Analysisofasymmetriesin theAfricanfruitbatsEldolonhelvum
and Rousettusegyptlacus(Mammalia:Megachiroptera)
fromtheislandsoftheGulfofGuinea.II.Integrationand levels ofmultivariatefluctuatingasymmetryacross
ageographicalrange
J. JUSTE, C. LO’PEZ-GONZA’LEZ R.E.STRAUSS
*EstacionBiologicadeDonaana(CSIC),Sevilla,41080,Spain
DepartamentodeBioquimicay BiologiaMolecularIV,FacultaddeVeterinaria, Universidad ComplutensedeMadrid,Madrid,Spain
Departmento BiologicalSciences,eeaaseechUniversity,uubboc ,e 9440431111,US
Keywords:
Eidolonhelvum; fluctuating asymmetry; fruit bats;
geographicalvariation; GulfofGuinea; Rousettusegyptiacus.
Abstract
Usingasetofcranialmorphometric characters, trendsofvariation in multivariate fluctuating asymmetrywereevaluatedandcompared inpopu- lationsofAfricanfruit bats Rousettusegyptiacus andEidolonhelvumfromthe GulfofGuineaislands,and theadjacentmainland.Levelsofasymmetrywere comparedacrosspopulations andspecies, andsignificant differenceswere found in both comparisons. Differences coincided with species-specific patternsofmorphologicalandgeneticdifferentiation. Concordanceofcorre- lation matricesofasymmetrywasalsocompared.Results weresignificant; concordanceishypothesizedtobeaby-productofdevelopmentalprocesses thatproducethe‘fox-like’ morphologysharedby thesespecies.Consistency of asymmetrypatternssuggeststhatthedevelopmentalpathwayproducingitis highlycanalized.Apredictionoftheabovehypothesisisthataradicalchange inthe‘fox-like’ structuralpatternwouldresultinbreakageoftheasymmetry parameterassociatedwith it.
Introduction
Fluctuatingasymmetry(FA)isapatternofvariation in which asymmetry values are normally distributed around ameanof zero(Palmer,1994).Traditionally, FAhas beenconsideredtheproductofrandom‘errors’in development,andthusameasure ofdevelopmental stability(PalmerStrobeck,1986,1992;Palmer,1994). ThebiologicalsignificanceofFAisamatterunderdebate in thecurrent literature onasymmetry(Grahametal.,
1993; HutchinsonCheverud,1995; Kraak,1997; Grahametal., 1998).Thebasicassumptionhasbeen thathighlevelsofFAareassociatedwith disruptionsof genetic or environmental origin, and with a lower
Correspondence:DrJavierJuste,Estacio’nBiolo’gicadeDonana(CSIC), AvdaM Luisas/n,Sevilla41013,Spain.
Tel.:+34954232340;fax:+34954621125;
e-mail:
capacityof individuals to copewith the environment (PalmerStrobeck,1986;Parsons,1990;Palmer,1994; Markov,1995).Morerecently,ithasbeensuggestedthat increasedFAmaynotbenecessarilyanexpressionoflow
‘quality’, but rather another expressionofphenotypic variability, specificallyabet-hedging strategy(Simons
Johnston, 1997). Notwithstanding methodological
problems (Palmer Strobeck, 1986, 1992; Palmer,
1994),thereisalargebodyofevidence indicatingthat geneticorenvironmental variation amongpopulations oftenisreflectedintheamountanddirectionofvariation betweensidesofbilateralcharacters(PalmerStrobeck,
1986;Markov,1995;MøllerThornhill, 1998). Determinedorstochastic, theexpressionofasymmetry
isultimately theresultofageneticprogramtranslating into anintegrated,functional whole, andassociations with genetic,developmentalormorphological patterns canbeexpected. Morphological patternsresultfrom developmentalprocessesworkinginconcertondifferent
bodypartstoproduceafunctional structure.Perturba- tionsinthedevelopmentalenvironment affectingspeci- fic bilateral charactersin a similar way should be reflectedin asignificant associationor integration of theasymmetriesforthesecharacters,regardlessofthe sourceoftheperturbation (Soule’,1967;Leamy,1994). Characters that aredevelopmentally orfunctionally relatedareexpectedtobemorehighly correlatedthan characterswithoutthisrelationship,thisbeingthebasic hypothesisofmorphological integration (Leamy,1977,
1994;Cheverud,1982).Concordance ofFAsatthe individual level indicates the presenceof ‘individual asymmetry parameters’or IAPs(Soule’,1967;Leamy,
1984,1994).Analogously,concordanceofcharacterFAs
acrosspopulationsisindicative ofa‘population asym- metryparameter’orPAP(Soule’ Baker,1968;Leamy,
1992,1994).PresenceofaPAPwouldindicatethatlevels
of asymmetry of asetof charactersin population A would bearthesamerelationshiptooneanotherasin population B(Soule’,1967).SignificantIAPsarerarein theliterature,butsignificantPAPs havebeendetectedfor somepopulations (Soule’,1967;Soule’ Baker,1968; Leamy,1994;butseeJackson,1973).AsignificantPAP wouldsuggestthatwhicheverthemechanismcontrolling thelevelsofdevelopmentalnoisewithinapopulation,it actsacrossthewholephenotype(orasectionofit),and indirectly in all componentsofthe genepool (Soule’,
1967).
