Species distribution modeling and molecular markers ... · 2013; Willis, Rudner, & Sumegi, 2000). In the traditional view, glacial refugia are known to be southern refugia for temperate
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Ecology and Evolution. 2017;7:1919–1935. | 1919www.ecolevol.org
Species distribution modeling and molecular markers suggest longitudinal range shifts and cryptic northern refugia of the typical calcareous grassland species Hippocrepis comosa (horseshoe vetch)
Martin Leipold | Simone Tausch | Peter Poschlod | Christoph Reisch
Inthetraditionalview,glacial refugiaareknowntobesouthernrefugiafortemperatespeciesfromallgroupsoforganisms:theIberian,Italian, Balkan peninsulas in Southern Europe (Hewitt, 1999, 2004;Taberlet, Fumagalli,Wust-Saucy,&Cosson, 1998),where the influ-enceof theglacial cycleswasalleviated (Tzedakis, Lawson,Frogley,Hewitt,&Preece,2002)andspeciescouldescapefromcolddrycli-matesandpersistuntiltheycouldrepopulateEuropeintheintergla-cialsandafterthelastglacial.Nevertheless,recentstudiessuggestedseveral additional refugia for temperate species beyond thesepen-insulas.Thesecrypticnorthernrefugiaarepostulatedforhigherlat-itudesthantheexpectedsouthernrefugia(Bhagwat&Willis,2008;Bylebyl, Poschlod, & Reisch, 2008;Magri etal., 2006;Willis &vanAndel,2004)andaredefinedasclimaticislandswithfavorablecon-ditions(Stewart&Lister,2001),surroundedbyunsuitableconditions.Inthecaseofcalcareousgrasslands,thismayhavebeenrockyout-cropsorsteepsunnyslopeswithshallowdrysoils (Ellenberg,1988;Poschlod,Baumann,&Karlik,2009)indeeplyincisedvalleysprovidingmicroclimatesfortemperatespecies,allowingtosurvivewheretheynormallywouldhaveperished(Flojgaard,Normand,Skov,&Svenning,2009; Stewart& Lister, 2001).As soon as climate becamewarmerin the postglacial, recolonization of the surrounding steppe–tundravegetationmayhavestarted from there. In this context, itmustbementionedthatthefollowingreforestationmighthavebeenheldbackbeforetheNeolithic(seeBush,1988),eitherbyhumanwithfireorbymegaherbivoresenlargingthepotentialhabitatsofcalcareousgrass-landspeciesbesidesnaturallytreelesssiteslikecliffs(Pokorný,Chytrý,&Juřičková,2015;Svenning,2002).Consequently,bothwildanimalsandhumanswithdomesticatedanimalsmayhavecontributedtospe-cies’rangesandcontributions.
Stewart, Lister, Barnes, andDalen (2010) postulated a longitu-dinaloceanic-continentalgradientthat isoften ignoredwhenspec-ulating about recolonization of species along the latitudinal axis.The longitudinal gradient explains the expansionof steppe speciesand their inclusion in the Late Pleistocene ‘steppe–tundra’ biome.Accordingly, theoccurrenceofcurrentpostglacialsteppespecies islimited to eastern continental interglacial refugiawhich are deter-mined by the longitudinal gradient.Occurrences in thewest couldtherefore be interpreted as cryptic refugia (compare Kunes etal.,2008;Schmitt&Varga,2012).On theotherhand, thereshouldbecounterpartexamplesforoceanicspecies,becauseextensionofaridclimatesduringthelatePleistocenewouldhavebeenanimpedimentto some taxa likewise cold climates.The rangesofoceanic specieswouldhaveexpandedduringthemoisterinterglacialsandcontractedto Western Europe during the glacial periods. Together with the
Inordertogaininsightintheoriginofatypicalcalcareousgrasslandspecieswith submediterranean and oceanic requirements,we choseHippocrepis comosa L. (horseshoevetch). Ithasalreadybeendemon-stratedthathumanactivitieshavecontributedatleastsincetheearlyNeolithic to the migration of crops, weeds, and animals (Beebee &Rowe,2000;Fjellheim,Rognli,Fosnes,&Brochmann,2006;Poschlod&Bonn,1998;Rosch,1998;Willerding,1986).ItseemsthereforequitepossiblethatthemigrationofH. comosaisalsorelatedtohumanmigra-tionprocesses.TheoccurrenceofH. comosawasfirsttimedocumentedfortheRomanageinthelowerRhineValley(Knörzer,1996).Therefore,thequestioniswhetherornotthespeciescametoCentralEuropeviaRomansettlers.AsMediterraneanspecies,thereisalsothepossibilityof spreading from Iberian orBalkanPeninsula. Exemplarily Poschlod(2014)claimsthemigrationofdrygrasslandspeciesfromtheEasternMediterranean regionor southeast Europe through themigration ofthe first farmers of the linearware ceramic culture (LBK) toCentralEuropeorfromWesternEuropethroughtheLaHoguetteculture.
Consideringtheclimaticconditions inCentralEuropeduringthePleistocene,wepostulatethatH. comosashiftedwestwardsinthegla-cial periods due to the lateral expansionof continental climate andadditional to its submediterranean character also southwards. Weassume thatH. comosa survived glaciations inwestern or southernrefugia,butitcannotfullybeexcludedthatthespeciesalsooccurredincrypticrefugiainCentralEurope.
Ouraimwasto identifyglacial refugiaandpostglacial immigra-tion routesofH. comosa toCentralEurope,andweapplied, there-fore,twoscientificapproaches.Firstly,weusedspeciesdistributionmodeling (SDM) to predict suitable refugia during the Pleistocenewith climate data.WithinMaxEnt, amachine-learning application,we initially calibrated a model containing actual distribution dataofH. comosa in combinationwith a setof today’s climateparame-ters(Elith&Leathwick,2009).Thismodelwasthenusedtoprocessclimate data prevailing during the last glacial maximum to predictsuitable refugia. Secondly, we applied amplified fragment lengthpolymorphisms(AFLPs)asmolecularmarkerstoanalyzethegeneticvariationwithin and among 38 populations ofH. comosa from thewholedistributionrangetogaininformationaboutglacialrefugiaandrecolonizationroutesofthespecies.Morespecificallyweaskedthefollowingquestions:(i)Whichrefugialareasservedassourceforthepostglacial immigrationofH. comosa toCentral Europe? (ii)WherewerethemainmigrationroutesfromtherefugiatoCentralEurope?(iii)Isthereevidenceforthelong-termsurvivalofH. comosaincryp-ticnorthernrefugia?
