differed distinctly from the ingroup taxa, possibly indicating a different pattern of molecular evolution. Therefore, additional likelihood parameters were determined excluding these taxa. In that case, a simpler model (TrNef + G) was favored. For each outgroup combination, 1000 bootstrap replicates were calculated for all analyses as indicated above. Under the ML criterion, the different Modeltest parameters were applied in all four combinations. For each data set, branches found in all analyses (e.g., different outgroup, indel treatment, optimality criterion,

353

model of molecular evolution) and even in parsimony strict consensus (PSC) trees and all bootstrap 50% majority rule consensus (MRC) trees were determined. Major conclusions were based solely on concordant results. 3. Results 3.1. Chloroplast trnT–trnL intergenic spacer All analyses (different optimality criteria and models, different treatment of indels, inclusion or exclusion of the poly(A) part) resulted in basically the same tree topology. A representative ML tree is shown in Fig. 1. Congruent branching patterns of all analyses are summarized. According to these results, subgenus Pilosella is polyphyletic and forms two well-differentiated groups. One of them (referred to as Pilosella I) includes the supposed outgroup taxon Hispidella hispanica and is characterized by two unique substitutions; the other one (called Pilosella II) includes genus Andryala and shows four synapomorphic substitutions plus one duplication. The latter clade is always

Fig. 1. Phylogenetic tree based on the chloroplast trnT–trnL intergenic spacer. A ML tree is shown with bootstrap percentages above branches. Branches found in all analyses are indicated as bold lines; bootstrap support for parsimony (in italics) and distance trees (regular) are in grey below bold branches. Species sharing identical sequences are listed sequentially. Those comprising more than one haplotype are underlined in grey.

354

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

statistically well-supported, while two differentiating characters of the former clade are not sufficient to obtain higher bootstrap values (Kluge and Wolf, 1993), although these substitutions are not homoplastic. In Pilosella I, two species (Hieracium castellanum and Hieracium argyrocomum) show the ancestral trnT–trnL type. In Pilosella II, a total of 11 species share the ancestral condition, eight of which even have identical sequences including the poly (A). Andryala is also polyphyletic with its haplotypes being derived from the most widespread ancestral Pilosella II type. Three lineages occur: separate ones for the two endemic relict species Andryala agardhii (Spain) and Andryala levitomentosa (Romania), and a well-supported cluster of the Macaronesian and the rest of the Mediterranean taxa. A. agardhii falls within the Pilosella II range of variability. A. levitomentosa shows a rather high number of autapomorphic mutations. According to the tree, the A. levitomentosa haplotype branched off earlier than the other Andryala and Pilosella II lineages diversified, but the single transition supporting this could also be a reversal of an earlier mutation. The hawkweed subgenera Hieracium and Chionoracium are as similar or even more similar to Pilosella I than both Pilosella clades are to each other (Fig. 1). Chionoracium is only characterized by a single transversion, which—as not entirely consistent—is probably due to ancient polymorphism and lineage sorting. Only two branches are stable: one comprises most of the South American species, and another one consists of three North American taxa (Hieracium albiflorum, Hieracium venosum and Hieracium scabrum). Subgen. Hieracium is polyphyletic and has no synapomorphic character states in the trnT–trnL region. One particular sequence dominates and is identical for nine species (Hieracium umbellatum etc.) which span a geographic area from eastern, central and western Europe to Canada (Table 1). Another haplotype is shared by species of the Hieracium alpinum group (Fig. 1). Subgen. Hieracium clades usually received low bootstrap support. While several species even share identical haplotypes (including identity of the poly(A) part), in others, some intraspecific polymorphism is present (underlined species in Fig. 1), either as a 1-bp difference in the number of A’s (Hieracium pilosella), or likely resulting from introgression of another haplotype lineage by hybridization (H. lactucella GR, Hieracium peleterianum Bay., Hieracium caespitosum 5x). In the middle part of the trnT–trnL intergenic spacer region, several phylogenetically informative nested inserts occur as duplications from either side of the insertion point. They were used to reconstruct an order of successive mutational events, providing information about the reading direction. According to these sequential duplications, Hieracium/Chionoracium are also intermediate between Pilosella I and II, and Pilosella II and Andryala haplotypes show derived conditions. Hieracium s.l., Andryala and Hispidella constitute a wellsupported monophyletic group with unresolved basal rela-

tionships. No other species of the Hieraciinae sensu Bremer (1994) cluster within this assemblage, and the applied outgroup taxa are the most closely related ones to the clade under consideration (Gemeinholzer and Kilian, 2005). Some of these results are counter-intuitive: (1) especially the split of the well-defined subgenus Pilosella into two clusters, one of which includes Hispidella, the other one Andryala, and (2) the monophyly of the whole clade with this particular combination of (sub) genera. 3.2. Chloroplast matK gene For a subset of species, we analyzed sequences of the matK gene in order to test whether the unexpected results are artifacts of the trnT–trnL intergenic spacer region. Variability in the 970–979 bp of matK (up to 1.3% in the ingroup) is only about half of the sequence divergence of the trnT–trnL intergenic spacer (up to 2.5%). MatK generally confirms the results based on trnT–trnL, in separate as well as in combined analyses. Pilosella is split into the same two clusters, Hispidella belongs to the Pilosella I group, and Andryala is closely related to Pilosella II (Fig. 2). Further details match the trnT–trnL results, e.g., the cluster of H. lactucella, Hieracium vahlii and Hieracium breviscapum,

Fig. 2. Phylogenetic tree based on chloroplast trnT–trnL and matK sequences. A ML tree is shown with bootstrap values given above branches (TVM + I/F81 + G), for parsimony (in italics) and distance (regular) analyses in grey below branches. Bold branches are found in all analyses (see Methods).

