isolation by instability: historical climate change shapes

1748 emsp|emsp wileyonlinelibrarycomjournalmec Molecular Ecology 2019281748ndash1764copy 2019 John Wiley amp Sons Ltd

Received 26 May 2017emsp |emsp Revised 17 January 2019emsp |emsp Accepted 23 January 2019

DOI 101111mec15045

O R I G I N A L A R T I C L E

Isolation by instability Historical climate change shapes population structure and genomic divergence of treefrogs in the Neotropical Cerrado savanna

Mariana M Vasconcellos1 emsp| Guarino R Colli2 emsp| Jesse N Weber3 emsp| Edgardo M Ortiz14 emsp| Miguel T Rodrigues5 emsp| David C Cannatella1

1Department of Integrative Biology The University of Texas at Austin Austin Texas2Departamento de Zoologia Universidade de Brasiacutelia Brasiacutelia Brazil3Department of Biological Sciences University of Alaska Anchorage Anchorage Alaska4Plant Biodiversity Research Department of Ecology and Ecosystem Management Technical University of Munich Freising Germany5Departamento de Zoologia Instituto de Biociecircncias Universidade de Satildeo Paulo Satildeo Paulo Brazil

CorrespondenceMariana M Vasconcellos City College of New York City University of New York New York NYEmail marimiravascgmailcom

Funding informationNational Science Foundation Division of Environmental Biology (NSF-DEB) GrantAward Number 1311517 United States Agency for International Development GrantAward Number AID-OAA-A-11-00012

AbstractAlthough the impact of Pleistocene glacial cycles on the diversification of the tropical biota was once dismissed increasing evidence suggests that Pleistocene climatic fluctuations greatly affected the distribution and population divergence of tropical organisms Landscape genomic analyses coupled with paleoclimatic distribution models provide a powerful way to understand the consequences of past climate changes on the present-day tropical biota Using genome-wide SNP data and mito-chondrial DNA combined with projections of the species distribution across the late Quaternary until the present we evaluate the effect of paleoclimatic shifts on the genetic structure and population differentiation of Hypsiboas lundii a treefrog en-demic to the South American Cerrado savanna Our results show a recent and strong genetic divergence in H lundii across the Cerrado landscape yielding four genetic clusters that do not seem congruent with any current physical barrier to gene flow Isolation by distance (IBD) explains some of the population differentiation but we also find strong support for past climate changes promoting range shifts and struc-turing populations even in the presence of IBD Post-Pleistocene population persis-tence in four main areas of historical stable climate in the Cerrado seems to have played a major role establishing the present genetic structure of this treefrog This pattern is consistent with a model of reduced gene flow in areas with high climatic instability promoting isolation of populations defined here as ldquoisolation by instabil-ityrdquo highlighting the effects of Pleistocene climatic fluctuations structuring popula-tions in tropical savannas

K E Y W O R D S

ddRADseq Hypsiboas lundii isolation by distance landscape genomics phylogeography Pleistocene climatic fluctuations

emspensp emsp | emsp1749VASCONCELLOS Et AL

1emsp |emspINTRODUC TION

The climatic fluctuations of Pleistocene glacial cycles have greatly impacted the distribution of organisms imprinting on their popu-lations a genetic signature of such changes (Davis amp Shaw 2001 Hewitt 2000) Even though the genetic consequences of the ice ages are well documented for many organisms in the Northern Hemisphere (Hewitt 2000 2004) less evidence exists for the Southern Hemisphere (but see Cabanne et al 2016) possibly due to the reduced extent of glaciations in this region Although gla-ciers covered only small portions of South America most notably in the extreme southern region and in the Andes (Ehlers Gibbard amp Hugues 2011 Vuilleumier 1971) cycles of contraction and ex-pansion of forests and savannas during glacial and interglacial pe-riods are thought to have occurred (Van der Hammen 1974) and to have significantly affected vegetation structure and distribu-tion of organisms throughout the continent (Bonaccorso Koch amp Peterson 2006 Saia Pessenda Gouveia Aravena amp Bendassolli 2008)

