CHLOROPLAST DNA PHYLOGENY OF THE WOODY …plantbiology.ucr.edu/faculty/Leeetal(AJB2005).pdf · 2072...

2072

American Journal of Botany 92(12): 2072–2085. 2005.

CHLOROPLAST DNA PHYLOGENY OF THE WOODY

SONCHUS ALLIANCE (ASTERACEAE: SONCHINAE) IN THE

MACARONESIAN ISLANDS1

CHUNGHEE LEE,2 SEUNG-CHUL KIM,2,5 KAREN LUNDY,3 AND

ARNOLDO SANTOS-GUERRA4

2Department of Botany and Plants Sciences, University of California, Riverside, California 92521 USA; 3Department of Ecology andEvolutionary Biology, University of California, Irvine, California 92697 USA; and 4Jardın de Aclimatacion de La Orotava,

Calle Retama Num. 2, 38400, Puerto de La Cruz, Tenerife, Canary Islands, Spain

Determining accurate phylogenetic relationships among the members of the woody Sonchus alliance presents challenges because ofan insufficient level of molecular variation and the convergent evolution of similar morphological traits in island settings. To obtain abetter resolved phylogeny and to test the potential role of hybridization and introgression, we sequenced all members of the alliancewith multiple populations for the ITS of nrDNA and over 4000 base pairs of coding and noncoding regions of cpDNA. The cpDNAphylogeny is not well resolved in the core members of the alliance (i.e., subg. Dendrosonchus and genus Taeckholmia), but like theITS tree, it has identified basal lineages of monotypic genera. The cpDNA data set was not significantly different from that of ITS,and subsequent combined analysis provided a better resolved and supported phylogeny within the alliance. The combined ML treeidentified the same basal lineages, suggested nonmonophyly of Dendrosonchus and Taeckholmia, and did not support either Boulos’or Aldridge’s infrasubgeneric classification system. Assessment of the role of hybridization and introgression was limited due to poorresolution in the cpDNA phylogeny. The combined analysis supports a Gran Canaria origin for the alliance and two subsequent longdistance dispersal events to Madeira and Cape Verde islands.

Key words: adaptive radiation; cpDNA phylogeny; Dendrosonchus; hybridization; ITS; Macaronesia.

High in endemism and rich in morphologically unusual or-ganisms, the special character of oceanic islands was recog-nized by Carlquist (1974), who described them as ‘‘naturallaboratories’’ for the interpretation of plant evolutionary pro-cesses. Numerous molecular phylogenetic studies have sub-sequently shed further light on the historical patterns of or-ganismal evolution and their underlying mechanisms in islands(see examples in Baldwin et al., 1998). A lack of accumulatedmolecular sequence variation and the convergence of morpho-logical traits have, however, limited the ability to make senseof the evolutionary past of many endemic plants, especiallythose suspected to be the products of adaptive radiation. Thisinadequacy is further confounded by the possibly misleadinginferences that have been based on single-marker phylogenies(i.e., ‘‘gene tree’’ vs. ‘‘species tree’’ differences), as well asthe poorly documented role of hybridization and introgressionin adaptive radiations (for an exception, see Howarth andBaum, 2005).

The woody Sonchus alliance (Asteraceae: Sonchinae),which includes six genera and ca. 31 species, is a premierexample of adaptive radiation among angiosperms in Maca-ronesia (Aldridge, 1975, 1979). It comprises 19 species ofwoody pachycaulous members of Sonchus (subg. Dendroson-

1 Manuscript received 19 January 2005; revision accepted 25 August 2005.The authors thank Dan Crawford for helpful comments on an earlier version

of the manuscript; Dr. Julie Caujape-Castells, Jacinto Leralta, Ana Calero,Olga Baeta, and Paulo Gouveia for assistance with field support and samplecollections; Drs. Roberto Jardim and Susana Fontinha for issuing permits tocollect plants in Madeira. We are also greatly indebted to Jutta Burger for herGerman translation and considerations and Pesach Lubinsky for careful ed-iting. This paper was greatly improved by suggestions from two anonymousreviewers. This work was supported in part by an Academic Senate Grant,Regent’s Faculty Fellowship, The Genomics Institute Grant from UC-River-side to S.-C.K. and NSF DEB 9521017 to D.J.C. and S.-C.K.

5 Author for correspondence (e-mail: [email protected])

chus Sch. Bip. ex Boulos), seven species of Taeckholmia Bou-los, four monotypic genera [Babcockia Boulos, Lactucoson-chus (Sch. Bip.) Svent., Sventenia Font Quer, and Chryso-prenanthes (Sch. Bip.) Bramwell], and one species of subg.Sonchus (S. tuberifer Svent.). The entire alliance is endemicto the three archipelagoes (i.e., the Canaries, Madeira, andCape Verde) in the north Atlantic, with the one exception inSonchus pinnatifidus Cav., which occurs in both the Canariesand Morocco. All but four species of the alliance are endemicto the Canaries (Fig. 1). The members of the alliance displayextensive morphological, ecological, and anatomical diversity,yet all taxa have a uniform chromosome number (Aldridge,1977, 1978; Ardevol Gonzalez et al., 1993). Natural interspe-cific and even intergeneric hybrids have often been observedand documented in the islands (Aldridge, 1976; Perez de Paz,1976; Hansen and Sunding, 1993). Furthermore, artificial in-terspecific and intergeneric hybrids are easily produced ingreenhouse conditions (Aldridge, 1975; S.-C. Kim, unpub-lished data). This indicates that, like some other island endem-ic plants that underwent rapid radiation, there are no strongreproductive barriers within the alliance and that the primaryisolation mechanism is geographical and/or phenological iso-lation.

Several initial phylogenetic analyses revealed the monophy-ly of the woody Sonchus alliance despite its extensive mor-phological and ecological diversity and geographical proxim-ity of the Macaronesian Islands to a large continental sourcearea (Kim et al., 1996a, b, 1999a; S.-C. Kim, unpublisheddata). This implies that all members of the alliance were de-rived from a common ancestor in the Macaronesian Islands,primarily from the Canaries. These studies also suggested thatall members were derived from possibly a single dispersalevent from a more widely distributed European/African her-baceous Sonchus species. The low average sequence diver-

December 2005] 2073LEE ET AL.—CPDNA PHYLOGENY OF THE WOODY SONCHUS ALLIANCE

Fig. 1. The three Macaronesian archipelagos and distribution of the members of the woody Sonchus alliance on each islands. The asterisk indicates a singleisland endemic species in the Canaries. S. 5 Sonchus; T. 5 Taeckholmia.

gence and the polytomy in the phylogenetic trees suggest arapid radiation of major lineages early in the history of thealliance. An enzyme electrophoresis study was also conductedto assess genetic diversity within and divergence among spe-cies of the alliance (Kim et al., 1999b). This study found rel-atively high mean genetic identities (i.e., 0.804) for all species,supporting the genetic cohesiveness of the alliance. In addi-tion, species in the alliance have about 50% higher total ge-netic diversity (HT) than the mean for other oceanic endemics

(see reviews de Joode and Wendel, 1992; Crawford et al.,2001). Lastly, these previous studies hypothesized that the ra-diation of the alliance took place in the Canary Islands, mostlikely either Gran Canaria or Tenerife, during the late Mioceneor early Pliocene.

Despite the previous studies, several important issues andquestions remain to be considered in the evolution of thewoody Sonchus alliance. First, resolution of phylogenetic re-lationships within the alliance has proved a difficult challenge,

2074 [Vol. 92AMERICAN JOURNAL OF BOTANY

in part because of a lack of sequence variation in the ITS ofnrDNA. As a result, relationships among taxa are poorly re-solved, with the exception of identifying lineages that radiatedearly (Kim et al., 1996b). Second, very low bootstrap anddecay values, with the exception of one clade, supported re-lationships within the alliance. Third, a single nuclear marker,especially the ITS region of nrDNA, may mislead phyloge-netic inference due to several molecular genetic processes (Al-varez and Wendel, 2003). Fourth, taxon sampling was limitedin the previous phylogenetic study. Lastly, the ITS phylogenyindicated potential hybridization and introgression events be-tween subg. Dendrosonchus and the genus Taeckholmia: twospecies of subg. Dendrosonchus, S. gonzalezpadronii and S.ortunoi, were clustered with several species of Taeckholmia ina very strongly supported clade (98% bootstrap value). There-fore, it was necessary to obtain additional molecular markers,especially from the chloroplast genome, to estimate accuratelythe phylogenetic relationships within the alliance and also totest for hybridization and introgression. Furthermore, it is alsocritical to sample multiple populations, especially from speciesoccurring in more than one island and also for widely distrib-uted ones on a single island.

The objectives of this study were (1) to reconstruct phylo-genetic relationships within the alliance, based on ITS ofnrDNA, with complete and detailed geographic sampling, (2)to independently reconstruct a cpDNA phylogeny based on thematK gene and noncoding regions, (3) to evaluate the infra-subgeneric or infrageneric classifications of the subg. Dendro-sonchus and genus Taeckholmia, and (4) to assess hybridiza-tion and introgression hypotheses based on ITS and cpDNAphylogeny.

