Phylogeny,biogeography anddiversificationpatterns · Phylogeny,biogeography anddiversificationpatterns ofside-neckedturtles (Testudines:Pleurodira) GabrielS.Ferreira1,2,3,MarioBronzati4,5,

rsosroyalsocietypublishingorg

ResearchCite this article Ferreira GS Bronzati MLanger MC Sterli J 2018 Phylogenybiogeography and diversification patterns ofside-necked turtles (Testudines Pleurodira)R Soc open sci 5 171773httpdxdoiorg101098rsos171773

Received 1 November 2017Accepted 26 February 2018

Subject CategoryBiology (whole organism)

Subject Areasbiogeographytaxonomy andsystematicspalaeontology

KeywordsPleurodira historical biogeographyBioGeoBEARS transoceanic dispersal diversity

Author for correspondenceGabriel S Ferreirae-mail gsferreirabiogmailcom

Electronic supplementary material is availableonline at httpsdxdoiorg106084m9figsharec4035212

Phylogeny biogeographyand diversification patternsof side-necked turtles(Testudines Pleurodira)Gabriel S Ferreira123 Mario Bronzati45

Max C Langer1 and Juliana Sterli61Biology Department FFCLRP University of Satildeo Paulo Ribeiratildeo Preto Brazil2Senckenberg Center for Human Evolution and Palaeoenvironment (HEP) at EberhardKarls Universitaumlt Sigwartstraszlige 10 72076 Tuumlbingen Germany3Fachbereich Geowissenschaften der Eberhard Karls Universitaumlt TuumlbingenHoumllderlinstraszlige 12 72074 Tuumlbingen Germany4Bayerische Staatssammlung fuumlr Palaumlontologie und GeologieRichard-Wagner-Strasse 10 80333 Munich Germany5Department of Earth and Enviromental Sciences LudwigndashMaximiliansndashUniversitaumltRichard-Wagner-Strasse 10 80333 Munich Germany6CONICET-Museo Paleontoloacutegico Egidio Feruglio Fontana 140 9100 Trelew ChubutArgentina

GSF 0000-0003-1554-8346 MB 0000-0003-1542-3199

Pleurodires or side-necked turtles are today restrictedto freshwater environments of South America AfricandashMadagascar and Australia but in the past they weredistributed much more broadly being found also on EurasiaIndia and North America and marine environments Twohypotheses were proposed to explain this distributionin the first vicariance would have shaped the currentgeographical distribution and in the second extinctionsconstrained a previously widespread distribution Herewe aim to reconstruct pleurodiran biogeographic historyand diversification patterns based on a new phylogenetichypothesis recovered from the analysis of the largestmorphological dataset yet compiled for the lineagetesting which biogeographical process prevailed duringits evolutionary history The resulting topology generallyagrees with previous hypotheses of the group andshows that most diversification shifts were related to theexploration of new niches eg littoral or marine radiationsIn addition as other turtles pleurodires do not seemto have been much affected by either the CretaceousndashPalaeogene or the EocenendashOligocene mass extinctions Thebiogeographic analyses highlight the predominance of bothanagenetic and cladogenetic dispersal events and support

2018 The Authors Published by the Royal Society under the terms of the Creative CommonsAttribution License httpcreativecommonsorglicensesby40 which permits unrestricteduse provided the original author and source are credited

rsosroyalsocietypublishingorgRSocopensci5171773

the importance of transoceanic dispersals as a more common driver of area changes than previouslythought agreeing with previous studies with other non-turtle lineages

1 IntroductionThe turtle crown group Testudines is composed of two lineages with extant taxa [1] cryptodiresor hidden-necked turtles and pleurodires or side-necked turtles Although early studies (eg [23])proposed a Triassic origin for the crown group more recent analyses suggest a maximum age of ca 165 Ma(Middle Jurassic) [45] Stem-pleurodires are known from Late Jurassic and Early Cretaceous deposits [6]but the oldest records of the crown group come from the Barremian (Early Cretaceous) of Brazil [78]

Pleurodires today represent a small fraction of the diversity of Testudines (93 of 356 species [9])and are restricted to freshwater environments (although some chelids seem to tolerate higher levels ofsalinity [1011]) of Africa Australia Madagascar and South America [12] Their fossil record howeverreveals a much broader distribution including Eurasia India and North America (eg [12ndash14]) as wellas taxa adapted to marine (at least coastal) environments (eg [121516]) The crown group includesthree lineages with extant representatives Chelidae Pelomedusidae and Podocnemididae and threeextinct groups Araripemydidae Euraxemyidade and Bothremydidae [121317] Podocnemididae andBothremydidae are by far the most abundant pleurodires in the fossil record and a proposed peak inpleurodire diversity during the Cretaceous and Paleocene seems to be related to the diversification ofthose two groups [12]

The biogeographic distribution of extant pleurodires which are restricted to some areas of theSouthern Hemisphere has been the subject of many investigations A classical viewpoint (eg [1819]) isthat the disjunct Pelomedusoides (Pelomedusidae + Podocnemididae see the electronic supplementarymaterial for definitions of pleurodiran clades) distribution represents the outcome of vicariant eventsin this case caused by the break-up of the supercontinent Gondwana Contrary to those interpretationsNoonan [20] using a phylogenetic hypothesis based on molecular data combined with that of Meylan[21] from fossil taxa suggested that the distribution of extant pelomedusoids is a remnant of a muchmore widespread pattern shaped by large-scale extinctions Later Romano amp Azevedo [22] based on areanalysis of the morphological data matrix of de la Fuente [23] adding more fossil taxa corroboratedthe idea of vicariant events shaping the biogeography of pelomedusoids As for chelids even thoughmolecular (eg [524ndash28]) and morphological (eg [172930]) results disagree regarding the position ofthe Australian and South American taxa both hypotheses return similar biogeographic interpretationswhere the group starts to diversify prior to the separation of Australia from the remaining of Gondwanasuggesting a widespread distribution for the group before this vicariant event [3132]

The above-mentioned studies were however conducted prior to several fossil findings that greatlyincreased our knowledge of pleurodiran taxonomic morphological and distributional diversities (eg[81416]) Additionally although some large phylogenetic analyses have been conducted (eg [12ndash14]) sampling a variety of fossil taxa these were restricted to some pleurodiran subclades and nophylogenetic analysis including a comprehensive sample of all lineages of the group has been toour knowledge so far conducted Considering that reliable phylogenetic frameworks are necessaryto conduct numerical diversity and biogeographic analyses [33ndash35] a well-sampled phylogenetichypothesis for Pleurodira is needed to test the previously proposed evolutionary scenarios

To fulfil the above-mentioned requirement we compiled the largest morphological matrix includinga broad taxon sample of all pleurodiran lineages obtaining to our knowledge the most inclusivephylogenetic hypothesis so far proposed for the clade Based on this new hypothesis we conducteddiversification and historical biogeography analyses to explore the evolution of the side-necked turtles

2 Material and methods21 Phylogenetic analyses and time-scaled treesA new taxonndashcharacter matrix (101 taxa times 245 characters) for Pleurodira was built using MESQUITE

v 30 [36] The character list is largely based on the extensive studies of Gaffney et al [1213] withadditions from other sources (eg [142337ndash42]) especially those focusing on chelids [172943] and 18new characters proposed here Special effort was made to better sample post-cranial structures resultingin 97 characters from that partition (395 of the total) more than in any previous study (eg [12] and

[14] had 297 and 364 of post-cranial characters respectively) Twelve characters were interpreted asforming morphoclines and ordered accordingly (see the electronic supplementary material for additionalinformation about the phylogenetic analysis)

The taxon sample was conceived to incorporate all pleurodiran lineages namely ChelidaePelomedusidae Araripemydidae Euraxemydidae Bothremydidae and Podocnemididae including 98crown pleurodires as terminal taxa (see the electronic supplementary material for the complete taxonlist) Previously the largest matrices for pleurodires comprised 43 [8] and 91 [14] in-group taxa (althoughthe latter study employed a reduced version with 70 in-group taxa in the main analyses) The non-Testudines Testudinata Proganochelys quenstedti and the stem-pleurodires Notoemys laticentralis andPlatychelys oberndorferi composed the outgroup taxa

The resulting matrix was analysed under the parsimony criterion in TNT v 11 [44] via a heuristicsearch with the following settings 2000 replicates of Wagner trees random seed = 0 tree bisectionreconnection (TBR) for branch-swapping hold = 20 and collapse of zero-length branches accordingto rule lsquo1rsquo of TNT The most parsimonious trees (MPTs) found in this first round of the analysiswere the subject of a second round of TBR A strict consensus tree decay (Bremer support) andresampling (bootstrap and jackknife) values were obtained using implemented functions on TNT Theresampling values were calculated using 1000 replicates for absolute and difference of frequencies (grouppresentcontradicted or GC in [45]) Consistency indexes (CI) and retention indexes (RI) were calculatedusing the script statsall (designed by Peterson L Lopes v 13) The IterPCR script [46] was used to identifyunstable taxa during preliminary analyses (ie taxa with multiple alternative positions) and to re-evaluate our scoring when the instability was caused by conflict of information rather than missing dataAdditionally considering that molecular-based phylogenetic analyses retrieve distinct arrangements forseveral extant taxa (eg [20242527]) we ran a second analysis (referred to hereafter as the lsquomolecularconstrainedrsquo analysis) following the same settings as the previous (referred to hereafter as the lsquooriginalrsquoanalysis) except for setting constraints (see the electronic supplementary material) for the relations ofthe extant taxa based on the molecular phylogenetic hypothesis of Guillon et al [25] We also conductedthree additional constrained analyses to evaluate how many steps were needed to achieve arrangementsfound in other alternative hypotheses

