A likelihood inference of historical biogeography in the world's most diverse terrestrial vertebrate genus: Diversification of direct-developing frogs (Craugastoridae: Pristimantis) across the Neotropics

The geology of the northern Andean region has driven the evolutionary history of Neotropical fauna through the creation of barriers and connections that have resulted in speciation and dispersal events, respectively. One of the most conspicuous
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A likelihood inference of historical biogeography in the world’s mostdiverse terrestrial vertebrate genus: Diversification of direct-developingfrogs (Craugastoridae:  Pristimantis ) across the Neotropics Ángela María Mendoza a,b,c, ⇑ , Oscar E. Ospina a,d , Heiber Cárdenas-Henao a,e , Juan C. García-R  a,f  a Grupo de estudios en genética, ecología molecular y fisiología animal, Universidad del Valle, 760032 Cali, Colombia b Grupo de investigación en ecología y conservación neotropical, Samanea Foundation, 760046 Cali, Colombia c Conservation Genetics Laboratory, Instituto de Investigación de Recursos Biológicos Alexander von Humboldt, 6713 Cali, Colombia d Department of Biology, University of Puerto Rico, Rio Piedras 00931-3360, Puerto Rico e Departamento de Biología, Universidad del Valle, 760032 Cali, Colombia f  Institute of Agriculture and Environment, Massey University, Palmerston North, Private Bag 11-222, New Zealand a r t i c l e i n f o  Article history: Received 25 September 2014Revised 26 January 2015Accepted 1 February 2015Available online 11 February 2015 Keywords: Ancestral range reconstructionAndesBiogeographyDiversificationTerrarana a b s t r a c t The geology of the northern Andean region has driven the evolutionary history of Neotropical faunathrough the creation of barriers and connections that have resulted in speciation and dispersal events,respectively. One of the most conspicuous groups of anuran fauna in the Andes and surrounding areasis the direct-developing species of the genus  Pristimantis.  We investigated the molecular phylogeneticplacement of 12 species from the montane Andes of Colombia in a broader geographical context withanewgenus-levelphylogenyinordertoidentifytheroleofAndeanorogenyoverthelast40millionyearsand the effect of elevational differences in diversificationof   Pristimantis . We examined the biogeographichistory of the genus using ancestral range reconstruction by biogeographic regions and elevationalranges. We recognized the middle elevational band (between 1000 and 3000m) in the NorthwesternAndes region of Colombia andEcuador as a focal point for the srcin and radiation of   Pristimantis  species.Additionally, we found several Andean migrations toward new habitats in Central Andes and MeridaAndes for some species groups. We suggest that the paleogeological changes in the Northwestern Andeswerethemainpromoterofspeciationin Pristimantis ,andmayhaveservedasacorridorforthedispersionof lowland species.   2015 Elsevier Inc. All rights reserved. 1. Introduction High biodiversity in the Neotropics is a consequence of severalfactors driving genetic divergence (Beheregaray and Caccone,2007; Feral, 2002; Hewitt, 2004; Schluter, 2009). Prolonged geolo-gicalactivityyieldingcomplextopography(valleysandcordilleras)is thought to have caused vicariance events through one of themost species rich areas in the world (Antonelli et al., 2009;Brumfield and Edwards, 2007; Gentry, 1982; Kattan et al., 2004;Lynch and Duellman, 1997; Weir and Price, 2011).During thelast 50millionyears thelandscapeof the Neotropicshas been continuously modified. Prior to the uplift of the North-western Andes of Colombia and Ecuador in the early Paleocene,lowland forests likely extended from the Pacific to the Atlantic(Hooghiemstra and Van der Hammen, 1998). Uplift of the North-western Andes produced new vicariant barriers for many organ-isms and a multitude of new niches providing ecologicalspeciation opportunities (Antonelli et al., 2009; Chaves et al.,2011). Uplift of the Peruvian Central Andes between 5   and 15  Sduring the middle Miocene (Gregory-Wodzicki, 