Phylogeny and systematics of Achillea (Asteraceae ...

53 (3) • August 2004: 657–672

Guo & al. • Phylogeny and systematics of Achillea

Phylogeny and systematics of Achillea (Asteraceae-Anthemideae) inferred from nrITS and plastid trnL-F DNA sequences Yan-Ping Guo, Friedrich Ehrendorfer & Rosabelle Samuel Department of Higher Plant Systematics and Evolution, Institute of Botany, University of Vienna, Rennweg 14, A-1030 Vienna, Austria. [email protected]; [email protected] (author for correspondence); [email protected] The N Hemisphere genus Achillea includes about 130 perennial and allogamous species, is centered in SE Europe/SW Asia, and exhibits a complex phyletic structure due to excessive hybridization and polyploidy. About half of the species and five of the six traditional sections together with several outgroup genera were studied using nrITS and plastid trnL-F DNA sequences. In spite of some discordance, these markers were shown by Maximum Parsimony and Bayesian Inference to be suitable for revealing relationships with generic allies and for distinguishing the main lineages within Achillea. With the inclusion of Otanthus (and possibly Leucocyclus) Achillea s.l. becomes monophyletic and appears as sister to Anacyclus. A basal clade is formed by the xerophytes of Achillea sections Babounya and Santolinoideae in SW Asia together with the Mediterranean coastal Otanthus. Achillea sect. Ptarmica s.l. has to be divided into the meso- to hygrophytic herbs of A. sect. Ptarmica s.s. in the N Hemisphere and the mountain species of A. sect. Anthemoideae. The latter differentiated in the mountains from NW Anatolia to the Pyrenees, possibly originating from ancestors related to the extant A. ligustica. Finally, taxa of sect. Achillea s.l. (to be merged with A. sect. Filipendulinae) radiated from a center in SE Europe, occupied very different open habitats, and reached an extensive distribution with the very polymorphic polyploid and reticulate complex A. millefolium agg. Here and in other groups of Achillea, various instances of conflicting evidence from nrITS, plastid trnL-F, and morphology point to hybridization and lineage sorting. This means that reticulate evolution is not only involved in recent radiations but must have been active already in the early diversification of the genus.

KEYWORDS: Achillea, Asteraceae-Anthemideae, nrDNA ITS, phylogeny, plastid trnL-F, systematics.

INTRODUCTION Achillea (Asteraceae-Anthemideae) is a genus of about 130 species centered in SE Europe and SW Asia with extensions through Eurasia to North America. It demonstrates remarkable ecological adaptability (Fig. 1), ranging from deserts and sea coasts to nival pioneer biota, and from rock fissures and talus (often with relict species) to ruderal habitats (with species of wide distribution, e.g., A. millefolium, as a worldwide weed). Plants in this genus are perennial, entomophilous and predominantly outbreeding. The present study is primarily concerned with the phylogeny and systematics of Achillea. There are three problems which we hope to clarify in light of nuclear ribosomal ITS and plastid trnL-F DNA sequence data: (1) Phylogenetic position of Achillea within Anthemideae. — This problem has recently been approached by cladistic studies. Bremer & Humphries (1993) and Bremer (1994) have placed Achillea into a new subtribe Achilleinae together with eight other Mediterranean genera (Santolina, Chamaemelum, Cladanthus, etc.), and have regarded it as close to the

monotypic Mediterranean coastal Otanthus, the S Turkish Leucocyclus (1–2 species), and the more widespread, predominantly annual Anacyclus (12 species). Recent DNA sequence analyses on the tribe Anthemideae have provided new information with respect to the intergeneric relationships of Achillea, i.e., Watson & al. (2000) with plastid ndhF, Oberprieler & Vogt (2000) with nuclear ribosomal ITS and plastid trnLF, and Francisco-Ortega & al. (2001) with ITS. These DNA sequence data support a close relationship between Achillea, Otanthus, and Anacyclus. They also suggest unsuspected affinities with Tanacetum, Anthemis, and Matricaria, but place Santolina, Chamaemelum, Cladanthus and Aaronsohnia, etc., in a more distant position to Achillea. Nevertheless, only a single representative Achillea species was included in these DNA studies. Thus, the relationships and position of this genus remain of relevance for the present study. (2) Circumscription and monophyly of Achillea. — Linnaeus (1753) circumscribed his genus Achillea in a wide sense, whereas Candolle (1838), following an older concept, recognized Ptarmica Mill. as a separate additional genus. In recent years this narrow cir657


