Morphologicalconsequencesofrangefragmentation andpopulation declineonthe endangeredIberianlynx (Lynxpardinus)
C.Pertoldi1,2,3*,R.Garc´ıa-Perea4*,J.A.Godoy2,M.Delibes2 V.Loeschcke1,5
1DepartmentofEcologyandGenetics,InstituteofBiologicalSciences,UniversityofAarhus,NyMunkegade,AarhusC,Denmark
2DepartmentofAppliedBiology,Estacio´nBiolo´gicadeDon˜ana,CSIC.Pabello´ndelPeru´,Seville,Spain
3DepartmentofWildlifeEcologyandBiodiversity,NationalEnvironmentalResearchInstitute,KaløGrena˚vej,Rønde,Denmark
4DepartmentofBiodiversityandEvolutionaryBiology,MuseoNacionaldeCienciasNaturales,Madrid,Spain
5InstituteofAdvancedStudy,LaTrobeUniversity,Vic.,Australia
Keywords
Iberianlynx;Lynxpardinus;principal componentanalysis;inbreeding;habitat fragmentation;bottleneck;admixture analysis.
Correspondence
CinoPertoldi,DepartmentofEcologyand Genetics,InstituteofBiologicalSciences, UniversityofAarhus,NyMunkegade, building540,8000AarhusC,Denmark. Email:
RosaGarc´ıa-Perea, Departmentof BiodiversityandEvolutionaryBiology, MuseoNacionaldeCienciasNaturales, CSIC.J.GutierrezAbascal2,28006Madrid, Spain.Email:
*Theseauthorscontributedequallytothis work.
Abstract
TheIberianlynxLynxpardinusisoneoftheworld’smostendangered felidsandis vulnerabletohuman-inducedmortality andhabitat loss,whichreducepopulation sizeandacceleratethelossofgeneticvariation.Twenty-fivemetrictraitsofIberian lynxskullshavebeenmeasuredon95skullscollectedbetween1872and2003.The skulls belong to three geographically distinct areas/populations, which have recentlydivergedfromeachotherasaconsequenceofincreasedhabitat fragmen- tation: Donanaarea,SierraMorena mountainsandMontesdeToledoarea.The morphometricstudywasundertakenusingunivariate, multivariate andadmixture analysisapproaches,andallthreetechniquesprovidedevidenceformorphometric differentiation,bothinskullsizeandshape,amongthethreepopulationsforboth malesandfemales.Environmentalandgeneticforcesthatmayhaveshapedthese patterns arediscussed.ThemalesofthepopulationoftheDonana areashoweda different degreeofreduction insizeinnineoftheskulltraits withtime,which hasbeensuggestedtobepartlybecauseofworsenedhabitat conditions. However, the heterogeneity ofthe degreeofmean sizereduction and the relatively high degreeofreduction ofsomeoftheskulltraits investigated (44%),whichhave alteredtheoriginalproportionsbetweentheskullvariables, couldalsopartly be attributedtoinbreeding depression intheDonanapopulation.Thephenotypic variabilityoftheskulltraitsshowedsignificantincreases(twotraits)ordecreases (ninetraits)withtime,andthisdifferentpatternof changewithtimehasbeen suggestedtobebecauseofadifferentnumber ofgenescontrolling thetraitswith different degreesofdominance and epistatic interactions. Theincreased pheno- typicvariabilityoftwoofthetraitshasalsobeenattributedtoapossibledecreased levelofdevelopmental stability,whichcanbeproduced byenvironmentaland/or geneticstress.Thefindingsofthisinvestigation contributetothediscussionabout theutilityandthelimitsofquantitative geneticstechniquesforconservation purposes.
Introduction
Effectof inbreedingand population bottlenecksongeneticvariabilityand fitness-relatedtraits
Among populations, the consequences of isolation and smallpopulationsizeincludegeneticdifferentiationdriven bygeneticdrift,whoseeffectis inverselyrelatedtothe effectivepopulationsize(Ne).Within thepopulations, loss of geneticvariationandinbreedingdepressioncanoccur. Numerous studies conducted both innatural and incon-
trolledconditionshaveshowna directcorrelationbetween geneticdiversityandmeasuresoffitness(e.g.Koehn, Diehl
Scott,1988;ReedFrankham,2003).Onereasonisthe impactof inbreedingdepression(Saccheri,Brakefield& Nichols,1996; KellerWaller,2002).Inbreeding can contributetoachangeofthemorphometrictraits because of the increased expression of genetic load, as well as reducedopportunitiestoexpressoverdominance(see e.g. DeRose Roff, 1999;Keller Waller, 2002).However, therearefewdirectdemonstrationsofsuchphenomenain naturalpopulations.Aproblem,whichaffectsthesecorrela- tionalstudies,isthelackofaccurate knowledgeofthepast
andpresentdemographictrendsanddistribution ofthe populationsunder study. Furthermore,environmental factors also can lead to variation in fitness components andmorphometrictraits,becauseofgenotype–environment interactions (GXE) (Turelli, 1988). Thus, to find a link betweenthedecline inafitnesscomponentormorphometric traitandinbreedingdepressioncanbe difficultin natural populations(KellerWaller,2002).Itisevenmorechallen- gingtotrytoassessthecontributionofinbreedingdepres- siontotheriskofpopulationextinction. Reports fromthe fieldofconservationbiologyonpopulations whichhave sufferedseverebottlenecks, butneverthelesscurrentlypros- per(Ellegrenetal.,1993), couldquestionthatinbreeding depressionis aseverethreat forlong-termsurvivalof populations.However, theseexamplesareisolated studies thatarenotunexpectedbecauseof thelargevariationin inbreedingdepressionbetweenpopulations. Consequently, empirical studies wherethe species’demographicchanges havebeenextensivelystudiedarenecessary,inordertodraw strongerconclusionsabouttheeffects ofhabitat lossand populationbottlenecks.
