This item is only available as the following downloads:

Bruna_etal_2011_PLoSONE ( PDF )

PAGE 1

AsymmetricDispersalandColonizationSuccessof AmazonianPlant-AntsQueensEmilioM.Bruna1,6* ,ThiagoJ.Izzo2,6,BrianD.Inouye3,6,MariaUriarte4,6,HeraldoL.Vasconcelos5,61 DepartmentofWildlifeEcologyandConservationandCenterforLatinAmericanStudies,UniversityofFlorida,Gainesville,Florida,UnitedStates ofAmerica, 2 DepartamentodeBota ˆ nicaeEcologia,UniversidadeFederaldeMatoGrosso,Cuiaba ,MatoGrosso,Brazil, 3 DepartmentofBiologicalScience,FloridaStateUniversity, Tallahassee,Florida,UnitedStatesofAmerica, 4 DepartmentofEcology,EvolutionandEnvironmentalBiology,ColumbiaUniversity,NewYork,NewYork,UnitedStatesof America, 5 InstitutodeBiologia,UniversidadeFederaldeUberla ˆ ndia,Uberla ˆ ndia,MinasGerais,Brazil, 6 BiologicalDynamicsofForestFragmentsProject,InstitutoNacional dePesquisasdaAmazo ˆ nia&SmithsonianTropicalResearchInstitute,Manaus,Amazonas,BrazilAbstractBackground:Thedispersalabilityofqueensiscentraltounderstandingantlife-historyevolution,andplaysafundamental roleinantpopulationandcommunitydynamics,themaintenanceofgeneticdiversity,andthespreadofinvasiveants.In tropicalecosystems,speciesfromover40generaofantsestablishcoloniesinthestems,hollowthorns,orleafpouchesof specializedplants.However,littleisknownabouttherelativedispersalabilityofqueenscompetingforaccesstothesame hostplants.Methodology/PrincipalFindings:Weusedempiricaldataandinversemodeling—atechniquedevelopedbyplant ecologiststomodelseeddispersal—toquantifyandcomparethedispersalkernelsofqueensfromthreeAmazonianant speciesthatcompeteforaccesstohost-plants.Wefoundthatthemodalcolonizationdistanceofqueensvaried8-fold,with thegeneralistantspecies( Crematogasterlaevis )havingagreatermodaldistancethantwospecialists( Pheidoleminutula Azteca sp.)thatusethesamehost-plants.However,ourresultsalsosuggestthatqueensof Azteca sp.havemaximal distancesthatarefour-sixteentimesgreaterthanthoseofitscompetitors.Conclusions/Significance:Wefoundlargedifferencesbetweenantspeciesinboththemodalandmaximaldistanceant queensdispersetofindvacantseedlingsusedtofoundnewcolonies.Thesedifferencescouldresultfrominterspecific differencesinqueenbodysize,andhencewingmusculature,orbecausequeensdifferintheirabilitytoidentifypotential hostplantswhileinflight.Ourresultsprovidesupportforoneofthenecessaryconditionsunderlyingseveralofthe hypothesizedmechanismspromotingcoexistenceintropicalplant-ants.Theyalsosuggestthatforsomeantspecieslimited dispersalcapabilitycouldposeasignificantbarriertotherescueofpopulationsinisolatedforestfragments.Finally,we demonstratethatinversemodelsparameterizedwithfielddataareanexcellentmeansofquantifyingthedispersalofant queens.Citation: BrunaEM,IzzoTJ,InouyeBD,UriarteM,VasconcelosHL(2011)AsymmetricDispersalandColonizationSuccessofAmazonianPlant-AntsQueens.PLoS ONE6(8):e22937.doi:10.1371/journal.pone.0022937 Editor: WayneM.Getz,UniversityofCalifornia,Berkeley,UnitedStatesofAmerica Received February22,2011; Accepted July8,2011; Published August3,2011 Copyright: 2011Brunaetal.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,whichpermits unrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited. Funding: ThisresearchwassupportedbyGrantsDEB-0453631andDEB-0452720fromtheUSNationalScienceFoundation(http://nsf.gov).Publicationofthis articlewasfundedinpartbytheUniversityofFloridaOpen-AccessPublishingFund.Thefundershadnoroleinstudydesign,datacollectionandanaly sis,decision topublish,orpreparationofthemanuscript. CompetingInterests: Theauthorshavedeclaredthatnocompetinginterestsexist. *E-mail:embruna@ufl.eduIntroductionTheapproximately14,000speciesofants(familyFormicidae) accountforroughlyone-thirdoftheworld’sinsectbiomass[1]. Thedispersalabilityoffoundressqueensiscentraltounderstandingantlife-historyevolution,andalsoplaysafundamentalrolein antpopulationandcommunitydynamics,themaintenanceof geneticdiversity,andthespreadofinvasiveantspecies[2,3,4,5]. However,withtheexceptionofafewwell-studiedspecies[5,6], littleisknownregardingthedistancesqueenstypicallydisperse whentheyleavetheircoloniestofoundnewnestsorthemaximum distancestheyarecapableofdispersing(butseee.g.,[4,7,8]).This isbecausetechniquescommonlyusedtotrackdispersinganimals (e.g.,mark-recapturemethods,radiotransmitters)arerarely applicabletoantsgiventheirsize,thestructuralcomplexityof thehabitatthroughwhichtheydisperse,andthedifficultyin identifyingandsurveyingallpotentialnestingsites.Though genetictechniquesforestimatingdispersalappearpromising[4,8], theirapplicationmaybelimitedowingtotheirstringent assumptionsandchallengesinsamplingintensivelyenoughto accuratelyestimatedispersal. Thedominanceofantsisparticularlypronouncedinthetropics, whereinadditiontotheirnumericalsuperioritytheyarecritical predators,herbivores,ecosystemengineers,andagriculturalpests [1].Speciesfromatleast40generaoftropicalantsalsoestablish coloniesinthespecializedstems,hollowthorns,leafpouches,or petiolesofplantsknownas‘myrmecophyes’;theseantsdefend host-plantsagainstherbivoresandpruneencroachingvegetation [9,10].Multipleantspeciesoftenvieforthesamespeciesofhostplant[11,12],andvacantplantsinwhichqueenscanestablish PLoSONE|www.plosone.org1August2011|Volume6|Issue8|e22937

