Accepted Article 3 2 1 M. The casethe of widespread endemic as refugia andsources Paleo-islands of ge Article type : Original Article Accepted Date : 17-Jun-2015 Revised: 16-Jun-2015 Date Received: Date 05-Mar-2015 This article This protectedis by Allcopyright. rights reserved. 10.1111/mec.13282doi: lead to differences betweenversionthis and the Versionof Record. Please cite throughthebeen typesetting,copyediting, pa This article hasbeen accepted for publicationand undergone fullpeer review but has not † Tel: +34914203017; Fax: Running head: DIVERSIFICATIONWITHIN- IN VOLCANIC ISLANDS Keywords: ancestral areas, extinction, oceanic islands,paleo-islands, volcanic refugia. M. University College London, WC1E6BT London, United Kingdom Institut Botànic deBarcelona(IBB-CSIC-ICUB), 08038Barcelona, Spain Real 28014Madrid, JardínSpain Botánico (RJB-CSIC),

These senior authors contributed equally contributed authors senior These Correspondencetobe sent to: RealJardínBotánico (RJB-CSIC), 28014 Madrid, Spain;

A M LARCÓN AIRAL *1 2 , I. S †

ANMARTÍN +34914200157; e-mail: [email protected] 1 † , J.

J.

A canariensis LDASORO

netic volcanicdiversity within archipelagos: gination and process, which proofreading may 2 , V. CULSHAW (). 1 , I.

MANOLOPOULOU this article as

3 AND

Accepted Article diversification (Carson & Templeton 1984; Gille 1984; &Templeton (Carson diversification the of ecological role to examine adaptations driversgeographicversus isolationas of heterogeneity buffered and climates, oceanic islands represent excellent natural laboratories INTRODUCTION sources of genetic diversityto other areaswithin the archipelago. paleo-islandsTenerife, the with refugia acting as against extinction, andascradles and million years 3.5 islands played ago genetic rolesingenerating key withinboundaries a eruptions volcanic that argue We islands. of levels intra-island with variation, genetic structuring in role lesser contrast,clustering.tohavein appearplayed a Oceanic barriers, or islands us nearbythe within thewere also as ancestralinferred of migrant location areas towards alleles other disturbed areas these andRoquedelConde; of Anaga, Teno "paleo-islands" stable geologically in the divides Canarian populations into eastern andwe and axis as itshas central Tenerife that Ma 0.8 around divergence a first with structure, laurel the forestendemic of Canary astrong Islands. found geographicWe population in variation genetic (cpDNAchloroplast haplotypes) andthenucle diversification within-island of ingeological time. study This employed data thefrom studies haveeffectfew focused thatvolcanic the catastrophic have events on onpatterns had as some of the factorsmain driving diversification within volcanic archipelagos. However, This article This protectedis by Allcopyright. rights reserved. A BSTRACT Geographical isolation by barriers Geographical a oceanic Due to their small size, small geographic discrete substantial boundaries, to their environmentalDue

Canarina canariensis , an iconic species associated with the ing aBayesianapproach tophylogeographic nd landslides after the paleo- merging the the after nd landslides of genetic diversity larger than thosebetween- geneticdiversity largerthan ar (AFLPs) genomes to examine patterns of examine genomes to ar (AFLPs) spie 2004). Patterns ofspie Patterns are 2004). variation genetic stern stern Geneticclades. diversity wasgreatest nd climatic stability has been postulated