Acase study:theIberianlynx
The Iberian lynx Lynx pardinus (Temminck, 1827)isat present oneoftherarest mammals onEarth, theonlycat listedinCategory1 ofvulnerability toextinctionand considered as critically endangered (CE) by the IUCN (IUCN,2003).TheIberian lynxisalsoawell-documented example ofacarnivore suffering theconsequences ofhu- man-inducedmortality, scarcity ofprey and habitat loss. Lynxhabitat hasbeenseverely modifiedandreducedby extensivedestruction(Delibes,RodrıguezFerreras,2000). Thewildpopulationisbelievedtoconsistoflessthan 200 individuals(Guzman etal.,2003).Bytheearlyyearsofthe
20th century, the Iberian lynx had become very rare in northernSpain,although itwasstillcommonincentraland southern Spain. By the 1960s, its range was essentially limitedtothesouth-westernquarterof thepeninsula (Rodrıguez Delibes,1992)(seeFig.1).
The declineofthelynxpopulationsincethe1950shas beenprimarilycausedbyhabitat lossandadeclineoftheir mainpreyspecies,theEuropeanrabbit Oryctolaguscunicu- lus.Infact,therewasadrasticpopulationbottleneckduring the1950sand1960s,whenthemyxomatosisviraldiseasehit therabbit populations (Villafuerteetal.,1993).Recent estimates suggestthat there arejusttwo populations left, theDonana(D)andtheSierraMorena (SM)populations,
inhabiting anarea largerthan 2000km2 and separated by
more than 300km. The D population,with about 40–50 lynx,seemstohavebeenisolatedfromtheothersurround- ingandnowextinctpopulations formorethan50years, becauseofanexpansionofcroplandstothenorthanddense humansettlementstothewest(Rodriguez Delibes,1992). About 150individuals remain at western Sierra Morena. TheMontesdeToledo(TM)population,whichisprobably extinct,wasneartheSMpopulation. Thetwosmalland isolated remaining populations of the Iberian lynx are
(a)
(b)
TM
(c)
SM D
Figure1Threestagesofthe declineoftheIberianlynxLynxpardinus populationsinthesecondhalfofthe20thcentury:(a)estimated distributioninthe1960s(basedonRodr´ıguezDelibes,1990);(b) estimateddistributioninthe1990s(basedonRodr´ıguezDelibes,
1992);(c)breedingpopulationsatpresent(basedonGuzma´netal.,
2003).TM,locationoftheextinctMontesdeToledopopulation;SM, SierraMorenapopulation;D,Don˜anapopulation.
theoretically vulnerable to inbreeding and genetic drift, where alleles with low frequency are likelyto disappear fromthepopulation’s genepool.Beltran Delibes(1993) foundpreliminaryevidence forthisoccurrenceintheD population. Threepelagepatterns werepresentinthe populationbefore 1960,but now no animals exhibit the rarer fine-spotted pattern.Additionally,thegeneticvaria- tionestimatedinmtDNAgenesandnuclearmicrosatellites wasfoundtobereducedintheIberianlynxrelativetomost otherfelidspecies,suggestingthattheyexperiencedafairly severedemographicbottleneckandthatthetwoinvestigated populationsofDandSMweregeneticallydifferentiated at thegenomiclevel andshowedheterozygosity deficiency (Johnson etal.,2004).
Given the well-documented present and past demo- graphictrendsof theIberianlynx, thisinvestigationcould constituteanempiricalexampleof themorphometrical changesofaspecies whichbecameendangered andfrag- mentedwithinarelativelyshorttimespan.
Methodologicalapproachand aim
Severalmethodologicalapproacheshavebeenappliedtothe studyofthegeneticconsequencesofhabitat loss, habitat fragmentationandpopulationbottlenecks. Studiesoneco- logically relevant traits, such as craniometrical investiga- tion,havebeenappliedtoyieldinformationondifferences amongpopulationsandtheirstructure (e.g.Huson Page,
1980; Clutton-Brock,Kitchener Lynch,1994; Lynch& Hayden, 1995; Simonsenetal.,2003).Someresearchers believethat morphologicalsimilaritycannot beinterpreted toindicategenomicsimilarity(BaverstockAdams,1987), astheselectivepressures actingonmolecular markers are different:themolecularmarkersarebydefinitionconsidered neutral, andtheadaptive traits non-neutral(Lynch,1996). However,inpolygeniccharacters (asmostofthemorpho- metriccharactersare),theforcesofselectionaredistributed overalargenumberofloci,renderingtheselectiveforceson specific loci sufficiently overwhelmed by random genetic drift.Furthermore,foraNe smallerthanfew hundred individuals,theexpectedamount ofvariation foraquanti- tativecharacter is nearlyindependent ofthestrengthof selection and largely a result of mutation-drift balance (Lynch,1996).Inthesmall-sizedlynxpopulations under study,theselective pressuresshouldbeoverwhelmedby geneticdrifteffects, makingall thetraitsandthegenes selectivelyneutral, ornearlyso.Hence,theeventualdetec- tionofmorphometrical differencesbetweenthedifferent populationsstudiedismainlyaconsequenceofgeneticdrift and/orenvironmentalvariability.
Theaimofthisinvestigation wastotestwhetherthereis temporal and spatial variation (oftheskulltrait sizeand shape) among the three populationsinvestigated, which haverecentlybeenisolatedfromeachother.
Predictions
Isolation associated withsmallpopulationsizedifferences couldhaveproducedgeneticdivergenceofthethreepopula- tionsstudied (D,SM,TM), andtheconcomitanteffectof inbreedingdepressioncouldhaveproduced achangeofthe skulltraits,alteringtheproportions betweenthem.Size differencesmayreveal differenthabitatconditionsthe individuals collectedhavebeenexposedto. Thedetection ofshapedifferencescouldindicatethepopulations under studyaregeneticallydifferentiated (Simonsenetal.,2003), asseveralstudiesincontrolled conditions haveshownthat genesregulatetheshapeofthetraitsmoretightlythanthey regulatesize(Birdsallet al.,2000;Workmanet al.,2002)and thatmoregenesareinvolvedintheregulationofshapethan size(Workmanetal.,2002).
