Proc. Natl. Acad. Sci. USAVol. 93, pp. 7743-7748, July 1996Evolution

A common origin for woody Sonchus and five related generain the Macaronesian islands: Molecular evidencefor extensive radiation

(adaptive radiation/molecular evolution/biogeography)

SEUNG-CHUL KIM*, DANIEL J. CRAWFORD*t, JAVIER FRANCISCO-ORTEGAt, AND ARNOLDO SANTOS-GUERRA§*Department of Plant Biology, Ohio State University, Columbus, OH 43210-1293; tDepartment of Botany, The University of Texas, Austin, TX 78713-7640;and 3Jardin de Aclimataci6n de La Orotava, 38400 Puerto de La Cruz, Tenerife, Canary Islands, Spain

Communicated by Michael T. Clegg, University of California, Riverside, CA, March 15, 1996 (received for review January 4, 1996)

ABSTRACT Woody Sonchus and five related genera (Bab-cockia, Taeckholmia, Sventenia, Lactucosonchus, and Prenan-thes) of the Macaronesian islands have been regarded as anoutstanding example of adaptive radiation in angiosperms.Internal transcribed spacer region of the nuclear rDNA (ITS)sequences were used to demonstrate that, despite the extensivemorphological and ecological diversity ofthe plants, the entirealliance in insular Macaronesia has a common origin. Thesequence data place Lactucosonchus as sister group to theremainder of the alliance and also indicate that four relatedgenera are in turn sister groups to subg. Dendrosonchus andTaeckholmia. This implies that the woody members ofSonchuswere derived from an ancestor similar to allied genera nowpresent on the Canary Islands. It is also evident that thealliance probably occurred in the Canary Islands during thelate Miocene or early Pliocene. A rapid radiation of majorlineages in the alliance is consistent with an unresolvedpolytomy near the base and low ITS sequence divergence.Increase of woodiness is concordant with other insular en-demics and refutes the relictural nature of woody Sonchus inthe Macaronesian islands.

Ascertaining the origin and evolution of plants endemic tooceanic islands is both fascinating and frustrating. Endemicsmay assume the typical insular woody habit and become verydistinct in other suites of characters such that determiningtheir continental relatives is difficult if not impossible withmorphology (1-4). Once a colonizer becomes established onan oceanic island, extensive diversification may occur as plantsmove into a variety of open habitats. This process has beenviewed as a good example of adaptive radiation (2-9), and onceit occurs, the different descendent lines may exhibit a widearray of characters, making it difficult to determine whetherthey evolved from a common ancestor. Thus, difficult prob-lems posed by insular endemics include whether they resultfrom a single introduction, estimating the time of radiation,and identifying the continental relative(s) and source area(s)of the original colonizer(s) (10-12). As noted above, compar-ative morphology may be of limited value for addressing thesequestions because of the difficulty in distinguishing featuresshared by common ancestry as opposed to parallelisms. Mo-lecular sequences, however, have proven useful in the study ofsome insular groups because, contrasted with morphology,base substitutions of determined regions (mutations) may beneutral or nearly so (11, 13-16).One of the basic assumptions in the study of the origin and

evolution of endemics in oceanic islands is that the remotenessof the islands from possible source areas (combined with theirsmall land masses) acts like a sieve and the arrival of prop-

agules to the islands is a very rare event. This in turn makes itmore likely that related endemics on remote archipelagos havea common origin, i.e., represent a monophyletic group. Unlikemore remote volcanic islands in the Pacific (Juan Fernandez,Hawaiian, and Galapagos Islands) where most evolutionarystudies of insular endemic plants have been carried out, theMacaronesian islands (Fig. 1) are very close to possible sourceareas and exhibit a broad range of geological ages (17-22). (Inthis paper, "Macaronesian" will refer to the islands of this area,and the Canary Islands will be specified when referring to thatarchipelago alone.) The proximity of the islands to the Africancontinent and their different geological ages make it muchmore likely that multiple colonization events could haveoccurred for some closely related taxonomic groups and thatsome taxa could be much older than others. This means, ofcourse, that closely related endemic taxa may not be mono-phyletic. There also has been a long controversy over whethersome of the woody Macaronesian endemics are relict elementsof a flora that extended along the Mediterranean basin duringthe Tertiary period or are recent derivatives from continentalancestors (3, 6, 17, 23-26).There are 34 taxa of ca. 130 taxa in the subtribe Sonchinae