In addition to the lsquooriginalrsquo and lsquomolecular constrainedrsquo trees we obtained two additional topologiesto be employed in the subsequent analyses The lsquonon-marine taxa treersquo was built for the biogeographyanalyses by pruning taxa previously considered marine or adapted to brackish water ie Bothremydini ampStereogenyini [12151647] from the strict consensus tree of the lsquooriginalrsquo analysis Further an informallsquosupertreersquo was built for the diversity and diversification analyses by adding extinct and extant taxanot included in the lsquooriginalrsquo phylogenetic analysis to the strict consensus tree The four topologieswere time-scaled with the R [48] package strap [49] using information from the literature to define timeranges for each taxon (see the electronic supplementary material) and dividing branch lengths equallyalong the tree to avoid zero or close-to-zero values [50] As the biogeographic analysis requires fullydichotomous topologies the polytomies of the lsquooriginalrsquo tree were manually resolved by deliberatelychoosing particular arrangements (see the electronic supplementary material for additional topologies)

22 Diversification analysis and diversity curvesThe diversification analyses were conducted on the software SYMMETREE [51] using time-sliced treesaccording to the procedure presented by Tarver amp Donoghue [52] (but see [53] for an alternativeapproach) Accordingly each time-sliced tree is composed of a subset of the original tree containingtaxa of the same age or older than the given period in addition to the ghost lineages of younger taxaSeven different intervals were created Early Cretaceous Late Cretaceous Paleocene Eocene OligoceneMiocene and Pliocene-Recent Still we followed the procedure outlined in Bronzati et al [54] when adiversification shift was detected in a time interval younger than the clade it is referred to in order todifferentiate real shifts (ie detected shifts caused by diversification) from artefacts (ie detected shiftsresulting from speciation and extinction) We conducted two analyses one with the lsquooriginalrsquo tree and asecond with the lsquosupertreersquo in order to evaluate the effect of the sampling of our matrix in the resultingshifts

We also built two types of diversity curves across geological time for natural (ie clades) and artificial(ie ecological guilds) groups of Pleurodira to compare the diversity variation with biogeographic andclimatic events In this procedure we used the lsquosupertreersquo to increase our sample lsquoTaxic diversity curvesrsquowere obtained using the total number of taxa per time bin Considering that this kind of data is muchaffected by the incompleteness of the fossil record and the number of fossiliferous deposits on a given

period a second type of curve was created for comparison called lsquophyletic diversity curversquo that alsotakes into account the number of ghost lineages in each of the intervals

23 Estimation of ancestral ranges and number and types of biogeographical eventsWe conducted a series of ancestral area reconstructions on the R package BioGeoBEARS whichimplements three of the most used models in historical biogeography analyses in a common likelihoodframework [55] namely the LAGRANGE dispersalndashextinctionndashcladogenesis (DEC) model [3456] alikelihood version of DIVA (DIVALIKE) [57] and a likelihood version of the range evolution model(BAYAREA) implicit to the methods BayArea [58] and Bayesian binary model (BBM) of RASP [59]Each of those methods makes different assumptions about anagenetic and cladogenetic range changeprocesses [35] and those assumptions usually have a large impact on the results [60] The multi-modelapproach implemented in BioGeoBEARS allows the evaluation of competing hypotheses potentiallygenerated by those models by comparing the fit of their assumptions to the observed data [5560] Tocompare and choose between the different models we conducted a likelihood ratio test (LRT) for thenested models and the Akaike information criterion corrected for sample size (AICc) for the non-nestedmodels

We ran a first set of time-stratified analyses using a time-scaled version of the lsquooriginalrsquo treeaccounting for nine models the standard models (herein named M0) DEC DIVALIKE and BAYAREA[55] and two additional versions to each of those including x [61] and j [60] as free parameters (hereinnamed M1 and M2 respectively) The x parameter is used to estimate the relative probability of dispersalas a function of distance modifying the dispersal rates from area A to area B by (distance [A toB])x [61] The distances between the areas are established by user-defined matrices (see the electronicsupplementary material) When x is set as a free parameter the distance between the areas may influenceeach branch on the tree differently With j as a free parameter founder-event speciation is added to themodel allowing thatmdashduring cladogenesismdashone of the descendants jumps to a new range outside thatof the ancestor without prior range expansion [60]

Considering that marine taxa are not affected by the oceanic barriers that separate continental areasas terrestrial or freshwater turtles are distances may influence differently their dispersal capabilitiesand jumps to new areas without prior range expansion may not be uncommon This justifies testingthe influence of x and j in additional models [5562] As marine taxa are usually excluded frombiogeographic inferences (eg [63]) we ran a second set of stratified analyses using the lsquonon-marine taxarsquotree for comparison Moreover phylogenetic hypotheses for pleurodires using molecular data usuallyresult in different arrangements for AustralianSouth American chelids and South AmericanMalagasypodocnemidids (eg [5202527]) To test if those different topologies imply distinct ancestral areareconstructions we ran a third set of stratified analyses using the lsquomolecular constrainedrsquo tree Thetwo latter sets of analyses were conducted using only the nested models (M0 M1 and M2) of DEC andDIVALIKE because BAYAREA performed worse than those for the first set (see Results)

Three time ranges with their respective distance matrices were considered as they are thought torepresent intervals with a similar configuration of landmasses 170ndash911 911ndash555 and 555ndash0 Ma Weconsidered 10 possible areas South America Africa North America Madagascar Australia EuropeIndia ArabiaMiddle East East Asia and Antarctica Even though there is no record of pleurodires inAntarctica it could act as a land bridge between closely related areas in some time bins and as suchcould appear as a reconstructed ancestral area Three distance matrices between the areas were definedone for each time slice

We also conducted biogeographical stochastic mapping (BSM) implemented in BioGeoBEARS [64] onthe lsquooriginalrsquo tree dataset using M1 and M2 models of DIVALIKE and DEC to evaluate the impact of agiven model on the estimated frequencies of each type of range change event BSM simulates possiblebiogeographic histories constrained to produce the observed data of a given dataset estimating the timesand positions of all events on the tree [65] Means and standard deviations of event counts from 50 BSMwere used to estimate event frequencies

3 Results31 Phylogenetic analysesThirty-six MPTs of 1128 steps (CI = 0290 RI = 0750) were found by the unconstrained analysis of thelsquooriginalrsquo matrix (best score hit 388 times) Bremer support (BS) bootstrap and jackknife values lists

Bothremys cooki

Bairdemys sanchezi

Chelus fimbriatus

Araiochelys h

irayamai

Podocnemis erythrocephala

mys bl

Bothremys maghrebiana

Podocnemis bassleri

Yaminuechelysvgasparinii

Peiropemys mezzalirai

Dirqadim

schaefferi

Taphrosphys sulcatusShweboemys pilgrim

Azabbaremys moragjonesi

Podocnemis lew

Araripem

ys barretoi

Bauruemys elegans

Podocnemis expansa

Proganochelys quenstedti

Chelus colombianus

Pseudemydura umbrina

Bairdemys thalassica

Caninem

ys tridentata

Zolhafah bella

Acanthochelys radiolata

Platychelys oberndorferi

Taphrosphys congolensis

Bonapartemys bajobarrealis

Bothremys kellyi

Euraxem

ys essweini

Phosphatochelys tedfordi

lsquoNeo

ysrsquo f

Rosasia

soutoi

Laganemys tenerensis

Yaminuechelysvm

leChelodina colliei

Emydura macquarii

Arenila krebsi

Cambaremys langertoni

Notoemys laticentralis

Chedighaii hutchisoni

Podocnemis vogli

Prochelidella cerrobarcinae

Bairdemys winklerae

Phrynops geoffroanus

Bothremys arabicusMyuchelys latisternum

Latentem

ys plowden

Taphrosphys ippolitoi

Mesoclemmys nasuta

mys tr

lsquoBaird

emysrsquo

healey

Platemys platicephala

Hydrom

edusa tectifera

Podocnemis unifilis

Pricemys caiera

Labrostochelys galkini

Brasilemys josaiPortezueloemys patagonica

Rhothonemys brinkmani

Brontochelys gaffn

Elseya dentata

Lapparentemys vilavilensis

Cerrejonem

ys wayuunaiki

Lemurchelys diasphax

Bairdemys hartsteini

Phrynops hilarii

Hamadachelys escuilliei

Acleistochelys maliensis

Cordic

Nigeremys gigantea

Hydrom

edusa maxim

illiani

Podocnemis sextuberculata

Ummulisani rutgersensis

Chedighaii barberi

Bairdemys venezuelensis

Stereogenys cromeri

3 1 1 1

Stereogenyini

Erymnochelyinae

Podocnemidinae

Pelomedusoides

Taphrosphyini

Bothremydini

Cearachelyini

Pleurodira

Chelini

Chelina

Pan-Pleurodira

Peiropemydidae

Chelidae

Bothremydidae

Euraxemydidae

Pelomedusidae

Araripemydidae

Podocnemididae

Figure 1 Strict consensus of 36MPTs of 1128 steps each (CI= 0290 RI= 0750) Numbers belownodes represent Bremer support valueshigher than 0 Names refer to node- (circles) and branch-based (triangles) clades

of characters and common synapomorphies as well as phylogenetic definitions for the clades and amore detailed description of the results are provided in the electronic supplementary material The strictconsensus tree (figure 1) is well resolved (93 of the possible nodes) and mostly agrees with previousmorphology-based analyses recovering the monophyly of the main pleurodiran lineages with highsupport The lsquoconstrainedrsquo analysis found 756 MPTs of 1175 steps Although much longer the resultingstrict consensus tree is very similar to the lsquooriginalrsquo (see the electronic supplementary material) with theexception of the forced relations between extant taxa