53 (3) • August 2004: 657–672

Fig. 1. Representative species of Achillea and their habitats. A, B, A. fragrantissima (A. sect. Babounya), desert ravine, Sinai peninsula, 400 m; scale 10 cm. C, D, A. ceretanica-2x (A. sect. Achillea, A. millefolium agg.), upper montane Pinus uncinata stands and rocky grassland, Eastern Pyrenees, 1800 m; plant about 20 cm tall. E, F, A. nana (A. sect. Anthemoideae), glacier moraines, Alpi Bernina, 2800 m; plant about 10 cm tall.

658

53 (3) • August 2004: 657–672

cumscription has been adopted again, e.g., by Klokov & Krytzka (1984) and Sytnik & Androshchuk (1984). Otherwise, the broader Linnaean concept, recognizing Ptarmica only as a section of Achillea, was mostly followed. Furthermore, Boissier (1849) created the genus Arthrolepis, but subsequently he (1875) treated it only as a section of Achillea. In this wider circumscription the genus has been accepted in several general and multidisciplinary surveys of Asteraceae (Heywood & Harborne, 1977; Bremer, 1994; Caligari & Hind, 1996; Hind & Beentje, 1996) as well as in numerous regional floristic investigations (e.g., Afanasyev & Bochantsev, 1961; Wagenitz, 1968/1979; Huber-Morath, 1975). Nevertheless, up to now, and particularly with respect to circumscription and monophyly of the genus, no general survey based on DNA data is available. (3) Intrageneric clades in Achillea and their relationships. — Traditionally, three to six sections are recognized in Achillea (Koch, 1837; Boissier, 1875; Heimerl, 1884; Hoffmann, 1894; Afanasyev & Bochantsev, 1961; Huber-Morath, 1975). These sections are listed below with approximate species numbers in parentheses and general eco-geographical comments. Species studied and sections accepted are listed in Appendix 1. Achillea sections Babounya (2), Arthrolepis (4), and Santolinoideae (36) include species from xeric, often semidesert habitats, radiating from SW Asia through the S Mediterranean area and North Africa to Morocco and S Spain. The heterogeneous A. sect. Ptarmica so far has been circumscribed in a wide sense (s.l.) as proposed by Candolle (1838) and Heimerl (1884) who already suggested further subdivisions. Here we want to test a recent proposal (Saukel & al., 2004) to separate the N Hemisphere and predominantly lowland members as a more narrowly circumscribed A. sect. Ptarmica s.s. (20) from the predominantly montane to high alpine taxa in the central and southern European mountains that would then be recognized as A. sect. Anthemoideae (25). The last traditional sections are Achillea sect. Filipendulinae (18) with yellow ray florets and sect. Achillea (≡ A. sect. Millefolium; 25) with whitish or pink ray florets. Members of both sections are found in a great variety of xeric to wet habitats from the lowlands into the mountains, and A. sect. Achillea even into the alpine belt. But whereas taxa of A. sect. Filipendulinae are centered in SE Europe and range only from the Mediterranean to C Asia, some members of A. sect. Achillea, particularly from the polyploid aggregate A. millefolium agg. (17), have occupied an extensive N Hemisphere area, and A. millefolium s.s. even has become a cosmopolitan weed. As many of the Achillea sections mentioned above are still problematic (e.g., the separation of A. sect. Achillea and A. sect. Filipendulinae), we hope to clarify