MATERIALS AND METHODS

Taxon sampling and classification—In this study, we sampled multipleindividuals and populations of all species of the alliance (Appendix). Specif-ically, we included multiple populations of all Madeiran Dendrosonchus spe-cies and several species with restricted distributions in the Canaries (i.e., Son-chus tuberifer, S. brachylobus, S. gandogeri, S. bornmuelleri, S. fauces-orci,T. capillaris, T. canariensis, T. arborea, and T. microcarpa). We also sampledmultiple populations of species that have wide geographic distributions onsingle or multiple islands: Babcockia, Chrysoprenanthes, Lactucosonchus, S.congestus, S. acaulis, S. canariensis, S. palmensis, S. hierrensis, T. pinnata,and T. heterophylla. We also included one newly described species, S. wild-pretii, as well as S. bornmuelleri and the S. gummifer complex sensu Aldridge(S. gummifer, S. radicatus, and S. tectifolius).

There have been many different classifications, name changes, and sugges-tions for members of the woody Sonchus alliance (e.g., Aldridge, 1975, 1976;Reifenberger and Reifenberger, 1996; Sennikov and Illarionova, 1999; Bram-well and Bramwell, 2001; Bramwell, 2003). It seemed premature to makename changes and to propose new classification with a lack of strong molec-ular and/or morphological data. Therefore, we adopted the classfication ofBoulos (1967, 1972, 1974a, b), which represents the first complete revisionof the genus Sonchus and several other segregate genera.

PCR reaction and DNA sequencing—Total genomic DNA was isolatedfrom leaf tissue using the CTAB method of Doyle and Doyle (1987) andDNeasy plant mini kits (QIAGEN, Valencia, California, USA). The ITS ofnrDNA was amplified as described previously (Kim et al., 1996a, b), andcpDNA trnT-L-F noncoding regions were separately amplified using the uni-versal primers (Taberlet et al., 1991) under the same conditions as ITS. PCRproducts of approximately 100 mL of each sample were purified with theQiaQuick PCR Purification kit (QIAGEN). Direct sequencing of PCR prod-ucts was performed with ABI PRISM BigDye Terminator v3.1 Ready Reac-tion Cycle Sequencing kit (Applied Biosystems, Foster City, California,

USA), and extension products were purified and subsequently separated onABI377 and 3730 automated sequencing machines (Applied Biosystems). Forsequencing reactions for ITS, we used two additional internal primers: ITS3and ITS2 (White et al., 1990). For trnT-L-F regions, we used the same PCRamplification primers for the sequencing reaction. In the case of the matKgene and adjacent noncoding regions (59trnK-matK, matK-39trnK, and trnK-psbA), we used two primers designed by Johnson and Soltis (1995) for PCRamplification (trnK-3914F(dicot), 59-GGG GTT GCT AAC TCA ACG G-39and psbA-R, 59-CGC GTC TCT CTA AAA TTG CAG TCA T-39). Reactionconditions were an initial 5 min at 948C followed by 40 cycles of 1 min at948C, 1 or 2 min at 508C, and 2 min at 728C, with a final extension time of10 min at 728C. For the sequencing reaction, we designed eight internal prim-ers based on Sonchus species and used them with one PCR primer (psbA-R).The eight internal primers are as follows: matK665R (59-TGA TAC ATA GTGCGA TAC AGT-39), matK665F (59-ACT GTA TCG CAC TAT GTA TCA-39), matK892R (59-CGT TTC ACA ATT AGT AAG C-39), matK1282F (59-GTC ATA ATT GGG ATA GTC-39), matK1939R (59-GCC CAA AGC GGTCAA TAA-39), matK1934F (59-TGA TAT TAT TGA CCG CTT TGG GC-39), matK2430R (59-ACT CAG TTC CTT CAT TAG A-39), and matK2378F(59-TGA TGT AGT ACG GAG TAA GGA TG-39).

Base calling and sequence editing were performed with Sequencher 4.1(Gene Codes, Ann Arbor, Michigan, USA). Newly obtained ITS sequenceswere manually entered and visually aligned to the previously existingMacClade (version 4; Maddison and Maddison, 2000) data matrix. Both trnT-L-F and matK gene sequences were combined and manually aligned usingMacClade due to low sequence variation. All data matrices are available uponrequest from the corresponding author.

Phylogenetic analyses—For each ITS and combined cpDNA data set, phy-logenetic analyses using Fitch parsimony were performed with PAUP* (ver-sion 4.0; Swofford, 2001) using the HEURISTIC search option with TBRbranch swapping and MULPARS on. Sonchus palustris was used as an out-group based on the previous studies (Kim et al., 1996a, b). Phylogeneticanalyses of the subtribe Sonchinae based on ITS (73 species and three sub-species; a total of 123 accessions) and matK gene sequences (58 species andtwo subspecies; a total of 95 accessions) strongly suggest that the entire al-liance is monophyletic (100% and 88% bootstrap support in the ITS andcpDNA tree, respectively; S.-C. Kim, unpublished data). Thus, we used S.palustris, one of the sister species based on the previous ITS tree, to root thetrees. Gaps were treated as missing data because all but one were phyloge-netically uninformative (see Results, trnT-L-F and matK data set). Supportfor groups was examined by 1000 bootstrap replicates (Felsenstein, 1985)using the HEURISTIC search option from a simple addition sequence withTBR branch swapping. Pairwise sequence divergence was calculated usingthe Kimura 2-parameter method (Kimura, 1980) and a neighbor-joining (Saitoand Nei, 1987) tree was constructed using PAUP*. Each data set was alsoanalyzed using likelihood methods to determine the stability of the parsimonyresults with an explicit model-based approach (Felsenstein, 1981). ‘‘Optimal’’models of molecular evolution were chosen using the likelihood ratio test(Goldman, 1993; Whelan and Goldman, 1999) implemented in ModelTest(Posada and Crandall, 1998). Model parameters then were imported intoPAUP*, and heuristic searches were executed. Model parameters were re-estimated from the initial maximum likelihood (ML) tree and the processrepeated until the topology remained constant. ML bootstrap analyses with50 replicates were conducted using the same parameter values obtained fromthe ModelTest. Congruence between ITS and cpDNA data sets were testedusing the incongruence length difference (ILD) test (Farris et al., 1995) asimplemented by the partition homogeneity test in PAUP* for 50 replicates(heuristic search, simple addition, TBR branching swapping), each saving amaximum of 1000 most parsimonious trees per replicate.

Character optimization was performed using Fitch parsimony (ACCTRANoptimization) for island distribution using MacClade 4.0 (Maddison and Mad-dison, 2000). Character state for island distribution of the outgroup specieswas coded as ‘‘0,’’ and it was different from the eight states assigned to theingroup island endemics. Character optimization was mapped onto one of thetwo ML trees (Fig. 4) based on the combined data set.


RESULTS

ITS data set—Complete overlapping sequences for the en-tire ITS region, including 5.8S, were obtained using the fourITS sequencing primers (White et al., 1990). Lengths of ITS1and ITS2 in the newly sequenced members of the woody Son-chus alliance are within the size range reported previously(Kim et al., 1996a, b, 2004). For the initial phylogenetic anal-ysis, we included a total of 101 populations (100 ingroup andone outgroup), representing 30 species, two subspecies, andthree varieties (65 of which were newly sequenced in thisstudy). Species with identical ITS sequences were excluded,leaving a total of 58 ingroup populations, with the same num-ber of species as in the initial analysis, to be further analyzed.

Average pairwise sequence divergence between outgroup(S. palustris) and ingroup taxa was 7.39%, while within in-group divergence was 1.74%. Average sequence divergenceamong segregate genera (excluding Taeckholmia) was 2.11%,and 2.43% between segregate genera and subg. Dendroson-chus, including Taeckholmia. Within subg. Dendrosonchusand Taeckholmia, average sequence divergence was 1.57%and 0.68%, respectively. These average values were somewhatlower than those previously reported (Kim et al., 1996b) be-cause we included many conspecific populations in this study.

A total of 479 aligned characters, excluding the 5.8S region,were used for phylogenetic analyses. We found 394 constantcharacters (82%), 39 variable parsimony-uninformative char-acters (8%), and 46 parsimony-informative characters (9.6%).The heuristic search resulted in more than 500 000 trees, witha tree length (TL) of 112, a consistency index (CI) of 0.8304(excluding uninformative characters 5 0.7324), and a reten-tion index (RI) of 0.9181. The strict consensus tree (notshown) of this search is poorly resolved, except it shows earlydivergence of Babcockia from the remaining taxa and recog-nizes two major clades within the alliance (see Fig. 2, MLtree). Because the maximum likelihood (ML) trees are a subsetof the maximum parsimony (MP) trees, we present and baseour discussion on the ML result (Fig. 2). The LRT modelfound 34 trees (2ln 5 1367.8416), while the AIC modelfound 120 trees with higher likelihood scores (2ln 51362.0342). These trees represent a subset of MP trees andthus one of 120 trees based on AIC will be used for discussion.