2 | MATERIALS AND METHODS
2.1 | Study species
For this study, we selected Hippocrepis comosa (horseshoe vetch),whichisatypicalcalcareousgrasslandspecieswithsubmediterraneanandoceanicrequirements.AsmentionedbySchmidtetal.(2007),the
Information containing georeferenced occurrences of H. comosa was downloaded from the Global Biodiversity Information Facility(GBIF).Thetotalnumberofdownloadeddatawas17,934withabout7,000 locations clustered in the northern half of France. Thereforeandbecauseof the fact that thedatasetshowedamixtureofgridbaseddata(mainlyinGermany,France,Spain,andUK)andpinpointoccurrences, an uniform raster was created with a point distanceof2.5min inanunprojectedcoordinate referencesystem (WGS84)encompassingthetotaldistributionareaoftheH. comosa.Withthisapproach,samplingbiascanbeavoided(Wisz,Hijmans,&Li,2008).Thenewdistributionmapwasreducedto2,794points,38ofthemwere additionally added from another study focusing on the samespecies(unpublisheddata).Geologicalparametershadtobeexcludedfromoursurvey,asduetoourgridbasedapproach,theoccurrencedataofH. comosawouldhavebeen linkedto incorrectedaphicval-ues. Furthermore, to our knowledge, geologicalmaps that describetheedaphicconditionsduringtheLGM,especiallyinregardtocurrentunderseaareas,arenotavailable.
TodescribetheclimaticcircumstancesofthepresentageandtheLGM(about22,000yearsago),weused19bioclimaticvariables(listedinTableA1,Appendix).Thevariablesarederivedfrommonthlymeantemperature andprecipitation and represent climatic annual trends,seasonality,andextremeconditions.Providedasseparateclimatelay-ersatWorldClim(http://worldclim.org,Version1.4,release3,Hijmans,Cameron, Parra, Jones, & Jarvis, 2005), data were downloaded asgrid(raster)format.Theresolutionofthedatawas2.5min(WGS84,unprojected).Thecurrentconditionsinvolveinterpolationofobserveddatafrom1950to2000.Asclimaticdataforthelastglacialmaximumconditions,twodifferentreconstructionswereused:CCSM4andMPI-ESM-P.AllgatheredfromWorldClimtheoriginaldatawereprovidedby the CoupledModel Intercomparison Project (CMIP5). The reso-lutionwas 2.5min (WGS84, unprojected).AsH. comosa exclusivelyoccurs in Europe, geographic data were reduced to the Europeanregion.Toavoidgeographicbias,datawereprojectedtoanequalareaprojection(EuropeAlbersEAC).AllGIS-relatedworkwascarriedoutinArcGis10.2.2(ESRI,Redlands,CA,USA).
Ecologicalnichemodelingandthesubsequentcreationofthegeo-graphicdistributionmapsofH. comosaatpresentandpasttimewascomputedwiththeprogramMaxent,version3.3.3(Phillips&Dudik,2008).TheMaxentsoftwareusesamaximumentropyalgorithmwhichiswell suited for specieshabitatmodelingusingpresence-onlydata(Elith,Graham,&Anderson,2006).Therefore,itisapropermethodfor
predictingspeciesdistributions forbothpast-and future-orientatedscenarios(Hijmans&Graham,2006).Inafirststep,wecalibratedthemodelwiththeactualoccurrencedataofH. comosatogetherwithallcurrent 19 bioclimatic variables (TableA1,Appendix). The resultingpotentialspeciesdistributionmodelwasthenprojectedontothecli-mateconditionsprevailingduringthelastglacialmaximum.Tovalidatetheinformativevalueofthemodelregardingspeciesdistribution,weusedtheareaunderthereceiveroperatingcharacteristic(ROC)curve(AUC)(Fielding&Bell,1997),whichisanimplementedvalidationrou-tinewithinMaxent.Theoccurrencedatawere randomlypartitionedintotwogroups:Onegroupcontaining75%ofthedatawasusedforthemodelcalibration;theremaining25%wereusedformodeltesting(Phillips,Anderson,&Schapire, 2006).HighAUCvalues (>0.7) indi-cateagoodmodelperformance(Fielding&Bell,1997).Withfollowingexceptions,weusedthedefaultsettingsinMaxent(Phillips&Dudik,2008).Theconvergencethresholdwassetto10−5,themaximumnum-berofiterationswas5,000,and15replicateswiththereplicatedruntype “subsample”weremade.The selection of the relevant climatedatawasautomated.Asthresholdrule,wechosemaximumtestsensi-tivityandspecificity(MTSS)tooptimizethecorrectdiscriminationofpresencesandpseudoabsencesinthetestdata(Hernandez,Graham,Master,&Albert,2006;Jimenez-Valverde&Lobo,2007).Thecontin-uouslogisticoutputofMaxentwastransformedinabinarypresence–absencemap.ThethresholdvalueforpresencewasbasedonMTSSvalueswhichwereaveragedover15runs.
2.3 | AFLP analysis
Forthemolecularanalysis,plantmaterialofH. comosawassampledthroughout the whole species range distribution on the Europeancontinentandembracedintotal588individualsfrom38populations(Table1).Eachsampleencompassedmultiplefreshandhealthyleaveswhichweredriedinsilicagel.
DNAextractionfollowedCTAPprotocolfromRogersandBendich(1994)adaptedbyReisch (2007)using15mgofthedried leafsam-ples. DNA contentswere photometrically determined and adjustedto7.8ngDNAper1μl H2O.Wechosethedominantmarkeranaly-sisofamplified fragment lengthpolymorphisms (AFLP,Vos,Hogers,& Bleeker, 1995; Zabeau & Vos, 1993) to produce loci over thewholegenome(standardizedAFLPprotocolfromBeckmannCoulter(Brea, USA)).Molecular analyseswere conducted as described pre-viously (Bylebyl etal., 2008). For the selective DNA amplification,we chose three pairs of primer (D2: GATGAGTCCTGAGTAACTA-GACTGCGTACCAATTCAAC,D3:GATGAGTCCTGAGTAACAC-GACT GCGTACCAATTCAGG,andD4:GATGAGTCCTGAGTAACAC-GACTG CGTACCAATTCACA). PCR products were separated using capillaryelectrophoreses (CEQ 8000, Beckmann Coulter, USA). Data wereexportedascurve-filesandmanuallyanalyzedfortheoccurrenceofstrong, well-defined fragments in Bionumerics 6.6 (Applied Maths,Kortrijk,Belgium).Thepresenceorabsenceoffragmentswastrans-formedintoabinary(1-0)matrix,whichservedasbasisforallfurtheranalysis. Individualsshowingnoclearbandingsignalswererepeatedorultimatelyexcluded.
Based on the allele frequencies from the 0/1 matrix Nei′s GeneDiversity (Nei,1972),Shannon′s InformationIndex(Shannon1948),
thenumberandpercentageofpolymorphic lociwerecalculatedforeachpopulationusingPOPGENEv.1.31(Yeh,Yang,&Boyle,1999).Inorder tomake the resultsofNei’sgenediversitymorecompara-ble,anadditionalcalculationwasconductedatwhichthenumberoftestedindividualswassetto12foreachpopulation(lowestavailableamount).The12sampleswerechosenrandomlywith50,000 itera-tions,andameanvaluewascalculated.Theresultswereplottedontothegeographiccoordinatesofthesamplelocations.Abasemappro-videdby“NaturalEarth”servedasbackgroundforthisandallfollow-ingmaps.
As an additional measure of divergence, the rarity of markerswas calculated by frequency-down-weighted marker values (DW)(Schönswetter&Tribsch,2005).Thecalculationof theDWvalueswas performed via the r-script AFLPdat (Ehrich, 2006). To grantequal sample sizes, for each population, 12 individuals were ran-domlyselectedwith10iterationsandameanvalueofDWwascal-culated.The resultswereplottedonto thegeographiccoordinatesof the sample locations.The value ofDW is expected to be highin long-termisolatedpopulationswhereraremarkersshouldaccu-mulateduetomutationswhereasnewlyestablishedpopulationsareexpected to exhibit low values, thus helping in distinguishing oldvicariancefromrecentdispersal(Schönswetter&Tribsch,2005).WecalculatedaPearsoncorrelationcoefficientforNei′sGeneDiversityandDWvalue.