PAGE 2

coloniesarealimitingresourceforwhichthereisintense competition[13,14,15].Theorysuggeststhatinterspecificdifferencesinthedispersalcapabilityofantqueensplayakeyroleinthe maintenanceofdiversityinthesecommunities,eitherviatradeoffs betweendispersalabilityandotherlife-historytraits(e.g., competitiveability,colonyfecundity),orfromtheinteractionof dispersallimitationwithspatialheterogeneityinhost-plantdensity (reviewedin[3]).Studiesinmultipleplant-antsystemshave demonstratedinequitiesinthecompetitiveabilityofantqueensor colonies[16,17],plantandcolonydistributionconsistentwith habitatpartitioningandpatchdynamics[18,19],andpatternsof colonizationthatimplydispersallimitation[8,12,20]orinterspecificvariationindispersalability[20,21].Nevertheless,drawing generalconclusionsregardingtheimportanceofdispersalfor plant-antcoexistencerequiresquantitativedescriptionsofdispersal formultipleantscompetingforaccesstothesamehost-plants. Thebiologyofmyrmecophytesprovidesauniqueopportunity tocircumventthechallengesassociatedwithquantifyingantqueen dispersalinothersystems.Theantspeciesthatnestintheseplants dosoobligately,andeachisassociatedwithalimitedsubsetof plantspecies[11].Consequently,allantcoloniesinasite,aswell asallnestingsitestowhichqueenscouldpotentiallydisperse,can bereadilyidentifiedbymappingthedistributionofhostplants [8,22].Wemappedallindividualsoftheunderstoryshrubs Maieta guianensis and Tococabullifera (bothMelastomataceae)in9hectares ofprimaryforestinthecentralAmazon(Figure1).Thesetwo plantspeciesserveashostsforthreespeciesofantsymbiontsthat nestexclusivelyintheirdomatia: Crematogasterlaevis Pheidole minutula ,andanundescribedspeciesof Azteca [11,18,22]. Crematogasterlaevis competesforaccesstohostplantswithboth Azteca sp.and P.minutula (Figure2),andithasbeenhypothesized [21]thatsuperiordispersalabilitypromotesitspersistenceinthis systemdespitetheinferiorcompetitiveabilityofqueenscompeting foraccesstohost-plantseedlings[16],thepoordefensecolonies provideshost-plantsagainstherbivores[23],itslowratesofcolony persistence[18],andthehighmortalityratesofthehostplantsit occupies[18].Aftermappingallcoloniesofthethreeantspecies, wetransplantedvacant,greenhouse-grownseedlingsoftheirhost plants(N=50individualsofeachspecies)intothecentralhectare oftheplotandrepeatedlysurveyedthemforcolonizationbyant queens(seeMaterialsandMethods).Thesedata,coupledwiththe locationandsizeofestablishedcolonies,allowedustoestimatea probabilitydensityfunctiondescribingthespatialredistributionof successfullydispersingqueens(i.e.,the‘dispersalkernel’)ofeach antspeciesusing‘inversemodeling’–atechniquedevelopedby plantecologiststoestimatethedistancesseedsaredispersedfrom fruitingtrees[24,25,26].Toourknowledgethisisthefirst applicationofinversemodelingtechniquestocalculatethe dispersalkernelsofanimals.ResultsThemediandistancefromcolonizedseedlingstothenearest potentiallyreproductivecolonywassignificantlydifferentamong antspecies(Kruskal-Wallis,H=13.96,df=2, p 0.001,Figure3); experimentallytransplantedvacantseedlings(hereafter,‘‘trap plants’’,analogoustoseedtrapsusedinplantecology)colonized by Crematogatserlaevis queensweresignificantlyfurtherfrom reproductivecoloniesthanthosetrapplantscolonizedbyqueens ofeither Azteca sp.or Pheidoleminutula (Steele’sNonparametric MultipleComparisonTest[27],Table1).However,thisisnot becauseestablished C.laevis colonieswerelocatedfurtherfrom trapplants.Therewasasignificantdifferenceamongcoloniesof thedifferentspeciesintheirproximitytotrapplants(Table2),but P.minutula colonieswereactuallyfurtherfromtrapplantsthan thoseof C.laevis (averagedistancesfromtrapplantstocolonies: Figure1.Mapofestablishedantcoloniesandtargetseedlings. Locationandsizeofplantshostingcoloniesof Azteca sp., Crematogaster laevis ,and Pheidoleminutula andthelocationofexperimentallyplantedseedlings(‘‘trapplants’’)of Maietaguianensis and Tococabullifera doi:10.1371/journal.pone.0022937.g001 DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org2August2011|Volume6|Issue8|e22937