The ML analysis resulted in a tree similar to the previousMP tree (Kim et al., 1996b). Babcockia is suggested to havediverged first within the alliance. The phylogenetic relation-ships of the three monotypic genera, Chrosyprenanthes, Sven-tenia, and Lactucosonchus, and Sonchus tuberifer (hereafterreferred to as the ‘‘basal lineages’’) are unresolved. A majorradiation of the remaining taxa, the subg. Dendrosonchus andgenus Taeckholmia, follows the radiation of basal lineages(Fig. 2). Within Dendrosonchus and Taeckholmia, four majorgroups can be recognized. Group I includes only S. fauces-orci from Tenerife, Canary Islands. Group II is comprised ofthree species of Dendrosonchus (S. pinnatifidus, S. wildpretii,and S. brachylobus) and one species of Taeckholmia (T. ar-borea). Group III includes all Madeiran Dendrosonchus spe-cies in group III-A and strongly supported group III-B withmajority of Dendrosonchus species (i.e., S. congestus, S. acau-lis, S. canariensis, S. palmensis, S. gandogeri, S. hierrensis,and S. bornmuelleri). Group IV is composed of the majorityof Taeckholmia species with several Dendrosonchus species(i.e., S. radicatus, S. gummifer, S. tectifolius, S. ortunoi, and

S. gonzalezpadronii), and this group is one of the few stronglysupported clades in the ITS ML tree.

The trnT-L-F and matK data set—The aligned cpDNA se-quences of 73 accessions representing 31 species and 1 sub-species, including S. palustris as an outgroup, resulted in 4249base pairs (bp). The size of the trnT-L intergenic spacer rangedfrom 579 to 610 bp, and one 13-bp deletion was found inLactucosonchus, and one 17-bp deletion was found in all in-group taxa. The trnL intron ranges from 437 to 449 bp; one11-bp indel was shared by all ingroup taxa, except Sventeniaand Babcockia. The length of trnL-F varied from 335 to 336bp, and no indels were found in this region. In case of thematK gene and adjacent noncoding regions, the length variedfrom 2828 to 2846 bp, and one 17-bp direct repeat insertionwas found in T. pinnata (Gran Canaria population). Becauseall chloroplast markers are inherited as a single unit and themarkers we sampled had low substitution rates, we combinedall trnT-L-F and matK sequences for phylogenetic analyses.

Of the 4249 aligned base pairs, 4187 bp (98.5%) were con-stant and 62 bp (1.5%) were variable. Of the 62 variable sites,33 (0.78%) were parsimony uninformative, while 29 (0.68%)were parsimony informative. MP analysis found 486 equallyparsimonious trees with a TL of 68, a CI of 0.9118 (0.8286excluding uninformative), and a RI of 0.9577. The strict con-sensus tree (not shown) and the NJ tree were similar to theML tree (Fig. 3); thus our discussion will be based on the MLtree. ML analyses employing LTR (TrN 1 I 1 G) and AIC(GTR 1 I) criteria recovered a single tree (Fig. 3). ThecpDNA ML tree shows early divergence of the three mono-typic genera, Sventenia, Babcockia, and Chrysoprenanthes,without any resolution, within the alliance. The monotypic ge-nus Lactucosonchus, then diverged, and it turns out to be sisterto the strongly supported (94%) clade that includes subg. Den-drosonchus and genus Taeckholmia. Within this clade, rela-tionships, however, are poorly resolved and, unlike in the ITSML tree (Fig. 2), S. tuberifer is deeply nested within this poor-ly resolved clade. The group II, weakly supported in the ITSML tree, is no longer monophyletic, but the four species inthis group are in turn sisters to the remainder of the subg.Dendrosonchus and genus Taeckholmia.

Partition homogeneity test and combined analysis—Thepartition homogeneity test for the ITS and cpDNA indicatedthat the partitions were not significantly different from randompartitions (P 5 0.12). Therefore, we combined the two datasets to explore whether resolution and support would be im-proved by increasing the amount of sequencing data. Only 48accessions (representing 30 species and one subspecies), whichare completely overlapping between the two data sets, wereanalyzed. The MP analysis found 69 418 equally parsimonioustrees with a TL of 172, a CI of 0.8605 (0.75 excluding un-informative), and a RI of 0.9080. The strict consensus tree isvery similar to the tree based on ML analysis, and thus ourdiscussion will be based on the ML tree (Fig. 4). ML analysesof combined data sets employing different models (TrN 1 I1 G and GTR 1 I for LRT and AIC criteria, respectively)recovered two trees with identical branching patterns in eachinstance, and thus we present the ML topologies based on AICcriteria only. The only difference between the two trees is inthe relationship between T. arborea and S. wildpretii (cladeII; Fig. 4). The ML phylogeny based on a combined data set(Fig. 4) is similar to both the ITS and cpDNA trees. However,


Fig. 2. New ITS phylogeny of the woody Sonchus alliance. This is one of 120 maximum likelihood trees (2ln 5 1361.8695). Collapsed branches in thestrict consensus tree are indicated by dashed lines. Bootstrap supports .50% are shown above and below branches. S. ust. 5 Sonchus ustulatus.

it is more similar to the ITS and provides better resolutionthan separate analysis. Like the ITS tree, it identified basallineages, such as Babcockia, Sventenia, Chrysoprenanthes,and Lactucosonchus: Babcockia diverged first, then Sventeniaand Chrysoprenanthes, and finally Lactucosonchus. Themonotypic genus Lactucosonchus is shown to be sister to theclade containing all members of subg. Dendrosonchus and ge-nus Taeckholmia, including S. tuberifer. The relationshipswithin basal lineages were very weakly supported (64 and53%), but the combined analysis resolved the relationships

within them, except in Sventenia and Chrysoprenanthes. Son-chus tuberifer, the only species of subg. Sonchus in the alli-ance, is deeply embedded within the clade containing subg.Dendrosonchus and genus Taeckholmia; it is sister to S. fau-ces-orci, which occurs sympatrically (Teno, Tenerife), but withvery weak bootstrap support (,50%) (hereafter, these two spe-cies will be referred to as clade II). Monophyly of core mem-bers of the alliance (i.e., Dendrosonchus and Taeckholmia),including S. tuberifer, was strongly supported (92%) in theML tree, and four clades were identified with moderate to


Fig. 3. The cpDNA phylogeny of the woody Sonchus alliance. This is a single maximum likelihood tree based on trnT-L-F, matK gene and its adjacentregions (2ln 5 6255.0550). Bootstrap supports .50% are shown above and below branches.

strong support (clades I and II were very weakly supported,,50%). The relationships within the core members of the al-liance were very similar to those in the ITS phylogeny; allfour major groups in the ITS tree were also recognized in thetree of combined analysis.

Overall, the combined data set provides a more highly re-solved and strongly supported tree compared to separate anal-ysis of each data set. The average bootstrap value in ITS,cpDNA, and combined analyses was 59, 70, and 77%, re-spectively. In addition, the number of resolved internal nodesfor each data set and combined data set analysis was 3, 2, and

12, respectively. Therefore, the ML tree (Fig. 4) is used as theworking phylogenetic hypothesis for the woody Sonchus alli-ance.

Optimization of biogeographic data—The most parsimo-nious character optimization for island distribution require 17steps (Fig. 5). The character reconstruction suggests that an-cestral lineages of the woody Sonchus alliance originated inthe Canary Islands, followed by colonization of the Madeiraand Cape Verde Archipelagos. Gran Canaria is the most likely


Fig. 4. Combined ITS and cpDNA phylogeny of the woody Sonchus alliance in the Macaronesian Islands. This tree is one of the two maximum likelihoodtrees (2ln 5 7716.1533). Bootstrap supports .50% are shown above branches. S. ust. 5 Sonchus ustulatus.

ancestral area of the entire woody Sonchus alliance, and thusthe Canary Islands represent the ancestral archipelago.

DISCUSSION

The woody Sonchus alliance phylogeny and taxonomicimplications—This study provides the best estimate to date of

the phylogenetic relationships of the woody Sonchus alliancein the Macaronesian Islands. It is the first to include all mem-bers of the alliance and has the advantage of more extensivesampling from multiple populations and more clearly resolvedrelationships than reported in previous studies (Kim et al.,1996b). However, we are still far from a complete understand-


Fig. 5. Fitch parsimony (ACCTRAN) optimization of island distribution onto one of the two maximum likelihood trees based on the combined ITS andcpDNA data set.

ing of the alliance phylogeny because several internal branchesremain weakly supported and relationships between closelyrelated species are still unresolved. Nevertheless, this studyprovides important novel relationships among the membersthat have implications for further taxonomic and phylogeneticstudies of the woody Sonchus alliance.

Monophyly of the alliance and relationships within basallineages—The monophyly of the woody Sonchus alliancewithin the subtribe Sonchinae sensu Bremer was strongly sup-ported in this as well as in previous studies (Kim et al., 1996a,b; S.-C. Kim, unpublished data). This clearly indicates that allmembers of the alliance diverged rapidly on the islands aftera common ancestor dispersed from a continental source. Themonophyly of the woody Sonchus alliance contrasts with cases

of multiple introductions of the Macaronesian endemics (seeexamples in Carine et al., 2004). We still know little about theputative continental progenitor(s) or the closest relatives, butit is evident that all members of the alliance on the three ar-chipelagoes share a recent common ancestor. Very low se-quence divergences in the ITS and cpDNA and the many un-resolved relationships in the phylogenetic tree suggest that theentire alliance has undergone a recent and rapid radiation.

