© 2018 Naturalis Biodiversity Center

You are free to share - to copy, distribute and transmit the work, under the following conditions:Attribution: Youmustattributetheworkinthemannerspecifiedbytheauthororlicensor(butnotinanywaythatsuggeststhattheyendorseyouoryouruseofthework).Non-commercial: Youmaynotusethisworkforcommercialpurposes.Noderivativeworks: Youmaynotalter,transform,orbuilduponthiswork.Foranyreuseordistribution,youmustmakecleartoothersthelicensetermsofthiswork,whichcanbefoundathttp://creativecommons.org/licenses/by-nc-nd/3.0/legalcode.Anyoftheaboveconditionscanbewaivedifyougetpermissionfromthecopyrightholder.Nothinginthislicenseimpairsorrestrictstheauthor’smoralrights.

Blumea63,2018:93–101 ISSN(Online)2212-1676www.ingentaconnect.com/content/nhn/blumea https://doi.org/10.3767/blumea.2018.63.02.01RESEARCH ARTICLE

INTRODUCTION

RafflesiaR.Br.(Rafflesiaceae)isaSoutheastAsianparasiticplantgenusofc.30species(Nickrent1997onwards).MostoftheseareonlyknownfromindividualislandsintheMalesianarchipelago(Nais2001,Barcelonaetal.2009b,Bendiksbyetal.2010,Davidetal.2011).Thispatternofislandendemismisperhaps most pronounced in the Philippines, where 12 of the 13currentlyrecognisedspecies(Pelseretal.2011onwards,Galindonetal.2016,Hidayati&Walck2016)areonlyfoundonasingle island(Pelseretal.2017).Withintheseislands,several species have very narrow distributions and two are onlyknownfromasinglepopulation(i.e.,R. aurantia Barce-lonaetal.,R. manillanaTeschem.,Barcelonaetal.2009a,b, Pelseretal.2017).OtherPhilippineRafflesia species are more widespread within an island, but have relatively few and iso-latedpopulations.Forexample,eightspeciesareonlyknownfromtwoorthreepopulations(i.e.,R. baleteiBarcelonaetal., R. consueloaeGalindonetal.,R. leonardiBarcelona&Pel-ser, R. mixta Barcelonaetal.,R. verrucosaBaleteetal.,and R. schadenbergianaGöpp.exHieron.;Baleteetal.2010,Bar-celonaetal.2006,2008,2009b,2011,2014,Galindonetal.2016).TherarityofRafflesia species and their disjunct distribu-tionmakesthemparticularlyvulnerabletoextinctionasaresultof the on-going destruction and degradation of their habitat (Nais2001,Barcelonaetal.2009b).Theconversionofrain-

forests for agriculture, mining, and other uses has undoubtedly resultedinlocalextinction,areductionofthesizeofremainingRafflesia populations, and lower genetic connectivity among them(Pelseretal.2017).Theobligateendo-holoparasiticlifestyleoftheseplantsandtheirpronouncedhostspecificityarefactorsthatfurtherincreaseextinctionrisk,becausetheycan-not live in the absence of the few Tetrastigma(Miq.)Planch.(Vitaceae)speciesthatarewithintheirhostrange(Pelseretal.2016).Informationaboutthehostspecificity,habitatpreferences,andreproductive biology of Rafflesia species can assist their con- servation management by revealing the environmental con-ditions that the populations need to remain viable, thereby providing a better understanding of the factors that explaintheircurrentdistributionpatterns.TheresultsofarecentstudyofhostspecificityandhostpreferenceofPhilippineRafflesia species (Pelseretal.2016) indicate thatall sixknownhostspecieshavewidespreaddistributionswithinthePhilippines.Tetrastigma host plants are also more common and widespread than their parasites in other parts of the distribution area of Rafflesia(Nais2001).Thissuggeststhathostrangesarenota limiting factor in the distribution of Rafflesia, although local differencesinhostpresenceandabundancemightexplainitsdistributionpatternsatthepopulationlevel(Pelseretal.2016).Althoughdetailedresearchintothehabitatrequirementsofindi-vidual Rafflesia species has, to our knowledge, not been carried out,itappearsthattheyarenotveryhabitatspecific.Membersofthegenusareconfinedtotropicalrainforestecosystems,butgenerally grow in primary as well as secondary forests, have wide elevational ranges, and their hosts are found on various substrates(Nais2001,Barcelonaetal.2009b,2011).Habitat

A conservation genetic study of Rafflesia speciosa (Rafflesiaceae): patterns of genetic diversity and differentiation within and between islandsP.B.Pelser1,D.L.Nickrent 2,J.F.Barcelona1

1 SchoolofBiologicalSciences,UniversityofCanterbury,PrivateBag4800,Christchurch8140,NewZealand;

correspondingauthore-mail:pieter.pelser@canterbury.ac.nz.2 DepartmentofPlantBiology,SouthernIllinoisUniversityCarbondale,1125LincolnDrive,Carbondale,Illinois62901-6509,USA.

Key words

conservationmicrosatellitesNegrosPanayPhilippinespopulation genetics

Abstract Rafflesia speciosaisathreatenedendo-holoparasiticspecies.IthasseveralpopulationsintheCentralPanayMountainRange(CPMR)ofPanayislandandasinglepopulationonNegrosIsland.BecauseR. speciosa is the only Philippine species of the genus that is not endemic to an individual island, it is a suitable species for improving our understanding of the factors underlying the high island endemism of Philippine Rafflesia.Forthispurpose and to inform the conservation management of R. speciosa, patterns of genetic diversity and differentia-tionwerestudiedusing15microsatellitelociandsamplesfromninepopulations.Noneofthesepopulationsshowsevidence of inbreeding and R. speciosahassimilarlevelsofheterozygosityasgenerallyobservedinoutcrossingorperennialplantspecies.TheresultsofAMOVAandBayesianclusteranalysesindicatethattheNegrospopulationisgeneticallydifferentiatedfromtheCPMRpopulations.Inaddition,ithaslowergeneticdiversitythansimilar-sizedR. speciosapopulations.Thesefindingssuggestthatseastraitspotentiallyprovidesignificantreproductivebarriersto Rafflesiaspecies,andareperhapsresponsiblefortheirhighislandendemism.Thegenerallackofgeneticdif-ferentiationamongtheCPMRpopulationsassuggestedbytheAMOVA,PCoA,andSTRUCTUREresultsindicatesrecentgeneflowamongthemandthisfindingimprovesourunderstandingofthegeographicalscaleandcontextatwhich gene flow between Rafflesiapopulationsoccurs.ConservationeffortsshouldbetargetedtowardsavoidingfurtherhabitatdegradationintheNegrospopulation.WealsorecommendprotectivestatusfortheentireCPMRandreforestationeffortstomitigatetheseverehabitatfragmentation,destruction,anddegradationinthisarea.