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

and the cluster of South American Chionoracia (H. aff. asplundii, Hieracium patagonicum, Hieracium trichodontum, Hieracium stachyoideum, Hieracium frigidum). Likewise, subgenera Hieracium and Chionoracium are polyphyletic. In addition, the same final ingroup clade is statistically well-supported. A major difference is the monophyly of Andryala. In most of the analyses, it forms a sister group to Pilosella II, however, the branch leading to Pilosella II is not supported by the bootstrap MRC tree (Fig. 2). Nevertheless, Andryala and Pilosella II haplotypes are derived from a common ancestral one with Pilosella II being hardly differentiated, and the same Andryala species like before being further derived. Another difference at first glance is the basal position of subgen. Hieracium, however, internal branches leading to the other subgenera are neither wellsupported nor present in all analyses. We conclude that the apparent pattern of haplotype sharing across species is not a product of misleading homoplastic changes in the trnT–trnL, but rather representative for their chloroplast DNA as such.

355

3.3. Nuclear ribosomal DNA internal transcribed spacer In order to discover if the chloroplast DNA adequately reflects the evolutionary history of Pilosella, we compared the plastid results to a phylogeny based on the nuclear ribosomal ITS region. Phylogenetic analyses reveal the same well-supported final ingroup (Fig. 3). No further taxa of the Cichorieae tribe were more closely related in preliminary analyses (Gemeinholzer and Kilian, 2005). Hieracium intybaceum falls outside the well-supported assemblage of Hieracium s.l., Hispidella and Andryala. According to its chloroplast DNA, it belongs to subgenus Hieracium, which is in accordance with its current taxonomic placement. In order to exclude PCR or sequencing artifacts, we represented the species by three individuals, using material from different sources (Table 1). All have very similar or identical sequences. H. intybaceum-specific ITS character states are scattered all over the sequence. The same is true for character states either shared with or distinguishing the taxon from each of the three major ingroup clades. PCR recom-

Fig. 3. Phylogenetic tree based on the nuclear ribosomal ITS region. A ML tree of all taxa is shown with bootstrap support given above branches. Polyploid species are marked with an asterisk (*). Branches present in all trees are in bold, bootstrap percentages for parsimony and distance analyses are in grey italics and regular font below bold branches. Bootstrap support for the uppermost Pilosella clade is composed of percentages for this cluster plus support for a subset without the polyploids (ML and MP), together they match the support found with the distance criterion. The Koelpinia branch is truncated by about 20% of its original length.

356

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

bination as a possible explanation for the unusual placement of H. intybaceum can, therefore, be excluded. Analyses were subsequently done with different outgroup combinations: (1) using all species, and (2) using only H. intybaceum as an outgroup. Roughly the same tree topologies emerge, but bootstrap values of the ML tree comprising all species are markedly lower than those for parsimony and distance analyses (Fig. 3). This is not an effect of different substitution models, but depends upon inclusion or exclusion of the most divergent outgroup taxa. Using only H. intybaceum as an outgroup, bootstrap values are similar for all analyses and models of molecular evolution, indicating robustness with respect to underlying assumptions. We therefore used H. intybaceum for the subsequent analyses as the most appropriate outgroup due to its unexpected sister relation to the ingroup. Basically, three well-supported clusters constitute the final ingroup: monophyletic Pilosella (with Hispidella as sister taxon), Hieracium/Chionoracium, and Andryala (Fig. 3). Basal relationships among them are hardly resolved. Andryala shows a weakly supported sister group relationship to Hieracium s.l./Hispidella, but some analyses even result in a trifurcation of the major lineages. Within Pilosella, two major branches split further into two lineages, each. One of these major branches comprises the well-supported clade of species of the section Pilosellinae, characterized by an unbranched stem and a single large capitulum (H. argyrocomum and allies), with H. castellanum as sister taxon. The other major branch comprises one group with species of the closely related sections Auriculina and Alpicolina (H. lactucella and allies), while the second contains exclusively species of chloroplast haplotype group II. Subgen. Hieracium is also polyphyletic in the ITS trees; as far as comparable, clusters of the diploids roughly correspond to those based on the trnT–trnL region. Likewise, subgen. Chionoracium forms a very weakly supported cluster not distinctly different from subgen. Hieracium, and the same South American clade occurs as in the chloroplast DNA analyses. The only major difference in the ITS analyses is the well-supported monophyly of the latter two subgenera. Andryala is also distinctly differentiated, but again, within the cluster, the same lineage of Mediterranean-Macaronesian taxa matches the chloroplast data. Overall, the ITS tree is in accordance with morphology and other evidence and is therefore considered to reflect the true phylogenetic species relationships. 4. Discussion 4.1. Phylogeny of Hieracium subgen. Pilosella and close relatives Our study presents a phylogeny based on a complete sampling of basic species of the hawkweed subgenus Pilosella. Three gene regions are used to reconstruct relationships: the chloroplast trnT–trnL intergenic spacer (for all species), the chloroplast matK gene (for a representative