2000; Ramos andFolguera, 2009) created a dispersal barrier for lowland species aswell as new ecological space available for highland species. Plio-cene uplift of the Eastern Cordillera and the Ecuadorian Andeanregion (Coltorti and Ollier, 2000; Hoorn et al., 1995) marked aphase of separation occurring in the Pacific and Atlantic regionalbiotas. Simultaneous with the latter geological change, two newhabitat connections were made: the Northwestern cordillerasmerged with the Merida Andes, and the Central Andes with theNorthern Andes (Gregory-Wodzicki, 2000; Hooghiemstra and Vander Hammen, 1998). All these changes contribute to the geologicalcomplexity of the Andean region.   2015 Elsevier Inc. All rights reserved. ⇑ Corresponding author at: Conservation Genetics Laboratory, Instituto deInvestigación de Recursos Biológicos Alexander von Humboldt, Parque CientíficoAGRONATURA – CIAT, Km. 17, Recta Cali-Palmira, Colombia. E-mail address: (Á.M. Mendoza).Molecular Phylogenetics and Evolution 85 (2015) 50–58 Contents lists available at ScienceDirect Molecular Phylogenetics and Evolution journal homepage:  Several studies have illustrated the importance of topographicand geographic dynamics, and associated changes in environmen-tal conditions (e.g. temperature and precipitation) on the srcin of Neotropical diversity. They appear to have strongly influenced thedistribution of vertebrate species on each cordillera (e.g. Antonellietal.,2009;Grahametal.,2010;Hoornetal.,2010;Rull,2011).Forinstance, the complex geological history of the Neotropical Andesmay have significantly limited gene flow between populations,leading to allopatric and peripatric speciation in birds (e.g.Puebla-Olivares et al., 2008; Ruggiero and Hawkins, 2008), palms(e.g. Trénel et al., 2008), butterflies, rodents (e.g. Kattan et al., 2004), and amphibians (e.g. Santos et al., 2009; Castroviejo-Fisher et al., 2014).The amphibian communities in the Colombian, Ecuadorian andPeruvian Andean regions consist mostly of   Pristimantis  frogs in thefamily Craugastoridae (Lynch and Duellman, 1997; Pyron andWiens, 2011; Padial et al., 2014), the most diverse genus of terres-trial vertebrates. More than 420 species are currently described(AmphibiaWeb, 2005), which are characterized by terrestrialbreeding and direct embryonic development without a tadpolestage (Padial et al., 2014). By these characteristics, the reproduc-tive biology of   Pristimantis  is not associated with aquatic environ-ments and consequently can utilize all terrestrial environmentsthat have sufficient moisture for the survival of eggs, hatchlings,and adults (Lynch and Duellman, 1997). Consequently, this com-plex group can occupy a huge variety of microhabitats whilefacilitating their wide distribution (Heinicke et al., 2007). Todaythey occur through most of the Neotropics from Central America,throughout the North and Central Andes, Amazonia, and theGuyana forests and Cordillera de la Costa in Venezuela(AmphibiaWeb, 2005).Despite the prevalence of   Pristimantis  in community assem-blagesoftheAndeanfaunabyspeciescompositionandabundance,knowledge of the biogeographical history of the genus is in itsinfancy (e.g. Crawford et al., 2007; Crawford and Smith, 2005;Pinto-Sánchez et al., 2012, 2014). Pinto-Sánchez et al. (2014) men- tionthatTerraranancladeshaveradiatedinbothMiddleandSouthAmerican regions. Therefore, successful colonisations increasedlocal richness, mainly in some Andean communities with slightlyhigher richness and located where Terraranans have occurred thelongest. On the other hand, previous vicariance models providethe most parsimonious explanation for the pattern of distributionin  Pristimantis  (Lynch and Duellman, 1997) comprising uniqueassemblages inside each mountain range and high  b  diversity(BernalandLynch,2008;García-Retal., 2012)correlatedwithlim-ited gene flow inside a given elevational band (Lynch, 1999a,b).This hypothesis is highly related to niche conservatism patterns(Kozak and Wiens, 2010; Wiens and Graham, 2005; Wiens et al.,2010) and predicts that within each montane system, highlandtaxa are more