Estimating Ancestral Ranges

12 Pages·2011·0.53 MB·English

Estimating Ancestral Ranges: Testing Methods with a
Clade of Neotropical Lizards (Iguania: Liolaemidae)
Juan Manuel Dı´az Go´mez*
Ca´tedradeDiversidadBiolo´gicaIV,FacultaddeCienciasNaturales,CONICET-IBIGEO,UniversidadNacionaldeSalta,Salta,Argentina
Abstract
Establishing the ancestral ranges of distribution of a monophyletic clade, called the ancestral area, is one of the central
objectivesofhistoricalbiogeography.Inthisstudy,Iusedthreecommonmethodologiestoestablishtheancestralareaof
an important clade of Neotropical lizards, the family Liolaemidae. The methods used were: Fitch optimization, Weighted
AncestralAreaAnalysisandDispersal-VicarianceAnalysis(DIVA).Amaindifferencefrompreviousstudiesisthattheareas
used in the analysis are defined based on actual distributions of the species of Liolaemidae, instead of areas defined
arbitrarilyor basedonothertaxa.Theancestralarea ofLiolaemidaefound byFitchoptimizationisPrepuna onArgentina,
CentralChileandCoastalPeru.WeightedAncestralAreaAnalysisfoundCentralChile,Coquimbo,Payunia,AustralPatagonia
and Coastal Peru. Dispersal-Vicariance analysis found an ancestral area that includes almost all the areas occupied by
Liolaemidae,exceptAtacama,CoquimboandAustralPatagonia.Theresultscanberesumedontwoopposinghypothesis:a
restrictedancestralareafortheancestorofLiolaemidaeinCentralChileandPatagonia,orawidespreadancestordistributed
alongtheAndes.Somelimitationsofthemethodswereidentified,forexampletheexcessiveimportanceofplesiomorphic
areas inthe cladograms.
Citation:D´ıazGo´mezJM(2011)EstimatingAncestralRanges:TestingMethodswithaCladeofNeotropicalLizards(Iguania:Liolaemidae).PLoSONE6(10):
e26412.doi:10.1371/journal.pone.0026412
Editor:BrianGratwicke,Smithsonian’sNationalZoologicalPark,UnitedStatesofAmerica
ReceivedJune1,2011;AcceptedSeptember26,2011;PublishedOctober20,2011
Copyright:(cid:2)2011JuanManuelD´ıazGo´mez.Thisisanopen-accessarticledistributedunderthetermsoftheCreativeCommonsAttributionLicense,which
permitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalauthorandsourcearecredited.
Funding: The workwas funded by theConsejoNacional de InvestigacionesCient´ıficas y Te´cnicas (CONICET -www.conicet.gob.ar),in form of a doctoral
fellowshiptotheauthor,andtheConsejodeInvestigacio´ndelaUniversidadNacionaldeSalta(CIUNSa),ProyectoNu1783.Thefundershadnoroleinstudy
design,datacollectionandanalysis,decisiontopublish,orpreparationofthemanuscript.
CompetingInterests:Theauthorhasdeclaredthatnocompetinginterestsexist.
*E-mail:[email protected]
Introduction different than the sum of the individual areas of the species. As
such, ancestral area analysis is not ad hoc or unscientific [3], but
Inferring the ancestral area of distribution for a clade of another way to make hypotheses to explain the distribution of
organismsisoneoftheclassicgoalsofhistoricalbiogeography[1], taxa.
andispartofthenaturalhistoryoftheorganisms.Instudiesthat The main procedure to study the biogeography of individual
trytoassesstherelativeimportanceofvicarianceanddispersalin groups is the ancestral area methodology. Ancestral area analysis
the distribution of a group of organisms and its speciation, an was proposed by Bremer [6] as a way to identify the area of
important subject is the reconstruction of the ancestral ranges of distribution of the ancestor of a monophyletic group, which he
distribution for thetaxaanalyzed [2]. termed ancestral area.
Historical biogeography deals with two kinds of problems, as Themainassumptionoftheancestralareaapproachisthatthe
pointedbyHovenkamp[3]:EarthhistoryandTaxonhistory.The ancestral area of a taxon can be inferred from the topological
firstapproachattemptstoestablisharearelationshipsbasedonthe informationinitsareacladogram[8],giventheassumptionsthat
phylogenies of at least two taxa inhabiting the areas of interest. (1)plesiomorphicareasinacladogramaremorelikelypartofthe
The taxon-history approach seeks to elucidate the biogeographic ancestral area than apomorphic areas; and (2) areas represented
history of particular taxa. The utility of the latter approach has onmorethanonebranchhaveahigherprobabilityofbeingpart
been criticized [4,5] because inferences are restricted to general oftheancestralareathanareaslessrepresented.Forancestralarea
patterns. However, as noted by Bremer [6], the search for the analysis,Iappliedthreemethods:Fitchoptimization[7],weighted
historical biogeography of individual groups is a valid procedure, ancestral area Analysis [8], and Dispersal-Vicariance analysis
andispartofthestudyofthenaturalhistoryoftheorganisms.In (DIVA) [9]. These methods use optimizations with reversible
many cases, the main assumption of vicariance biogeography, parsimony for estimating ancestral areas. Fitch optimization was
namelythattheancestralareaofataxonisidenticaltothepresent proposed by Ronquist [7] to avoid the problems of Camin-Sokal
distribution, maynot apply.For example,widespread (cosmopol- (irreversible) parsimony originally proposed by Bremer [6].
