Proc. Natl. Acad. Sci. USAVol. 93, pp. 4085-4090, April 1996Evolution

Chloroplast DNA evidence of colonization, adaptive radiation, andhybridization in the evolution of the Macaronesian flora

(oceanic island flora/Atlantic islands/biogeography/Angiosperms/Argyranthemum)JAVIER FRANCISCO-ORTEGA*, ROBERT K. JANSEN*t, AND ARNOLDO SANTOS-GUERRA:*Department of Botany, University of Texas, Austin, TX 78713-7640; and tJardin de Aclimataci6n de La Orotava, Puerto de La Cruz, Tenerife 38400,Canary Islands, Spain

Communicated by Thomas N. Taylor, University of Kansas, Lawrence, KS, December 14, 1995 (received for review June 16, 1995)

ABSTRACT Most evolutionary studies of oceanic islandshave focused on the Pacific Ocean. There are very few exam-ples from the Atlantic archipelagos, especially Macaronesia,despite their unusual combination of features, including aclose proximity to the continent, a broad range of geologicalages, and a biota linked to a source area that existed in theMediterranean basin before the late Tertiary. A chloroplastDNA (cpDNA) restriction site analysis of Argyranthemum(Asteraceae: Anthemideae), the largest endemic genus ofplants of any volcanic archipelago in the Atlantic Ocean, wasperformed to examine patterns of plant evolution in Maca-ronesia. cpDNA data indicated that Argyranthemum is amonophyletic group that has speciated recently. The cpDNAtree showed a weak correlation with the current sectionalclassification and insular distribution. Two major cpDNAlineages were identified. One was restricted to northernarchipelagos-e.g., Madeira, Desertas, and Selvagens-andthe second comprised taxa endemic to the southern archipel-ago-e.g., the Canary Islands. The two major radiationsidentified in the Canaries are correlated with distinct ecolog-ical habitats; one is restricted to ecological zones under theinfluence of the northeastern trade winds and the other toregions that are not affected by these winds. The patterns ofphylogenetic relationships in Argyranthemum indicate thatinterisland colonization between similar ecological zones isthe main mechanism for establishing founder populations.This phenomenon, combined with rapid radiation into distinctecological zones and interspecific hybridization, is the pri-mary explanation for species diversification.

Oceanic islands provide model systems to assess organismicevolution following colonization and isolation. Classical stud-ies of the Galapagos finches (1) and Hawaiian Drosophila (2)have had a major impact on the history of evolutionary thoughtin the 19th and 20th centuries. Oceanic islands are relativelysimple systems where both patterns of dispersal and naturalselection can be more easily assessed than in continentalsystems. Most studies of island biota have been restricted to thePacific Ocean archipelagos of Hawaii (3-6), Galapagos (7, 8),and Juan Fernandez (9-11). Few studies (12, 13) have focusedon the Atlantic Ocean archipelagos that constitute the bio-geographical region known as Macaronesia (Fig. 1). Thesearchipelagos have several unique features that make themideal systems for understanding the origin and evolution ofisland floras and faunas. Macaronesia comprises five majorarchipelagos (Azores, Canaries, Cape Verde, Madeira, andSelvagens) that exhibit a broad range of geological ages from0.8 to 21 million years (14). The region is in close proximity tothe continent, unlike most volcanic archipelagos. In addition,the Macaronesian biota has been linked to one that spreadfrom the Mediterranean basin in the late Miocene and Plio-

cene, when the first northern hemisphere glaciation and thedesertification of most of northern Africa led to massiveextinction and migration of plants and animals (15).

Macaronesia, like other tropical and subtropical volcanicarchipelagos, is under the influence of the trade winds. Thesewinds, in combination with altitudinal gradients on the islands,have produced several distinct ecological zones (15). Two mainclimatic regions can be distinguished. The first is under theinfluence of humid and cool northeastern trade winds andmainly covers a band in the north of the islands between 400and 1200 m. This region has three main ecological zones:humid lowland scrub, laurel forest, and heath belt (Table 1).In contrast, the second climatic region is not under theinfluence of the trade winds and is more arid. This area isrestricted to the southern slopes of the islands, in the north ataltitudes higher than 1200 m and in the coastal belt below 400m. Four major ecological zones are recognized in the aridclimatic region: coastal desert, arid lowland scrub, Canary pineforest, and high altitude desert (Table 1). In many instances,the northeastern trade winds can also have an influence on thesouthern slopes of some of the islands (18). This is especiallytrue in the islands where the summit barely reaches 1200 m. Inthese cases, the laurel forest can also spread toward thesouthern slopes and a humid lowland scrub is found below thatforest in the south or west.The high diversity of habitats appears to be one of the main