Thedetection ofsignificant differencesintrait sizeand trait shape between different periods ofcollection (before andafter thepotential populationbottleneck happened in theDpopulationinthe1950–1960s)couldalsorevealthat environmentaland/orgeneticchangeshaveoccurredinthis population. Thedetectionofchangesinthephenotypic variability (Vp)oftheskulltraits couldalsorevealgenetic
changesasinaconstant environment (Vp)areroughly correlated to the additive genetic variance (Va) in the absence of dominant and epistatic interactions and to Ne (Podolski, 2001). The loss of genetic variability could also have disrupted developmental stability (DS) and thereby increased thevariability ofquantitativetraits (Hoelzel,1999).
Materialsandmethods
Iberian lynxskullswereavailablefromthescientificcollec- tionsofthefollowingInstitutions: EstacionBiologicade Donana-CSIC,Seville,Spain;MuseoNacional deCiencias Naturales-CSIC,Madrid, Spain;TheNaturalHistory Mu- seum,London,United Kingdom; Naturhistorisches Mu- seum Basel, Basel, Switzerland; American Museum of Natural History, NewYork,USA;NationalMuseumof NaturalHistory,Washington DC,USA;ZoologischesFor- schungsinstitut undMuseumAlexanderKoenig,Bonn, Germany; Zoologische StaatssammlungMunchen, Munchen,Germany; UniversidadComplutense,Madrid, Spain.Westudied95skulls,representing about 90%ofthe skullsofthisspecies keptinscientificcollections.These individuals werecollectedbetween1872and2003(Iberian lynxareprotected bySpanishlawsince1973; specimens collectedafterthatdatewerefounddeadorwereconfiscated byconservationofficers).
Twenty-fivemetrictraitswere measuredoneachskull. Thirteen of these traits are related to the cranium or mandibleand12ofthesemeasurements arerelatedtoteeth (seeFig.2fordescriptionofthetraitsandabbreviations).
Theskullsbelongtothreegeographically distinctareas: D (58specimens, 35males,23females,collectedbetween
1872 and 2003), SM (12 specimens, eight males, four females,collectedbetween1889and1997) andTM(25 specimens,15 malesand10 females,collectedbetween1960 and1985).
Theskullswereofknownsexorsexwasestimatedonthe basisofadiscriminant function. Skullswereidentified as adults (fullygrown)andsubadults (notfullygrown)based on tooth sections (following the method described by Zapata et al., 1997), or by checking the ossification of cranialsutures(Garcıa-Perea,1996).
Someskullsweredamaged,butwheneverpossibleallthe
25measurements weretakenoneachskull.Allthemeasure- ments were taken with a digital calliper, to the nearest
0.1mm,andareexpressedinmminthetables.
For thestatistical treatmentofmeasurements related to cranium and mandible, only adult skulls were included, sincetheyshowagevariation (Garcıa-Perea,1991).How- ever,allthetooth measurements weredoneonboth adults and subadults, asthesemeasurements arenot affected by theageoftheindividual(Garcıa-Perea,1991).
Multivariate and univariate analyses have been per- formed contemporarily,inorder toutilizeallthemeasure- mentsoftheavailabledataset.Thesexeshavebeenanalysed separately, because ofthe sexual dimorphism ofthe lynx
RW
RL
CBL
TL
BAL
BUL
ZWMW
OL
WC1
WP3
WP4
Lp3
Lp4
Lm1
p3–m1
ML
Wp3
Wp4
Wm1
MRH
Figure2 Descriptionofthetraitsmeasured. Thirteentraitsare relatedtothecraniumor mandible:TL, totallengthofcranium;CBL, condylobasallengthofcranium;BAL,basicra- nialaxislength;RL,rostrallength;OL,orbital length;ZW,zygomatic width;MW,mastoid width; RW,rostralwidth; BUL,bullarlength; ML,mandible length;MRH,mandible ramus height;C1–P4,C1 toP4length;p3–m1=p3to m1.Twelvemeasurementsarerelatedtoteeth: Lp3,p3maximumlength;Wp3,p3maximum width;Lp4,p4maximumlength;Wp4,p4max- imumwidth;Lm1,m1maximumlength; Wm1, m1maximumwidth;LP3,P3maximumlength; WP3,P3maximumwidth;LP4,P4maximum length; WP4, P4maximum width; LC1,C1
maximumlength;WC1,C1maximumwidth.
skull(Garcıa-Perea,Gisbert& Palacios,1985; Beltran& Delibes,1993).
Admixtureanalysis
An admixture analysis was conducted pooling all the fullygrownskulls(holdingthesexesseparated).Thisanaly- sis wasconducted inordertoascertainiftheapriori subdivisionofourdata setintothreeseparated geographic groups(D,SMandTM)wascorrect.Thefittingofnormal ort-componentmixturemodelstomultivariate data, using maximum likelihood viatheEMalgorithm, iswellknown andcommonlyused.However,forasetofdatacontaininga group orgroups ofobservationswithlonger than normal tailsoratypicalobservations,theuseofnormalcomponents may unduly affectthe fitofthe mixture model. Further- more,thisapproachnormally requirestheinitialspecifica- tion of an initial estimate of the vector of unknown
parameters,or, equivalently, ofan initial classification of the data with respect to the components of the mixture modelunderfit.Therefore,a morerobustapproachby modelling the data by a mixture of t distributions is provided.Theuse oftheECMalgorithm tofitthistmixture modeliscalledEMMIXandisdescribedinMcLachlanKrishnan(1997)andMcLachlanPeel(1998).Addition- ally,thealgorithm utilizedforthispurposedoesnotrequire any specification ofthe above-mentionedparametersand automatically clusters each skull into its most probable group, andtheaposteriori probabilityofanindividual to belong to the cluster in which it has been assigned is estimated.Theoptionofunrestrictedcomponentcovariance matrices forthedata sethasbeenchosen. Thesignificant level(P-value)isproduced bytheoptional bootstrapanaly- sis.Bysequentially testingn,n+1,n+2,etc.andstopping whenthestepbecomesinsignificant,thenumberofcompo- nentscanbeassessed.
Multivariateanalysis
Sinceour collection isoflimited size,wedecided to em- ploy resamplestatisticsforourmultivariateandunivariate investigation, whichutilizes100%oftheinformationavail- able and is less sensitive towards small sample sizesand deviations from normality (Davison Hinkley,
1997). A resample program was designed in order to comparethedatasamples.All theresampletestswere conducted asMonte Carlostylepermutation testswith replacement.