Published on13July2018

94 Blumea–Volume63/2,2018

preferenceisthereforeanunlikelyexplanationforthenarrowand disjunct distribution patterns of Rafflesia species and their highislandendemism.Instead, it is perhaps more likely that these distribution patterns are a result of poor seed dispersal beyond relatively small dis-tances(Barkmanetal.2017)andtheverylowprobabilitythatRafflesiaseedsaredepositedinsufficientproximityofasuitablehostplanttoresultininfection.Verylittleisknownaboutseeddispersal in Rafflesia.Themorphologyoftheirseedssuggeststhepresenceofanelaiosome(Kuijt1969,Pelseretal.2013),indicatingmyrmecochory, andPelser et al. (2013, presentstudy) indeedobservedants transporting seedsaway fromRafflesiafruits(Fig.1).Ifantsaretheprimaryseeddispersers,dispersal over large distances, particularly between islands, mightbevery infrequent.However,evidenceofmammalianfrugivory(Emmonsetal.1991,Bänziger2004)suggeststhatantsmightnotbetheexclusivedispersersofRafflesia seeds and that seeds might be able to be more readily transported overgreaterdistances(Pelseretal.2017).Populationgeneticdatacanprovideusefulcontextforunder-standing the factors that have shaped the distribution patterns of Rafflesia species, as well as for more directly informing conservation management by identifying populations that are ofconservationinterest,suchasthoseshowinguniquegeneticvariation(i.e.,privatealleles),evidenceoflowgeneticdiversityand inbreeding, or poor genetic connectivity with other popu-lations(Ellstrand&Elam1993,Frankham2005).Populationgenetic analyses of a microsatellite data set previously revealed low genetic connectivity between populations of the Philippine R. lagascaespeciescomplexthatconsistsofR. lagascaes.str.,MtLaboR. lagascae, and R. manillana(Pelseretal.2017).Furthermore, these analyses indicated that, despite their very close morphological similarity, the allopatric R. lagascae s.str.(LuzonIsland)andR. manillana(SamarIsland)arereproduc-tivelyisolated.Inaddition,ageneticallydistinctbutmorphologi-cally indistinct population of R. lagascaewasidentifiedfromMtLabointheBicolPeninsulaofLuzon,whichmightrepresentanundescribedcrypticspecies.Thesefindingssuggestlimitedeffective dispersal of pollen and seeds over larger distances and particularlybetweenislands.Limitedseeddispersalwasalsoconcluded for two RafflesiaspeciesfromPeninsularMalaysiaandThailand(R. cantleyiSolms)andBorneo(R. tuan-mudae Becc.)inaseparatestudy(Barkmanetal.2017).Thegeneticpatterns obtained for the R. lagascaespeciescomplexsuggestthat such limitationsmight subsequently result in allopatricspeciation,potentiallyexplainingthehighislandendemismof

Rafflesia.Thediscoveryofstronggeneticdifferentiationbe-tween morphologically similar populations of the R. lagascae complexondifferentislandsalsoraisesthequestionofwhetherother currently recognised Rafflesia species with populations on different islands need to be more narrowly circumscribed (Pelseretal.2017).Rafflesia speciosaBarcelona&Fernando(2002)issuchaspeciesandthereforeasuitablecandidateforfurtherexploringpatternsofgeneticdiversityofthisgenusinageographicalcontext.Rafflesia speciosa is the only Philippine species in the genus thatisnotendemictoasingleislandwithinthearchipelago.Itisfoundatc.350–1100ma.s.l.intheCentralPanayMountainRange(CPMR)ofPanayIslandandonMtKanlaoninnorthernNegrosIsland(Map1).TheforeststhatarehometoR. spe-ciosa are rapidly being fragmented and degraded as a result of landconversionforcattlegrazingandotherfarmingpractices,roadconstruction,andhousing(Fig.2).Athigherelevations,recreational housing, agritourism and ecotourism have recently decimatedfragileultramaficandlimestonehabitatssuchthat

Fig. 1TissueofopenedmaturefruitofRafflesia speciosa(Fig.2d)showingananttransportingaseed.Verticalarrowindicatesaseed.Horizontalarrowindicatesaseedthatisbeingmovedbyanant.—Scalebars=1mm.

Map 1DistributionofallcurrentlyknownpopulationsofRafflesia speciosa (underlined).Faintdotted lines indicateprovincialboundaries.Blackdotsshowpopulationsthatweresampledforthisstudy.GreydotsindicatetwomunicipalitieswithunconfirmedreportsofR. speciosa.ThegreyareashowstheareaoftheCentralPanayMountainRange(CPMR).ThedarkergreyareaindicatesSibalomNaturalPark.

95P.B.Pelseretal.:ConservationgeneticstudyofRafflesia speciosa

Fig. 2a–d.Rafflesia speciosa.a.Flower;b.flowerinhabitat;c.flowerbud;d.openedfruitshowingsmallorange-brownseeds.—e–h.Threatstotheconservation of R. speciosahabitat.e.Mammal trap (MtKanlaonNaturalPark); f.conversion intoagricultural land (barangayAningalan,SanRemigio); g.slash-and-burnpracticesinValderrama(smokeoffiresinthebackground,burnthillslopeintheforeground)leavesmallRafflesia-inhabited forest remnants amidstunproductivegrassland;h.alandslidefollowingdeforestation(Valderrama).

c

b

d

a

g h

e f

96 Blumea–Volume63/2,2018

now only small forest fragments remain.These and otheractivities have also been reported to negatively impact other species with which R. speciosasharesitshabitat(e.g.,Oliver2014,Oliveretal.1991,1993,Hamannetal.1999,Ferneretal.2000,Gaulke2010,Mould2012,Sammleretal.2012),there- bypromptingappealstodeclaretheCPMRaprotectedarea(e.g.,Oliveretal.1991,1993).WhileSibalomNaturalPark(Panay) andMtKanlaonenjoy legal protection, even theseareas are under the constant pressure of, among others, illegal firewoodcollecting,charcoalmaking,andhunting(Fig.2).IfR. speciosa shows similar patterns of low genetic connectivity as was observed in the R. lagascaecomplex,thisspeciesispotentially at risk of losing genetic diversity through inbreeding and genetic drift, and might be in need of urgent conservation management.Theobjectiveof thisstudy is toprovide thefirst informationabout the patterns of genetic connectivity and diversity of R. speciosawiththeprimaryaimsof1)improvingourunder-standing of the factors contributing to the narrow distributions and high island endemism of Rafflesiaspecies;and2)informingthe conservation management of R. speciosa.Forthis,weusepopulation genetic analyses of a microsatellite data set in which specimensfromninepopulationsareincluded.