The combined ML tree indicated that the monotypic genusBabcockia diverged first. Our previous work (Kim et al.,1996a, b) suggested that the herbaceous monotypic genus Lac-tucosonchus diverged first. Babcockia platylepis, a small shrubwith very large capitula that is endemic to Gran Canaria, isevidently the oldest member of the alliance. Babcockia iswidely distributed in higher elevations (800–1600 m) of Gran


Canaria, which is approximately 14–16 million years old. Ithas been considered to be one of subg. Dendrosonchus spe-cies, Sonchus platylepis, by various authors (Aldridge, 1975,1976, 1979; Sennikov and Illarionova, 1999; Bramwell andBramwell, 2001). If we, however, recognize the three mono-typic genera, Sventenia, Lactucosonchus, and Chrysoprenan-thes, the current molecular phylogenetic study strongly sug-gests that Babcockia should also be retained as a monotypicgenus because it is not closely related to any members of subg.Dendrosonchus and genus Taeckholmia (Fig. 4). After initialdivergence of Babcockia, two monotypic genera in Gran Can-aria, Sventenia and Chrysoprenanthes, diverged. Sventenia bu-pleuroides is a small caudex perennial (up to 30 cm) and veryrare on shady, humid, vertical rocks in the western region ofGran Canaria. It has several unique morphological features,such as entire leaves, beaked cypselas, and yellow-tipped glan-dular hairs covering the inflorescence stalk, peduncles, andinvolucral bracts. Chrysoprenanthes pendula, once placed inthe genus Prenanthes, commonly occurs in the mountain cliffson the south and west sides of Gran Canaria. It is a smallshrubby cliff plant and is unique in having heads with only5–6 florets. The close relationship between the two genera wasunexpected because they do not share any morphological char-acters within the alliance. The geographic distribution patternsof the two taxa, however, suggest they could share a mostrecent common ancestor.

The monotypic genus Lactucosonchus is a herbaceous pe-rennial with extremely long tuberous roots (but see Bramwelland Bramwell, 2001) and is sister to the core members of thealliance, including S. tuberifer. The early divergence of Lac-tucosonchus within the alliance has been weakly supported inprevious studies. Because it also occurs locally on a geologi-cally young island, La Palma (,2 Myr), it is plausible thatLactucosonchus is not the oldest member of the alliance. Thecloser relationship of Sonchus tuberifer to the core alliancemembers rather than to the basal lineages was unexpected(Figs. 3 and 4). Because the bootstrap support for the corealliance is very weak (31%) in the ITS ML tree (Fig. 2) anda branch was collapsed in the strict consensus MP tree (notshown), it remains possible that S. tuberifer is closely relatedto the core alliance members. Sonchus tuberifer occurs in ageologically old part of western Tenerife, Teno (7.4 Myr), withS. fauces-orci, and it is possible that it shares its most recentcommon ancestor with the core alliance rather than to the taxain the basal lineages. If this is true, the herbaceous habit withtuberous roots shared by Lactucosonchus and S. tuberifer musthave evolved independently; the core alliance members areshrubs or small shrubs and have tap-roots or rhizomes. In ad-dition, the circumscription of the subg. Dendrosonchus andgenus Taeckholmia has to be re-evaluated (discussed later).

Monophyly of subg. Dendrosonchus and genus Taeckhol-mia—The taxonomic treatments of the subg. Dendrosonchushave all more or less followed the original workers in theseparation of S. leptocephalus and similar taxa, in the retentionof S. pinnatus and similar taxa within the body of the broad-leaved group, and finally in the groups of S. fruticosus withthe pinnatus group. Boulos’ treatment (1967, 1974a, b) dif-fered only in the rank accorded to each group and in the rec-ognition of the large headed S. platylepis as distinct by placingit in a separate genus, Babcockia. In addition, Boulos (1967)pointed out that S. leptocephalus and some allied species (withnarrow leaves and small capitula) are different in many re-

spects from the rest of Dendrosonchus species (broader leavesand large capitula) and placed them in a new genus Taeck-holmia. However, Aldridge (1976a, b), based on her extensivemorphological investigation, strongly argued that the generaBabcockia and Taeckholmia cannot be distinguished from thesubg. Dendrosonchus and placed them within it.

The molecular phylogenetic study presented here stronglysuggests that neither subg. Dendrosonchus nor genus Taeck-holmia are monophyletic (Fig. 4). For example, clade II in-cludes two broader leaved Dendrosonchus species with T. ar-borea and S. wildpretii (a species of highly dissected leaveswith small capitula, but placed in the subg. Dendrosonchusbased on the authors’ broader view of the subgenus; Reifen-berger and Reifenberger, 1992). In addition, four taxa of thesubg. Dendrosonchus (S. gonzalezpadronii, S. orutnoi, S. rad-icatus, and S. tectifolius) are strongly clustered in clade IVwith several typical Taeckholmia taxa. Unfortunately, thecpDNA tree (Fig. 3) was not resolved enough to determinewhether this is due to hybridization and introgression in cladeIV.

The combined ML tree (Fig. 4) allows us to reevaluate thedifferent classification systems and species delimitations ofBoulos (1972) and Aldridge (1976). Aldridge treated threespecies of Dendrosonchus, S. pinnatus, S. canariensis, and S.palmensis, as three subspecies of S. pinnata. This study, incontrast, suggests that S. pinnatus, an endemic to Madeira, isnot closely related to any of the two species. Rather, it is close-ly related to other Madeiran Dendrosonchus species, S. ustu-latus and S. fruticosus, which indicates the common origin ofall Madeiran species. This supports Boulos’ species delimita-tion, recognizing them as three distinct species. Aldridge alsorecognized three subspecies in S. radicatus: subsp. radicatus(north coast, Tenerife), subsp. gummifer (south coast, Tener-ife), and subsp. tectifolius (south side of Anaga, east Tenerife).In the cpDNA tree (Fig. 3), the Sonchus radicatus complexsensu Aldridge (only two subspecies radicatus and tectifoliuswere included) represents the highly unresolved part of theclade. However, the ITS tree suggests that this complex is notmonophyletic: subsp. radicatus and tectifolius are more close-ly related to S. ortunoi and S. gonzalezpadronii than to subsp.gummifer. Subspecies gummifer seems to be closely related toother Taeckholmia species in the group IV (Fig. 2). This doesnot support the broader view of Aldridge for the S. radicatuscomplex.

Some of Aldridge’s species delimitations are supported bythe current phylogenetic study. For example. T. capillaris, andT. microcarpa were treated as the same subspecies of S. lep-tocephalus subsp. capillaris. The combined tree (Fig. 4) showsthat they are sister taxa that have an identical cpDNA haplo-type. However, the monophyly of S. leptocephalus, includingtwo subspecies, subsp. leptocephalus (T. pinnata and T. can-ariensis) and subsp. capillaris, is left unresolved in this study.

Neither Boulos’ nor Aldridge’s infrasubgeneric and infra-generic classfication was supported by the current study. Forexample, Boulos recognized three sections within the subge-nus Dendrosonchus, and they were polyphyletic. He also rec-ognized two subgenera within Taeckholmia, Taeckholmia (T.pinnata, T. capillaris, T. canariensis, and T. microcarpa) andPseudodendrosonchus (T. heterophylla, T. regis-jubae, and T.arborea), but the two subgenera are also polyphyletic (Fig. 4).Aldridge (1976) divided the subg. Dendrosonchus into twosections, Dendrosonchus and Atalanthus, but these two sec-tions are highly polyphyletic. Therefore, this study strongly


suggests that the previous classification systems are artificialand need revision.

Monophyly and origin of Dendrosonchus species in Ma-deira and Cape Verde islands—This study suggests that Den-drosonchus species in Madeira are monophyletic (Fig. 4; cladeIII, A) and that they originated most likely from Tenerife, theCanary Islands (Fig. 5). Although the Madeira archipelago isvery old (ca. 30 Myr), a single long-distance dispersal event(ca. 400 km; Fig. 1) of a common ancestor from Tenerife islikely responsible for the origin of the four Dendrosonchustaxa in Madeira. Colonization of Madeira from the Canariesinvolves dispersal against the prevailing trade winds, suggest-ing that birds carried seed to Madeira. The monophyly of theMadeiran clade is somewhat weak (66%), but all MP and MLtrees support the monophyly, with an exception in the cpDNAML tree (Fig. 3). Nonmonophyly of the Madeiran clade in thecpDNA tree is due to very low sequence divergence amongspecies in the clade. Within the Madeiran clade (Fig. 4), twoclades are further recognized with strong support; S. pinnatus-S. fruticosus and S. ustulatus (subsp. ustulatus and subsp. mad-erensis). Two subspecies of S. ustulatus, subsp. ustulatus andsubsp. maderensis, are common on dry rocky and sunny areason the south and north coast of Madeira, respectively. In con-trast, S. fruticosus occurs in the Laurisilva and moist ravinesin the interior of Madeira, mainly at altitudes of 800–1200 m.Sonchus pinnatus occurs primarily on rocky slopes, mainly ataltitudes of 1000–1400 m (but it also occurs at lower alti-tudes). The combined tree (Fig. 4) suggests that the commonancestor of these taxa arrived in the coastal areas of Madeiraand speciated into higher altitudes. There is also a gradualchange in their habit from herbaceous caudex perennial (S.ustulatus subsp. ustulatus and subsp. maderensis; reaching 30cm in height) to small shrub (S. pinnatus; reaching 2 m) andeventually to tall tree (S. fruticosus; reaching 4 m).

There is only one species of Dendrosonchus, S. daltonii, inthe Cape Verde Islands, which is 1600 km southwest of theCanaries (Fig. 1). Sonchus daltonii, a rosette shrub, is a crit-ically endangered species in the island, with only ca. 30 in-dividuals remaining (Gomes et al., 1995, 1999). The combinedtree shows that S. daltonii represents a highly derived specieswithin the alliance and is closely related to S. hierrensis, S.bornmuelleri, and S. gandogeri, which occur on geologicallyyoung islands (i.e, La Palma, El Hierro). Therefore, the currentstudy strongly suggests that S. daltonii originated from a Ca-nary Island ancestor. This requires that the long-distance dis-persal from the western Canary Islands was a single eventfacilitated by prevailing northeasterly trade winds, which wereaccentuated during the Pleistocene (Rognon and Coude-Gaus-sen, 1996) and potentially by birds migrating between the twoarchipelagoes.