(Asteraceae) endemic to Macaronesia (27), with most of themin the Canary Islands. A previous phylogenetic analysis of theSonchinae (ref. 28; Fig. 2) recognized a Macaronesian cladethat includes the woody members of Sonchus and five relatedgenera (hereafter for convenience often referred to as thewoody Sonchus alliance), but this was based on limited taxonsampling from Macaronesia. This alliance is composed pri-marily of 19 species of woody Sonchus (i.e., subg. Dendroson-chus), the genera Babcockia, Taeckholmia, Sventenia, Lactu-cosonchus, one species of Sonchus subg. Sonchus (i.e., S.tuberifer), and one species of Prenanthes. Two taxa, Lactu-cosonchus and S. tuberifer, are the only members of the groupthat do not have a true woody habit; they are herbaceousperennials with tuberous roots. As may be inferred by therecognition of six genera, these taxa display great morpholog-ical, ecological, and anatomical diversity (29, 30). Despite thisdiversity, the fertility of several intergeneric hybrids suggestsgenetic cohesiveness within the alliance (31, 32). The basicquestion is whether this remarkable array of plants results froma single introduction followed by extensive radiation anddiversification or from several introductions from nearby andclosely related source areas over a long period of time. It hasbeen shown, for example, that the three genera of the verydiverse silversword alliance in the Hawaiian Islands came froma single dispersal event (10, 33), but the extreme isolation of

Abbreviations: ITS, internal transcribed spacer region of the nuclearrDNA; Mya, million years ago.Data deposition: The sequences reported in this paper have beendeposited in the GenBank data base (accession nos. L48115-L48176and L48287-L48338).tTo whom reprint requests should be addressed.

7743

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996)

a

bCANARY ISLANDS

PA(1.5)

TE( 11.6)

gS GO(12.5)

HI(>0.7)GC(13.9)

LA(15.5)'

FU(20.7)

0 10OKm

FIG. 1. (a) The dashed line encloses the biogeographical region ofMacaronesia (17), which includes northwestern Africa and five archi-pelagos. (b) The Canary archipelago and the oldest published radio-metric ages from subaerial volcanics of each island in parenentheses(17-22). Abbreviations of islands are as follows: LA, Lanzarote [15.5million years ago (Mya)]; FU, Fuerteventura (20.7 Mya); GC, GranCanaria (13.9 Mya); TE, Tenerife (11.6 Mya); GO, La Gomera (12.5Mya); PA, La Palma (1.5 Mya); HI, Hierro (>0.7 Mya).

Hawaii contrasts with the situation in Macaronesia where asource area is within 100 km of certain islands (Fig. 1).

In this study, sequences from the internal transcribed spacerregion of the nuclear ribosomal DNA (ITS) were used to assessfurther the monophyly of the woody Sonchus alliance in theMacaronesian islands. The sequence data were also used toelucidate phylogenetic relationships within the group and therole of colonization and adaptive radiation in the evolution ofthe alliance and to examine the origin of woody members ofSonchus in Macaronesia.

MATERIALS AND METHODSA phylogenetic analysis of subtribe Sonchinae was conductedpreviously, with limited sampling from the Macaronesian taxa(ref. 28; Fig. 2). After this analysis, we examined the ITSsequences of 35 additional accessions representing 16 speciesof subg. Dendrosonchus, five species of Taeckholmia, and thefour related genera endemic to the Canary Islands. Total DNAwas isolated from leaf tissues by using the CTAB method (34)and purified in CsCl/ethidium bromide gradients. Methods forPCR amplification, purification of PCR products, and se-quencing reactions of the ITS regions are given in detailelsewhere (11, 28). Both coding and noncoding strands wereread and all sequences were manually aligned using MACCLADE

version 3.0 (35). Aligned sequences are available upon requestfrom the first author.

Variable nucleotide sites were analyzed by unweightedWagner parsimony using PAUP version 3.1.1 (36). On the basisof previous analyses of the Sonchinae (28), Sonchus palustriswas used as an outgroup. The heuristic algorithm with stepwiseaddition option was used to fiad the shortest trees. Bootstrap(100 replicates) and decay analyses were performed to assessthe relative support for clades found in the analysis (37).Pairwise sequence divergences were calculated by the Jukesand Cantor one-parameter method using PHYLIP version 3.52c(38). To assess rate homogeneity over different lineages, therelative-rate test (39, 40) was conducted using the genusLactucosonchus as the reference taxon because this lineage issister to the remainder of the alliance.