An analysis of molecular variance (AMOVA, Excoffier, Smouse,& Quattro, 1992) should give information about the genetic vari-ance within and between populations. The two-level AMOVA wasperformed within the program GENALEX v6.5 (Peakall & Smouse,2012) and included588 individuals of all 38populations.BasedonEuclideanpairwisegeneticdistances, thesumsofsquareswerecal-culated(SSWP)anddividedbythedegreesoffreedom(SSWP/n−1).The resultingAMOVA-SS diversityvalues per sample locationwerealso presented cartographically. Permutation tests (9,999 iterations)wereconductedtoshowsignificance.
The genetic structure and group assignment of the populationswas investigated with Bayesian clustering in STRUCTURE v 2.3(Pritchard, Wen, & Falush, 2009; Pritchard, Stephens, & Donnelly,2000).TheprogramperformsaMarkovchainMonteCarlo (MCMC)algorithmtoassignthetestedindividualsintokgroupsbasedonlyonitsgeneticdata,andnotonpopulationaffiliation.Theprogramwasrunwith followingparameters:noadmixtureancestrymodel, correlatedallele frequencymodel, k from2 to 40, a burn-in period of 10,000followedby10,000iterations,10replicateruns.Themostlikelynum-berofgroups in thedata setwasdeterminedvia the calculationofΔkfollowingthemethodofEvanno,Regnaut,andGoudet(2005).Theresultswereplottedonto thegeographiccoordinatesof thesamplelocations.
Toidentifyspatialgeneticpatternswithinthedataset,amultivari-ateapproachwasconductedusingspatialprincipalcomponentanaly-sis(sPCA).sPCAwascarriedoutinR(DevelopmentCoreTeam,2014)usingpackageadegenet(Jombart,Devillard,Dufour,&Pontier,2008).Forthisanalysis,all588individualsof38populationswereused.Thepopulations’ geographic coordinates (WGS1984)wereprojected to
ETRS1989LCC.Toavoidsamecoordinatesofindividualsinthesamepopulation,thecoordinateswereshiftedrandomlybyafactorof0.5.UnliketheanalysiswithSTRUCTURE,thedataforasPCAdonothavetomeetHardy–Weinbergexpectationsorlinkageequilibrium.Forthemethod,twomatricesarenecessary.Thefirstonecontainstherelativeallelefrequenciesofallindividualsandthesecondembracesallspa-tialproximityinformationfromtheprojectedcoordinates.ThespatialproximityinformationmatrixwasgainedfromaconnectionnetworkusingDelaunaytriangulation.Also,thesecondmatrixwasusedtocal-culateaspatialautocorrelationusingMoran’sI(Moran,1948).Moran’sIrangesfrom+1to−1,indicatingastrongpositiveornegativespatialautocorrelation,respectively.Incaseofapositivespatialautocorrela-tion, a global structure in the data can be assumed.AMoran’s I ofzeroindicatesatotallyrandompattern.Foravisualverificationoftheoccurrence of spatial structures, a screeplotwas drawn by plottingthevarianceof the sPCAagainst spatial autocorrelation (Moran’s I).Supplementary,tostatisticallystrengthenthepreviousvisualfindings,twoMonteCarlotestswith9,999permutationseachwereconductedinordertodetectglobalorlocalstructuresinthedataset.Asdisplay,the genetic differentiationof theprincipal componentswasplottedontothegeographiccoordinates.
3 | RESULTS
3.1 | Species distribution modeling
Speciesdistributionmodelingresultedinthreemaps,whichshowthepredictedgeographicdistributionofHippocrepis comosa forpresenttime and the twodifferent climatic assumptions (CCSM4andMPI-ESM-P)fortheLGM.Themodelfortoday’sdistributionofH. comosa displayed a good prediction of the reported locations (FigureA1,Appendix).Aberrationscanbearesultofanimprecisesamplingdesign,duetothefactthatfirstwehadtorasterizealldataandsecondanunsteadyparticipationofEuropeancountriesinprovidingoccurrencedataoftheinvestigatedplantspecies.Also,geologicalaspectswerenot included in themodel. If taken into consideration, they wouldrule out areaswith no occurrences of calcareous substrates like inNorthernGermanyandtheNetherlandsorsilicateaffectedsubsoilsforexampleinFranceortheCzechRepublic.Nevertheless,whenthefocusliesonclimaticfactorsonly,theseregionsprovidesuitablecli-matichabitatsforH. comosa.
Of all 19 tested bioclimatic parameters, “Precipitation of DriestQuarter” (BIO17) had the highest influence on the prediction ofsuitablehabitatsof theactualoccurrenceofH. comosa,witha con-tribution to themodel of 45%.Togetherwith “TemperatureAnnualRange” (BIO7, percent contribution: 36%) and the “Isothermality”(BIO3,percentcontribution:8%,Isothermality=MeanDiurnalRange/TemperatureAnnualRange×100) rankedsecondand third the firstthreeparameterscontributewith89%tothefinalmodel.
In theAFLPanalysisof588 individualswith threeprimercombina-tions,271unambiguousfragmentswereselectedrangingbetween60and420bpandofwhich98.16%werepolymorphic (D2CTA-AAC:111fragments,D3CAC-AGG:87fragments,D4CAC-ACA:73frag-ments).Thewithin-populationgeneticvariationwascalculatedasfourdifferentmeasures(Table2).Allgeneticvariationvalueswereconsist-entlylowestinpopulationno.11inFranceandhighestinpopulationno.29inGermany.Meanpercentageofpolymorphicloci(%PPL)was56.7%, ranging between 46.3 and 64.7. Mean Shannon’s informa-tionindex(SI)was0.28,rangingbetween0.21and0.32.MeanNei’sgenediversityyieldedforallindividualswas0.18,rangingfrom0.14to0.21.MeanNei’sgenediversityfor12individualswas0.18,rang-ingfrom0.13to0.21.Highervaluesforgeneticvariation(He≥0.19)wererecordedonlyinsouthernpopulations,liketheIberian(He=0.2and 0.21), the Italian (He=0.19), the Balkan Peninsula (He=0.20),southoftheAlps(populationno.14,He=0.2),exceptfromonepopu-lationinthenorthernAlps(populationno.29,He=0.21),andnorthoftheAlpsinGermany(populationno.31).Lowervalues(He≤0.16)onlyoccurredinnorthernpopulationsandontheItalianpeninsula,butnotinthetwootherpeninsulas(Figure2).