factors responsible for the rich Macaronesian flora, whichincludes -700 endemic species (19). The largest endemic plantgenus (Argyranthemum) of all Atlantic Ocean volcanic islandsoccurs in these archipelagos (20, 21). Its 23 species provide themost spectacular example of speciation in Macaronesia (22),and at least one species is endemic to each of the majorecological zones (Table 1). Thus, Argyranthemum is an idealgroup to examine patterns of speciation in Macaronesia.Further justification for studying this genus is that the closestrelatives (Chrysanthemum and Heteranthemis) have beenidentified from the continent (refs. 23-27; J.F.-O., R.K.J.,A. Hines, and A.S.-G., unpublished data).We used Argyranthemum as an example to assess the origin

and evolution of the Macaronesian flora. Phylogenies gener-ated from chloroplast DNA (cpDNA) data were used toexamine the role of colonization, adaptive radiation, andhybridization in the evolution of endemic plants from theAtlantic archipelagos. The approach is similar to previousstudies of endemic genera from three volcanic archipelagosfrom the Pacific Ocean: Dendroseris (9) and Robinsonia (10) inthe Juan Fernandez Islands; the silversword alliance (3) andCyanea (6) in the Hawaiian Islands; and Gossypium in theGalapagos (8) and Hawaiian (28) Islands. In all of these cases,cpDNA restriction site analyses were informative and providedevidence for rapid speciation, hybridization, and colonizationin the flora of Pacific archipelagos.

Abbreviation: cpDNA, chloroplast DNA.tTo whom reprint requests should be addressed.

4085

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

Aug

ust 5

, 202

0

4086 Evolution: Francisco-Ortega et al.

FIG. 1. The Macaronesian islands (inset) and the distribution ofArgyranthemum in the Madeira, Selvagen, and Canary archipelagos.Blackened islands are those where Argyranthemum occurs.

Our investigation identified a strong and unexpected cor-

relation between the two major climatic regions of Macaro-nesia and the groups in the cpDNA phylogeny. Patterns ofphylogenetic relationships in Argyranthemum indicate thatinterisland colonization between similar ecological zones,adaptive radiation, and hybridization have all played an im-portant role in speciation in the genus.

MATERIALS AND METHODSOne or more populations of each taxon of Argyranthemum,Heteranthemis, and Chrysanthemum were studied (see Fig. 2).Total genomic DNA was isolated from living material (29) andpurified in CsCl/ethidium bromide gradients. Restriction en-

zyme digestions, agarose gel electrophoresis, transfer of DNAfragments to filters, labeling of probes by nick translation with32p, and subsequent hybridization and autoradiography were

carried out as described elsewhere (30). Fourteen clonedrestriction fragments of lettuce cpDNA were used (31). Twen-ty-one restriction endonucleases were employed, including 146-bp enzymes (Ava I, Ava II, Ban II, Bgl II, BstNI, Cla I, DraI, EcoO109I,EcoRI, HincII, Nci I, Nsi I,Xba I, andXmn I) and7 4-bp enzymes (BstUL, Hae III, Hha I, Hinfl, Msp I, Rsa I, andTaq I). Complete restriction site maps were constructed foreach taxon for the 6-bp enzymes. The large number and smallsize of fragments generated by the 4-bp enzymes did not permitmapping. Mapped restriction sites for the 6-bp enzymes were

used in phylogenetic analyses, and restriction site changes wereinferred from fragment patterns for the seven 4-bp enzymes.

Phylogenetic analyses were performed on two data sets; oneincluded all 55 populations, and a second included only 15 taxafrom all major clades identified in the larger analysis. BothWagner (32) and weighted (33) parsimony were performed ona Macintosh Quadra 700 microcomputer using PAUP version3.1.1 (34). The 55-taxon analysis used the TREE BISECTIONRECONNECTION, MULPARS, ACCTRAN, and COLLAPSE optionswith 100 random entries of the taxa. The 15-taxon data matrixused the BRANCH AND BOUND, COLLAPSE, and ACCTRAN op-tions. Weighted parsimony was performed using a range of

weights from 1.01:1 to 2.5:1 of site gains/losses with ancestralstates designated as Os (35). Bootstrap values (36) from 100replicates were calculated to assess the amount of support formonophyletic groups. Maximum likelihood values were esti-mated for a subset of trees in the 15-taxon analysis using theRESTML program of PHYLIP (37).

Estimates of nucleotide sequence divergence were alsocomputed using equations 9 and 10 of Nei and Li (38). Valueswere calculated only for the 6-bp enzymes because of theavailability of complete restriction maps. Sequence divergencewas calculated among 11 taxa, including the four outgroupsand seven Argyranthemum species from the three major lin-eages identified in the phylogenetic analyses of the completedata set.