their congruence with natural clades, their circumscription, phylogenetic relationships, and their relative evolutionary position within the context of a molecular framework. The DNA sequence data of nrITS and the plastid trnL intron, plus trnL-F intergenic spacer (hereafter referred to as trnL-F) have contributed tremendously to our understanding of phylogenetic relationships among congeneric species in many angiosperm genera (see, e.g., Baldwin, 1992; Baldwin & al., 1995; Sang & al., 1995; Soltis & Kuzoff, 1995; Soltis & al., 1995, 2003; Noyes & Rieseberg, 1999; Hodkinson & al., 2000, 2002; Koch & Al-Shehbaz, 2000; Samuel & al., 2003). Concerted evolution among the multigene copies of ITS often homogenized DNA in groups affected by hybridization (Sang & al., 1995; Wendel & al., 1995a, b; Franzke & Mummenhoff, 1999; Hodkinson & al., 2002; Chase & al., 2003; Koch & al., 2003). In young hybrid taxa the two different ITS copies from the parental progenitors often have not yet been homogenized by concerted evolution (Kim & Jansen, 1994; Sang & al., 1995; Wendel & al., 1995b; Hodkinson & al., 2002; Koch & al. 2003; contrary evidence: Soltis & al., 2003). The plastid trnLF sequences have provided useful information to reveal maternal parentage in reticulate evolutionary relationships (Soltis & Kuzoff, 1995; Chase & al., 2003). As hybridization and allopolyploidy are suspected to be rampant within Achillea (Ehrendorfer, 1959a, b), we have chosen these two markers to trace the major intrageneric affinities of the genus and some of their close generic allies.

MATERIALS AND METHODS Sampling. — This study is based on the analyses of 91 accessions of Achillea and outgroup taxa (see Appendix 1). Most samples were collected as silica gel dried leaf materials from natural populations throughout the N Hemisphere. Sequencing from herbarium material so far has failed. Vouchers are deposited in the herbarium of the Institute of Botany, University of Vienna, Austria (WU). Appendix 1 lists all taxa studied, from the outgroups and the generic allies (six genera, nine species) to the ingroup genus Achillea (63 taxa and cytotypes from five of the six traditional sections) with information about ploidy level, localities, collectors and GenBank accession numbers. In most cases, one individual was sequenced per taxon. Only in a few species with a wide distributional range (e.g., A. setacea and A. millefolium) have two or more individuals from well separated populations been sampled to evaluate intraspecific variation. DNA extraction. — Total genomic DNA was 659


extracted from 0.02 g silica gel desiccated leaf material with the 2× CTAB protocol of Doyle & Doyle (1987) with slight modifications: After precipitation with isopropanol and subsequent centrifugation, the DNA pellet was washed with 70% ethanol at 37° C for 0.5 hr and dried in a vacuum centrifuge. This was followed by resuspension in TE buffer and incubation at 37° C for 0.5 hr with RNAse. Amplification and sequencing. — The complete ITS (ITS1, 5.8S rDNA gene, ITS2) and plastid trnL-F regions (trnL 5'-exon, intron, 3'-exon, intergenic spacer and trnF gene) were amplified with universal primers ITS5 / ITS4 (White & al., 1990) and trnL-c / trnF-f (Taberlet & al., 1991), respectively. Polymerase chain reaction (PCR) was carried out in a volume of 50 µL, containing 10–50 ng of genomic DNA, 20 pmol of each primer, 45 µL 1.1× Reddy MixTM PCR Master Mix (2.5 mM MgCl2; Abgene) and 1 µL 0.4% dimethyl sulfoxide (DMSO for ITS) or 0.4% bovine serum albumin (BSA for trnL-F). The amplification reaction included the following cycles: (1) 95° C, 3 min; (2) 95° C, 1 min 30 s - 45° C, 1 min - 72° C, 1 min; (3–37) 95° C, 1 min - 48° C, 1 min - 72° C, 1 min; (38) 72° C, 10 min. The amplification product was gel purified by QIAquick® Gel Extraction Kit and then used for direct sequencing. Cycle sequencing was conducted on a GeneAmp® PCR System 9700 with ABI PRISM® BigDyeTM Terminator following manufacturer’s protocols. For ITS, in most cases, two cycle sequencing reactions were performed for each template using the two primers employed for PCR. Internal primers (ITS3 and ITS2, White & al. 1990) were used for additional sequencing reactions when necessary. For trnL-F, two cycle sequencing reactions were conducted by two reverse primers A2 (5'AGGATTTTCAGTCCTCTGCTC-3') and B2 (5'GGGGATAGAGGGACTTGAAC-3'), which are located on the trnF gene and the trnL 3'-exon, respectively. Additional cycle sequencing reactions were conducted to obtain sequences for ambiguous base calls using three forward primers A1 (5'-GGTTCAAGTCCCTCTATCCC-3', at the trnL 3'-exon), B1 (5'-CTACGGACTTAATTGGATTGAGC-3', at the trnL 5'-exon) and the specifically designed internal primer AchA3 (5'-GTAAGAGTCCCGTTCTAGATG-3', at the up-stream region of the trnL 3'-exon which is conserved in Achillea). The sequencing products were separated electrophoretically in 6% acrylamide gel on an ABI PRISMTM 377 DNA Sequencer (PE Applied Biosystems). Phylogenetic analyses. — Sequences were assembled with AutoAssembler version 1.4.0 (Applied Biosystems), aligned with ClustalX (1.5b), and then visually improved. All sequences obtained for this study have been deposited in GenBank (see Appendix 1). Phylogenetic analyses were performed separately on the 660