PAGE 3

Azteca sp.:116.09m 6 59.84SD, C.laevis :130.88m 6 60.60SD, P. minutula :137.59m 6 53.69SD,Figure4).Instead,ourinverse modelssuggest C.laevis queensestablishcoloniesfurthestfrom natalcolonies.Assumingalog-normalkernel(seeMaterialsand Methods),themodaldispersalandcolonizationdistanceof Crematogasterlaevis queensisdoublethatof Pheidoleminutula queens (40.1mand18.9m,respectively)andeight-foldthatof Azteca sp. (5m;Table3,Figure5).Thekernelsalsohadverydifferentshapes (Figure5),suggestingthatthemaximalcolonizationdistanceof C.laevis isapproximately80m,whilequeensof Azteca sp.maybe capableofinfrequentmovementsinexcessof400m.DiscussionWewereabletoestimatetheshapesofeffectivedispersalkernals forqueensofthreeantspecies.Thedispersalkernelfor Crematogasterlaevis hadthegreatestmode,suggestingthatit generallydispersesfurtherthaneitherantspecieswithwhichit competesforaccesstohostplants.However,ourresultsalso revealedthepotentialforlong-distancedispersaleventsby Azteca sp.Thatpotentialpartnerantspeciesdiffersignificantlyintheir capacitytodispersetoandcolonizehost-plantsplantsmayhelp explainpatternsofcolonizationandantcolonydistribution previouslyobservedinthis[18]andother[19,28,29]ant-plant systems.Alongwiththelackofspecializedentrancestodomatia (i.e.,‘‘lock-and-key’’mechanisms,sensu[30]),interspecificdifferencesindispersalandcolonizationsuccessmayalsobeimportant mechanismsinhibitingtheevolutionoffurtherspecializationin ant-plantsystems,inwhichthereareoftenlargedifferencesinthe qualityofdefenseandhostplantfitnessassociatedwithdifferent antpartnerspecies[18,23,31,32]. Ithaspreviouslybeensuggested[21]thatsmallerbodysize,and henceflightmuscles,mayexplainwhythe Pheidoleminutula queens havelowerdispersaldistancesthanthoseof Crematogasterlaevis ;the sameappearstobetrueinotherant-plantsystems[20].However, the Azteca sp.queenshavethelowestmodaldispersaldistanceof thesethreespecies,despitebeingsimilarinsizeto C.laevis .Given thepotentialforlongdistancedispersalbyqueensof Azteca sp.,we hypothesizethatthisshortermodaldispersaldistanceinstead Figure2.Focalcommunityofant-plantmutualists. Graphical depictionoftheAmazonianplant-antcommunityusedtoquantify dispersalcapabilityofantqueens.Valuesbyarrowsarethepercentage ofhost-plantscolonizedbyeachspeciesofantinour9-hastudysite. doi:10.1371/journal.pone.0022937.g002 Figure3.Distancefromcolonizedseedlingstothenearestreproductiveantcolony. Histogramsofthepairwisedistancesfromeach colonizedtrapplanttothenearestreproductivecolonyoftheantspeciesthatcolonizedit.A) Pheidoleminutula :meanpairwise distance=10.91m 6 5.26SD,B) Crematogasterlaevis :meanpairwisedistance=37.49m 6 25.92SD,C) Azteca sp.:meanpairwisedistance=12.30m 6 6.53SD. doi:10.1371/journal.pone.0022937.g003 DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org3August2011|Volume6|Issue8|e22937