RESULTSThe 21 restriction enzymes identified '3190 restriction sites(978 from 6-bp enzymes and 2112 from 4-bp enzymes) incpDNAs of each species. This represents 14,716 bp, or 11% ofthe chloroplast genome of each taxon. A total of 151 variablerestriction sites was detected, 84 of which were phylogeneti-cally informative. Wagner parsimony yielded 54 equally par-simonious trees of 172 steps with a consistency index of 0.80(excluding uninformative changes) and a retention index of0.967. The trees provided strong support (14 characters, 100%bootstrap value) for the monophyly of Argyranthemum. Twomajor clades were resolved in the genus (Fig. 2A). One cladeincluded taxa endemic to the Canaries (Canarian clade),whereas the second one comprised only species from Madeira,Selvagens, and Desertas (Madeiran clade). The Canarian cladeconsisted of two clades (A and B), which collapsed in the strictconsensus tree (Fig. 2A, dashed line). These three lineages(one Madeiran and two Canarian) were detected in all 54equally parsimonious trees and their relationships were re-solved in three different ways: (i) Madeira sister to the CanaryIslands, (ii) one of the two canary clades sister to the Madeiraand the other Canarian clade, and (iii) an unresolved trichot-omy. The bootstrap analysis provided weak support for theresolution of the trichotomy; the Madeiran lineage was sisterto the Canarian taxa in only 12 of the 100 replicates. Insummary, Wagner parsimony did not resolve the basal rela-tionships in Argyranthemum.Weighted parsimony is preferable to Wagner parsimony

because of the known asymmetry in the probability of gainsversus losses of restriction sites (33, 35). Thus we performedweighted parsimony using a range of weights from 1.01:1.0 to2.5:1.0 of gains/losses. All weighted parsimony analyses, evenwith a low weight of 1.01 for gains, generated trees with theMadeira lineage sister to the Canarian taxa (Fig. 2). The twoequally parsimonious trees at a weight of 1.01 for gains hadidentical topologies to 2 of the 54 equally parsimoniousWagner trees; the only difference concerned the position ofA.foeniculaceum, which collapsed into a large polytomy with otherCanarian taxa in one of the two trees (Fig. 2A, dashed arrow).Weighted parsimony, and 23 of the Wagner trees, suggested

that the Madeira lineage may be sister to the Canarian taxa, butsupport for this result was weak. We performed maximumlikelihood analysis to determine which resolution of basalrelationships is more likely. The maximum likelihood programis too intensive computationally to use all 55 taxa. Thus weselected 11 Argyranthemum species representing all majorlineages and four outgroup taxa for these analyses (indicatedin boldface type in Fig. 2A). Wagner and weighted parsimonyanalyses of these 15 taxa demonstrated that the reduction inthe number of taxa had no effect on the major groups. Wagnerparsimony generated two equally parsimonious trees (data notshown) of 144 steps (consistency index of 0.867 and retentionindex of 0.940), one with Madeira sister to the Canarian taxaand the other with a Canary Island group sister to the

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

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 4087

Table 1. Distribution of Argyranthemum in Macaronesia

Age, Altitude, Surface, No. of taxa per ecological zone*Age, Altitude, Surface,Island Myr m km2 CD ALS HLS LF HB PF HD

Canary IslandsEl Hierro 0.8 1520 307 1 0 1 0 1 1La Palma 2 2423 789 1 0 2 0 1 1 1La Gomera 12.5 1484 425 2 0 1 1 0Tenerife 12 3714 2355 3 2 3 1 1 2 1Gran Canaria 14 1950 1625 3 1 2 1 0 1Fuerteventura 21 807 1717 0 0 1Lanzarote 15.5 670 717 0 0 1

Madeira IslandsMadeira 5 1861 730 2 0 1 1 1Deserta Grande 5 479 20 1 0 -

Selvagen IslandsSelvagem Pequena 12 49 0.5 1

Myr, million years.*Ecological zones (altitude ranges are given in parentheses) are as follows: CD, coastal desert (0-400 m); ALS, arid lowlandscrub (400-600 m, only in zones which do not receive the influence of the northeastern trade winds); HLS, humid lowlandscrub (400-500 m, only in zones which receive the influence of the northeastern trade winds); LF, laurel forest (600-1000m, and only on northern face zones); HB, heath belt (1000-1200 m, only on northern face zones); PF, pine forest (1200-2000m on northern face zones; 600-2000 on southern face zones); HD, high altitude desert (2000-3700 m). A - indicates thatecological zone is absent. Geographical and geological data were compiled from refs. 14, 16, 17, and 19. Each taxon is endemicto a particular zone and a particular island, the exception beingA. haemotomma, which is found both in Madeira and DesertaGrande.