subset), and the internal transcribed spacer (ITS) region of nuclear ribosomal DNA (for diploids). Generally, basic tree topologies are rather robust given the low overall genetic variability and the broad variety of criteria employed for phylogenetic analyses. Although chloroplast and nuclear DNA trees are not congruent in several respects, all three markers show a good resolution of major clades and also reveal a coarse substructure of relationships. A framework of major processes concerning the diversification and evolutionary patterns of Pilosella and its closest relatives could be established. 4.1.1. Pilosella In a previous study involving Pilosella species from a restricted geographic area (Fehrer et al., 2005), we discovered that trnT–trnL sequences divided Pilosella into two well-separated haplotype groups named Pilosella I and II. Comprehensive sampling of species from a broad geographic range now confirms that the previous finding is not due to sampling artifact, but representative for the whole subgenus (Fig. 1). The low overall variability and sharing of trnT–trnL sequences between species implies that our samples approximate the total genetic variability. The two haplotype groups are not derived from each other—otherwise nested clusters would be expected—nor do they have a common ancestor without Hieracium/Chionoracium in the trnT–trnL analyses. MatK confirms the Pilosella clusters revealed by trnT–trnL analyses, in separate (results not shown) as well as in combined analyses of chloroplast markers (Fig. 2). This is also supported by the work of Trewick et al. (2004) who analyzed the trnL intron and trnL–trnF intergenic spacer. Our Pilosella groups I and II correspond to the representative haplotypes D and A of these authors (GenBank Accession Nos. AY342314 and AY342315). They differ from each other by six substitutions and four indels, corresponding to a sequence divergence of 1.1%. Thus, differentiation into two divergent haplotype groups is a general feature of Pilosella chloroplast DNA. The Pilosella I haplotype group is presumably the older one. In the nested insert region, it shows the plesiomorphic condition without duplications. Differentiation within the Pilosella I clade is mostly due to substitutions and indels (trnT–trnL), and divergence from the ancestral haplotype is higher compared to Pilosella II. In the latter, most species share the basic ancestral haplotype, and most of the variation occurs only in the hypervariable poly(A) region. This indicates that speciation in Pilosella II is more recent and not accompanied by major chloroplast DNA evolution. Also, Pilosella II shows a derived condition in the nested insert region as it contains two independent duplications. Within the haplotype groups, there is little substructure elucidating species relationships due to low variability. In contrast to the cpDNA-based trees, the nuclear ITS phylogeny (Fig. 3) shows that subgenus Pilosella is monophyletic as was expected for a taxon behaving like a single hybridogenous complex. This is in concordance with the

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

general observation that ITS trees have considerable congruence with prior hypotheses about phylogenetic relationships based on morphological, biogeographical or other ´ lvarez and Wendel, 2003), which is also data (reviewed in A reflected by numerous examples from other Asteraceae at the generic level and below (summarized in Mavrodiev et al., 2004). 4.1.2. Hispidella An intriguing finding is the very close relationship of the monotypic west-Mediterranean genus Hispidella to Pilosella. According to ITS sequences (Fig. 3), they are sister taxa, well-separated from both other major clades. CpDNA associates Hispidella and the Pilosella I group. The position of Hispidella additionally suggests that the Pilosella I haplotype group is the older one as it is shared by the sister of the whole Pilosella clade. 4.1.3. Hieracium/Chionoracium Based on nuclear ITS sequences, both subgenera together form a well-supported clade (Fig. 3). In the nested insert region, all surveyed Hieracium species and all but one Chionoracium species show exclusively the first duplication. Within this clade, however, all three markers reveal the polyphyly of subgen. Hieracium. Also, Chionoracium species cluster together with only weak bootstrap support, i.e., the monophyly of this subgenus is doubtful, too. The same holds for the chloroplast markers. Truly synapomorphic character states defining the group are generally lacking. Thus, relationships between these subgenera remain unresolved, and the molecular markers do not clearly distinguish them. Low nuclear (only 2% on average) and chloroplast sequence divergence and an unresolved polytomy at the base are consistent with a rapid recent diversification of their major lineages. 4.1.4. Hieracium intybaceum Taxonomically, it currently belongs to subgen. Hieracium, but was once treated as the monotypic genus Schlagintweitia (Grisebach, 1852) that was adopted by a few authors in the 19th century, but later abandoned. Due to its unexpectedly distinct position revealed by the nuclear marker (Fig. 3), it requires special attention. No ITS character states associate H. intybaceum closer to other species of subgen. Hieracium than to any other ingroup lineage. In contrast, chloroplast DNA unequivocally places H. intybaceum to subgen. Hieracium. The most straightforward explanation in this case is to assume hybridization with subsequent capture of a Hieracium chloroplast genome. The existence of several morphologically ‘intermediate’ polyploid Hieracium species with probable contribution of H. intybaceum (Zahn, 1921–23; de Retz, 1975) suggests that genetic exchange between H. intybaceum and Hieracium species might be possible. After initial hybridization, repeated backcrosses to H. intybaceum may have eliminated subgen. Hieracium ITS copies and/ or concerted evolution has deleted any evidence of previous