53 (3) • August 2004: 657–672

two datasets. Complete ITS and trnL-F data matrices are available electronically on request from the authors. Maximum Parsimony (MP), as implemented in PAUP* version 4.0b10 (Swofford, 2003) was conducted with nucleotide substitutions equally weighted (Fitch parsimony, Fitch, 1971), and gaps treated as “missing” data. Heuristic searches were performed initially using 1000 random addition replicates, ACCTRAN optimization, TBR branch-swapping and MulTrees option. For ITS analysis, as a first step, no more than 10 trees were saved per replicate to minimize swapping on large numbers of sub-optimal trees. After that, all trees thus obtained were used as starting trees for a further search (swapping to completion) with MulTrees option in effect and a limit of 15000 trees saved. To assess support for each node, bootstrap analyses (Felsenstein, 1985) were performed with 1000 bootstrap replicates, TBR branchswapping and simple sequence addition. Bayesian Inference (Huelsenbeck & Ronquist, 2001; Lewis, 2001) under the model determined by MODELTEST were performed with MrBayes 3.0b4 (Huelsenbeck & Ronquist, 2001). Modeltest 3.06 (Posada & Crandall, 1998) was used to find the best-fitting substitution model, and the GTR+I+G model (nst = 6, rate = invgamma) was selected for both data matrices. The Markov chains, three heated and one cold, ran simultaneously starting from a random tree for 1×106 generations with trees sampled every 100 generations. Trees that preceded the stabilization of the likelihood value were discarded as burn-in. The majority-rule consensus tree containing posterior probabilities (PP; Larget & Simon, 1999) was built from the remaining sampled trees. All analyses were run twice and results from each of these runs were compared. Since hybridization and polyploidy are strongly involved in the evolution of Achillea, the reconstruction of phylogenies by parsimony based on DNA sequence variation (especially with the multi-copy nrITS) might be problematic because of reticulation (McDade, 1995). To assess the influence of polyploidy on tree topology, comparable phylogenetic analyses of the ITS data matrix were conducted for all taxa and for diploid taxa alone (Figs. 2 and 3). In all phylogenetic analyses, trees were rooted with species of Tanacetum (incl. sequences of Tanacetum corymbosum from GenBank). The inclusion of the stepwise more and more remote Brocchia, Aaronsohnia and Santolina (Watson & al., 2000; Oberprieler & Vogt, 2000; Francisco-Ortega & al., 2001) into the outgroups had no effect on the tree topologies. To help define the length of the intergenic spacers, sequences of ITS1/ITS2 and the trnL-F IGS of A. millefolium, A. ptarmica and some species of Otanthus, Anacyclus and Tanacetum from GenBank were used as

53 (3) • August 2004: 657–672


A 64/.88 62/.97

1 = Sect. Babounya 2 = Sect. Santolinoideae 3 = Sect. Ptarmica s.s. 4 = Sect. Anthemoideae 5 = Sect. Achillea s.l. 5a = “Sect. Filipendulinae” 5b = Sect. Achillea s.s. (excl. 5c) 5c = A.millefolium agg.