RESULTS

Size of ITS, Sequence Divergence, and Relative-Rate Test.Lengths of ITS1 and ITS2 in the woody Sonchus alliance fallwithin the size range reported for Asteraceae (41). Percentpairwise sequence divergence between species ranges from 0.0to 4.8, and average divergences within subg. Dendrosonchusand Taeckholmia are 1.6% and 1.0%, respectively. Averagesequence divergence between subg. Dendrosonchus (excludingTaeckholmia) and other genera in the alliance varies from2.5% to 3.4%.

Results of the relative rate tests indicate no significantdifferences for any of the lineages, and thus the molecularclock cannot be rejected.

Phylogenetic Analyses of ITS Sequences. After the previousphylogenetic study of the Sonchinae, which suggests the mono-phyly of the woody Sonchus alliance in Macaronesia (Fig. 2),additional phylogenetic analyses of ITS sequences with a muchbroader sampling within the group were carried out. Theheuristic search option for 35 accessions found 8123 equallymost parsimonious trees, one of which is shown in Fig. 3. Thistree is identical to the 50% majority-rule consensus tree andsuggests that Sventenia, Babcockia, Prenanthes pendula, andSonchus tuberifer are in turn sister groups to subg. Dendroson-chus sensu Aldridge. It also suggests that Lactucosonchus, an

endemic genus from the island of La Palma (Fig. lb), is sisterto the remaining members of the woody Sonchus alliance (Fig. 3).

DISCUSSION

Origin and Evolution of the Woody Sonchus Alliance inMacaronesia. The molecular data strongly confirm and extendconsiderably our previous preliminary work and indicate that,despite the extensive morphological and ecological diversity ofthe plants and the geographical proximity of the Macaronesianislands to a continental source area, the entire alliance wasderived from a single colonization event (Figs. 2 and 3). Bothbootstrap and decay analyses support strongly the monophylyof the group. The presence of intergeneric hybrids (32, 42) andresults of crossing experiments (A. Aldridge, personal com-munication) likewise suggest genetic cohesiveness despite theconsiderable morphological diversity of the alliance. Theseresults are similar to those for other insular groups such as thegenus Argyranthemum in the Macaronesian islands (43) andthe silversword alliance (10, 33) and Tetramolopium (44) inHawaii, where highly fertile interspecific hybrids can be pro-duced despite pronounced morphological differences. Withinthe woody Sonchus group, the ITS tree suggests that Lactu-cosonchus diverged first, followed by the radiation of the fourother genera, as well as subg. Dendrosonchus sensu Aldridge(Fig. 3). It is also likely that, based on the previous study of theSonchinae (ref. 28; Fig. 2), the entire group was derived froma single dispersal event from a more widely distributed Euro-pean taxon, such as S. palustris. The low average sequence

Azores Islands

Europe

Madeira Islands

Selvagens Islands N, Africa

Canary Islands,-,,- --- Tropic of Cancer

Cape VerdeLslands

7744 Evolution: Kim et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 7745

Krigia montanaMicroseris laciniataPyrrhopappus multicaulis OutgroupsLactuca perenisLactuca sativaPrenanthes purpurea - PrenanthesTaraxacum officinale - TarxacumPrenanthes altissima - PrenanthesReichardia picroides -Reichardia tinginata RelchardlaReichardia ligulata _Launaea arborescens _Launaea nudicaulis j LaunaeaAetheorhiza bulbosa -AetheorhizaSonchus kirkiiSonchus asper Sonchus subg.Sonchus oleraceus SonchusSonchus bourgeauiSventenia bupleuroidesBabcockiaplatylepisPrenanthes pendula(N)Prenanthes pendula(S)Sonchus tuberiferSonchus canariensis The woodySonchus congestus SonchusSonchus fruticosus alliance In theSonchus gonzalezpadroni MacaroneslanSonchus ortunoi IslandsTaeckholmia pinnataTaeckholmia canariensisTaeckholmia heterophyllaTaeckholmia arboreaLactucosonchus webbliSonchus palustris SSonchus arvensis Sonchusaubg.Sonchus maritimus J SonchusKirkianella novae-zelandiae KirkianellaEmbergeria grandifolia - EmbergerlaDendroseris litoralisDendroseris marginataDendrosenis macrantha

rDendrosefis micrantha (TDendroerDendroseris pruinata Fhe JuanDendroseris neriifolia FernancdezhieDendrosenis pinnata Felande)Dendroseris berteroanaDendrosenis regia _Sonchus luxurians U Sonchussubg.Sonchus schweinfurthii .. Sonchus