Accepted Article al. populations; andtheypopulations; ultimately drive differen al. landslides,eruptions, merging paleo-isla of is usually by followed a large numberofde systems (Gillespie & However,Claguethe development 2009). of birth and volcanic islands insular atpromotingin thantopographic generally isolation barriers effective are more inter-population differentiation in comparison incomparison differentiation inter-population is flora and the 2001) (Santos-Guerra endemics are diversity plant 40%ofCanarianfor &Quézel vascular 1997): (Médail approximately al. biodiversitycradles ofmultiple events in-situ where diversification have place taken (Juan continental lineages that have survived theclimatic changes of Latethe Cenozoic, but also as connected to the mainland. The Canary Islandshave long been considered refugia for al. pattern of chronological emergence due to a mantle plume (Carracedoplume mantle of chronological toa pattern due emergence aneasttowest the with years (Ma) in last islands formed 21 million The were 1855). Lyell This article This protectedis by Allcopyright. rights reserved. within-islands (Bottin than islands, hierarchical be with expected to in endemic biota have interested scientists since the scientists early19 have biota interested endemic 110km the from of north-westerncoast geological highly located history Their Africa. and (allopatric speciation, extirpations and recolonisations). temporal levels scales andwithin-islands)(between- over spatial different andalso different at diversification of patterns studying frameworkfor ideal an archipelagos represent volcanic 1990; Gillespie & Roderick 2002, 2014; 2002, 2014; Macías-Hernández &Roderick 1990;Gillespie 2011). These events habitat promote fragmentation; thesubsequent genetic isolation of 2000;Francisco-Ortega 2015) (Figure They 1a). areseparated by deep oceanic trenches and have never been The are a volcanic archipelago formed by a chain of seven islands, et al. et al. 2005; García-Verdugo 2005; 2000).In theseaddition, islands are regarded as ahotspot nds, etc (Carracedo 1994; Fernández-Palacios1994; (Carracedo nds, etc between-island genetic differentiation strongerdifferentiation between-island genetic tiation and speciation within-islands (Carson to archipelagos (Francisco-Ortega other structive events in the form ofsecondary generally characterised by generally characterised of high levels et al. th century 1814; Humboldt (von 2010), because2010), oceanic barriers et al. et et al. 2013). Therefore, 1998; Zaczek 1998; et al et et et et et . Accepted Article climate- orclimate- human-induced extinction (Harter mighthaveand variety micro-climates that of stability, the (volcanic) three paleo-islands of Tenerife topographic exhibit also complexity a sharethey severalendemicrestricted and species Table (see S1).Besidestheir geological preservedforests, laurel consideredancient,as an flora unique restricted to Macaronesia, and the CanarianArchipelagothe (Reyes-Betancourt highest phy the home to are presently – Pliocene the since stable geologically remained has also – which LaGomera with together Tenerife, of Trusty 1999; Guillou position inposition the lineage's phylogeny. position This arguedagainst these theidea of massifs as Tenerife continued to be active until 0.13 Ma (Ancochea (Ancochea Ma 0.13 tobeactivecontinued until Tenerife Fig. 1b). paleo-islands The remained thereafter relatively stable, whereas thecentral part of within-island diversification (Juan secondaryavalanches and manyof of eruptions 2008; Vitales these islandsthese 3.5 Ma and gave riseto volcanic central episodes Eruptive Ma) inthe northeast. Anagafused and (4.9-3.9 northwest, thein (6.2-5.6 Ma) southwest, Teno the Ma) in (11.9-8.9 RoqueMiocene: Conde Late del the dating asthreeseparate Itexisted at backto islands. islands, of all the first history complex diversification, particularly at particularly at diversification, This article This protectedis by Allcopyright. rights reserved. sources ofgenetic diversity(García-Verdugo de Paz&Caujapé-Castells2000; 2013). Recently, et al In reviewing the role of Tenerife paleo-islands as refugia across several plant lineages,plant several across refugia as paleo-islands Tenerife of role reviewing the In Most studies on Most the Canarian florahave focused onthepattern of inter-island . (2005) found that species endemic to the paleo-islands aderived occupied species tothe often endemic that (2005) found . et al. et al . 