closely related to other highland taxa than to low-land taxa.New DNA sequences of   Pristimantis  are reported here, and ‘‘su-permatrices’’ (Pyron and Wiens, 2011) enable further elucidatingthe complex taxonomy and biogeographic history of this group(Padial et al., 2014). We combine novel and published DNA datatotesttheroleofthegeologicalcomplexityoftheprincipalbiogeo-graphic provinces and elevational gradients in generation of   Pristi-mantis  diversity. Our study includes new DNA sequence data from12 highland species not previously included in any molecular phy-logenetic inferences. These new data reduce the sampling gap inthe Colombian Andean region, while improving resolution of thegeneric level phylogeny. We further address the following ques-tions: (1) are highland species of   Pristimantis  in the NorthwesternAndes more closely related to other highland or lowland species?and (2) is the cladogenesis of   Pristimantis  correlated with the timeof uplifting of each Andean region? 2. Materials and methods  2.1. Taxon sampling  We created a dataset of DNA sequences deposited in GenBankfor mitochondrial ( 12S  ,  16S  ,  COI  ) and nuclear ( Rag-1  and  Tyr  )regions (Faivovich et al., 2005; Elmer et al., 2007; Heinicke et al.,2007; Hedges et al., 2008; Padial and De la Riva, 2009; Padialet al., 2009; Fouquet et al., 2012; Pinto-Sánchez et al., 2012). Addi-tionally, we collected specimens of 12 species of   Pristimantis  fromthe Northwestern Andes not included in previous phylogeneticstudies, and generated representative sequence data. The resultingdataset included 160 species or 37% of the currently known diver-sity within  Pristimantis , making this study the most extensive ana-lysis for the genus to date. We rooted our phylogeny using asoutgroups species of   Phrynopus ,  Lynchius  and  Oreobates  (Hedgeset al., 2008; Heinicke et al., 2007; Pyron and Wiens, 2011). A listof specimens and GenBank accession numbers of the sequencesincluded in this study are presented in Table S1.  2.2. Molecular techniques Genomic DNA was extracted from liver using the QiagenDNeasy kit following the manufacturer recommendations. Geno-mic DNA was PCR amplified using the primers specified inTable S2. For mitochondrial genes we used 20 l L PCRs with2.5mM MgCl 2  and 0.3 l M of each primer (forward and reverse),0.625 U of Taq polymerase and 1–2 l L of template DNA. Thermalcycling conditions involved 3min at 95  C followed by 34 cyclesof 30s at 93  C, 40s at 53  C and 90 s at 72  C, with a final exten-sion of 8min at 72  C. Amplification of nuclear genes included thefollowing changes in conditions: 2mM MgCl 2 , and 0.6 l M eachprimer. Thermal cycling conditions involved 94  C for 5min and40 cycles of 30s at 94  C, 30s at 53  C, and 1min at 72  C, with afinal extension of 7min at 72  C. We purified the PCR productswith the Qiagen QIAquick PCR kit. Sequencing was performed inboth directions with a 3730XL DNA analyzer with the Big Dye Ter-minator sequencing kit from Applied Biosystems.  2.3. Data analysis 2.3.1. Phylogenetic reconstruction Resulting sequences were reviewed and edited manually inorder to identify and correct incongruence between both chainsbefore phylogenetic analysis. We constructed matrices for eachgene aligned using SATé-II (Liu et al., 2012) with the followingparameters: aligner: MAFFT; merger: MUSCLE, model: GTR+G20.The matrices were inspected and manual corrections were madeat the end of the chains with Mesquite 2.75 (Maddison andMaddison, 2011). Individual matrices were analyzed in jModelTest0.1.1 (Posada, 2008) to choose a substitution model for each genebased on the Bayesian Information Criterion. The best model forall three mitochondrial genes ( 12S  ,  16S   and  COI  ) was GTR+ C ,while the best models for  Rag-1  and  Tyr   were HKY+ C  andHKY+I+ C , respectively. Finally, we concatenated the gene matri-ces in Mesquite 2.75.Bayesian phylogenetic inference used a three-partition scheme(i.e. partition by each model) in MrBayes 3.2.1 (Ronquist andHuelsenbeck, 2003). We ran two independent analyses for 12 mil-lion generations, each sampling trees and parameter values every1000 generations. Burn-in was set to 25% and thus the first 3 mil-lion generations were discarded. Additionally, a Maximum Likeli-hood (ML) analysis was performed using RAxML (Stamatakis,2006). The substitution model for ML analysis was GTR+ C . Onehundred trees were inferred and the tree with the highest likeli-  Á.M Mendoza et al./Molecular Phylogenetics and Evolution 85 (2015) 50–58  51  hood was chosen for the posterior 1000 replicates bybootstrapping.  