FIG. 2. ITS phylogeny of subtribe Sonchinae (redrawn from ref. 28). This is one of the 144 equally most parsimonious trees (consistency index= 0.526; retention index = 0.743). Dashed lines indicate branches that collapse in strict consensus tree. Numbers above in parentheses and belowbranches represent decay and percentage of bootstrap values, respectively. Arrow indicates the clade of woody Sonchus alliance.

divergence and the polytomy in the ITS tree suggest a rapidradiation of major lineages early in the history of the allianceafter a single introduction.Adaptive radiation connotes the process by which a mono-

phyletic group of organisms adapts to a broad diversity ofhabitats (3, 4), and it has been of major importance in theevolution of woody Sonchus and its close relatives. The ITSphylogeny suggests that eight ecological shifts from mesic todry habitats and three shifts from dry to coastal habitats haveoccurred during the diversification of the group (Fig. 3). Twolineages of Taeckholmia (excluding two species of Sonchus)have radiated exclusively into dry habitats, whereas a majorclade of Sonchus has radiated primarily into mesic habitatswith two ecological shifts to dry habitats. The ITS phylogenyalso indicates that, during adaptive radiation, rosette-shrub/subshrub and rosette-tree habits have been quite successful incolonizing and adapting to the various habitats of the islands.In contrast, two lineages with tubers have only one specieseach, i.e., S. tuberifer and Lactucosonchus, and have not beensuccessful in radiating and speciating in the Macaronesianislands (Fig. 3).The ITS phylogeny also provides evidence that interisland

colonization events to similar ecological habitats and intrais-

land differentiation have played important roles in the evolu-tion of the alliance (Fig. 3). Note, for example, that S.canariensis occurs on two islands, as do S. hierrensis, S. acaulis,S. cogestus, and T. pinnata as well as other species (Fig. 3).Further, long distance dispersal has been responsible for theorigin of Sonchus species in other archipelagos (Fig. 3). Forexample, S. daltonii in the Cape Verde archipelago likelyresults from a recent dispersal event from the western CanaryIslands. Also, three species of Sonchus in Madeira appear tobe derived from a single colonization event, probably fromTenerife or Gran Canaria of the Canary Islands (Fig. 3).

Biogeographical Implications. The molecular data suggestthat, after the initial divergence of Lactucosonchus from acommon ancestor, the two genera Sventenia and Babcockia aswell as Prenanthespendula and S. tuberifer radiated early in theCanary Islands. These taxa are confined to mountains ormountain cliffs of the geologically oldest areas of Gran Canariaand Tenerife (with the exception of P. pendula, which is alsolocally common in the south and west of Gran Canaria), andsuggests that the woody Sonchus group arose in the CanaryIslands. Further, it indicates that subg. Dendrosonchus andTaeckholmia probably originated in the geologically oldestareas of Gran Canaria or Tenerife; remarkably these two

Evolution: Kim et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996)

I Sventenia bupleuroides (GC; cp)D Babockia platylepis (GC; s)

D 2 | Prenanthespendula (GC-N; s)81% I_. Prenanthespendula (GC-S; s)

Sonchus tuberifer (TE; hp) _

D mq|- S. pinnatifidus (LA; t) -

r S. brachylobus (GC; s)S. daltonii (CAPE VERDE; cp)S. hierrensis (GO; t)S. hierrensis (PA; t)S. gandogeri (HI; t)S. palmensis (PA; t)S. canariensis (TE; t)S. canarensis (GC; t)S. acaulis (GC; cp)S. acaulis (TE; cp)S. congestus (TE, GC; t)S. pinnatus (MADEIRA; t)S. fruticosus (MADEIRA; t)S. ustulatus (MADEIRA; cp)S. fauces-orci (TE; cp)S. gonzalezpadroni (GO; cp)S. ortunoi (GO; cp)T. pinnata (GC,TE; s)T. pinnata var. microcarpa (TE;T. canariensis (GO; s)T. heterophylla (GO-1; s)T. capillaris (TE; a)T. heterophylla (GO-2; a)T. arborea (TE; t)T. arborea (PA; t) J