MATERIALS & METHODS

Specimen samplingRafflesia speciosaisknownfromnineareas(Map1).OneoftheseisatthefoothillsofMtKanlaoninNegrosandtheothereight are found in theCPMR inPanay.Threeof thesearelocatedinthemunicipalitiesofIgbaras,Leon,andMiag-aoinIloiloprovinceattheeasternsideoftheRange.TheremainingfiveareasareinAntiqueProvinceatthewesternsideofthedivide.WithinthewiderSibalomNaturalParkarea,R. speciosa sites are concentrated in two areas that are separated by the TipuluanRiver.WehererefertotheareanorthoftheriverastheMtPoras-Aningalanarea,whichspansthebarangaysofCabladan and Imparayan of Sibalom municipality and baran-gayAningalanofSanRemigio.TheareasouthoftheTipuluanRiver is part of barangays Bad-as and Bululacao of Sibalom and referred toas thesouthernSibalomNaturalParkarea.Rafflesia speciosa is further known from the municipalities of Barbaza,Culasi,andValderrama.Anecdotalevidencesuggeststhat R. speciosaisalsopresentinLaua-anandTibiao(AntiqueProvince),butthesereportsremaintobeconfirmed.Forthepurposeofconvenience,becauseofdifficultiesincircumscrib-

ing the actual limits of populations, we refer to these nine areas as individual R. speciosapopulations.Following the sampling strategy described byPelser et al.(2017),tissuesamplesonsilicagelwerecollectedfromsmallparts of two to 16 flower buds or flowers from one or more sitesfromeachoftheninepopulations(Table2).Thenumberof samples per population is relatively low, because Rafflesia populations are often small and are sometimes even associ-atedwithjustasinglehostplant(Pelseretal.2017).However,because tissue samples were taken from all or nearly all host plantsthatwereidentifiedinthesepopulations,thepercentageofindividualssampledineachpopulationishigh.Regardless,ourestimatesofallelefrequenciesforsmallpopulationsshouldbe interpretedwith care.Voucher specimens (Table 1) aredepositedatCANU,PNH,andSIU.

DNA extraction and microsatellite genotypingUsingtheDNeasyplantminikit(Qiagen,Germantown,Mary-land),totalgenomicDNAwasextractedfromtissueofatotalof105R. speciosaflowersorflowerbuds.Theresultsofapilotstudyinwhichtheamplificationof17microsatellitemarkersde-veloped for R. lagascae (Pelseretal.2017)wastested(resultsnotreportedhere)showedthat15oftheselociamplifiedwellfor R. speciosa(Man78,Man80,Man109,Man111,Man120,Man142,Man144,Man166,Man171,Man273,Man553,Man714,Man866,Man1134,Man1169), but that the lociMan788 andMan1193 frequently failed to yield unambigu-ousgenotypingprofiles.These15loci(seePelseretal.2017

Population Voucher(herbarium)1

Culasi,AntiqueProv. Barcelona 3801(CANU,PNH)Barbaza,AntiqueProv. Barcelona 3791(CANU,PNH)Valderrama,AntiqueProv. Barcelona 3769 with Pelser(CANU,PNH)MtPoras-Aningalan,Sibalom& Barcelona 3701 with Pelser (SIU)SanRemigio,AntiqueProv.Southern Sibalom Natural Park, Barcelona 3990 with Pelser (PNH)Sibalom,AntiqueProvLeon,IloiloProv. NovoucherIgbaras,IloiloProv. Barcelona 4154 with Pelser(PNH)Miag-ao,IloiloProv. Barcelona 4133 with Pelser(PNH)MtKanlaon,Negros Barcelona 4013 with Pelser(PNH)OccidentalProv.1 CANU=UniversityofCanterburyHerbarium; PNH=PhilippineNationalHerbarium; SIU=SouthernIllinoisUniversityHerbarium.

Table 1VoucherinformationforRafflesia speciosapopulationssampled.

Population Protected # Rafflesia P Na(SE) Ne(SE) Na(p) Ho(SE) He(SE) uHe(SE) FIS(SE)(#sites) area samples (#hosts)

Culasi(5) Noprotectivestatus 7(3) 93% 4.13(0.49) 3.18(0.35) 0 0.65(0.08) 0.60(0.06) 0.65(0.07) -0.06(0.05)Barbaza(4) Noprotectivestatus 9(6) 93% 3.93(0.44) 3.07(0.33) 0 0.62(0.08) 0.59(0.06) 0.63(0.06) -0.01(0.05)Valderrama(10) Noprotectivestatus 16(16) 100% 5.67(0.57) 3.15(0.26) 5 0.58(0.06) 0.63(0.05) 0.65(0.05) 0.06(0.06)*MtPoras-Aningalan(6) SibalomNaturalPark 14(9) 93% 5.47(0.58) 3.38(0.45) 6 0.63(0.07) 0.61(0.06) 0.63(0.07) -0.03(0.04) pro parteSouthernSibalom SibalomNaturalPark 8(5) 100% 4.27(0.43) 3.00(0.28) 3 0.69(0.06) 0.62(0.04) 0.66(0.04) -0.12(0.06)NaturalPark(5)Leon(1) Noprotectivestatus 2(2) 87% 2.27(0.21) 2.10(0.20) 0 0.60(0.11) 0.46(0.06) 0.63(0.08) -0.30(0.17)Igbaras(6) Noprotectivestatus 9(9) 93% 4.40(0.63) 2.97(0.49) 2 0.58(0.08) 0.55(0.07) 0.58(0.07) -0.05(0.06)Miag-ao(5) Noprotectivestatus 12(11) 100% 4.60(0.43) 3.07(0.33) 3 0.58(0.06) 0.61(0.05) 0.64(0.05) 0.04(0.05)MtKanlaon(4) MtKanlaonNaturalPark 10(5) 87% 4.33(0.47) 2.75(0.28) 6 0.60(0.09) 0.55(0.07) 0.58(0.07) -0.07(0.07)

Totalsandmeans 87(66) 94% 4.34(0.18) 2.96(0.11) 3.57 0.61(0.03) 0.58(0.02) 0.63(0.02) -0.06(0.03)

Table 2Geneticdiversityandnumberofprivateallelesobservedat15microsatellitelociforninepopulationsofRafflesia speciosa.NumberofRafflesia samples(numberofTetrastigma host plants from which Rafflesiasampleswereobtained),percentageofpolymorphicloci(P),allelicrichness(Na),numberofeffectivealleles(Ne),numberofprivatealleles(Na(p)),observedheterozygosity(Ho),expectedheterozygosity(He),unbiasedexpectedheterozygosity(uHe),andinbreedingcoefficient(FIS).SE=StandardError.*PopulationwithasignificantdeviationfromHardy-WeinbergproportionsduetoheterozygotedeficitsafterB-Ycorrection(Narum2006).