32 Diversification shifts and diversity levelsShifts in three clades were recovered in all periods subsequent to time bin 1 Early Cretaceous (figure 2)Pleurodira (also in time bin 1) Podocnemidoidea and the clade including Cearachelyini + Bothremydini +Taphrosphyini On time bin 4 Eocene (figure 3) two additional shifts were recovered one for the cladeincluding Erymnochelys madagascariensis and Neochelys arenarum but not Peltocephalus dumerilianus andanother for the clade Peiropemydidae + Podocnemididae The latter is also seen in all subsequent timebins but the former is restricted to time bins 3 and 4 Two new shifts appear in time bins 6 and 7 onefor the Stereogenyini clade including Bairdemys venezuelensis and Stereogenys cromeri but not lsquoBairdemysrsquohealeyorum and one for Chelini On time bin 7 a shift for the clade including all chelids but Emyduramacquarii was recovered

0102030405060708090100

Upper Lower Upper Paleocene Eocene Oligocene Miocene

Jurassic Cretaceous Palaeogene Neogene Qu

Elseya dentataMyuchelys latisternumPseudemydura umbrinaAcanthochelys radiolataPlatemys platycephalaMesoclemmys nasutaPhrynops geoffroanusPhrynops hilariiChelus fimbriatus

Chelus colombianusChelodina collieiHydromedusa maximillianiHydromedusa tectifera

Prochelidella cerrobarcinaeBonapartemys bajobarrealis

Yaminuechelys gaspariniiYaminuechelys maior

Emydura macquarii

Laganemys tenerensisAraripemys barretoi

Dirqadim schaefferiEuraxemys essweini

Pelomedusa subrufaPelusios castanoidesPelusios castaneus

Sokatra antitraAtolchelys lepida

Kurmademys kallamedensisSankuchemys sethnaiKinkonychelys rogersi

Cearachelys placidoiGalianemys whiteiGalianemys emringeri

Polysternon provincialeFoxemys mechinorum

Foxemys trabantiRosasia soutoi

Zolhafah bellaAraiochelys hirayamen

Chedighaii barberiChedighaii hutchisoni

Bothremys kellyiBothremys arabicus

Bothremys cookiBothremys maghrebiana

Nigeremys giganteaArenila krebsi

Acleistochelys maliensisAzabbaremys moragjonesi

Labrostochelys galkiniRhothonemys brinkmani

Phosphatochelys tedfordiUmmulisani rutgersensis

Taphrosphys congolensisTaphrosphys ippolitoiTaphrosphys sulcatus

Platychelys oberndorferiNotoemys laticentralis

Podocnemidoidae

Africa

EuropeAsia

South America

North America

Madagascar

Australia

Figure 2 Ancestral area reconstructions and diversification shifts for the time-calibrated lsquooriginalrsquo tree excluding Podocnemidoidae(figure 3) Rectangles next to each terminal taxon represent its area distribution pie charts represent the probabilities for ancestral areaof nodes and yellow stars point to a node in which a diversification shift was found

The taxon diversity curves for Pleurodira show little perturbation with an increasing diversityduring the Early Cretaceous a peak during the Campanian and Maastrichtian a slight drop duringthe Early Paleocene and a fast recovery by the end of this period (figure 4) This pattern is replicated inthe curve for Pelomedusoides whereas Chelidae shows an increase by the Early Cretaceous followed

0102030405060708090100

Brasilemys josaiHamadachelys escuilliei

Portezueloemys patagonicaCambaremys langertoni

Bauruemys elegansPeiropemys mezzalirai

Lapparentemys vilavilensisPricemys caiera

Caninemys tridentataCerrejonemys wayuunaiki

Podocnemis vogliPodocnemis erythrocephalaPodocnemis expansaPodocnemis unifilisPodocnemis lewyana

Podocnemis bassleriPodocnemis sextuberculata

Stupendemys geographicusCarbonemys cofrinii

Dacquemys paleomorphaUCMP 42008

Neochelys fajumensisPapoulemys laurenti

Kenyemys williamsiNeochelys arenarum

Neochelys franzeniTurkanemys pattersoni

Erymnochelys madagascariensisPeltocephalus dumerilianus

Mogharemys blackenhorniBairdemys healeyorumCordichelys antiqua

Latentemys plowdeniBrontochelys gaffneyiLemurchelys diasphax

Shweboemys pilgrimiStereogenys cromeri

Bairdemys winkleraeBairdemys hartsteini

Bairdemys sancheziBairdemys venezuelensis

Africa

Europe

South AmericaNorth America

Madagascar

Australia

Figure 3 Ancestral area reconstructions and diversification shifts for the time-calibrated lsquooriginalrsquo tree including only PodocnemidoidaeRectangles next to each terminal taxon represent its area distribution pie charts represent the probabilities for ancestral area of nodesand yellow stars point to a node in which a diversification shift was found

by a relative stasis (figure 4) which is surely a consequence of its poor fossil record The contrastbetween Bothremydidae and Podocnemidoidae (figure 4) reveals a higher and increasing diversity ofthe former from the Aptian to the Campanian decreasing afterwards until its extinction during the EarlyEocene Although the first records of Podocnemidoidae also extends back to the Aptian its diversityremains low during most of the Cretaceous to rise more quickly after the Santonian with a peakfrom the Late Paleocene to the Miocene when it also starts to decrease (figure 4) Lastly the contrastbetween the freshwater and littoralmarine taxa (figure 4) shows a predominance of freshwater formswith a curve almost identical to that of Pleurodira except for the Campanian peak that results fromaccounting the marine taxa The diversity curve of littoralmarine pleurodires reflects the Bothremydiniand Taphrosphyini diversification beginning in the Late Cretaceous and declining after the Campanianbut rising again from the Late Eocene to the Miocene (figure 4) with the diversification of Stereogenyini

kim tit ber

maa dan

pleurodiran phyletic diversity Pan-Chelidae Pan-Pelomedusoides

kim tit ber

maa dan

kim tit ber

maa dan

Bothremydidae Podocnemidoidae freshwater taxa littoralmarine taxa

kim tit ber

maa dan

pleurodiran taxic diversity

Figure 4 Diversity curves comparing distinct subsets of pleurodiran taxa Circles and squares identify phyletic or taxic diversity curvesrespectively The orange and yellow bands highlight the CretaceousndashPalaeogene and EocenendashOligocene mass extinctions respectively

Table 1 Pairwise comparison of the results of nested models using the lsquooriginalrsquo tree

alternative model LnL df null model LnL df likelihood-ratio test p

DEC-M1 minus1917 3 DEC-M0 minus2093 2 30times 10minus9

DIVALIKE-M1 minus2035 3 DIVALIKE-M0 minus2313 2 86times 10minus14

BAYAREA-M1 minus2459 3 BAYAREA-M0 minus2535 2 98times 10minus5

33 Biogeographical analysisThe LRT showed that M2 models were the best fitted across all three alternative topologies (table 1see the electronic supplementary material) AICc model selection supports DIVALIKE-M2 in all cases(although DEC-M2 also performs well) stressing the power of cladogenetic range changes ie vicarianceand jump-dispersal (table 2 [55]) in explaining our data (see the electronic supplementary material) Itis noteworthy that the inclusion of x as a free parameter improved model fit (table 1) supporting that thedistance between the considered areas distinctly affects the pleurodiran lineages Also the inclusion of jas a free parameter consistently improved the fit in all models and topologies suggesting that founder-event speciation is important to account for range changes in side-necked turtles

Except for three nodes ie Pan-Pleurodira Pleurodira and Platychelyidae widespread ancestors(ie ancestor occupying more than one defined area) are never favoured in our M2 models (seethe electronic supplementary material) The AfricanAustralian ancestral range for the Pleurodiranode is best supported (although with low probability figure 2) whereas for Pan-Chelidae anAustralian range is favoured The most probable ancestral area for Pan-Pelomedusoides and Pan-Pelomedusidae is Africa whereas the ancestors of Pan-Podocnemididae probably dispersed to SouthAmerica during the Early Cretaceous (figure 2) Other South AmericaAfrica pairs are found duringthis period for example in Araripemydidae Euraxemydidae Podocnemidoidea Cearachelyini andHamadachelys + other Podocnemidoidae (figures 2 and 3) and two also during the Late Cretaceous iePodocnemididnae + Erymnochelyinae and the clade including Dacquemys and Stupendemys (figure 3)However in the best fitted model none of these joint distributions is favoured making it unlikely that thedescendant ranges are a result of vicariance [3555] The most probable ancestral area of bothremydidsis Africa and several dispersal events occurred from there eg to India (Sankuchemys + Kurmademys)Madagascar (Kinkonychelys rogersi) South America (Cearachelyini) and Europe (Bothremydini)