Thesurveyof the rarityofmarkers revealedDWvalues rangingfrom4.83inanItalianpopulation(no.10)to10.5inaSpanishpop-ulation (no. 2),with an average of 7.66 (SE=0.2; seeTable2).Thehighest values were recorded either in populations of the IberianPeninsula (values between 8.35 and 10.5), the Balkan Peninsula(valuesbetween8.89and9.89),or insomenorthernpopulations, inCentralGermany (DW=10.3), BavarianAlps (DW=8.84 and8.36),andtheUnitedKingdom(DW=8.05).Lowvaluesoccurredinnorth-ernpopulationsandontheItalian,butnotontheIberianandBalkanpeninsulas (Figure2).Therewere highly significant positive correla-tions betweenDWand all geneticvariationmeasures (r=.729413,p<.0001, t=6.3977),meaning that a highnumber of raremarkers(highDW)wereassociatedwithhighgeneticvariation.
The analysis ofmolecularvariance (AMOVA) involving all popu-lations without classification of regions revealed a total molecularvariancewithin thepopulationsof68%.This leavesa strongdiffer-entiation among the populationswith amolecular variance of 32%(Figure3).Theresultswerehighlysignificant(p<.001).AllAMOVA-SSvaluesaregiveninTable2.
FigureA3).Besides thismost likelynumberof clusters, avery smallprobability for 13 groups was found, one group involving popula-tionsofWesternandCentralEurope(Spain,France,UnitedKingdom,Belgium,Switzerland,Germany),andasouthEasternEuropeangroup(Italy,Slovenia,Croatia,Macedonia).Onlyat theboundarypointsofthetwogroupsinSpain(no.3and5),Germany(no.28,29,and33),andHungary (no.37), the100%assignment intooneof thegroupswasdiluted,resultinginsomeadmixedpopulations(Figure4).
Likewise,theresultsofthesPCAintheassignmentofthepopu-lationsintotwogroupsrevealedasimilarpattern.All588individualswere included in thespatialanalysis.ThespatialPCAwasbasedonDelaunay triangulation as connection network. The eigenvalues ofthesPCAindependenceofitsMoran’sIandvarianceareshowninascreeplot(FigureA4).Theeigenvalueofthefirstglobalscoreλ1couldclearlybedistinguished fromallothereigenvaluesdue to itshigherlevelsofvarianceandspatialautocorrelation.Thisindicatestheexis-tenceofspatialstructuresinthedata,whichwassubsequentlytestedwithglobalandlocalMonteCarlotests(9,999iterations).Astheglobaltestshowedasignificant(p<.0001)result,andwiththescreeplotinmind,aglobalspatialstructurewasassumedforthedataset.ThelocalMonteCarlotestwasnotsignificant.
Figure5 shows the eigenvalues of the first global score plottedagainst geographic coordinates. Black squares indicate positive andwhitesquaresnegativevaluesofthescores.Thesizeofthesquaresrepresents different absolute values. Therefore, large-sized squaresfrombothcolorsarehighlydifferentiated,while small-sizedsquaresindicate only small differentiation. A clear distinction between twoclusters can be drawn, one involving populations of Western andCentralEurope(Spain,France,UnitedKingdom,Belgium,Switzerland,Germany)andtheotheroneincludingpopulationsfromsouthEasternEurope(Italy,Hungary,Slovenia,Croatia,Macedonia).AMonteCarloManteltest(10,000iterations)foracorrelationbetweengeographicandgeneticdistanceswashighlysignificant(p<.0001).
4 | DISCUSSION
4.1 | Species distribution modeling
As thepredictedpresent-daydistributionofH. comosa showed agoodmatchwiththeactualdistribution,weconsideredthefollow-ingSDMspredictionssuitablehabitatsduringthelastglacialmaxi-mum as convincing.Our findings support the generally accepted
F IGURE 1 SpeciesdistributionmodelprojectionofH. comosaatthelastglacialmaximum(21,000ya)basedontheoutputoftheMPI-ESM-Pscenario.Darkgrayareasindicatesuitablehabitatswithintheecologicalniche;lightgrayareaareunsuitablehabitatsforH. comosa. Ice shieldsareshowninwhitewithadarkoutline.Nationalboundariesrepresenttoday’sEuropeanlandarea
| 1925LEIPOLD Et aL.
assumptionof northern temperate speciesoutlasting the LGM insouthlocatedEuropeanrefugia(Bennett,Tzedakis,&Willis,1991;Hewitt,1999,2000;Huntley&Birks,1983;Taberletetal.,1998).Additionally, results of the species distribution models suggest
theexistenceofpossible refugia inFranceandalongtheAtlanticcoastup to theUK.Bothmodels (CCSM4andMPI-ESM-P) showdifferences in this area.While CCSM4 predicts suitable habitatsfor almost entire France and even parts of southwest Germany,
the MPI-ESM-P model draws the restriction further west. Thereason liesmostprobably in the fact thatbothmodelsmakedif-ferent assumptions regarding annual precipitation.WhileCCSM4model only shows drier summer in Central Spain and NorthernItaly but not in Central Europe, the opposite conditions are pre-dicted by theMPI-ESM-Pmodel (PMIP3, Braconnot,Harrison,&Kageyama, 2012).Given that themain contributing parameter inourmodelswas“PrecipitationofDriestQuarter”,thisassumptionseemsreasonable.Nevertheless,bothSDMpredictedvastareasofnowadayssubmerged landas suitablehabitats forH. comosa asaspecies adjusted to oceanic climate. The up to 110m lower sealevel(Ruddiman&Thomson,2001)revealedice-freelandwestofFranceandtheUKofseveral100kmwide.Beyondclimaticparam-eters,itisuncertainwhetherthislandmassesprovidedthepropercalcareous substrates for H. comosa and therefore could haveserved as refugia. Otherwise several studies including flora andfauna(Boston,Montgomery,Hynes,&Prodohl,2015;Ohlemüller,Huntley, Normand, & Svenning, 2012; Svenning, Normand, &Kageyama,2008)aresupportingtheexistenceofthesenorthwest-ernpotentialrefugia.
4.2 | Eastern and western lineages and postglacial migration
AFLPanalysisisapowerfultoolinrevealingglacialrefugiaandpost-glacial immigration processes, as we show here on an example ofthecalcareousgrasslandspeciesH. comosa.Baseduponourresults,weassumethatH. comosafollowedacontraction–expansionmodel,whereby the species was restricted to traditional southern refugiaduringtheLGMfollowingalatitudinaltemperatureandalongitudinalhumidity gradient andexpanded into the rest of Europe afterward.Ourresults indicateadditionalcrypticrefugiaatthewesternshoresofFranceandUK.