remaining taxa. Likelihood values were calculated for the twomost parsimonious topologies and 20 additional trees up to twosteps longer than the shortest trees. Maximum likelihoodanalysis requires the specification of the number of base pairsin the recognition site of the enzymes. Our study used acombination of 4-bp and 6-bp enzymes. We performed thelikelihood estimates using both values. The two shortest trees(data not shown, 144 steps) had the highest likelihood values,regardless of whether 4 bp (tree 1, Ln = -7393.21524; tree 2,Ln = -7395.15716) or 6 bp (tree 1, Ln = -10654,42282; tree2, Ln = - 10656.79832) was specified. Although the likelihoodof all 22 tree topologies did not differ significantly by theKishino and Hasegawa paired-sites test (39), the tree with theMadeira lineage sister to the Canarian taxa had the highestlikelihood and will be used hereafter.The groups in the cpDNA tree did not correlate well with the

current sectional classification in that no section was mono-phyletic (Fig. 2A). There was also a poor correlation betweeninsular distribution and the cpDNA clades because none of themajor classes were restricted to a single island (Fig. 2B).cpDNA divergence in Argyranthemum (Table 2) ranged

from 0.056% (between A. pinnatifidum and A. haemotomma) to0.185% (between A. pinnatifidum and A. haouarytheum). Diver-gence between Argyranthemum species and continental generaranged from 0.155% (between A. maderense and H. viscidehirta)to 0.298% (between A. haouarytheum and C. segetum).

DISCUSSIONColonization and Adaptive Radiation. The cpDNA phylog-

eny (Fig. 2A) provides strong support for the monophyly ofArgyranthemum, in agreement with previous morphological(23) and DNA (27) evidence. This result, combined with thefact that the genus is absent from the continent, suggests thatArgyranthemum has experienced a single colonization intoMacaronesia. The cpDNA tree also resolves two major lin-eages in genus (Fig. 2A). The first is restricted to the northernarchipelagos in the islands of Madeira, Selvagens, and Deser-tas, and the second occurs in the Canary Islands. The northernarchipelagos are ecologically less diverse than the CanaryIslands. The island of Madeira has four ecological zones,whereas Selvagem Pequena and Deserta Grande have one andtwo, respectively. The low number of ecological zones and

islands is reflected in the patterns of the cpDNA phylogeny(Fig. 2B): evolution of Argyranthemum in the northern archi-pelagos appears to have resulted from interisland colonizationbetween the coastal areas, and radiation in the largest andmost ecologically diverse island of Madeira. Each of the fourecological zones other than the coastal desert on Madeira hasbeen exploited by at least one taxon ofArgyranthemum (Table1), and all of these taxa were derived from a single ancestor orfounder. Radiation apparently occurred rapidly without fur-ther enhancement of genetic variation from new colonizers,and it is likely that it followed a sequence from the coasttowards the rest of the island.Argyranthemum in the Canary Islands had a more complex

evolutionary history that involved more islands and at leastthree ecological zones in each island (Table 1). This is re-flected in the cpDNA phylogeny. The two major clades in theseislands were not correlated with insular distribution or sec-tional classification (Fig. 2). However, there was a clearrelationship between the cpDNA phylogeny and the ecology ofArgyranthemum (Fig. 2B). Most taxa (22 of 25) from ecologicalzones not influenced by the trade winds-i.e., coastal desert,arid lowland scrub, pine forest, and high altitude desert-wererestricted to clade A from the Canary Islands. The secondclade (clade B) consisted primarily of taxa (12 of 18) that occurin ecological zones under the influence of the trade winds-i.e., humid lowland scrub, laurel forest, and heath belt. ThecpDNA phylogeny suggests that at least two independentintroductions have occurred in Tenerife, La Palma, El Hierro,and La Gomera, and that most interisland colonization oc-curred between similar climatic regions. This finding is incontrast with the pattern of ecological radiation in the Ha-waiian silversword alliance (40), where major ecological shiftsbetween wet and dry habitats accompanied speciation in eachof the major island-endemic lineages.

Restriction of phylogenetic lineages to particular climaticregions has not been reported for other oceanic island groups.This correlation may occur in other groups in the CanaryIslands because several genera [i.e., Crambe and Descurainia(Brassicaceae), Cheirolophus (Asteraceae), and Limonium(Plumbaginaceae)] have most of their taxa in the same habitaton different islands (15). Our results support the hypothesisthat interisland colonization has taken place between areaswith similar ecology.