itan)groupsconsistingofmanytaxaofverylimiteddistributions.If Weighted ancestral area analysis uses Fitch parsimony with a
extant taxa are limited in their distribution, it does not seem weightingschemethatweightsfavorablyplesiomorphic,andmore
probablethatitscommonancestorwascosmopolitan[6].Another common areas. With this method, a probability index (PI) is
example is when all relatives of widespread taxa have limited calculatedtogiveameasureofthelikelihoodofaparticulararea
distributions,asisthecaseofthehumansandthegreatapes[7].In being part of the ancestral area. DIVA searches ancestral areas
these cases, it may seem logical to search for an ancestral area using a three-dimensional cost matrix that gives different costs to
PLoSONE | www.plosone.org 1 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
events, minimizingthe dispersal events needed for explaining the scenario assumes that the southern populations of Phymaturus
distributions.Usingthisapproach,vicarianceeventshavenocost, would have experienced more drastic climatic and vegetation
whereasdispersalsandextinctionshaveacostofoneperareaunit changesthanthenorthernpopulations,whichwouldhavecaused
addedtothedistribution.Theoptimalreconstruction(s)arethose extinctions of several of the original southern populations.
requiring theminimal number of dispersal events. Recently, Lobo & Quinteros [24] studied the historical biogeog-
Ancestral area methods have been criticized, mainly on the raphy of Phymaturus, assigning to their terminal taxa the areas
basis that these methods are strictly a dispersalist approach proposed by other authors [36–38]. They discussed the congru-
[10,11], or because of their basic assumption, namely that more ence of the various phylogenies that they obtained with the area
plesiomorphicareaswillbemorelikelytobepartoftheancestral cladograms of the aforementioned authors. They also compared
area, comparing it to the progression rule of Hennig [12], or the area relationships inferred from the Phymaturus phylogenies
because of the impossibility of identifying one basal area in a with the biogeographic analysis of the relationships between
symmetrical cladogram [13,14,15]. Dispersal-Vicariance Analysis provinces of the Andean subregion made by Morrone [39,40].
has also been criticized for its bias towards an all-vicariance However,theydidnotperformanyformalbiogeographicanalysis.
explanation [16], and for its inability to model extinction and D´ıaz Go´mez [16] made the first analysis of the historical
range expansions [17]. biogeography of Phymaturus using quantitative methodology. In
Recently, a new method for estimating geographic range thatstudy,anancestralareaanalysiswasmadeonatreethatalso
evolution [18,19] named Dispersal-extinction-cladogenesis model includedCtenoblepharysandasingleterminalrepresentingLiolaemus,
(DEC)havebeenproposed.Thismethodenablestheinferenceof usingthreedifferentmethodsofanalysis.However,theareasused
ancestralrangesinalikelihoodframework,rangecontractionsand in that study were previously defined, and based on the
expansions are caused by dispersal to an unoccupied area and distribution of arthropods [36–38].
local extinction within an area. This method requires an explicit In the case of Liolaemus, there have been biogeographic studies
description of likelihood of dispersal between areas and estimates using formal analyses: Young-Downey [41] made a Brooks
oflineagedivergencetimes.Givenaphylogeny,thedistributionof Parsimony Analysis (BPA) on a phylogeny of Liolaemus; Lobo
thetaxainvolved,andanexplicitmodelofDispersal-extinctinand [42] in a phylogenetic analysis of the Chilean group of Liolaemus
cladogenesis, dispersal and extinction rates are calculed using assignedtheareasdefinedbyRoig-Jun˜ent[43]tothespecieslisted
maximumlikelihood. asterminals;Schulteetal.[44]optimizedthedistributionofspecies
Ancestral area methods however, despite all its shortcomings, of Liolaemus on a molecular phylogeny, making an ancestral area
remain being widely used as ways toinfer the history of a taxon. analysis, although it was not explicit. D´ıaz Go´mez & Lobo [45]
DIVA in particular, remain very popular in the literature (more madeanancestralareaanalysisoftheChileangroupofLiolaemus
than 340 citations since it was published [17]), hence, it is very using Fitch optimization, Weighted Ancestral Area Analysis and
importanttoevaluatethebehaviorofthemorecommonmethods Dispersal-Vicariance analysis(DIVA).
used forreconstructing ancestral ranges. Acommonproblemofallthosestudiesisthattheareasusedas
units for the analyses were based on other taxa’s distribution
Liolaemidae [16,24,42,45], or arbitrarily defined [44]. As a result, the areas
Liolaemid lizards are the most common reptiles of southern maynotdescribeadequatelythedistributionofLiolaemidlizards.
South America. Members of this clade are distributed from the Mostoftheselizardshaverestricteddistributions,orareendemic
highAndesofcentralPerutotheshoresofTierradelFuegoand [46], so choosing an area much larger (such as geopolitical units,
fromsealeveltomorethan5000m[19–24].Liolaemidaeconsists or areas based on vegetation) than the distribution of liolaemid
of three genera: Ctenoblepharys, Liolaemus, and Phymaturus [25–28], species will cause unwanted situations: i.e: species that are
which currently include approximately 240 species [23,28,29]. allopatric having assigned the same area, or species that are
ThemonotypicCtenoblepharysisknownonlyfromcoastalsouthern presentonlyinasmallpartoftheareaassignedtothem,effectively
Peru [26 and is the sister taxon of the clade Liolaemus plus overestimating theactualdistributions.