2.3.2. Divergence times Molecular dates were estimated using a Lognormal relaxedBayesian clock implemented in BEAST 1.6.2 (Drummond andRambaut, 2007). In the absence of a fossil record for the genus, webased our analysis on previously published divergence times. Ascalibration constraints we used the stem age of   Pristimantis  24.45(C.I.=17.30–34.82) Million years ago (Ma) (Heinicke et al., 2007).The second calibration node was placed at 8.6Ma (C.I.=5.5–12.0),indicatingthestemageofdivergenceofthe P.pardalis speciesgroup(represented hereby P. altae ,  P. pirrensis  and P. pardalis ) fromother Pristimantis (Pinto-Sánchezetal.,2012).Finally,thenodecontainingthe  P. taeniatus  clade provided our third calibration point with anage of 8.3Ma (C.I. =5.6–11.2). The topology was inferred  de novo by the software using the same partition substitution models. WechoseaYulespeciationprocess(Drummondetal.,2006)withvaria-tion in the substitution rates. The analysis was run for 10milliongenerations sampled every 1000 generations.  2.3.3. Ancestral range reconstruction Dataforspecies’rangeswereobtainedfromtheGlobalAmphib-ian Assessment, Global Biodiversity Information Facility (GBIF),and recent publications of new records and range extensions(Padial and de la Riva, 2009; Lehr et al., 2004). Eight different geo-graphical regions were considered in our analysis based on previ-ous delimitations applied in previous biogeographic analysis fordendrobatids (Santos et al., 2009) and centrolenids (Castroviejo- Fisher et al., 2014), whose current distributions in the Neotropicsare coincident with current  Pristimantis  species distributions. Theregions were refined by major geological events that accompaniedtheir formation (Hoorn et al., 1995; Hooghiemstra and Van derHammen, 1998; Coltorti and Ollier, 2000; Gregory-Wodzicki,2000; Ramos and Folguera, 2009). The regions included were asfollows: A – Amazonas, B – Central Andes, C – Northwest Andes,D – Eastern Cordillera, E – Central America, F – Guiana Shield, G–Choco-Darien, H–AntillesArchipelagoandtheCaribbeancoastalcordilleras(Fig.1).Weusedtheposteriorconsensustreeestimatedin BEAST to leave only one sample for each species, eliminatingnon-specific taxa from all the analyses (sequences found as  Pristi-mantis  sp., aff. and cf. in GenBank). We inferred the most probableancestral distributions for the internal nodes within the phyloge-netictree usingthe likelihood Dispersal–Extinction–Cladogenensis(DEC) model (Ree et al., 2005; Ree and Smith, 2008), implementedin LAGRANGE 2.7. Antilles/Caribbean coastal cordillera and North-westernAndes wereclumpedin singleregions, restrictingthe pos-sible regions by species to less than three regions each, in order toease the designation of species ranges in LAGRANGE 2.7. Connec-tivity and dispersal parameters are summarized in Table S3. Thisapproachwasusedbecauseitallowedatime-stratifiedmodelidealfor the  Pristimantis  case in the Andean region, for which someregions were just starting to develop at the Oligocene. The DECmodel also permits the designation of multiple areas occupied byeach species, and the inheritance of ancestral ranges by daughterspecies (Kodandaramaiah, 2010). For this analysis, Python scriptswere generated using the online LAGRANGE configurator ( DEC model incorporates information on the absence of thedefined areas during certain periods in time. Hence, uplift of Andean regions during middle Miocene and early Pliocene can beincluded. In this sense, three time periods were differentiated:(1) From the srcin of   Pristimantis  during the Oligocene to middleMiocene. During this time, only the Northwestern