Themultivariate analysiswasconducted onfullygrown skulls, andbecausemultivariatemethodsdonotallow observations tohaveincompletedata,themultivariate analysiswasperformedonasmallersubsetof measurements (24 insteadof25measurements, asthetraitp3–m1was excludedfromtheanalysis)andspecimens(30femalesand
40 males). Skull specimens with more than fivemissing values were excluded from further multivariate analyses andmissingvalueswereestimated bymeansofastepwise regressionanalysis(Zar,1999).
A principal componentanalysis (PCA; Marcus, 1990) wascarriedoutonthecovariance matrix derivedfromthe skullmeasurements. Thisanalysisclassifiesphenotypic var- iation into independent components that can beused to dissect geneticnetworksregulatingcomplexbiologicalsys- tems(Chaseetal.,2002).
Thus,ifsize variation ispresentinthedataandthe coefficientsofprincipalcomponent (PC)1areeitherall positive or all negative, then this PC can be said to summarize the within-sample size variation (Bookstein,
1989).Shapeisthusdefinedasthatsubspaceofdimensions onelessthan thenumber ofmeasured variablesandquan- tifiesthevariation thatcannotbeexplainedbysizevariation andallometricrelationships.Theexpectationfromtheoryis thatfunctionally independent partsoftheskullshouldvary independentlyamongthelocicontrollingquantitativetraits, andthereforeshouldbe associatedwithdifferentPCs (KlingenbergLeamy,2001).
Alltheskullsweregrouped intothreedistinctgroupson thebasisof theirgeographicalorigin.Anon-parametric MANOVA(NPMANOVA)(5000 permutations)(Ander- son,2001)wasconducted inordertotestthesignificanceof multivariate differentiation amongthethreegeographic groups and anon-parametricANOVA (NPANOVA)was conducted on the first three principal components (PC1–PC3).
Multiple comparison tests weremade with a Scheffe’s F-test(Zar,1999),forcomparingthedifferencesbetweenthe PCsinthepopulations.
TheDmales’skullswereseparatedintotwotimeperiods, from1872to1961,whichiscalledthepre-bottleneckperiod, andfrom1971to1998,whichiscalledthepost-bottleneck period.AnNPMANOVAwasmadefortestingmultivariate differences between the two periods of collection. This analysishasnotbeenperformed onfemalesbecauseofthe toosmallsamplesizeoftheskullsofthissexavailablefrom thepre-bottleneckperiod.
Univariateanalysis
Anon-parametricANOVA(NPANOVA)was conductedin ordertotestthesignificanceofeachsingletraitdifferentia- tionamongthe threegeographicgroups,andthepairwise comparisonsbetweenthegroupswere madewithaScheffe’s F-test. Aresample t-test (1000permutations),whichtakes intoaccountdeviationfromnormaldistribution,inequality ofsamplesizeandvariancesamong thegroups tested,was conducted for each trait for testing differences in size betweenthetwoperiodsofcollectionofthemalesfromthe Dpopulation.Thedegreeofchangeofmeanandvariance betweentheperiodsof collectionhavebeenestimatedand expressedin percent.Thesignificanceof thedifferences between variances has been tested by a non-parametric F-test (1000permutations),which takes into account the non-normaldistributionofthe data and unequal sample size.
Because of the large number of tests that we have performedin thisinvestigation,anoverallBonferronicor- rection (Rice,1989)wasapplied toalltheresamplet-and F-teststoavoidsignificantresultsarisingasaconsequence ofalargenumberofrelatedtests.FollowingMiller’s(1981) suggestions,we madeaseparate probability statement groupingallthetraitswithanisometricrelationship, which hadacorrelationcoefficientR40.5,bothfortheskull-and mandible-relatedtraits group and the teeth traits group. This grouping ended up with three groups for the skull- related traits (k=3;threshold P=0.167)and fivegroups for the tooth traits (k=5;threshold P=0.01) (data on correlationcoefficientsandgrouping oftraitsareavailable onrequestfromthecorrespondingauthors).
Results
Admixtureanalysis
The admixture analysis gave four clusters for the males skulls (k=4, P=0.012) and three clusters (k=3, P=0.045)forthefemaleskulls.Thepercentage ofspeci- mens correctly assigned to the geographic group they belongedtowas100%forbothmalesandfemales,withthe exception of the comparison TM versus SM in females, whichgave 88.89%ofcorrectlyclassifiedindividuals.The malesfromtheDpopulationwereseparated bytheadmix- ture program into twoclusters(k=2),whichwerecoinci- dentwiththetemporal apriorisubdivisions(pre-andpost- bottleneck periods), with the exception of two skulls as- signedtothepre-bottleneckperiodbutcollectedinthepost- bottleneckperiod.
Multivariateanalysis
The first three PCs explained 83.68% of the males and
82.79%ofthefemalestotal variation; therefore, themulti- variatestatisticshavebeenconducted onthefirstthreePCs (seeTable1).
Table1Vectorsoftheprincipalcomponents(PC1–PC3)andtheper centofvarianceexplainedbythesinglePC formalesandfemales
MalesFemales
2).Infemales,nosignificant pairwisedifferentiationswere found forPC1;thePC2ofDwassignificantlybiggerthan thatofSMandthePC2ofTMwassignificantlybiggerthan that ofSM;finally,thePC3ofDwassignificantly bigger
PCVector
%ofvariance
explainedPCVector
%ofvariance
explained
thanthoseofTMandSM(Table2).
TheNPMANOVAthatwasperformed fortestingmulti-
1 / 56.87 / 73.35 / 1 / 60.190 / 69.653 / variate differences between males from the D population2 / 6.410 / 8.270 / 2 / 6.252 / 7.241 / collected in the two periods was significant (F=3.94,
3 3.920 5.061 3 5.111 5.901
PC,principalcomponent.
PC1explained73.35and69.65%ofthetotalvariabilityin malesandfemales,respectively(seeTable1).
