Hybrid ancestry of an island subspecies of Galápagos mockingbird explains discordant gene trees

12 Pages·2013·1.9 MB·English

by  Pirmin Nietlisbach

MolecularPhylogeneticsandEvolution69(2013)581–592
ContentslistsavailableatScienceDirect
Molecular Phylogenetics and Evolution
journal homepage: www.elsevier.com/locate/ympev
Hybrid ancestry of an island subspecies of Galápagos mockingbird
explains discordant gene trees
Pirmin Nietlisbacha,⇑, Peter Wandelera, Patricia G. Parkerb, Peter R. Grantc, B. Rosemary Grantc,
Lukas F. Kellera, Paquita E.A. Hoecka,1
aInstituteofEvolutionaryBiologyandEnvironmentalStudies,UniversityofZurich,Winterthurerstrasse190,8057Zurich,Switzerland
bDepartmentofBiology,UniversityofMissouri–St.Louis,R223ResearchBuilding,OneUniversityBoulevard,St.Louis,MO63121,USA
cDepartmentofEcologyandEvolutionaryBiology,PrincetonUniversity,106AGuyotHall,Princeton,NJ08544,USA
a r t i c l e i n f o a b s t r a c t
Articlehistory: Introgressionofgenesthroughhybridizationhasbeenproposedtobeanimportantdriverofspeciation,
Received15April2013 butinanimalsthishasbeenshownonlyinrelativelyfewcasesuntilrecently.Additionally,introgressive
Revised17July2013 hybridizationamongnon-sisterspeciesleadstoachangeinthegenetreetopologyoftheconcernedloci
Accepted18July2013
and thus complicates phylogenetic reconstruction. However, such cases of ancient introgression have
Availableonline30July2013
beenverydifficulttodemonstrateinbirds.Here,wepresentsuchanexampleinanislandbirdsubspe-
cies,theGenovesamockingbird(Mimusparvulusbauri).Weassessedphylogeneticrelationshipsandpop-
Keywords:
ulation structure among mockingbirds of the Galápagos archipelago using mitochondrial and nuclear
Introgression
DNAsequences,autosomalmicrosatellites,andmorphologicalmeasurements.Mitochondrialhaplotypes
Hybridization
ofGenovesamockingbirdsclusteredcloselywiththehaplotypesfromtwodifferentspecies,SanCristóbal
Phylogeny
Speciation (M.melanotis)andEspañola(M.macdonaldi)mockingbirds.Thesamepatternwasfoundforsomehaplo-
Mimus typesoftwonucleargeneintrons,whilethemajorityofnuclearhaplotypesofGenovesamockingbirds
Nesomimus weresharedwithotherpopulationsofthesamespecies(M.parvulus).At26autosomalmicrosatellites,
GenovesamockingbirdsgroupedwithotherM.parvuluspopulations.ThispatternshowsthatGenovesa
mockingbirdscontainmitochondriaandsomeautosomalallelesthathavemostlikelyintrogressedfrom
M.melanotisintoalargelyM.parvulusbackground,makingGenovesamockingbirdsalineageofmixed
ancestry, possibly undergoing speciation. Consistent with this hypothesis, mockingbirds on Genovesa
aremore clearlydifferentiated morphologically from otherM. parvuluspopulations thanM.melanotis
isfromM.parvulus.
(cid:2)2013ElsevierInc.Allrightsreserved.
1.Introduction tiveradiationofDarwin’sfinches(GrantandGrant,2008a,b,2009).
Apartfromitsroleinspeciation,hybridizationcancausecomplica-
Speciationisthekeyprocesscreatingtheastonishingbiological tionswhenattemptingtoreconstructphylogenies,especiallywhen
diversity that can be found on earth. Speciation may happen analysingonlyfewgeneticloci(FunkandOmland,2003).Ifhybrid-
through several routes, including hybridization (Schwenk et al., izationamongnon-sisterlineagesisatsomelocifollowedbythe
2008; Butlin et al., 2009). Although the potential importance of replacementofallelesinonelineagebyallelesfromtheotherline-
hybridization for speciation in plants has long been recognised, age,thetopologyofthegenetreesoftheselocidoesnotreflectthe
relatively few examples were known to exist in animals until speciestree(RheindtandEdwards,2011).Althoughthisphenom-
recently(Arnold,2006;Mallet,2007;MavárezandLinares,2008; enonofancientintrogressionispotentiallywidespread,ithasbeen
Schwenk et al., 2008). For example, hybridization appears to be extremely difficult to conclusively demonstrate in birds (Peters
one of the reasons for the rapid speciation in African cichlid fish etal.,2007;RheindtandEdwards,2011;Warrenetal.,2012).Here
(Salzburgeretal.,2002;Seehausen,2004,2006)andfortheadap- wepresentacaseofsuchtopology-changingancientintrogressive
hybridizationinGalápagosmockingbirdsanddiscussitspotential
⇑ contributiontomorphologicallineagedivergenceandspeciation.
Correspondingauthor.
The Galápagos Islands and their mockingbirds have become
E-mailaddresses:[email protected](P.Nietlisbach),peter.wandeler@
ieu.uzh.ch (P. Wandeler), [email protected] (P.G. Parker), [email protected] famousduetotheirkeyroleinDarwin’sformulationofhistheory
(P.R.Grant), [email protected] (B.R. Grant), [email protected](L.F.Keller), of evolution by natural selection (Chancellor and Keynes, 2006):
[email protected](P.E.A.Hoeck). Darwin’s Galápagos mockingbird specimens triggered his ideas
1 Currentaddress:InstituteforConservationResearch,SanDiegoZooGlobal,15600
about adaptive radiation, with a mainland form speciating into
SanPasqualValleyRoad,Escondido,CA92027,USA.
1055-7903/$-seefrontmatter(cid:2)2013ElsevierInc.Allrightsreserved.
http://dx.doi.org/10.1016/j.ympev.2013.07.020
582 P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592
(a) Galápagos mockingbird distribution and sampled islands (b) autosomal FIB7
Wolf
2
Pinta
6
Mimus parvulus Mimus parvulus bauri
Genovesa
Marchena
Santiago
N = 1 N = 25 N = 40
Rábida Mimus melanotis (c) autosomal TGF
Fernandina
San Cristóbal
Isabela Santa Cruz
Santa Fé
5
Mimus macdonaldi
Gardner-by-Floreana
Gardner-by-Española
Mimus trifasciatus
50 km
Champion Española
(d) factorial correspondance analysis of 26 autosomal microsatellite loci
5
0.
0.5 0
%) %) -0.5
7 4
2 ( 0 3 (
s s
axi axi -1
5 5
0. 1.
- -
2
-
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
axis 1 (9%) axis 1 (9%)
Fig.1. Autosomalgeneticdata.(a)MapoftheGalápagosarchipelagowiththedistributionofthefourrecognisedmockingbirdspeciesandcolour-codedislandssampledin
thisstudy(forsamplesizesseeTable2).(b)and(c)Median-joiningnetworksforthenuclearintronsFIB7andTGF,respectively.Piesrepresenthaplotypesandarecoloured
accordingtotaxon(seeFig.1a;blackpiesrepresenttheoutgroupMimuslongicaudatus)andsizedrelativetotheabundanceoftherespectivehaplotype(seegreyreference
pies;notethatNrepresentsthenumberofhaplotypes,notthenumberofindividuals).Numbersalonglinesrepresentthenumberofmutationsoccurringbetweentwonodes,
ifhigherthan1.Notethatforbothdatasets,individualsfromGenovesadisplayhaplotypesgroupingwithSanCristóbalandEspañola,aswellaswithotherM.parvulus.(d)
Plotsofthefirstthreeaxesofafactorialcorrespondenceanalysisof26microsatelliteloci.NotethatGenovesagroupsamongtheotherM.parvuluspopulations.
severalislandformsuponcolonisationofthearchipelago(Darwin, ingbirdtaxa.Theirtaxonomyhasexperiencedvariouschangesover
1839). The Galápagos archipelago, situated in the Pacific Ocean thelastcentury,butbasedonphenotypesmockingbirdsinGalápa-
960km west of mainland Ecuador, is inhabitedby several mock- gosarecurrentlyconsideredtoformfourdifferentspecies(Fig.1a;
P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592 583
Swarth, 1931; Remsen et al., 2012). After extinction on the large groupGenovesamockingbirdswithotherM.parvuluspopulations.
islandofFloreanaaroundtheyear1880,theFloreanamockingbird, (iv)GenovesamockingbirdsaremostcloselyrelatedtoM.parvulus
Mimustrifasciatus,isnowrestrictedtothetwosmallisletsCham- but dispersal of M. melanotis females from San Cristóbal has re-
pion and Gardner-by-Floreana (Curry, 1986; Grant et al., 2000). placedtheancestralmtDNAlineagesintheGenovesapopulation.