Table 2 Summary of biogeographical stochastic mapping counts for Pleurodira using DIVALIKE-M1 and -M2 models showing the meanand standard deviations (sd) of different types of events estimated by those models

mode type mean (sd) mean (sd)

dispersal range expansions 281(10) 220 0 (0) 0

range contractions 0 (0) 0 0 (0) 0

founder events 0 (0) 0 289 (09) 292

within-area speciation sympatry 719 (12) 565 683 (09) 690

subset speciation 0 (0) 0 0 (0) 0

vicariance vicariance 271 (12) 213 18 (05) 18

The summary of the BSM counts (table 2) shows that founder-event and vicariance representrespectively 292 and 18 of all events under the DIVALIKE-M2 On the other hand under DIVALIKE-M1 vicariance accounts for 213 All kinds of dispersal events (range expansion and founder events)represent 221 and 292 under DIVALIKE-M1 and -M2 models respectively The summary of BSMcounts for DEC-M1 and -M2 models provides similar results (see the electronic supplementary material)

4 Discussion41 Pleurodiran relationshipsAs previously mentioned the analysis conducted here represents the largest (101 taxa245 characters)and most inclusive phylogenetic study so far to our knowledge conducted for side-necked turtlesThe resulting arrangement of the main lineages generally agrees with previous analyses (eg [12ndash142123]) with some exceptions that reveal taxa with unstable relationships The major extant (ChelidaePelomedusidae Podocnemididae) and extinct (Araripemydidae Euraxemyididae Bothremydidae)lineages were mainly retrieved with relations similar to those previously proposed although somediscrepancies are seen

The long-necked chelids are united into the clade Chelina (for phylogenetic definitions see theelectronic supplementary material) as in previous morphology-based studies [172930] whereasmolecular data studies suggest monophyletic Australasian and South American chelids instead[524ndash28] Although seemingly more intuitive (but see below the discussion about ancestral areareconstructions) this result could not be replicated in any morphological study even with the expansionof character and taxon samples seen here as well as in other studies [293066] Indeed based on our datamatrix constraining South American and Australasian chelid clades resulted in 72 trees of 1162 stepsmuch longer than the original MPTs Nine synapomorphies support the long-necked Chelina clade andonly half of those are related to the cervical vertebrae Thus the morphological support for this cladeis not exclusively related to their long necks On the other hand more recent molecular analyses haveincreased the sample of chelid taxa and gene sequences with unchanging results [525] Indeed thereseems to be currently no objective reason to choose between those alternatives

The position of Araripemydidae and Euraxemydidae (figure 1) forming the sister clade toPelomedusoides (sensu [1]) agrees with a proposal by Meylan [21] that has never been replicated sincethen Euraxemydids were retrieved inside Pelomedusoides in all other analyses and Araripemydidaehas been alternatively placed outside that group [1223] inside it and closer to Podocnemididae [867]closer to Pelomedusidae [13] or even on the stem lineage of Pleurodira [14] Forcing Euraxemydidaecloser to Podocnemididae than to Araripemydidae or Pelomedusidae resulted in a tree only two stepslonger (1131 steps) Similarly forcing Araripemydidae inside Pelomedusoides alternatively closer toPelomedusidae or Podocnemididae resulted in trees with 1133 and 1131 steps respectively Thusalthough the position of Araripemydidae and Euraxemydidae in our strict consensus tree (figure 1) iswell supported forcing previous hypotheses does not result in much longer trees

The position of Atolchelys lepida the oldest known crown Pleurodira is also controversial It wasretrieved inside Bothremydidae [8] or inside a clade with euraxemydids and Sokatra antitra [14]Conversely our results support both S antitra and A lepida as Pan-Podocnemididae (figure 1) but forcing

A lepida inside Bothremydidae requires only one additional step Indeed new data are needed to betterevaluate the relations of this taxon

Among Podocnemididae alternative arrangements have distinct biogeographical implications FirstCaninemys tridentata was retrieved as the sister taxon to the South American Cerrejonemys + Podocnemisclade [68] in contrast to previous analyses that supported an Erymnochelyinae affinity [131441]Another point of divergence between molecular and morphological datasets is seen for the extantPodocnemis Erymnochelys and Peltocephalus Molecular analyses [520252869] support the arrangementPeltocephalus + (Erymnochelys + Podocnemis) whereas morphological data (including that presentedhere) suggest that Erymnochelys and Peltocephalus are closer to one another than to Podocnemis [13ndash15213738416768] In almost all morphological analyses Erymnochelys and Peltocephalus are sister taxawhereas here Peltocephalus is sister to Stereogenyini and Erymnochelys is nested in a clade including otherAfrican and Europeans taxa (see also [1568]) A close relationship between Erymnochelys Turkanemysand Kenyemys was already suggested before (united in the so-called Erymnochelys group [70ndash74]) buthas never been recovered in a phylogenetic analyses In addition Neochelys and Papoulemys were forthe first time also included into that clade (figure 1) This arrangement has important implications forbiogeographic analyses as Erymnochelys is positioned closer to an AfricanndashEuropean rather than to aSouth American group

42 Diversification of side-necked turtlesMolecular divergence time estimates suggest that the evolutionary history of Pleurodira began in theLate Jurassic between 165 and 150 Ma [4575] but the oldest unequivocal records of the group comefrom the Early Cretaceous (Barremian ca 125 Ma) of Brazil [78] The phyletic diversity in the last stagesof the Early Cretaceous is low in comparison to that of the Late Cretaceous and Cenozoic (figure 4)As suggested for tetrapods in general [76] poor sampling might explain the low Early Cretaceousphyletic diversity of Pleurodira especially taking into account the much older molecular clock estimatesfor the origin of the group Nevertheless the results of our analysis indicate that the two mainpleurodiran lineages Pan-Chelidae and Pan-Pelomedusoides did not experience diversification ratesthat significantly differ from one another However the ghost lineages associated to Early Cretaceoustaxa indicate that they were already established during this period (figures 2 and 3)

Two diversification shifts have been recognized in the Late Cretaceous (time bin 2) one related to thePodocnemidoidea clade and another for a clade within Bothremydidae Because of the poorly sampledEarly Cretaceous fossil record the shifts detected for time bin 2 could be artefacts Indeed the LateCretaceous fossil record of Podocnemidoidea is richer than that of other pleurodiran lineages includingChelidae However the phylogenetic position of Early Cretaceous taxa such as Prochelidella cerrobarcinaeand Araripemys barretoi indicates that a large number of non-podocnemidoidean Pelomedusoides andChelidae lineages were already present in the Early Cretaceous (figures 2 and 3) Accordingly as shiftsare related to the balance of the tree the appearance of further Podocnemidoidea lineages duringthe Late Cretaceous is interpreted as indicating a higher rate of diversification of this group relatedto other pleurodiran lineages rather than as an artefact caused by the poor Early Cretaceous fossilrecord In any case this diversification shift can also be associated with the real diversification ofCearachelyini + Bothremydini + Taphrosphyini (figure 2)

The CretaceousPalaeogene boundary does not seem to correspond to a critical period of Pleurodiraextinction or diversification (figure 4) With a few exceptions (eg Araripemyidae and Galianemys) mostlineages crossed the CretaceousPalaeogene boundary (figures 2 and 3) albeit with a phyletic diversityreduction in some groups (eg Bothremydini) Furthermore the lack of additional shifts in the Paleoceneshows that the extinction of some lineages in the Late Cretaceous was not a trigger for diversificationsin the first stage of the Cenozoic Finally the CretaceousndashPalaeogene mass extinction does not seem toaffect the diversity of Pleurodira Bothremydids experienced a decrease in diversity in the Mesozoic toCenozoic passage but this was already dropping after the Campanian and the trend continued until theSelandian (figure 4) suggesting that other factors were related to this decrease Our results add to thoseon the North American turtle fauna (eg [77ndash79]) supporting the hypothesis that turtles were not muchaffected by the end-Cretaceous mass extinction

The shift recovered for the clade Peiropemydidae + Podocnemididae in the Eocene is here interpretedas a real pattern rather than an artefact caused by the extinction of lineages in previous intervalsespecially given the relatively rich fossil record of bothremydids in the early Palaeogene ThePodocnemidoidea fossil record (approx 20ndash30 known taxa) during the Paleocene and Eocene is alsomuch richer than that of Chelidae (less than 5 taxa) However only few lineages reach the end of the

Eocene and there is so far no record of pleurodires in the first stage of the Oligocene (Rupelianmdashca34 Ma figure 4) Yet it is not possible to assert if this is related to sampling biases or to the EocenendashOligocene extinction event [80] and the reduction of phyletic diversity in the Eocene associated to thetemperature decrease during the Eocene (eg [81]) The phyletic diversity curve however suggests thatpleurodires were not much affected by this event either (figure 4)