PAGE 4

reflectstheirsuperiorefficiencyatfindinghostplants.Testingthis hypothesiswillbechallenging–itremainsamysteryhowplantantqueensinflightidentifyhost-plantseedlingsagainstabackdrop ofhundredsofotherplantspecies[9,33].However,itislikelythey useacombinationofvisualandolfactorycues,asisthecasewith phytophagousinsects[34].Indeedatshortdistances,queenshave beenshowntousevolatilesemittedbyplantstodiscriminatehostplantsfromcloselyrelatedbutnon-myrmecophyticspecies [33,35,36].Itmaybethat Azteca sp.queenshavetheabilityto detectthesevolatilesatgreaterdistancesthantheircompetitors, superiorabilitiestoidentifytheshapeofplantsandthe characteristicvenationpatternsofhost-plantleaves,orboth. Itisnotablethatthemodalantqueendispersaldistanceswe estimatedwithinversemodelsareshorterthantheaverage distancesinferredusingothertechniques[4,8,20]andwellbelow thepotentialdispersalcapacitysuggestedbyobservationsofantsin novelorexperimentallycreatedhabitatpatches[5,7].Ifhostplant densityisgreaterinoursitesthaninothersystems,thenqueens mightonlyberequiredtodisperseshortdistancestofindvacant hostplants.Amorelikelyexplanation,however,isthatprevious studieshaveoverestimateddispersal.Thiscouldresultfromnot exhaustivelymappingallpotentialsourceanddestinationhostplantsinasite[8,20],therebymissingmanyshort-distance dispersalevents. Ourstudyhastwoimportantcaveats.First,itwasconducted entirelyduringathreemonthperiodduringthedryseason.Little isknownregardingtheenvironmentalcuesthatstimulatethe nuptialflightsofantqueensintropicalforests[37],butthe colonizationofseedlingsby Pheidoleminutula inourfieldsites appearstobecloselylinkedtoprecipitation([21];seealso[37]for evidencefromPeruofsimilarseasonalityincolonizationof Cordia by Allomerusoctoarticulatus ).Ifthisseasonalvariationinhostplant colonizationby P.minutula iscommon,thencautionshouldbe takeninestimatingthetotalnumberofcolonizationsperyear usingourdata.Second,wecouldbeoverestimatingdispersal distancesforallthreespeciesifqueensarriveatexperimental seedlingsbutleftwithoutattemptingtocolonizethemordiedprior toenteringdomatia.Thelowdensityofvacantplants[15,22] probablymakesitextremelycostlyforaqueentodisperseagain onceshehasarrivedatahost-plantseedling,andextensivefield observationsindicatethatuponarrivingataseedlingqueensofall threefocaltaxaimmediatelyshedtheirwingsandattempttoenter domatia(HLVandTJI, personalobservation ).Somequeenswill probablydiepriortocolonizingtheseedlingonwhichtheyland, however,andthereissomeexperimentalevidencethat P.minutula successfullyentersdomatiaatahigherratethan C.laevis [16].Itis thereforepossiblethatusingcolonizationoftrap-plantsbyqueens asaproxyfordispersalmeansthatourresultsareconservative estimatestruedispersalability,especiallyfor C.laevis .Ifso,our estimatesofdispersalmightbestbecalled‘realizeddispersal’,i.e., dispersalfollowedbysuccessfulcolonization[8,20]. Inconclusion,ourresultshaveimplicationsforthestudyof plant-antdiversityintropicalecosystems.First,tropicalforestsare increasinglyfragmentedbyhumanactivities,whichisolates populationsofant-plantpartners[22].Thematingsystemof socialinsectsmakesthemparticularlysusceptibletoinbreeding [38],andisolatedpopulationsarefrequentlysmallerthanthosein unbrokenforest[22].Ifthedistanceseparatingfragmentsprovesa barriertodispersalforqueensofsomespecies,thiswillincreasethe likelihoodthatisolatedpopulationsofantsandtheirhost-plants couldsufferthedetrimentaleffectsofdemographic,environmental,orgeneticstochasticity[8].Second,acriticalbutrarely documentedrequirementofsomemechanismsthatpromote coexistenceinant-plantcommunitiesisthatpoorercompetitorsor habitatspecialistsaresuperiordispersers.Ourresultsare consistentwiththishypothesis,butalsosuggestthatattempting tocategorizespeciesas‘‘good’’or‘‘poor’’disperserswhentesting modelsofcompetition-colonizationtradeoffsisoverlysimplistic– isthebestdispersertheonethathasthegreatestpotentialdispersal distanceortheonethatdispersersfurtheronaverage?Finally,we showthataninversemodelingapproachcanhelpovercomethe challengesinquantifyingantdispersalinstructurallycomplex habitats,nottheleastofwhichisthedifficultyindocumentingrare long-distancedispersalevents[26].MaterialsandMethods EthicsStatementAllresearchwasconductedwiththeapprovalofBrazil’s NationalCouncilofScientificandTechnologicalDevelopment (CNPq,PermitNumber276/2005)andtheBrazilianInstituteof EnvironmentandRenewableNaturalResources(IBAMA,Permit Number226/2005).FieldSiteanddatacollectionFieldworkwasconductedJanuary–September2007inReserve # 1501oftheBiologicalDynamicsofForestFragmentsProject (BDFFP).This1,000hareserveislocated70kmnorthofManaus, Brazil(2 u 30 9 S,60 u W)andisembeddedinalarge( 10,000ha) expanseofprimaryforest.Thehabitatisnon-floodedlowlandrain forest,witha30–35mtallcanopyandanunderstorydominated bystemlesspalms.Soilsinthesitesarehighlyacidicandnutrient poorxanthicfarralsolswithpoorwaterretentioncapacity[39]. Annualrainfallrangesfrom1,900–3,500mmperyear,andthere isapronounceddryseasonfromJune–October[40]. Tococabullifera (Melastomataceae)isanunderstoryshrubthat growstoamaximumheightof2–3m.Ithastwopouchesatthe baseofeachleafinwhichantqueensestablishcolonies[18,41]. Table1. ResultofSteel’sTestcomparingthemedian distanceofcolonizedtrapplantstothenearestreproductive colonyforallpairwisecomparisonsofantspecies. Comparison RelativeEffect, ^ p p (lower-upper95%confidencelimits) pvalue Azteca sp.vs. Crematogasterlaevis 0.82(0.58–1.07)0.005 Azteca sp.vs. Pheidole minutula 0.45(0.21–0.68)0.87 C.laevis vs. P.minutula 0.15( 2 0.09–0.40)0.002 doi:10.1371/journal.pone.0022937.t001 Table2. NestedAnalysisofVariancecomparingtheaverage distanceoftrapplantstocoloniesofthethreeantspecies ( Crematogasterlaevis Azteca sp., Pheidoleminutula )mapped inour9hastudysite. Source df MS F PAntSpecies2496230169.68 0.0001 Trap(AntSpecies)3740.0250.99 Error1669448822953 (NestedANOVA;MaineffectofAntspecies:F2,16694=169.68,P= 0.0001). doi:10.1371/journal.pone.0022937.t002 DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org4August2011|Volume6|Issue8|e22937