357

crosses. In the latter case, the assumed hybridization(s) was probably not very recent. Given the apparent reliability of the ITS to correctly reveal species relationships in Hieracium, we suggest that the taxonomic position of H. intybaceum should be investigated by further nuclear genetic markers as well as by experimental crosses with subgen. Hieracium species, and that its morphological, cytological and other features may be reassessed. 4.1.5. Andryala While Stebbins (1953) considered Andryala to be the next closest relative to Hieracium s.l., a view that is widely accepted in recent literature, we were puzzled when our trnT–trnL data suggested that Andryala chloroplast haplotypes are derived from Pilosella haplotype group II. In contrast, the ITS tree clearly reveals monophyly of all Andryala species, which is in concordance with traditional taxonomic views. The striking discrepancy between nuclear and chloroplast markers for this group will be discussed below. The trnT–trnL chloroplast haplotypes of the Macaronesian species (Andryala pinnatifida, Andryala varia, Andryala glandulosa) are ancestral to those of the Andalusian populations (Andryala integrifolia, Andryala ragusina). This is in concordance with a re-colonization of the continent from Macaronesia as has been described for other plants, e.g., Tolpis (Mort et al., 2003), and Convolvulus (Carine et al., 2004). More comprehensive sampling of Andryala species and populations will be necessary in order to identify continental relatives, source area and particular lineage of the original colonizer(s), and route(s) of back-colonization. 4.1.6. Ingroup and outgroup As outgroup, we successively included taxa of decreasing presumed relatedness to Hieracium subgen. Pilosella. Many of the initial outgroup taxa turned out to be part of the ingroup. Our ‘final’ ingroup, consisting of Hieracium subgenera Pilosella, Hieracium and Chionoracium, and genera Hispidella and Andryala, is strongly supported by all molecular markers. The three major lineages (Pilosella/Hispidella, Hieracium/Chionoracium, Andryala) apparently became established in relatively rapid succession following divergence from a common ancestor. Regarding Hieracium s.l., the nearly indistinguishable assemblage of subgenera Hieracium and Chionoracium is no more closely related to subgen. Pilosella than to Hispidella or Andryala, with H. intybaceum being sister to the whole clade. Apart from the two Tolpis species that expectedly cluster with high support in all analyses, no unequivocal relationships among the ‘final’ outgroup taxa can be detected, probably as a consequence of long branch attraction (Felsenstein, 1978). 4.2. Intergeneric chloroplast capture 4.2.1. Major incongruencies and potential reasons Two apparent major contradictions between chloroplast and nuclear markers occur in the final ingroup clade: (1)

358

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

two rather divergent Pilosella clusters (cpDNA) versus their monophyly (ITS), and (2) chloroplast haplotypes of Andryala derived from one of the Pilosella haplotypes versus Andryala being monophyletic and clearly separated from Pilosella according to ITS. In these cases, results from chloroplast DNA are contradictory to all other evidence. Reticulation/chloroplast capture is considered to be the most likely reason (apart from sampling error, convergence, rate heterogeneity, long branch attraction, or ancient polymorphism/lineage sorting) that potentially contributes to phylogenetic discordance at lower taxonomic levels (Rieseberg and Soltis, 1991; Soltis and Kuzoff, 1995; Sang and Zhong, 2000). This is especially true if the chloroplast phylogeny suggests an almost haphazard arrangement of species that is at odds with other findings. To some extent, incomplete lineage sorting of trnT–trnL seems to have occurred within subgenera Hieracium and Chionoracium. However, this is not a likely explanation for the major discrepancies observed. Incomplete lineage sorting of the major haplotypes would imply that evolution of most of their diversity preceded speciation of the whole clade. Especially the distribution of the Pilosella II type that occurs at the tips of the Pilosella clade as well as in genus Andryala (Fig. 4) contradicts a lineage sorting scenario. Recent haplotype distribution on the ITS-based phylogeny also seems much too structured or non-random to favor that possibility. Some presumed parallel mutations in trnT–trnL (four substitutions) occur independently in different lineages. Given the low overall sequence divergence (below 2.5%), these might reflect potential cases of non-independent mutation caused by physical sequence constraints (secondary structure), but at present not enough is known about the molecular evolution of the trnT–trnL spacer to draw further conclusions. While the mentioned homoplasious mutations caused some noise in phylogenetic analyses, the signal provided by homoplasy-free mutations is much stronger and also largely unaffected with regard to the questionable clusters. No other potential causes for incongruence seem to be applicable to our data sets. We therefore consider hybridization and subsequent chloroplast capture as the most likely explanation for the major discrepancies between chloroplast and nuclear DNA trees.

Fig. 4. Distribution and assumed transfer of chloroplast DNA lineages on the ITS-based phylogeny. The major chloroplast haplotype groups are mapped on an ITS tree (green: Pilosella I/Hispidella type, red: Pilosella II/ Andryala type). Colors of branches reflect the assumed ancient trnT–trnL haplotype distribution. Acquisition of the two Pilosella I/Hispidellacharacterizing substitutions under this scenario is indicated by green asterisks. The first intergeneric chloroplast transfer is indicated by a black arrow, the second by a red arrow. Dashed red lines show 3–4 more recent transfers within Pilosella. A ML tree produced with the TrNef + G model is shown with bootstrap support >50% above branches and for parsimony (italics) and distance (regular) analyses in grey below branches for Pilosella and deeper nodes.