Iy: Lactucosonchus webbii (PA-1; hp)Lactucosonchus webbii (PA-2; hp)Sonchus palustris (EUROPE; hp)

Sonchus subg. Sonchus

Sonchussubg.Dendro-sonchus

Taeckholmia

Sonchussubg.Dendro-sonchussensuAldridge

FIG. 3. Preferred ITS phylogeny of the woody Macaronesian Sonchus and their alliance. This is one of the 8123 equally most parsimonious trees(consistency index = 0.821; retention index = 0.821) and is identical to the 50% majority-rule consensus tree. Dashed lines and numbers aboveand below branches as in Fig. 2. Abbreviations of islands as in Fig. 1. Abbreviations of habits are as follows: hp, herbaceous perennial; cp, caudexperennial; s, rosette subshrub; t, rosette tree/shrub. Abbreviation of habitats are as follows: M, mesic; D, dry; C, coastal.

islands are also the center of diversity for subg. Dendrosonchusand Taeckholmia (45).Because the molecular clock cannot be rejected, it is possible

to estimate the time of the radiation of the woody Sonchusalliance. The genus Dendroseris, which is endemic to the JuanFernandez Islands, was used to estimate the rate of nucleotidesubstitution in ITS because both chloroplast (cp) DNA restric-tion sites (46) and ITS sequence data (28) suggest that it isclosely related to the Macaronesian group. In addition, Den-droseris is an insular endemic and has life forms similar to subg.Dendrosonchus (i.e., paliform and rosette trees to rosetteshrubs). Thus, generation-time effect on the substitution rateof the ITS sequences should be minimized (16). Two differentaverage rates of nucleotide substitutions per site per year inITS in the genus Dendroseris, r = (3.94 ± 0.10) x 10-9 and(6.06 ± 0.15) x 10-9, were used to estimate the time ofradiation. The slower rate assumes radiation shortly after theformation of the island of Masatierra (one of the JuanFernandez Islands), whereas the faster rate assumes a laterradiation and was estimated from cpDNA divergence. Theslower rate for the sequences using Dendroseris as the standardwas 0.78% per million years (11). The average sequence

divergence between Lactucosonchus and the other Macaron-

esian taxa is 3.34%, indicating that the divergence betweenthem may have occurred 4.2 million years ago (Mya). Theaverage sequence divergence between subg. Dendrosonchus(including Taeckholmia) and allied genera is 2.8%. Thus, thedivergence of subg. Dendrosonchus from the other genera mayhave occurred approximately 3.6 Mya. Therefore, the origin ofgenera in the woody Sonchus alliance may have taken placeabout 4.2 Mya or earlier on the Canary Islands, and theradiation of subg. Dendrosonchus accordingly took place be-tween 4.2 and 3.6 Mya on Gran Canaria or Tenerife. However,if we use the faster rate, i.e., 1.20% per million years, then theorigin of the alliance and the radiation of subg. Dendrosonchusmay have occurred 2.8 Mya or earlier and 2.3 Mya, respec-tively. Although all the colonization events of Sonchus speciesin the Macaronesian islands postdate the geological origin ofthe islands (47), the calculated time of occurrence of Lactu-cosonchus in La Palma predates the origin of the island (notolder than 1.5 Mya; Fig. lb). This suggests that the immediateancestor of Lactucosonchus evolved for some time on anotherisland in the Canaries, either Tenerife or Gran Canaria,followed by extinction of Lactucosonchus or its ancestor onsuch a source island.The estimated divergence times for the allied genera and the

radiation of subg. Dendrosonchus are long after the formation

15%

M

100,

240%6

94%

7746 Evolution: Kim et aL

)m

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 7747

of the Canary Islands, except the two westernmost ones, LaPalma and El Hierro. Even assuming some uncertainty indating the ages of the older islands, it seems highly likely thattheir ages are still considerably older than the calculated timeof radiation for Sonchus and its relatives. One question then iswhy Sonchus was so successful if radiation occurred after theislands were several million years old and presumably openhabitats were not plentiful. One hypothesis is that extinctionsin the Canaries and northwestern Africa during the firstglaciation in the Northern Hemisphere (2.8 Mya; ref. 48) andthe beginning of Sahara desertification (2.5 Mya; ref. 49) mayhave provided many open habitats for the radiation of thealliance. The estimated divergence times, using the fastermutations rate for Dendroseris in the calculations, coincideclosely with glaciation and desertification.The Origin of the Woody Sonchus in Macaronesia and