Theskulltraitstotallengthofcraniumandcondylobasal
length ofcranium showedthehighestloadings onPC1in both sexes, whereas zygomatic width (ZW) showed the highestloadingonPC2androstrallengthonPC3in both sexes.Thetoothtraitsshowedrelativelysmallerloadingson all the three PCs as compared with the skull traits and showedageneraltendencytohavehigherloadingsonPC2 andPC3.Thetraits’loadingsforthefirstthreePCsforboth malesandfemales,computed fromthepooledwithin-group variance–covariancematrixof traits’measurements, andthe percentage of variation accounted for by each PC are availableonrequestfromthecorrespondingauthors.
TheNPMANOVA conducted inordertotestthesig- nificanceofmultivariate differentiation amongthethree geographic groups was highly significant for both males andfemales(males:F=6.33,P=0.0002,females:F=3.39, P=0.0058).
TheNPANOVAsconducted fortestingwhetherthefirst threePCsweredifferentamongpopulationsweresignificant forbothsexes andforallthreePCs(Table2).Pairwise Scheffe’stestshowedasignificantlybiggerPC1formalesof theDpopulationascomparedwiththeSMpopulation,that thePC2ofDwassignificantly biggerthan TM’sand the PC3ofSMwassignificantlybiggerthanthatofTM(Table
P=0.034).
Univariateanalysis
TheNPANOVAsthat wereconducted inorder totestfor differencesofeverysingletraitamongthethreegeographic groupsweresignificantfor50%oftheskull-andmandible- relatedtraitsinmalesand38.5%infemales,whereasforthe tooth traits, the percentages ofsignificant testswere16.7 and25%formalesandfemales,respectively(Table2).
Thepairwisecomparisonsbetweenthegroups (Scheffe’s test) showed ageneral tendency for the D populationto haveabiggermeansize oftheskulltraitsinbothsexes, followedby TMandfinallySM,whichhadthesmallestsize. Scheffe’stestconducted onthemales’tooth traits showed thatthetraitsWP3ofSMandTMwerebiggerthanintheD population.TheSMpopulationalsohad biggerLp4than D.ThefemalesofDshowedsignificantlybiggerWm1and LC1thanthatofTM,andboththeDandTMpopulations showedabiggerLp3thanSM(Table2).
Thetestfordifferencesinsizebetweenthetwoperiodsof collectionofthemaleskullsfromtheDpopulationwiththe resamplet-testshowedasignificantreduction insizeinthe post-bottleneckperiod in30.8% oftheskulltraits and in
25% ofthetooth traits (Table3).Alltheother traits also showed a reduction in sizein the post-bottleneckperiod although thiswasnot significant (withtheonlyexception beingtraitLm1).Themeanreduction intheskulltraits’size was2.12%(range:0.21–4.29%),whereasinthetooth traits
Table2Non-parametricANOVA(NPANOVA)conductedfortestingdifferencesamong thethreepopulations(D,Don˜ana;SM,SierraMorena; TM, Toledomountains)ofthe meanoftheprincipalcomponents (PC1–PC3)
MalesFemales
PCs Populations
nMeanSE
NPANOVA
(F-value)P
Scheffe’s
F-testnMeanSE
NPANOVA
(F-value)PScheffe’sF-test
PC1 D26 253.70 1.536.850.0029 (D4SM)**21 237.78 1.73 3.730.037
TM10 249.02 1.006 229.39 2.48
SM4 241.32 0.793 230.51 2.09
PC2 D261.51 0.42 15.940.0001 (D4TM)*** 21 —13.32 0.46 5.790.0081 (D4SM)**,(TM4SM)*
TM10—2.48 0.376 —13.17 0.65
PC3 / SMD / 4
26 / —0.16
—25.82 / 0.65
0.39 / 5.79 / 0.0065 / (SM4TM)** / 3
21 / —17.83
—33.07 / 2.23
0.48 / 6.35 / 0.0055 / (D4TM)*,(D4SM)*
TM / 10 / —27.22 / 0.42 / 6 / —35.07 / 0.60
SM / 4 / —23.72 / 0.50 / 3 / —36.23 / 0.57
Pairwisetestbetweenthe populationswereperformedwithaScheffe’sF-test.Both testswereperformedseparatelyfor thetwosexes.
*Po0.05,**Po0.01,***Po0.001.
NS,non-significant;PC,principalcomponent.
Table3Non-parametricANOVA(NPANOVA)conductedfortestingdifferencesamongthethreepopulations(D,Donana;SM,SierraMorena;TM, Toledomountains)ofthemeanofthe skull,mandibleandtoothtraits (expressedinmm)
MalesFemales
Traits Populations
nMeanSE
NPANOVA
(F-value)P
Scheffe’s
F-testnMeanSE
NPANOVA
(F-value)P
Scheffe’s
F-test
Skullandmandibletraits
TLD28 136.120.884.380.018820 127.361.062.350.1165
TM12 135.180.826 123.301.22
SM4 129.700.362 123.602.60
CBLD25 122.250.852.580.090317 113.501.020.860.4371
TM10 122.660.856 111.251.10
SM3 117.270.131 114.90
BALD2540.800.342.750.07831837.160.520.170.8500
TM939.820.47636.717 0.305
SM238.450.35136.4
RLD2755.730.525.280.0095 (D4SM)*2151.490.682.750.0823
TM1053.840.68548.141.12
SM451.800.95350.170.27
OLD2635.430.242.0920.13862234.330.360.160.8559
TM1035.810.18634.580.77
SM234.100.10333.930.14
ZWD2797.300.577.9310.0013 (D4SM)**, (D4TM)*
2291.660.616.530.0045 (D4SM)*
TM1194.850.60788.641.43
SM391.671.05385.572.05
MWD2759.860.29 12.720.0001 (D4SM)***, (D4TM)*, (TM4SM)*
1956.450.444.410.0234 (D4TM)*
TM1158.380.37653.830.77
SM355.970.48254.901.40
RWD2753.130.260.4300.65542250.770.360.380.6859
TM953.560.35750.140.58
SM553.581.02450.520.92
BULD2428.190.41 11.850.0001 (D4TM)***1826.820.2627.540.0001 (D4TM)***, (D4SM)***
TM924.760.39422.670.86
SM227.250.05222.351.05
MLD2690.831 0.592.400.10512084.960.782.200.1301
TM1089.790.65782.770.67
SM487.800.80482.001.78
MRHD2639.110.329.490.0005 (D4SM)*, (D4TM)**
2035.830.477.250.0029 (D4TM)*, (D4SM)*
TM1037.070.30733.040.69
SM436.900.55432.870.76
P3–M1 Thetestcouldnotbeperformedbecausenomeasurements wereavailableforTM
1828.840.2501.930.176
128.10
130.81
C1–P4 D2539.240.241.960.1552237.682 0.246.010.0067 (D4TM)**
TM1038.620.29635.817 0.402
SM539.740.393371.097
Toothtraits
Lp3 / D / 29 / 7.57 / 0.08 / 0.83 / 0.4435 / 18 / 7.256 / 0.084 / 4.56 / 0.0205 / (D4SM)*, (TM4SM)*TM / 11 / 7.74 / 0.11 / 7 / 7.329 / 0.087
SM / 5 / 7.74 / 0.21 / 3 / 6.533 / 0.524
Wp3 / D / 29 / 4.10 / 0.04 / 0.99 / 0.3806 / 18 / 3.95 / 0.04 / 1.17 / 0.3255
TM / 11 / 4.21 / 0.06 / 7 / 3.90 / 0.07
SM / 5 / 4.16 / 0.18 / 3 / 3.77 / 0.18
Lp4 / D / 31 / 10.25 / 0.09 / 1.61 / 0.2117 / 21 / 9.87 / 0.09 / 0.339 / 0.7153
Table3Continued.