Evolution: Francisco-Ortega et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

4088 Evolution: Francisco-Ortega et al.

Taxon2fru

1 frufrLIu c

2 fruAl

1 88 gra68 -- fru LIuC

8 fru gra3 lem

fru foe4 tru par

2r hao1 8Z2 hao 1

fru 13 fru purn

frUl 2tru canllid

I lid 1.1 d1ada jac59 --- ada gra

ada canada dugfilesc

8 tbro gom-- hao 2

ten- foe

1cal2 web76 ada pal

broada

_-]1 ---ada erv

2 sve

A 2

3 hie2

hie2-~ mad

2

L win2 piIl Sticdis

-1 dis

-0 72 pin mntpinhaetha

5--- hae15. Het vis

_, - Chr car100°-° Chr cor

1001_11 Chrseg

Section

Arg

Pre

] Mon- Sti

Arg- Stig- Pre- Sti

] Pre

Arg

Sti

]Arg- Sph]ArgSph

- Sti-Arg-Sti

a)

COCO

I-

uc6r,

C-ctCOUtu

B

0)

C3

EcologicalIsland zone

I .. J.{....i L. t...L... t_ k I I.,.l

---b rCDI(1;5)D NlMadeira I-Il S

C t [)Deserta Grand e - CDSelvagem Pequ. - CD

Madeira CDIberia and Morocco

MoroccoOutgroup Mediterranean basin

Mediterranean basin

FIG. 2. One of the two equally parsimonious trees from weighted parsimony analyses (weight of 1.01 in favor of site gains). Dashed lines inA indicate branches that collapse in the strict consensus tree of 54 equally parsimonious Wagner trees. Arrow inA indicates the branch that collapsedin the second weighted parsimony tree. Number of changes is given above each branch. Bootstrap value (100 replicates) >50 is shown below eachbranch. Taxa in boldface type were used in maximum likelihood comparisons. Taxa are coded as follows: ada (Argyranthemum adauctum subsp.adauctum), ada can (Argyranthemum adauctum subsp. canariae), ada dug (Argyranthemum adauctum subsp. dugourii), ada ery (Argyranthemumadauctum subsp. erythrocarpon), ada gra (Argyranthemum adauctum subsp. gracile), ada jac (Argyranthemum adauctum subsp. jacobifolium), adapal (Argyranthemum adauctum subsp.palmensis), bro (Argyranthemum broussonetii subsp. broussonetii), brogom (Argyranthemum broussonetii subsp.gomerensis), cal (Argyranthemum callichrysum), Chr car (Chrysanthemum carinatum), Chr cor (Chrysanthemum coronarium), Chr seg (Chrysanthe-mum segetum), cor (Argyranthemum coronopifolium), dis (Argyranthemum dissectum), esc (Argyranthemum escarrei) fil (Argyranthemum filifolium),foe (Argyranthemum foeniculaceum), fru (Argyranthemum frutescens subsp. frutescens), fru can (Argyranthemum frutescens subsp. canariae), fru foe(Argyranthemum frutescens subsp. foeniculaceum), fru gra (Argyranthemum frutescens subsp. gracilescens), fru par (Argyranthemum frutescens subsp.parviflorum), frupum (Argyranthemum frutescens subsp. pumilum), fru suc (Argyranthemum frutescens subsp. succulentum), fru 1 (Argyranthemumfrutescens subsp. nov. 1), fru 2 (Argyranthemum frutescens subsp. nov. 2), gra (Argyranthemum gracile), hae (Argyranthemum haemotomma), hao(Argyranthemum haouarytheum subsp. haouarytheum), hao 1 (Argyranthemum haouarytheum subsp. nov. 1), hao 2 (Argyranthemum haouarytheumsubsp. nov. 2), Het vis (Heteranthemis viscidehirta), hie (Argyranthemum hierrense), lem (Argyranthemum lemsii), lid (Argyranthemum lidii subsp. lidii),lid 1 (Argyranthemum lidii subsp. nov. 1), mad (Argyranthemum maderense), A 1 (Argyranthemum sp. nov. 1),A 2 (Argyranthemum sp. nov. 2), pin(Argyranthemumpinnatifidum subsp.pinnatifidum),pin mon (Argyranthemumpinnatifidum subsp. montanum),pin suc (Argyranthemumpinnatifidumsubsp. succulentum), sun (Argyranthemum sundingii), sve (Argyranthemum sventenii), ten (Argyranthemum tenerifae), tha (Argyranthemumthalassophilum), web (Argyranthemum webbii), and win (Argyranthemum winteri). The distribution of the species among the five recognized sections(22) ofArgyranthemum (Arg), Monoptera (Mon), Preauxia (Pre), Sphenismelia (Sph), Stigmatotheca (Sti) is given (A), as is the geographic distributionamong islands (B). Ecological zones in B are coded as in Table 1. Solid circles in B indicate ecological habitats not influenced by the trade winds.