97P.B.Pelseretal.:ConservationgeneticstudyofRafflesia speciosa

forprimerandmicrosatellite information)weresubsequentlygenotyped for all available R. speciosasamples.MultiplexPCRanalysesusingtheQiagenType-itmicrosatellitePCRkitanduptofourprimercombinationsperPCRsamplefollowedanM13protocolasdescribedinPelseretal.(2017).ThePCRproductswererunonanABI3130xLGeneticAnalyzerattheUniversityofCanterbury.Geneious6.1.7(BiomattersLtd,Auckland,NewZealand)wasusedtodeterminefragmentlengths.

Data analysesTestsforthepresenceofnullalleleswereperformedwithMicro-checkerv.2.2.3(VanOosterhoutetal.2004;95%confidenceinterval;1000repetitions).Onlylocithatwereflaggedasdis-playing null alleles across all populations were considered to be truly producing null alleles after a re-inspection of the associated genotypingelectropherograms.GENEPOPv.4.2(Raymond&Rousset1995)asimplementedinGenePopontheweb(http://genepop.curtin.edu.au/)wasusedforexacttestsfordeviationsfromHardy-Weinbergequilibrium.SignificancelevelsforthiswereadjustedformultipletestsusingtheB-YmethodFDRatp=0.05(Narum2006).Thepercentageofpolymorphicloci,allelicrichness(meannum-berofallelesandnumberofeffectivealleles),observed(Ho)and expected (He) heterozygosity, unbiased expected heterozy-gosity (uHe), and the inbreeding coefficientwere calculatedinGenAlExv.6.5(Peakall&Smouse2012)todeterminethegenetic diversity of each R. speciosapopulation.GenAlExwasalso used to assess patterns of genetic differentiation among these populations by determining the number of private alleles of each population and by performing an analysis of molecular variance (AMOVA; FST, 999 permutations). Furthermore, aPrincipalCoordinateAnalysis(PCoA)wascarriedoutinGen-AlEx,inwhichacovariancematrixofco-dominantgenotypicpairwise distances between individual sampleswas used.Genetic differentiation was also studied by performing Bayes-ianmodel-basedclusteringanalysesinSTRUCTUREv.2.3.4(Pritchardetal.2000,Falushetal.2003,2007)forwhichanadmixturemodelandcorrelatedallelefrequencieswereused.TheclusteringanalyseswererunwithKvaluesfrom1–16andwith20iterationsforeachKvalue.Eachanalysiswasrunfor200 000 generations, including 100 000 generations that were discarded as burn-in. STRUCTUREHARVESTER (Earl &VonHoldt2012)wasusedtoselectKusingboththeEvannoetal.(2005)method(KwiththehighestvalueofΔK)andthemethodofPritchardetal.(2000;KwiththehighestPr(X|K)).TheSTRUCTUREresultsweresummarisedfortheselectedvaluesofKinCLUMPAKv.1.1(Kopelmanetal.2015).GenAlExwasused toperformMantel tests (999 replicates)with the aim of testing the correlation between non-transformed geographic distance and co-dominant genotypic distance be-tweenindividuals.Thesetestswererepeatedwithadatasetin which geographic distances were transformed using the naturallogarithm.

RESULTS

TheR. speciosa flowers or flower buds that were sampled for this study were obtained from 66 Tetrastigma host plants from ninepopulations.Atotalof20ofthesehostsbelongtoT. har-mandiiPlanch.(30%)and45toT.cf.magnumMerr.(68%).Onehostplantcouldnotbeidentifiedtospecieslevel.Ofthe83Rafflesiaspecimensforwhichthesexcouldbedetermined,58(70%)werestaminateand25(30%)werepistillate.Be-cause we sampled buds or flowers from nearly all infected host plants that we encountered in each population and we did not notice substantial differences in the proportion of infected host plants and those that did not show signs of infection, population samplesizesareassumedtoberoughestimatesoftherelativesizesofR. speciosapopulations(Table2).Genotypinganalysesusing15microsatellitelocishowedthat18ofthe105DNAsampleshadidenticalgenotypestothoseofother flowersorbudsobtainedfromthesamehostplant.These18samplescouldpotentiallyrepresentthesameRaf-flesiaindividualandwerethereforeexcludedfromthegeneticanalyses.Atotalof85oftheremaining87samplesweregeneti-callyunique.Thetwosamplesthatweregeneticallyidenticaltoeach other were obtained from different but nearby host plants atthesamesiteofthesamepopulation.The15microsatellitelociyielded118alleles.Acrossallpopu-lations, thenumber of alleles variedbetween4and13perlocus,withameanof7.87.Noneofthelocishowedevidenceofproducingnullalleles.OnlytheValderramapopulationdevi-atedsignificantly fromHardy-Weinbergproportionsowingtoadeficiencyofheterozygotes(Table2).Theobservedmeanheterozygosity (Ho=0.61)was slightly higher or lower thantwoestimatesofthemeanexpectedheterozygosity(He=0.58,uHe=0.63).TheMtPoras-Aningalan,SouthernSibalomNaturalPark,andValderramapopulationshavethehighestgeneticdiversityasmeasured by the percentage of polymorphic loci, allelic rich-ness,numberofeffectivealleles,andheterozygosity(Table2).ThepopulationsinCulasi,andMiag-aoshowsomewhatlessgeneticdiversity.The lowest levelsofgeneticdiversitywerefoundinBarbazaandIgbaras,butparticularlyinLeonandMtKanlaon.Theinbreedingcoefficientsarelowforallpopulations(FIS=-0.30–0.06;Table2).Thenumberofprivateallelesperpopulationvariesfrom0inBarbaza,Culasi,andLeon,to6inMtKanlaonandMtPoras-Aningalan.TheresultsoftheAMOVAanalysisindicatethattheoverall genetic differentiation within R. speciosa is statistically significant(F’ST=0.11,p<0.001).Atotalof4%ofthegeneticvariationisfoundamongpopulationsand5%amongindividu-als.TheAMOVAfurthershowsstatisticallysignificantpairwiseFSTvalues(atP=0.05aftertheB-Ycorrection)betweenMtKanlaonandallotherpopulations,exceptLeon(Table3).Atotal of seven of the pairwise FST values between populations

Valderrama Barbaza Culasi Leon Miag-ao Igbaras MtPoras- SouthernSibalom Aningalan Natural Park

Barbaza 0.013

Culasi 0.000 0.017

Leon 0.000 0.042 0.006

Miag-ao 0.001 0.032* 0.001 0.000

Igbaras 0.037* 0.039 0.015 0.000 0.036*

MtPoras-Aningalan 0.009 0.017 0.011 0.045 0.015 0.060*

SouthernSibalomNaturalPark 0.032 0.045 0.034 0.010 0.050* 0.046* 0.052*

MtKanlaon 0.105* 0.114* 0.094* 0.076 0.084* 0.086* 0.097* 0.108*

Table 3 Pairwise FST values between populations of R. speciosa.*SignificantpairwisecomparisonsafterB-Ycorrection(Narum2006).