Both spatial analyses (STRUCTURE and sPCA) identified twoalmostcompletelydistinctlineagesofH. comosa:onegroupinvolv-ingpopulationsoftheIberianpeninsulaandWesternEurope(Spain,France, United Kingdom, Belgium, Switzerland, Germany) and theother includingpopulations fromthe ItalianandtheBalkanpenin-sulas (Italy, Croatia, The Former Yugoslav Republic of Macedonia,Slovenia).Only a fewadmixedpopulationsweredetected, locatedinHungary,Germany,andSpain.Accordingtothesespatialresults,
F IGURE 2 MapofNei’sgenediversity(leftsemicircle)andfrequency-down-weightedmarkervalues(DW,rightsemicircle)foreachsurveyedpopulation.Thedifferentsizesofthecirclesindicatedifferentabsolutevalues.Iceshieldsareshowninwhitewithadarkoutline.Nationalboundariesrepresenttoday’sEuropeanlandarea
| 1927LEIPOLD Et aL.
two traditional southern refugia (Hewitt, 1999, 2004; Taberletetal., 1998) could be confirmed: Iberia as a southwestern refugiaand Italy and the Balkans as southeastern refugia, enclosing theadmixedpopulationsinSpain,SouthGermany,andHungaryascon-tactzones.ContactzoneshavebeenreportedtoclusterintheAlps,CentralEurope,northernBalkans,andthePyrenees(Hewitt,2000;Taberlet etal., 1998), resulting in an accumulation of genotypesfrombothgroups.Therefore,weassumethatH. comosarepopulatedCentral Europe from Iberia (Pyrenees) to France, Britain, Belgium,Switzerland,andGermanyuntil itseasternborder,wherethepop-ulations admixedwith populations that were migrating from Italyand the Balkans up north. It has been shown before that in con-trasttoIberianlineages,ItaliangenomesrarelypopulatedNorthernEurope, as the ice-cappedAlps prevented their northward expan-sion (Hewitt, 2000; Taberlet etal., 1998). This barrier is regardedasexplanationoftherelativelylowspecies’andgeneticdiversityofnorthern populations compared to southern populations (Hewitt,2000).ConsideringthevaluesforNei’sgenediversityandtherarityindex (DW)-values,H. comosa populationsof theBalkanPeninsulamay have served as refugium fromwhere the Italian populationswerefoundedsubsequently.
4.3 | Southern and cryptic northern refugia
Themostimportantrefugialareasaregeographicregionswherespe-ciespersistedthroughoutseveralfullglacial/interglacialcycles(each100–120kyr in duration). These so-called true refugia (Stewart &Dalen,2008)areexpectedtopossesshighergeneticvariabilitycom-paredwithsurroundingrecolonizedregions(Comes&Kadereit,1998;Taberlet etal., 1998; Tzedakis etal., 2013). In contrast, recent dis-persalmight lead to genetic depauperation due to founder effects.Supporting this theory, the highest values for genetic variation ofH. comosawererecordedalmostentirelyinsouthernpopulations,likethe Iberian, the Italian, theBalkanPeninsula,andsouthoftheAlps.Otherabove-averagegeneticallydiversepopulationswerelocatedintheAlpsandinGermany.Theformercanbeascribedtotheabove-mentionedhybridizationofthewesternandthesoutheasternlineage(Petit, Aguinagalde,& deBeaulieu, 2003; Provan&Bennett, 2008;Tzedakisetal.,2013).Wefounddifferentexplanations for thehighgeneticvariationofthe latter.Firstly, itmaybetheresultofrecentgeneticexchangedue tograzingmanagement,which ismoreprev-alent in the areaof the Jurassicmountains inGermany than in thenorthernGermanpopulations.Paun,Schonswetter,Winkler,Tribsch,
F IGURE 3 MapofAMOVA-SSvaluesforeachsurveyedpopulation.Thedifferentsizesofthecirclesindicatedifferentabsolutevaluesofmolecularvariance
1928 | LEIPOLD Et aL.
and Consortium (2008) criticized the use of genetic variation foridentificationofrefugiaastheymayinfactreflectcurrentprocesses(geneticexchangeandpopulationsizes)insteadofhistoricalprocesses.Secondly,thehighgeneticvariationsmayalsohaveresultedfromaphalanxwayof recolonization fromsouth tonorthor canbe inter-pretedasa legacyfromYoungerDryascoldreversal (Hewitt,1999,2000, 2004). A reduction of northern populations during this coldperiodcouldhaveledtohighdiversitypopulationsduetothemixturewithrecolonizinglinagesduringtheHolocene(Tzedakisetal.,2013).Inordertocircumventtheseconfusions,asecondparameter,therarityindex(DW)wasused,asitisabetterindicatorofhistoricalprocesses(Paunetal.,2008).ThevalueofDWisexpectedtobehighinlong-termisolatedpopulationswhererarefragmentscouldaccumulateduetomutations,whereasyoungpopulationsareexpectedtoshowlowvalues,thushelping indistinguishingoldvicariancefromrecentdis-persal.Refugialpopulationsandrecolonizedregionswouldontheonehandcontainidenticalfragmentsbutduetodriftrarefragmentswouldaccumulateanddistinguish themfromother refugialareasandsur-roundingyoungerpopulationsthatwouldbelessdivergent(Provan&Bennett,2008;Schönswetter&Tribsch,2005;Tzedakisetal.,2013).According to this, we identified additional northern populations of
H. comosa thatare located far from the traditional southern refugiawhich possessed highDW values, supposing cryptic refugia inUK,theAlps,andCentralGermany.Suchcrypticrefugiahavepreviouslyalso been identified for other grassland species likeBromus erectus (Sutkowska,Pasierbinski,Warzecha,Mandal,&Mitka,2013)orgrass-landrelatedpineforestspecies likePolygala chamaebuxus.Althoughitmustnotbeignoredthattherarityindexmightbeoverestimated,asrelatedsouthernpopulationswiththeseallelesmaynothavebeeninvestigatedordistinctgenepatchesmightresultbygenesurfingontheleadingedge(Tzedakisetal.,2013).Nevertheless,wefoundasig-nificantcorrelationoftherarityindex(DW)andNei’sgenediversity(He),whichmeansthatinourstudy,rarefragmentsaccumulationandhighgeneticdiversity camealongwitheachother, pointing toward“true refugia”.TheGermanpopulationwithaveryhigh rarity indexandrelativelylowgeneticvariationmayindicateanisolatedrefugiumduringtheLGM,similar toalpinepopulationsofRanunculus glacialis (Paunetal.,2008).TheseresultsalsocoincidewithDengler,Janisova,Torok,andWellstein(2014),whoproposedacontinuousexistenceofpalearcticgrasslandat leastsincethePleistocene(2,400ka).Duringglaciations, grasslands covered most of the continent as steppe–tundraoverpermafrostandasxerothermicgrassland further in the
F IGURE 4 ResultsfromSTRUCTUREanalysis.ThesurveyedpopulationsofHippocrepis comosawereplottedontogeographiccoordinatesof.AsSTRUCTUREproposedatwo-groupsolution,eachpopulationwasassignedaccordingitsassociatedgroup
Besides, there are additional parameters influencing the accessibil-ityofspecieslikelifeform,dispersalabilityandLGMrefugialocation(glacialcontraction),generationtime,habitatadaptation,competitionwith established vegetation, soil development, geographic barriersandhumanhabitatfragmentation(Normandetal.,2011),whichmayhaveresultedinacurrentdisequilibrium(postglacialcolonization)ofspecies rangeswithin theactualclimate (migrational lag).RegardingthedistributionofH. comosa inEurope,weassumethat ithas fullyexpandedto itspotential range.Thenortherndistribution limitationasitistodaywasinvestigatedonaregionalscalebyHennenbergandBruelheide(2003),showingareducedfitness(reducedseedsetting),whichcorrelatedwitheffectiveairtemperaturemeasuredataheightof10cmaboveground.Theareaofclimaticallysuitablehabitatswasalso shown in the present time species distribution model for thisspecies. In addition, soil composition impedes an expansion furthernorthandeastthanitspresentstatus(Hartmann&Moosdorf2012).Therefore,wedepicted that thecurrentoccurrenceofH. comosa ismainlylimitedduetoclimateandsoilfactorsandthatthedistributionpatternisnotlimitedbyreducedaccessibility.