Comparativestudies ofFAamongpopulationsin island—mainlandsystemsarescarce.Isolation by dis- tance,immigration of‘foreign’genesintoislands,higher levelsofinbreedingininsularpopulations,andvarying degreesof habitat heterogeneity acrossthesesystems may affect coadaptation, and hence asymmetry, in islandpopulations.Nonetheless,theevidenceavailable is inconclusive and no general trends have been detected(Soule’,1967;SarreDean,1991).Thisstudy focuseson patternsof variation in cranial FA across populationsintwospeciesofAfricanfruit bats(Eidolon helvumandRousettusegyptiacus). Bothspeciesarewide- spreadin continental Africa, and alsooccur in the islands oftheGulfofGuinea,WestAfrica.Previous researchonthesespecieshasshownuniquepatternsof morphological andgeneticvariation acrosstheislands (JusteIba’nez,1993;Juste etal.,1996,2000).Weused asetofbilateral morphometric characters tocompare multivariate patternsofFAwithin andbetweenspecies, andwith other datasets.Weaddressedthefollowing specificquestions:(1)ArelevelsofFAcorrelatedwithin andacrosspopulations?(2)Aretheresignificantdiffer- encesin levelsof FAamongpopulations or between species?(3) Are there any geographicalpatterns of variation inFA,andiftheyexist,dotheyreflectother morphologicalandgenetic patternsofvariationinthese species?(4)Areasymmetrypatternsconsistent across species,or havethe patternsin eachspeciesevolved independentlyinadistinctway?
Fig.1Geographicalsettingofpopulationsofthefruit batsEidolon helvumandRousettusegyptiacusexamined, andgeographical distances betweenthem.CA=Cameroon (Mount Cameroon area);
RM=Ri’oMuni; B=Biokoisland;P=Pri’ncipeisland;ST=Sao
Tome’ island;A=Annobo’nisland.R.egyptiacusdoesnot occurinAnnobo’n.
Material and methods
Studyarea
TheislandsofBioko,Pri’ncipe,SaoTome’ andAnnobo’n constitutetheonly insularsystem offshorethesub- SaharanAtlanticcoastofAfrica(Fig.1).Theislandsare alignedalong avolcanicfracturestretchingNE—SWfrom Mont Kupe in inland Cameroonto St.Helenaisland (Tera’n,1962;Eisentraut,1963).Pri’ncipe,SaoTome’ and Annobo’naretypicaloceanicislands,whereasBiokoisa
‘land-bridge’ island,which hasbeenperiodically con- nectedtothemainlandduringpastclimaticcycles.As a result,Bioko’sfloraandfaunaarerelativelyspecies-rich andpoorinendemisms.Theoceanicislands,incontrast, showlowrichness,buthighlevelsofendemismatthe specificandgenericlevels(Exell,1968;Jones, 1994).The smallareaandrelatively longdistancesamongoceanic islands(Fig.1)havepredominantly determinedcolon- izationeventsfromthemainlandtoeachisland,gener- ating distinctinsularcommunities.Followingthis general trend,batcommunitiesshowa highproportion of endemismson eachof the oceanicislands,andonly three species(Hipposideros ruber,Eidolonhelvum,and Rousettusegyptiacus)aresharedbymorethantwoofthem (JusteIba’nez,1994).Samplesfrominsularpopulations
ofthefruit batsEidolonhelvumandRousettusegyptiacus were examined.R.egyptiacus occursin all the islands exceptthemost-distantAnnobo’n,whereasE.helvumis spreadthroughout the system.Foreachspecies,two populationsfromthemainlandwereusedasreference populations,onefromtheMount Cameroon area, originated from thesame volcanicfracturethan the islands, andtheotherfromRi’oMuni,locatedsouth-west of Cameroon, and approximately equidistant to all islands(Fig.1).
Datacollection
Landmarkcoordinatesweredigitizedfromatotalof224 skullsofEidolonhelvum(Cameroon,44;R’io Muni, 49; Bioko,37;Pri’ncipe,38;SaoTome’,29;Annobo’n,27)and
241Rousettus egyptiacus (Cameroon,66;Ri’o Muni, 54;
Bioko,44;Pri’ncipe,39;SaoTome’,38)usinga3-DReflex Microscope (ReflexMeasurementLtd,Butleigh,UK). Fromeachskull,asetof82landmarkpointswastaken acrossall areasandon threedimensions.Fromthese landmarks,74pairsofbilateraldistanceswerecalculated following theprotocoldescribedinJusteetal.(2001).