98 Blumea–Volume63/2,2018

inPanayaresignificantaswell.AnalysesoftheSTRUCTUREdataforK=1–16inSTRUCTUREHARVESTERindicatethepresenceoffourprimarygeneticclusters(K=4;Fig.3).OneofthesegeneticclustersalignswellwiththeMtKanlaonpopu-lation.TheothergeneticclustersdonotalignwithindividualpopulationsorgroupsofpopulationsinPanay.Mostofthesepopulationsdisplayalargediversityofadmixturepatternsatthelevelofindividualplants(Fig.3a).Thefourgeneticclustersalso do not align with the two Tetrastigma species that were identifiedashostsforR. speciosa(Fig.3b).TheresultsofaPCoA of pairwise genetic distances show little genetic differ-entiationamongpopulations(Fig.4).TheresultsoftheManteltestsdonotrevealastatisticallysignificantpositivecorrelationat p=0.05betweengeographicandgeneticdistancesbetweenindividuals.

DISCUSSION

Usingdatafrom15microsatelliteloci,thisstudypresentsthefirst information about the patterns of genetic diversity andgenetic differentiation of R. speciosa.Althoughthisspeciesisthe second-most common species of Philippine Rafflesia after R. lagascaes.str. fromLuzonIslandintermsofthenumberof knownpopulations (9 vs13;Pelseret al. 2017), its totaldistributionarea(CPMRinPanayandMtKanlaoninNegros)isrelativelysmall(Map1)andallpopulationsareexposedtohabitat degradation and fragmentation as a result of slash-and-burnagricultureandotherdestructivepractices(Barcelonaetal.2009b;Fig.2).ManyoftheWorld’stropicalforestsexperienceon-goingdegradationandfragmentationoftheirhabitat(Turner&Corlett1996,Laurance1998,Haddadetal.2015),whichcan result in populations that are smaller and more isolated fromeachother.Suchpopulationsaremoreproneto losing

Fig. 3K=4STRUCTUREresultsforR. speciosa.Eachbarrepresentsanindividualplantandbarcoloursindicatetheproportionofmembershiptoeachgeneticcluster(qvalues).a.Individualsgroupedbypopulation;b.individualsgroupedbyTetrastigmahostspecies.

Fig. 4ThePCoAordinationplotofR. speciosausingacovariancematrixofco-dominantgenotypicpairwisedistancesbetweenindividualsampleswithdatastandardization.ThefirstandsecondPCoAaxesexplain9.0%and7.2%ofthevariation,respectively.

99P.B.Pelseretal.:ConservationgeneticstudyofRafflesia speciosa

genetic diversity through genetic drift and inbreeding depres-sion, increasing theirextinctionrisk(Ellstrand&Elam1993,Youngetal.1996,Loweetal.2005,Chávez-Pesqueiraetal.2014).However,theresultsofourstudyindicatethathabitatdegradation and fragmentation have not yet had a measurable detrimental effect on the genetic diversity of R. speciosa popu-lations, because we did not observe evidence of inbreeding as measuredbytheinbreedingcoefficientFIS (Table2).Inaddition,R. speciosa hasverysimilarlevelsofheterozygosity(Ho=0.61,He=0.58,uHe=0.63)astheR. lagascaecomplex(Ho=0.60,He=0.53,uHe=0.64;Pelseretal.2017).Thesevaluesarecomparable to those reported for outcrossing or perennial plant species in a meta-analysis of genetic diversity in plants (Nybom2004).Undertheassumptionthatoursamplesizesareapproximatelyproportionaltorelativepopulationsizes,smallerpopulations of R. speciosa,particularlyinLeon,appeartohavesomewhat lower levels of genetic diversity, as measured by the percentage of polymorphic loci and allelic richness, than larger populations(e.g.,Valderrama).However,thispatternismuchless pronounced than what was observed for the R. lagascae complex(Pelseretal.2017).TheonlyR. speciosa population known fromNegros (MtKanlaon)deviates from thispatternby being relatively large, but showing among the lowest levels of genetic diversity when compared to populations from which similarnumbersofsampleswereobtained(Table2).In addition to having relatively low genetic diversity, the results ofourAMOVA(Table3)andSTRUCTURE(Fig.3a)analysesindicatethattheMtKanlaonpopulationisgeneticallydifferenti-ated from the Panay populations of R. speciosa.Thelackofastatisticallysignificantpositivecorrelationbetweengeographicdistances and genetic distances of R. speciosa individuals sug-geststhatthismightnotbeduetoisolationbydistance.Verysimilar patterns of genetic diversity and genetic differentiation werepreviouslyidentifiedforR. manillana of the R. lagascae complex(Pelseretal.2017).ThisspeciesresemblestheMtKanlaonpopulationofR. speciosa in constituting a single popu-lation on a different island than where its most closely related populationsare found.Whereasallotherpopulationsof theR. lagascaecomplexarefoundonLuzonIsland,R. manillana is only known from a single population on the nearby island ofSamar.LiketheMtKanlaonpopulation,R. manillana has relatively low genetic diversity and is genetically differentiated from populations on the other island, although these patterns are more pronounced in the R. lagascaecomplexthaninR. spe- ciosa(Pelseretal.2017).Thesefindingsarecongruentwiththe hypothesis that the sea straits that separate islands form significantbarrierstogeneflowforRafflesiaspecies.AlthoughourresultsindicatethattheMtKanlaonpopulationofR. spe-ciosa is genetically differentiated from populations on Panay, thelackofdistinctclusteringofMtKanlaonindividualsinthePCoAplot(Fig.4)andtheadmixturepatternsoftheMtKanlaonpopulation (Fig. 3a) suggest that somegene flowbetweenbothislandshasoccurred.ThispossiblytookplaceinthelatePleistocene,whentheislandsofCebu,Guimaras,Masbate,Negros, and Panay were connected during times with lower sea levels(Negros-PanayPleistoceneAggregateIslandComplex;Brownetal.2013).Wecan,however,notexcludethepossibilitythatthegeneticdifferentiation between the Panay and Negros populations of R. speciosa is instead, or in part, due to the natural absence ofsuitablehabitatbetweentheCPMRandMtKanlaon.Raf-flesia speciosa populations have only been found between c.350and1100ma.s.l.,butalllandbetweentheCPMRandMtKanlaon(includingtheislandofGuimaras,whichislocatedintheGuimarasStraitbetweenPanayandNegros;Map1)isbelow200mandmostofitwellbelow100m.However,becauseother Rafflesiaspecieshavebeenfoundaslowas50ma.s.l.