One conclusion that can be drawn from our results concernsthe pattern of dispersal followed by Argyranthemum. In theMadeiran, clade,A. haemotomma from Madeira is sister to the

remaining taxa, whereas taxa from Selvagens and Desertasoccur in a more derived position (Fig. 2B). Thus, taxa from thenorthern archipelago may have followed a "stepping stone"

A

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

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 4089

Table 2. Nucleotide sequence divergence (equations 9 and 10 from ref. 38) expressed as percentage (above diagonal) and number ofrestriction site differences (below diagonal) between pairs of taxa of Argyranthemum and continental relatives

Taxa

Chr car Chr cor Chr seg Het vis ada fru hae* hao 3 pin ten madChr car 0.012 0.080 0.128 0.209 0.233 0.217 0.282 0.225 0.225 0.177Chr cor 3 0.048 0.128 0.209 0.233 0.217 0.282 0.225 0.225 0.177Chr seg 9 6 0.144 0.209 0.249 0.234 0.298 0.235 0.241 0.193Het vis 17 16 16 0.160 0.201 0.168 0.234 0.176 0.193 0.155ada 24 23 26 20 0.104 0.104 0.152 0.121 0.112 0.160fru 30 32 28 19 13 0.128 0.064 0.136 0.136 0.088hae 24 23 25 19 12 16 0.177 0.056 0.120 0.072hao 3 32 35 33 26 20 8 21 0.185 0.153 0.153pin 25 26 27 22 13 16 7 23 0.128 0.118ten 24 24 25 22 13 16 14 19 15 0.098mad 24 23 25 21 7 17 14 21 13 13Taxa are abbreviated in Fig. 1.

*A. haemotomma from Deserta Grande.

model (41). Alternatively, there could have been independentcolonizations of Desertas and Salvagens from Madeira. Ineither case, the north-to-south sequence Madeira-Desertas-Selvagens is the prevailing direction of dispersal. In contrast,colonization between adjacent islands is not the predominantpattern in the Canary Islands. There is evidence of dispersalbetween Lanzarote and Fuerteventura, two adjacent islandssituated furthest east. However, the two species involved (A.maderense and A. winteri) do not occur in the same lineage asthe taxa on the closest island, Gran Canaria (Fig. 2B). Indeed,the Gran Canarian taxa form a distinct group in clade B of theCanarian lineage, that comprises taxa exclusively from thewestern islands. This suggests long distance dispersal of Argy-ranthemum between the western and the eastern CanaryIslands or, perhaps, extinction of taxa on the central islands.

Hybridization. The cpDNA phylogeny confirms earlier sug-gestions that hybridization has played an important role in theevolution ofArgyranthemum (42-44). The best example can befound among the seven subspecies of the A. adauctum com-plex. Two subspecies occur in the pine forest (A. adauctumsubspp. canariae and dugourii); one is restricted to the aridlowland scrub of southern Gran Canaria (A. adauctum subsp.gracile), and the remaining four subspecies (A. adauctumsubspp. adauctum, erythrocarpon, jacobifolium, and palmensis)are found in the heath belt area under the influence of thenortheastern trade winds. The cpDNA phylogeny revealed thatthe subspecies from the pine forest were part of clade A of theCanarian lineage (Fig. 2B), whereas those from the heath belt(except A. adauctum subsp. jacobifolium from Gran Canaria)grouped in clade B. The three subspecies of A. adauctum inGran Canaria represent a case of radiation into three of themajor ecological zones of this island. This is not the case ofA.adauctum in the heath belt of Tenerife and La Palma, perhapsbecause early colonizers from this complex could have becomesympatric with the endemics A. webbii in La Palma or A.broussonetii in Tenerife. This could have led to the opportunityfor hybridization and the consequent enhancement of thevariation and adaptation of A. adauctum. The two cpDNAgroups of the A. adauctum complex differ by a minimum of 21restriction site changes. It is unlikely that convergent restric-tion site changes or ancestral polymorphisms could account forthe separation of the subspecies of A. adauctum into the twodivergent clades. The coherence of the seven subspecies issupported by five unique morphological characters (22). Thisevidence, in combination with the known occurrence of hybridswarms inArgyranthemum (42,43), indicates that introgressionfrom other species from the area ilnfluenced by the trade windsis the most likely explanation for the paraphyly ofA. adauctumin the cpDNA tree.

Argyranthemum hierrense and A. haouarytheum are alsoparaphyletic in the cpDNA tree (Fig. 2A) but morphologically(22) they are coherent taxonomic entities. They could alsorepresent examples of ancient introgression and interspecificgene flow within the genus. In contrast, the paraphyly of thetwo subspecies of A. broussonetii is likely due to the presenceof two independent evolutionary lineages. In each case, theyare also distinct in their morphology (J.F.-O. and A.S.-G., un-published data), and each should be considered a distinct species.

Hybridization has generally been considered to play a minorrole in the evolution of insular plants (19, 45), even thoughthere is usually a lack of genic barriers to hybridizationbetween congeneric endemic species (46). In most cases,founder effect coupled with selection in different ecologicalzones has been regarded as the main evolutionary force involcanic archipelagos (46,47). However, hybridization has alsobeen implicated in the evolution of the silversword alliance,Bidens (Asteraceae), Scaevola (Goodeniaceae), and Pipturus(Urticaceae) in the Hawaiian Islands (48-50). Thus it seemsappropriate to suggest that interspecific hybridization has beena factor in the evolution of endemics in oceanic islands.