F IGURE 5 Graphicaldisplayofthespatialdistributionofallsurveyedpopulationswiththevaluesofthefirstpositive(global)sPCAscore.Thedifferentsizesofthesquaresindicatedifferentabsolutevalues.Largeblacksquaresarewelldifferentiatedfromlargewhiteones,whilesmallsquaresshowlessdifferentiation.Onthemap,thegenotypesdifferentiateintwodistinctclusters,oneintheinthenorthwestandoneinthesoutheast.TheusedconnectionnetworkbasedonDelaunaytriangulationisshownwithgraylines.Onthetoprightpositionofthemap,thefirst25sPCA-positivescoresareshown
1930 | LEIPOLD Et aL.
Thedecliningtemperatureatthebeginningofthelasticeagewaswithnodoubtthemaindrivingfactorinfluencingmigrationofplantsfollowingalatitudegradienttotheirsouthernrefugia(Taberletetal.,1998). Nevertheless, likewise Stewart etal. (2010), we emphasizethat not only this latitudinal gradient should be taken into consid-eration.The longitudinal gradient representing an increasing conti-nental climate fromwest toeastmighthavehadan impacton themigrationtosuitablehabitatsespeciallyforspeciesclassifiedasoce-anic likeH. comosa. Due to declining precipitations in combinationwithalowersealevel,whichcomprisedgreatlandmassesbetweenEngland and Europe, the oceanity of Central Europe decreased.Regardingthemigrationofplantsatthattime,itisouropinionthatoceanic plantswere not only forced tomove southwards but alsowestwards to maintain their climatic niche. Furthermore, the AlpsandtheCarpathianMountainshaveactedasnaturalbarrierhinderingexchangeandmigrationofplantstowardItaly.ThenorthandCentralEuropean populations ofH. comosa therefore might have followeda combination of bothvectors to the south and thewest pointingtoward apotential refugium inSpain,which canbe anexplanationfor our cluster including Spain, France, Belgium, UK, Switzerland,andGermany.AftertheLGM,H. comosamigratedtoCentralEurope,butno furthereastbecauseof theclimatic limitationsandedaphicbarriers.Nevertheless, followingtheresultsof thespeciesdistribu-tion modeling describing climatic niches during the LGM, there isalsoapossibilityof crypticnorthern refugia located in theUKandFrance,whichcouldhaveservedasadditionalplacesoforiginforarecolonization.
ForasuccessfulrecolonizationfollowingtheLGM,H. comosacouldrelyonatleasttwomechanismsofseeddispersal.Endozoochoryviaherbivores (Fischer, Poschlod, & Beinlich, 1996; Müller-Schneider,1938;vonOheimb,Schmidt,Kriebitzsch,&Ellenberg,2005)andepi-zoochory, implyingatransportationwithsoilmaterial inthehooves,whichwasreportedforsheepandcattle(Fischeretal.,1996;Poschlod&Bonn,1998)andaffectsthegeneticstructureofplantpopulations(Willerding & Poschlod, 2002). However, thisway of transport canmostprobablyalsobetransferredtootherhoofedanimalslikeredorroedeer.Therefore,wewoulddesignateH. comosaasaspecieswithahighlong-distancedispersalpotential,especiallybecauseofthefact,that hoofed animals can bridge distances of several kilometers perday (Pépin,Adrados,Mann,&Janeau,2004).ThisbecomesobviouswhenconsideringtheworkofSkog,Zachos,andRueness(2009)andMeiri, Lister, andHigham (2013), presenting the recolonization andphylogeographyofEuropeanreddeer(Cervus elaphus).Basedonmod-ernandancientDNA,thisstudystrikinglyresemblesourfindingsforH. comosa,showingarestrictionofCervus elaphustosouthernrefugiaduringtheLGMandarecolonizationofWesternandNorthernEuropeoriginatingfromIberia.Thisverysimilarphylogeographicpatternmaylead to the assumption thatH. comosa expanded toCentral EuropeviaC. elaphus or other equivalent herbivores (wild horse, roe deer;Pakeman,2001).
existedsince theLGM (Bush&Flenley,1987;Bush,1988;Pokornýetal.,2015),alsobecauseofhumanpractices(Poschlod,2014).TheLaHoguetteculture(~7,500y.a.)showedfirststepstowardnomadicgoatand sheep breeding (Gronenborn, 2003) and therefore could havedirectlyinfluencedthedispersalofH. comosaviaepizoochory(Müller-Schneider,1938)fromSouthernFrancetoGermany.Thisassumptiondemonstratesthatamoreinterdisciplinaryapproachtothissubjectisnecessary inorder to fullyunderstand the recolonizationofCentralEuropebyplants,animals,orhumans.Astrengthenedcooperationofphylogeneticsandarcheologycoulddeliverintriguingnewinsightsinthegenesisofourenvironment.
5 | CONCLUSIONS
Basedonthepresentstudyonthepreviousandcurrentdistribu-tion of H. comosa, we could demonstrate that a comprehensiveclimatic approach including a second driving factor can lead toa better understanding of historical and present developments.The traditional latitudinal temperature gradient as major param-eterwasextendedbyalongitudinalhumiditygradientwhichbothworkintandemdefiningthesuitablehabitatsduringtheLGM.Ofthe two detected clearly distinguished phylogeographic clusters,one inWestern Europe ranging from Spain to Germany and theotherembracingpartsofsouthEasternEurope,onlythewesternrefugiacontributedtotherecolonizationofCentralEurope.AstheclimatebecamedrierinCentralEuropeduringtheLGM,H. comosa evaded tomoister climates, which prevailed inWestern Europe.Theresultsofoursurveyfurthermoreprovideevidencefornorth-ernlocationsinFranceandtheUKthatcouldhaveservedascryp-tic refugia. For the postglacial recolonization, H. comosa couldhave benefited from habitats shapes by humans and zoochory,which provided a long ranged dispersal to Central Europe. Thusintegrated approaches incorporating multidisciplinary knowledgemightbethebestwaytoapproximateandilluminatehistoricalandpresentprocesses.
ACKNOWLEDGMENTS
We would like to thank John Dickie, Langlois Estelle, ChristineFrohnauer, Melanié Harzé, Daniela Listl, Anton Mayer, RichardMichalet,NicolaSchoenenberger,SebastianTeufel,andMolnarZsoltforcollectingplantmaterialfromsinglelocationsandPetraSchitkoforherassistanceinthelaboratory.
Beebee,T.J.C.,&Rowe,G. (2000).Microsatellite analysis of natterjacktoad Bufo calamita Laurenti populations: Consequences of dispersalfromaPleistocene refugium.Biological Journal of the Linnean Society,69,367–381.
Bennett,K.D.,Tzedakis,P.C.,&Willis,K.J.(1991).QuaternaryRefugiaofNorthEuropeanTrees.Journal of Biogeography,18,103–115.
Bhagwat,S.A.,&Willis,K.J.(2008).SpeciespersistenceinnortherlyglacialrefugiaofEurope:Amatterofchanceorbiogeographicaltraits?Journal of Biogeography,35,464–482.