Statisticalanalyses
SignificanceofAAandDApatternswereexaminedfor eachofthe74characters prior toall multivariate analyses.A subsampleof 31 Eidolonhelvumand 30
Rousettus egyptiacus drawn at random from the total
samplewasanalysedfollowing themethodologyrecom- mendedbyPalmer(1994).Presenceof AAinthesamples wastestedontherawdata(right-minus-leftdifferences), byspecies,usingtheShapiro—Wilk’stestfornormality (SASInstitute, 1995).Thesubsamplesweremeasured twice,andamixed-model sides-by-individualsanalysis (PalmerStrobeck, 1986)wasperformedforeach species.SignificanceofDA(the‘sides’component), as wellasthemagnitudeofmeasurementerrorwithrespect tonondirectionalasymmetry (thesides-by-individuals interaction component) weretestedoneachofthe74 variables. DAwasstatisticallysignificantonlyfor10 charactersofE.helvumand10ofR.egyptiacus.Thesides- by-individuals interaction term oftheANOVA wasnot significant for nine charactersofE.helvumand11of R.egyptiacus.Additionally, correlationanalysisfoundno significantrelationshipbetweenasymmetryandsize,and AA wasnot significant for any character.Resultsof univariate analysesarediscussedindetailinJusteetal. (2001).
Themainobjectiveofthispaper istoexaminepatterns ofvariationinmultivariatefluctuating asymmetry across islandpopulationsandbetweenspecies.Therefore,only variableswith nosignificantDA,andsignificantinterac- tion term in the sides-by-individuals ANOVA (Palmer,
1994)wereusedinfurther analyses(Justeetal.,2001). Thusthedatasetconsistedof55variablesforE.helvum
and57forR.egyptiacus.Because nosignificantrelation- shipwithsizewasdetected,datawerenotscaledbysize, butonlylog-transformedtolinearizepossibleallometric relationships.Inordertokeepadequatesamplesizesin multivariatecomparisons,missingvalueswereestimated usingtheexpectation-maximization algorithm ofLittle
Rubin(1987).
Becauseantisymmetry wasnotsignificant,univariate asymmetryin thissamplerepresents‘true’ fluctuating asymmetrysensuPalmer(1994),anapparentlyrandom patternofvariationbetweensidesforacharacterwithin apopulation,normallydistributed,andwithameannot significantlydifferentfromzero.Whenconsideringmul- tivariatefluctuatingasymmetrywe extend theconceptof FAtomore thanonevariableandtheircorrelations,with noaprioriassociationofpatterntoanyprocess.
Absolutedeviationofasymmetryvalues(logr— logl) from their meanwascalculatedforallindividuals, by character,population and species.Thesetransformed asymmetryvalueswereused inallfurther analyses. Although themeanabsolute deviationforanyone characteris notequivalenttoavariance(meansquared deviationofasymmetryvalues),itisameasureofspread aroundthemeanthatallowsforparametricunivariate andmultivariate comparisonsofvariancesviaLevene’s test(Schultz,1985).Apreliminary two-waymultivariate analysisofvariance(MANOVA)wascarriedouttotestfor significantdifferences inFAbetweensubadultsand adults(specimenswith fully growndentition andwith bone sutures fused) and between sexes,pooling all populationswithineachspecies.Nosignificant differences weredetected,eitherformainfactorsorinteractionterm foranyspecies(E.helvum:sex,F65,152 =0.82,P=0.78; age,F65,152 =1.29,P=0.05;interaction, F65,152 =1.02, P=0.43. R. egyptiacus: sex, F63,172 =0.82, P=0.80; age,F63,172 =1.32,P=0.08;interaction, F63,172 =0.9, P=0.64). Therefore,sexesand ageswere pooled in further tests.
Toestimatethedegreeofassociation(integration) of FAin the skull, Spearman’srank correlation and its significancewasestimatedfor allpossible pairsof characters,bypopulation. Correspondenceofcharacter asymmetriesacrosspopulationswasassessed bycalcu- lating theSpearman’srank correlation ofmeanasym- metry valuesofall charactersfor all possiblepairsof populationswithinaspecies.Significanceofcorrelations indicatesthe presence ofapopulation asymmetry parameter(PAP).Additionally, we testedfor pairwise associationsin multivariate FA between populations usingMantel’s tests,significanceofthesetestsalsois indicative of a PAP.MANOVA was used to test for differencesinlevelsofmultivariate FAbetweenpopu- lations.Analyseswererunbyspecies,with populations astreatments.Toobtainavisualizationofthetrendsof asymmetrybeingcompared, thegrandmeanofthe asymmetryvaluesforeachcharacterwas calculated,by population andspecies.
Forspeciescomparisons, populationswerepooledinto onesamplebyspecies (excluding Annobo’n, in which R.egyptiacusdoesnotoccur).Concordanceofcharacter asymmetrieswastestedon the variablescommon to bothspecies(48)bycalculatingthemean asymmetry valueofeachcharacterforeachspeciesandthen the rank correlation betweenthem. Usingthesamesetof characters,matricesofpairwise correlationsofcharacters werecompiledforeachspeciesandthesignificanceof theirassociationtestedusingMantel’stests.Differences inlevelsofasymmetrybetweenspeciesweretestedvia MANOVA (multivariate Levene’stest);atwo-wayanalysis species-by-population wasperformedtocontrolfor differencesbetweenpopulationswithineachspecies.