Phymaturus[28,29].Phymaturusarerobust,saxicolouslizards,which This paper aims to evaluate the behavior of three common
are distributed from the high Andes of western Argentina and methodsforancestralareaanalysis:Fitchoptimization,Weighted
eastern Chile, to the Patagonian tablelands of Argentina [24]. Ancestral Area Analysis and Dispersal-Vicariance Analysis, using
Liolaemus is the most diverse genus of lizards in the southern as example the lizard family Liolaemidae, making a historical
hemisphere,include223recognizedspecies(secondonlytoAnolis biogeographyanalysisthataddressestheshortcomingsofprevious
in species richness), and an average of 4–5 new species are contributions, and using for the first time for this family, areas
described per year[23,30,31]. definedbasedonactualdistributionsratherthanpredefinedorad
Despite the importance of the group and the publication of hocareas.
several recent phylogenies, there are few explicit, quantitative
studies dealing with its historical biogeography. Cei [32] Materials and Methods
characterized the Patagonia as an active centre for speciation
anddispersalforthePatagonianherpetofauna,includingLiolaemus Phylogeny
as an example of recent adaptive radiation. Later [21], an The ancestral area methods require a phylogeny of the taxa
Andean-Patagonian origin was proposedfor Phymaturus,based on understudy.However,todatethereisnocompletephylogenyof
the refuge theory for geographic speciation, where the patagonic the three genera published. Recent molecular based phylogenies
tablelands would have acted as refuges and neo-dispersal centres [29] found Liolaemus and Phymaturus as sister taxa, with Cteno-
[33].Pereyra[34]basedonapheneticanalysisusingmeristicand blepharysassistertothatclade.Followingthatthehypothesisfrom
chromosomicdata,supportedadispersalscenarioproposedbyCei that study, a cladogram including the three genera was
[35] that placed Patagonia as the centre of origin for Phymaturus constructed (called a metatree [47]) (Fig. 1). For Liolaemus, recent
with the northern range of Phymaturus in Catamarca province, phylogenies were used for the Chilean group [45], and
Argentina, where it is ecologically replaced by a species of Argentinian (Eulaemus) group [28,29]. Phymaturus phylogeny was
LiolaemuswhichinhabitssimilarrockyhabitatsasPhymaturus.This takenfromLobo&Quinteros[24].Ctenoblepharyswasthenadded
PLoSONE | www.plosone.org 2 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
After the analysis, to each of the 170 species included in the
phylogeny depicted in the metatree, an area unit was assigned
following the results of the NDM analysis. The species that were
notrecoveredasendemictoanyareahadtobeassignedtooneor
more areas examining its distribution and comparing it to the
areas foundintheNDManalysis.
Ancestral area analysis
For the ancestral area analysis three different methods were
used: Fitch optimization [7], Dispersal-Vicariance Analysis
(DIVA) [9] and Weighted Ancestral Area Analysis (WAAA
henceforth) [8]. These methods use reversible parsimony to
optimize the areas on a tree, finding an estimate of the ancestral
distribution of a monophyletic clade, the ancestral area. Fitch
optimizationwasproposedbyRonquist[7]asanalternativetothe
Camin-Sokal (irreversible) optimization proposed originally by
Bremer [6]. DIVA uses a three-dimensional cost matrix that
assigns different costs to events (extinctions, dispersals and
vicariance) in order to minimize dispersal events. Using this
Figure1.Cladogramdepictingthephylogeneticrelationships approach, vicariance events have no cost, whereas dispersals and
of the family Liolaemidae, constructed joining recent partial extinctions have a cost of one per area unit added to the
phylogeniesforeachofthecladesincluded(calledaMetatree). distribution.Theoptimalreconstruction(s)arethoserequiringthe
Eachterminal,withtheexception ofCtenoblepharysrepresentseveral
minimalnumberofdispersalevents.WAAAusesFitchparsimony
species: chiliensis group, 86 species; boulengeri group, 40 species;
with a weighting scheme that weights favourably plesiomorphic
montanusgroup,9species,lineomaculatusgroup,12species.
doi:10.1371/journal.pone.0026412.g001 areas,andareasmorecommonasterminals.Withthismethod,a
probability index (PI) is calculated to give a measure of the
likelihood ofa particular area being part of theancestral area.
as the most basal taxon. The total number of species included in
For the DIVA analysis an additional adjustment was needed.
the phylogeny is 170 (147 Liolaemus, 22 Phymaturus and Cteno-
Duetoalimitationofthesoftware[9]nomorethan15areaunits
blepharys).
canbeusedintheanalysis.Inordertoapplythemethod,theareas
obtained by the NDM analysis were examined and areas which
Area selection
overlapped extensively (i.e. more than 50 percent) were joined
In order to improve over the previous contributions, an
forming onearea tobeusedintheDIVA analysis.