The Española mockingbird, M. macdonaldi, inhabits Española and Asthisscenariorequiresfemaleimmigrants,itiscompatiblewith
itsadjacentisletGardner-by-Española.TheSanCristóbalmocking- thegenerallyfemale-biaseddispersalinbirds(GreenwoodandHar-
bird, M. melanotis, is restricted to San Cristóbal Island, whereas vey,1982),including,onasmallgeographicscale,Genovesamock-
most islands in the north-west are inhabited by a widespread ingbirds (Curry and Grant, 1989). Under such a scenario, the
fourthspecies,M.parvulus,whichissplitintoseveralsubspecies. majority,butnotnecessarilyall,ofthenucleargeneswouldshow
Plumagediffersamongthesespecies(Swarth,1931;Cody,2005). agroupingofGenovesamockingbirdswithotherM.parvuluspopu-
Floreana mockingbirds show large dusky patches on each side of lations. A fifth explanation, proposed by Hoeck et al. (2010), is
thebreastanddarkspotsacrossit,andawhitishauricularregion. incomplete lineage sorting (Nichols, 2001; Funk and Omland,
Españolamockingbirdshaveabandofbrownblotchesacrossthe 2003). Under this scenario, ancestral polymorphisms are present
breastandaduskyauricularregion,whileSanCristóbalmocking- intheparentalpopulationpriortodivergence.
birdsalsohaveaduskyauricularpatchandnoorindistinctspot- Itisnotpossibletodistinguishbetweenthefivescenariosbased
ting on the breast. The fourth species, M. parvulus, has a whitish on one genetic locus alone, and thus more unlinked genetic data
unmarkedbreastandablackauricularregion.Mostofitssubspe- are needed in addition to the maternally-inherited mtDNA se-
ciesshowaratherdarkbackandcrown,whilethesubspecieson quencesanalysedbyArbogastetal.(2006).Hoecketal.(2010)ana-
Genovesa and Santa Fé have a lighter back with more distinct lysed variation at 16 microsatellite loci and found that Genovesa
streaksorspotsandagreyishcrown(Swarth,1931;Cody,2005). mockingbirdsgroupedwithotherM.parvuluspopulations,making
Theseallopatricspeciesalsodifferinmorphologicalmeasurements scenarios of introgression or incomplete lineage sorting most
(Swarth, 1931; Bowman and Carter, 1971; Abbott and Abbott, likely. Another recent study analysed mitochondrial and nuclear
1978; Curry, 1989; Curry and Grant, 1990), as well as in social lociofallMimidae(Lovetteetal.,2012)andconfirmedknownphy-
organisation (Curry, 1989), song (Gulledge, 1970), and foraging logeneticpatternsamongGalápagosmockingbirds,butitsonlyM.
behaviour (Bowman and Carter, 1971; Curry and Anderson, parvulussampleswerefromIsabelaandSantaCruz,preventingclo-
1987).Furthermore,theyrarelydisperseamongislands(P.R.Grant seranalysisofmanyrelationshipsamongGalápagosmockingbirds.
andB.R.Grant,ownobservations;R.L.Curry,personalcommunica- Herewedeterminewhichofthesescenariosmostlikelyexplains
tion)andthuspopulationsondifferentislandsaregeneticallywell theevolutionaryhistoryofGenovesamockingbirds,usingacombi-
differentiated(Hoecketal.,2010). nationofunlinkedanddifferentiallyinheritedgeneticmarkers.We
Arbogastetal.(2006)investigatedthephylogenyofGalápagos compare the gene trees of mitochondrial NADH dehydrogenase
mockingbirds and their non-Galápagos relatives using molecular subunit2(ND2)andthecytochromeb(CYTB)geneswiththeauto-
techniques.TheirmitochondrialDNA(mtDNA)sequencedatasug- somal beta-fibrinogen gene intron 7 (FIB7) and the transforming
gestedasinglecolonisationeventoftheislandsfollowedbydiver- growthfactorb2intron(TGF)sequences,alongwithdataonfrag-
sification within the archipelago, and revealed an unexpected ment length polymorphism of 26 microsatellite loci. In addition,
pattern:fourlineagesofmockingbirdsintheGalápagoswereiden- weincludemorphometricdatainthisstudytoinvestigateifGeno-
tifiedthatcouldbeconsidereddistinctatthespecieslevel,butthese vesamockingbirdsarephenotypicallydistinctfromtheirpotential
fourlineagesdidnotcorrespondtothefourphenotypicallydefined parental species. Thus, we use new, independent, and extensive
species (described above). Instead, the 7 examined mockingbird samplesfrommostGalápagosmockingbirdpopulationstorepeat
individualsfromGenovesaIsland(M.parvulusbauri)clusteredclo- the ND2 analysis reported by Arbogast et al. (2006) and confirm
selywiththe13birdsfromEspañola(M.macdonaldi)andthe2from their major findings. We provide entirely new data on CYTB and
SanCristóbal(M.melanotis),i.e.birdsphenotypicallyclassifiedas nucleargeneticvariation.Thecombineddata,togetherwithmor-
belongingtothreedistinctspecies.Fourdifferentscenariosexplain- phological analyses, allow us to tackle the reasons for Genovesa
ing the discrepancy between morphology-based taxonomy and mockingbird’s enigmatic position in mtDNA gene trees (Arbogast
mitochondrialgenealogywereputforwardbytheauthors.(i)The etal.,2006;Štefkaetal.,2011).Ourgeneticdataaremostcompat-
mitochondrial gene tree reflects the species tree, with extensive ible with the hypothesis (iv) of female-mediated introgression
morphologicalconvergenceexplainingthesimilarityofGenovesa from San Cristóbal, making the Genovesa mockingbird a popula-
mockingbirdswithM.parvulusfromotherislands,leadingtotheir tionofmixedancestry.
designation as a subspecies of M. parvulus. (ii) The plumage of
M. parvulus represents the ancestral state, while M. macdonaldi
andM.melanotisderivedtheircharacteristicplumageonlyrecently. 2.Materialsandmethods
Under this scenario, Genovesa mockingbirds are identified as
M.parvulusduetoplumagesymplesiomorphies,whiletheyactually 2.1.Samples
aremorecloselyrelatedtoM.macdonaldiandM.melanotisthanto
otherpopulationsofM.parvulus.Underbothscenarios(i)and(ii), We(PEAH,LFK,PGP,PRGandBRG)capturedthebirdsforthis
the mitochondrial gene tree reflects the species tree and hence, studywithmistnetsorpottertraps,andtookbloodsamplesfrom
mitochondrial as well as nuclear genes should reveal the same individuals collected on 15 islands between 2004 and 2008
clades.Alternatively,twoscenariosofdifferentialintrogressionat (Fig.1a).Samplesizesfromthedifferentlocationsvariedbetween
nuclearandmitochondriallocicouldproducepatternswherethe mtDNAandnuclearsequencinganalyses(101individualsintotal,
mitochondrial gene tree does not reflect the species phylogeny. comprisingasubsetoftheindividualsusedformicrosatelliteanal-
(iii)GenovesamockingbirdsaremostcloselyrelatedtoM.melanotis ysis; Table 1), microsatellite analysis (543 samples from Hoeck
andM.macdonaldi,butnuclearlociaffectingmorphologyhaveen- et al. (2010), plus 4 additional ones from Santa Cruz; Table 2),
tered the Genovesapopulationvia male-mediated dispersalfrom and morphological analysis (801 individuals in total; Table 2).
other M. parvulus populations without a corresponding influx of Wecollectedbloodsamplesonfilterpaperafterasmallpuncture
maternal mtDNA. Under this scenario, mtDNA and most nuclear of the wing vein of live birds. We extracted genomic DNA using
genes should show a grouping of Genovesa mockingbirds with the QIAamp DNA Mini Kit (Qiagen; Hoeck et al., 2009) and
M.melanotisandM.macdonaldi,whilesomenucleargeneswould standardised concentrations of DNA extracts at 20ng/lL
584 P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592
Table1
Samplesizesanddiversitystatistics,aswellasGenBankaccessionnumbersforthetwomitochondrialgenes(ND2=NADHdehydrogenasesubunit2;CYTB=Cytochromeb)and
thetwonuclearintrons(followingpage;FIB7=Beta-fibrinogengene;TGF=Transforminggrowthfactorb2)analysedinGalápagosmockingbirds.