Insular Woodiness. The origin of woody Sonchus in Macaro-nesia is controversial. Boulos (50) proposed that subg. Den-drosonchus evolved from subg. Origosonchus (endemic toAfrica), which he considered the most primitive group ofSonchus. This hypothesis is supported by pollen morphology(51). In contrast, Aldridge (45) suggested that subg. Den-drosonchus is very primitive and that the two subgeneraOrigosonchus and Sonchus were derived from it. Her hypoth-esis is congruent with Takhtajan's view (52) and also agreeswith Bramwell (24, 25). Neither of these two hypotheses issupported by the ITS phylogeny. The ITS phylogeny indicatesthat subg. Dendrosonchus is a relatively derived group but wasnot derived directly from subg. Origosonchus (Figs. 2 and 3).Rather, it is more likely that it originated from an ancestorsomewhat like one of its allied genera in the Canary Islands(Fig. 3).The woody life-forms in the Canaries in genera such as

Sonchus, Echium,Argyranthemum, Pericallis, and Crambe havebeen considered by several authors to represent tertiary relicts(6, 17, 24, 53-55). Bramwell (24) suggested that taxonomic,cytological, morphological, palaeobotanical, distributional,and phytogeographical data are consistent with the endemicflora being of considerable age and probably ancestral to manymodern Mediterranean genera and species. Meusel (53)viewed the woody Macaronesian species as ancestral forms ofmodern Mediterranean herbaceous species, with the herba-ceous forms derived by reduction in lignification and adapta-tion to more extreme conditions. In contrast, Carlquist (2-4)argued that the endemic frutescent species found on manyoceanic islands are the result of an increase in woodiness inresponse to the uniformity of insular climates. More specifi-cally, Carlquist (3) claimed that a group of true shrubbySonchus in the Macaronesian Islands evolved from plantssimilar to the European weedy sow thistle (S. oleraceus). Thus,he did not consider these plants to be relicts but rathersecondary derivatives of herbaceous ancestors. The ITS treeclearly shows that neither subg. Sonchus nor subg. Origoson-chus was derived from the woody members of Sonchus inMacaronesia (Fig. 2). Rather, it is evident that the woodySonchus and its relatives were derived ultimately from conti-nental herbaceous perennial ancestors (Fig. 3). Further, theITS tree shows a general trend toward increased woodiness.For example, it is likely that the ancestor of the entire alliancewas an herbaceous perennial, and then there was evolutiontoward caudex perennials, shrubs, and trees in different lin-eages during radiation in the Macaronesian islands (Fig. 3). Inparticular, it seems reasonable to assume that the immediateancestors of the three Madeiran Sonchus species and thelineage consisting of most of the woody Sonchus species wereprobably either herbaceous or caudex perennials and thenbecame trees after dispersal to Madeira and various CanaryIslands (see asterisked clade in Fig. 3). This trend is awell-known feature of adaptive radiation in many insular

plants (4). Thus, the molecular data do not support subg.Dendrosonchus being relictually woody plants.

In conclusion, the woody Sonchus alliance in the Macaro-nesian islands apparently resulted from a single introduction.This founder event was likely followed by several radiations.The ITS phylogeny refutes the relictural nature of the woodylife form of Sonchus in Macaronesia and suggests that itrepresents secondary derivation from herbaceous ancestors.The ITS sequence data also suggest that evolution of the woodySonchus alliance may have occurred during Late Tertiary (i.e.,the late Miocene or early Pliocene) from Gran Canaria orTenerife in the Canary Islands. The rapid radiation of majorlineages in the alliance is consistent with an unresolved poly-tomy at the base of the cladogram and low sequence divergence.