58/.81

55/.81

72/.88

65/.78 52/.89

61/1.00

--/.87 87/1.00

72/1.00

6

-- /.84

61/.95

89/1.00

90/1.00

5 76/.98 73/1.00

-- /.68

98/1.00

4

70/.97 65/.99

85/.81

90/1.00

3b

50/.95

1

-- /.98

3 64/.98 97/1.00

-- /.72

2

60/--

2b 100/1.00

2a 58/1.00

91/-83/.99

89/.99

A. pannonica 107 A. millefolium 155-8x A. setacea 69 A. setacea 93 A. setacea 21 A. setacea 76 A. setacea 60 A. setacea 23 A. setacea 09 A. millefolium 44 A. millef.ssp.sud. 55 A. distans 97 A. schmakovii 152 A. styriaca 91 A. collina 01 A. pratensis 45 A. roseoalba 03 A. asplenifolia 14 A. euxina 84 A. ceretanica-2x 16 A. ceretanica-4x 17 A. asiatica-2x 20 A. asiatica- 4x 150 A. lanulosa 132 A. alpina 06 A. wilsoniana 120 A. monticola 189 A. micrantha 300 A. virescens 62 A. clypeolata 74 A. nobilis 37 A. crithmifolia 35 A. grandifolia 67 A. clypeolata 40 A. biebersteinii 12 A. pseudopectinata 314 A. leptophylla 82 A. chamaemelifolia 291 A. absinthoides 68 A. ageratum 117 A. holosericea 70 A. filipendulina 36 A. clavennae 24 A. clavennae 175 A. fraasii 73 A. atrata 280 A. oxyloba 48 A. clusiana 176 A. ochroleuca 186 A. ligustica 07 A. nana 191 A. abrotanoides 53 A. multifida 88 A. erba-rotta 100 A. ×morisiana 99 A. moschata 25 A. macrophylla 286 A. schurii 281 A. schurii 313 A. pindicolassp.integ. 72 A. ptarmicifolia 157 A. biserrata 158 A. cretica 296 Otanthus maritimus 27 A. fragrantissima11 A. teretifolia 104 A. acuminata 118 A. salicifolia 340 A. salicifolia 85 A. impatiens 355 A. pyrenaica 190 A. ptarmica 369 A. wilhelmsii 26 A. wilhelmsii 95 Anacyclus clavatus 31 Tanacetum macrophyllum 310 T. corymbosum T. millefolium 192 T. vulgare 08 Brocchia cinerea 163 Aaronsohnia pubescens 164 Santolina africana 34

A. pannonica 107 A. millefolium 155-8x A. millef. ssp.sud.111 A. millefolium 108 A. styriaca 161 A. collina 01 A. pratensis 04 A. roseoalba 03 A. asplenifolia 14 A. euxina 84 A. ceretanica-2x 16

B 54/.96

A. virescens 146 A. schmakovii 152 A. setacea 21 A. asiatica-2x 124 A. lanulosa 132 A. wilsoniana 120 A. alpina 06 A. asiatica-4x 150 A. monticola 189 A. micrantha 300 A. nobilis 138 A. ochroleuca 186 A. leptophylla 82 A. filipendulina 36 A. clavennae 175 A. clavennae 24 A. distans 97 A. setacea 106 A. setacea 23 A. setacea 93 A. grandifolia 67 A. clypeolata 40 A. fraasii 73 A. chamaemelifolia 291 A. pindicola ssp integ. . 72 A. pseudopectinata 314 A. ligustica 07 A. crithmifolia 35 A. absinthoides 68 A. ageratum 117 A. biebersteinii 12 A. holosericea 70 A. coarctata 293 A. oxyloba 48 A. atrata 280 A. multifida 88 A. clusiana 176 A. abrotanoides 53 A. nana 191 A. moschata 25 A.× morisiana 99 A. erba-rotta 100

62/.91

V 63/.94

Insertion A

63/.90

-- /.92

IV Deletion B 94/1.00

II

50/.95 51/.99

III Indel F

A. macrophylla 286 A. schurii 313 A. schurii 281 A. acuminata118 A. salicifolia 340 A. ptarmica 369 A. impatiens 355 A. pyrenaica 190 A. biserrata 158 A. teretifolia 104 A. wilhelmsii 95 A. wilhelmsii 26 A. fragrantissima 11 Otanthus maritimus 27 Anacyclus clavatus 31 Tanacetum.vulgare 08 T. corymbosum Santolina africana 34

Insertion D

89/.99 78/1.00

64/.99

I 87/1.00

Outgroups

Fig. 2. A, strict consensus tree (> 15000 equally most parsimonious trees) from nrITS data (tree length = 349, CI = 0.60, RI = 0.78) for taxa of Achillea and related Anthemideae. Bootstrap percentages (> 50%) and posterior probabilities (from Bayesian Inference) are shown above branches; important clades are designated by arabic numerals below branches. B, strict consensus tree of 24 equally most parsimonious trees from trnL-F data (tree length = 52, CI = 0.92, RI = 0.98) for taxa of Achillea and related Anthemideae. “--” are values