4.2.2. Reconciling chloroplast and nuclear sequence data In order to better understand these discrepancies, we mapped the major chloroplast haplotype groups on the ITS tree (Fig. 4). The Pilosella/Hispidella clade comprises both chloroplast haplotype groups, and type II is additionally found in the rather distant Andryala clade. Assuming that haplotype group II evolved first would imply that it had to be present in the ancestor of the whole ingroup clade. This would require quite a number of parallel independent mutational steps to explain haplotype evolution along the tree and does not fulfill minimal evolution requirements. It is also not in concordance with the pat-

terns of trnT–trnL molecular evolution and the reading direction derived from the nested insert region. Assuming instead that haplotype group I evolved first is the more parsimonious explanation and also in concordance with the higher number of mutations that distinguish the Pilosella II lineage from others (Fig. 1). However, haplotype group II cannot be derived from group I without assuming the parallel loss of two synapomorphic type I-specific mutations (Fig. 4, asterisks) as well as the independent gain of an identical insert along the ITS tree, which seems rather

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

unlikely. Instead, the assumption of an early introgression of chloroplast DNA from an ancestor of the Hieracium/ Chionoracium lineage into one of the Pilosella lineages would resolve these character conflicts. A similar scenario, transmission of the ancestral Pilosella II chloroplast haplotype to the Andryala ancestor, can easily explain the presence of haplotype group II in a distant clade. 4.2.3. Reconstruction of major events In the following, we summarize what we believe is the clade’s most likely history: The ancestral cpDNA condition (without insert) is maintained in Pilosella I, Hispidella, and H. antarcticum (subgen. Chionoracium). Two unique mutations characterizing Pilosella I and Hispidella are not present in Pilosella II indicating that the Pilosella II cpDNA evolved separately. The first insert is present in all but one Hieracium/Chionoracium species and also in Pilosella II/Andryala indicating cpDNA transfer from a Hieracium/Chionoracium ancestor to a Pilosella II or Andryala ancestor. The transfer resulted in rapid evolution of cpDNA in the introgressed lineage (second insert plus four synapomorphic substitutions in trnT–trnL). All Andryala chloroplast haplotypes are derived from the ancestral Pilosella II type. Thus, cpDNA was not at first transferred to Andryala, but to a Pilosella lineage, most likely the one containing exclusively Pilosella II type species (Fig. 4). From that lineage, the ancestral Pilosella II type (still present in the majority of present-day species) was transferred to the Andryala ancestor where it diverged further. Finally, several more recent hybridization events within Pilosella can explain additional group II haplotypes at the tips of two other Pilosella clades (Fig. 4, dashed arrows) as the genus is noted for its hybridizing potential. In order to decide which chloroplast genome is the ‘true’ and which one the ‘captured’ genome, Soltis and Kuzoff (1995) suggested considering those taxa having a chloroplast genome not well differentiated from other taxa most likely to represent the ‘captured’ haplotype. All four recently introgressed species indeed show the most widespread Pilosella II haplotype whereas the majority of Pilosella I species within the same ITS clades have more derived haplotypes. 4.2.4. Likelihood of ancestral hybridizations At first, the idea of hybridization between Pilosella and Hieracium/Chionoracium seemed very far-fetched due to their different modes of reproduction, morphology, ecology, DNA content etc. Subgenera Pilosella and Hieracium are noted for their hybridizing potential within their respective groups. Pilosella species are linked by extensive recent hybridizations between basic species and intermediates, sexual and facultatively apomictic ones alike (Krahulcova´ et al., 2000; Krahulec et al., 2004; Fehrer et al., 2005). In subgen. Hieracium, hybridizations are extremely rare at present (Mra´z et al., 2005), but are likely to have been very frequent in the past, giving rise to the vast number of polyploids that makes it currently one of the world’s largest