Hybridization and introgression—The initial ITS phylog-eny of the woody Sonchus alliance (Kim et al., 1996b) indi-cated potential hybridization and introgression events betweensubg. Dendrosonchus and Taeckholmia. For example, two spe-cies of subg. Dendrosonchus, S. gonzalezpadronii and S. or-tunoi, were clustered with several species of Taeckholmia ina very strongly supported clade (98% bootstrap value and adecay index of 3; Fig. 3, Kim et al., 1996b). Although eachof these Dendrosonchus and Taeckholmia species occurs onthe island of Tenerife and La Gomera, they are found in rad-ically different environmental settings (Taeckholmia species

are xerophytes and the two Dendrosonchus species are pachy-caulous perennials or robust shrubs in mesic habitats) and havedivergent phenotypes. The incongruence between the taxono-my (based on morphological similarities among the Taeckhol-mia species) and the placement of those taxa in the ITS phy-logeny has several possible explanations. One explanation forthis nonconcordance between the phenotypic and ITS char-acters is that subgenus Dendrosonchus is paraphyletic. A sec-ond explanation is that there has been a reticulation event sincethe origin of subg. Dendrosonchus, possibly involving the pro-genitors of the two Dendrosonchus and five Taeckholmia spe-cies that fall into the common clade. Lastly, lineage sorting ofancient polymorphic ITS sequences for rapidly radiated line-ages in the alliance can also explain this pattern.

The new ITS and combined ML trees shows that three ad-ditional species in the Sonchus radicatus complex in Tenerife,S. radicatus, S. gummifer, and S. tectifolius, were also clus-tered in the same unresolved group IV (Fig. 2) and clade IV(Fig. 4). Within the group and clade, all Taeckholmia speciesare highly unresolved, but members of subg. Dendrosonchus,with the exception of S. gummifer, form a moderately sup-ported clade, indicating closer relationships among them thanto other species of Taeckholmia. In addition to the species inclade IV, a newly sequenced S. wildpretii and T. arborea spe-cies were placed in clade II containing the two Dendrosonchusspecies, S. pinnatifidus and S. brachylobus. Sonchus wildpretii,one of the critically endangered species in the island of LaGomera, does not belong to either Dendrosonchus or Taeck-holmia (but is phenotypically more similar to Taeckholmiathan to Dendrosonchus; Reifenberger and Reifenberger, 1992,1996). Taeckholmia arborea is similar to other typical speciesof Taeckholmia, but with somewhat broader leaf lobes (up to5 mm) and slightly more florets per head (15–20 florets). Un-like the species in clade IV, the species in clade II occur ondifferent islands, Lanzarote, La Gomera, Gran Canaria, and LaPalma. These two cases indicate the potential effects of con-temporary and/or more ancient hybridization in the evolutionof the woody Sonchus alliance.

Despite the long cpDNA sequences we sampled (.4000bp), the cpDNA phylogeny (Fig. 3) is highly unresolved andweakly supported, especially the clade containing the coremembers of the alliance. Therefore, we could not rigorouslyassess the role of hybridization and introgression. However,given low resolution and weak support, we obtained some in-sights into the potential role of hybridization and introgression.First, the four species in the group II in the ITS ML tree (Fig.2) are sister to the remaining of the core alliance members,sharing similar haplotypes (Fig. 3). This molecular congruencebetween ITS and cpDNA markers suggests that Taeckholmia-like phenotypes are due to convergent evolution. However, wecannot completely rule out the possibility of initial hybridiza-tion between Dendrosonchus and Taeckholmia, with Dendro-sonchus acting as the maternal parent in repeated backcrossingtoward Taeckholmia, and the preferential fixation of Dendro-sonchus type ITS sequences and/or selection against Taeck-holmia ITS carrying individuals. Another possible explanationis lineage sorting of ancestral polymorphic ITS sequences andcpDNA haplotypes for rapidly radiated lineages within the al-liance, which was suggested by isozyme study (Kim et al.,1999b). As for the case of the hybridization in group IV (Fig.2), the cpDNA ML tree was highly unresolved, and we werenot able to test the potential hybridization and introgression


hypotheses. More variable cpDNA markers for the core mem-bers of the alliance are needed to further test this hypothesis.

Biogeography—The molecular phylogenetic study of thewoody Sonchus alliance suggests that the Canary Islands arethe ancestral archipelago (Fig. 5). All but one taxon of thebasal lineages occur in mountains or mountain cliffs of thegeologically old areas of Gran Canaria, the Canaries, suggest-ing that the common ancestor of the entire alliance first arrivedin Gran Canaria and subsequently dispersed to other islandsof the Canaries. This study also indicates that the MadeiranDendrosonchus species are derived from a common ancestorin Tenerife, while the Cape Verde island endemic S. daltoniievolved from El Hierro. Colonization patterns from the Ca-naries to Madeira and Cape Verde are similar to the majorityof Macaronesian endemics (e.g., Echium, Bohle et al., 1996;Pericallis, Panero et al., 1999; Sideritis, Barber et al., 2000;subg. Crambe, Francisco-Ortega et al., 2002; Teline monspes-sulana group, Percy and Cronk, 2002; Convolvulus, Carine etal., 2004). The Madeira archipelago appears to be ancestral ina few other cases (e.g., Argyranthemum, Francisco-Ortega etal., 1996, 1997; Tolpis, Moore et al., 2002). Thus, the dispersalroute from the Canaries to Madeira and Cape Verde (and alsoto Azores) appears to be a typical pattern for majority of flow-ering plants in the Macaronesian Islands.

The detailed geographic sampling of this study allowed usto determine the role of interisland colonization events andintra-island differentiation during the evolution of the woodySonchus alliance. The best example of intra-island differenti-ation is on the island of Madeira (Fig. 4; clade IIIA). The basalposition of coastal taxa, S. ustulatus subsp. ustulatus andsubsp. maderensis, suggests that radiation on this island ini-tiated in this ecological zone. Likewise, intra-island speciationmay have been important in the basal lineages (Gran Canaria,except in Lactucosonchus from La Palma) and clade IV (allmembers of this unresolved clade are from La Gomera andTenerife). For the remaining clades, interisland colonizationrather than intra-island differentiation probably has been theprimary manner by which Sonchus has exploited the manyecological niches in the Macaronesian Islands. The importantrole of interisland colonization has been observed in other Ma-caronesian endemics (e.g., Crambe sect. Dendrocrambe, Fran-cisco-Ortega et al., 2002; Aeonium, Mes 1995), whereas thepredominant role of intra-island speciation was documented inother taxa (e.g., Chamaecytisus proliferus complex, Francisco-Ortega et al., 1992; the silversword alliance, Baldwin et al.,1990, Baldwin and Robichaux, 1995; Hawaiian Lobeliads,Givnish et al., 1995). Thus, both interisland dispersal eventsto similar ecological habitats and radiation within each islandhave played important roles in the evolution of the alliance.

Closing remarks—In this study, the combined analysis ofnuclear and chloroplast DNA sequences provided improvedresolution and identification of the four basal lineages withinthe Sonchus alliance. Subgenus Dendrosonchus and genusTaeckholmia are not suggested to be monophyletic and S. tub-erifer, the only member of subg. Sonchus within the alliance,is very closely related to core members. The radiation of thealliance from Gran Canaria is given strong support, as are twodistinct instances of long-distance dispersal giving rise sepa-rately to the origin of Dendrosonchus species in Maderia andthe Cape Verde islands.

Poor resolution in the cpDNA phylogeny for the core mem-

bers of the alliance unfortunately limited the assessment of therole of hybridization and introgression in adaptive radiation.The previous ITS phylogeny of the subtribe Sonchinae (Kimet al., 1996a, 1999a) and other preliminary data (S.-C. Kim,unpublished data) suggest that genus Sonchus is highly poly-phyletic and new circumscription and reclassification of sub-tribe Sonchinae are needed. It seems also inevitable that sev-eral segregate monotypic genera of Sonchinae in the Pacificand Atlantic islands (i.e., Embergeria, Kirkianella, Actites,Babcockia, Sventenia, Lactucosonchus, Chrysoprenanthes)and two genera of Dendroseridineae (i.e., Dendroseris andThamnoseris) evolved within the genus Sonchus. These lin-gering uncertainties point to the need for a comprehensiveclassification of this plant group based on both phylogeneticreconstruction and a reevaluation of traditional taxonomiccharacters.

Understanding plant evolution and speciation in oceanic is-lands has been hindered by a lack of molecular variation, theconvergent evolution of similar morphological traits in islandsettings, and an insufficient analysis of the role of hybridiza-tion and introgression. The phylogenetic relationships amongspecies of the woody Sonchus alliance in the MacaronesianIslands, a premier example of adaptive radiation in the AtlanticOcean, have thus far been difficult to determine accurately.Future research needs to be focused on finding several alter-native nuclear intron regions to fully resolve phylogenetic re-lationships within the alliance and also to determine the roleof hybridization and introgression in adaptive radiations.

LITERATURE CITED

ALDRIDGE, A. 1975. Taxonomic and anatomical studies in Sonchus L. sub-genus Dendrosonchus Webb ex Schultz Bip. and related genera. Ph.D.dissertation, University of Reading, Reading, UK.