2004; Carracedo 2014; see Fig. 1b). Intesee2004; 2014; Fig.1b). Carracedo . 2014a, topographies b).Nevertheless, complex and the of long histories the species levelspecies (Francisco-Ortega the et al. the present (Ancochea island ofTenerife 2000; Brown et al et al. et al favoured their role as micro-refugia against micro-refugia as role their favoured logenetic diversity and endemic richness diversity andendemic of logenetic . 2015; Patiño these islands are likely to have favoured islandsto likely are these . 2015). 2008). These areas also harbour the best- the 2008).alsoThese areas harbour they have been proposed as reservoirs and et al et al et restingly, the three paleo-islandstherestingly, three . 2006). Tenerife has themost Tenerife has 2006). . . 1990, 1999; Cantagrel . 1990, et al . 2015). et al. et 2002; Kim 2002; et al . 1990; et al et al et . . Accepted Article This article This protectedis by Allcopyright. rights reserved. infra-species level (Juan divergence events between taxa these to endemic paleo-islands,thespecies either at the or at refugia.ancient However studies, other especi 2013; Puppo a wide, continental-scale disjunction of 7000 km spanning across the Sahara. Mairal Sahara. the 7000 kmspanning across of continental-scale disjunction wide, a precursor paleo-islands 3.5 million years paleo-islands million (Ancochea ago precursor 3.5 (Gómez endemics" or endemics" areespeciallyMIEs) relevant geological island that Species the history. presentare island inmultipleislands ("multiple of lineageestimates patterns which to times, divergence necessary within-island is torelate variation Canariangenetic for widespread endemics, and noneof them have provided climbs onnearbyclimbs pollinated by itis plants; 2015). In addition to butbetween-islands. within- areasas reservoi thathave acted undisturbed Afromontane forests of Eastern Africa, Eastern of forests Afromontane 2008). Genus Campanulaceae (Olesen 2011) and its fleshy2011) andits dispersed are (Valido fruits byvertebrates Gomera, and . and ElHierro. La Palma Gomera, Tenerife, Canaria, Gran Islands: Canary western and the central in occurs presently It in cleared that growsmostly Canary Islands, elected asits “national multiple island of endemic, a Here, we study patterns geneticof diversit et al. et al 2003;García-Verdugo Canarina . 2014;) that are contemporaneous orpredate the age merging of the of C. canariensis et al et al belongs to tribe Platycodoneae, a basal group family basal within a Platycodoneae, to tribe belongs Canarina canariensisCanarina . 2012; Mansion . 1996, 2000; Dimitrov 2000; 1996, . C. canariensis , the, genus comprises twoother species inhabiting the et al. C.eminii flower” (Kunkel 1991),is a (2n= diploid 34) generalist birds (Rodríg birds generalist 2010) have onpatternsfocused within-island of areas surrounding the endemic laurisilva forest. laurisilva endemic the surrounding areas rs and sources of genetic diversity not only of genetic only diversityrs andsources not et al. to understand the role understand the to and ally in animals, haveagesanimals, ally in reported for y demographic and spatialthe and history is a herbaceous plant that occasionallyplant that a herbaceous is (L.) Vatke. This "flagship" species of the the of species "flagship" This Vatke. (L.) 2012;Wang C. abyssinica C. et al et alet . 2008;Macías-Hernández et al. . 1990). Few plant studies , beingan of this example , et al. uez-Rodríguez Valido & 2003, Rodríguez 2003, 2013; Mairal of paleo-islands as et al. et al. et al et al. La .