Island Locus N #ht hd±SD p(MPD)±SD pn±SD GenBank
Champion ND2 10 1 0±0 0±0 0±0 KF411070
Española ND2 6 2 0.533±0.172 0.5333±0.5077 0.00053±0.00058 KF411074,75
Fernandina ND2 1 1 0±0 0±0 0±0 KF411078
Gardner-by-Española ND2 6 1 0±0 0±0 0±0 KF411075
Gardner-by-Floreana ND2 10 1 0±0 0±0 0±0 KF411070
Genovesa ND2 10 1 0±0 0±0 0±0 KF411079
Isabela ND2 6 4 0.800±0.172 3.6667±2.1551 0.00365±0.00248 KF411078,80–82
Marchena ND2 6 2 0.333±0.215 0.3333±0.3801 0.00033±0.00044 KF411083,84
Pinta ND2 5 2 0.600±0.175 2.4000±1.5567 0.00239±0.00181 KF411085,86
Rábida ND2 2 2 1.000±0.500 8.0000±6.0000 0.00797±0.00845 KF411083,87
SanCristóbal ND2 7 2 0.286±0.196 0.2857±0.3409 0.00029±0.00039 KF411071,72
SantaCruz ND2 7 1 0±0 0±0 0±0 KF411073
SantaFé ND2 7 2 0.286±0.196 0.2857±0.3409 0.00029±0.00039 KF411076,77
Santiago ND2 9 6 0.917±0.073 2.3333±1.4034 0.00232±0.00159 KF411088-93
Wolf ND2 3 1 0±0 0±0 0±0 KF411078
Total ND2 95 24 0.918±0.015 39.5306±17.3232 0.03937±0.01911 KF411070-93
Champion CYTB 7 1 0±0 0±0 0±0 KF411094
Española CYTB 6 1 0±0 0±0 0±0 KF411096
Fernandina CYTB 6 3 0.733±0.155 0.8667±0.7008 0.00152±0.00142 KF411099-101
Gardner-by-Española CYTB 5 1 0±0 0±0 0±0 KF411096
Gardner-by-Floreana CYTB 6 1 0±0 0±0 0±0 KF411094
Genovesa CYTB 5 1 0±0 0±0 0±0 KF411097
Isabela CYTB 5 2 0.600±0.175 0.6000±0.5622 0.00105±0.00115 KF411102,03
Marchena CYTB 4 1 0±0 0±0 0±0 KF411104
Pinta CYTB 5 2 0.400±0.237 0.4000±0.4351 0.00070±0.00089 KF411105,06
Rábida CYTB 5 2 0.600±0.175 3.0000±1.8741 0.00525±0.00384 KF411107,08
SanCristóbal CYTB 5 1 0±0 0±0 0±0 KF411095
SantaFé CYTB 7 1 0±0 0±0 0±0 KF411098
Total CYTB 66 15 0.908±0.017 19.3161±8.6588 0.03383±0.01681 KF411094-108
Champion FIB7 22 2 0.091±0.081 0.0909±0.1682 0.00011±0.00023 KF411109,14
Española FIB7 16 2 0.325±0.125 0.6500±0.5305 0.00079±0.00072 KF411113,16
Fernandina FIB7 4 1 0±0 0±0 0±0 KF411112
Gardner-by-Española FIB7 14 2 0.264±0.136 0.5275±0.4674 0.00064±0.00064 KF411113,16
Gardner-by-Floreana FIB7 20 3 0.195±0.115 0.3000±0.3263 0.00036±0.00044 KF411109,18,19
Genovesa FIB7 16 2 0.325±0.125 2.2750±1.3174 0.00276±0.00179 KF411112,17
Isabela FIB7 12 2 0.530±0.076 0.5303±0.4735 0.00064±0.00065 KF411112,21
Marchena FIB7 12 1 0±0 0±0 0±0 KF411112
Pinta FIB7 10 1 0±0 0±0 0±0 KF411112
Rábida FIB7 4 3 0.833±0.222 2.1667±1.4988 0.00263±0.00217 KF411110,12,21
SanCristóbal FIB7 16 1 0±0 0±0 0±0 KF411111
SantaCruz FIB7 14 2 0.143±0.119 0.1429±0.2187 0.00017±0.00030 KF411112,15
SantaFé FIB7 12 1 0±0 0±0 0±0 KF411112
Santiago FIB7 24 2 0.083±0.075 0.0833±0.1602 0.00010±0.00022 KF411112,20
Wolf FIB7 6 1 0±0 0±0 0±0 KF411112
Total FIB7 202 11 0.685±0.027 3.0532±1.5963 0.00370±0.00214 KF411109-21
Champion TGF 6 1 0±0 0±0 0±0 KF411123
Española TGF 4 2 0.500±0.265 0.5000±0.5191 0.00090±0.00112 KF411124,25
Fernandina TGF 4 2 0.667±0.204 1.3333±1.0249 0.00241±0.00221 KF411128,31
Gardner-by-Española TGF 2 1 0±0 0±0 0±0 KF411125
Gardner-by-Floreana TGF 6 1 0±0 0±0 0±0 KF411129
Genovesa TGF 6 3 0.733±0.155 1.1333±0.8472 0.00205±0.00177 KF411122,27,28
Isabela TGF 8 3 0.714±0.123 1.0000±0.7481 0.00181±0.00154 KF411122,28,31
Marchena TGF 12 2 0.409±0.133 0.4091±0.4031 0.00074±0.00082 KF411122,28
Pinta TGF 4 2 0.500±0.265 0.5000±0.5191 0.00090±0.00112 KF411122,28
Rábida TGF 2 2 1.000±0.500 2.0000±1.7321 0.00361±0.00442 KF411128,31
SanCristóbal TGF 14 1 0±0 0±0 0±0 KF411126
SantaCruz TGF 10 3 0.511±0.164 0.5556±0.4943 0.00100±0.00101 KF411122,28,30
SantaFé TGF 8 1 0±0 0±0 0±0 KF411122
Santiago TGF 6 2 0.600±0.129 0.6000±0.5477 0.00108±0.00114 KF411122,28
Wolf TGF 6 2 0.533±0.172 0.5333±0.5077 0.00096±0.00106 KF411122,31
Total TGF 98 6 0.714±0.026 1.0940±0.7244 0.00198±0.00145 KF411122-31
N=number of haplotypes (i.e. the number of birds for mitochondrial data, or twice the number of birds for nuclear data); #ht=number of distinct haplotypes;
hd±SD=haplotype(gene)diversity±standarddeviation;p(MPD)±SD=meanpairwisesequencedifference±standarddeviation;pn±SD=nucleotidediversity±standard
deviation;GenBank=GenBankaccessionnumber(s)ofsequencesfromeachisland;notethatthefollowingpairsortripletsofnuclearsequencesareidentical,buthave
different GenBank accession numbers because they occur in more than one species: KF411109=KF411110, KF411116=KF411117, KF411122=KF411123,
KF411125=KF411126=KF411127,KF411128=KF411129.
P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592 585
Table2 2.2.2.Sequenceanalysis
Samplesizesformicrosatelliteanalysisandmorphologicalmeasurements,andisland Basecallingandvisualinspectionoftracefileswasdonewith
abbreviationsusedinFig.3.