ALDRIDGE, A. 1976. Macaronesian Sonchus subgenus Dendrosonchus s.l.(Compositae-Lactuceae), including a reappraisal of the species conceptand new combinations. Botanica Macaronesica 2: 81–93.

ALDRIDGE, A. 1977. Anatomy and evolution in the Macaronesian Sonchussubgenus Dendrosonchus (Compositae-Lactuceae): nodal and petiolarvascular patterns. Botanica Macaronesica 3: 41–59.

ALDRIDGE, A. 1978. Anatomy and evolution in the Macaronesian Sonchussubgenus Dendrosonchus (Compositae-Lactuceae). Botanical Journal ofLinnean Society 76: 249–285.

ALDRIDGE, A. 1979. Evolution within a single genus: Sonchus in Macaro-nesia. In D. Bramwell [ed.], Plants and islands, 270–290. AcademicPress, New York, New York, USA.

ALVAREZ, I., AND J. F. WENDEL. 2003. Ribosomal ITS sequences and plantphylogenetic inference. Molecular Phylogenetics and Evolution 29: 417–434.

ARDEVOL GONZALEZ, J. F., L. BORGEN, AND P. L. PEREZ DE PAZ. 1993.Checklist of chromosome numbers in Canarian vascular plants. Som-merfeltia 18: 1–59.

BALDWIN, B. G., D. W. KYHOS, AND J. DVORAK. 1990. Chloroplast DNAevolution and adaptive radiation in the Hawaiian silversword alliance(Asteraceae-Madiinae). Annals of the Missouri Botanical Garden 77: 96–109.

BALDWIN, B. G., AND R. H. ROBICHAUX. 1995. Historical biogeography andecology of the Hawaiian silversword alliance (Asteraceae): new molec-ular phylogenetic perspectives. In W. L. Wagner and V. A. Funk [eds.],Hawaiian biogeography: evolution on a hot spot archipelago, 259–287.Smithsonian Institution Press, Washington, D.C., USA.

BALDWIN, B. G., D. J. CRAWFORD, J. FRANCISCO-ORTEGA, S.-C. KIM, T.SANG, AND T. F. STUESSY. 1998. Molecular phylogenetic insights on theorigin and evolution of oceanic island plants. In D. E. Soltis, P. S. Soltis,and J. J. Doyle [eds.], Molecular systematics of plants II: DNA sequenc-ing, 410–441. Kluwer, Boston, Massachusetts, USA.

BARBER, J. C., J. FRANCISCO-ORTEGA, A. SANTOS-GUERRA, A. MARRERO,AND R. K. JANSEN. 2000. Evolution of endemic Sideritis (Lamiaceae)


in Macaronesia: insights from a chloroplast DNA restriction site analysis.Systematic Botany 25: 633–647.

BOHLE, U. R., H. H. HILGER, AND W. F. MARTIN. 1996. Island colonizationand evolution of the insular woody habit in Echium L. (Boraginaceae).Proceedings of the National Academy of Sciences, USA 93: 11740–11745.

BOULOS, L. 1967. Taeckholmia, a new genus of Compositae from CanaryIslands. Botaniska Notiser 120: 95–108.

BOULOS, L. 1972. Revision systematique du genre Sonchus L. s.l. I. Intro-duction et classification. Botaniska Notiser 125: 287–305.

BOULOS, L. 1974a. Revision systematique du genre Sonchus L. s.l. V. Sous-genre 2. Dendrosonchus. Botaniska Notiser 127: 7–37.

BOULOS, L. 1974b. Revision systematique du genre Sonchus L. s.l. VI. Sous-genre 3. Origosonchus. Genres Embergeria, Babcockia et Taeckholmia.Botaniska Notiser 127: 402–451.

BRAMWELL, D. 2003. The correct generic names for Sonchus webbii Sch.Bip. and Prenanthes pendula Sch. Bip. Botanica Macaronesica 24: 179–182.

BRAMWELL, D., AND Z. BRAMWELL. 2001. Wild flowers of the Canary Is-lands, 2nd ed. Editorial Rueda, Madrid, Spain.

CARINE, M. A., S. J. RUSSEL, A. SANTOS-GUERRA, AND J. FRANCISCO-OR-TEGA. 2004. Relationships of the Macaronesian and Mediterranean flo-ras: molecular evidence for multiple colonizations into Macaronesia andback-colonization of the continent in Convolvulus (Convolvulaceae).American Journal of Botany 91: 1070–1085.

CARLQUIST, S. 1974. Island biology. Columbia University Press, New York,New York, USA.

CRAWFORD, D. J., E. RUIZ, T. F. STUESSY, E. TEPE, P. AQEVEQUE, F. GON-ZALEZ, R. J. JENSEN, G. J. ANDERSON, G. BERNARDELL, C. M. BAEZA,U. SWENSON, AND M. SILVA O. 2001. Allozyme diversity in endemicflowering plant species of the Juan Fernandez Archipelago, Chile: eco-logical and historical factors with implications for conservation. Ameri-can Journal of Botany 88: 2195–2203.

DE JOODE, D. E., AND J. F. WENDEL. 1992. Genetic diversity and origin ofthe Hawaiian Islands cotton, Gossypium tomentosum. American Journalof Botany 79: 1311–1319.

DOYLE, J. J., AND J. L. DOYLE. 1987. A rapid DNA isolation procedure forsmall quantities of fresh leaf material. Phytochemical Bulletin 19: 11–15.

FARRIS, J. S., M. KALLERSJO, A. G. KLUGE, AND C. BULT. 1995. Construct-ing a significance test for incongruence. Systematic Biology 44: 570–572.

FELSENSTEIN, J. 1981. Evolutionary trees from DNA sequences: a maximumlikelihood approach. Journal of Molecular Evolution 17: 368–376.

FELSENSTEIN, J. 1985. Confidence limits on phylogenies: an approach usingthe bootstrap. Evolution 39: 783–791.

FRANCISCO-ORTEGA, J., M. T. JACKSON, J. P. CATTY, AND B. V. FORD-LLOYD.1992. Genetic diversity in the Chamaecytisus proliferus complex (Fa-baceae: Genisteae) in the Canary Islands in relation to in situ conser-vation. Genetic Resources and Crop Evolution 39: 149–158.

FRANCISCO-ORTEGA, J., R. K. JANSEN, AND A. SANTOS-GUERRA. 1996. Chlo-roplast DNA evidence of colonization, adaptive radiation, and hybrid-ization in the evolution of the Macaronesian flora. Proceedings of theNational Academy of Sciences, USA 93: 4085–4090.

FRANCISCO-ORTEGA, J., D. J. CRAWFORD, A. SANTOS-GUERRA, AND R. K.JANSEN. 1997. Origin and evolution of Argyranthemum (Asteraceae: An-themideae) in Macaronesia. In T. J. Givnish and K. J. Sytsma [eds.],407–431, Molecular evolution and adaptive radiation. Cambridge Uni-versity Press, Cambridge, UK.

FRANCISCO-ORTEGA, J., J. FUERTES-AGUILAR, S.-C. KIM, A. SANTOS-GUER-RA, D. J. CRAWFORD, AND R. K. JANSEN. 2002. Phylogeny of the Ma-caronesian endemic Crambe section Dendrocrambe (Brassicaceae) basedon internal transcribed spacer sequences of nuclear ribosomal DNA.American Journal of Botany 89: 1984–1990.

GIVNISH, T. J., K. J. SYTSMA, J. F. SMITH, AND W. J. HAHN. 1995. Molecularevolution, adaptive radiation, and geographic speciation in Cyanea (Cam-panulaceae), the largest plant genus endemic to Hawaii. In W. L. Wagnerand V. A. Funk [eds.], Hawaiian biogeography: evolution on a hot spotarchipelago, 288–337. Smithsonian Institution Press, Washington, D.C.,USA.

GOLDMAN, N. 1993. Statistical tests of models of DNA substitution. Journalof Molecular Evolution 36: 182–198.

GOMES, I., S. GOMES, N. KILIAN, T. LEYENS, W. LOBIN, AND M. T. VERA-

CRUZ. 1995. Notes on the flora of the Cape Verde Islands, W Africa.Willdenowia 25: 177–196.

GOMES, I., T. LEYENS, B. DA LUZ, J. COSTA, AND F. GONCALVES. 1999. Newdata on the distribution and conservation status of some angiosperms ofthe Cape Verde Islands, W Africa. Willdenowia 26: 105–114.

HANSEN, A., AND P. SUNDING. 1993. Flora of Macaronesia. Checklist of vas-cular plants, 4th revised ed. Sommerfeltia 17: 1–296.

HOWARTH, D. G., AND D. A. BAUM. 2005. Genealogical evidence of hom-oploid hybrid speciation in an adaptive radiation of Scaevola (Gooden-iaceae) in the Hawaiian Islands. Evolution 59: 948–961.

JOHNSON, L. A., AND D. E. SOLTIS. 1995. Phylogenetic inference in Saxifra-gaceae sensu stricto and Gilia (Polemoniaceae) using matK sequences.Annals of the Missouri Botanical Garden 82: 149–175.

KIM, S.-C., D. J. CRAWFORD, AND R. K. JANSEN. 1996a. Phylogenetic rela-tionships among the genera of the subtribe Sonchinae (Asteraceae): ev-idence from ITS sequences. Systematic Botany 21: 417–432.