The effect of Pleistocene climatic fluctuations on the South American biota has long been a controversial topic especially since it was first hypothesized to explain the large number of spe-cies in the Amazon (Bush 1994 Haffer 1969) The forest refugia hypothesis (Haffer 1969) posits that glacial periods promoted fragmentation of forest into isolated patches serving as faunal refuges which then spurred allopatric speciation Given that addi-tional studies did not support fragmentation of the Amazon forest (Colinvaux Oliveira amp Bush 2000) elevated speciation rates in the Pleistocene (Rull 2008) or refugia locations (Nelson Ferreira Silva amp Kawasaki 1990) the impact of Pleistocene climatic fluc-tuations was generally dismissed for Neotropical regions (Willis amp Whittaker 2000) Nonetheless climatic changes have been since then invoked to explain the marked genetic structure in Neotropical species in the Brazilian Atlantic Forest (Cabanne et al 2016 Carnaval Hickerson Haddad Rodrigues amp Moritz 2009) and the Cerrado (Diniz-Filho et al 2015 Prado Haddad amp Zamudio 2012) for example Historically stable habitats since the Pleistocene also seem to be an important predictor of endemic diversity (Carnaval et al 2014) species richness (Graham Moritz amp Williams 2006 Werneck Nogueira Colli Sites amp Costa 2012b) and genetic struc-ture in several other tropical regions (Bell et al 2010 Carnaval amp Bates 2007 He Edwards amp Knowles 2013) suggesting that range shifts associated with climatic cycles must have occurred in the tropics Moreover due to their narrow thermal tolerance tropical ectotherm species are expected to have been particularly affected by temperature shifts during climatic cycles (Deutsch et al 2008)

Fluctuation in species ranges associated with Pleistocene cli-matic cycles might strongly affect the configuration of popula-tions in space limiting gene flow and affecting genetic structure Although the current spatial genetic structure of populations can reflect both historical and contemporary ecological factors his-torical factors such as paleoclimate dynamics or barriers to gene flow that are not currently present (eg paleodrainages) seem to

have a long and lasting contribution to the current pattern of ge-netic differentiation (Epps amp Keyghobadi 2015) In many cases historical factors can better explain genetic variation across the landscape than current climate or recent barriers to gene flow (He et al 2013 Nascimento et al 2013 Ortego Gugger amp Sork 2014 Thomaz Malabarba Bonatto amp Knowles 2015) Therefore current low levels of admixture might not overwrite the genetic signatures of demographic and distributional changes due to past environmental fluctuations which have to be carefully considered whenever inferring current ecological correlates of gene flow (Dyer Nason amp Garrick 2010)

Understanding the factors limiting gene flow and structuring populations in space can also provide important insights about driv-ers of diversification and their relative contribution to high biodi-versity regions In particular the Cerrado which is the largest and most species-rich tropical savanna and a global hotspot of biodi-versity (Klink amp Machado 2005 Myers Mittermeier Mittermeier Fonseca amp Kent 2000) is remarkable for high endemicity levels reported for plants (44 Klink amp Machado 2005) frogs (53 Valdujo Silvano Colli amp Martins 2012) and squamate reptiles (39 Nogueira Ribeiro Costa amp Colli 2011) Yet the evolutionary and ecological processes that led to this high species diversity are still not well understood (Silva amp Bates 2002 Werneck 2011) This sa-vanna spreads over the central part of an open dry diagonal in South America separating the Amazon and the Atlantic forest and is com-posed of a mosaic of heterogeneous vegetation ranging from grass-land to riverine gallery forest While adaptation to fire seems to be an important and recent promoter of plant diversification in Cerrado (Simon et al 2009) geological events in particular the uplift of the Central Brazilian Shield are regarded as more important for diver-sification in vertebrates (Domingos et al 2014 Santos Nogueira Giugliano amp Colli 2014 Werneck Gamble Colli Rodrigues amp Sites 2012a) Even though some studies have associated intraspecific levels of differentiation with past climatic changes in the Cerrado (Diniz-Filho et al 2015 Novaes Lemos Ribeiro amp Lovato 2010 Prado et al 2012 Ramos Lemos-Filho Ribeiro Santos amp Lovato 2007) only one study (of an endemic tree) has explicitly tested for and found significant association between climate change and ge-netic structure (Diniz-Filho et al 2015)