Two diversification shifts were detected in the Miocene time bin one related to Chelini and anotherone associated with a clade within Stereogenyini The fossil record of Pleurodira in the second stageof the Oligocene (Chattian ca 25 Ma) is solely composed of stereogenyins These Oligocene taxa andtheir respective ghost lineages have an influence in the tree balance of Stereogenyini in the Oligocenetime bin but no shift is detected in this interval On the other hand the subgroup of Stereogenyini forwhich the shift in the Miocene was detected (the least inclusive clade containing Brontochelys gaffneyiand Bairdemys thalassica) is majorly composed of Miocene representatives and ghost lineages from thePliocene + Recent interval Thus even with the appearance of lineages of Stereogenyini in the Oligocenethe further diversification (ie changing in tree balance) of this subclade was still detected as a shift in ouranalysis and is thus here interpreted as a real shift Regarding the shift in the Miocene associated withChelini there are no Chelini taxa known from the Oligocene in a way that the tree balance of Chelinishows no alteration between time bins 4 and 5 Eocene and Oligocene Thus the Miocene shift shouldhave been seen with caution because of the low phyletic diversity of the group in previous stages whichmight be related to sampling biases However there is a second aspect which indicates that the shiftassociated to this group is an artefact The clade for which the shift was recognized also encompasses apart of the tree containing a great number of older representatives of Chelini that were already extinct bythe Miocene (eg Bonapartemys Prochelidella Chelodina alanrixi) Thus this shift is here understood as anartefact Regarding the shift associated to Chelidae in time bin 7 there is also the presence of Chelidaetaxa that were already extinct by the Pliocene + Recent interval However most of these also composethe Chelidae clade in the preceding time bin Miocene for which no shift was detected Thus we hereinterpret this as a real shift associated to the radiation of chelid modern lineages (ie Myuchelys ElseyaAcanthochelys)

43 Dispersal not vicariance nor large-scale extinctions drove the distribution of pleurodiresPrevious studies dealing with the biogeographic history of pleurodires (eg [51920326370]) sufferedfrom two main drawbacks First except for that of Joyce et al [63] they were conducted when fewerfossil taxa were known or did not include fossil taxa at all [5] For example the studies of Noonan [20]and Romano amp Azevedo [22] included only eight and six fossil taxa respectively Now many more fossilpleurodires are known (eg [812ndash1582]) providing a more detailed account of past distribution patterns(eg [1683]) and more accurate age for the cladogenetic events which largely influence the results ofbiogeographical reconstructions [3384]

The second drawback is related to the theoretical and methodological turnover in historicalbiogeographic analyses from a search for common patterns and causes to the reconstruction of ancestralranges with a broader consideration of several evolutionary processes the event-based approaches[34355557] The latter approach became dominant in the last two decades [55] assigning coststo evolutionary processesmdasheg vicariance dispersal and extinctionmdashand considering biogeographicpatterns that minimize such total cost as optimal solutions [34] Also the advent of parametricapproaches allowed the incorporation of information other than only distribution and topology suchas branch lengths and distance between areas into complex models [5345561] Finally implementationof those models in a common likelihood framework enabled the choice of best fitted models and the testof competing biogeographic scenarios [5560]

Our analyses employed those recent methodological advancements on an up-to-date inventory offossil taxa and the results do not support either of the previously proposed explanatory hypothesesfor the geographical distribution of pleurodires By contrast dispersal events are the dominant processunderlying range changes under the best model (tables 1 and 2) This result is clearly dependent onthe model under M1 models (that exclude founder events) vicariance accounts for a similar proportionof range changes (table 2 see the electronic supplementary material for additional results) This isexpected because every model makes explicit assumptions about biogeographic processes and it hasbeen shown that they largely affect ancestral area reconstructions [55] Fortunately the implementationunder a common framework on BioGeoBEARS allows the comparison of competing models based onLRT to objectively choose the best fitted for each dataset [60] DIVALIKE-M2 is clearly the best model forour data (table 1) so that we consider its reconstructions (figures 2 and 3) as the best representation

of pleurodiran biogeographic history Also the BSM shows that dispersal events (including rangeexpansion and founder events) occurred more commonly from South America to Africa and fromAfrica to South America Europe and North America This holds true even for M1 models (see theelectronic supplementary material) reinforcing the high frequency of movements between those areasfor pleurodires

It is evident from the fossil record that at least during the Cretaceous pan-chelids and pan-pelomedusoids were restricted to southern and northern Gondwana respectively [12313282] eventhough we did not define those as distinct areas for the analyses We agree with Joyce et al [63]that a desert zone the Botucatu Desert [85] is the more likely barrier preventing the expansion ofthe southern Pan-Chelidae and the northern Pan-Pelomedusoides to one anotherrsquos areas because ourhypothesis suggests that those lineages were already separated prior to the formation of the Paranaacute-Etendeka Volcanic Province (ca 137ndash127 Ma [86]) considered a possible barrier by Romano amp Azevedo[22] remaining separated longer than the duration of that event It is well recorded that in the northernpart of the Paranaacute Basin desert conditions continued to prevail after the volcanic event as indicatedby the aeolian sandstones of the Caiuaacute Group [87] The ancestral area reconstructions using the lsquooriginalrsquotree support an Australian origin for pan-chelids which dispersed to South America still during the EarlyCretaceous Chelodina ancestors diverged from this South American lineage dispersing back to Australiastill during the Early Cretaceous (figure 5) The lsquomolecular constrainedrsquo analyses also support a SouthAmerican origin for Pan-Chelidae (see the electronic supplementary material) and all dispersal eventsbetween this area and Australia are also reconstructed to the Early Cretaceous Although the distributionpatterns of several other groups including meiolaniid turtles [638889] suggest that southern SouthAmerica and Australia remained connected via Antarctica through the entire Cretaceous [90] our datasuggest that by the end of the Early Cretaceous some kind of barrier prevented the dispersal of chelidturtles between those areas The southward movement of the Antarctic continent reaching latitudeshigher than 70degS during the Aptian [91] and decreasing temperatures after the Cenomanian [92] mayhave been some factors limiting the presence of chelids in that continent because turtles are rarer at suchhigh latitudes and lower temperatures [6392]

The analyses of the lsquooriginalrsquo or lsquomolecular constrainedrsquo trees do not result in drastic changes inchelid ancestral area reconstructions (see the electronic supplementary material) It was previouslysupposed that chelids had diversified prior to the break-up of southern Gondwana if results delivered bymolecular analyses were favoured or that lsquoextensive dispersalrsquo occurred more recently when favouringthe morphology-based topologies [82] However as fossil chelids are found deeply nested inside Chelina(figure 1) the divergence times between Australian and South American clades have been pushed backto the Early Cretaceous and only two dispersal events are needed to explain their current geographicaldistribution both occurring prior to the separation of those continents (figure 2)

The biogeographic history of Pan-Pelomedusoides by contrast was dominated by the occurrence ofareas of endemism for each clade with several dispersal events to other areas (figure 5) The exceptionis Pelomedusidae which was always endemic to continental Africa Currently some pelomedusids arefound in Madagascar the Arabic peninsula the Seychelles archipelago and other small islands [993]but the absence of fossil records other than very scarce and fragmentary remains in continental Africa[72] preclude a more detailed account of the Pelomedusidae biogeographic history Given the currentdata we hypothesize that Pan-Pelomedusidae were always restricted to the African continent and onlyrecently dispersed transoceanically to those other areas

Our results also show that the most recent ancestors of Araripemydidae Euraxemydidae andPan-Podocnemididae originally inhabited Africa dispersing to South America during the EarlyCretaceous (figure 5) The ancestors of Podocnemidoidae remained in the latter region whereas thoseof Bothremydidae returned to the African continent Bothremydids greatly diversified in this region(figure 2) but several taxa dispersed independently to other areas at least once to Europe IndiaMadagascar and back to South America and at least three times to North America (figure 5) Our resultshighlight the great dispersion capability of bothremydids in accordance to the inferred marine or littoralhabits of these pleurodires [164763] Bothremydidae was the most widespread group of side-neckedturtles during the Cretaceous and Paleocene when they started to decline in diversity until their completeextinction by the Ypresian (figure 4)

The Podocnemidoidae have also been more widespread during some periods (eg Paleocene toMiocene) but our results show that as bothremydids they were mainly restricted to a few areas fromwhich they dispersed to others (figure 5) The group was initially endemic to South America but theancestors of Hamadachelys escuilliei and Erymnochelyinae dispersed to Africa respectively during theEarly and Late Cretaceous The latter group greatly diversified in this continent dispersing twice to

PelomedusidaePan-Podocnemididae

Chelidae

Araripemydidae

Podocnemidoidae Euraxemydidae

StereogenyiniBothremydidae

10 myr

35 myr

70 myr

105 myr

Chelidae

Araripemydidae

10 myr

35 myr

70 myr

105 myr

Figure 5 Palaeomaps summarizing the main dispersal events (arrows) of different pleurodiran groups (circles)

Europe and once back to South America during the Paleocene (figure 5) The latter event originatedPeltocephalus dumerilianus and Stereogenyini which returned to Africa later There are however almostequal probabilities for a dispersal event including only the ancestor of P dumerilianus (figure 3) Inthis context as with bothremydids and also in accordance to their inferred marine lifestyles [159495]stereogenyins would have dispersed several times out of Africa to North America South America andEast Asia (figure 5)