PAGE 5

Maietaguianensis (Melastomataceae),alsoanunderstoryshrub, growstoaheightof1.5m[18,21].Ithashighlydimorphicpaired leaves,withapairoffoliarpouchesatthebaseofthelargerleaves inwhichantsnest.Inourstudysitestwoantspeciesareassociated with M.guianensis ;mostplantscontaincoloniesof Pheidoleminutula (95%),withtheremainderoccupiedby Crematogasterlaevis (5%). Theantassociatesof T.bullifera areanundescribedspeciesof Azteca (67%)and Crematogasterlaevis ( 33%)(Figure2).These frequenciesaresimilartothosereportedinprevioussurveys conductedinourfieldsites[18].Althoughapreviousstudy conductedinourstudysites[11]hastreatedthe Azteca speciesthat colonizes T.bullifera and M.guianensis asthesameonecolonizing thesympatricmyrmecophyte Cordianodosa (Boraginaceae),this appearstobeamisidentificationresultingfromtheuseofworker morphologytodifferentiatespecies.Thecomplextaxonomyof Azteca requiresusingqueenstodistinguishspecies[42];differences between Azteca queensfrom C.nodosa andthosefrom T.bullifera in size,coloration,theshapeofthepropodeum,andthenumberof propodealhairsstronglysuggestthesearedistinctspecies(T.Izzo, unpubl.data ).Althoughseedlingsofbothplantspeciescanharbor incipient(i.e.,non-reproductive)coloniesofmorethanoneant species,adultplantshousejustasinglecolonyofonlyonespecies. Inadditiontoscavengingforinsectsontheleafsurface,resident antstendcoccidsforhoneydewinsidedomatia[43,44]. Figure4.Pairwisedistancesfromestablishedcoloniestotrapplants. Histogramsofthedistancefromtrapplantstocoloniesforeachofthe threefocalantspecies.TheXaxisshowsthepercentageofallcolony-trappairwisecomparisons.A) Pheidoleminutula :meancolony-trap distance=137.59m 6 53.69SD,B) Crematogasterlaevis :meancolony-trapdistance=130.88m 6 60.60SD,C) Azteca sp.:meancolony-trap distance=116.09m 6 59.84SD. doi:10.1371/journal.pone.0022937.g004 Table3. Maximum-likelihoodparameterestimates(MLE)and95%supportintervals(SI)forinversemodelsestimatingthe dispersalkernelsofthreemutualistantspeciesnestingintwospeciesofAmazonianant-plants.PheidoleminutulaAztecasp.Crematogasterlaevis Parameter1MLE (lowerSI-UpperSI)MLE (lowerSI-UpperSI)MLE (lowerSI-UpperSI) X018.85(17.42–20.47)5(5–5.05)40.16(31.54–52.16) Xb0.14(0.09–0.23)1.85(1.21–1.93)0.27(0.15–0.66) a 1.58(1.00–4.00)11.40(7.45–22.57)14.84(8.75–23.36) b 55.07(36.50–96.20)114.15( 2 10.00–250.00)1.91(0–10)1Parameters: X0=Modeofthelog-normaldispersalkernel, Xb=Varianceofthelog-normaldispersalkernel, a =Slopeofthelinedescribingtherelationshipbetween plantsizeandqueenproduction, b =Interceptoftherelationshipbetweenplantsizeandqueenproduction. doi:10.1371/journal.pone.0022937.t003 DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org5August2011|Volume6|Issue8|e22937

PAGE 6

FromJanuary–July2007wedemarcateda9-haplotatreserve 1501andthenmarkedandmappedall Maietaguianensis and Tococa bullifera intheplot(Figure1).Foreachplantwerecordedthe identityofitsantresidentestimateditssizebycountingthe numberofdomatiabearingleaves.Wemappedatotalof217 M. guianensis ( n =10with Crematogasterlaevis colonies, n =207with Pheidoleminutula colonies)and79 Tococabullifera ( n =26with C.laevis colonies, n =53with Azteca sp.colonies). BecausecolonysizeinAmazonianplant-antsislimitedbythe numberofhost-plantdomatia[15],weuseddomatianumberasa proxyforcolonysize.Toestimatequeenproductionasafunctionof colonysizewedestructivelysampled67 Tococabullifera with Crematogasterlaevis n =83 T.bullifera with Azteca sp., n =87 Maieta guienensis with C.laevis and n =101 M.guianensis with Pheidoleminutula allfromnearbylocationsoutsideoft hefocalstudyarea.Ofthese,9,9, 11,and36colonies(respectively),werereproductive.Weusedthese reproductivecoloniestoestimatetherelationshipbetweencolonysize andqueenproduction(Table4);linearregressionprovidedabetterfit tothedatathannon-linearmodels( resultsnotshown ). Wethenestablishedanarrayofgreenhouse-grownseedlingsin thecenterofthe9-haplot(Figure1).Thearraywascomposedof n =50 M.guianensis (forcolonizationby Pheidoleminutula or Crematogasterlaevis )and n =50 T.bullifera (forcolonizationby Azteca sp.or C.laevis ).Seedlingshadatleasttwofullyexpandedleaves withdomatiaandwerearrangedinagridwithspeciesalternating andplantsseparatedfromeachotherby10m.Seedlingsof T. bullifera weregrownfromseedscollectedinReserve1501and germinatedinashadehouseinmoistsand;becauseofthe difficultyingerminating M.guianensis seedswecollectedvacant M. guianensis seedlingsinthereserveandtransplantedthemto containersfilledwithlocalsoilandmaintainedinthesame shade-house.FromJuly–September2007wesurveyedthetarget Figure5.DispersalkernelsforthreespeciesofAmazonianplant-ants. Dispersalkernels(i.e.,probabilitydensityfunctionsdescribingthe spatialredistributionofqueensaroundreproductivecolonies)forantqueensobligatelynestingin Tococabullifera or Maietaguianensis .Thesekernels arescaledforacolonyhousedinaplantofthemediansizeobservedinour9-hastudyplot. doi:10.1371/journal.pone.0022937.g005 Table4. Resultsoflinearregressionstestingforarelationshipbetweenthenumberofdomatiaaplanthasandthenumberof queenscountedinthataplant. Antspecies Hostplant df SS SS Fvalue Pvalue R2 Regressionequation (regression) (residual) CrematogasterlaevisMaietaguianensis 1,1018.432.5771.651 0.00010.88 queens p ~ 0 : 404 domatia p PheidoleminutulaMaietaguianensis 1,35228.5440.46197.72 0.00010.85 queens p ~ 0 : 726 domatia p Azteca sp. Tococabullifera 1,830.913.0980.09 0.00010.91 queens p ~ 0 : 333 domatia p CrematogasterlaevisTococabullifera 1,8120.2694.7210.160.0130.56 queens p ~ 1 : 242 domatia p Notethattheinterceptofallthreeregressionsiszerobecausequeensareonlyfoundinplantswithatleastonedomatium. doi:10.1371/journal.pone.0022937.t004 DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org6August2011|Volume6|Issue8|e22937