We thank Aguedo Marrero, Pedro Ortega-Machin, and FranciscoJose Gonzalez Artiles for assistance during a field trip in the CanaryIslands. Also, Instituto Canario de Investigaciones Agrarias in Tener-ife, through the advice of Manuel Fernandez-Galvan, provided finan-cial support for field studies to one of us (J.F.O.). We are indebted toCharles Jeffrey and Kare Bremer for helpful suggestions aboutoutgroups and are grateful to Robert Jansen, Eric Knox, Tom Myers,David Glenny, P.J. Garnock-Jones, Loutfy Boulos, and other peoplefor providing plant and DNA materials. We thank Angela Aldridge forher helpful discussion about this project. We are especially indebtedto Sherwin Carlquist for many helpful comments and suggestions onan earlier version of this paper. This work was supported by Sigma XiGrants-in-Aid of Research, Graduate Student Research Grant fromAmerican Society of Plant Taxonomists, Tinker Foundation (LatinAmerican Studies Program, Ohio State University), Janice BeatleyHerbarium Award (Ohio State University Herbarium) to S.-C.K., andNational Science Foundation Doctoral Dissertation ImprovementGrant DEB-9521017 to D.J.C. and S.-C.K.

1. Carlquist, S. (1965) Island Life (The Natural History Press, NewYork).

2. Carlquist, S. (1970) Hawaii: A Natural History (The NaturalHistory Press, New York).

3. Carlquist, S. (1974) Island Biology (Columbia Univ. Press, NewYork).

4. Carlquist, S. (1995) in Hawaiian Biogeography: Evolution on a HotSpotArchipelago, eds. Wagner, W. & Funk V. (Smithsonian Inst.,Washington, DC), pp. 1-13.

5. Lack, D. (1947) Da,win's Finches (Cambridge Univ. Press, NewYork).

6. Bramwell. D. (1975) Anal. Inst. Bot. Cavanilles 32, 241-254.7. Baldwin, B. G., Kyhos, D. W. & Dvorak, J. (1990) Ann. Mo. Bot.

Gard. 77, 96-109.8. Carr, G. D. & Kyhos, D. W. (1981) Evolution 35, 543-556.9. Helenurm, K. & Ganders, F. (1985) Evolution 39, 753-765.

10. Baldwin, B. G., Kyhos, D. W., Dvorak, J. & Carr, G. D. (1991)Proc. Natl. Acad. Sci. USA 88, 1840-1843.

11. Sang, T., Crawford, D. J., Kim, S.-C. & Stuessy, T. F. (1994)Am. J. Bot. 81, 1494-1501.

12. Sanders, R. W., Stuessy, T. F., Marticorena, C. & Silva, 0. M.(1987) Opera Bot. 92, 195-215.

13. Givnish, T. J., Sytsma, K. J., Smith, J. F. & Hahn, W. J. (1994)Proc. Natl. Acad. Sci. USA 91, 2810-2814.

14. Sang, T., Crawford, D. J., Stuessy, T. F. & Silva, 0. M. (1995)Syst. Bot. 20, 55-64.

15. Baldwin, B. G. (1992) Mol. Phylogenet. Evol. 1, 3-16.16. Soltis, P. S. & Soltis, D. E. (1995) in Evolutionary Biology, eds.

Hecht, M. K., Macintyre, R. J. & Clegg, M. T. (Plenum, NewYork), pp. 139-194.

17. Sunding, P. (1979) in Plants and Islands, ed. Bramwell, D.(Academic, New York), pp. 13-40.

18. McDougall, I. & Schmincke, H. U. (1976-1977) Bull. Volcanol.40, 57-77.

19. Banda, E., Danobeitia, J. J., Surinach, E. & Ansorge, J. (1981)Earth Planet. Sci. Lett. 55, 11-24.

20. Carracedo, J. C. (1994) J. Volcanol. Geotherm. Res. 53, 251-274.21. Abdel-Monem, A., Watkins, N. D. & Gast, P. W. (1971) Am. J.

Sci. 271, 490-521.22. Abdel-Monem, A., Watkins, N. D. & Gast, P. W. (1972) Am. J.

Sci. 272, 805-825.