Weassessedtherelationshipsofasymmetryvariation with other morphological (averagetrait size),genetic (allozymic variation) andgeographicalpatternsofvari- ationacrosstheislands.Generalpatternsofmultivariate FAwereevaluated byspeciesbycalculatingpair-wise Mahalanobisdistancesamongallpossiblepairsofpopu- lations.Fromthese,adistancematrix wasconstructed andusedwith aneighbour-joiningalgorithm(Swofford etal.,1996)toconstructunrootedtrees,whichsummar- izerelationshipsofmultivariate FAsamongpopulations for eachspecies.In thesetrees,branch lengths are proportional tothedistanceamongpopulations.Similar matricesofgeneticdistancesbetweenpairsofpopula- tions,byspecies,werebuilt from allozymicdatausing Rogers’(1972) distances(Justeetal.,1996,2000).To
specified, all statistical analyseswere performed in
Matlab for Windows ver. 4.2c(TheMathWorks Inc.,
1994).
Results
Concordanceofasymmetries
CorrelationsbetweencharacterFAswerenotsignificant for most pairsof charactersin the populations after Bonferroni correction(Table1).Inallcasesexceptone, lessthan 1% of all possiblepair-wise correlations of characterasymmetries (1485combinationsforE.helvum,
1596forR.egyptiacus) weresignificant. Thereisalow
degreeofintegration incharacterasymmetryforthese species,andnoevidencefor anindividual asymmetry parameter.Incontrast,meancharacterFAswerestrongly correlated between pairs of populations; all testsof
Table2Testsforsignificanceofassociationsofasymmetriesacross populations.Uppermatrices:valuesofSpearman’srank-correlations ofmeanvaluesofmultivariate FAsbetweenpairsofpopulations. Lowermatrices:standardizedMantel’sstatistics(r)forallpossible pair-wisecomparisonsofpopulationmatricesofmultivariate FA correlations.Allcomparisonswerehighlysignificant(P0.001) after Bonferroniadjustmentof significancelevel( =0.05)persetof testsandspecies.CA=Cameroon;RM=Ri’oMuni; B=Bioko;
P=Pri’ncipe;ST=SaoTome’;A=Annobo’n.
assessvariation inmeantraitsize,amatrix ofmorpho-CA / RM / B / P / ST / Alogicaldistanceswasgeneratedbytakingtheaverageof therawmeasurementsonrightandleftsidesoftheskull (r+l/2)foreachobservationandcalculatingtheMaha- lanobisdistancesbetweenpopulationsforeachspecies. Finally,amatrix ofgeographicaldistancesamonglocal- itieswasconstructed.Genetic,morphometric andgeo- graphicaldistance matriceswerecomparedagainst multivariate FAmatricesusingMantel’stests.Asequen- tial Bonferroni adjustment(Rice,1989)of (expected TypeIerrorforanumber ofsimultaneoushypotheses being tested) was used to decide on the statistical significanceofmultiplesimultaneoustestsataglobal
Eldolonhelvum
CA—0.9300.8850.8830.9080.840
RM0.151—0.9120.9530.9320.908
B0.1640.120—0.9040.9260.904
P0.1970.1490.194—0.9270.920
ST0.1660.1620.0990.100—0.867
A0.1310.1510.1220.1110.151—
Rousettusegyptlacus
ST0.1830.1900.1980.153—
levelof0.05bysetoftestsandspecies.Unlessotherwise
Table1Minimum,maximumandmeanvaluesofSpearman’srank-correlations(integration) ofcharactermultivariate FAsforallpossible pair-wisecomparisonsofasymmetryvariables,bypopulationandspecies.Lastrow(%vars.)isthepercentageofsignificantcorrelations(table- wide =0.05)aftertheBonferroniadjustment(table-widea=0.05perspecies)for1484and1596comparisonsforE.helvum andR.egyptiacus, respectively.CA=Cameroon;RM=Ri’oMuni; B=Bioko;P=Pri’ncipe;ST=SaoTome’;A=Annobo’n.
EldolonhelvumRousettusegyptlacus
CARMBPSTACARMBPST
Min. / —0.440 / —0.410 / —0.553 / —0.468 / —0.592 / —0.587 / —0.458 / —0.382 / —0.429 / —0.486 / —0.636Max. / 0.679 / 0.639 / 0.700 / 0.737 / 0.849 / 0.774 / 0.737 / 0.845 / 0.889 / 0.886 / 0.747
Mean / 0.021 / 0.045 / 0.021 / 0.038 / 0.034 / 0.017 / 0.028 / 0.025 / 0.035 / 0.042 / 0.020
%vars. / 0.404 / 0.135 / 0.135 / 0.202 / 0.337 / 0.135 / 0.934 / 1.002 / 0.627 / 0.439 / 0.313
association were highly significant in both species (Table2), suggesting the presence of a population asymmetryparameter.Similarly, pair-wisecomparisons ofcorrelationmatrices(Mantel’stests)acrosspopulations were highly significant (Table2). The relationships amongcharacterFAs weresimilarforbothspecies; correspondenceofaverage asymmetriesbetween specieswasalsohighly significant in both the rank correlation (r=0.94,P0.001,n=48) andMantel’s tests(r=0.53,P0.001,n=48).