Even excluding the marinelittoral adapted bothremydids and stereogenyins our results suggestseveral events of transoceanic dispersal during pleurodiran biogeographic history Except for thoselineages which undertook long distance oceanic dispersals (eg from Africa to North America orEast Asia) all the other events occurred across relatively short distances Africa and South Americanexchanges do not occur after the Paleocene (except for Stereonyini) and other events are short dispersalsfrom Africa to Europe or Madagascar This hints at the possibility that even not usually occupyingbrackish waters pleurodires could tolerate transoceanic crossings maybe carried out or rafted by oceancurrents Oceanic dispersals may have been more common causes of biogeographic range changesthan previously thought as already suggested for other groups such as tortoises [532] lizards [96]amphibians [62] and even invertebrates [97] For pleurodires this should not come as a surprisegiven the island distribution of some pelomedusids and chelids [993] and the results of a recentstudy [11] in which Chelodina expansa and Emydura macquarii individuals were exposed to salineconditions for long periods (50 days) without showing physiological problems Finally consideringthat stem-pleurodires could also have a certain tolerance to salty waters [6] this could be a morewidespread feature in the group that could have facilitated the origin of groups more adapted tomarine environments

5 ConclusionOur phylogenetic hypothesis is to our knowledge the most inclusive and well-sampled ever proposedfor extinct pleurodiran turtles Although pleurodiran relationships can be said to be stable owing to thegeneral agreement between different hypotheses some points of conflict still exist especially betweenmorphology and molecular data Our diversification analysis suggests that pleurodires were not muchaffected by the CretaceousndashPalaeogene and EocenendashOligocene mass extinction events This result agreeswith patterns observed for other testudine lineages and may represent a general trend for turtles

Pelomedusoid extinct subclades show a greater number of diversification shifts in relation to Chelidaelineages including the diversification shift related to bothremydids in the Early Cretaceous the shiftassociated to the clade composed by Peiropemydidae + Podocnemididae in the Late Cretaceous andthe one related to the clade within Stereogenyini in the Oligocene Freshwater pleurodires apparentlyexperienced steady diversification rates whereas marine taxa peaked during the Late Cretaceousand OligocenendashMiocene periods (figure 4) In this sense most pleurodire diversification shifts can beassociated with the invasion of different niches eg bothremydids and stereogenyins invading littoral ormarine environments

The current distribution of pleurodires cannot be fully understood using vicariance or large-scaleextinctions as sole explanations Although those may have affected the observed patterns dispersalevents seem to be the most important factor shaping the biogeography of side-necked turtles (table 2)Even though most dispersals occurred across relatively short distances long distance dispersals werealso common especially among the littoralmarine-adapted bothremydids and stereogenyins (figure 5)As such our hypothesis adds to recent results (eg [629697]) indicating that transoceanic dispersals maybe a much more common biogeographic event than previously thought

Data accessibility The input files for the biogeographic analyses as well as the scripts used are available through GitHubRepository httpsgithubcomgsferreirabiopleurodira-biogeogitAuthorsrsquo contributions GSF conceived the ideas GSF MB and JS collected the data GSF JS MCL and MBanalysed the data GSF and MB led the writing with revisions from MCL and JS All co-authors read and gavetheir final approval for the articleCompeting interests The authors declare no competing interestsFunding This work was supported by Fundaccedilatildeo de Amparo agrave Pesquisa do Estado de Satildeo Paulo (grant nos 201603934-2 and 201425379-5 to GSF and 201403825-3 to MCL) Conselho Nacional de Desenvolvimento Cientiacutefico eTecnoloacutegico (CNPq)mdashCiecircncia sem Fronteiras (grant 2466102012-3 to MB) and grant PICT 2013 -0095 Preacutestamo BIDto JSAcknowledgements We thank EAB Almeida I Werneburg and M Rabi for discussions and suggestions that contributedto this project and N Matzke for support with BioGeoBEARS as well as the reviewers E Cadena and P Upchurch fortheir insightful reviews and the editors of Royal Society open science J B Desojo and K Padian

References1 Joyce WG Parham JF Gauthier JA 2004 Developing

a protocol for the conversion of rank-based taxonnames to phylogenetically defined clade names asexemplified by turtles J Paleontol 78 989ndash1013(doi1016660022-3360(2004)078lt0989DAPFTCgt20CO2)

2 Gaffney ES 1988 A cladogram of the pleurodiranturtles Acta Zool Cracov 31 487ndash492

3 Gaffney ES Meylan PA 1988 A phylogeny of turtlesIn The phylogeny and classification of the tetrapodsvol 1 amphibians reptiles birds (ed MJ Benton)pp 157ndash219 Oxford UK Clarendon Press

4 Joyce WG Parham JF Lyson TR Warnock RCMDonoghue PCJ 2013 A divergence dating analysis ofturtles using fossil calibrations an example of bestpractices J Paleontol 87 612ndash634(doi10166612-149)

5 Pereira AG Sterli J Moreira FRR Schrago CG 2017Multilocus phylogeny and statistical biogeographyclarify the evolutionary history of major lineages ofturtlesMol Phylogenet Evol 113 59ndash66(doi101016jympev201705008)

6 Cadena E Joyce WG 2015 A review of the fossilrecord of turtles of the clades platychelyidae anddortokidae B Peabody Mus Nat Hist 56 3ndash20(doi1033740140560101)

7 Ferreira GS Langer MC 2013 A pelomedusoid(Testudines Pleurodira) plastron from the lowerCretaceous of Alagoas Brazil Cretaceous Res 46267ndash271 (doi101016jcretres201310001)

8 Romano PSR Gallo V Ramos RRC Antonioli L 2014Atolchelys lepida a new side-necked turtle from theEarly Cretaceous of Brazil and the age of crownPleurodira Biol Lett 10 20140290 (doi101098rsbl20140290)

9 Rhodin AGJ Iverson JB Bour R Fritz U Georges AShaffer HB van Dijk PPJ 2017 Turtles of the worldannotated checklist and atlas of taxonomysynonymy distribution and conservation status(8th edn) In Conservation biology of freshwaterturtles and tortoises a compilation project of theIUCNSSC tortoise and freshwater turtle specialistgroup vol 7 (eds AGJ Rhodin JB Iverson PP vanDijk RA Saumure KA Buhlmann PCH Pritchard RAMittermeier) pp 1ndash292 Chelonian ResearchMonographs New York NY Chelonian ResearchFoundation and Turtle Conservancy

10 Pritchard PCH Trebbau P 1984 The turtles ofVenezuela Oxford OH Society for the Study ofAmphibians and Reptiles

11 Bower DS Scheltinga DM Clulow S Clulow JFranklin CE Georges A 2016 Salinity tolerances oftwo Australian freshwater turtles Chelodinaexpansa and Emydura macquarii (TestudinataChelidae) Conserv Physiol 4 cow042(doi101093conphyscow042)

12 Gaffney ES Tong H Meylan PA 2006 Evolution ofthe side-necked turtles the familiesBothremydidae Euraxemydidae andAraripemydidae B Am Mus Nat Hist 300 1ndash698(doi1012060003-0090(2006)300[1EOTSTT]20CO2)

13 Gaffney ES Meylan PA Wood RC Simons E DeAlmeida Campos D 2011 Evolution of theside-necked turtles the family podocnemididae BAm Mus Nat Hist 350 1ndash237 (doi1012063501)

14 Cadena E 2015 A global phylogeny ofPelomedusoides turtles with newmaterial ofNeochelys franzeni Schleich 1993 (TestudinesPodocnemididae) from the middle Eocene MesselPit of Germany PeerJ 3 e1221 (doi107717peerj1221)

15 Ferreira GS Rincoacuten AD Soloacuterzano A Langer MC2015 The last marine pelomedusoids (TestudinesPleurodira) a new species of Bairdemys and thepaleoecology of Stereogenyina PeerJ 3 e1063(doi107717peerj1063)

16 Joyce WG Lyson TR Kirkland JI 2016 An earlybothremydid (Testudines Pleurodira) from the LateCretaceous (Cenomanian) of Utah North AmericaPeerJ 4 e2502 (doi107717peerj2502)

17 Gaffney ES 1977 The side-necked turtle familyChelidae a theory of relationships using sharedderived characters Am Mus Novit 26201ndash28

18 Laurent RF 1979 Herpetofaunal relationshipsbetween Africa and South America In The SouthAmerican herpetofauna its origin evolution anddispersal (ed WE Duellman) pp 55ndash71 LawrenceKS Museum of Natural History University ofKansas Monograph 7

19 Maisey JG 1993 Tectonics the SantanaLagerstatten and the implications for lateGondwanan biogeography In Biologicalrelationships between Africa and South America (edP Goldblat) pp 435ndash454 New Haven CT YaleUniversity Press

20 Noonan BP 2000 Does the phylogeny ofpelomedusoid turtles reflect vicariance due tocontinental drift J Biogeogr 27 1245ndash1249(doi101046j1365-2699200000483x)

21 Meylan PA 1996 Skeletal morphology andrelationships of the early Cretaceous side-neckedturtle Araripemys barretoi (TestudinesPelomedusoides Araripemydidae) from theSantana Formation of Brazil J Vertebr Paleontol16 20ndash33 (doi10108002724634199610011280)

22 Romano PSR Azevedo SAK 2006 Are extantpodocnemidid turtles relicts of a widespreadCretaceous ancestor S Am J Herpetol 1 175ndash184(doi1029941808-9798(2006)1[175AEPTRO]20CO2)