Sequence Divergence. The low levels of nucleotide sequencedivergence inArgyranthemum (Table 2) suggest that the genusoriginated and radiated recently. cpDNA sequence divergencevalues for oceanic island endemics and their mainland coun-terparts is only known for the Hawaiian silversword alliance(3). Differentiation of this Pacific group from the closestcontinental relatives is much higher than between Argyranthe-mum in Macaronesia and its mainland relatives (0.310-0.610in the silversword alliance/California tarweeds vs. 0.155-0.298in Argyranthemum/mainland species), even though vigorousintergeneric hybrids between the silversword alliance speciesand their mainland relatives (Madia bolanderi and Raillardi-opsis muirii from California) can be generated (3). Fertilehybrids between species ofArgyranthemum have been reported(42, 44, 51). We have also produced hybrids among variouscombinations of A. coronopifolium, A frutescens, and A. ma-derense. However, all attempted crosses between Argyranthe-mum and the two mainland relatives have failed to yield anyseeds (unpublished data). Thus, despite the fact that thechloroplast genomes of the Macaronesian genus and its main-land relatives are less divergent than those of the silverswordalliance and its two closest North American genera, they areapparently cross-incompatible.

In this study, we have demonstrated that despite its ecogeo-graphical peculiarities, the Macaronesian flora has experi-enced some of the same evolutionary phenomena as otheroceanic islands. A common pattern based on rapid morpho-logical differentiation accompanied by little genetic diver-gence emerges for the evolution of the flora ofvolcanic islands.

Evolution: Francisco-Ortega et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0

4090 Evolution: Francisco-Ortega et al.

Our results, however, also show that interisland colonizationbetween similar ecological zones may be one of the primaryfactors involved in the evolution of the endemic flora ofoceanic archipelagos. In Argyranthemum this is particularlyevident in the Canary Islands where the two cpDNA clades arecorrelated strongly with the two major climatic areas. Thesetwo lineages evolved in parallel by exploiting different eco-logical zones. Occasionally these two climatic clades inter-sected via hybridization-e.g., the seven subspecies of A.adauctum-which may have provided a new source of varia-tion. These hybrid genotypes may have provided new oppor-tunities for exploitation of the extremely diverse range ofhabitats in Macaronesia.

John Clement and K. E. Holsinger assisted with the maximumlikelihood analysis. Technical assistance was provided by A. Strongand A. Hines. We are grateful to G. J. Anderson, B. G. Baldwin, C.Ferguson, J. Fuertes-Aguilar, A. L. Hempel, D. Levin, L. A. Prather,J. F. Wendel, and the Systematics Discussion Group at the Universityof Texas at Austin for critically reading the manuscript. Support forthis research was provided by grants from the Ministerio de Educaci6ny Ciencia, Spain (PF92 42044506 to J.F.-O.) and from the NationalScience Foundation (DEB-9318279 to R.K.J.). The Instituto Canariode Investigaciones Agrarias, Canary Islands, through the advice of M.Fernandez-Galvan, provided a specific contract to J.F.-O. to conductthis study. The Instituto Nacional de Investigaci6n y TecnologiaAgraria y Alimentaria, Spain, provided financial support for germ-plasm collection.

1. Grant, P. R. (1986) Ecology and Evolution of Darwin's Finches(Princeton Univ. Press, Princeton, NJ).

2. Carson, H. L. & Kaneshiro, K. Y. (1976)Annu. Rev. Ecol. Syst. 7,311-345.

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

4. DeSalle, R. & Grimaldi, D. A. (1991) Annu. Rev. Ecol. Syst. 22,447-475.

5. Gillespie, R. G., Croom, H. B. & Palumbi, S. R. (1994) Proc.Natl. Acad. Sci. USA 91, 2290-2294.

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

7. Yang, S. Y. & Patton, J. L. (1981) Auk 98, 230-242.8. Wendel, J. F. & Percival, A. E. (1990) Plant Syst. Evol. 171,

99-115.9. Crawford, D. J., Stuessy, T. F., Cosner, M. B., Haines, D. W.,

Silva, M. & Baeza, M. (1992) Syst. Bot. 17, 676-682.10. Crawford, D. J., Stuessy, T. F., Cosner, M. B., Haines, D. W. &

Silva, M. (1993) Plant Syst. Evol. 184, 233-239.11. Sang, T., Crawford, D. J., Kim, S. C. & Stuessy, T. F. (1994)Am.