Sequencing Analysis Software v5.1 (Applied Biosystems). We
Island Microsatellites Morphology Abbreviation alignedmitochondrialandnuclearsequencesmanuallyinBioEdit
Champion 48 74 Cham (Hall,1999)andtrimmedsequencestoconstantlength(Table3).
Española 87 182(PGP:94) Esp We used FinchTV v1.4.0 (Geospiza Research Team, http://
Fernandina 24 www.geospiza.com/Products/finchtv.shtml) for visual control of
Gardner-by-Española 10 10 G.Esp
polymorphicsitesandambiguousbases.
Gardner-by-Floreana 69 118 G.Flo
BothnuclearsequencesFIB7andTGFrevealedseveralheterozy-
Genovesa 37 103(PGP:103) Geno
Isabela 62 88(PGP:56) Isab goussites.TheirphaseswerereconstructedusingPhasev2.1(Ste-
Marchena 38 39 March phensetal.,2001;StephensandDonnelly,2003)intheDnaSPv5
Pinta 27 package (Librado and Rozas, 2009). Phase reconstruction was
Rábida 21 21 Ráb
unambiguous, withphase probabilities of at least0.98 at allhet-
SanCristóbal 37 37 S.Cris
SantaCruz 39 82(PGP:20) S.Cruz erozygous sites. Finally, we detected identical sequences using
SantaFé 21 20 S.Fé CleanCollapse v1.0.5 (Ray, 2006). Arlequin v3.5 (Excoffier and
Santiago 27 27 Santi Lischer,2010)wasusedtocalculatethediversitystatisticsshown
Total 547 801
inTable1.
PGPindicatesthesizeofthesubsetofmorphologicalsamplesmeasuredbyP.G.
Parker’sgroup.TheothermeasurementsarefromthegroupofP.E.A.HoeckandL.F.
Keller. 2.2.3.Phylogeneticreconstruction
Onesampleofeachhaplotypewasusedforphylogeneticanal-
ysis. We reconstructed two separate phylogenies for each of the
(Quant-iTPicoGreendsDNAQuantitation,Invitrogen).Allsamples twomitochondrialgenes,withgenespartitionedaccordingtoco-
arestoredinfreezersattheUniversityofZurich,Switzerlandand don position. Stationarity of base frequencies among sequences
stillinuseforongoingstudies. was assessed with a v2-test using PAUP* v4.0b10 (Swofford,
2002). None of the partitions showed differing base frequencies
amongsamplesandexaminedgenes(allp>0.9),norhadanyco-
don position reached substitution saturation as examined using
2.2.Mitochondrialandnucleargeneanalysis thetestbyXiaetal.(I wasalwayssignificantlysmallerthanI
SS SS.C
for symmetrical and asymmetrical trees, p<0.00005) (Xia et al.,
2.2.1.PCRamplificationandsequencing 2003; Xia, 2009), as implemented in the software DAMBE v5.2.9
Polymerase chain reactions (PCR) were conducted in 10lL (Xia and Xie, 2001). We applied MrModeltest v2.3 (Nylander,
reactionvolumeandthefollowingconcentrations:0.25lMofeach 2004)todeterminethebestmodelofnucleotideevolutionbased
primer (Table 3), 0.1U GoTaq polymerase (Promega), 0.2mM of ontheAkaikeInformationCriterion(Akaike,1974)(selectedmod-
each dNTP, GoTaq reaction buffer, and 20ng of template DNA. els:ND2position1:HKY+I,ND2position2:HKY+I,ND2position
The PCR profile included 2min preheating at 94(cid:3)C followed by 3:HKY+G,CYTBposition1:GTR+I,CYTBposition2:K80,CYTBpo-
35cyclesof30sdenaturationat94(cid:3)C,30sannealing(foranneal- sition3:HKY).MrBayesv3.1.2(HuelsenbeckandRonquist,2001;
ingtemperaturesseeTable3)and45sextensionat72(cid:3)Candafi- RonquistandHuelsenbeck,2003)wasusedtogenerateaMetrop-
nalextensionat72(cid:3)Cfor10min.PCRwasdoneinaGeneAmpPCR olis coupled Markov chain Monte Carlo search and to determine
System9700oraVeriti96WellThermalCycler(bothAppliedBio- Bayesian posterior probabilities under the best-fit model (parti-
systems). We checked for PCR amplification on a 1% agarose gel tioned by codon position) using Mimus longicaudatus as an out-
and subsequently cleaned successful amplifications from non- group for ND2 (GenBank accession number EF468200; Lovette
incorporatedprimersbyapplyingastandardExoSAPprotocol.Fi- and Rubenstein, 2007) and Toxostoma redivivum as an outgroup
nally,wedidforwardandreversesequencingona3730DNAAna- for CYTB (GenBank accession number AY124543; Barhoum and
lyser(AppliedBiosystems)usingBigDyeTerminator(v3.1;Applied Burns, 2002), which were close relatives of Galápagos mocking-
Biosystems) chemistry, Better Buffer (Web Scientific) and the birds for which sequences were available on GenBank (Lovette
primers from Table 3 for mitochondrial sequences or universal et al., 2012). Searches of 8,000,000 generations with 50% burn-in
M13primers(Schuelke,2000)fornuclearsequences. forND2and3,000,000generationswith25%burn-inforCYTBwith
Table3
mtDNAandnuclearlocisequencedinthisstudy.Theirprimersequences,annealingtemperature(TA),analysedsequencelength(inbasepairs)andforwardandreverseprimer
sequencesareshown.
Locus Primersequence50-30 TA Length
mtDNA:NADHdehydrogenasesubunit2gene(ND2) L52151:TATCGGGCCCATACCCCGAAAAT 55 1004
H10642:CTTTGAAGGCCTTCGGTTTA
mtDNA:Cytochromeb(CYTB) L14996_FW23:AAYATYTCWGYHTGATGAAAYTTYGG 46a 571
H160643:CCTCANTYTTTGGYTTACAAGRCC
Nuclear:Beta-fibrinogengeneintron7(FIB7) FIB-BI7U4:tgtaaaacgacggccagtGGAGAAAACAGGACAATGACAATTCAC 55 825
FIB-BI7L4:caggaaacagctatgaccTCCCCAGTAGTATCTGCCATTAGGGTT
Nuclear:Transforminggrowthfactorb2intron(TGF) TGFB2-5F5:tgtaaaacgacggccagtGAAGCGTGCTCTAGATGCTG 66 554
TGFB2-5F5:caggaaacagctatgaccAGGCAGCAATTATCCTGCAC
Universalprimersequences(Schuelke,2000)addedtothelocus-specificsequencesareprintedinlowercaseletters.
Superscriptnumbersafterprimernamesrefertothesereferences:1J.CracraftandK.Helm-BychowskiinHackett(1996);2Drovetskietal.(2004);3Sorenson(2003);
4PrychitkoandMoore(1997);5Primmeretal.(2002).
TA=annealingtemperaturein(cid:3)C;Length=analysedsequencelengthinbasepairs.
a WeusedtheQiagenMultiplexMasterMix,0.25lMofeachprimer,and40ngtemplateDNAatcyclingconditionsaccordingtomanufacturer’sinstructions.
586 P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592
samples every 100 generations were performed. Runs converged group(forsamplesizessee Table2). Due todifferingtechniques,
later for ND2 than CYTB and hence, number of generations and measurements from the two groups differed systematically. We
burn-indifferforthetwogenestoensurethatonlyconvergedruns thususedmorphologicaldataofbirdsfromEspañola,SantaCruz,
weresampled,leadingtoeffectivesamplesizeswellabove200for and Isabela Islands that were visited by both groups in order to
allparameters.TreeswerevisualisedwithFigTreev1.2(Rambaut, correct for systematic measuring differences between observers.