endemism analysis was made to delimit or describe area units to
FitchoptimizationwasmadewiththesoftwareTNTversion1.1
be used in the ancestral area analysis, based on actual Liolaemid
[54], constructing a matrix consisting on the tree and one
distributions. For the endemism analysis, distributional data were
characterrepresentingthedistributionoftheterminals.Weighted
collectedforallthespeciesincludedinthemetatree,frommuseum
Ancestral Area Analysis was made with the help of an Excel
collections and from theliterature. A data matrix was constructed
spreadsheet,andDispersalVicarianceAnalysiswasmadewiththe
and analyzed using the software NDM (Endems) [49,50]. NDM
softwareDIVA,version1.2[9].Thefollowingoptionswereused:
searches for areas of endemism using an optimality criterion that
settings: hold=32767, weight=1.000,age=1.000.
includesaspatialcomponent.NDMhasbeenshowntooutperform
Dispersal-evolution-cladogenesis model (DEC) was not applied
other common methods for identifying areas of endemism
inthisstudybecausethemethodrequiresmolecularphylogeniesto
[49,51,52]suchasParsimonyAnalysisofEndemicity(PAE;[53]), estimate likelihoods, and two of the phylogenies used (Chilean
that consists on scoring on a grid presences/absences of a set of group of Liolaemus and Phymaturus) are strictly morphology-based,
speciesinamatrix,andthenanalyzingitunderparsimonyusingthe andarethemostcompletepublishedtodate,includingmorethan
gridsasterminalsandthespeciesascharacters.Cladessupportedby half of thespecies included in thispaper, making it impossible to
twoormoretaxaareconsideredtorepresentareasofendemism.A evaluatethisnewandpotentiallyusefulapproachtotheestimation
gridsizeof0.75u60.75uwasused;asthereisnoformalargumentto of ancestral ranges.
select a ‘better’ grid size, different grid sizes were evaluated, and
selected the size that produced more areas, defined by more Results
taxa(with higher endemicity index). Grid origins were fixed at
X=280, Y=5. Radius size used were: to fill: X=40, Y=40; to Areas
assume X=80, y=80. Searches for endemism areas were The NDM analysis gave as a result 40 areas, and after the
conductedusingthefollowingoptions:savesetswithtwoormore consensus procedure 32 remained (Table 1). These areas are
endemicspecies,withscoreof1.5orhigher;swaponecellatatime; showninFigure2.TheareasusedfortheDIVAanalysisarelisted
discard superfluous sets; keep overlapping subsets only if 50% of in Table 2. Ctenoblepharys is not represented in any of the original
species unique; use edge proportions. In order to improve the areas,anewareawasdefinedcorrespondingtothedistributionof
support of the areas found, twenty replicates of the analysis were this genus, in order to be able to include Ctenoblepharys in the
made,eachusingadifferentseednumber,andtheresultingareas analysis.
weresavedina file.Later,duplicateareas were deleted usingthe
command ‘d’, and a consensus was calculated using a cut-off of Ancestral area analysis
40% (percent of similarity in species), and including areas in the The ancestral area analyses were applied on the complete
consensusonlyifitsharesthatpercentageofsimilaritywithallother phylogenies at species level, optimizing the distribution of
areasintheconsensus. individual species, but for clarity, only the results at higher level
PLoSONE | www.plosone.org 3 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Table1. Areacodification.
# Area # Area
0 CentralChile(Maule,O’Higgins) H NorthernAtacama(Chile)
1 CumbresCalchaquies I LosLagos(Chile)
2 Atacama J CentralMonte(LaRioja)
3 Arica K CentralChile(Metropolitana,O’Higgins)
4 PrepunaofCatamarca L CentralPatagonia(SantaCruz)
5 CentralBolivia M NorthernPatagonia
6 Araucan´ıa,B´ıo-B´ıo N CentralPatagonia(R´ıoNegro)
7 Payunia O Coquimbo
8 CentralR´ıoNegro P CentralChile(Coquimbo)
9 PayuniaandCentralChile Q PayuniaandMonteCentral
A PrepunaofSaltaandJujuy R CentralMonte(Mendoza)
B Atacama(Chile) S Prepuna(JujuyandBolivia)
C PunaofJujuy T AustralPatagonia
D Cuyo U SierrasSubandinasandCumbresCalchaqu´ıes
E AtacamaandCoquimbo V PrepunaandMonteBoreal
F Maule W CoastalCentralPeru
G AtacamaandPunaofBolivia
Thetableshowthelettersusedtorepresentdifferentareasfoundbytheendemismanalysis.