23 de La Fuente MLS 2003 Two new pleurodiran turtlesfrom the Portezuelo Formation (Upper Cretaceous)of northern Patagonia Argentina J Paleontol 77559ndash575 (doi101017S0022336000044243)

24 Georges A Birrell J Saint KM McCord WPDonnellan SC 1998 A phylogeny for side-neckedturtles (Chelonia Pleurodira) based onmitochondrial and nuclear gene sequence variationBiol J Linn Soc 67 213ndash246 (doi101111j1095-83121999tb01862x)

25 Guillon JM Gueacutery L Hulin V Girondot M 2012 Alarge phylogeny of turtles (Testudines) usingmolecular data Contrib Zool 81 147ndash158

26 Krenz JG Naylor GJP Shaffer HB Janzen FJ 2005Molecular phylogenetics and evolution of turtlesMol Phylogenet Evol 37 178ndash191 (doi101016jympev200504027)

27 Seddon JM Georges A Baverstock PR McCord W1997 Phylogenetic relationships of Chelid turtles(Pleurodira Chelidae) based on mitochondrial 12S

rRNA gene sequence variationMol PhylogenetEvol 7 55ndash61 (doi101006mpev19960372)

28 Sterli J 2010 Phylogenetic relationships amongextinct and extant turtles the position of Pleurodiraand the effects of the fossils on rootingcrown-group turtles Contrib Zool 79 93ndash106

29 Bona P de La Fuente MS 2005 Phylogenetic andpaleobiogeographic implications of Yaminuechelysmaior (Staesche 1929) new comb a largelong-necked chelid turtle from the early Paleoceneof Patagonia Argentina J Vertebr Paleontol 25569ndash582 (doi1016710272-4634(2005)025[0569PAPIOY]20CO2)

30 de la Fuente MS Maniel I Jannello JM Sterli J RigaBG Novas F 2017 A new large panchelid turtle(Pleurodira) from the Loncoche Formation (upperCampanian-lower Maastrichtian) of the MendozaProvince (Argentina) Morphologicalosteohistological studies and a preliminaryphylogenetic analysis Cretaceous Res 69 147ndash168(doi101016jcretres201609007)

31 de la Fuente MS Umazano AM Sterli J CarballidoJL 2011 New chelid turtles of the lower section ofthe Cerro Barcino formation (Aptian-Albian)Patagonia Argentina Cretaceous Res 32 527ndash537(doi101016jcretres201103007)

32 de la Fuente MS Sterli J Maniel I 2014 Originevolution and biogeographic history of SouthAmerican turtles Cham Switzerland SpringerInternational Publishing

33 Meseguer AS Lobo JM Ree R Beerling DJSanmartiacuten I 2014 Integrating fossils phylogeniesand niche models into biogeography to revealancient evolutionary history the case of Hypericum(Hypericaceae) Syst Biol 64 215ndash232(doi101093sysbiosyu088)

34 Ree RH Moore BR Webb CO Donoghue MJ 2005 Alikelihood framework for inferring the evolution ofgeographic range on phylogenetic trees Evolution59 2299ndash2311 (doi10155405-1721)

35 Ronquist F Sanmartiacuten I 2011 Phylogenetic methodsin biogeography Ann Rev Ecol Evol Syst 42441ndash464 (doi101146annurev-ecolsys-102209-144710)

36 Maddison WP Maddison DR 2017Mesquite amodular system for evolutionary analysis Version331 See httpmesquiteprojectorg

37 Cadena EA Bloch JI Jaramillo CA 2010 Newpodocnemidid turtle (Testudines Pleurodira) fromthe middlendashupper Paleocene of South America JVertebr Paleontol 30 367ndash382 (doi10108002724631003621946)

38 Cadena EA Bloch JI Jaramillo CA 2012 Newbothremydid turtle (Testudines Pleurodira) fromthe Paleocene of northeastern Colombia JPaleontol 86 688ndash698 (doi10166611-128R11)

39 Cadena EA Ksepka DT Jaramillo CA Bloch JI 2012New pelomedusoid turtles from the late PalaeoceneCerrejoacuten Formation of Colombia and theirimplications for phylogeny and body size evolutionJ Syst Palaeontol 10 313ndash331 (doi101080147720192011569031)

40 Dumont Jr MV 2013 Um novo podocnemiacutedeo foacutessilde grande porte da Formaccedilatildeo Solimotildees(Mioceno-Plioceno) Acre Brasil e as relaccedilotildeesfilogeneacuteticas entre os Podocnemidae Unpublished

MSc dissertation Universidade de BrasiacuteliaBrasilia Brazil

41 Meylan PA Gaffney ES De Almeida Campos D 2009Caninemys a new side-necked turtle(Pelomedusoides Podocnemididae) from themiocene of Brazil Am Mus Novit 3639 1ndash26(doi1012066081)

42 Peacuterez-Garciacutea A de Lapparent de Broin F 2013 A newspecies of Neochelys (Chelonii Podocnemididae)from the Ypresian (Early Eocene) of the south ofFrance Comptes Rendus Palevol 12 269ndash277(doi101016jcrpv201305011)

43 Thomson S Georges A 2009Myuchelys gennovmdasha new genus for Elseya latisternum andrelated forms of Australian freshwater turtle(Testudines Pleurodira Chelidae) Zootaxa 205332ndash42

44 Goloboff PA Farris JS Nixon KC 2008 TNT a freeprogram for phylogenetic analysis Cladistics 24774ndash786 (doi101111j1096-0031200800217x)

45 Goloboff PA Farris JS Kaumlllersjouml M Oxelman BRamıacuterez MJ Szumik CA 2003 Improvementsto resampling measures of group support Cladistics19 324ndash332 (doi101111j1096-00312003tb00376x)

46 Pol D Escapa IH 2009 Unstable taxa in cladisticanalysis identification and the assessment ofrelevant characters Cladistics 25 515ndash527(doi101111j1096-0031200900258x)

47 Rabi M Tong H Botfalvai G 2011 A new species ofthe side-necked turtle Foxemys (PelomedusoidesBothremydidae) from the Late Cretaceous ofHungary and the historical biogeography of theBothremydini Geol Mag 149 662ndash674(doi101017s0016756811000756)

48 R Core Team 2014 R a language and environment forstatistical computing Vienna Austria R Foundationfor Statistical Computing See httpwwwr-projectorg

49 Bell MA Lloyd GT 2014 strap an R package forplotting phylogenies against stratigraphy andassessing their stratigraphic congruencePalaeontology 58 379ndash389 (doi101111pala12142)

50 Brusatte SL Benton MJ Ruta M Lloyd GT 2008 Thefirst 50 Myr of dinosaur evolutionmacroevolutionary pattern and morphologicaldisparity Biol Lett 4 733ndash736 (doi101098rsbl20080441)

51 Chan KMA Moore BR 2004 SYMMETREEwhole-tree analysis of differential diversificationrates Bioinformatics 21 1709ndash1710 (doi101093bioinformaticsbti175)

52 Tarver JE Donoghue PCJ 2011 The trouble withtopology phylogenies without fossils provide arevisionist perspective of evolutionary history intopological analyses of diversity Syst Biol 60700ndash712 (doi101093sysbiosyr018)

53 Ruta M Pisani D Lloyd GT Benton MJ 2007 Asupertree of Temnospondyli cladogenetic patternsin the most species-rich group of early tetrapodsProc R Soc B 274 3087ndash3095 (doi101098rspb20071250)

54 Bronzati M Montefeltro FC Langer MC 2015Diversification events and the effects of massextinctions on Crocodyliformes evolutionary historyR Soc open sci 2 140385 (doi101098rsos140385)

55 Matzke NJ 2013 Probabilistic historicalbiogeography newmodels for founder-eventspeciation imperfect detection and fossils allow

improved accuracy and model-testing FrontBiogeogr 5 242ndash248

56 Ree RH Smith SA Baker A 2008 Maximumlikelihood inference of geographic range evolutionby dispersal local extinction and cladogenesisSyst Biol 57 4ndash14 (doi10108010635150701883881)

57 Ronquist F Cannatella D 1997 Dispersal-vicarianceanalysis a new approach to the quantification ofhistorical biogeography Syst Biol 46 195ndash203(doi101093sysbio461195)

58 Landis MJ Matzke NJ Moore BR Huelsenbeck JP2013 Bayesian analysis of biogeography when thenumber of areas is large Syst Biol 62 789ndash804(doi101093sysbiosyt040)

59 Yu Y Harris AJ He X-J 2013 RASP (ReconstructAncestral State in Phylogenies) 21 beta See httpmnhscueducnsoftblogRASP

60 Matzke NJ 2014 Model selection in historicalbiogeography reveals that founder-eventspeciation is a crucial process in island cladesSyst Biol 63 951ndash970 (doi101093sysbiosyu056)

61 Van DamMH Matzke NJ 2016 Evaluating theinfluence of connectivity and distance onbiogeographical patterns in the south-westerndeserts of North America J Biogeogr 431514ndash1532 (doi101111jbi12727)

62 Pyron RA 2014 Biogeographic analysis revealsancient continental vicariance and recent oceanicdispersal in amphibians Syst Biol 63 779ndash797(doi101093sysbiosyu042)