2009). Tothisend,weperformedlinearregressionanalysisofthemean
We performed molecular dating only for the longer and geo- traitvaluesperisland(separatelyformeanwing,beak,andtarsus
graphicallymorecompletelyrepresentedND2sequencedata.We measurements)asmeasuredbytheHoeck/Kellergroupagainstthe
appliedarelaxedclockmodelwithestimateduncorrelatedlognor- correspondingmeasurementsoftheParkergroup.Then,measure-
malratesinBEASTv1.5.4(DrummondandRambaut,2007)using ments of the Parker group were modified by multiplying them
inputfilesgeneratedwithBEAUtiv1.5.4(DrummondandRambaut, withtheslopeoftheregressionline(0.66forbeak,0.93forwing,
2007). We simulated a birth–death speciation process starting and 0.64 for tarsus measurements, respectively) and adding the
froma randomtree and unlinkedtheHKY+G+I modelparame- y-axisintercept(1.84forbeak,1.06forwing,and10.49fortarsus
ters between codon positions. The gamma distribution was measurements, respectively). The high R2 values (0.97 for beak,
approximated with 6 categories. We applied a lognormal prior 0.97forwing,and0.85fortarsusmeasurements,respectively)of
withameanof1,500,000yearsandstandarddeviation(logtrans- the regression lines indicate that systematic differences in mea-
formed) of 0.5 on the height of the root (95% of values within surementscanbeefficientlycorrectedinthisway.Forfurtheranal-
581,600–3,013,000years).Thispriorcoverstheestimatedagesof ysesweusedParker’scorrecteddataandtherawmeasurementsof
most islands currently above water in the Galápagos archipelago theHoeck/Kellergroup.
(Geist, 1996; D. Geist, 2005–2008, unpublished data; Poulakakis Totestformorphologicaldifferencesamongislandpopulations,
et al., 2012) and includes the likely age estimates from Arbogast we conducted a MANOVA on the standardised measurements
etal.(2006).Wedidnotimposeanyfurtherconstraintsonother (standardization through subtraction of mean and division with
node ages. We performed four independent runs of 10,000,000 standard deviation), using the function ‘‘manova’’ in R. To get an
generations each and checked for convergence of runs in Tracer idea of the similarities among island populations, we performed
v1.5.0(RambautandDrummond,2009).ThesoftwareLogCombin- a cluster analysis of the mean beak, wing, and tarsus measure-
erv1.5.4andTreeAnnotatorv1.5.4(bothDrummondandRambaut, ments per island, using Euclidean distances and the average dis-
2007)wereusedtocombinetheparameterestimatesandtreesof tance among clusters, using the R function ‘‘hclust’’. To further
thefourrunsafterexclusionof10%burn-inatthestartofeachrun, investigatehowgeneticdifferencesamongislandpopulationsare
leadingtoeffectivesamplesizeswellabove200forallparameters. related to morphological traits, and to conduct this analysis on
AsthenuclearintronsFIB7andTGFshowedlowlevelsofpoly- uncorrelated linear combinations (correlations among standard-
morphism within the Galápagos mockingbirds and high levels of ised traits: r =0.73, r =0.75, r =0.54), we
wing-beak wing-tarsus beak-tarsus
haplotypesharingamongislands,weconstructedmedian-joining performedaprincipalcomponentsanalysisofwing,beak,andtar-
networks (Bandelt et al., 1999) with Network v4.5 (http:// sus length using the R function ‘‘prcomp’’. We tested the three
www.fluxus-engineering.com/sharenet.htm) and Network Pub- principal components for differences between islands with ANO-
lisher v1.1.0.7 (http://www.fluxus-engineering.com/nwpub.htm). VA.WedeterminedsignificantpairwisecomparisonsusingTukey’s
Allcharacterswereweightedequallyandanepsilonvalueofzero honestsignificantdifferencesinR.
waschosen.Epsilonspecifiesatoleranceuptowhichmutational
distances among haplotypes are considered equal during the
searchformedianvectors(Bandeltetal.,1999). 3.Results
2.3.Microsatelliteanalysis 3.1.MitochondrialandnuclearDNAsequences
We used microsatellite genotypic data for 26 loci from Hoeck We sequenced four loci to investigate the genetic diversity of
et al. (2010) and Hoeck and Keller (2012). To visualise patterns Galápagos mockingbirds. The two mitochondrial genes revealed
ofvariationinmicrosatelliterepeatlength,weconductedafacto- high nucleotide diversity over all islands (p =0.04 for ND2;
n
rial correspondence analysis using Genetix v4.05.2 (Belkhir, p =0.03 for CYTB) in comparison to the two nuclear introns
n
2004). Plots including the first three axes were drawn in R (p =0.004 for FIB7; p =0.002 for TGF; Table 1). The birds from
n n
v2.12.1 (R Development Core Team, 2010). Additionally, we ana- Genovesa ranked among the most variable at the nuclear se-
lysedpopulationstructureusingtheadmixturemodelwithuncor- quences(Table1),althoughtheyonlyshowedasinglemitochon-
related allele frequencies in Structure v2.3.3 (Pritchard et al., drialhaplotype.
2000). We sampled 600,000 repetitions after excluding the first InordertoinvestigatethegeneticrelatednessamongGalápagos
100,000 as burn-in. Runs were conducted for cluster numbers K mockingbirds,weconstructedphylogenetictreesofthemitochon-
ranging from 2 to 14 (i.e. the number of sampled island popula- drial sequences and median-joining networks of the nuclear in-
tions) and we determined the numberof clusters best fitting the trons. Furthermore, we estimated node ages for mitochondrial
databasedonDK(Evannoetal.,2005)usingStructureHarvester ND2.ThetreesderivedfromCYTBandND2weretopologicallycon-
(Earl and vonHoldt, 2012). Structure output plots were coloured sistent;wethereforeonlyshowthetreebasedonthemorecom-
andorderedinDistruct(Rosenberg,2004). plete and variable ND2 dataset (Fig. 2). We found four well
differentiated and statistically supported clades. The oldest split
2.4.Morphologicalanalysis withinGalápagosmockingbirdsoccurredca.500,000(95%credible
interval: 145,957–1,388,173) years ago and separated the mito-
Wemeasuredbeakandtarsuslengthsofallcapturedbirdswith chondrial haplotypes of birds from San Cristóbal, Genovesa, and
digitalcalipers(tothenearest0.1mm),andwinglengthfromthe EspañolatogetherwithGardner-by-Españolafromallotherislands
carpaljointtothetipoftheunflattenedlongestprimaryfeatherto (node1inFig.2).Mockingbirdsontheformerislandsareconsid-
thenearest1mmusingaruler(accordingtoSvensson,1992).Most ered to belong to three distinct species (Mimus melanotis on San
birdsweremeasuredbyP.E.A.HoeckandL.F.Keller.Measurements Cristóbal, M. macdonaldi on Española, and M. parvulus bauri on
of Genovesa mockingbirds were only available from P.G. Parker’s Genovesa;Swarth,1931;Remsenetal.,2012),buttheirmitochon-
P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592 587
Fig.2. Mitochondrialphylogeny.PhylogenetictreederivedfrommitochondrialND2sequences(consistentwithphylogenyderivedfromasmallerdatasetofCYTBsequences;
datanotshown).ShownabovebranchesareBayesianposteriorprobabilitiesandbelowbranchesinitalicsthemodeand95%highestposteriordensityintervalsofsplitdate
estimatesasderivedfromarelaxedclockmodel.Numbersinbracketsbehindislandnamesrefertothenumberofbirdswiththerespectivehaplotype.Notethatwedetected
onlyonehaplotypeinbirdsfromGenovesawhichclusteredwithhaplotypesofbirdsfromSanCristóbalandEspañola.