doi:10.1371/journal.pone.0026412.t001
cladesarereported.Allthreemethods(Table3)recoveredCentral (CoastalPeru),whichisthemostbasalareainthecladogram,and
Chile and Payunia as part of the ancestral area for Liolaemidae. appears as part of the ancestral area for the family, rendering a
TheareaCoastalPerualsoappearedaspartoftheancestralarea, disjunctancestralarea.Theproblemwithdisjunctsancestralareas
yielding a disjunct ancestral area (Figure 3). Both Fitch isthatthisdisjunctionhastobeexplainedbydeficientsamplingor
optimization and WAAA recovered an ancestral area smaller undetected extinctions, as one monophyletic clade cannot have
than the distribution of Liolaemidae, which implies dispersalist originatedinmorethanoneareasimultaneously,aswouldbethe
explanations.DIVAontheotherhandrecoveredanancestralarea case of a disjunct ancestral area. Unsurprisingly, this is the same
almost equal as the actual distribution of Liolaemidae, preferring result as the one from the historical biogeography analysis of
mostly vicariant explanations. Phymaturus [16], even though in this study a complete phylogeny
Fitch optimization assigned to the family Liolaemidae as with more appropriate areas for Liolaemus was used. The area
ancestral the following areas: Prepuna of Catamarca, Payunia Coastal Peru will appear as part of the ancestral area in every
and Central Chile, Maule, Central Chile (Regio´n Metropolitana reconstruction,eventhoughitistheareaofonlyonespecies,just
andO’Higgins)andCoastalPeru(correspondingtoCtenoblepharys) becauseofitsbasalposition,showingthatthispositionofanarea
(Fig. 4). WAAA assigns to the family Liolaemidae (Fig. 5) an in the cladogram will outweigh any other criteria WAAA tries to
ancestral area formed by Central Chile, Maule, O’Higgins, solve the problem of giving excessive weight to plesiomorphic
Coquimbo, Payunia, Austral Patagonia and Coastal Peru. areas counting how many times a particular area appears in the
Dispersal Vicariance analysis assigned to the family Liolaemidae cladogram.Thisway,anareathatisnotplesiomorphiccanstillbe
an ancestral area which encompasses almost all the actual recoveredasancestralifisoccupiedbyseveraltaxa.However,in
distribution of the family, only Atacama, Coquimbo and part of practice the most plesiomorphic area will have the highest
Austral Patagonia areexcluded (Fig.6). probability index of all, making very difficult for other areas to
reacha similar index.
Discussion Dispersal Vicariance Analysis uses reversible parsimony, but
uses a cost scheme that favors vicariance, giving as a result
About the methods ancestralareareconstructionsthatusuallyincludemost(ifnotall)
Fitchoptimizationmayrecoveroneormoreareasasancestral, of the areas, or giving several equally optimal reconstructions
ashappensforLiolaemidae.However,theinterpretation ofthese (Table 4; i.e. more than 100 reconstructions for Eulaemus).
results is not direct. In the case of this study, Fitch optimization Ronquist [9] proposed two possible solutions: one is add more
recovers four areas as ancestral, but these should not be outgroups,makingthebasalorrootnodenolongerroot;andlimit
interpreted as one ancestral area formed by four units, but as themaximumnumberofareasallowedtobepartoftheancestral
fourequallyprobableancestralareas.Assuch,Fitchoptimization area. The first solution, at least for Liolaemidae, is difficult to
will produce an all-dispersal scenario, no matter how many area implement. In the phylogenetic proposal of Frost and Etheridge
units arerecovered as ancestral. [55]LiolaemidaeisthemostbasalsubfamilyofTropiduridae,but
Regarding plesiomorphic areas being more likely part of the thereisnoconsensusaboutitssistertaxon,soaddingoutgroupsfor
ancestralarea,bothFitchoptimizationandWAAAgiveexcessive Liolaemidae isproblematic. Even ifoutgroups could be added, if
importance to the basal position of a particular area in the those outgroups were distributed on an area not occupied by
cladogram.ThisisevidentforthedistributionareaofCtenoblepharys Liolaemidae species, DIVA would add this new area to the
PLoSONE | www.plosone.org 4 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Figure2.AreasfoundbyNDM(Endems),asoftwaretosearchandindentifyareasofendemism.2A:0:CentralChile(Maule,O’Higgins);
1:CumbresCalchaqu´ıes;2:Atacama;3:Arica;5:CentralBolivia;6:Araucan´ıaandB´ıoB´ıo;8:CentralR´ıoNegro;A:PrepunaofSaltaandJujuy;D:Cuyo;
H:NorthernAtacama(Chile).-I:LosLagos;L:CentralPatagonia(SantaCruz);W:CoastalCentralPeru.2B:4:PrepunaofCatamarca;7:Payunia;B:
Atacama(Chile);E:AtacamaandCoquimbo;F:Maule;N:CentralPatagonia(R´ıoNegro);T:AustralPatagonia.2C:9:PayuniaandCentralChile;C:Puna
ofJujuy;J:CentralMonte(LaRioja);K:CentralChile(Metropolitana,O’Higgins);M:NorthernPatagonia.2D:G:AtacamaandPunaofBolivia;U:Sierras
SubandinasandCumbresCalchaqu´ıes;O:Coquimbo;R:CentralMonte(Mendoza).2E:S:Prepuna(JujuyandBolivia);V:PrepunaandMonteBoreal;P:
CentralChile.2F:Q:PayuniaandMonteCentral.
doi:10.1371/journal.pone.0026412.g002
ancestral reconstruction (as happens with the area of Ctenoble- isn’tacriteriontoselectanumberofareastorestricttheancestral
pharys). The second proposal allows restricting the maximum area, and this procedure should be used carefully. Kodandar-
number of areas recovered as ancestral [16,51]. However, there amaiah[19]suggestedusingdifferentlevelsofconstraintinsteadof
no constraining, or constraining to two or three areas; and
avoiding the use of DIVA when large scale extinctions are
Table2. AreasforDIVA.
suspected, giventhat DIVAdoesnot model extinctions well.
TherearemorethingstoconsideraboutDIVA.Theprogramis
no longer maintained (it wasn’t available to download from
Originalareas DIVAareas Originalareas DIVAareas
Ronquist’swebsiteatthetimeofwritingthisarticle),andalsohas
1-4-D-I A L I some serious limitations: it cannot accept polytomies in the
0-F-K-R B V J cladograms,themaximumnumberoftaxathatcanbeincludedis
3-G-H C 8 K 180,andnomorethan15areaunitscanbeusedintheanalysis.