MalesFemales
Traits Populations
nMeanSE
NPANOVA
(F-value)P
Scheffe’s
F-testnMeanSE
NPANOVA
(F-value)P
Scheffe’s
F-test
TM1110.340.1169.770.20
SM410.670.2129.650.15
Wp4D314.940.040.180.8376214.760.052.770.081
TM114.900.0764.550.034
SM44.950.0624.70.2
Lm1D3312.790.083.060.05552112.440.0900.230.7993
TM1312.960.10712.370.13
SM713.260.21312.270.49
Wm1 D335.680.042.510.0917215.470.045.090.0128 (D4TM)*
TM135.520.0575.200.03
SM75.710.1245.400.16
LP3D2910.120.091.680.1986199.770.090.560.5787
TM1210.320.0959.960.09
SM410.450.0629.850.15
WP3D294.880.05 10.060.0003 (SM4D)**, (TM4D)*
194.600.050.470.6326
TM125.120.0554.680.12
SM45.420.1724.750.15
LP4 D 33 14.96 0.09 5.75 0.0056 (SM4D)** 22 14.37 0.08 0.18 0.8331
TM1315.260.16714.430.21
SM815.610.14414.520.45
WP4 D 34 7.37 0.07 1.42 0.2514 22 6.78 0.08 0.89 0.4218
TM137.150.1176.630.14
SM87.270.1546.970.31
LC1 D 34 7.46 0.06 2.47 0.0947 21 7.17 0.01 3.65 0.0395 (D4TM)*
TM137.720.1266.670.12
SM57.680.1736.970.18
WC1 D 33 6.10 0.06 1.70 0.1929 21 5.94 0.07 1.80 0.1843
TM136.310.0765.630.21
SM56.100.1835.870.18
PairwisetestbetweenthepopulationswereperformedwithaScheffe’sF-test.Bothtestswereperformedseparatelyforthetwosexes(males left,femalesright).
*Po0.05,**Po0.01,***Po0.001.
TL,totallengthofcranium;CBL,condylobasallengthofcranium;BAL,basicranialaxislength;RL,rostrallength;OL,orbitallength;ZW,zygomatic width;MW,mastoidwidth;RW,rostralwidth;BUL,bullarlength;ML,mandiblelength;MRH,mandibleramusheight;C1–P4,C1toP4length; p3–m1=p3tom1.Lp3,p3maximumlength;Wp3, p3maximumwidth;Lp4,p4maximumlength;Wp4,p4maximumwidth;Lm1,m1maximum length;Wm1,m1maximumwidth;LP3,P3maximumlength;WP3, P3maximumwidth;LP4,P4maximumlength;WP4,P4maximumwidth;
LC1,C1maximumlength;WC1,C1maximumwidth.
the mean reduction was 2.57% (range: 0.32–5.4%) (Table3).
TheresampleF-testshowedasignificantreduction ofVp
oftheskulltraits inthepost-bottleneckperiod compared
withthepre-bottleneckperiod in38.6%, and asignificant increasein7.6% (trait basicranial axislength)oftheskull traits.Intheteeth,asignificantreductionof Vpwasfoundin
33.3%ofthetraits,whereasasignificantincreasewasfound
in8.33%(traitWP3)ofthetraits(Tables3and4).