Andes wasdeveloped, and there was a continuum between the Choco andAmazonian regions, (2) From middle Miocene to early Pliocene,when the uplift of the Central and Caribbean-Merida Andesoccurred and (3) From early Pliocene to present, when the CentralAndes connected with the Northwestern Andes in South America.The connection occurred because of the uplift of the EcuadorianAndes,andtheNortheasternAndescreatedacorridortotheNorth-western Andes and the Merida Andes. These two events limit fau-nal exchanges between Amazonas and Choco, whereas Choco isnow connected with Central America by the uplift of the Isthmusof Panamá (Fig. 2). For connectivity between regions, we generatea matrix (Table S3) with values of 1.0 for pair of regions havingdirect connection; 0.5 values for pair of regions separated by oneregion and for regions which are not directly connected but cur-rently have the same species living in both regions; and 0.1 valuesfor pair of regions with no direct connection, not emerged in thetimeperiodanalyzedandnoevidenceofsharingcurrent Pristiman-tis  species. Values of 0.0 were not used in this matrix because itgenerated crashes in the ML search.A similar analysis was performed for species grouped by eleva-tions in order to evaluate the elevational band hypotheses. In thiscase, three states were recognized as follows: L=species below1000m; M=species between 1000 and 3000m and N=specieshigher than 3000m (Fig. 2). The number of intervals was chosenconsidering the complexity of restriction matrices in LAGRANGE2.7 equal to regional categories. In this sense, the elevation inter-vals follow Caldas (1951) and Cuatrecasas (1934) life zones classi- fication. Multiple regions and elevations were defined forwidespread species: LM=species occurring from zero to 3000m,MN=species occurring from 1000 m to higher than 3000m andLMN=species occurring along all elevation categories. 3. Results  3.1. Phylogenetic relationships and divergence times We generated an alignment of 2841 base pairs (bp) for the  12S  and  16S   regions, 635 for  COI  , 634 for  Rag-1  and 531 for  Tyr   for atotal of 4638bp. No stop codons in the three protein-coding genes( COI  ,  Rag-1  and  Tyr  ) were detected. Bayesian and ML topologiesobtained were congruent with one another, and we present herethe Bayesian consensus tree (Fig. 3). Posterior probabilities (pp)higher than 0.8 were obtained for 75% of the nodes (183 of 243)in Bayesian reconstruction and bootstrap values above 70% wereobtained in 53% of the nodes (130 of 243) in ML reconstruction.In order to ease interpretation of the results, four large groups(A-D) were recognized in the phylogeny. Among these, group A isthe best supported clade, and comprises 23 species. Group Bincludes 32 species; group C has seven species; and group D theremaining 93 species that were divided into two subgroups (D1and D2; Fig. 4). Four of the new species included ( P. calcaratus, P.kelephas, P. orpacobates, P. thectopternus ) were placed in the groupA, seven ( P. jubatus, P. hectus, P. acatallelus, P. capitonis, P. myops, P.quantus and P. breviforns )wereclumpedwithinthegroupD1,sixof them( P. acatallelus, P. capitonis, P. myops, P. angustilineatus, P. quan-tus  and  P. breviforns ) forming a single clade (Fig. 4). Our phyloge-netic reconstruction reveals that  P. lymani  was placed among  P.achatinus  samples, andsomeofthe P.  aff.  taeniatus  specimenswereincluded in clades containing  P. palmeri .  Pristimantis rozei (represented a non-monophyletic taxon) generated extremelylowsupportforbranches(pp<0.70),andtherefore,itwasremovedfor ancestral range reconstruction analysis. Diversification timesof the identified groups were during early Oligocene andMiocene (Fig. 4). A crown group A has a likely originabout 28Ma (95% HDP=18.11–31.67), group B originated  27Ma (95% HDP=19.80–33.05), group C dates from 26Ma(95% HDP=17.86–31.17) and, finally, the most recent group Dsrcinated  22Ma (95% HDP=15.88–28.22). 