drial haplotypes only split around 30,000 (5797–124,036) years 3.2.Autosomalmicrosatelliteloci
ago (node 6). A second clade was formed by the 380,000
(110,766–1,186,787)yearoldsplit(node2)oftheFloreanamock- Toobtainamorerepresentativepictureoftheautosomalgen-
ingbird(M.trifasciatus)fromallM.parvuluspopulationsexceptthe ome,weinvestigatedvariationatadditionalnuclearloci:26auto-
one on Genovesa Island. Another deep split (node 3) occurred somalmicrosatellites.Afactorialcorrespondenceanalysis(Fig.1d)
around 250,000 (70,856–831,935) years ago within M. parvulus, revealedfourgroupsthatcorrespondtothefourcurrentlyrecogni-
separating the mitochondria of the birds on the western islands sedspecies(Swarth,1931;Remsenetal.,2012).Thegenotypesof
of Isabela, Fernandina, and Wolf, from the mitochondria of the Genovesa mockingbirds grouped with the other M. parvulus on
birdsonthecentralandnorthernislandsofSantaCruz,Santiago, axes1and2.Onthethirdaxis,genotypesofGenovesamocking-
Marchena, Pinta, Santa Fé, and Rábida. Note that the modal esti- birds were more divergent, but still grouped most closely with
mate of the age of this intraspecific split (node 3 in Fig. 2) was the other M. parvulus (Fig. 1d). Hence, the microsatellite data
6.3timesolder(calculatedfromtheMCMCsamples;95%credible showed that at these 26 autosomal markers, Genovesa mocking-
interval:2.6–17.2timesolder)thantheageoftheseparationofthe birdsaresimilartotheotherM.parvuluspopulations.Thisfinding
mitochondrialhaplotypesofthethreespeciesM.melanotisonSan isconsistentwiththeStructureanalysisforthebest-fittingnum-
Cristóbal, M. macdonaldi on Española, and M. parvulus bauri on berofclusters(i.e.K=4,SupplementaryFig.S1),whereGenovesa
Genovesa (node 6). The age of the first split within Galápagos mockingbirds also clustered with most M. parvulus populations
mockingbirds (node 1) was 11.9 (4.8–29) times older than the (SupplementaryFig.S2).Contrarytotheresultsofthefactorialcor-
ageofthemostrecentsplitatnode6. respondence analysis (Fig. 1d), the Structure analysis for K=4
Median-joining networks based on both nuclear introns, FIB7 groupedM.melanotis(SanCristóbal)andM.macdonaldi(Española
(Fig.1b)and TGF(Fig.1c),showedmuchlessresolutionthanthe andGardner-by-Española)inthesamecluster,whiletheM.parvu-
mitochondrial sequences due to their lower variation (Table 1). luspopulationsonIsabelaandFernandinaformedanowncluster
However, they did not contradict the mitochondrial phylogenies, (SupplementaryFig.S2).
except for the placement of Genovesa mockingbirds: The birds
from Genovesa showed two and three divergent haplotypes for 3.3.Morphologicalanalysis
FIB7andTGF,respectively.OneoftheFIB7haplotypesfromGeno-
vesamockingbirdswassharedwithbirdsonEspañolaIslandand Measurementsofwing,beak,andtarsuslengthdifferedamong
one mutational step away from birds on San Cristóbal Island island populations for each trait separately (ANOVA; wing,
(Fig. 1b). Similarly, one of the TGF haplotypes from Genovesa F =81.9, p<0.001; beak, F =214.0, p<0.001; tarsus,
11,789 11,789
mockingbirdswassharedwithbirdsofEspañolaandSanCristóbal F =100.4, p<0.001), as well as combined (MANOVA,
11,789
Islands (Fig. 1c). This pattern was consistent with the mitochon- F =71.0, p<0.001). In the cluster analysis (Fig. 3), Española
33,2367
drial data, where haplotypes of birds from these three islands mockingbirds(M.macdonaldi)andFloreanamockingbirds(M.tri-
groupedinthesameclade,with1–2substitutionsinND2between fasciatus) grouped separately from the other populations. San
birds from Genovesa and San Cristóbal Islands and 5–6 substitu- Cristóbal mockingbirds (M. melanotis) grouped together with the
tions between birds from the latter two islands and Española Is- birdsfromSantiagoamongM.parvuluspopulations.Interestingly,
land. Contrary to the mitochondrial data, the second FIB7 Genovesamockingbirds(M.parvulusbauri)didnotclusterclosely
haplotypefromGenovesamockingbirdswassharedwithM.parv- withanyotherpopulation.Wealsoconductedaprincipalcompo-
ulus(Fig.1b).Similarly,thesecondandthirdTGFhaplotypeswere nentanalysisofmorphologicalmeasurements(Fig.4,Supplemen-
shared with M. parvulus and M. trifasciatus (Fig. 1c). These mito- tary Figs. S3 and S4). Principal component 1 (representing size
chondrialandnuclearsequencedataindicateamixedgeneticcom- differences, as all variables load in the same direction) explained
positionofGenovesamockingbirds. 78% of the variation and separated M. macdonaldi (Española and
588 P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592
Santiago M. parvulus
San Cristóbal M. melanotis
Rábida
Santa Cruz M. parvulus
Isabela
Genovesa M. parvulus bauri
Marchena
M. parvulus
Santa Fé
Gardner-by-Española
M. macdonaldi
Española
Gardner-by-Floreana
M. trifasciatus
Champion
2 1 0
Euclidean distance
Fig.3. Morphologicaldendrogram.DendrogramconstructedwithEuclideandistancesofmeanmorphologicalmeasurements(wing,beak,andtarsuslength)perisland
populationandaveragedistancesamongclusters.
PC1= −0.61*Wing−0.55*Beak−0.56*Tarsus
%)
8
7
1 (
nt
e
n
o
p
m
o
C
al
p
ci
n
Pri
Fig.4. Morphologicalmeasurements.Boxplotsshowingthefirstprincipalcomponentofmorphologicalmeasurements,representingsizedifferences.Notethatlargervalues
representsmallerbirds.Thelettersbelowtheabbreviatedislandnames(seeTable2fortranslation)representTukey’shonestsignificantdifferences;pairsofislandsthatare
labelledwithatleastoneidenticalletterdonotdiffersignificantlyintheirrespectiveprinciplecomponent;thosethatdonotshareatleastoneletteraresignificantly
different.78%ofthevariationisexplainedbythisprincipalcomponent.Forprincipalcomponents2and3seeSupplementaryFigs.S3andS4.
Gardner-by-Española)andM.trifasciatus(ChampionandGardner- liteloci,andmorphologicalmeasurements.Thesedatashowthat
by-Floreana)fromallotherpopulations.Themeanprincipalcom- Genovesamockingbirds(M.parvulusbauri)areofmixedancestry
ponent 1 of the birds from Genovesa Island differed significantly with mitochondria from M. melanotis and a nuclear genome lar-
from all populations of M. parvulus, which were not significantly gely,butnotentirely,fromM.parvulus.Theyarealsomorpholog-
different among themselves. The birds from San Cristóbal were icallydistinct.
on average intermediate between, but not significantly different
from,thebirdsfromGenovesaIslandandmostotherM.parvulus.
4.1.Geneticdatarevealancientintrogression
Taken together, wing, beak, and tarsus measurements separated
M.trifasciatusandM.macdonaldifromtheotherGalápagosmock-
Phylogenies derived from the mitochondrial genes ND2 and
ingbirds and, to a lesser extent, from each other. Among the
CYTB revealed two important patterns. (a) The mitochondria of
remainingtaxa,theGenovesamockingbirdsweremoreclearlydif-
themockingbirdsfromGenovesa(M.parvulusbauri),SanCristóbal
ferentfromotherM.parvuluspopulationsthanM.melanotisfrom
(M.melanotis),andEspañola(M.macdonaldi)sharearecentcom-
M.parvulus.