aswellasathigherelevations(e.g.,R. lagascae)andbecausealmost all forest in Panay and Guimaras at lower elevation has been cleared in historic times (Hamannet al. 1999,Ferneretal.2000,Sammleretal.2012),itisalsopossiblethatthedisjunct distribution pattern of R. speciosa is more recent and aresultoftheanthropogenicdestructionofitsforesthabitat.Thelatterexplanationhasbeenproposedtoexplainthege-netic differentiation between Panay and Negros populations of two hornbill species, Penelopides panini and Aceros waldeni (Sammleretal.2012).Regardless of whether the genetic differentiation between the Panay and Negros populations is a result of a lack of gene flow across the Guimaras Strait or across unfavourable ter-restrial habitats, the results of this study and that of Pelser etal. (2017) improveourunderstandingof thegeographicalcontextandscaleatwhichgeographicisolationmanifestsasgenetic differentiation between Rafflesiapopulations.Furthercontextforthisisprovidedbythepatternsofgeneticdifferen-tiation among the Panay populations of R. speciosa.Althoughthe presence of private alleles in some of these populations (Table2)suggestthattheyaregeneticallydifferentiated,thelack of population-level genetic differentiation revealed by the STRUCTURE (Fig.3) andPCoA (Fig.4) analyses, the lowpercentageofgeneticvariationfoundamongpopulations(4%;AMOVA),andtheabsenceofstatisticallysignificantpairwiseFSTvaluesbetweenmostPanaypopulations (Table3),sug-gests that they recently had relatively high genetic connectivity amongthem.Thiscontrastswiththelowgeneticconnectivityamong populations of R. lagascaes.str.(Pelseretal.2017).Considering that all Panay populations are associated with the CPMR,itispossiblethatthelackofgeneticdifferentiationinPanay indicates a previously more continuous distribution of R. speciosaontheisland.Ifgeneflowbetweenthesepopu-lations is still on-going, the contrast between the patterns of genetic differentiation in R. speciosa and R. lagascae s.str.mightalsobeexplainedbythegenerallysmallergeographi-cal distances between R. speciosa populations than between populations of R. lagascae s.str.Thesedistancesmight besmallenoughforitsseeddisperserstocross.However,sincehabitatdestructionandfragmentationintheCPMRarequitesevere (e.g.,Oliver et al. 1991, 1993,Hamannet al. 1999,Ferneretal.2000,Gaulke2010,Mould2012,Sammleretal.2012,Oliver2014),andbecauseRafflesia seeds are most likely dispersedbyants(Fig.1)ormammals,itisperhapsmorelikelythat gene flow across deforested terrain primarily happens by meansofpollendispersalbyflies.Rafflesia pollen is dispersed bycarrionflies(Calliphoridae),whichareattractedtotherottingodour of Rafflesia flowers and the visual appearance of their largeredflowers(Bänziger1991,Nais2001).Thesefliesarelong-lived and strong fliers, and can travel distances of 22 km inafewdays(Bishopp&Laake1921,Bänziger1991).Theyhave therefore been considered to be able to disperse Raffle-siapollenoverlargerdistances(Bänziger1991,Nais2001).If theunconfirmedreportofR. speciosa inTibiao iscorrect,allpopulationsintheCPMRarewithin22kmofoneormoreof the other populations, so it is possible that pollen dispersal currently plays a more important role in sustaining gene flow between populations of R. speciosathanseeddispersal.Thishypothesis,however,remainstobetested.Rafflesia speciosa exclusively parasitizes twoTetrastigma species(T. harmandii and T.cf.magnum),althoughitissym-patricwithfouradditionalspecies(Pelseretal.2016).Whendifferent populations of a parasite species evolve preferences for different species within their host range, this can result in genetically distinct groupswithin the parasitic species (i.e.,host-raceformation;DeVegaetal.2008,Norton&Carpenter1998,Románetal.2007,Pelseretal.2016).Thesehostraces

100 Blumea–Volume63/2,2018

mighteventuallyevolveintodifferentspecies(Amico&Nickrent2009).TheSTRUCTURE results (Fig. 3b), however, fail toshow that R. speciosa individuals growing on T. harmandii are genetically differentiated from those infecting T.cf.magnum plants and therefore demonstrate that host-race formation cannotexplainthegeneticpatternsobservedforR. speciosa.In conclusion, our study resulted in a better understanding of the patterns of genetic diversity and differentiation of R. spe-ciosainthecontextofitscurrentdistribution.Whencomparedto those of the R. lagascaecomplex(Pelseretal.2017),ourresults indicate that although Rafflesia species are typically rare, have restricted and fragmented distributional ranges, show highislandendemism,andhavesmallpopulationsizes,theymay still retain similar levels of genetic diversity as outcrossing andperennialplantspecies.Inaddition,closelyrelatedpopu-lations on different islands are genetically differentiated from eachother.ThisfindingiscompatiblewiththehypothesisthatseastraitsformsignificantbarrierstogeneflowinRafflesia, although, in the case of R. speciosa, the genetic differentia-tionbetweenPanayandNegroscanalsobeexplainedbythecurrent presence of unfavourable habitat between these two is-lands.Finally,ourresultscontributetoabetterunderstandingofpatterns of genetic differentiation among Rafflesia populations indifferentgeographicalcontexts,withR. speciosa showing stronger genetic connectivity among populations within a single mountain range, and R. lagascae displaying less genetic con-nectivity between generally more distant populations belonging todifferentmountainsandmountainranges(Pelseretal.2017).OurresultsshowthatparticularlytheR. speciosa populations intwoprotectedareas(SibalomNaturalParkandMtKanlaonNaturalPark)andValderramacontainprivatealleles(Table2).Theseareasarethereforeofnotedconservationsignificanceassourcesofuniquegeneticdiversity.TheMtKanlaonpopulation,however, has relatively low genetic diversity and this together with its isolated location make it vulnerable to further reduc-tions in genetic diversity as a result of inbreeding and genetic drift. It is located close to settlements andagricultural landandthereforeexposedtohumanactivitiesthatresultinhabitatdegradation.Thispreviouslyincludedtheconversionofforestintoplantationsofexotictreespeciesforforestry,butcurrentlyperhapsmostly illegal firewood collecting, charcoalmaking,logging,andhunting(Fig.2).Themostsignificantthreatstothe populations in Panay, including those in Sibalom Natural Park, are habitat destruction and fragmentation resulting from land conversion to farms, roads, and settlements, as well as wildfiresthatspreadfromburningnearbygrasslands(Fig.2).Cattle are alsooftenallowed to graze inside the remainingpatches of natural forest and this has resulted in severe habi-tatdegradation.Insomeareas,grazingofunderstoryplantsappears to have negative impacts on natural forest regenera-tion.ParticularlyinValderrama,deforestationonsteepslopespromptedlandslidesthatdamagednearbyforestpatches(Fig.2).Rafflesia speciosa populations in Panay that enjoy partial legal protection currently still display genetic connectivity and donotshowevidenceofinbreeding.However,on-goinghabitatdestruction, fragmentation, and degeneration are undoubtedly bringingthisspeciesclosertoextinction,asisthecasewithother threatenedspecies thatshare itsnaturalhabitat (e.g.,Walden’sHornbill,theVisayanspotteddeer,thePanaymoni-tor,andtheVisayanwartypig).Inouropinion,thisdownwardtrendcanonlybehaltediftheentireCPMRreceiveseffectivelegal protection and the loss of genetic connectivity between R. speciosa populations is prevented by creating corridors of nativeforestbetweentheremainingforestfragments.