ForLiolaemidae,thenumberofspeciescurrentlyismorethan240
6-7-9-Q D I L
[30,31] and more are described each year. When updated
10-B-C-S-U E C M
phylogenies are published, it will not be possible analyze them
2-E F 5 N
with DIVA because of this limitation. The maximum number of
M-N G Ctenoblepharys* O* areasallowedforcesonetomakeadhocdecisions,asjoiningareas
O-P H to reach that number, which negates the advantage of make an
endemismanalysistoidentifyareaswhichmoreaccuratelyreflect
OriginalareasfromtheNDManalysisandresultingareasusedintheDIVA
thedistribution of thespecies included inthestudy.
analysis,obtainedbyjoiningtogethertheoriginalareas.ForCtenoblepharys,
notrepresentedinanyoftheoriginalareas,anewareawasdefinedasO. Fortheanalysisofthehistoricalbiogeographyofataxon,Fitch
doi:10.1371/journal.pone.0026412.t002 optimizationshouldbeavoided,unlessadispersalistexplanationis
PLoSONE | www.plosone.org 5 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Table3. Ancestralareas.
Fitch WAAA DIVA
Liolaemidae 49FKW 09FPQTW ABCDEGHIJKLMO
Ctenoblepharys W W W
Phymaturus 49FK 9N BDGIK
Liolaemus 49FKQT 06FPT *
Chiliensis 9FKQ 0679FKPQR BDGHLK
Eulaemus 4T 4T *
lineomaculatus T T ADGHILM
(montanus+boulengeri) 4 18ACJV *
montanus 4 AC ABCEH
boulengeri 4 4JV ABDEGIJK
AncestralareaassignationsforLiolaemidaeandincludedclades.Groupswithanasteriskhavemultipleoptimalreconstructions,listedinTable4.
doi:10.1371/journal.pone.0026412.t003
preferred. Weighted ancestral area analysis and DIVA remain as Liolaemidae
alternatives, bearing in mind their limitations, particularly in the InapreviousstudyofthehistoricalbiogeographyofPhymaturus,
caseofDIVA.Usingfewerareaswillfacilitatetheanalysisandthe D´ıaz Go´mez [16] included two more terminals representing
interpretationoftheresults,butthismayrequirecompromiseslike LiolaemusandCtenoblepharys.Inthatstudy,thesamemethodologies
joining areas together, or constraining the maximum number of used here were applied, but the area units used were those
areas recovered asancestral. proposedbyotherauthorsandwerebasedonarthropods[37–39].
Moreover, the area assigned to Liolaemus was not based on an
explicit analysis, rather assigned based on a partial study of a
subcladeofLiolaemus[46]andpaleontologicaldata.Inthatstudy,
Patagonia Central and Coastal Peru were proposed as ancestral
areasforthefamilyLiolaemidae.Withtheexceptionoftheareain
Peru, the ancestral areas found in this study do not include
Patagonia Central.
The area Patagonia Central appeared in that study mainly
becauseofitsbasalpositiononthecladogram.Theancestralarea
methods favor more plesiomorphic areas as part of the ancestral
areas.Also,theareaPatagoniaCentralasdefinedbyRoig-Jun˜ent
et al. [39] also does not correspond with theareas found here for
Liolaemidae, being awide area including most of the south of
Argentina.
Fitch optimization founds a disjunct ancestral area. Thiscould
be interpreted in two ways: Only one of each of the areas is the
ancestral area (as in character reconstruction), which implies an
all-dispersalscenario,oracceptadisjunctancestralarea.However,
as a monophyletic group can only have one origin, one must
assumethatthedisjunctioniscausedbyextinctionorlackofdata.
Also it is evident that the areas Coastal Peru and Prepuna of
Catamarcaarerecoveredmainlybecauseoftheirbasalpositionin
thecladogram.
The results of this study can be resumed on two different and
opposinghypotheses.Onepostulatesarestrictedancestralareafor
theancestorofLiolaemidae,locatedinCentralChileandPayunia,
and the current distribution would be explained by dispersal to
Patagonia and, following the Andes to the north including Puna
andPrepuna.Thepaleontologicalevidenceavailableiscongruent
with this hypothesis; the oldest and currently only fossil of a
memberoftheLiolaemidaefamilyisaLiolaemusfromtheMiocene
of Patagonia, at the Gantman formation in Chubut, Argentina
Figure 3. Cladogram of Liolaemidae with ancestral area [56].
assignations. The triangles indicate that each terminal represent
The other hypothesis (from DIVA analysis) postulates a
severalspecies.Thenumbersinsidetrianglesshownumberofspecies.
widespread ancestor, distributed from Peru to the Patagonia,
Normal: Fitch optimization, Italics: Weighted Ancestral Area Analysis.