63 Joyce WG Rabi M Clark JM Xu X 2016 A toothedturtle from the Late Jurassic of China and the globalbiogeographic history of turtles BMC Evol Biol 16236 (doi101186s12862-016-0762-5)

64 Matzke NJ 2016 See Stochastic mapping underbiogeographical models PhyloWiki BioGeoBEARSwebsite 2016 See httpphylowikidotcombiogeobearsstochastic_mapping

65 Dupin J Matzke NJ Saumlrkinen T Knapp S OlmsteadRG Bohs L Smith SD 2016 Bayesian estimation ofthe global biogeographical history of theSolanaceae J Biogeogr 44 887ndash899 (doi101111jbi12898)

66 de la Fuente M Maniel I Jannello J Sterli J GarridoA Garciacutea R Salgado L Canudo J Bolatti R 2017Unusual shell anatomy and osteohistology in a newLate Cretaceous panchelid turtle from northwesternPatagonia Argentina Acta Palaeontol Pol 62585ndash601 (doi104202app003402017)

67 Franccedila MAG Langer MC 2006 Phylogeneticrelationships of the Bauru Group turtles (LateCretaceous of south-central Brazil) Rev BrasPaleontol 9 1ndash9 (doi104072rbp2006101)

68 Menegazzo MC Bertini RJ Manzini FF 2015 A newturtle from the upper Cretaceous Bauru Group ofBrazil updated phylogeny and implications for ageof the Santo Anastaacutecio Formation J S Am EarthSci 58 18ndash32 (doi101016jjsames201412008)

69 Vargas-Ramiacuterez M Castantildeo-Mora O Fritz U 2008Molecular phylogeny and divergence times ofancient South American and Malagasy river turtles(Testudines Pleurodira Podocnemididae) OrgDivers Evol 8 388ndash398 (doi101016jode200810001)

70 Broin FD 1988 Les tortues et le Gondwana Examendes rapports entre le fractionnement du Gondwanaau Creacutetaceacute et la dispersion geacuteographique des

tortues pleurodires agrave partir du Creacutetaceacute StudPalaeocheloniol 2 103ndash142

71 Lapparent de Broin F 2000 African chelonians fromthe Jurassic to the present phases of developmentand preliminary catalogue of the fossil recordPalaeontol Africana 36 43ndash82

72 Lapparent de Broin F 2000 The oldestpre-Podocnemidid turtle (Chelonii Pleurodira)from the early Cretaceous Cearaacute state Brasil andits environment Treballs del Museu de Geologia deBarcelona 9 43ndash95

73 Peacuterez-Garciacutea A de Lapparent de Broin F 2015 Newinsights into the anatomy and systematic oflsquoPapoulemysrsquo laurenti a representative of Neochelys(Chelonii Podocnemididae) from the early Eoceneof the south of France Palaumlontol Z 89 901ndash923(doi101007s12542-015-0259-3)

74 Peacuterez-Garciacutea A Lapparent de Broin F Murelaga X2017 The Erymnochelys group of turtles (PleurodiraPodocnemididae) in the Eocene of Europe new taxaand paleobiogeographical implications PalaeontolElectron 20 1ndash28 (doi1026879687)

75 Shaffer HB McCartney-Melstad E Near TJ MountGG Spinks PQ 2017 Phylogenomic analyses of 539highly informative loci dates a fully resolved timetree for the major clades of living turtles(Testudines)Mol Phylogenet Evol 115 7ndash15(doi101016jympev201707006)

76 Benson RBJ Mannion PD Butler RJ Upchurch PGoswami A Evans SE 2013 Cretaceous tetrapodfossil record sampling and faunal turnoverimplications for biogeography and the rise ofmodern clades Palaeogeogr Palaeocl 372 88ndash107(doi101016jpalaeo201210028)

77 Lyson TR Joyce WG 2009 A revision of Plesiobaena(Testudines Baenidae) and an assessment ofbaenid ecology across the KT boundary JPaleontol 83 833ndash853 (doi10166609-0351)

78 Lyson TR Joyce WG Knauss GE Pearson DA 2011Boremys (Testudines Baenidae) from the latestCretaceous and early Paleocene of North Dakota an11-million-year range extension and an additionalKT survivor J Vertebr Paleontol 31 729ndash737(doi101080027246342011576731)

79 Holroyd PA Wilson GP Hutchison JH 2014 Temporalchanges within the latest Cretaceous and earlyPaleogene turtle faunas of northeastern MontanaGeol S Am S 503 299ndash312

80 Ivany LC Patterson WP Lohmann KC 2000 Coolerwinters as a possible cause of mass extinctions atthe EoceneOligocene boundary Nature 407887ndash890 (doi10103835038044)

81 Zachos JC Dickens GR Zeebe RE 2008 An earlyCenozoic perspective on greenhouse warming andcarbon-cycle dynamics Nature 451 279ndash283(doi101038nature06588)

82 Maniel IJ de la Fuente MS 2016 A review of thefossil record of turtles of the clade Pan-Chelidae BPeabody Mus Nat Hist 57 191ndash227(doi1033740140570206)

83 Ferreira GS Rincoacuten AD Soloacuterzano A Langer MC2016 Review of the fossil matamata turtles earliestwell-dated record and hypotheses on the origin oftheir present geographical distribution Sci Nat103 28 (doi101007s00114-016-1355-2)

84 Wood HM Matzke NJ Gillespie RG Griswold CE2012 Treating fossils as terminal taxa in divergencetime estimation reveals ancient vicariance patterns

in the palpimanoid spiders Syst Biol 62 264ndash284(doi101093sysbiosys092)

85 Assine ML Piranha JM Carneiro CDR 2004 Ospaleodesertos Piramboacuteia e Botucatu In Geologia docontinente Sul-americano evoluccedilatildeo da obra defernando flaacutevio marques de almeida (eds V MantessoNeto A Bartorelli CDR Carneiro BB Brito Neves) pp77ndash92 Satildeo Paulo Brazil Editora Beca

86 Renne PR Ernesto M Pacca IG Coe RS Glen JMPreacutevot M Perrin M 1992 The age of Paranaacute floodvolcanism rifting of Gondwanaland and theJurassic-Cretaceous boundary Science 258975ndash979 (doi101126science2585084975)

87 Fernandes LA de Castro AB Basilici G 2007Seismites in continental sand sea deposits of theLate Cretaceous Caiuaacute Desert Bauru Basin BrazilSediment Geol 199 51ndash64 (doi101016jsedgeo200512030)

88 Sterli J de la Fuente MS 2012 New evidence fromthe Palaeocene of Patagonia (Argentina) on theevolution and palaeo-biogeography ofMeiolaniformes (Testudinata new taxon name) J

Syst Palaeontol 11 835ndash852 (doi101080147720192012708674)

89 Sterli J de la Fuente MS Krause JM 2015 A newturtle from the Palaeogene of Patagonia(Argentina) sheds new light on the diversity andevolution of the bizarre clade of horned turtles(Meiolaniidae Testudinata) Zool J Linn Soc 174519ndash548 (doi101111zoj12252)

90 Sanmartiacuten I Ronquist F 2004 Southern hemispherebiogeography inferred by event-based modelsplant versus animal patterns Syst Biol 53 216ndash243(doi10108010635150490423430)

91 van Hinsbergen DJJ de Groot LV van Schaik SJSpakmanW Bijl PK Sluijs A Langereis CGBrinkhuis H 2015 A paleolatitude calculator forpaleoclimate studies PLoS ONE 10 e0126946(doi101371journalpone0126946)

92 Nicholson DB Holroyd PA Valdes P Barrett PM2016 Latitudinal diversity gradients in Mesozoicnon-marine turtles R Soc open sci 3 160581(doi101098rsos160581)

93 Fritz U Kehlmaier C Mazuch T Hofmeyr MD duPreez L Vamberger M Voumlroumls J 2015 Important new

records of Pelomedusa species for South Africa andEthiopia Vertebr Zool 65 383ndash389

94 Weems RE Knight JL 2013 A new species ofBairdemys (Pelomedusoides Podocnemididae)from the Oligocene (Early Chattian) ChandlerBridge Formation of South Carolina USAand its paleobiogeographic implications for thegenus InMorphology and evolution of turtles (edsDB Brinkman PA Holroyd JD Gardner)pp 289ndash303 Dordrecht The NetherlandsSpringer

95 Wood RC 1970 A review of the fossil Pelomedusidae(Testudines Pleurodira) of Asia Breviora 3571ndash24

96 Carranza S Arnold EN 2003 Investigating the originof transoceanic distributions mtDNA showsMabuya lizards (Reptilia Scincidae) crossed theAtlantic twice Syst Biodivers 1 275ndash282(doi101017s1477200003001099)

97 Toussaint EFA Taumlnzler R Balke M Riedel A 2017Transoceanic origin of microendemic and flightlessNew Caledonian weevils R Soc open sci 4 160546(doi101098rsos160546)

Introduction

Material and methods

Phylogenetic analyses and time-scaled trees

Diversification analysis and diversity curves

Estimation of ancestral ranges and number and types of biogeographical events

Results

Phylogenetic analyses

Diversification shifts and diversity levels

Biogeographical analysis

Discussion

Pleurodiran relationships

Diversification of side-necked turtles

Dispersal not vicariance nor large-scale extinctions drove the distribution of pleurodires

Conclusion

References