52  Á.M Mendoza et al./Molecular Phylogenetics and Evolution 85 (2015) 50–58   3.2. Ancestral range reconstruction The log-likelihood of the first model (by regions) was  369.4, adispersal rate three times higher than extinction rates (disper-sal=0.02578, extinction=0.00828), and the log-likelihood for sec-ond model (by elevations) was   286.4 with a dispersal rateextremelyhighandaverylowextinctionrate(dispersal=0.03829,extinction=1.65e  09). The DEC analysis suggests the Northwest-ern Andean region as the most probable ancestral range of themost recent common ancestor (MRCA) of   Pristimantis  (RelativeProbabilities: C|BC=0.361, C|CG=0.2156, C|C=0.1223). Theinferred character state of the ancestral node is a montane eleva-tional range between 1000 and 3000m (Relative Probabilities:M|M=0.6594, M|LM=0.1807; Fig. 5).The ancestral range of group A occupied the NorthwesternAndes andsubsequent dispersal events tookit tothe Chocoregion.ThedistributionalrangefortheMRCAofgroupBincludedtheCho-co region. Species diversification within group B was accompaniedby several dispersal events into the Amazonian forests (mainly inthe subgroup containing  P. stichogaster, P. aniptopalmatus, P. rhab- Fig. 1.  Distribution and biogeographic areas of   Pristimantis  species designated for this study. These areas where included in the reconstruction of ancestral ranges. Time1: 35-23 MyaEarly Oligocene A Time2: 23-6 MyaOligocene to Miocene B Time3: 6-0 MyaPliocene to present C Fig. 2.  Time periods used for calibration in the DEC analysis using LAGRANGE 2.7. (A) Only Northwestern Andes are present in the early Paleocene. (B) Uplift of the CentralAndes during the middle Miocene and (C) with the uplift of the Eastern cordillera and the Ecuadorian Andean region during Pliocene time, the Northwestern cordillerasmerged with the Merida Andes, and Central Andes connects with Northern Andes. See Fig. 1 for the elevational code.  Á.M Mendoza et al./Molecular Phylogenetics and Evolution 85 (2015) 50–58  53  dolaemus, P. toftae, P. pluvicanorus  and  P. sagittulus ), Central andMerida Andes, and the Caribbeancoastal cordilleras (Fig. 6). Diver-sification of group B is inferred as occurring in the elevation bandM (1000–3000m), except for some species that occurred in L (be-low1000m),correspondingtoapossibleChocoantoMeridaAndesdispersion (31% of the species within this group). Group C shows asingleoriginanddiversificationinNorthwesternAndesandamainvicariance event in elevations >3000m. The MRCA of group D wasprobably present in Northwestern Andes, with most basal nodesassociated with this region. A dispersal event to Central Andesoccurring 22Ma was identified for the MRCA of subgroup D2(95% HDP: 14.81–25.46). Finally, group D likely dispersed fromNorthwestern and Central Andes to Amazonian and Choco regions(Fig. 6). Althoughmostof the events areassociatedwiththe1000–3000 elevation band, this group shows more elevational variationand dispersion between all elevations than those found for groupsA-C. 4. Discussion Our results identify the middle elevation band of NorthwesternAndes in Colombia and Ecuador as the most important region forsrcin and diversification of   Pristimantis . Phylogenetic evidencesupporting the hypothesis of   in situ  divergence has been reported P_capitonis_UVC15948P_capitonis_UVC15947P_capitonis_UVC15848 Fig. 3.  Bayesianinference(BI) consensus treewithposteriorprobabilityfor Pristimantis  basedona5-geneanalysisof244frogs. Numbersabovebranchesrepresent posteriorprobabilities. Fig. 4.  Chronogramof  Pristimantis  includingoutgroupinferredfromBayesianrelaxedclockdatingmethodsinBEAST1.6.2.Thetopologyincludesdivergencetimesandshowsthe four main groups. Three calibration constraints were used: (1) divergence of   Pristimantis  from other Terraranans 24.45Ma (C.I. =17.30–34.82), (2) divergence of the  pardalis  species group 8.6Ma (C.I.=5.5–12.0), and (3) the node containing the  P. taeniatus  clade 8.3Ma (C.I. =5.6–11.2). For each node the estimated time of divergence isindicated with a bar representing the 95% HPD intervals of node ages. The time scale is in millions of years ago (Ma).54  Á.M Mendoza et al./Molecular Phylogenetics and Evolution 85 (2015) 50–58
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks