1

SupportingInformation(SI)Appendix

PredictableallelefrequencychangesduetohabitatfragmentationintheGlanville

fritillarybutterfly

TobyFountain,MarkoNieminen,JukkaSirén,SweeChongWongandIlkkaHanski

DNAextraction.Inthecaseoffield-collectedspecimens,larvaltissuewashomogenizedpriorto

extractionusingTissueLyser(Qiagen)at30/sfor1.5minswithTungstenCarbideBeads,3mm

(Qiagen).DNAwasextractedusingtheNucleoSpin96TissueCoreKit(Macherey-Nagel).Where

DNAyieldwaslow,extractedDNAunderwenttworoundsofWholeGenomeAmplification(WGA)

(LGCGenomics).Inthecaseofmuseumspecimens,DNAwasextractedfromaleg(inafewcases

fromasmallwingbiopsy)usingtheQIAmpMicrokit(Qiagen).WhereDNAyieldwaslow,an

additionalextractionwasperformedandtheproductsofthetwoextractionswerepooled.To

furtherincreaseDNAyield,WGAwasperformedusingtheGenomePlexCompleteWholeGenome

AmplificationKit(Sigma).AfterWGA,sampleswerecleanedusingQIAquickPCRpurificationkit

(Qiagen)tooptimizedownstreamPCRperformance.Sterilemethods,andpositiveandnegative

controls,wereusedtoensurenocross-contaminationbetweenmuseumandcontemporary

samples.

InitialSNPselectionandvalidation.SNPmarkerswereselectedfromcandidategenes,putatively

neutralregionsofthegenome,andotherwiseuncoveredchromosomes(seemaintext).SNP

callingwasperformedontheRNA-seqdata,including40unrelatedindividualssampledacrossthe

ÅlandIslands,usingthe“mpileup”functionfrom“SAMtools”package(1)withthedefault

parametervalues.These40ÅlandindividualswereonlyusedforSNPvalidationandwerenot

includedinthesubsequentanalyses.SNPswithminorallelefrequency(MAF)>0.2,callrate>0.9,

andSNPqualityscore>100wereretainedascandidateSNPs,whichweremappedtothegenome

(2)toobtainthecorrespondinggenesandexons.ASNPwasexcludediftherewerelessthan60

nucleotidesflankingbothupstreamanddownstreamfromtheSNPintheexon.Thegap-filling

SNPswerefromcodingregions,andtheyweresubjectedtothesamefilteringcriteriaasthe

candidategenes.TheneutralSNPswereobtainedfromtheSOLiDmatepair-1genomesequences

(2)usinganin-houseSNPcallingmethod(3,4).Asthegenomicsequenceswereattainedby

sequencingonlyonemaleindividual,onlyheterozygotesSNPsspanningacrossall31

chromosomesfromnon-codingregionswereselected.SNPsfromcandidategenesandnon-coding

2

regionswerecombinedintotheinitialcandidateSNPset.A121base-pair(bp)flankingregionwas

extractedforeachSNPintheinitialSNPset.IfthereareanyflankingSNPslocatedwithinboththe

upstreamandthedownstreamregionsforthecandidateSNP,theSNPwasexcluded.Thestepwas

adheredtoensureacleanregionforprimersandprobesdesign.ThefilteredSNPsweresubmitted

toLGCGenomicsforprimerdesignandinsilicotesting.BLASTwasperformedfortheprimerpairs

withthereferencegenometoconfirmthattheprimersbindonlytotheregionwherethe

correspondingSNPislocated.SNPsthatdidnotfulfillthiscriterionwereexcluded.TheSNPsthat

passedallthequalitycontrolstepswereselectedforvalidation.Avalidationpanelwas

constructedusing48individualsfromeightfamiliessampledacrosstheÅlandIslandsin2007-11.

ASNPpassedthevalidationstepifitproducedclearlydefinedgenotypeclustersinascatterplot

withlessthantwoMendelianerrors,andhadahighSNPcallrate(>0.9).Followingthevalidation

process,320SNPs(18SNPsfrompriorstudies,182SNPsfrom164candidategenes,15SNPsfrom

sexchromosomalscaffolds,45neutralSNPs,and60gapfillingSNPs)wereselectedasthefinal

genotypingpanelimplementingKASPchemistry(LGCGenomics).

Potentialascertainmentbiasinmuseumsamples.Onepotentiallimitationofthisdatasetisthat

thecandidateSNPswerevalidatedwiththecontemporaryÅlandsamplesonly,introducingthe

possibilityofascertainmentbiasinthemuseumsamples.Thereasonisthatgeneticdifferences

betweentheothersamplesandtheÅlandsamplecouldreducemarkerperformanceandlevelsof

polymorphisminmoredistantpopulations.However,suchabiasisveryunlikelyinthepresent

case,forseveralreasons.AllmarkerswerepolymorphicintheÅlandmuseumsamples,and>94%

ofthemarkerswerepolymorphicinSWFinnishmuseumsamples.Thealternativealleleatloci

thatweremonomorphicintheSWFinnishsamplestendedtobeatverylowfrequencyinthe

museumÅlandsamples.Moreover,theSWFinnishsamplesshareancestralgeneticvariationwith

thecontemporaryÅlandmetapopulation,andthenowextinctSWFinnishpopulationswere

isolated.

Validationofmuseumsamplegenotypes.Totestrepeatabilityofgenotypingasubsetofthe

museumsamples(n=8)wasgenotypedforasubsetofSNPsusingSequenomiPLEXGold

genotypingplatform.Oneindividualwasgenotypedat20SNPs,sixindividualsat14SNPs,andone

individualatsevenSNPs.IncaseswhereasamplewassuccessfullycalledinbothSequenomand

KASPgenotypeconcordancewas92.1%(70/76genotypesimilarity).Moreover,weexaminedthe

relationshipbetweenthecallrateandthelevelofheterozygositytoensurethatreduced

3

heterozygosityinmuseumsamplesisnotaresultoflowersamplequalityofmuseumthan

contemporarysamples(Fig.S7).Thereisgeneralreductionofheterozygositywhencallrate

decreasesinbothcontemporaryandmuseumsamples,butatthehighestcallrates(>0.8),SW

FinnishmuseumsampleshavethelowestheterozygositycomparedwithmuseumÅland(Tukey

test,P<0.0001)andcontemporaryÅland(Tukeytest,P<0.0001)samples.Asanalternative

analysis,wetestedthedifferencesbetweenthepopulationsinthefulldatasetwhileincludingcall

rateasacovariate.Inthisanalysis,thepopulationeffectwashighlysignificant(population:F=33.09,P<0.0001,callrate:F=74.29,P<0.0001),withSWFinlandhavingsignificantlylower

heterozygositythanbothcontemporaryÅland(Tukeytest,P<0.0001)andmuseumÅland

samples(Tukeytest,P<0.0001).Nonetheless,toavoidintroducinganypotentialbiasduetolow

samplequalityweonlyretainedindividualswithanaveragecallrate>0.8acrossallthe222SNPs.

FSTvalues.Usingthe222loci,wecalculatedtheFSTvalue(5)betweeneachmuseumspecimenand

thecontemporarySaltvikpopulation,sampledin2007(n=530).TheFSTvaluewasusedto

computetheprobabilityforeachmuseumspecimen(genotype)separately,usingtheobserved

allelefrequenciesintheÅlandpopulationastheexpectationandauniformpriordistribution

between0and1.TheposteriormeanoftheFSTwasusedasanestimateoftheevolutionary

distanceofthespecimenfromthereferencepopulation.Theeffectsofyearofsampling,

populationtypeandmarkertype(candidatevsneutral)weretestedusinglinearmodelsinR.We

selectedaprioriasetofbiologicallyplausiblemodelsandusedthefunctionmodel.selinthe

packageMuMInv.1.13.4torankthemodelsbasedontheirAkaikeinformationcriterionforfinite

samplesizes(AICc)(6,7).

Allelefrequencychangesintheoutlierloci.Tocharacterizeallelefrequency(AF)changesdueto

populationturnoverwecalculatedthedifferenceinallelefrequenciesbetweennewly-established,

isolatedpopulations(AF(new))andold,well-connectedpopulations(AF(old)).Thedifference

AF(new)-AF(old)iscorrelatedwithAF(old)(P=0.06;Fig.S6).Wethereforerepeatedtheanalysis

afterremovingthisbiasbyregressingAF(new)-AF(old)againstAF(old),andusedtheresidualfrom

thisregressionasthemeasureofallelefrequencydifferencesbetweenthepopulationtypes.

Similarly,wecalculatedcorrespondingresidualsforthevariablesAF(Sottunga)-AF(old)andAF(SW

Finland)-AF(old)inFig.4andFig.S5.Inallcases,theresultswereonlylittleaffectedbythis

correctionandtheconclusionsremainedunchanged(TableS6).Forsimplicity,weshowtheresults

4

basedontheuncorrectedvaluesinthemaintext,withtheexceptionoftheanalysisassociated

withFigS5inwhichcorrectedvalueswereusedduetoverysmallsamplesize.

Tocharacterizeallelefrequenciesintheoutlierlociinbutterfliesfromfragmentedvs

continuouslandscapes,weextractedtheallelefrequenciesfromanRNA-seqdataset(8)forthe

fourregionalpopulationsinFig.1.Astherewerefourregionalpopulations,twoofeachlandscape

type,wesummarizedvariationinallelefrequencieswithaprincipalcomponentanalysis(Table

S4).PC2,whichexplained33%oftotalvariance,wasstronglycorrelatedwithlandscapetype

(TableS4).PC2wascorrelated,thoughnotsignificantly,withcorrectedAF(Sottunga)–AF(old)(R2

=0.14,P=0.17)andcorrectedAF(SWFinland)–AF(old)(R2=0.29,P=0.08).Tocombinethese

twoanalyses,weranaprincipalcomponentanalysisonAF(SWFinland)-AF(old)andAF(Sottunga)-

AF(old)andusedPC1astheaverageallelefrequencychangeinthetwopopulations.PC1

accountedfor79%ofthetotalvariance.ThreeoutlierlociwerenotavailablefromtheRNA-seq

datasetandwerethereforeexcludedfromthisanalysis.

References

1. LiH,etal.(2009)Thesequencealignment/mapformatandSAMtools.Bioinformatics25(16):2078–2079.

2. AholaV,etal.(2014)TheGlanvillefritillarygenomeretainsanancientkaryotypeandrevealsselectivechromosomalfusionsinLepidoptera.NatComms5:1–9.

3. RastasP,PaulinL,HanskiI,LehtonenR,AuvinenP(2013)Lep-MAP:fastandaccuratelinkagemapconstructionforlargeSNPdatasets.Bioinformatics29(24):563–3134.

4. KvistJ,etal.(2015)FlightinducedchangesingeneexpressionintheGlanvillefritillarybutterly.MolEcol24(19):4886-4900.

5. GaggiottiOE,FollM(2010)QuantifyingpopulationstructureusingtheF-model.MolEcolResour10(5):821–830.

6. BartońK(2015)MuMIn:Multi-modelinference.Rpackagev.1.13.4.https://cran.r-project.org/web/packages/MuMIn/MuMIn.pdf.

7. JohnsonJB,OmlandKS(2004)Modelselectioninecologyandevolution.TrendsEcolEvol19(2):101–108.

8. SomervuoP,etal.(2014)Transcriptomeanalysisrevealssignatureofadaptationtolandscapefragmentation.PlosOne9(7): e101467

5

TableS1:FSTvaluesforthespatio-temporallypooledsamples(seeMaterialandMethodsinmaintext)calculatedusingallthe222SNPs(lowerdiagonal).Significantvalues(P<0.05)arehighlightedinbold.Valuesbasedonthe34neutralSNPs(upperdiagonal)gavequalitativelysimilarresults.

Åland1905 Åland1945 Åland1965 Saltvik2010 Sottunga2010 SWFinland1900 SWFinland1940 SWFinland1965

Åland1905 0.033 0.043 0.014 0.074 0.013 0.148 0.227Åland1945 0.012 0.014 0.035 0.052 -0.003 0.142 0.177Åland1965 0.021 0.013 0.054 0.094 0.009 0.168 0.204Saltvik2010 0.014 0.026 0.040 0.075 -0.022 0.134 0.197Sottunga2010 0.051 0.059 0.074 0.056 -0.053 0.147 0.191SWFinland1900 0.060 0.041 0.050 0.050 0.045 -0.043 -0.015SWFinland1940 0.131 0.116 0.122 0.125 0.149 0.129 0.059SWFinland1965 0.204 0.165 0.190 0.182 0.198 0.226 0.136

6

TableS2.ThesetofaprioriselectedmodelsexplainingFSTvaluesinFig.S1.Kisthenumberofparametersinthemodel.ModelsarerankedbasedonthedifferenceinAICcvalues(ΔAICc)betweenthefocalmodelandthebestmodelintheset.Akaikeweightsreflectthelikelihoodofamodelrelativetoallothermodelsintheset.

model K AICc ΔAICc Akaikeweight

year+pool+markertype+markertype*pool 6 -238.7 0 0.409year+pool+markertype+markertype*pool+year*pool 7 -238.3 0.44 0.328year+pool+markertype 5 -236.2 2.49 0.118year+pool+markertype+year*pool 6 -235.8 2.95 0.094pool+markertype 4 -234.5 4.16 0.051year+pool 4 -223.1 15.59 0.000pool 3 -221.7 16.99 0.000type 3 -181.1 57.61 0.000year+markertype 4 -179.7 58.99 0.000interceptonly 2 -172.0 66.72 0.000year 3 -170.6 68.12 0.000

7

TableS3.Asummaryoftheresultsfromthethreeanalysesinvolvingtheoutlierloci.Inthecaseoftherepeatabilityanalysisandcomparisonwithpopulationturnoverrate,allelefrequencydifferencebetweenthefocalpopulationandtheoldlocalpopulationsfromtheSaltvikreferencepopulationaregiven.Inthecomparisoninvolvingthedegreeoffragmentationatthelandscapelevel,thetwoprincipalcomponentsusedintheanalysisaregiven.Positivevaluesarehighlightedingreen,negativevaluesarehighlightedinred.

OutlierSNPs SWFinland Sottunga SWFinland New Sottunga SWFinland/Sottunga(pc1) Fragmentedlandscapes(pc2)

Mc1:1041:122591 0.462 0.066 0.462 -0.040 0.066 1.176 1.870

Mc1:2666:34531 0.518 0.010 0.518 -0.040 0.010 0.852 -0.663

Mc1:2966:24907 -0.384 0.076 -0.384 -0.135 0.076 0.604 0.011

Mc1:1061:35594 -0.052 -0.259 -0.052 0.049 -0.259 -1.199 0.534

Mc1:1873:36910 -0.585 -0.439 -0.585 -0.171 -0.439 -2.006 -1.853

Mc1:2025:177786 -0.382 -0.197 -0.382 -0.073 -0.197 -0.448 0.639

Mc1:1124:71239 -0.244 -0.055 -0.244 0.040 -0.055 -1.115 -1.375

Mc1:1206:26737 0.238 0.096 0.238 -0.009 0.096 NA NA

Mc1:752:33517 0.196 -0.296 0.196 0.008 -0.296 0.351 0.199

Mc1:2673:141336 0.745 0.323 0.745 0.136 0.323 1.785 0.639

Mc1:1129:26699 -0.252 -0.144 -0.252 -0.195 -0.144 NA NA

Mc1:658:68226 0.091 -0.193 0.091 -0.092 -0.193 NA NA

Repeatability(Fig.3) Habitatfragmentation(Fig.S5)Populationturnover(Fig.4)

8

TableS4.Correlationsbetweenthefirstfourprincipalcomponentsandtheallelefrequenciesinthefourregionalpopulationsinhabitingeitherfragmented(bold)orcontinuouslandscapes.

TableS5.PairsofSNPswithsignificantLD(P<0.05afterFDR).KASPID,chromosomeandpositionareshown.

TableS6.Resultsfromlinearmodelsofallelefrequencydifferencesusingvaluescorrectedversusnotcorrectedforaweakandnon-significantcorrelationwithAF(old)(seeSectionAllelefrequencychangesintheoutlierlociinSIAppendix).

PC1 PC2 PC3 PC4

Åland -0.43 0.61 -0.45 -0.49Uppland -0.46 0.43 0.73 0.27Öland -0.62 -0.28 -0.44 0.59Saaremaa -0.47 -0.61 0.27 -0.58

Proportionofvariance 0.47 0.33 0.16 0.04

Locus#1 Chromosome Position Locus#2 Chromosome Position

Åland

K3-82 1 1620 K4-5 24 100694K2-111 2 194473 K5-7 2 192234K2-25 3 5112 K3-17 3 15952K3-12 13 16716 K3-62 8 103903K2-60 13 37262 K3-150 13 33875K3-185 15 10977 K8-82 10 47209K3-192 15 73713 K5-133 15 70120K2-79 17 168063 K2-80 17 166435K2-86 25 24287 K2-88 25 19949K2-54 25 15219 K2-86 25 24287K3-162 NA 1329 K5-128 10 13465

SWFinland

K3-134 4 2340 K3-137 4 3654

R2 P R2 P R2 P R2 P

Uncorrected 0.36 0.02 0.14 0.13 0.36 0.02 0.15 0.16Corrected 0.30 0.04 0.05 0.24 0.30 0.04 0.30 0.07

Repeatability(Fig.3) Habitatfragmentation(Fig.S5)

SWFinland-Sottunga Sottunga-New SWFinland-New

Populationturnover(Fig.4)

SWFinland/Sottunga

9

Fig.S1.FSTvaluesofeachmuseumspecimencomparedtoSaltvikreferencepopulation(in2007)plottedagainstthetimeofsampling,separatelyforcandidate(solidtriangles)andneutralmarkers(opentriangles)for(a)Ålandand(b)SWFinnishmuseumsamples.TheregressionlinesoftheFSTvaluesagainsttimeareplottedseparatelyforthetwomarkertypes(solidlineforcandidatemarkers,dashedlineforneutralmarkers).FortheanalysisseeTableS2.

Fig.S2.Theallelefrequencyshiftsofoutlierloci(n=12)intheSWFinnishpopulationinrelationtocontemporaryÅlandsamplesaresignificantlycorrelatedwiththecorrespondingallelefrequencyshiftsintheSottungapopulationinrelationtocontemporaryÅlandsamples(R2=0.36,P=0.02)

1880 1920 1960 2000

0.0

0.2

0.4

0.6

0.8

1.0

Aland

Year

Drif

t val

ue

1880 1920 1960 2000

0.0

0.2

0.4

0.6

0.8

1.0

SW Finland

Year

Drif

t val

ueYear Year

F ST

F ST

a) b)

AllelefrequencydifferencebetweenSo4ungaandÅland

Allelefreq

uencydiffe

rencebe

tween

SWFinland

and

Åland

10

Fig.S3.Allelefrequencydifferencesofneutralloci(n=34)betweensamplesfromnowextinctpopulationsinSWFinland(black)andtheintroducedmetapopulationinSottunga(white),comparedtoold,well-connectedlocalpopulationsinSaltvik.Dashedlinesshowlociwithallelefrequencieslessthan0.2orgreaterthan0.8intheoldlocalpopulations.

Allelefreq

uencydiffe

rencefrom

contem

poraryÅland

AllelefrequencyincontemporaryÅland

11

Fig.S4.Allelefrequencyshiftsofneutralloci(n=20)intheSWFinnishmuseumsamplesandinSottungainrelationtothecontemporaryÅlandsamples.AstherewasahighlysignificantrelationshipbetweentheallelefrequencydifferenceAF(Sottunga)-AF(old)inrelationtoAF(old)intheneutralmarkers(R2=0.22,P=0.003),allelefrequencychangesareexpressedasresidualsfromalinearregressionofAF(Sottunga)-AF(old)againstAF(old)andAF(SWFinland)-AF(old)againstAF(old)(seetextinSI).Locithathadallelefrequenciesgreaterthan0.2orlessthan0.8inoldwell-connectedpopulationsareincluded(Fig.S3).Theremainingmarkerswereexcludedastheyfelloutsidetheminorallelefrequencycutoffsusedintheselectionofcandidatemarkers.Thecorrelationisnotsignificant(R2=-0.004,P=0.35).

Allelefreq

uencydiffe

rence

betw

eenSW

Finland

and

Åland

AllelefrequencydifferencebetweenSo8ungaandÅland

12

Fig.S5.Thesecondprincipalcomponentoftheallelefrequenciesintheoutliers(n=9)infourregionalpopulations(Fig.1)plottedagainstthefirstprincipalcomponentoftheallelefrequencydifferencesbetweenSottungaandSWFinlandfromoldlocalpopulationsintheSaltvikreferencepopulation(forthecalculationseeAllelefrequencychangesintheoutlierlociinSItext).DataforthreeoutlierswerenotavailableforallthefourregionalpopulationsintheRNA-seqdataset,hencen=9.PC2ontheverticalaxisispositivelycorrelatedwithhabitatfragmentationintheregionalpopulations(TableS4).Theregressionisclosetosignificant(R2=0.30,P=0.07).

●

●

●

●

●

●

●

●

●

−2 −1 0 1−2−1

01

2

pc1cor

pc2

Allelefrequencychangeinisolatedmetapopula6onandex6nctpopula6ons(PC1)

Allelefreq

uenciesc

haracterizingfragmen

ted

land

scapes(P

C2)

13

Fig.S6.Thereisanegativerelationshipintheoutlierloci(n=13)betweentheallelefrequencydifferenceAF(new)-AF(old)andAF(old)(R2=0.22,P=0.06).Toassesswhetherthiswasinfluencingourresultswerepeatedtheanalysesonallelefrequenciesusingresidualsfromthisregression(seetextinSI).Theresultsremainedqualitativelythesameandtheconclusionswerenotaffected(TableS6).

Fig.S7.Therelationshipbetweenthecallrateandtheproportionofheterozygousloci.SamplesaresplitamongstthetwoKASPgenotypingplatesthatcontainedmuseumsamples.FortheanalysesseethetextinSI.

AllelefrequencyincontemporaryÅland

Allelefreq

uencydiffe

renceinold

comparedtonew

pop

ula6

ons

Plate&15,&museum&Åland&Plate&15,&museum&SW&Finland&

Plate&16,&museum&Åland&Plate&16,&museum&SW&Finland&

Plate&16,&contemporary&So:unga&Plate&16,&contemporary&Åland&

Call&Rate&

Prop

or>o

n&of&heterozygou

s&loci&