KIM, S.-C., D. J. CRAWFORD, J. FRANCISCO-ORTEGA, AND A. SANTOS-GUER-RA. 1996b. A common origin for woody Sonchus and five related generain the Macaronesian Islands: molecular evidence for extensive radiation.Proceedings of the National Academy of Sciences, USA 93: 7743–7748.

KIM, S.-C., D. J. CRAWFORD, R. K. JANSEN, AND A. SANTOS-GUERRA. 1999a.The use of a non-coding region of chloroplast DNA in phylogeneticstudies of the subtribe Sonchinae (Asteraceae: Lactuceae). Plant System-atics and Evoltion 215: 85–99.

KIM, S.-C., D. J. CRAWFORD, J. FRANCISCO-OOTEGA, AND A. SANTOS-GUER-RA. 1999b. Adaptive radiation and genetic differentiation in the woodySonchus alliance (Asteraceae: Sonchinae) in the Canary Islands. PlantSystematics and Evolution 215: 101–118.

KIM, S.-C., C. T. LU, AND B. J. LEPSCHI. 2004. Phylogenetic positions ofActites megalocarpa and Sonchus hydrophilus (Sonchinae: Asteraceae)based on ITS and chloroplast non-coding DNA sequences. AustralianSystematic Botany 17: 73–81.

KIMURA, M. 1980. Simple method for estimating evolutionary rate of basesubstitution through comparative studies of nucleotide sequences. Jour-nal of Molecular Evolution 16: 111–120.

MADDISON, W. P., AND D. R. MADDISON. 2000. MacClade, version 4. Sinauer,Sunderland, Massachusetts, USA.

MES, T. H. M. 1995. Origin and evolution of the Macaronesian Sempervi-voideae (Crassulaceae). Ph.D. dissertation, University of Utrecht,Utrecht, Netherlands.

MOORE, M. J., J. FRANCISCO-ORTEGA, A. SANTOS-GUERRA, AND R. K. JAN-SEN. 2002. Chloroplast DNA evidence for the roles of island coloniza-tion and extinction in Tolpis (Asteraceae: Lactuceae). American Journalof Botany 89: 518–526.

PANERO, J. L., J. FRANCISCO-ORTEGA, R. K. JANSEN, AND A. SANTOS-GUER-RA. 1999. Molecular evidence for multiple origins of woodiness and aNew World biogeographic connection of the Macaronesian island endem-ic Pericallis (Asteraceae: Senecioneae). Proceedings of the NationalAcademy of Sciences, USA 96: 13886–13891.

PERCY, D. M., AND Q. C. B. CRONK. 2002. Different fates of island brooms:contrasting evolution in Adenocarpus, Genista, and Teline (Genisteae,Fabaceae) in the Canary Islands and Madeira. American Journal of Bot-any 89: 854–864.

PEREZ DE PAZ, P. L. 1976. Observaciones sobre la biologıa y relaciones deSventenia bupleuroidesF. Q. Botanica Macaronesica 1: 51–65.

POSADA, D., AND K. CRANDALL. 1998. Modeltest: testing the model of DNAsubstitution. Bioinformatics 14: 817–818.

REIFENBERGER, U., AND A. REIFENBERGER. 1992. Sonchus wildpretii (Com-positae), ein neuer endemit der insel La Gomera (Kanarische Inseln).Willdenowia 22: 49–53.

REIFENBERGER, U., AND A. REIFENBERGER. 1996. Vereinigung der gattungTaeckholmia Boulos mit der gattung Sonchus L. (Compositae) und emen-dation des subgenus DendrosonchusSch. Bip. Feddes Repertorium 107:311–319.

ROGNON, P., AND G. COUDE-GAUSSEN. 1996. Paleoclimates off NorthwestAfrica (288–358N) about 18,000 yr B.P. based on continental eolian de-posits. Quaternary Research 46: 118–118.

SAITO, N., AND M. NEI. 1987. The neighbor-joining method: a new methodfor reconstructing phylogenetic trees. Molecular Biology and Evolution4: 406–425.

SENNIKOV, A. N., AND I. D. ILLARIONOVA. 1999. Reclassification of Pren-anthes pendula (Asteraceae: Lactuceae). Composite Newsletter 34: 53–61.


SWOFFORD, D. L. 2001. PAUP*: phylogenetic analysis using parsimony(*and other methods), version 4.0. Sinauer, Sunderland, Massachu-setts, USA.

TABERLET, P., L. GIELLY, G. PAUTOU, AND J. BOUVET. 1991. Universal prim-ers for amplification of three non-coding regions of chloroplast DNA.Plant Molecular Biology 17: 1105–1109.

WHELAN, S., AND N. GOLDMAN. 1999. Distributions of statistics used for the

comparison of models of sequence evolution in phylogenetics. MolecularBiology and Evolution 16: 1292–1299.

WHITE, T. J., T. BRUNS, S. LEE, AND J. TAYLOR. 1990. Amplification anddirect sequencing of fungal ribosomal RNA genes for phylogenetics. InM. Innis, D. Gelfand, J. Sninsky, and T. White [eds.], PCR protocol: aguide to methods and amplification, 315–322. Academic Press, San Di-ego, California, USA.

APPENDIX. Collection data and GenBank accession numbers for the woody Sonchus alliance ITS, trnT-L-F, and matK sequences. Abbreviations of islands areas follows: LA, Lanzarote; FU, Fuerteventura; GC, Gran Canaria; TE, Tenerife; GO, La Gomera; PA, La Palma; and HI El, Hierro. Accession numberswith asterisk were reported in a previous study (Kim et al., 1996b). We followed the classification of Boulos (1972), with the exception of S. wildpretii,which was described as a new species in 1992 based on a Sonchus s.l. view by the authors. Voucher specimens are deposited in the herbarium of Jardınde Aclimatacion de La Orotava, Tenerife. Voucher specimen abbreviations: KSC 5 Kim, Seung-Chul; SGA 5 Santos-Guerra, Arnoldo. Herbariumabbreviation: OS, Ohio State University Herbarium. Species that were not sequenced for a particular DNA region are indicated by ‘‘—.’’

Taxon; Voucher; Locality; Accession numbers: ITS1, ITS2; trnT-L-F; matK.

Outgroup speciesSonchus palustris L.; —; DNA provided by R. Jansen; L48311*, L48312*;

DQ068479–81; DQ022984.

Ingroup speciesBabcockia Boulos—B. platylepis (Webb) Boulos; KSC et al. 1028 (OS); be-

tween Tasarto and Tasartico (GC1); L48137*, L48318*; DQ068500–2;DQ022964. B. platylepis (Webb) Boulos; SGA & KSC 110; Roque deNublo view place (GC2); DQ072470–1; DQ0685003–5; DQ022965. B.platylepis (Webb) Boulos; SGA & KSC 111; between Tejeda and Artenaria(GC3); DQ072472–3; DQ068506–8; DQ022966. B. platylepis (Webb)Boulos; SGA & KSC 112; Cortijo, Crespos (GC4); —; DQ068509–11;DQ022992.

Chrysoprenanthes (Sch. Bip.) Bramwell—C. pendula (Sch. Bip.) Bramwell;KSC 1051 (OS); Fataga north (GC1); L48157*, L48158*; DQ068512–4;DQ022993. C. pendula (Sch. Bip.) Bramwell; SGA & KSC 115; ParralilloDam (GC2); DQ072464–5; DQ068515–7; DQ022994. C. pendula (Sch.Bip.) Bramwell; SGA & KSC 116; Fataga south (GC3); L48155*,L48156*; DQ068518–20; DQ022995. C. pendula (Sch. Bip.) Bramwell;SGA & KSC 117; Barranco de Fataga (GC4); DQ072466–7; DQ068521–3; DQ022996.

Lactucosonchus (Sch. Bip.) Svent.—L. webbii (Sch. Bip.) Svent.; KSC et al.1033 (OS); San Antonio (PA1); —; DQ068482–4; DQ022986. L. webbii(Sch. Bip.) Svent.; SGA & KSC 100; San Antonio (PA2); DQ072460–1;DQ068485–7; DQ022987. L. webbii (Sch. Bip.) Svent.; —; DNA providedby Crawford(PA3); L48161*, L48162*; DQ068488–90; DQ022988.