Discussion
Despitetherelativelysmallsamplesizeoftheskullsinvesti- gated and the relatively long time period of collection, certainsizeandshapedifferencesweredetectedamong the
a prioridefinedgeographicgroups.Thisgeographicsub- division was also supported by the admixture analysis, whichhasproventobe ausefultoolforthedetectionof temporal and/or spatialpatterns ofmorphometricdiffer- ences.Morphologicaldifferencesinsize-relatedtraits(PC1) werenotsignificant,exceptwhencomparing malesfromD andSMpopulations. However,significantdifferencesin shape-relatedtraits(PC2,PC3)werefoundamongthethree populations.Thissimilarityinsize andthesmallshape differencesinasamplecoveringatimespanofalmostone century suggest that no large sizedifferences have ever existedamong thethreepopulations,although theslightly largersizeofDonana’s lynx(especiallybeforethehypothe- ticalbottleneckhadoccurred)couldreflect apeculiar characteristic ofthispopulation,whichismostcommonly
Table4Acomparisonofthemeansize(inmm)andvarianceofthetraitsofskullsofmalescollectedinthepre-bottleneckandinthe post-bottleneckperiodintheDon˜anaarea
Skulland mandible traits
Pre-bottleneck period(mean) (Var) (n)
Post-bottleneck period(mean) (Var) (n)
Resample
t-test
(t-value)Pt-test
Changesin
%ofthe
meantraitsize
Presample
F-test
Changesin
%ofthe variance ofthetrait
TL 138.00(17.63)(10) 135.08(21.97)(18) 1.69 NS (pre4post)2.1 NS (post4pre)19
CBL 125.46(12.34)(7) 121.01(15.17)(18) 2.76 * (pre4post)3.5 NS (post4pre)18.65
BAL 41.67(1.78)(7) 40.46(3.13)(18) 1.85 NS (pre4post)2.47 * (post4pre)43.11
RL 57.38(4.43)(9) 54.91(7.16)(18) 2.61 * (pre4post)4.29 NS (post4pre)38.12
OL 35.66(1.36)(8) 35.32(1.66)(18) 0.66 NS (pre4post)0.95 NS (post4pre)18.07
ZW 98.08(16.17)(9) 96.91(5.27)(18) 0.81 NS (pre4post)1.19 ** (pre4post)67.41
MW 59.94(3.75)(10) 59.81(1.55)(17) 0.18 NS (pre4post)0.21 * (pre4post)58.65
RW 53.84(2.38)(10) 52.71(1.22)(17) 2.035 NS (pre4post)2.1 * (pre4post)48.86
BUL 28.30(5.51)(8) 28.14(3.63)(16) 0.169 NS (pre4post)0.57 NS (pre4post)34.11
ML 91.94(11.73)(10) 90.14(6.54)(16) 1.43 NS (pre4post)1.96 * (pre4post)44.27
MRH 39.38(3.81)(10) 38.94(2.04)(16) 0.612 NS (pre4post)1.10 * (pre4post)46.65 p3–m1 31.04(1.17)(9) 29.8(0.71) (11) 2.82 * (pre4post)4.00 *,# (pre4post)39.48
C1–P4 40.10(1.56)(8) 38.84(0.91)(17) 2.52 ** (pre4post)3.14 *,# (pre4post)41.67
Toothtraits
Lp3 7.82(0.12)(12) 7.39(0.20) (17) 2.98 ** (pre4post)5.40 *,# (post4pre)40
Wp3 10.47(0.31)(8) 4.08(0.03)(17) 0.76 NS (pre4post)1.21 NS (pre4post)19.05
Lp4 4.13(0.04)(12) 10.26(0.14)(16) 0.97 NS (pre4post)2.00 NS (pre4post)20.91
Wp4 4.96(0.02)(10) 4.93(0.06) (21) 0.38 NS (pre4post)0.61 NS (pre4post)59.75
Lm1 12.75(0.32)(12) 12.82(0.18)(21) (—0.37) NS (post4pre)0.53 * (pre4post)45.37
Wm1 5.73(0.05)(12) 5.64(0.05) (21) 1.13 NS (pre4post)1.57 NS (post4pre)3.40
LP3 10.42(0.32)(10) 9.96(0.12) (19) 2.35 *,# (pre4post)4.43 ** (pre4post)63.58
WP3 4.89(0.04)(10) 4.87(0.09) (19) 0.17 NS (pre4post)0.32 * (post4pre)55.43
LP4 15.17(0.24)(12) 14.84(0.24)(21) 1.87 NS (pre4post)2.18 NS (post4pre)1.24
WP4 7.59(0.08)(12) 7.24(0.17) (22) 2.89 ** (pre4post)4.57 ** (pre4post)51.78
LC1 7.64(0.15)(12) 7.36(0.11) (22) 2.13 *,# (pre4post)3.63 NS (post4pre)27.39
WC1 6.27(0.06)(12) 6.00(0.14) (21) 2.56 ** (pre4post)4.38 ** (pre4post)56.20
Thetestsconductedarearesamplet-test(1000permutations),whichtakesintoaccountunequalvariancesandaresampleF-test(1000 permutations),whichtakesintoaccountdeviationsfromthenormaldistribution.Theskulland mandibletraitsweremeasuredonlyonadult individuals,toothtraits havebeenmeasuredonboth adultsandsub-adults.Thechangesofthemeanandvarianceofthetraits betweenthetwo periodsofcollectionhavebeen estimatedinpercent.AnoverallBonferronitest,followingMiller’ssuggestions(1981),wasappliedmakinga separateprobabilitystatementgroupingallthetraitsthathadacorrelationcoefficient40.5forboththeskullrelatedtraits(k=3,threshold: P=0.0167)andtoothtraits (k=5,threshold:P=0.01).
*Po0.05,**Po0.01.
NS,non-significant;TL,totallengthofcranium;CBL,condylobasallengthofcranium;BAL,basicranialaxislength;RL,rostrallength;OL,orbital length;ZW,zygomatic width;MW,mastoidwidth;RW,rostralwidth;BUL,bullarlength;ML,mandiblelength;MRH,mandibleramusheight; C1–P4,C1toP4length;p3–m1=p3tom1.Twelvemeasurementsarerelatedto teeth:Lp3,p3maximumlength;Wp3,p3maximum width;Lp4, p4maximumlength;Wp4,p4maximumwidth;Lm1,m1maximumlength;Wm1,m1maximumwidth;LP3, P3maximumlength;WP3,P3 maximumwidth;LP4,P4maximumlength;WP4,P4maximumwidth;LC1,C1maximumlength;WC1,C1maximumwidth.
#,result nolongersignificantafterasequentialBonferronitest.
reported toincludeungulatesinitsdiet(Delibesetal.,1975; Delibes,1980;Aymerich,1982).
Inrecenttimes,we knowthattheIberian lynxhasa polygynousmatingsystem,withmalesdefendingterritories, whichcanincludethebreedingterritoriesofseveralfemales (Ferreras etal.,1997).InapreservedarealikeDonanaN.P., alargemalesize shouldbeadvantageousbecauseofthe increasedprobabilities ofoccupying anddefendingaterri- tory.However,thisadvantage maydecreaseifhighhuman- inducedmortalityreduceslynx densityandthusthestrength ofcompetition forterritories, asitlikelyhappened inTM andSM.Thiscouldpartlyexplainwhygeographical differ-
encesinsizearelargerformalesthan forfemales,andwhy malesarelargerintheDpopulation.