monancestor,withmitochondriaofthemockingbirdsfromGeno-
vesabeingmostsimilartothoseofbirdsfromSanCristóbal.Two
4.Discussion previousstudieshavedescribedthissurprisinggroupingofmito-
chondria from three species (Arbogast et al., 2006; Štefka et al.,
WeinvestigatedtheancestryofGenovesamockingbirdsusinga 2011).Itshowsthatthemockingbirdsonthesethreeislandsmust
combination of sequence data from two mitochondrial and two have exchanged genes relatively recently, until between 124,036
nucleargenes,lengthpolymorphismsat26autosomalmicrosatel- and 5797years ago (modal estimate of 28,027years ago; node 6
P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592 589
inFig.2).Otherwisetheycouldnothavesharedacommonances- (modal estimate of 28,027years ago; node 6 in Fig. 2) at least
toratthattime(butgeneflowmayhavepersisteduntilmorere- onefemale(ormore,andpossibly,butnotnecessarily,alsosome
cently; Nichols, 2001). (b) The mitochondrial haplotypes of the males) from San Cristóbal dispersed to Genovesa. The immigrant
birds on Genovesa, San Cristóbal, and Española separated from female(s)introducedtheirmitochondrialhaplotypesintothepop-
those of other M. parvulus populations very long ago (145,957– ulationonGenovesa,wherethehaplotypeobservedtodayeventu-
1,388,173yearsagowithamodalestimateofnearly500,000years ally became fixed. In addition to their mitochondrial haplotypes,
ago;node1inFig.2).Thissplit(node1inFig.2)occurred4.8–29 these birds also introduced distinct nuclear haplotypes into the
times(modalestimateof11.9times)earlierthanthesplitamong population on Genovesa, but none of them reached fixation at
the mtDNA of the three species on Genovesa, San Cristóbal, and the examined loci. Mitochondrial haplotypes are more strongly
Española. In this study we investigated if the close clustering of influencedbygeneticdriftthanautosomalgenesduetotheirfour
Genovesamockingbirdswithotherspeciesandthelongseparation timessmallereffectivepopulationsize,andarethereforeexpected
of Genovesa mockingbirds from conspecific populations was a toreachfixationfasterthanautosomalpolymorphisms(Takahata
peculiarity of the mitochondrial DNA. To that end we analysed andSlatkin,1984;FunkandOmland,2003).Thescenariooutlined
the sequences of two nuclear introns and the length polymor- above is consistent with our microsatellite data, which show a
phismsat26microsatelliteloci.Wedetectedtwodivergenthaplo- closerelationshipbetweenthebirdsfromGenovesaandotherM.
typesforFIB7andthreedivergenthaplotypesforTGFinbirdsfrom parvuluspopulations.Themostparsimoniousexplanationforthis
Genovesa(M.parvulusbauri).Onehaplotypeofeachlocusgrouped patternisthatitwasthematernallyinheritedmitochondrialhap-
withhaplotypesofM.melanotis(SanCristóbal)andM.macdonaldi lotypethatintrogressedintothepopulationandreplacedtheres-
(Española and Gardner-by-Española), while the other haplotypes identhaplotype.
grouped with M. parvulus haplotypes (Fig. 1b and c). Autosomal Introgressionofgenestransmittedbyfemalesmaygenerallybe
microsatellitedata,ontheotherhand,groupedthemockingbirds more likely, as females are usually the dispersing sex in birds
on Genovesa together with other M. parvulus populations, while (GreenwoodandHarvey,1982),includingGenovesamockingbirds
the remaining three species (M. melanotis, M. macdonaldi, and M. (CurryandGrant,1989).However,itisunknownifthesepatterns
trifasciatus)formedthreedistinctgroupsinthefactorialcorrespon- offemale-biaseddispersalwithinislands(CurryandGrant,1989)
dence analysis (Fig. 1d), or two clusters in the Structure analysis alsoholdforlong-distancedispersalbetweenislands.Chancesfor
(SupplementaryFig.S2).Thisresultisconcordantwithaprevious thecompletereplacementofM.parvulusmtDNAwithmtDNAfrom
analysis of our samples at 16 microsatellite loci (Hoeck et al., M.melanotiswouldbemuchincreasedbypopulationbottlenecks.
2010),withtheexceptionthatthetenadditionallociincludedhere Interestingly,declinesingeneticdiversityofthegreywarbler-finch
allowedtoseparateM.melanotisfromM.macdonaldiinthefacto- (Certhidea fusca) population on Genovesa may be explained by
rialcorrespondenceanalysis,butnotintheStructureanalysis.To periodiccyclesofpopulationincreasesduringwetElNiñoevents
sumup,themicrosatellitesandthemajority,butnotall,ofthenu- followedbybottlenecksduringaridtimes(FarringtonandPetren,
clearhaplotypesgroupGenovesamockingbirdswithotherM.parv- 2011).Similarly,thesharp-beakedground-finch(Geospizadifficilis)
uluspopulations. populationonGenovesaseemstohaveundergoneageneticbottle-
This genetic pattern provides a resolution of the evolutionary neckwhichaidedtheintrogressionofsmallground-finch(G.fuli-
history of Genovesa mockingbirds. In line with Hoeck et al. ginosa) genes (Grant and Grant, 2008b). Thus, it is plausible that
(2010),twoofthepotentialexplanationsofArbogastetal.(2006) bottlenecksalsoaffectedthemockingbirdsonGenovesa.Addition-
fortheunexpectedmitochondrialclusteringofGenovesamocking- ally,itispossiblethatselectiononmtDNAmayhavecontributedto
birdswithM.melanotisandM.macdonaldicanberejected:(i)plum- the spread of the introgressed mitochondrial lineage. Although
ageconvergenceofGenovesamockingbirdswithM.parvulus,and controversial(Karletal.,2012),selectiononmtDNAhasbeensug-
(ii) plumage symplesiomorphy of Genovesa mockingbirds and gested in various taxa (Bazin et al., 2006); (but see Berry, 2006;
M.parvulus.Bothscenarioswouldrequirethatallgeneticlocireveal Mulliganetal.,2006;Albuetal.,2008;Nabholzetal.,2008)includ-
thesameclades(exceptthemarkerslinkedtolocicodingforplum- ing parasitic wasps (genus Nasonia; Oliveira et al., 2008), killer
agecharacteristics).Scenario(iii),whichimpliesdifferentialintro- whales (Orcinus orca; Foote et al., 2011), and humans (Mishmar
gressionofplumage-determininggenesfromM.parvulusintothe etal.,2003),aswellasinsimulationstudies(Bonnet,2012).
Genovesa mockingbird population (Arbogast et al., 2006), is also Analternativeexplanationtointrogressionfortheobservedge-
unlikely because the majority of the analysed autosomal genetic neticpatternscanbe(v)incompletelineagesorting(Nichols,2001;
variation, and not just a few loci, group Genovesa mockingbirds FunkandOmland,2003;Hoecketal.,2010).However,thisexpla-
clearlywithM.parvulus.Itisconceivable,albeitunlikely,thatstrong nationseemsveryunlikelyinourcase.Forincompletelineagesort-
geneflowfromM.parvulusintotheGenovesamockingbirdpopula- ingtoexplainourresults,atleastallpopulationsofM.parvulus,M.
tionmighthavereplacedmostofthenucleargenomeofGenovesa macdonaldi, and M. melanotis would have had to exchange genes
mockingbirds,butnottheirmitochondrialgenome.Suchascenario during almost 145,957–1,388,173years (modal estimate of
wouldrequireconsiderableintrogressionatmanyunlinkedautoso- approximately460,000years;fromnode1untilnode6inFig.2).
malloci,butnoneatthemitochondrion,andthusdoesnotappearto Startingbetween124,036and5797yearsago(modalestimateof
be parsimonious. Similar effects could come from an invasion of 28,027years ago; node 6 in Fig. 2), exchange of mitochondrial
M. parvulus onto Genovesa Island, followed by replacement of a genes among species would have stopped and complete lineage
residentM.melanotispopulation.Curratetal.(2008)showedthat sorting according to species affiliation would have occurred on
in such cases of range expansion, introgression is most likely to allislandsexceptonGenovesa.Hence,lineagesortingwouldhave
occurintotheinvadingspeciesandpreferentiallyinvolveslocilike hadtooccurinaperiodoftime(sincenode6inFig.2)thatwas
mtDNAthataresubjecttoreducedgeneflow(Birkyetal.,1989). 11.9 (4.8–29) times shorter than the long period of time (from
Suchascenarioofspeciesreplacementwithintrogressionissimilar node1tonode 6in Fig.2) duringwhichallmajorlineagesmust
toitsreversescenario(iv)(Arbogastetal.,2006),female-mediated havebeensegregatingwithoutlineagesorting.Althoughsomeof
introgressionofmitochondrialandsomenucleargenesfromM.mel- thecentralislandswerefusedduringpartsofthePleistocene,this
anotisintoGenovesamockingbirds. wasneverthecaseforallislandsinhabitednowbyM.parvulus,M.