J. Bot. 81, 1494-1501.12. Lems, K. (1960) Ecology 41, 1-17.13. Thorpe, R. S., McGregor, D. P., Cumming, A. M. & Jordan,

W. C. (1994) Evolution 48, 230-240.14. Carracedo, J. C. (1994) J. Volcanol. Geotherm. Res. 60, 225-241.15. Bramwell, D. (1972) in Taxonomy, Phytogeography and Evolution,

ed. Valentine, D. H. (Academic, London), pp. 141-159.16. Galopim De Carvalho, A. M. & Brandao, J. M. (1991) Geologia

doArchipelago da Madeira (Museu Nacional de Hist6ria Natural,Mineralogia e Geologia da Universidade de Lisboa, Lisbon).

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

18. Fernandez-Galvan, M. (1983) in Comunicacoes II Congreso In-ternacional Flora Macaronesica, ed. Anonymous (Funchal, Ma-deira, Spain), pp. 269-293.

19. Humphries, C. J. (1979) in Plants and Islands, ed. Bramwell, D.(Academic, New York), pp. 171-199.

20. Hansen, A. & Sunding, P. (1993) Sommerfeltia 17, 1-295.21. Howard, R. A. (1973) in Vegetation and Vegetational History of

Northern Latin America, ed. Graham, A. (Elsevier, Amsterdam),pp. 1-38.

22. Humphries, C. J. (1976) Bull. Br. Mus. (Nat. Hist.) Bot. 5,147-240.

23. Bremer, K. & Humphries, C. J. (1993) Bull. Nat. Hist. London(Bot.) 23, 71-177.

24. Greger, H. (1977) in The Biology and Chemistry of the Compositae,eds. Heywood, V. H., Harborne, J. B. & Turner, B. L. (Academic,London), Vol. 1, pp. 899-941.

25. Christensen, L. P. (1992) Phytochemistry 31, 7-49.26. Fancisco-Ortega, J., Crawford, D. J., Santos-Guerra, A. &

Sa-Fontinha, S. (1995) Am. J. Bot. 82, 1321-1328.27. Francisco-Ortega, J., Jansen, R. K., Crawford, D. J. & Santos-

Guerra, A. (1995) Syst. Bot. 20, 413-422.28. DeJoode, D. R. & Wendel, J. F. (1992)Am. J. Bot. 79,1311-1319.29. Doyle, J. J. & Doyle, J. L. (1987) Phyt. Bull. 19, 11-15.30. Palmer, J. D. (1986) Methods Enzymol. 118, 167-186.31. Jansen, R. K. & Palmer, J. D. (1987) Curr. Genet. 11, 553-564.32. Farris, J. S. (1970) Syst. Zool. 19, 83-92.33. Albert, V. A., Mishler, B. D. & Chase, M. W. (1992) in Molecular

Systematics of Plants, eds. Soltis, P. S., Soltis, D. E. & Dolke, J. J.(Chapman & Hall, New York), pp. 369-403.

34. Swofford, D. L. (1993) PAUP (Illinois Nat. Hist. Survey, Cham-paign, IL).

35. Holsinger, K. E. & Jansen, R. K. (1993) Methods Enzymol. 224,439-455.

36. Felsenstein, J. (1985) Evolution 39, 783-791.37. Felsenstein, J. (1990) Cladistics 5, 164-166.38. Nei, M. & Li, W. H. (1979) Proc. Natl. Acad. Sci. USA 76,

5269-5273.39. Kishino, H. & Hasegawa, M. (1989) J. Mol. Evol. 29, 170-179.40. Baldwin, B. G. & Robichaux, R. H. (1995) in Hawaiian Biogeog-

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

41. MacArthur, R. H. & Wilson, E. 0. (1967) The Theory of IslandBiogeography (Princeton Univ. Press, Princeton, NJ).

42. Borgen, L. (1976) Norw. J. Bot. 23, 121-137.43. Brochmann, C. (1984) Nord. J. Bot. 4, 729-736.44. Brochmann, C. (1987) Nord. J. Bot. 7, 609-630.45. Ganders, F. R. & Nagata, K. M. (1984) in Plant Biosystematics,

ed. Grant, W. F. (Academic, Toronto), pp. 179-194.46. Crawford, D. J., Whitkus, R. & Stuessy, T. F. (1987) in Patterns

of Differentiation in Higher Plants, ed. Urbanska, K. (Academic,London), pp. 183-199.

47. Carson, H. L. & Templeton, A. R. (1984) Annu. Rev. Ecol. Syst.15, 311-346.

48. Baldwin, B. G., Kyhos, D. W. & Dvorak, J. (1990) Ann. Mo. Bot.Gard. 77, 96-109.

49. Gillet, G. W. (1972) in Taxonomy, Phytogeography and Evolution,ed. Valentine, D. H. (Academic, London), pp. 205-219.

50. Carr, G. D. (1995) Am. J. Bot. 79, 1574-1581.51. Humphries, C. J. (1975) Bot. Not. 128, 239-255.

Pro~c. Natl. Acad. Sci. USA 93 (1996)

Dow

nloa

ded

by g

uest

on

Aug

ust 5

, 202

0