Accepted Article distribution of genetic within variation distribution thegeographic history Our geological thearchipelago. to:aims i)determine of were main tocandidate evaluate patterns of within-island diversification in relation tothe recent islands ofislands Tenerife as refugiaof genetic diversity, relictualboth andrecent. paleo- the of putativeiv) inter-island role the events, migration reconstruct examine and vicariance within Tenerife (Mairal (Mairal Tenerife within vicariance Pleistocene) Mainvolved east-west (Late and only isdated 1 divergence an at around population event of population: the earliest African North closer and geographically canariensis distribution.widespread of colonization the The Canary Islandsby theancestors of climate-drivenvicariance and resultingfragmentation the extinction in ofanancient spatial disjunctionwas and resultofThis extraordinary explained temporal asthe Ma). This article This protectedis by Allcopyright. rights reserved. the phylogeny(2015) recently reconstructed spatiotemporal evolution of and and diversification processes that may be related togeological events related be may that processes diversification and unobserved events such as local population grow evaluation of alternative phylogeographic scenarios of alternativeevaluation phylogeographic (Bloomquist to associated uncertainty network inference, the mutation of differingevents among haplotypes. et al et Templeton (SP, Parsimony Statistical using inferred present.are often is These networks flow gene especially tree, branching when reticulatea than between pools pictureclearer relationships genetic of the a provide inferred that that inferred . 2000),which allows estimationthehapl of Haplotype networks areHaplotype networks commonly usedin This age and the presencethe This age and of apparently much later, occurred in the Pleistocene, probablya nowextinct from C. canariensis divergeditsAfricanrelatives from ofMiocene the the (c.at end 7 et al Canarina canariensis et al . 2015). Canarina canariensis . 1992) implemented in the software TCS (Clement TCS software the in 1992) implemented . However, this fails method theto incorporate otype whilenetwork minimising thenumber and therefore does not allow for statistical th or past extinction of haplotypes may population-level studiesbecause they in several islands makes it an ideal an it makes islands several in , ii) find evidence of evidence extinction , ii) find et al. , iii) find ancestral areas iii)findancestral 2010). Moreover, 2010). Canarina . They C. Accepted Article 2015). We 432generated sequences: new This article This protectedis by Allcopyright. rights reserved. Manolopoulou and Ecological Clustering (BPEC, those obtained amodel-based, Bayesianfrom statistical method, Bayesian Phylogeographic inference in parsimony-basedmislead methods Population sampling and DNA extraction M useto this formethod island phylogeography. study first this the is our knowledge, To 2012). & Emerson (Manolopoulou model mutation tree haplotype for probabilities posterior (144 sequences) and have proven to be useful for intraspecific analyses intraspecific usefulto be proven population for have structure of (Mairal Chloroplast DNA sequencing expeditions. the field leaves 20-25mg ofsilica-gelfrom dried obtaine was gel extracted silica USA), California, Inc., (QIAGEN Kit Mini Plant DNeasy using the whole occupancythe areaof each population. sampling (Caujapé-Castells 2010),samples were collected individualsfrom scattered across biased inflationin geneto descriptors due To ofsamples 10 reduce perpopulation. minimum 2012: 4 in 2 inLG,GC, 8 and in 2 TF, LP possible,1 inEH.Where in we collected a Seventeen populations of (LG), La with Gomera (LP) Palma La ElHierroin (EH). (TF), and comparison Tenerife ATERIALS AND Canarina canariensis We selectedWe three cpDNA intergenic spacers regions for sequencing; markersthese M ETHODS pet B C. canariensis