Paleodistribution models for the Cerrado savanna predict a dy-namic history of vegetation fluctuation over the late Quaternary in-cluding both the Pleistocene and the Holocene (Werneck Nogueira et al 2012b) These models indicate that climate conditions during the Last Interglacial (LIG ~120 k years ago) which was 05ndash2degC warmer and slightly drier than present in the Cerrado were some-what favourable to the occurrence of Cerrado vegetation They also indicate a contraction in the extent of the Cerrado during the Last Glacial Maximum (LGM ~21 k years ago) which was 1ndash35degC colder and drier than the present More recently during the mid-Holocene (~6 k years ago) the climate and extent of the Cerrado vegetation did not substantially differ from present conditions (Werneck Nogueira et al 2012b) Although this response departs from the expectation of savanna expansion towards adjacent forested regions during the

1750emsp |emsp emspensp VASCONCELLOS Et AL

LGM with subsequent contraction (Mayle Beerling Gosling amp Bush 2004) it highlights the dynamic climate and vegetation changes in the Cerrado throughout the late Quaternary

To evaluate the contribution of the dynamic late Quaternary climate as a significant promoter of intraspecific diversification in the Cerrado savanna we studied the phylogeography and popula-tion divergence of a treefrog (Hypsiboas lundii) that inhabits gallery forest habitats We sampled populations throughout the region and obtained genetic data from a mitochondrial marker and from ge-nome-wide SNPs to assess the phylogeographic structure and popu-lation differentiation across the landscape We first determined the effect of simple geographic distance on population differentiation a widespread phenomenon known as isolation by distancemdashIBD (Wright 1943) Then to evaluate whether climatic stability since the Pleistocene might also explain population differentiation and ge-netic structure in H lundii we built species distribution models for the present and three successively older time periods We then com-bined these models to generate a stability surface predicting areas where populations might have persisted over time This surface al-lowed us to measure the resistance to gene flow among populations specifically the resistance caused by climatic instability imposed by local changes in environmental suitability for the species over time Finally we used regression models to determine the individual and combined effects of two models of genetic divergence isolation by distance and isolation due to historical climate fluctuations which we refer to as ldquoisolation by instabilityrdquo

2emsp |emspMATERIAL S AND METHODS

21emsp|emspGeographic sampling and DNA extraction

We sampled 214 individuals of Hypsiboas lundii from 47 widespread localities in the Cerrado region (Table 1) encompassing the entire known distribution of the species (Figure 1) We also sampled two individuals of the sister species H pardalis from two localities in the Atlantic Forest (Supporting Information Table S1) to use as outgroup Liver or muscle samples were collected from all individuals and pre-served in 99 ethanol We isolated genomic DNA using a DNeasy Blood amp Tissue Kit (Qiagen Inc CA) confirmed the quality of extrac-tions by visualizing high molecular weight bands on a 1 agarose gel and quantified DNA concentrations using a Quibitreg 20 fluorometer (Life Technologies NY)

22emsp|emspddRAD library preparation sequencing assembly and genotyping (SNP data set)