359

genera. Cross-breeding between the subgenera nowadays seems not to be possible; pollination with Pilosella pollen elicits a mentor effect in Hieracium species resulting in autogamous reproduction (Mra´z, 2003). An attempted reciprocal cross resulted in neither autogamy nor hybridization (Mra´z, personal communication). In fact, their failure to hybridize contributed to their treatment as different genera by Greuter (2003). Guppy (1978) observed a rare occurrence of natural hybrids among Chionoracia, but hybrids with Pilosella or Hieracium species were never observed. Furthermore, the large differences in DNA content (roughly twice as high in Hieracium/Chionoracium as compared to Pilosella and Andryala; Bra¨utigam and Bra¨utigam, 1999; Krahulcova´, Suda, and Tra´vnı´cˇek, unpublished data) may hamper recent hybridizations. But genera that do not currently hybridize could have done so at some point in their evolutionary past. The deduced introgression event must have pre-dated most of the subgenera’s divergence, transmitting a now extinct haplotype to a Pilosella lineage. From a morphological point of view, a Chionoracium ancestor might have donated its plastid genome, as the present-day New World taxon comprises most of the morphological variation seen in both Hieracium and Pilosella. Hybridization between Pilosella and the Andryala ancestor is much easier to acknowledge. Both share a number of morphological, ecological and cytological features in which they differ from Hieracium/Chionoracium. Chloroplast capture by the Andryala progenitor is also easier to deduce. The ancient transferred haplotype is still present in many Pilosella II species, although it is now superimposed by a comparably high number of mutations in different Andryala lineages. Andryala plastid DNA provides an example of ‘stepping-stone’ chloroplast transfer that enables chloroplast genomes to travel rather complex routes (Rieseberg and Soltis, 1991; Soltis et al., 1991). Both cases provide evidence for cpDNA introgression in the absence of analogous ITS gene introgression. Differential cytoplasmic exchange can be triggered by a variety of processes and with remarkable speed (reviewed by Rieseberg and Soltis, 1991). Similar cases of chloroplast capture between rather divergent taxonomic groups are also known from other plants. Soltis and Kuzoff (1995) found a strong likelihood that several ancient events of chloroplast capture occurred between lineages of the Heuchera group (Saxifragaceae) including genera that are not known to hybridize at present. A high frequency of historical hybridization (at least six ‘unexpected and unlikely cases’) was found between cotton lineages whose modern descendants are strongly isolated by geography and intrinsic genetic barriers (Cronn and Wendel, 2004). A study in Achillea (Asteraceae, Anthemideae) revealing conflict between nuclear and plastid markers showed that reticulate evolution was not only involved in recent radiations but must have been active already in the early diversification of the genus (Guo et al., 2004). Early hybridizations among genera have also been suggested to explain the conflict between chloro-

360

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361

plast and nuclear markers in the Veroniceae (Albach and Chase, 2004). The assumption of wide hybridizations between the precursors of the (sub)genera as inferred from detailed molecular character analysis resolves the major contradictions between chloroplast and nuclear markers and reveals basic evolutionary processes in Hieracium subgen. Pilosella and its closest relatives. 4.3. Future prospects In spite of the (1) broad non-random evolutionary patterns of Pilosella, (2) identification of introgression events and distinction between ‘true’ and ‘captured’ chloroplast haplotypes, and (3) surprisingly high uniformity of cpDNA haplotypes for a given species in view of the group’s hybridizing potential, future research might reveal still more complex patterns even within basic species. Sampling of many individuals from across each species’ respective distribution area will be necessary to confirm the basic patterns presented in this paper. Species relationships within crown groups remain largely unresolved due to low genetic variability and very recent speciation. Different molecular techniques or markers will have to be applied for this level of relatedness. For this first survey of hawkweed phylogenetic relationships at the molecular level, we focused on diploid species and ITS sequences for nuclear DNA analyses. Further analyses including polyploid basic species and using single or low-copy nuclear genes for elucidation of their origin are currently in progress. Taxonomic implications from this study, corroborated by morphological and floristic data, shall be presented elsewhere (Gemeinholzer and Kilian, 2005). Acknowledgments Many thanks to all who provided samples: C. Cojocariu, G. Grosskopf, M. Hajman, P. Mra´z, F. Mu¨ller, Z. Palice, R. Rabeler, M. Severa, Z. Szela˛g, L. Wilson, and especially F. Schuhwerk for the majority of Chionoracium samples and for the determination of Ecuadorian H. frigidum. We are indebted to people who contributed chromosome counts (D. Fitze, A. Krahulcova´, P. Mra´z, M. Severa) and flow cytometry data (E. Bra¨utigam, A. Krahulcova´, J. Suda, P. Tra´vnı´cˇek), to M. Kretschmer, M. Loncova´, P. Oswald, and B. Wohlbier for considerable parts of the labwork, to R. Dvorˇa´kova´ and L. Kirschnerova´ for help with the DNA isolations, and to several gardeners who helped taking care of the plants. K. Bachmann, F. Blattner, F. Schuhwerk, T. Sharbel, and two anonymous reviewers helpfully commented the manuscript, and R. Wilck is acknowledged for checking the English. The Deutsche Forschungsgemeinschaft (Grant Fe 512/ 1-1), the Ministry of Environment of the Czech Republic (Grant VaV/610/3/00), and the Academy of Sciences of