Evolution: Kim et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996)

23. Carlquist, S. (1962) Phytomorphology 12, 30-45.24. Bramwell, D. (1972) in Taxonomy, Phytogeography, and Evolution,

ed. Valentine, D. H. (Academic, New York), pp. 141-159.25. Bramwell, D. (1976) in Biogeography and Ecology in the Canary

Islands, ed. Kunkel, G. (Dr. W. Junk Publishers, Hague), pp.207-240.

26. Bohle, U.-R., Hilger, H., Cerff, R. & Martin, W. F. (1994) inMolecular Ecology and Evolution: Approaches and Application,eds. Schierwater, G. P., Streit, B., Wagner, G. P. & DeSalle, R.(Bukhauser, Basel, Switzerland), pp. 391-403.

27. Bremer, K (1994) in Asteraceae: Cladistics & Classification(Timber, Portland, OR), pp. 157-201.

28. Kim, S.-C., Crawford, D. J. & Jansen, R. K. (1996) Syst. Bot., inpress.

29. Aldridge, A. (1977) Bot. Macar. 3, 41-59.30. Aldridge, A. (1978) Bot. J. Linn. Soc. 76, 249-285.31. Aldridge, A. (1976) Bot. Macar. 2, 25-57.32. Perez de Paz, J. (1976) Bot. Macar. 1, 51-65.33. Baldwin, B. G. & Robichaux, R. H. (1995) in Hawaiian Biogeog-

raphy: Evolution on a Hot Spot Archipelago, eds. Wagner, W. &Funk, V. (Smithsonian Inst., Washington, DC), pp. 259-287.

34. Doyle, J. J. & Doyle, J. L. (1987) Phytochem. Bull. 19, 11-15.35. Maddison, W. P. & Maddison, D. R. (1992) MACCLADE, Analysis

of Phylogeny and Character Evolution (Sinauer, Sunderland,MA), Version 3.0.

36. Swofford, D. L. (1993) PAUP, Phylogenetic Analysis Using Par-simony (Smithsonian Inst., Washington, DC), Version 3.1.1.

37. Felsenstein, J. (1985) Evolution 39, 783-791.38. Felsenstein, J. (1986-1993) PHYLIP, Phylogeny Inference Package

(Univ. of Washington, Seattle), version 3.52c.39. Li, W.-H. & Tanimura, M. (1987) Nature (London) 326, 93-96.

40. Wu, C.-I. & Li, W.-H. (1985) Proc. Natl. Acad. Sci. USA 82,1741-1745.

41. Baldwin, B. G., Campbell, C. S., Porter, J. M., Sanderson, M. J.,Wojciechowski, M. F. & Donoghue, M. J. (1995) Ann. Mo. Bot.Gard. 82, 247-277.

42. Hansen, A. & Sunding, P. (1993) Sommerfeltia 17, 1-295.43. Francisco-Ortega, J., Jansen, R. K., Crawford, D. J. & Santos-

Guerra, A. (1995) Syst. Bot. 20, 413-422.44. Lowrey, T. K. (1995) in Hawaiian Biogeography: Evolution on a

Hot SpotArchipelago, eds. Wagner, W. & Funk, V. (SmithsonianInst., Washington, DC), pp. 195-220.

45. Aldridge, A. (1979) in Plants and Islands, ed. Bramwell, D.(Academic, New York), pp. 279-290.

46. Whitton, J., Wallace, R. S. & Jansen, R. K. (1995) Can. J. Bot. 73,1058-1073.

47. Mitchell-Thome, R. C. (1985) Bol. Mus. Munic. Funchal 37,52-85.

48. Flint, R. R. (1971) in Glacial and Quaternary Geology (Wiley,New York).

49. Williams, M. A. J. (1982) in Quaternary Environments in NorthernAfrica, eds. Williams, M. S. J. & Adamsom, D. A. (Balkema,Rotterdam), pp. 13-22.

50. Boulos, L. (1967) Bot. Not. 120, 95-108.51. Saad, S. I. (1961) Pollen Spores 3, 247-260.52. Takhtajan, A. (1969) Flowering Plants: Origin and Dispersal

(Smithsonian Inst., Washington, DC).53. Meusel, H. (1953) Flora 139, 333-393.54. Takhtajan, A. (1978) The Floristic Region of the World (Nauka,

Leningrad, U.S.S.R.).55. Cronk, Q. C. B. (1992) Biol. J. Linn. Soc. 46, 91-103.

7748 Evolution: Kim et aL

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

0