Most morphological differences observed among the threepopulationsareshaperelated(PC2,PC3),andconcern the following measurements: ZW, mastoid width, bullar lengthandmandibleramusheight(MRH).Twoof these variables (ZW and MRH) are related to the origin and insertionofthemainmasticatory muscles,masseterand temporalis,whicharealsoresponsibleforthestrengthofthe lethalbite(Smith,1993).Significantlyhighervaluesofthese variablesintheDpopulationarelikelyrelatedtotheability tohuntlargerprey.Relatedtothetoothtraits,although the
PCArevealedshapedifferencesincarnassials (P4andm1), Scheffe’stest failed to find significant differences when making pairwise comparisons between populations,and thefewdifferencesidentifieddidnotrevealaclearpattern ofvariation.Thedifferencesfoundamongthethreepopula- tionsappearnottobestronglylinkedtoanyecologicalspe- cialization norhaveaspecificanatomical-functionalmean- ing.Therefore, consideringthat shapedifferencesaremore geneticallydeterminedthansizedifferences(Atchley,Rutle- dgeCowley,1981;Klingenberg Leamy, 2001;Work- man etal., 2002)and giventhat shape ofmorphological traitsareingeneralassumedtobemorestablethansizeand submitted tostrongdevelopmental constraints(Debat etal.,
2003),our resultssuggestthat thethree geographic groups investigated are also genetically differentiated.These data agreewiththeprevious findingsofageneticdifferentiation betweenDandSMatthegenomiclevel(Johnsonetal.,2004). Related toTM and SMpopulations,Rodrıguez De- libes(1992)estimated that both populationswerestillcon- nectedinthe1980s,justbeforetheprobable extinction of theTMpopulation.Ourresultsindicateageneticdifferen- tiation betweenTMandSM,suggestingthat eitherRodrı- guezDelibes(1992)weretoooptimisticintheirevaluation ofthelynxsituation, orthat veryloweffectivepopulation sizesdrove a fast genetic structuringof the populations. These two possibilities willbe evaluated in an extended
studyofcurrentandhistoricalvariation,nowinprogress. Differences observed between D and TM populations
(Table2)arenotconsistentwithdatapreviouslyreportedon both populations,asthoseofGarcıa-Perea(1991)onTM and of Beltran Delibes (1993)on D show. Values of similartraitsmeasuredbytheseauthors indicated that TM maleswerelargerthanDmales,while TMfemaleswere smallerthanDfemales.Thislackofcongruenceisprobably aconsequence ofthesmallsizeoftheDsampleanalysed, likelycomposedbyamixtureofspecimenscollectedindif- ferentperiods (pre-andpost-bottleneck).Thisfactreveals howimportantthesekindsofstudiesareforspecieshaving sufferedsevere declineswithinrelativelyshorttimespans.
Thepattern ofmorphometricdifferentiationwasmoreor lessconsistent forboth malesandfemales,although males seem tohaveahigherphenotypic plasticitythanfemales,as has been shown by the relatively bigger percentage of variance explained inPC1andPC2inmalesascompared withfemales.Thechangesinsizeandshapeinmaleskullsof theDpopulationcanbebecauseofseveralfactors:spatial andtemporalvariationin habitat qualityandpopulation density can affect adult body sizeand skull traits. For example, fragmented habitats usually play the role of islands,recreating sometimestheeffectofdirectional body sizechangeobservedforseveralmammalspeciesonislands, referredtoastheislandrule(VanValen,1973). Asmall number of captive-reared, free-born Iberian lynx from DonanaandSierraMorena havereachedlargerbodysizes than their siblings reared in nature (I. Sanchez, pers. comm.),whichsuggestsanimportanteffectofthetrophic conditions onadult sizeduring thegrowingperiod. There- fore,areduction ofthelynxsizeinDonanasincethe1960s
couldbeaconsequenceof thedecreaseinthenumberof rabbits, thelynx’s mainprey,followingthearrivalof myxomatosis and later the rabbit haemorrhagic disease (RHD).Asimilareffectcouldbeexpectedtohaveoccurred in SM.LynxfromTMhadnochancetoexperiencethese changesbecausetheywerealreadyextinct.
Sometraitshowevershowedstrongerreduction insizeas compared withothers,evenif thecompared traitswere isometricallyrelated.Therearethereforestrongindications thattheobservedheterogeneityin thedegreeof meansize reduction and the relatively high degree of reduction of someoftheskulltraitsinvestigated(44%) canalsobe attributedto inbreeding depression. The heterogeneity of thedegreeofmeansizereduction (whichhasconsequently altered the original proportionsbetween the skull traits) couldbecausedby thedifferentnumberofgenescontrolling thetraitswithdifferentdegreeof dominanceandepistatic interactions,conferringtraitsdifferentsusceptibilityto en- vironmental change. This hypothesis was supported by recentworkswhichsuggestthattraitswithdifferentdegree ofdominance and epistatic interactions havedifferent de- greesof susceptibility to the environment(e.g.Podolski,
2001).Wemustkeepinmindthatinbreedingdepressionwill only be observed ifdominance interactions of traits are present(Lynch,1996).
Underastrictlyadditivegeneticarchitecture, allelements of the genetic covariance matrix for a set of traits are expectedtoshrinkwiththesamefactor,1–1/(2Ne)(Wright,
1951;Lande, 1979).Thus, thegeneticcorrelationand the
meantraitsizearenotexpectedtochangewithdrift.Thisis notwhatwefoundasboththemeantraitsize andtrait proportionshavechanged.Thecomplexityoftheeffectsof inbreedingontraitmeansis furthermanifestedasbeing environmentspecific.Interactionsbetweeninbreeding and theenvironmentareespeciallyinvestigated inconservation biology, wherethere isagrowing awareness that itisthe combinedeffectofgeneticandenvironmental stressthat threatens long-termpopulationsurvival.
The Vp of the skull traits showed heterogeneity with respect to either an increase or a decrease with time for severaltraits. Thesefindingsreinforce thehypothesis that the different skull traits are controlled by several genes, wheretheadditiveinteraction ismoreorlessexpressed.