Boston,E. S.M.,Montgomery,W. I.,Hynes,R.,&Prodohl, P.A. (2015).NewinsightsonpostglacialcolonizationinwesternEurope:Thephylo-geographyoftheLeisler’sbat(Nyctalus leisleri).Proceedings of the Royal Society of London Series B: Biological Sciences,282:20142605.
Braconnot, P.,Harrison, S. P., Kageyama,M., et al. (2012). Evaluation ofclimate models using palaeoclimatic data.Nature Climate Change, 2,417–424.
Bush,M.B.(1988).EarlyMesolithicdisturbance:Aforceonthelandscape.Journal of Archaeological Science,15,453–462.
Bylebyl,K.,Poschlod,P.,&Reisch,C.(2008).GeneticvariationofEryngium campestre L. (Apiaceae) in Central Europe. Molecular Ecology, 17,3379–3388.
Comes,H.P.,&Kadereit,J.W. (1998).Theeffectofquaternaryclimaticchangesonplantdistributionandevolution.Trends in Plant Science,3,432–438.
Cornish,M.W.(1954).TheoriginandstructureofthegrasslandtypesofthecentralNorthDowns.The Journal of Ecology,42,359–374.
Dengler, J., Janisova, M., Torok, P., & Wellstein, C. (2014). Biodiversityof Palaearctic grasslands: A synthesis. Agriculture Ecosystems & Environment,182,1–14.
Dutoit,T., &Alard,D. (1995). Permanent seed banks in chalk grasslandunder various management regimes – their role in the restorationof species-rich plant-communities. Biodiversity and Conservation, 4,939–950.
Elith,J.,Graham,C.H.,Anderson,R.P.,etal. (2006).Novelmethods im-prove prediction of species’ distributions from occurrence data.Ecography,29,129–151.
Elith,J.,&Leathwick,J.R.(2009).Speciesdistributionmodels:Ecologicalexplanation and prediction across space and time.Annual Review of Ecology, Evolution, and Systematics,40,677.
Ellenberg, H. (1988). Vegetation ecology of central Europe. Cambridge:CambridgeUniversityPress.
Ellenberg,H.,&Leuschner,C.(2010).Vegetation Mitteleuropas mit den Alpen in ökologischer, dynamischer und historischer Sicht.Stuttgart:Ulmer.
Excoffier, L., Smouse,P.E.,&Quattro,J.M. (1992).Analysisofmolecu-lar variance inferred frommetric distances amongDNA haplotypes:Application to human mitochondrial DNA restriction data.Genetics,131,479–491.
Fielding,A.H.,&Bell, J. F. (1997).A reviewofmethods for the assess-mentof prediction errors in conservationpresence/absencemodels.Environmental Conservation,24,38–49.
Fischer,S.F.,Poschlod,P.,&Beinlich,B. (1996).Experimentalstudiesonthedispersalofplantsandanimalsonsheepincalcareousgrasslands.Journal of Applied Ecology,33(5),1206–1222.
Fjellheim, S., Rognli, O. A., Fosnes, K., & Brochmann, C. (2006).Phylogeographicalhistoryofthewidespreadmeadowfescue(Festuca pratensisHuds.) inferred fromchloroplastDNAsequences.Journal of Biogeography,33,1470–1478.
Flojgaard,C.,Normand,S.,Skov,F.,&Svenning,J.C.(2009).Iceagedistri-butionsofEuropeansmallmammals:Insightsfromspeciesdistributionmodelling.Journal of Biogeography,36,1152–1163.
GBIF (Global Biodiversity Information Facility).GBIF Backbone Taxonomy. http://www.gbif.org/species/5353750.
Gronenborn,D.(2003).Migration,acculturationandculturechangeinwest-ern temperate Eurasia, 6500–5000 cal BC.Documenta Praehistorica,30,79–91.
Hartmann, J., andN.Moosdorf (2012). The new global lithologicalmapdatabase GLiM: A representation of rock properties at the Earthsurface, Geochemistry, Geophysics, Geosystems, 13, Q12004,doi:10.1029/2012GC004370.
Hejcman, M., Hejcmanova, P., Pavlu, V., & Benes, J. (2013). Origin andhistory of grasslands inCentral Europe– a review.Grass and Forage Science,68,345–363.
Hewitt,G.M.(2004).GeneticconsequencesofclimaticoscillationsintheQuaternary. Philosophical Transactions of the Royal Society of London Series B: Biological Sciences,359,183–195.
Hijmans,R.J.,Cameron,S.E.,Parra,J.L.,Jones,P.G.,&Jarvis,A.(2005).Veryhighresolutioninterpolatedclimatesurfacesforgloballandareas.International Journal of Climatology,25,1965–1978.
Hijmans,R.J.,&Graham,C.H.(2006).Theabilityofclimateenvelopemod-els to predict the effect of climate change on species distributions.Global Change Biology,12,2272–2281.
Huntley,B.,&Birks,H.J.B.(1983).An atlas of past and present pollen maps for Europe: 0–13000 Years Ago.Cambridge:CambridgeUniversityPress.
Janišová,M.,Bartha,S.,Kiehl,K.,&Dengler,J.(2011).Advancesintheconser-vationofdrygrasslands:IntroductiontocontributionsfromtheseventhEuropean Dry GrasslandMeeting. Plant Biosystems – An International Journal Dealing with all Aspects of Plant Biology,145,507–513.
Kahmen,S.,&Poschlod,P.(2004).Plantfunctionaltraitresponsestograss-landsuccessionover25years.Journal of Vegetation Science,15,21–32.
Kahmen, S., Poschlod, P., & Schreiber, K.-F. (2002). Conservation man-agement of calcareous grasslands. Changes in plant species com-position and responseof functional traits during25years.Biological Conservation,104,319–328.
Karlik, P.,&Poschlod, P. (2009).History or abiotic filter:Which ismoreimportantindeterminingthespeciescompositionofcalcareousgrass-lands?Preslia,81,321–340.
Korneck,D.,Schnittler,M.,Klingenstein,F.,etal. (1998).WarumverarmtunsereFlora?AuswertungderRotenListederFarn-undBlütenpflanzenDeutschlands.Schriftenreihe für Vegetationskunde,29,299–444.
Kunes,P.,Pelankova,B.,Chytry,M.,etal.(2008).Interpretationofthelast-glacialvegetationofeastern-centralEuropeusingmodernanaloguesfromsouthernSiberia.Journal of Biogeography,35,2223–2236.
Lang, G. (1994). Quartäre Vegetationsgeschichte Europas: Methoden und Ergebnisse.Heidelberg:SpektrumAkademischerVerlag.
Magri,D.,Vendramin,G.G.,Comps,B.,etal.(2006).AnewscenariofortheQuaternary history of European beech populations: Palaeobotanicalevidenceandgeneticconsequences.New Phytologist,171,199–221.
Normand, S., Ricklefs, R. E., Skov, F., et al. (2011). Postglacialmigrationsupplements climate in determining plant species ranges in Europe.Proceedings of the Royal Society of London Series B: Biological Sciences,278,3644–3653.
vonOheimb,G.,Schmidt,M.,Kriebitzsch,W.-U.,&Ellenberg,H. (2005).DispersalofvascularplantsbygameinnorthernGermany.PartII:Reddeer (Cervus elaphus).European Journal of Forest Research,124,55–65.