Bold:DispersalVicarianceAnalysis(DIVA).Nodeswithanasteriskhave following the Andes and arid regions in South America (Fig. 5).
multipleoptimalreconstructions. The current distribution of the family would be explained by
doi:10.1371/journal.pone.0026412.g003 successive vicariant events that fragmented the distributions and
PLoSONE | www.plosone.org 6 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Figure4.AncestralareaforLiolaemidaefoundbyFitchoptimization.NumbersorlettersrefertoFig.2.Formedby:PrepunaofCatamarca
(4),PayuniaandCentralChile(9),Maule(F),CentralChile(Regio´nMetropolitanaandO’Higgins-K)andCoastalPeru´ (W).
doi:10.1371/journal.pone.0026412.g004
PLoSONE | www.plosone.org 7 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Figure5.AncestralareaforLiolaemidaefoundbyWeightedAncestralAreaAnalysis(WAAA).NumbersorlettersrefertoFig.2.Formed
by:CentralChile(0),PayuniaandCentralChile(9),Maule(F),Coquimbo(P),Payunia(Q),AustralPatagonia(T),andCoastalPeru´ (W).
doi:10.1371/journal.pone.0026412.g005
PLoSONE | www.plosone.org 8 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Figure6.AncestralareaforLiolaemidaefoundbyDispersalVicarianceanalysis(DIVA).NumbersorlettersrefertoFig.2.Includesallarea
units,exceptAtacama(B),Coquimbo(P)andAustralPatagonia(T).
doi:10.1371/journal.pone.0026412.g006
PLoSONE | www.plosone.org 9 October2011 | Volume 6 | Issue 10 | e26412
EstimatingAncestralRangesinNeotropicalLizards
Table4. Multiplereconstrucions forDIVA.
Node Reconstructions
(boulengerimontanus) EJACEJBCEJABCEJCDEJACDEJBCDEJABCDEJCEGJACEGJBCEGJABCEGJCDEGJACDEGJBCDEGJABCDEGJCEHJACEHJBCEHJ
ABCEHJCDEHJACDEHJBCDEHJABCDEHJCEGHJACEGHJBCEGHJABCEGHJCDEGHJACDEGHJBCDEGHJABCDEGHJCEIJACEIJBCEIJ
ABCEIJCDEIJACDEIJBCDEIJABCDEIJCEGIJACEGIJBCEGIJABCEGIJCDEGIJACDEGIJBCDEGIJABCDEGIJCEHIJACEHIJBCEHIJABCEHIJ
CDEHIJACDEHIJBCDEHIJABCDEHIJCEGHIJACEGHIJBCEGHIJABCEGHIJCDEGHIJACDEGHIJBCDEGHIJABCDEGHIJCEJKACEJKBCEJK
ABCEJKCDEJKACDEJKBCDEJKABCDEJKCEGJKACEGJKBCEGJKABCEGJKCDEGJKACDEGJKBCDEGJKABCDEGJKCEHJKACEHJK
BCEHJKABCEHJKCDEHJKACDEHJKBCDEHJKABCDEHJKCEGHJKACEGHJKBCEGHJKABCEGHJKCDEGHJKACDEGHJKBCDEGHJK
ABCDEGHJKCEIJKACEIJKBCEIJKABCEIJKCDEIJKACDEIJKBCDEIJKABCDEIJKCEGIJKACEGIJKBCEGIJKABCEGIJKCDEGIJKACDEGIJK
BCDEGIJKABCDEGIJKCEHIJKACEHIJKBCEHIJKABCEHIJKCDEHIJKACDEHIJKBCDEHIJKABCDEHIJKCEGHIJKACEGHIJKBCEGHIJK
BCEGHIJKCDEGHIJKACDEGHIJKBCDEGHIJKABCDEGHIJK
Eulaemus ACEJMABCEJMACDEJMABCDEJMACEGJMABCEGJMACDEGJMABCDEGJMACEHJMABCEHJMACDEHJMABCDEHJMACEGHJM
ABCEGHJMACDEGHJMABCDEGHJMACEIJMABCEIJMACDEIJMABCDEIJMACEGIJMABCEGIJMACDEGIJMABCDEGIJMACEHIJM
ABCEHIJMACDEHIJMABCDEHIJMACEGHIJMABCEGHIJMACDEGHIJMABCDEGHIJMACEJKMABCEJKMACDEJKMABCDEJKM
ACEGJKMABCEGJKMACDEGJKMABCDEGJKMACEHJKMABCEHJKMACDEHJKMABCDEHJKMACEGHJKMABCEGHJKMACDEGHJKM
ABCDEGHJKMACEIJKMABCEIJKMACDEIJKMABCDEIJKMACEGIJKMABCEGIJKMACDEGIJKMABCDEGIJKMACEHIJKMABCEHIJKM
ACDEHIJKMABCDEHIJKMACEGHIJKMABCEGHIJKMACDEGHIJKMABCDEGHIJKMACEJLMABCEJLMACDEJLMABCDEJLMACEGJLM
ABCEGJLMACDEGJLMABCDEGJLMACEHJLMABCEHJLMACDEHJLMABCDEHJLMACEGHJLMABCEGHJLMACDEGHJLMABCDEGHJLM
ACEIJLMABCEIJLMACDEIJLMABCDEIJLMACEGIJLMABCEGIJLMACDEGIJLMABCDEGIJLMACEHIJLMABCEHIJLMACDEHIJLM
ABCDEHIJLMACEGHIJLMABCEGHIJLMACDEGHIJLMABCDEGHIJLMACEJKLMABCEJKLMACDEJKLMABCDEJKLMACEGJKLM
ABCEGJKLMACDEGJKLMABCDEGJKLMACEHJKLMABCEHJKLMACDEHJKLMABCDEHJKLMACEGHJKLMABCEGHJKLMACDEGHJKLM
ABCDEGHJKLMACEIJKLMABCEIJKLMACDEIJKLMABCDEIJKLMACEGIJKLMABCEGIJKLMACDEGIJKLMABCDEGIJKLMACEHIJKLM
ABCEHIJKLMACDEHIJKLMABCDEHIJKLMACEGHIJKLMABCEGHIJKLMACDEGHIJKLMABCDEGHIJKLM
Liolaemus ACEHJLMABCEHJLMACDEHJLMABCDEHJLMACEGHJLMABCEGHJLMACDEGHJLMABCDEGHJLMACEHIJLMABCEHIJLMACDEHIJLM
ABCDEHIJLMACEGHIJLMABCEGHIJLMACDEGHIJLMABCDEGHIJLMACEHJKLMABCEHJKLMACDEHJKLMABCDEHJKLMACEGHJKLM
ABCEGHJKLMACDEGHJKLMABCDEGHJKLMACEHIJKLMABCEHIJKLMACDEHIJKLMABCDEHIJKLMACEGHIJKLMABCEGHIJKLM
ACDEGHIJKLMABCDEGHIJKLM
CladeswithmultipleoptimalreconstructionsfoundbyDIVA.Allreconstructionshavethesamecost,andareequallyprobable.