PAGE 7

seedlings15,35,and90daysaftertransplantingtorecordthe presenceandspeciesidentityofqueens.Allqueensfoundwere removedtoallowforsubsequentcolonization,whichprevious workhasshowndoesnotinfluencetheprobabilityofrecolonization[21].Therewere n =17colonizationsby C.laevis n =23by Azteca sp.,and n =25by P.minutula Crematogasterlaevis colonized n =15ofits100potentialhostplantseedlings(15%), while Pheidoleminutula colonized n =17outof50(34%)and Azteca sp.colonized n =17outof50(34%),whiletheremainingevents wererepeatcolonizationsofindividualseedlings.ModelingframeworkWeusedinversemodels[26,45]parameterizedwiththe observationalandexperimentaldatadescribedabovetocharacterize thecolonizationofhostplantsbyqueensofourthreefocalspecies. Thismethodassumesthatobserveds patialvariationincolonization ofhostplantsbyqueensisamultiplicativefunctionofqueen production,whichisbasedonthesizeofpotentialqueensources(i.e., host-plantsize),andlocaldispersal,whichismodeledwithadispersal kernelthataccountsforproximityofthesourcestoexperimentalhost seedlings.Forthoroughreviewsofinversemodelsandtheir construction,assumptions,andapplicationsee[24,26,45]. Thetotalnumberofdispersingqueens, t ,producedbyacolony wasestimatedasalinearfunctionofthenumberofdomatiaits hostplanthasasfollows: t ~ a domatia z b 1 wheretheparameter a determinesthesteepnessintheincreasein queenproductionwiththenumberofdomatia,and b determines theinterceptofthedomatia-queenproductionrelationship. Weusedalognormaldispersalfunction,whichconsiderable empiricalandtheoreticalworkhasfoundtobethemostappropriate functionforavarietyofdispersalmechanismsincludinganimal movement([46,47,48,49],reviewedin22).Thekerneltakestheform: f ( d ) ~ 1 g e{ 1 2 ln d = X0 Xb0 @ 1 A2 2 where d istheobserveddistancebetweenthecolonyandthe vacanthostplantseedling, X0isthedistanceatwhichmaximum recruitmentoccurs(i.e.,themodeofthedispersalkernel), Xbdeterminesthebreadthorspreadofthedispersalkernel,and g isa normalizationconstantequaltothearcwiseintegrationofthe dispersalkernel[25]. Combininglocalqueenproduction Q andthedispersalkernel resultsinamodelforthepotentialnumberofqueensintrapplant i overthecourseofoursamplinginterval: Qi~ Xn k ~ 1 a tk1 z ( a = b ) tkf ( dik) 3 where tkisthenumberofqueensof k =1… n colonieswithinthe maximaldispersaldistance(inmeters)suggestedbyourmodelin the9haplot, dikisthedistancefromhostplant i tosourceplant k and f() isthelognormaldispersalkernel.Forallanalyseswe assumedthattheexpecteddensityofqueensinahostplantfollows anegativebinomialdistribution,reflectingthehighdegreeof clumpingobservedinthedata[50].Weusedsimulatedannealing, aglobaloptimizationalgorithm,tofindtheparametervaluesthat maximizedthelikelihoodofobservedrecruitmentdensities.We alsocalculatedasymptotic95%supportlimitsforallthe parameters.TheseAnalyseswereconductedusingRv2.9.2 statisticalsoftware[51]andthepackages‘‘Likelihood1.3’’and ‘‘NeighLikeli1.0’’,aswereallstatisticalanalyses.AcknowledgmentsWethankScottPowell,W.Getz,andthreeanonymousreviewersfor commentsonthemanuscriptandEmiliaZoppasdeAlbuquerque,Wesley Da ttiloandOsmaildoFerreiradaSilvaforassistanceinthefield.The BDFFPstaffprovidedinvaluablelogisticalhelp;thisispublication582in