Ohlemüller, R., Huntley, B., Normand, S., & Svenning, J. C. (2012).Potential source and sink locations for climate-driven species rangeshifts in Europe since the LastGlacialMaximum.Global Ecology and Biogeography,21,152–163.
Pakeman, R. J. (2001). Plant migration rates and seed dispersal mecha-nisms.Journal of Biogeography,28,795–800.
Pärtel,M.,Bruun,H.H.,&Sammul,M. (2005).Biodiversity intemperateEuropean grasslands: Origin and conservation: 13th InternationalOccasional Symposium of the European Grassland Federation. In:Grassland Science in Europe, pp. 1–14. In: Integrating EfficientGrassland Farming and Biodiversity: Proceedings of the 13thInternational Occasional Symposium of the European GrasslandFederation.
Paun,O.,Schonswetter,P.,Winkler,M.,Tribsch,A.,&Consortium,I.(2008).Historical divergence vs. contemporary gene flow: Evolutionary his-toryofthecalcicoleRanunculus alpestrisgroup(Ranunculaceae)intheEuropeanAlpsandtheCarpathians.Molecular Ecology,17,4263–4275.
Peakall,R.,&Smouse,P.E.(2012).GenAlEx6.5:GeneticanalysisinExcel.Population genetic software for teaching and research – an update.Bioinformatics,28,2537–2539.
Pearson,R.G.,&Dawson,T.P. (2003).Predictingthe impactsofclimatechangeonthedistributionofspecies:Arebioclimateenvelopemodelsuseful?Global Ecology and Biogeography,12,361–371.
Pépin,D.,Adrados,C.,Mann,C.,&Janeau,G.(2004).AssessingrealdailydistancetraveledbyungulatesusingdifferentialGPSlocations.Journal of Mammalogy,85,774–780.
Petit, R. J.,Aguinagalde, I., de Beaulieu, J. L., et al. (2003). Glacial refu-gia:Hotspotsbutnotmeltingpotsofgeneticdiversity.Science,300,1563–1565.
Poschlod,P.(2014).Geschichte der Kulturlandschaft.Stuttgart:Ulmer.Poschlod,P.,Baumann,A.,&Karlik,P.(2009).Originanddevelopmentof
grasslandsinCentralEurope.InP.Veen,R.Jefferson,J.deSmidt,&J.Straaten(Eds.),Grasslands in Europe – of high nature value(pp.15–25).Utrecht,NL:KNNVPublishing.
Poschlod,P.,&WallisDeVries,M.F. (2002).Thehistorical and socioeco-nomicperspectiveofcalcareousgrasslands–lessonsfromthedistantandrecentpast.Biological Conservation,104,361–376.
Rogers,S.O.,&Bendich,A.J.(1994).ExtractionoftotalcellularDNAfromplants, algae and fungi. Plant Molecular Biology Manual, (183–190),SpringerNetherlands.
Römermann, C., Bernhardt-Römermann, M., Kleyer, M., & Poschlod, P.(2009).Substitutes forgrazing insemi-naturalgrasslands–domow-ingormulchingrepresentvaluablealternativestomaintainvegetationstructure?Journal of Vegetation Science,20,1086–1098.
Rosch,M.(1998).Thehistoryofcropsandcropweedsinsouth-westernGermany from the Neolithic period to modern times, as shown byarchaeobotanical evidence. Vegetation History and Archaeobotany, 7,109–125.
Schmidt,M.,Fischer,P.,&Becker,C.(2007).ZurHerkunftvonPflanzenartenanthropo-zoogener Kalk-Halbtrockenrasen Mitteleuropas-Überlegungen am Beispiel des nördlichen deutschen Hügel-undBerglandes.Hercynia- Ökologie und Umwelt in Mitteleuropa,40,257–267.
Schmitt,T.,&Varga,Z.(2012).Extra-Mediterraneanrefugia:Theruleandnottheexception?Frontiers in Zoology,9: 22.
Stewart, J. R., & Lister, A. M. (2001). Cryptic northern refugia andthe origins of the modern biota. Trends in Ecology & Evolution, 16, 608–613.
Stewart,J.R.,Lister,A.M.,Barnes,I.,&Dalen,L.(2010).Refugiarevisited:Individualistic responsesof species in spaceand time.Proceedings of the Royal Society of London B: Biological Sciences,277,661–671.
| 1933LEIPOLD Et aL.
Sutkowska,A.,Pasierbinski,A.,Warzecha,T.,Mandal,A.,&Mitka,J.(2013).RefugialpatternofBromus erectusincentralEuropebasedonISSRfin-gerprinting.Acta Biologica Cracoviensia Series Botanica,55,107–119.
Svenning,J.C.,Normand,S.,&Kageyama,M.(2008).Glacialrefugiaoftem-perate trees in Europe: Insights from species distributionmodelling.Journal of Ecology,96,1117–1127.
Taberlet, P., Fumagalli, L., Wust-Saucy, A. G., & Cosson, J. F. (1998).Comparative phylogeography and postglacial colonization routes inEurope.Molecular Ecology,7,453–464.
Tzedakis,P.C.,Emerson,B.C.,&Hewitt,G.M.(2013).Crypticormystic?Glacial tree refugia innorthernEurope.Trends in Ecology & Evolution,28,696–704.
Willis, K. J.,&Whittaker, R. J. (2002). Species diversity – scalematters.Science,295,1245–1248.
Wisz,M.S.,Hijmans,R.,Li,J.,etal.(2008).Effectsofsamplesizeontheperformanceofspeciesdistributionmodels.Diversity and Distributions,14,763–773.
Yeh,F.C.,Yang,R.,&Boyle,T.(1999).POPGENE Version 1.31 – Microsoft Window-based freeware for population genetic analysis – quick user guide. Canada:UniversityofAlberta.
How to cite this article:LeipoldM,TauschS,PoschlodP,ReischC.SpeciesdistributionmodelingandmolecularmarkerssuggestlongitudinalrangeshiftsandcrypticnorthernrefugiaofthetypicalcalcareousgrasslandspeciesHippocrepis comosa (horseshoevetch).Ecol Evol.2017;7:1919–1935.https://doi.org/10.1002/ece3.2811
F IGURE A1 DistributionmodelofH. comosainEuropebasedoncurrentclimatedata.Darkgrayareasindicatesuitablehabitatswithintheecologicalniche;lightgrayareaareunsuitablehabitatsforH. comosa.BlackdotsshowtheoccurrencedatagatheredfromGBIFandusedformodeling.Nationalboundariesaregiven
| 1935LEIPOLD Et aL.
F IGURE A2 SpeciesdistributionmodelprojectionofH. comosaatthelastglacialmaximumbasedontheoutputoftheCCSM4scenario.Darkgrayareasindicatesuitablehabitatswithintheecologicalniche;lightgrayareaareunsuitablehabitatsforH. comosa.Iceshieldsareshowninwhitewithadarkoutline.Nationalboundariesrepresenttoday’sEuropeanlandarea
F IGURE A3 AnalysisoftheSTRUCTUREruns.Relationshipbetweenthenumberofproposedgroupsk,ΔkandLn(k),respectively.ThemostlikelynumberofgroupswithhighestΔk=684.8wasfoundwithk=2