Levelsofasymmetry
MANOVAs for differences among populations yielded highly significant results in both species(Table3). Annobo’nhadhigherlevelsofasymmetrythanallother populations of E. helvum,whereas for R. egyptiacus
Table3Averagemultivariatefluctuatingasymmetry(FA)valuesby populationandspecies,andresultsofmultivariateLevene’stestsfor differencesinFAlevelsamongpopulations,byspecies.CA=
Cameroon;RM=Ri’oMuni; B=Bioko;P=Pri’ncipe;ST=Sao
Tome’;A=Annobo’n.
EldolonhelvumRousettusegyptlacus
CA 0.0307 0.0280
RM 0.0333 0.0308
B 0.0341 0.0309
P 0.0353 0.0335
ST 0.0336 0.0327
A 0.0384 —
d.f. 5/275/822.8 4/228/721.6
2 0.1873 0.25038
1. 1.1940 1.3118
P 0.032 0.005
average asymmetrywashigher for SaoTome’ and Pri’ncipe.Neighbour-joiningtreessummarizethe patterns of variation in multivariate FA levels (Fig.2). In E.helvumAnnobo’nisclearlyseparatedfromtherestof thepopulationsinthetree.Theremainingislandscluster togetherandthen with CameroonandRi’o Muni. In R.egyptiacusawell-resolvedtreeemerged,inwhichR’io Muniand Bioko(thecontinentalisland),and then Cameroon,aregroupedtogether, whereasSaoTome’ andPri’ncipe (the oceanicislands)areseparatedin a different branch (Fig.3). AverageFA washigher for E.helvum(Table3).Thespecies-by-populationtwo-way MANOVA yieldedhighly significant differencesbetween species(F48,381 =2.95,P0.001).Theinteraction term species-by-population was nonsignificant (F192,1523=
1.16,P=0.08),indicating similarvariation inlevelsof
asymmetryacrosspopulationsforbothspecies.
Fluctuatingasymmetry andothersources ofvariation
AssociationbetweenFAand morphologicalmatriceswas highlysignificantforR.egyptiacus.Concordancebecomes evidentwhentheneighbour-joiningmorphologicaltree andtheFAtreearecompared (Fig.3).Theformershows little differences betweenmainlandpopulationsand Bioko,whereastheoceanic islandsclusterseparatelyand, withinthiscluster,SaoTome’divergesfromPri’ncipe.The remainingcomparisonsbetweendatasetsaremarginally significant for R. egyptiacus and nonsignificant for E.helvum,butallbecomenonsignificantafterBonferroni adjustment (Table4). Nonetheless,the morphological treeforE.helvumiswellresolved.Itconsistsofacluster includingthemainlandpopulationsand Bioko;the remaining populations join this cluster following the
order of their geographicalpositions. The Annobo’n
ST
A)
B
P
CA
RM
A
B
B)RMC)
CA P
ST
A
Fig.2Neighbour-joining dendrograms showingpatternsofvariationinthepopula- tionsofEidolonhelvumexamined.
(A)MahalanobisD2distancesoncharacter
fluctuating asymmetries(FAs).(B)Maha-
STlanobisD2distancesonmorphologicalvari-
Pables.(C)WagnertreebasedonRogers’ RM(1972)geneticdistances;datafromJuste etal.(2000).CA=Cameroon;RM=Ri’o
BMuni; B=Bioko;P=Pr’incipe;ST=Sao
ATome’;A=Annobo’n.
Fig.3Neighbour-joining dendrograms showingpatternsofvariationinthe
populationsofRousettusegyptiacusstudied. (A)MahalanobisD2distancesoncharacter fluctuating asymmetries(FAs).(B)
MahalanobisD2distancesonmorphological
variables.(C)WagnertreebasedonRogers’ (1972)geneticdistance;datafromJusteetal. (1996).CA=Cameroon;RM=Ri’oMuni;
B =Bioko;P=Pr’incipe;ST=SaoTome’.
Table4Mantel’stestsforconcordanceofthefollowingmatricesof distancesbetweenpopulations,byspecies:FA:Matrix ofMahala- nobisD2distancesofFAs.MOR:Matrix ofMahalanobisD2on74 averageskullvariables(r+l/2).GEN:Matrix ofRogers’(1972) geneticdistances(populationfromCameroonmissing)basedondata
fromJusteetal.(1996)forR.egyptiacusandfromJusteetal.(2000) forE.helvum.GEO:Matrix ofgeographicaldistancesusing(great- circle) arcdistances (RMisnotconsideredbecauseitisequidistantto alltheislands).
developmentalnoise(Grahametal.,1993).Underthis model,geneticorenvironmental stressmayincreasethe levelsofnoise(asymmetry),butevenintheabsenceof perturbations,acertainlevelofFAis tobeexpectedfor anyparticulardevelopmentalpath.Small disturbances duringtheprocessmayquickly leadtodeviationsfrom the symmetrical condition due to cascadingeffects. Negativefeedbackamonggrowingcellsoneachside,or betweenright andleftstructures,would keepbilateral
anyfurtherstepduringtheprocesswouldbebufferedat
*SignificantconcordanceafterasequentialBonferroniadjustmentof table-wiseP-valuesto =0.05,byspecies.
populationdivergesfromtherest,inagreementwiththe
FAtree(Fig.3).