Sonchus L.—Subg. Dendrosonchus Sch. Bip. ex Boulos—S. acaulis Dum.-Cours.; KSC et al. 1027 (OS); Roque Bentaija (GC); L48287*, L48288*;DQ068590–2; DQ023029. S. acaulis Dum.-Cours.; SGA & KSC 152; Lad-era de Guimar (TE); L48289*, L48290*; DQ068593–5; DQ023028. S.bornmuelleri Pitard; SGA & KSC 136; La Fajana, Barlovento (PA1);DQ072506–7; DQ068563–5; DQ023021. S. bornmuelleri Pitard; SGA &KSC 138; near Airport, toward Santa Cruz (PA2); DQ072508–9;DQ068566–8; DQ023022. S. brachylobus Webb & Berth.; SGA & KSC228; Cuesta de Silva (GC1); L48133*, L48134*; DQ068653–5;DQ022977. S. brachylobus Webb & Berth.; SGA & KSC 229; AndenVerde (GC2); DQ072510–1; DQ068656–8; DQ022976. S. canariensis(Sch. Bip.) Boulos; KSC et al. 1021 (OS); Anden Verde (GC); L48293*,L48294*; DQ068584–6; —. S. canariensis (Sch. Bip.) Boulos; SGA &KSC 149; Guia de Isora toward Teide (TE); L48291*, L48292*;DQ068587–9; DQ022997. S. daltonii Webb; —; DNA provided by D.Crawford (Cape Verde); L48115*, L48116*; DQ068545–7; DQ023015. S.congestus Willd.; KSC et al. 1000 (OS); El Bailadera, Anaga (TE);L48173*, L48174*; DQ068596–8; DQ023030. S. congestus Willd.; SGA& KSC 154; Ladera Barranco de la Virgen Sobre Molino Chico (GC);L48175*, L48176*; —; —. S. fauces-orci Knoche; SGA & KSC 161;Masca, Teno (TE1); L48119*, L48120*; DQ068626–8; DQ023007. S. fau-ces-orci Knoche; SGA & KSC 162; Masca, Teno (TE2); DQ072474–5;DQ068629–31; DQ022974. S. fauces-orci Knoche; SGA & KSC 163;Masca, Teno (TE3); DQ072476–7; DQ068632–4; —. S. fruticosus L. fil.;—; DNA provided by D. Crawford (Madeira1); L48125*, L48126*;DQ068608–10; DQ023001. S. fruticosus L. fil.; KSC 2003; Encumeada(Madeira2); DQ072482–3; DQ068611–3; DQ023002. S. fruticosus L. fil.;

KSC 2004; Encumeada (Madeira3); DQ072484–5; DQ068614–6;DQ023003. S. gandogeri Pitard; SGA & KSC 141; between Guinea andLas Puntas (HI1); L48121*, L48122*; DQ068569–71; DQ023023. S. gan-dogeri Pitard; SGA & KSC 142; between Guinea and Las Puntas (HI2);DQ072512–3; DQ068572–4; DQ023024. S. gandogeri Pitard; SGA &KSC 143; between Guinea and Las Puntas (HI3); DQ072514–5;DQ068575–7; DQ023025. S. gonzalezpadronii Svent.; KSC et al. 1037(OS); San Sebastian (GO1); L48127*, L48128*; DQ068635–7; DQ023008.S. gonzalezpadronii Svent.; SGA & KSC 166; Peraca (GO2); DQ072492–3; DQ068638–40; DQ023031. S. gummifer Link.; KSC et al. 1022 (OS);Orticosa (TE); DQ072502–3; —; —. S. hierrensis (Pitard) Boulos; —;DNA provided by D. Crawford (GO); L48167*, L48168*; DQ068548–50;DQ023016. S. hierrensis (Pitard) Boulos; —; DNA provided by D. Craw-ford (PA1); L48169*, L48170*; DQ068551–3; DQ023017. S. hierrensis(Pitard) Boulos; SGA & KSC 132; Barranco de Franceses (PA2); —;DQ068554–6; DQ023019. S. hierrensis (Pitard) Boulos; SGA & KSC 133;Tabano (HI1); —; DQ068557–9; DQ023018. S. hierrensis (Pitard) Boulos;SGA & KSC 134; San Andres (HI2); DQ072504–5; DQ068560–2;DQ023020. S. ortunoi Svent.; —; DNA provided by D. Crawford (GO1);L48129*, L48130*; DQ068641–3; DQ023032. S. ortunoi Svent.; KSC etal. 1036 (OS); between Roque de Ojila and Zarzita (GO2); DQ072490–1;DQ068644–6; DQ023037. S. palmensis (Sch. Bip.) Boulos; SGA & KSC145; Los Tilos (PA1); L48123*, L48124*; DQ068578–80; DQ023026. S.palmensis (Sch. Bip.) Boulos; SGA & KSC 148; Mazo (PA2); —;DQ068581–3; DQ023027. S. pinnatifidus Cav.; SGA & KSC 125; Famara(LA1); L48131*, L48132*; DQ068536–8; DQ022971. S. pinnatifidusCav.; SGA & KSC 126; Mirado Riscos de Famara (LA2); DQ072480–1;DQ068539–41; DQ022972. S. pinnatifidus Cav.; SGA & KSC 127; Mir-ado Riscos de Famara (LA3); — DQ068542–4; DQ022973. S. pinnatusAit.; KSC 1996; Levada North after Espigao (Madeira1); L48171*,L48172*; DQ068599–01; DQ022998. S. pinnatus Ait.; KSC 1990; SaoVicente (Madeira2); DQ072478–9; DQ068602–4; DQ022999. S. pinnatusAit.; KSC 1991; Sao Vicente (Madiera3); — DQ068605–7; DQ023000. S.radicatus Aiton; SGA & KSC 169; Punta de Hidalgo (TE1); DQ072494–5; DQ068647–9; DQ02300. S. radicatus Aiton; SGA & KSC 170; Teno(TE2); DQ072496–7; —; —. S. tectifolius Svent.; SGA & KSC 175; Bar-ranco de Crrspin (TE1); DQ072500–1; DQ068650–2; DQ023010. S. tec-tifolius Svent.; SGA & KSC 176; Anaga, between San Andres and ElBailadero (TE2); DQ072498–9; —; —. S. ustulatus subsp. ustulatusLowe;—; DNA provided by D. Crawford (Madiera1); L48117*, L48118*;DQ068617–9; DQ023004. S. ustulatus subsp. ustulatus Lowe; KSC 2010;Gaula (Madeira2); DQ072488–9; DQ068620–2; DQ023006. S. ustulatussubsp. maderensis Aldr.; KSC 2011; between Sexial and Ribeira da Janela(Madeira); DQ072486–7; DQ068623–5; DQ023005. S. wildpretii U. & A.Reifenberger; SGA & KSC 180; Roque de Ojila (GO1); DQ072516–7;DQ068659–61; DQ022978. S. wildpretii U. & A. Reifenberger; SGA &KSC 181; Roque de Agando (GO2); DQ072518–9; DQ068662–4;DQ022979.

Sonchus L.—Subg. Sonchus L.—S. tuberifer Svent.; KSC et al. 1045 (OS);Masca, Teno (TE1); L48313*, L48314*; DQ068524–6; DQ022967. S. tub-erifer Svent.; SGA & KSC 120; Masca, Teno (TE2); DQ072468–9;DQ068527–9; DQ022969. S. tuberifer Svent.; SGA & KSC 121; Masca,


Teno (TE3); —; DQ068530–2; DQ022970. S. tuberifer Svent.; SGA &KSC 122; Masca, Teno (TE4); —; DQ068533–5; DQ022968.

Sventenia Font Quer—S. bupleuroides Font Quer; KSC et al. 1041 (OS);cultivated in Jardin Canario (GC1); L48315*, L48316*; DQ068491–3;DQ022989. S. bupleuroides Font Quer; KSC 1067; cultivated in JardinCanario (GC2); DQ072462–3; DQ068494–6; DQ022990. S. bupleuroidesFont Quer; KSC 1068; cultivated in Jardin Canario (GC3); —; DQ068497–9; DQ022991.

Taeckholmia Boulos—T. arborea (DC.) Boulos; KSC et al. 1047 (OS); Mirca(PA1); L48327*, L48328*; DQ068665–7; DQ022983. T. arborea (DC.)Boulos; SGA & KSC 220; Mirca (PA2); DQ072524–5; DQ068668–70;DQ022980. T. arborea (DC.) Boulos; SGA & KSC 225; Punta de Teno(TE1); L48325*, L48326*; DQ068671–3; DQ022981. T. arborea (DC.)Boulos; SGA & KSC 226; Punta de Teno (TE2); DQ072522–3;DQ068674–6; DQ022982. T. canariensis Boulos; KSC et al. 1043 (OS);El Camello, San Sebastian (GO1); L48323*, L48324*; DQ068686–8;

DQ023033. T. canariensis Boulos; SGA & KSC 207; between Valle GranRey and Arune (GO2); DQ072526–7; —; —. T. capillaris (Svent.) Boulos;SGA & KSC 209; Masca, Teno (TE1); L48329*, L48330*; DQ068689–91; DQ023013. T. capillaris (Svent.) Boulos; SGA & KSC 210; Masca,Teno (TE2); DQ072528–9; —; —. T. heterophylla Boulos; KSC et al. 1037(OS); between Agulo and Los Rosas (GO1); L48333*, L48334*;DQ068692–4; DQ023034. T. heterophylla Boulos; SGA & KSC 215;Above Hermigua (GO2); L48335*, L48336*; DQ068695–7; DQ023035. T.heterophylla Boulos; SGA & KSC 216; Agulo (GO3); DQ072530–1; —.;—. T. microcarpa Boulos; SGA & KSC 205; Ladeira de Guimar (TE1);L48321*, L48322*; DQ068683–5; DQ023012. T. microcarpa Boulos;SGA & KSC 206; Malpais de Guimar (TE2); DQ072520–1; —; —. T.pinnata (L. f.) Boulos; SGA & KSC 201; Icod de Los Vinos (TE);L48319*, L48320*; DQ068677–9; DQ023011. T. pinnata (L. f.) Boulos;SGA & KSC 202; Cuesta de Silva (GC); L48317*, L48318*; DQ068680–2; DQ023036.

CHLOROPLAST DNA PHYLOGENY OF THE WOODY …plantbiology.ucr.edu/faculty/Leeetal(AJB2005).pdf · 2072...

Documents

Transcript of CHLOROPLAST DNA PHYLOGENY OF THE WOODY …plantbiology.ucr.edu/faculty/Leeetal(AJB2005).pdf · 2072...