Thisscenario(iv)wouldinvolvethefollowingevents:Genovesa macdonaldi, and M. melanotis (Geist, 1996; D. Geist, 2005–2008,
Island was colonised by a population of birds originating from a unpublished data; Poulakakis et al., 2012). Thus, it is difficult to
M. parvulus population. Between 124,036 and 5797years ago imaginesuchadrasticchangeinthebiologyofGalápagosmock-
590 P.Nietlisbachetal./MolecularPhylogeneticsandEvolution69(2013)581–592
ingbirds that would have changed their dispersal behaviour to theirgeneralistfeedingbehaviourmaybeamoreimportantexpla-
causeashiftfromapopulationgeneticallymixingacrossmostof nation(Arbogastetal.,2006;GrantandGrant,2008b,chapter11).
the archipelago to many well-differentiated island populations,
asfoundtoday(Hoecketal.,2010).Furthermore,itisunlikelythat 4.3.Morphologicalanalysis
therandomprocessoflineagesortingwouldoccurexactlyaccord-
ingtospeciesaffiliationinallcasesexceptinGenovesamocking- Ouranalysisofmorphologicalmeasurementsshowedthatthe
birds. Hence, incomplete lineage sorting is extremely unlikely to introgressed population on Genovesa is more distinct from M.
account for the observed genetic pattern in Galápagos parvulus,thanthebirdsfromSanCristóbal(M.melanotis)arefrom
mockingbirds. manyM.parvuluspopulations(Figs.3and4).Infact,M.melanotis
clusteredwithM.parvulusfromSantiago(Fig.3).Itmusthavebeen
this morphological similarity that led both Charles Darwin and
4.2.ColonisationhistoryofGalápagosmockingbirds JohnGouldtoconsiderthepopulationonSantiagotobethesame
speciesasthatonSanCristóbal(Gould,1837,1841;Darwin,1839).
Our mitochondrial phylogeny is consistent with a previously All M. parvulus populations (including Genovesa mockingbirds)
publishedmitochondrialphylogeny(Arbogastetal.,2006).Alater and M. melanotis are more similar to each other than to any of
studyinvestigatedthephylogeniesofGalápagosmockingbirdsand theother,morphologicallydistinctpopulationsofthetwospecies
twooftheirparasites(Štefkaetal.,2011).Theirmitochondrialphy- M. trifasciatus and M. macdonaldi. Abbott and Abbott (1978) per-
logenies of the host and both parasites were largely consistent formedacanonicalvariatesanalysisofbeak,tarsus,andwingmea-
withourphylogeny,withoneexception.Štefkaetal.(2011)found surements. Similar to our results, they found that M. trifasciatus
thattheFloreanamockingbirdwasnestedwithinM.parvulus,mak- andM.macdonaldiaremorphologicallyclearlydistinctfromeach
ing the latter species paraphyletic. However, this paraphyly had other and from all other Galápagos mockingbirds. M. melanotis
low support and was not recovered by either our analysis or by was morphologically similar to some M. parvulus populations,
Arbogastetal.(2006)orLovetteetal.(2012). and Genovesa mockingbirds were at the edge of the M. parvulus
The orderof splits in the presented phylogeny can be used to cluster.AbbottandAbbott(1978)alsocalculatedforeachpopula-
speculate about the colonisation history of mockingbirds in the tion a measure of morphological dissimilarity towards all other
Galápagos archipelago (Supplementary Fig. S5), as has similarly Galápagos mockingbird populations. This dissimilarity measure
beendonebyArbogastetal.(2006).M.melanotisfromSanCristó- was higher for the birds from Genovesa than for those from San
bal and M. macdonaldi from Española formed the basal lineage Cristóbal,butseveralotherpopulationsofM.parvulusalsoshowed
amongGalápagosmockingbirdsandfittinglyinhabitthetwoold- highdissimilaritymeasures(AbbottandAbbott,1978).Takento-
est currently exposed islands (Geist, 1996; D. Geist, 2005–2008, gether,theresultsofourmorphologicalanalysisagreedwiththat
unpublished data), suggesting that the first mockingbirds on the ofAbbottandAbbott(1978)andshowedthatthebirdsonGeno-
archipelago arrived on San Cristóbal or Española. The next split vesaarenotphenotypicallyintermediatebetweentheirsuggested
(node2inFig.2)separatedtheFloreanamockingbirds(M.trifasci- parentaltaxaM.parvulusandM.melanotis.However,introgression
atus)fromM.parvulus,makingitlikelythatFloreanawasthethird ofM.melanotisintoM.parvulusbaurimayhavenonethelesscon-
island to be colonised. From there, mockingbirds would have in- tributed to the phenotypic distinctiveness of Genovesa mocking-
vadedthecentralislands,probablyincludingGenovesa.Depending birds.Inadditiontogeneflow,environmentalconditionsmaybe
ontheexacttimingofthissplit,someofthecentralislandswere shaping mockingbird morphology, leading to small and short-
thenconnectedbyland(Poulakakisetal.,2012). Lateron, mock- beaked birds on low and arid islands and large and long-beaked
ingbirdsfromthecentralislandscrossedtotheyoungislandofIsa- birdsonthemorehumid,higherandmorediverseislands(Gould,
bela(node3inFig.2),fromwheresoonafterwardspopulationson 1837,1841;Swarth,1931;BowmanandCarter,1971;Abbottand
Wolf and on the youngest island of the archipelago, Fernandina, Abbott,1978;Curry,1989;CurryandGrant,1990).Thus,conver-
were established. Roughly at the same time, introgression from gent evolution may be part of the reason why mockingbirds on
SanCristóbalintothepopulationonGenovesaoccurred.Thiscolo- lowandaridGenovesaaremorphologicallydifferentfrompopula-
nisationhistorymatcheswellwiththeorderofemergenceofthe tionsofM.parvulusonhigherandlargerislands,approaching,but
Galápagosislands(Geist,1996;D.Geist,2005–2008,unpublished not reaching, the beak and body size of mockingbirds on arid
data; Poulakakis et al., 2012) and follows the prevailing pattern Española or Champion and Gardner-by-Floreana. Similarly, San
of mainly southerly, southeasterly, and easterly winds (Jackson, Cristóbal mockingbirds may be morphologically similar to many
1991;Arbogastetal.,2006). M. parvulus populations because they also live on a higher, more
The Galápagos archipelago is inhabited by one of the best- humidandmorediverseisland.
knownadaptiveradiations,theDarwin’sfinches(GrantandGrant,
2008b).Itisintriguingwhyonegroup,theDarwin’sfinches,radi- 4.4.Conclusions
ated extensively with multiple species occurring in sympatry,
while in another group on the same archipelago, the Galápagos Generally in line with previous studies (Arbogast et al., 2006;
mockingbirds, only four species occur and none of them live in Hoecket al.,2010), weshowed that atnuclear loci the mocking-
sympatry with another mockingbird species (Grant and Grant, birdsonGenovesaIsland(M.parvulusbauri)arelargelysimilarto
2008b).Possibleexplanationsforthisobservationincludedifferent otherM.parvuluspopulations,butcontainmitochondriaandsome
timessincecolonisationoftheislandsorsincethestartofradia- autosomallocithatgroupthemwithM.melanotisandM.macdon-
tion.Althoughtheageestimatesarenotveryprecise,ourdataesti- aldi.Themostparsimoniousexplanationforthisgeneticpatternis
mate the first split within Galápagos mockingbirds (not the introgressive hybridization.The directionof hybridization cannot
colonisation of the archipelago) to have occurred between be unequivocally resolved, but it is much more likely that intro-
145,957–1,388,173years ago (modal estimate of 488,078years gression from San Cristóbal into an M. parvulus population on
ago)yearsago.Thisestimateismorerecentthanthecurrentesti- Genovesagaverise tothisgeneticpattern,becausethis direction
mate (obtained with different methods than our estimates) of ofintrogressiononlyimpliesthereplacementofonelocuswithrel-
1,650,000years for the first split within Darwin’s finches (Petren ativelysmalleffectivepopulationsize,themitochondrialDNA.In
etal.,2005).Thus,themorerecentstartoftheGalápagosmocking- any case, the genetic composition of Genovesa mockingbirds is
birdradiationmayinpartexplaintheirlowerdiversity.However, theresultofancienthybridization.Becausehybridizationoccurred