1365 has a significantly greater onGran Canariapresence has (GC)and asignificantly greater – pet D 738 (144sequences). PCR and sequencingprotocols were sampled in several fieldtrips between 2009 and several fieldtrips between sampled in were rpl networks underacoalescent-based migration- 32- DNA from 160 individuals and preserved in preservedindividuals 160 and DNA from . . Here, we compare results TCS from with d from the fresh plant trn et al. et L UAG 2011), allowswhich 2011), estimating the

(144 sequences), (144 sequences), tissue collectedfrom trn S GCU – trn et al G UCC .

Accepted Article Sequences for each region for were each Sequences analysesHaplotype accession numbers and full references are detailedinTable S2. statistical parsimony algorithm (Templeton parsimony (Templeton statistical algorithm sequenced regions. Genealogical relationships thethree andexamining populations different analysis, by individuals from using 6-12 each population usingDnaSP (version data. haplotypes was calculated 95%confidence with al.et was used to determine the best fitting model of sequence evolutio sequence of model fitting best the determine to used was toperform (Nylander, MrModeltest 2004) phylogenetic software v.2.2 partitions analyses. describedrules inKelchner We(2000). analysed the three sequenced regions as three data Sequences were checked andmanually adjustedwhere necessary by following alignment Pro5.4.softwarethe Geneious in implemented This article This protectedis by Allcopyright. rights reserved. Mairal of those followed parsimony in order to reduce the number of candidate trees to a manageable set. The BPEC set. amanageable candidate to trees of number the toreduce in order parsimony population clusters distinct geographical ancestral locations.and Like on TCS, BPECrelies toidentify prep.) genetically Hille, in inthe(Manolopoulou R package & implemented number of haplotypes H(n), H(n), haplotypes of number diversity nucleotide diversity haplotype (Hd), heterozygosity heterozygosity 2000). number The ofmutational steps resulting from single substitutions among The relationshipsThe amongst were lineages investigated through haplotype network The BPEC method (Manolopoulou BPEC The method Summary statistics for within-population gene within-population for statistics Summary θ , G ST and thenumber of migrants generation per (Nm) were estimated for et al . 2015.. The sources of the material examined, the GenBank 5.10;Librado & Rozas 2009). usingaligned MAFFT 6.814b(Katoh et al. et al. 4. Ltd., (Biomatters Auckland, New Zealand). viatheamonghaplotypes were inferred 2011; Manolopoulou &Emerson 2012) was limits, gaps were represented as as represented missing were gaps limits, 1992) implemented in TCS 1.21 (Clement inTCS 1.21 1992) implemented tic as: diversity werecalculated the n of each data partition. partition. data each n of π et al et , nucleotide . 2002), Accepted Article details see details see into exploration, wedividedthe dataset further upon ahaplotype tree, and used to ancestral estimate locations migration for events. As a phyloge The 5–cluster model. converging a to migrations onthehaving anyabove of and similarly, tree, inferred number 4 noeffect runwere chains MCMC ds>3 with stable, were resultsThe iterations. million three for of sequences of distance ds}{1,…, nucleotides will consideredbe as the “missing Twopath”. missing anunobservedhave intermediate seque isincreased, value is parsimonythe assumption ds As time. the greater computation much modelsbutrequire more include values Larger of migration 1). + hence(MaxMig ofevents number clusters number and maximum the thefor bound upper set the to user the allows MaxMig the software). as in (denoted “ds” software) and parameterthe parsimony relaxation the in (denoted as migrations number “MaxMig” of maximum inputs: user-defined the main assumes the that migration andthe popula rate method The corresponding clusters. events and migration of number trees, probability haplotype) migrates to a new geographical cluster. MCMC simultaneously estimates high occurwere assumed to parent (with orwithout amutation its whenahaplotype from This article This protectedis by Allcopyright. rights reserved. by Sanmartín explored throughare Markov Monte chain Ca to theled observed population substructure. Different scenarios of trees and migration events (Knowles 2008). Each possible defines settreepossible a that of m eventsmigration uncertainty inhaplotypefor which relationships, one is of main the criticisms of TCS fu in a withhigh posterior probability trees unlikemethod, Standard Parsimony, afits prior over all possible trees in orderto identify coalescent dating et al . (2008) for estimating rates of inter-island dispersal.events Migration inter-island rates (2008) of estimating for . section in results). Haplotypes sampled from Roque del results). Conde in Haplotypes Roquedel sampled from section lly framework,model-based thus accommodating ographic clustering obtained was superimposed was clustering obtained ographic rlo (MCMC), similar to the method proposed the method to similar rlo (MCMC), two groups (eastern and western cladeswestern twogroups –for (eastern and tion constant. growthare two BPEC requires nce (an unobserved mutation), then pair any (anunobservednce mutation), relaxed: if twoobserved sets ofsequences used to reconstruct th to used reconstruct e set of possibletreesof set e ay haveay Accepted Article This article This protectedis by Allcopyright. rights reserved. the for distribution accommodate the in change ratemutation from species to populations, with uniforma and(eastern western groups with without and Roque del Conde). the two locationfrom this groups. ran BPEC We analysis ofon each thesefour datasets separate analyses divergent; thus, were quite reference to these settings. Two MCMC chains were run for 50 million generations, sampling generations, 50 million run for were chains Two MCMC these settings. to reference to analyses from exploratory see results S3 to Factor comparisons usingpath the ucld.stdev constant populationconstant tree usingprior, a secondary (Mairal age estimate a Rambaut We 2007). out first carried analysis “ thenode of root the to 0.327 Ma)calibrate (SD)= deviation standard Ma, =0.76 mean prior: canariensis ( dataset population-level linked a of rate clock the wasused inform outgroup to taxa second analysis applyingthe“nested da presencepoint.(see and Results) calibration single a level of population out aWe carried age posterior estimates,ESS for due probably at the content information low to the gaveusanalysis very large95%HPD Density) (High credibility intervals Posterior and poor (Mairal (Mairal datasetpopulation-level stochasticbirth- and a al higher-level wasdataset calibrated with fossil-derived secondary age estimates (see Mairal which a higher-level dataset including dataset threespecieswhich ahigher-level representatives of of all C. canariensis C. . 2015), while. 2015), the tree was unlinked toprior apply acoalescent constant sizemodelthe to Haplotype divergence times were were es times Haplotype divergence et al ; the substitution; the model was set GTR+G; to choice wasbased ofpriors onBayes ) under a ) under a . 2015). dataset”; this included all haplotypes detected in our sampling (N = 10). This This 10). = sampling (N haplotypesin our all dataset”; detected included this mixed model Yule-coalescent ucld.mean

The clock priorto log-normal settoanuncorrelated The clock was model (10 sampling method in BEAST (Baele sampling method BEAST in -4 -10 ting approach” described in ting in described Mairal approach” -1 ) and a default distribution) andadefault theexponential for were run in which we included andexcluded were runinwhichweincluded assess the reliability of our date estimates with ourdateestimates the reliability of assess timated intimated and(Drummond BEAST v.1.7 under a strict clock model underastrictclock anda coalescent death (Yule) prior to the species-level one (Ho (Ho et al et . 2005; Pokorny 2005; . et al. et al. 2012); see Table et al et al Canarina 2015; normal 2015; . 2011).The . . (2015),in . and 9 C. et