AcknowledgementsThisprojectwasfundedbytheMarsdenFundCouncilfromGovernmentfundingadministeredbytheRoyalSocietyofNewZealand,andbytheLinneanSociety,PercySladenMemorialFund,andtheNationalGeographicFoundation.WearegratefultotheDepartmentofEnvironmentandNaturalResources(DENR)andinparticularM.S.Lim,DirectoroftheBiodiversityManagementBureau(BMB)andstaffoftheDENRregionalandlocalofficesforfacilitatingtheissuanceofcollecting,transportandexportpermits.PriorInformedConsents(PICs)wereissuedbytheProtectedAreasandManagementBoards(PAMB)ofSibalomNaturalParkandMtKanlaonNaturalPark,andtheMayorsofBarbaza,Culasi,Igbaras,Leon,Miag-ao,SanRemigio,Sibalom,andValderrama.WewouldalsoliketothankNoeGapasoftheNationalMuseum,L.J.PaguntalanandP.G.C.Jakosalem(PhilippineBiodiversityConservationFoundation)forhelpingussecureaPICfromMtKanlaonNaturalPark.R.P.M.LotillaprovidedadviceaboutlocalitynamesinPanay.Numerousfieldguides/assistantsandstaffofDENR,especially A.FernandoandA.Raña,andLGUdeserveoursincerestgratitude,forwith- outtheirhelp,ourfieldworkwouldhavenotbeenpossible.

REFERENCES

AmicoGC,NickrentDL.2009.Populationstructureandphylogeographyof themistletoesTristerix corymbosus andT. aphyllus (Loranthaceae)usingchloroplastDNAsequencevariation.AmericanJournalofBotany96:1571–1580.

BaleteDS,PelserPB,NickrentDL,etal.2010.Rafflesiaverrucosa(Raffle-siaceae),anewspeciesofsmall-floweredRafflesiafromeasternMindanao,Philippines.Phytotaxa10:49–57.

BänzigerH.1991.Stenchandfragrance:uniquepollinationlureofThailand’slargestflower,RafflesiakerriiMeijer.NaturalHistoryBulletinoftheSiamSociety39:19–52.

BänzigerH.2004.StudiesonhithertounknownfruitsandseedsofsomeRaf-flesiaceae, and a method to manually pollinate their flowers for research andconservation.LinzerBiologischeBeiträge36:1175–1198.

BarcelonaJF,CajanoMO,HadsallAS.2006.Rafflesiabaletei,anothernewRafflesia(Rafflesiaceae)fromthePhilippines.KewBulletin61:231–237.

BarcelonaJF,CoLL,BaleteDS,etal.2009a.Rafflesiaaurantia(Rafflesia-ceae):anewspeciesfromnorthernLuzon,Philippines.Gardens’BulletinSingapore61:17–27.

BarcelonaJF,FernandoES.2002.AnewspeciesofRafflesia (Rafflesia- ceae)fromPanayIsland,Philippines.KewBulletin57:647–651.

BarcelonaJF,FernandoES,NickrentDL,etal.2011.Anamendeddescrip-tionofRafflesialeonardiandarevisedkeytoPhilippineRafflesia(Raffle-siaceae).Phytotaxa24:11–18.

BarcelonaJF,MantingMME,ArbolonioRB, et al. 2014.Rafflesiamixta(Rafflesiaceae),anewspeciesfromSurigaodelNorte,Mindanao,Philip-pines.Phytotaxa174:272–278.

BarcelonaJF,PelserPB,BaleteDS,etal.2009b.Taxonomy,ecology,andconservationstatusofPhilippineRafflesia.Blumea54:77–93.

BarcelonaJF,PelserPB,CabutajeEM,etal.2008.Anothernewspeciesof Rafflesia (Rafflesiaceae) fromLuzon,Philippines:R. leonardi.Blumea53:223–228.

BarkmanTJ,KloosterMR,GaddisKD,etal.2017.Readingbetweenthevines:hostsasislandsforextremeholoparasiticplants.AmericanJournalofBotany104:1382–1389.

BendiksbyM,SchumacherT,GussarovaG, et al. 2010.Elucidating theevolutionary history of the Southeast Asian, holoparasitic, giant-flowered Rafflesiaceae: pliocene vicariance, morphological convergence and char-acterdisplacement.MolecularPhylogeneticsandEvolution57:620–633.

BishoppFC,LaakeEW.1921.Dispersionoffliesbyflight.JournalofAgri-culturalResearch21:729–766.

BrownRM,SilerCD,OliverosCH,etal.2013.Evolutionaryprocessesofdiversificationinamodelislandarchipelago.AnnualReviewofEcology,Evolution,andSystematics44:411–435.

Chávez-PesqueiraM,Suárez-MontesP,CastilloG, et al. 2014.HabitatfragmentationthreatenswildpopulationsofCaricapapaya(Caricaceae) inalowlandrainforest.AmericanJournalofBotany101:1092–1101.

DavidM,DeJesusAC,BarcelonaJF.2011.RafflesiaofthePhilippines:astoryofadventure,appetite,andaffinity.EnergyDevelopmentCorporation(EDC),Manila.

DeVegaC,BerjanoR,AristaM,etal.2008.Geneticracesassociatedwiththe genera and sections of host species in the holoparasitic plant Cytinus (Cytinaceae)intheWesternMediterraneanbasin.NewPhytologist178:875–887.

EarlDA,VonHoldtBM.2012.STRUCTUREHARVESTER:awebsiteandprogramforvisualizingSTRUCTUREoutputandimplementingtheEvannomethod.ConservationGeneticsResources4:359–361.