Discussion
Patterns ofconcordanceofFAs
Pair-wise correlationsofFAamongcranialcharacters suggestthatthereisnointegration ofFAwithinpopu- lations,i.e.thereisnoevidence foranindividual asymmetry parameter (IAP) sensu Soule’ (1967) and Leamy (1994). IAPsare seldom detected(e.g. Lens
vanDongen,1999).Thereasonmaybeaweaksignal, orfailuretolookattherightcharacters(Leamy,1994), butalso,absence ofan IAP maybethenormalcondition ifFAreflectsfluctuations consequent onthenormal developmentalprocessesofbilateralstructures,aswellas
process. Thenetresultwouldbeanasymmetrypatternin whichlevelsofFAononecharacterarenotnecessarily correlatedwith levelsofFAinothercharacters.
Underthe ideaofanormalor ‘base’levelofasymmetry
associatedwith a developmental process,we would expectthat,onaverage,individualsina populationhad similarlevelsofFAforagivencharacter.Allcharacters measuredinthesamesubsetofindividuals wouldhave anassociated levelofasymmetry,sothat avectorof averageFAsfor asetof characterswould reflect the developmentalprocessthatgivesrisetoamorphological pattern.Wefound asignificant population asymmetry parameter (PAP) in each speciesexamined, which furthermore isthesameforbothspeciesstudied.These resultssuggestthepresenceofaskull-wide patternof characterFAs,whichispreservedacrosspopulationsand, beyondthat,acrossspecies.
E.helvumandR.egyptiacus shareasimilar, ‘fox-like’ skullmorphology,characterizedbyhavingthecraniofacial axisalmosthorizontal,andthebraincaseslightlyvaulted
andraisedabovethefaceline (Andersen,1912).This similarity ledpreviousauthorstoclassifyRousettusand Eidolonas sistergenera(Andersen, 1912; Bergmans,
1997).However,moleculardataindicatethattheyare
phylogenetically separated (Hollar Springer,1997; Justeetal.,1999),andthatmorphologicalresemblance isduetoaveryconservative,non-specialized,primitive structural pattern for the family Pteropodidae(Kirsch
et al., 1995;A’lvarez etal., 1999). Theremay exist a
generalasymmetryparametercommon toall‘fox-like populations’ in Pteropodidaethat isaproperty ofthe developmental process that producesthat structural pattern.Underdifferentgeneticorenvironmentalcondi- tions, levelsof FAmay change(seebelow), but the generalasymmetrypatternwould persistaslongasthe morphologicalpatterntowhich itisassociatedpersists. Because amorphologicalpatternistheproductofgenetic as well as epigenetic processes,organism-wide FA patternsmayultimately haveageneticorigin,asimplied by Soule’(1967).However,thepatternwouldnot necessarilyappearas aco-ordinated response to increaseddisturbanceordecreasedbuffering,butwould ratheritbeaproductofthedevelopmentalsystem.
Differences influctuatingasymmetry levels
Levels ofmultivariate FAweresignificantly different amongpopulationswithin species,andbetweenspecies. Patternsof multivariate FAbased on Mahalanobis distances mostly reflectdifferencesin magnitude of asymmetry.ForE.helvum,theFAtreebasedonMaha- lanobisdistancesidentified theAnnobo’npopulation as being most distinct, whereas the remaining islands clusterwith themainlandpopulations.Clusteranalysis basedongeneticdistancesshowedapatterninwhichSao Tome’andPr’incipeclustered together,R’ioMuni clus- teredwith thesetwoinamajorgroup,andAnnobo’nis segregatedfromtherest(Fig.2).Morphologicalevidence alsoreflects this trend; E. helvumfrom Annobo’n is distinguishablefrom theremaining conspecificpopula- tions by its smaller overall size,to suchextent that subspecificstatusisrecognizedforthispopulation(Juste etal.,2000).