the Czech Republic (institutional long-term research plan no. AV0Z60050516) provided financial support. References Albach, D.C., Chase, M.W., 2004. Incongruence in Veroniceae (Plantaginaceae): evidence from two plastid and a nuclear ribosomal DNA region. Mol. Phylogenet. Evol. 32, 183–197. ´ lvarez, I., Wendel, J.F., 2003. Ribosomal ITS sequences and plant A phylogenetic inference. Mol. Phylogenet. Evol. 29, 417–434. Arnold, M.L., 1992. Natural hybridization as an evolutionary process. Ann. Rev. Ecol. Syst. 23, 237–261. Blattner, F.R., 1999. Direct amplification of the entire ITS region from poorly preserved plant material using recombinant PCR. Biotechniques 27, 1180–1186. Blattner, F.R., Weising, K., Baenfer, G., Maschwitz, U., Fiala, B., 2001. Molecular analysis of phylogenetic relationships among myrmecophytic Macaranga species (Euphorbiaceae). Mol. Phylogenet. Evol. 19, 331–344. Bra¨utigam, S., Bra¨utigam, E., 1996. Determination of the ploidy level in the genus Hieracium subgenus Pilosella (Hill) S.F. Gray by flow cytometric DNA analysis. Folia. Geobot. Phytotax. 31, 315–321. Bra¨utigam, S., Bra¨utigam, E., 1999. Kern-DNA-Gehalt und Ploidiestufe bei Hieracium L. 14. Symp. Biodiversita¨t & Evolutionsbiol. Jena, p. 26. Bremer, K., 1994. Asteraceae–Cladistics and Classification. Timber Press, Portland, Oregon. Carine, M.A., Russell, S.J., Santos-Guerra, A., Francisco-Ortega, J., 2004. Relationships of the Macaronesian and Mediterranean floras: molecular evidence for multiple colonizations into Macaronesia and backcolonization of the continent in Convolvulus (Convolvulaceae). Amer. J. Bot. 91, 1070–1085. Chapman, H., Bicknell, R., 2000. Recovery of a sexual and an apomictic hybrid from crosses between the facultative apomicts Hieracium caespitosum and H. praealtum. New Zealand J. Ecol. 24, 81–85. Cronn, R.H., Wendel, J.F., 2004. Cryptic trysts, genomic mergers, and plant speciation. New Phytol. 161, 133–142. de Retz, B., 1975. Hieracium. In: Jovet, P., de Vilmorin, R. (Eds.), Flore descriptive et illustre´e de la France par l’abbe´ H. Coste, troisie`me supple´ment. Librairie scientifique et technique Albert Blanchard, Paris, pp. 248–297. Fehrer, J., Sˇimek, R., Krahulcova´, A., Krahulec, F., Chrtek, J. jr., Bra¨utigam, E., Bra¨utigam, S., 2005. Evolution, hybridisation, and clonal distribution of apo- and amphimictic species of Hieracium subgen. Pilosella (Asteraceae, Lactuceae) in a Central European mountain range. In: Bakker, F.T., Chatrou, L.W., Gravendeel, B., Pelser, P.B. (Eds.), Plant Species-Level Systematics: New Perspectives on Pattern & Process. Regnum Vegetabile 143, Koeltz, Ko¨nigstein, pp.175–201. Felsenstein, J., 1978. Cases in which parsimony or compatibility methods will be positively misleading. Syst. Zool. 27, 401–410. Garcia-Jacas, N., Garnatje, T., Susanna, A., Vilatersana, R., 2002. Tribal and subtribal delimitation and phylogeny of the Cardueae (Asteraceae): a combined nuclear and chloroplast DNA analysis. Mol. Phylogenet. Evol. 22, 51–64. Gemeinholzer, B., Kilian, N., 2005. Phylogeny and subtribal delimitation of the Cichorieae (Asteraceae). XVII IBC 2005. XVII International Botanical Congress. Vienna, Austria, Europe. Abstract, p. 103. Gemeinholzer, B., Fehrer, J., Oberprieler, Ch., Lack, W., Kilian, N., (manuscript in preparation). Molecular phylogenetics and systematics of Cichorieae (Asteraceae). Greuter, W., 2003. The Euro + Med treatment of Cichorieae (Compositae)–generic concepts and required new names. Willdenowia 33, 229– 238. Grisebach, A., 1852. Commentatio de distributione Hieracii generis. Abh. ko¨nigl. Ges. Wiss. Go¨ttingen 5, 76.

J. Fehrer et al. / Molecular Phylogenetics and Evolution 42 (2007) 347–361 Guo, Y.P., Ehrendorfer, F., Samuel, R., 2004. Phylogeny and systematics of Achillea (Asteraceae–Anthemideae) inferred from nrITS and plastid trnL–F DNA sequences. Taxon 53, 657–672. Guppy, G.A., 1978. Species relationships of Hieracium (Asteraceae) in British Columbia. Can. J. Bot. 56, 3008–3019. Hall, T.A., 1999. BioEdit, a user-friendly biological sequence alignment editor and analysis program for Windows 95/98/NT. Nucl. Acids Symp. Series 41, 95–98. Houliston, G.J., Chapman, H.M., 2001. Sexual reproduction in field populations of the facultative apomict, Hieracium pilosella. New Zealand J. Bot. 39, 141–149. Jeffrey, C., 1966. Notes on Compositae, I: the Cichorieae in east tropical Africa. Kew Bulletin 18, 427–486. Johnson, L.A., Soltis, D.E., 1995. Phylogenetic inference in Saxifragaceae sensu-stricto and Gilia (Polemoniaceae) using matK sequences. Ann. Mo. Bot. Gard. 82, 149–175. Kluge, A.G., Wolf, A.J., 1993. Cladistics: what’s in a word? Cladistics 9, 183–199. Krahulcova´, A., Krahulec, F., 2000. Offspring diversity in Hieracium subgen. Pilosella (Asteraceae): new cytotypes from hybridisation experiments and from open pollination. Fragm. Flor. Geobot. 45, 239–255. Krahulcova´, A., Krahulec, F., Chapman, H.M., 2000. Variation in Hieracium subgen. Pilosella (Asteraceae): what do we know about its sources? Folia Geobot. 35, 319–338. Krahulec, F., Krahulcova´, A., Fehrer, J., Bra¨utigam, S., Placˇkova´, I., Chrtek Jr., J., 2004. The Sudetic group of Hieracium subgen. Pilosella in the Krkonosˇe Mts: a synthetic view. Preslia 76, 223–243. Krahulcova´, A., Papousˇkova´, S., Krahulec, F., 2004. Reproduction mode in the allopolyploid facultatively apomictic hawkweed Hieracium rubrum (Asteraceae, H. subgen. Pilosella). Hereditas 141, 19–30. Lloyd, D.G., Calder, V.L., 1991. Multi-residue gaps, a class of molecular characters with exceptional reliability for phylogenetic analysis. J. Evol. Biol. 4, 9–21. Mavrodiev, E.V., Edwards, C.E., Albach, D.C., Gitzendanner, M.A., Soltis, P.S., Soltis, D.E., 2004. Phylogenetic relationships in subtribe Scorzonerinae (Asteraceae: Cichorioideae: Cichorieae) based on ITS sequence data. Taxon 53, 699–712. Morgan-Richards, M., Trewick, S.A., Chapman, H.M., Krahulcova´, A., 2004. Interspecific hybridization among Hieracium species in New Zealand: evidence from flow cytometry. Heredity 93, 34–42. Mort, M.E., Crawford, D.J., Santos-Guerra, A., Francisco-Ortega, J., Esselman, E.J., Wolfe, A.D., 2003. Relationships among the Macaronesian members of Tolpis (Asteraceae: Lactuceae) based upon analyses of inter simple sequence repeat (ISSR) markers. Taxon 52, 511–518. Mra´z, P., 2003. Mentor effects in the genus Hieracium s. str. (Compositae, Lactuceae). Folia Geobot. 38, 345–350. Mra´z, P., Chrtek Jr., J., Fehrer, J., Placˇkova´, I., 2005. Rare recent natural hybridization in Hieracium s. str.–evidence from morphology, allozymes and chloroplast DNA. Plant Syst. Evol. 255, 177–192. Na¨geli, C., Peter, A., 1885. Die Hieracien Mittel-Europas, Piloselloiden. R. Oldenbourg, Mu¨nchen.