theBDFFPtechnicalseries.Datausedinthispaperhavebeenarchivedat Dryad(www.datadryad.org):doi:10.5061/dryad.h6t7g.AuthorContributionsConceivedanddesignedtheexperiments:EMBBDIHLVTJI.Performed theexperiments:TJI.Analyzedthedata:EMBMUBDIHLVTJI. Contributedreagents/materials/analysistools:MU.Wrotethepaper: EMBBDIHLVMUTJI.References1.WilsonEO,Ho ¨lldoblerB(2005)Theriseoftheants:Aphylogeneticand ecologicalexplanation.ProceedingsoftheNationalAcademyofSciencesofthe UnitedStatesofAmerica102:7411–7414. 2.Ho ¨lldoblerB,WilsonEO(1990)Theants.CambridgeMA:BelknapPressof HarvardUniversityPress. 3.PalmerTM,StantonML,YoungTP(2003)Competitionandcoexistence: exploringmechanismsthatrestrictandmaintaindiversitywithinmutualist guilds.AmericanNaturalist162:S63–S79. 4.SuniSS,GordonDM(2010)Fine-scalegeneticstructureanddispersaldistance intheharvesterant Pogonomyrmexbarbatus .Heredity104:168–173. 5.MarkinGP,DillierJH,HillSO,BlumMS,HermannHR(1971)Nuptialflight andflightrangesoftheimportedfireant Solenopsissaevissima (Hymenoptera, Formicidae).JournaloftheGeorgiaEntomologicalSociety6:145–156. 6.PintoSRR,MendesG,SantosAMM,DantasM,TabarelliM,etal.(2010) Landscapeattributesdrivecomplexspatialmicroclimateconfigurationof BrazilianAtlanticforestfragments.TropicalConservationScience3:389–402. 7.SimberloffDS,WilsonEO(1969)Experimentalzoogeographyofislands:the colonizationofemptyislands.Ecology50:278–296. 8.Tu ¨rkeM,FialaB,LinsenmairKE,FeldhaarH(2010)Estimationofdispersal distancesoftheobligatelyplant-associatedant Crematogasterdecamera .Ecological Entomology35:662–671. 9.HeilM,McKeyD(2003)Protectiveant-plantinteractionsasmodelsystemsin ecologicalandevolutionaryresearch.AnnualReviewofEcologyEvolutionand Systematics34:425–453. 10.TragerMD,BhotikaS,HostetlerJA,AndradeGV,Rodriguez-CabalMA,etal. (2010)Benefitsforplantsinant-plantprotectivemutualisms:Ameta-analysis. PLoSONE5:e14308. 11.FonsecaCR,GanadeG(1996)Asymmetries,compartmentsandnull interactionsinanAmazonianant-plantcommunity.JournalofAnimalEcology 65:339–347. 12.StantonML,PalmerTM,YoungTP(2002)Competition-colonizationtrade-offs inaguildofAfricanAcacia-ants.EcologicalMonographs72:347–363. 13.YuDW,WilsonHB,PierceNE(2001)Anempiricalmodelofspecies coexistenceinaspatiallystructuredenvironment.Ecology82:1761–1771. 14.LonginoJT(1989)Geographicvariationandcommunitystructureinan ant-plantmutualism: Azteca and Cecropia inCostaRica.Biotropica21:126– 132. 15.FonsecaCR(1999)Amazonianant-plantinteractionsandthenestingspace limitationhypothesis.JournalofTropicalEcology15:807–825. 16.IzzoTJ,BrunaEM,VasconcelosHL,InouyeBD(2009)Cooperativecolony foundingalterstheoutcomeofinterspecificcompetitionbetweenAmazonian plant-ants.InsectesSociaux56:341–345. 17.PalmerTM,YoungTP,StantonML,WenkE(2000)Short-termdynamicsofan acaciaantcommunityinLaikipia,Kenya.Oecologia123:425–435. 18.VasconcelosHL,DavidsonDW(2000)Relationshipbetweenplantsizeandant associatesintwoAmazonianant-plants.Biotropica32:100–111. 19.YuDW,DavidsonDW(1997)Experimentalstudiesofspecies-specificityin Cecropia-antrelationships.EcologicalMonographs67:273–294.DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org7August2011|Volume6|Issue8|e22937