ForR.egyptiacus, levelsofmultivariate FAwerealso significantly different. Thetreebasedon Mahalanobis distancesofFAvaluesclearlydistinguishedtheoceanic islandsfrom themainland andBiokopopulations.The patternofFAcorrespondedtothatofgenetic,morpho- logicalandgeographicaldistances.Lackofsignificancein testsofassociationwith geographicalandgeneticdis- tancesmay reflect areduction in power due to the exclusionofthepopulationfromCameroon inthese comparisons.Allozymic dataplacedBiokoclosertoR’io Muni, whereasPri’ncipe and SaoTome’ clusteredto- gether and separatelyfrom them (Fig.3). Basedon morphometricandqualitativeexternalcharacters,three subspeciesofR.egyptiacusaredistinguishedinthisarea,
onefortheCentralAfricanmainlandandBioko(Eisen- traut,1959),oneendemictoPri’ncipe,andonefromSao Tome’ (JusteIba’nez,1993).Clusteranalysisofcranial morphometric data(Fig.3)reflectedtheserelationships amongpopulations.
Studiesofgenetic andphenotypicvariationinisland— mainlandsystemsusuallyconcernthepurportedgenetic andmorphological changes that could result from isolation. Island populations might be under genetic stressproduced by lossof heterozygosity and poly- morphism,orenvironmental stressproduced bysigni- ficantly harsher environments on the islands. The higherthelevelofstress,thelowerthecapacityofthe organismto‘correct’forerrors,andthehighertheFA inthepopulation(VanValen,1962;PalmerStrobeck,
1986).Thereisnoevidenceofdecreasedorincreased heterozygosity intheislandsforeitherspecies.Levelsof genetic polymorphism are slightly lower only in E. helvumfrom Annobo’n, whereas in R. egyptiacus polymorphism ishigherintheislands thaninthe mainland (Justeet al.,1996,2000). Moreover, both speciesarerelatively abundantin theislands(JusteIba’nez,1994),indicatinglargeeffectivepopulationsizes. Forthesetwospecies,variation inlevelsofasymmetry doesnotseemtobetheproductofstressduetogenetic disruptions.Overallenvironmentalconditionsmay be moreextremeintheislands(Fuster, 1955),forinstance, availability of resourcesmay be diminished in the islandsascomparedtothemainland(Exell,1968).On theotherhand,however,insularfruit batpopulations may benefitfromlowercompetitioninmuchreduced communitiesofbats andbirdsascomparedtothe mainland(Jones,1994;JusteIba’nez,1994).Observed asymmetrypatterns,particularly thatofEidolonhelvum, donot correspondto what would beexpectedif the insularcondition wasthemaindeterminant oftheFA levels.Infact,forthisspecies,onlyAnnobo’npopulation branches offwhile theotheroceanicislandscluster togetherwith themainland populations(Fig.2).Addi- tionally, increasedenvironmental stressdoesnotaffect FAlevelsin recentexperimental studies(David etal.,
1998;Bjorkstenetal.,2000).
Weproposeanalternative explanation for the observeddifferencesinlevelsofFAbasedonthemain assumptionthatinbothspeciesa processofgeneticand morphologicaldifferentiation isoccurringorhasrecently occurredintheislandsystem. Thisassumptionis sustainedbytheshortgeneticdistancesfound among thesepopulations(Justeetal.,1996,2000).Wesuggest that higher levelsof FAin thesepopulations arethe result of arelaxation of canalization associatedwith theprocessofdifferentiation. Thisrelaxationisnotthe product ofenvironmental pressures,but aproperty of thedevelopmentalsystem, which mayfacilitatethe changesoccurringintheislandpopulations,ormaybe theconsequenceofthem.Ineithercase,itwouldmean lesscontrol ofthemechanismsthat regulatelevelsof
asymmetryduringdevelopment,resultinginanincrease inasymmetryinthedifferentiating populations.
Observedpatternsaremoreconsistentwiththis predic- tion.InE.helvum,alarge,vagilebat,differentiationoccurs onlyinAnnobo’n,thepopulationwithsignificantlyhigher levelsofFA.R.egyptiacusisasmallerbatwith specific roostingrequirementsandlowervagility;isolationofthe oceanicpopulations istherefore more effective(Juste etal.,1996),andpopulations on eachof the oceanic islands arefollowing their own evolutionary paths. Accordingly,higher levelsofFAwereobservedinPri’ncipe andSaoTome’.FAisusuallyviewedasameasureoffitness; nonetheless,anincrement in the phenotypic but not necessarilythegenotypicvariancecouldconferisland populationsahigherprobability ofsurvivalinthenew environmental conditions(Simons &Johnston,1997). Althoughincreasedasymmetry mayrepresent ‘lowqual- ity’,itmayalsobeanadvantageinnewenvironments,or evenbeirrelevanttothesurvivaloforganisms.
Generaltrends across islands
Twogeneraltrendsofasymmetryvariationacrossislands arediscerniblefrom thisanalysis.First,skull-wide FA patternsremain consistentacrosspopulations in both species,and acrossspecies.Second,multivariatelevelsof asymmetrydiffersignificantlyacrosspopulations;differ- encescorrespondtothepatternofmorphologicaldiver- genceamongpopulations,andtheyalsoareassociated with observedpatternsofgeneticdistances.