101P.B.Pelseretal.:ConservationgeneticstudyofRafflesia speciosa

EllstrandNC,ElamDR.1993.Populationgeneticconsequencesofsmallpopulation size: implications for plant conservation.AnnualReviewofEcologyandSystematics24:217–242.

EmmonsLH,NaisJ,BriunA.1991.ThefruitandconsumersofRafflesiakeithii(Rafflesiaceae).Biotropica23:197–199.

EvannoG,RegnautS,GoudetJ.2005.Detectingthenumberofclustersofindividualsusingthesoftwarestructure:asimulationstudy.MolecularEcology14:2611–2620.

FalushD,StephensM,PritchardJK.2003.Inferenceofpopulationstructure:extensionstolinkedlociandcorrelatedallelefrequencies.Genetics164:1567–1587.

FalushD,StephensM,PritchardJK.2007.Inferenceofpopulationstructureusingmultilocusgenotypedata:dominantmarkersandnullalleles.Mo-lecularEcologyNotes7:574–578.

FernerJW,BrownRM,SisonRV,etal.2000.TheamphibiansandreptilesofPanayIsland,Philippines.AsiaticHerpetologicalResearch9:1–37.

FrankhamR.2005.Geneticsandextinction.BiologicalConservation126:131–140.

GalindonJMM,OngPS,FernandoES.2016.Rafflesiaconsueloae(Raffle-siaceae),thesmallestamonggiants:anewspeciesfromLuzonIsland,Philippines.PhytoKeys16:37–46.

GaulkeM.2010.OverviewonthepresentknowledgeonVaranusmabitangGaulkeandCurio,2001,includingnewmorphologicalandmeristicdata.Biawak4:50–58.

HaddadNM,BrudvigLA,ClobertJ,etal.2015.HabitatfragmentationanditslastingimpactonEarth’secosystems.ScienceAdvances1:e1500052.

HamannA,BarbonEB,CurioE,etal.1999.Abotanicalinventoryofasub-montanetropicalrainforestonNegrosIsland,Philippines.Biodiversity&Conservation8:1017–1031.

HidayatiSN,WalckJL.2016.AreviewofthebiologyofRafflesia:whatdoweknowandwhat’snext?BuletinKebunRaya19:67–78.

KopelmanNM,MayzelJ,JakobssonM,etal.2015.CLUMPAK:aprogramforidentifying clustering modes and packaging population structure inferences acrossK.MolecularEcologyResources15:1179–1191.

KuijtJ.1969.Thebiologyofparasiticfloweringplants.UniversityofCaliforniaPress,BerkeleyandLosAngeles.

LauranceWF.1998.Acrisisinthemaking:responsesofAmazonianforeststolanduseandclimatechange.TrendsinEcology&Evolution13:411–415.

LoweAJ,BoshierD,WardM,etal.2005.Geneticresourceimpactsofhabitatlossanddegradation;reconcilingempiricalevidenceandpredictedtheoryforneotropicaltrees.Heredity95:255.

MouldA.2012.CavebatsofthecentralwestcoastandsouthernsectionoftheNorthwestPanayPeninsula,PanayIsland,thePhilippines.JournalofThreatenedTaxa4:2993–3028.

NaisJ.2001.Rafflesiaoftheworld.SabahParks,KotaKinabalu.NarumSR.2006.BeyondBonferroni:lessconservativeanalysesforcon-servationgenetics.ConservationGenetics7:783–787.

NickrentDLN.1997onwards.Theparasiticplantconnection(http://parasitic-plants.siu.edu/,lastaccessed3Sept.2017).

NortonDA,CarpenterMA.1998.Mistletoesasparasites:hostspecificityandspeciation.TrendsinEcology&Evolution13:101–105.

NybomH.2004.ComparisonofdifferentnuclearDNAmarkersforestimatingintraspecificgeneticdiversityinplants.MolecularEcology13:1143–1155.

OliverWLR.2014.Thetaxonomy,distributionandstatusofPhilippinewildpigs.JournalofMountainEcology3:26–32.

OliverWLR,CoxCR,DolarLL.1991.ThePhilippinespotteddeerconserva-tionproject.Oryx25:199–205.

OliverWLR,CoxCR,GonzalesPC,etal.1993.CloudratsinthePhilippines:Preliminaryreportondistributionandstatus.Oryx27:41–48.

PeakallR,SmousePE.2012.GenAlEx6.5:geneticanalysisinExcel.Popula-tiongeneticsoftwareforteachingandresearch–anupdate.Bioinformatics28:2537–2539.

PelserPB,BarcelonaJF,NickrentDL (eds). 2011 onwards.Co’s digitalFlora of thePhilippines (http://www.philippineplants.org, last accessed3Sept.2017).

PelserPB,NickrentDL,BarcelonaJF.2016.Untanglingavineanditspara- site: host specificity of PhilippineRafflesia (Rafflesiaceae).Taxon 65:739–758.

PelserPB,NickrentDL,CalladoJRC,etal.2013.Mt.Banahawreveals:TheresurrectionandneotypificationofthenameRafflesialagascae(Rafflesia-ceae)andcluestothedispersalofRafflesiaseeds.Phytotaxa131:35–40.

PelserPB,NickrentDL,GemmillCEC, et al. 2017.Genetic diversity andstructure in thePhilippineRafflesia lagascae complex (Rafflesiaceae)inform its taxonomic delimitationand conservation.SystematicBotany42:543–553.

PritchardJK,StephensM,DonnellyP.2000.Inferenceofpopulationstructureusingmultilocusgenotypedata.Genetics155:945–959.

RaymondM,RoussetF.1995.GENEPOP(version1.2):populationgeneticssoftwareforexacttestsandecumenicism.JournalofHeredity86:248–249.

RománB,SatovicZ,AlfaroC,etal.2007.HostdifferentiationinOrobanchefoetidaPoir.Flora202:201–208.

SammlerS,KetmaierV,HavensteinK,etal.2012.Mitochondrialcontrolregion I and microsatellite analyses of endangered Philippine hornbill spe-cies(Aves;Bucerotidae)detectgeneflowbetweenislandpopulationsandgeneticdiversityloss.BMCEvolutionaryBiology12:203.

TurnerIM,CorlettRT.1996.Theconservationvalueofsmall,isolatedfrag-mentsof lowlandtropicalrainforest.TrendsinEcology&Evolution11:330–333.

VanOosterhoutC,HutchinsonWF,WillsDPM,etal.2004.Micro-checker:software for identifying and correcting genotyping errors in microsatellite data.MolecularEcologyNotes4:535–538.

YoungA,BoyleT,BrownT.1996.Thepopulationgeneticconsequencesof habitat fragmentation for plants.Trends inEcology&Evolution 11:413–418.