PAGE 8

20.YuDW,WilsonHB,FredericksonME,PalominoW,DelaColinaR,etal. (2004)Experimentaldemonstrationofspeciescoexistenceenabledbydispersal limitation.JournalofAnimalEcology73:1102–1114. 21.VasconcelosHL(1993)Antcolonizationof Maietaguianensis seedlings,an Amazonant-plant.Oecologia95:439–443. 22.BrunaEM,VasconcelosHL,HerediaS(2005)Theeffectofhabitat fragmentationoncommunitiesofmutualists:Amazonianantsandtheirhost plants.BiologicalConservation124:209–216. 23.BrunaEM,LapolaDM,VasconcelosHL(2004)Interspecificvariationinthe defensiveresponsesofobligateplant-ants:experimentaltestsandconsequences forherbivory.Oecologia138:558–565. 24.CanhamCD,UriarteM(2006)Analysisofneighborhooddynamicsofforest ecosystemsusinglikelihoodmethodsandmodeling.EcologicalApplications16: 62–73. 25.RibbensE,SilanderJA,PacalaSW(1994)Seedlingrecruitmentinforests: calibratingmodelstopredictpatternsoftreeseedlingdispersion.Ecology75: 1794–1806. 26.NathanR,Muller-LandauHC(2000)Spatialpatternsofseeddispersal,their determinantsandconsequencesforrecruitment.TrendsinEcology&Evolution 15:278–285. 27.SteeleRGD(1961)Someranksummultiplecomparisontests.Biometrics17: 539–552. 28.DeboutGDG,DaleckyA,NgomiA,MckeyDB(2009)Dynamicsofspecies coexistence:maintenanceofaplant-antcompetitivemetacommunity.Oikos 118:873–884. 29.YuDW,WilsonHB(2001)Thecompetition-colonizationtrade-offisdead;long livethecompetition-colonizationtrade-off.AmericanNaturalist158:49–63. 30.BrouatC,GarciaN,AndaryC,McKeyD(2001)Plantlockandantkey: pairwisecoevolutionofa nexclusionfilterinanant-plantmutualism. ProceedingsoftheRoyalSocietyofLondonSeriesB-BiologicalSciences268: 2131–2141. 31.StantonML,PalmerTM,YoungTP,EvansA,TurnerML(1999)Sterilization andcanopymodificationofaswollenthornAcaciatreebyaplant-ant.Nature 401:578–581. 32.PalmerTM,DoakDF,StantonML,BronsteinJL,KiersET,etal.(2010) Synergyofmultiplepartners,includingfreeloaders,increaseshostfitnessina multispeciesmutualism.ProceedingsoftheNationalAcademyofSciencesofthe UnitedStatesofAmerica107:17234–17239. 33.EdwardsDP,HassallM,SutherlandWJ,YuDW(2006)Assemblinga mutualism:antsymbiontslocatetheirhostplantsbydetectingvolatilechemicals. InsectesSociaux53:172–176. 34.BernaysEA,ChapmanRF(1994)Host-plantselectionbyphytophagousinsects. NewYork:ChapmanandHall. 35.Da ttiloWFC,IzzoTJ,InouyeBD,VasconcelosHL,BrunaEM(2009)Volatile recognitionby Pheidoleminutula (Myrmicinae),anAmazonianant-plantspecialist. Biotropica41:642–646. 36.BlatrixR,MayerV(2010)CommunicationinAnt–PlantSymbioses.In: Balus kaF,NinkovicV,eds.PlantCommunicationfromanEcological PerspectiveSpringerBerlinHeidelberg.pp127–158. 37.FredericksonME(2006)ThereproductivephenologyofanAmazonianant speciesreflectstheseasonalavailabilityofitsnestsites.Oecologia149:418–427. 38.DarvillB,O’ConnorS,LyeGC,WatersJ,LepaisO,etal.(2010)Cryptic differencesindispersalleadtodifferentialsensitivitytohabitatfragmentationin twobumblebeespecies.MolecularEcology19:53–63. 39.FearnsidePM,LealFilhoN(2002)SoilanddevelopmentfromAmazonia: lessonsfromtheBiologicalDynamicsofForestFragmentsProject.In: BierregaardRO,GasconC,LovejoyTE,MesquitaR,eds.Lessonsfrom Amazonia:theecologyandconservationofafragmentedforest.NewHaven: YaleUniversityPress.pp291–312. 40.BierregaardRO,GasconC,LovejoyTE,MesquitaR,eds.Lessonsfrom Amazonia:theecologyandconservationofafragmentedforest.NewHaven: YaleUniversityPress.478p. 41.MichelangeliFA(2000)Acladisticanalysisofthegenus Tococa (Melastomataceae)basedonmorphologicaldata.SystematicBotany25:211–234. 42.LonginoJT(1991)Taxonomyofthe Cecropia -inhabiting Azteca ants.Journalof NaturalHistory25:1571–1602. 43.VasconcelosHL(1991)Mutualismbetween Maietaguianensis Aubl.,a myrmecophyticmelastome,andoneofitsantinhabitants:antprotection againstinsectherbivores.Oecologia87:295–298. 44.LapolaDM,BrunaEM,deWillinkCG,VasconcelosHL(2005)Ant-tended hemipterainAmazonianmyrmecophytes:patternsofabundanceandimplicationsformutualismfunction.Sociobiology46:433–442. 45.ClarkJS,SilmanM,KernR,MacklinE,HilleRisLambersJ(1999)Seed dispersalnearandfar:Patternsacrosstemperateandtropicalforests.Ecology 80:1475–1494. 46.TackenbergO(2003)Modelinglong-distancedispersalofplantdiasporesby wind.EcologicalMonographs73:173–189. 47.NathanR,KatulGG,HornHS,ThomasSM,OrenR,etal.(2002) Mechanismsoflong-distancedispersalofseedsbywind.Nature418:409–413. 48.GreeneDF,JohnsonEA(1992)Canthevariationinsamaramassandterminal velocityonanindividualplantaffectthedistributionofdispersaldistances? AmericanNaturalist139:825–838. 49.GreeneDF,CalogeropoulosC(2002)Dispersalofseedsbyanimalsandwind. In:BullockJ,KenwardR,HailsR,eds.Dispersalecology.Oxford:Blackwell Press.pp3–23. 50.ClarkJS,MacklinE,WoodL(1998)Stagesandspatialscalesofrecruitment limitationinsouthernAppalachianforests.EcologicalMonographs68:213–235. 51.RCoreDevelopmentTeam(2008)R:Alanguageandenvironmentfor statisticalcomputing.RFoundationforStatisticalComputing.Vienna,Austria.DispersalofAntQueensinTropicalForests PLoSONE|www.plosone.org8August2011|Volume6|Issue8|e22937