DifferencesinmagnitudeofFAmaybeassociatedwith uniqueenvironmental pressuresorgeneticchange(dif- ferentiationor stress),whereastheskull-wideasymmetry patternseemstobeaby-productofthedevelopmentofthe skull, andwill persistas longas thecranialstructural pattern does.Patternsof averageFA seemto beas canalizedasthebasiccranialmorphology; nonetheless, canalizationmaybreakdownwhentheontogenyswitches fromone developmentalpathwaytoanother,assome thresholdvalueofgenotypicorenvironmental effectsis attained(Zakharov,1992; SchlichtingPigliucci,1998). Forthespeciesanalysed,apredictionofthishypothesisis thataradicalchangeinthe‘fox-like’ patternwouldresult inthebreakageoftheasymmetryparameterassociated withthatmorphology.Itremainstobeinvestigatedhow radicalandofwhatnaturethechangewouldhavetobeto produceadifferentskull-wide asymmetrypattern.Also, wemaywanttoaskwhatkind ofasymmetrypatterns would beexpected ifwecouldgeneratesimilar morphologiesthroughdifferentdevelopmentalprocesses.
Acknowledgments
ThankstoR.D.OwenandM.Grimsonfortheirsupport duringthiswork,toR.Cifelliforhis encouragementand forkindly loaningtheReflexmicroscope,toC.Iba’nez, E.Costas,F.HerreraandK.Jarrettfor their constant
support.The following institutions kindly loanedspeci- mens:AmericanMuseumofNaturalHistory(NewYork, USA), The Alexander Koenig Museum (Bonn, Germany),TheCarnegie MuseumofNaturalHistory (Pittsburgh,USA),Estacio’nBiolo’gicadeDonana(CSIC) (Sevilla,Spain), NaturalHistoryMuseum(London,UK), The OklahomaMuseumofNaturalHistory(Norman, USA),TheTulaneMuseum(BelleChasse,USA),Texas Tech Museum(Lubbock,USA),TheSmithsonian Institution(Washington, USA)andTheRoyalOntario Museum(Ontario, Canada).TheSpanish Ministerio de Educacio’nyCienciaandtheDepartmentofBiological SciencesatTexasTechUniversity financedaPost-Doc positiontoJ.J.,andtheSpanish MaBCommitteeof UNESCOfundedtravelexpensesforC.L.G.
References
A’lvarez,Y.,Juste,J.,Tabares,E.,Garrido-Pertierra,A.,Iba’nez,C.
Bautista,J.M.1999.Molecularphylogenyandmorpholo- gicalhomoplasyinfruit bats.Mol.Biol.Evol.16:1061—1067.
Andersen,K.1912.Catalogo theChiroptera o theBritishMuseum. I.Megachiroptera.BritishMuseum(NaturalHistory),London. Bergmans,W.1997.TaxonomyandbiogeographyofAfrican fruit bats (Mammalia, Megachiroptera). V. The genera uissonycterisAndersen1912,MyonycterisMatschie,1899,and MegaloglossusPagestencher,1885;generalremarksandcon-
clusions;annex:keytoallspecies.Beauortia47:11—90. Bjorksten,T.,David,P.,Pomiankowski,A.Fowler,K.2000.
Fluctuatingasymmetryofsexualandnonsexualtraitsinstalk-
eyedflies:apoorindicatorofdevelopmentalstressandgenetic quality.J.Evol.Biol.13:89—97.
Cheverud,J.M.1982.Phenotypic,geneticandenvironmental morphological integration in the cranium. Evolution36:
499—516.
David,P., Hingle,A.,Fowler,K.Pomiankowski,A.1998.Male sexualornamentsizebutnotasymmetryreflectsconditionin stalkedeyedflies.Proc.R.Soc.uond. B265:2211—2217.
Eisentraut, M. 1959. Der Rassenkries Rousettus aegyptiacus
E. Geoff.Bonn.Zool.Beitr.10:218—235.
Eisentraut,M. 1963.DieWirbeltiere DesKamerungebirges. Paul
Parey,Hamburg.
Emlen,J.M.,Freeman,D.C.Graham,J.H.1993.Nonlinear growthdynamics andtheoriginoffluctuating asymmetry. Genetica89:77—96.
Exell,A.W.1968.Pr’incipe,SaoTome’ andAnnobo’n. ctaPhyt. Suec.54:132—136.
Fuster,J.M.1955.Vulcanolog’iadelAtla’nticoMeridional. rch. I.E. .33:67—79.
Graham, J.H., Freeman, D.C., Emlen, J.M., Leamy, L.J. Kiesler,J.A.1998.Directionalasymmetryandthemeasure- mentofdevelopmentalinstability.Biol.J.uinn.Soc.64:1—16. Graham,J.H.,Freeman.D.C.Emlen,J.M.1993.Antisymmetry, directionalasymmetry,anddynamicmorphogenesis.Genetica
89:121—137.
Hollar, L.J.M. Springer, S. 1997. Old World fruit bat phylogeny: evidence for convergent evolution and an endemicAfricanclade. Proc.Nat. cad.Sci.US 94:5716—5721. Hutchinson, D.W.Cheverud,J.M.1995.Fluctuating asym- metryintamarin(Sanguinus)cranialmorphology:intra and