doi:10.1371/journal.pone.0026412.t004
dispersalsinordertoexplainthespeciesfoundintheeastofSouth iguanian fauna of Patagonia, particularly for Liolaemus and
America.Morepaleontologicaldatacouldbeusefultosupportthis Phymaturus. Later, Pereyra [34], based on a phenetic analysis,
explanation, but unfortunately there are no fossil records for followed Cei’s hypothesis proposing Patagonia as the centre of
Ctenoblepharys or Phymaturus. For Iguanidae the fossil record is origin for Phymaturus, postulating a differentiation between
scarce [53], theearliest fossil that canbe referred to Iguanidae is northern and southern populations, and a posterior invasion to
Pristiguanabrasiliensis,fromtheUpperCretaceousBauru´ formation thePatagoniabythenorthernspecies.Neitherofthesehypotheses
of Brazil [58]. Interestingly, this fossil shows characters similar to are supported by the results of this study, that postulate as
thetropidurines,acladecloselyrelatedtoLiolaemidae[25].There ancestralareaforPhymaturusPayunia,CentralChileandnorthern
areotherrecordsforIguanianlizardsfortheCenozoicofBolivia Patagonia. In the case of the scenario proposed by Pereyra, only
[59] and Patagonia [60]. If these fossils could be related to Fitch optimization includes part of the northern distribution of
ancestorsofLiolaemidae,theywouldsupportawidespreadorigin Phymaturus(PrepunaofCatamarca),buttherearenodispersalsof
for thefamily. thenorthern speciestosouthern areas.
Some aspects of the distribution of Liolaemidae could be D´ıaz Go´mez [51] published an ancestral area analysis for
explained by a widespread ancestor. For example, most Liolaemus Phymaturus, applying the same methodology used here, but using
species are distributed in the arid regions of southern South areasdefinedbyarthropods[36].Inthatstudytheancestralarea
America,andintheAndescordillera,precordilleraandPatagonia. forPhymaturuswasCentralPatagonia(plusAndeanCordilleraand
However, a small group of species (of the weigmannii group) are Valle Central in Chile for DIVA analysis). The ancestral area
distributed forming a series of‘islands’ fromthecoasts of Buenos found here by Fitch optimization is not congruent with those
Aires, Uruguay and Brazil, up to Rio de Janeiro, associated with results. The WAAA results of this study are congruent with D´ıaz
sand dunes. These species are disjunct from the rest of Liolaemus Go´mez (2007), including Central Chile and Central Patagonia.
speciesthatarenotpresentintheChacoorinhumidforests.The However,theareaidentifiedasPatagoniainthepreviousstudyis
desertification process from the Miocene to Pliocene [57] may much bigger than the area defined here as Central Patagonia,
have allowed these sand systems to expand, followed by the making the results not directly comparable. The DIVA results of
expansionofthedistributionoftheancestorsofthosespecies.After this study are congruent with the area proposed by D´ıaz Go´mez
that, the arid/humid cycles following the glacial and interglacial [48], mainly because DIVA found an ancestral area that
periodsofthePlioceneandPleistocene producedexpansionsand encompass almost all the current distribution of Phymaturus,
retractionsofhumidandxerichabitats,actingasvicariantevents including completely the areas Payunia, Central Chile, Central
and causing the fragmentation and speciation of the extant taxa Patagonia andAraucan´ıa.
[61–64].
About Liolaemus
About Phymaturus Laurent[65–67]dividedthegenusLiolaemusintwogroups,the
Cei [21] postulated an Andean-Patagonian origin for Phyma- Chilenogroup(Liolaemussensustrictoorchiliensis)andtheArgentino
turus, based on a refuge theory, where the patagonic tablelands group(Eulaemus),pointingattheAndeanupliftasthecauseofthis
would have been refuges and neo-dispersal centres for the division. The results from this study support this hypothesis,
PLoSONE | www.plosone.org 10 October2011 | Volume 6 | Issue 10 | e26412