361

Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14, 817–818. Rieseberg, L.H., Carney, S.E., 1998. Plant hybridization. New Phytol. 140, 599–624. Rieseberg, L.H., Soltis, D.E., 1991. Phylogenetic consequences of cytoplasmic gene flow in plants. Evol. Trends Plants 5, 65–84. Sang, T., Zhong, Y., 2000. Testing hybridization hypotheses based on incongruent gene trees. Syst. Biol. 49, 422–434. Schuhwerk, F., 1996. Published chromosome counts in Hieracium. http:// www.botanik.biologie.uni-muenchen.de/botsamml/projects/ chrzlit.html. Sell, P.D., West, C., 1976. 181. Hieracium L. In: Tutin, T.G., Heywood, V.H., Burges, N.A., Moore, D.M., Valentine, D.H., Walters, S.M., Webb, D.A. (Eds.), Flora Europaea 4. Cambridge University Press, Cambridge, London, New York, Melbourne, pp. 358–410. Smissen, R.D., Breitwieser, I., Ward, J.M., 2004. Phylogenetic implications of trans-specific chloroplast DNA sequence polymorphism in New Zealand Gnaphalieae (Asteraceae). Plant Syst. Evol. 249, 37–53. Soltis, D.E., Kuzoff, R.K., 1995. Discordance between nuclear and chloroplast phylogenies in the Heuchera group (Saxifragaceae). Evolution 49, 727–742. Soltis, D.E., Soltis, P.S., Collier, T.G., Edgerton, M.L., 1991. Chloroplast DNA variation within and among genera of the Heuchera group (Saxifragaceae): evidence for chloroplast transfer and paraphyly. Amer. J. Bot. 78, 1091–1112. Stebbins, G.L., 1953. A new classification of the tribe Cichorieae, family Compositae. Madron˜o 12, 65–81. Stebbins, G.L., 1971. The role of hybridization in evolution. In: Processes of Organic Evolution. Prentice-Hall, Englewood Cliffs, New Jersey, pp. 116–135. Sˇtorchova´, H., Hrdlicˇkova´, R., Chrtek Jr., J., Tetera, M., Fitze, D., Fehrer, J., 2000. An improved method of DNA isolation from plants collected in the field and conserved in saturated NaCl/CTAB solution. Taxon 49, 79–84. Swofford, D.L., 2002. PAUP*. Phylogenetic Analysis Using Parsimony (*and Other Methods). Version 4. Sinauer Assoc., Sunderland, Massachusetts. Taberlet, P., Gielly, L., Pautou, G., Bouvet, J., 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Mol. Biol. 17, 1105–1109. Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997. The CLUSTAL_X windows interface, flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucl. Acids Res. 25, 4876–4882. Trewick, S.A., Morgan-Richards, M., Chapman, H.M., 2004. Chloroplast DNA diversity of Hieracium pilosella (Asteraceae) introduced to New Zealand: reticulation, hybridization, and invasion. Amer. J. Bot. 91, 73–85. Zahn, K.H., 1921–23. Hieracium. In: Engler, A. (Ed.), Das Pflanzenreich 4(280). Wilhelm Engelmann, Leipzig. Zahn, K.H., 1922–38. Hieracium. In: Ascheron, P., Graebner, P. (Eds.), Synopsis der mitteleuropa¨ischen Flora 12(1-3). Gebru¨der Borntra¨ger, Leipzig, Berlin.