We used the ddRADseq method (Peterson Weber Kay Fisher amp Hoekstra 2012) to construct and sequence DNA libraries of short fragments distributed across the entire genome For each sample we first digested 250 ng of freshly extracted DNA with the restriction enzymes SphI and MspI (New England Biolabs MA) for 3 hr at 37degC For enzymatic clean-ups we used Agencourt AMPure beads (Beckman Coulter CA) standardized DNA to 75 ng

per sample and performed ligations using sixfold excess of two customized Illumina adapters (molarity calculated using ddRAD online materials) We used 48 P1 adapters that each included a 5 bp individual-specific barcode in combination with a general tagged biotin-labelled P2 adapter including a 10 bp degenerate base region (DBR) developed following Schweyen Rozenberg and Leese (2014) The inclusion of this DBR (5prime NNNNNIIICC 3prime) allowed us to mitigate PCR duplication bias during library prepara-tion (see our bioinformatics pipeline next) one of the main criti-cisms of ddRADseq protocol (Puritz et al 2014) After ligation purified samples were pooled into libraries of up to 48 barcoded samples Fragment sizes of 448ndash496 bp were then isolated using a BluePippin (Sage Science MA) automated gel extraction We iso-lated biotin-containing fragments using Streptavidin-coated beads (Dynabeadsreg M-270 Life Technologies) and amplified all pools with a 12-cycle PCR using a Phusionreg High Fidelity polymerase kit (New England Biolabs) and primers with pool-specific indexes as described in Peterson et al (2012) The quality of each library was confirmed on a 2100 Bioanalyzer (Agilent Technologies CA) before sequencing on an Illuminareg HiSeq 2500 (paired-end 2 times 125 bp chemistry) at the Genomic Sequencing and Analysis Facility (GSAF) at the University of Texas at Austin All pools were sequenced using two HiSeq lanes Sequencing reads are available at NCBI SRA (BioProject PRJNA521095)

We ran the following bioinformatics pipeline on a High-Performance Computing Cluster at Texas Advanced Computing Center (TACC) The raw sequences of each indexed pool were demul-tiplexed by their individual barcodes using deML (Renaud Stenzel Maricic Wiebe amp Kelso 2015) based on the maximum likelihood of assignment using the default cut-off for quality scores We identified and filtered PCR duplicates (ie any reads with identical degener-ate base region ldquoDBRrdquo and 100 bp of identical sequence after the DBR) for each individual using a custom Python script available at httpsgithubcommarimiraPCRdupsRemover This step reduces the amount of over-represented fragments resulting from PCR du-plication during library preparation which could lead to erroneous genotype calls and skewed allele frequencies We then used the pyRAD 305 pipeline (Eaton 2014) for subsequent quality filtering de novo assembly (using paired-end reads with DBR trimmed out) and genotype calling In short after filtering out low-quality reads discarding those with more than four low-quality bases (PHRED score lt20) reads were clustered within and across individuals with a 90 similarity threshold using vsearch (httpsgithubcomtorognesvsearch) allowing for indels Loci with low coverage (lt6times) within individuals were excluded During the genotype-calling step a consensus sequence was generated for all loci within individuals considering an estimated sequencing error rate (euro) and heterozygos-ity (π) across all sites (euro = 000164 and π = 000776 for this study) A final alignment step for each locus across samples is then performed in muscle (Edgar 2004)

Post-genotyping filters were applied to remove potential pa-ralogs We excluded loci with more than two alleles per individ-ual (Hypsiboas lundii is diploid) loci for which within-individual


TA B L E 1 emsp Sampled localities of Hypsiboas lundii in the Brazilian Cerrado with geographic coordinates

Code Municipality State N ind Longitude Latitude Cluster

1 guima Chapada dos Guimaratildees MT 10 minus55804 minus15472 West

2 jacia Jaciara MT 9 minus55036 minus15983 West

3 rondo Rondonoacutepolis MT 1 minus54736 minus16669 West

4 altog Alto Garccedilas MT 5 minus53480 minus17043 CentralWest

5 garca Barra do Garccedilas MT 6 minus52254 minus15873 CentralWest

6 serra Serranoacutepolis GO 9 minus51963 minus18288 CentralWest

7 caiap Caiapocircnia GO 4 minus51833 minus16963 CentralWest

8 goias Goiaacutes GO 11 minus50115 minus16001 Central

9 campe Campestre de Goiaacutes GO 2 minus49697 minus16737 Central

10 petro Petrolina de Goiaacutes GO 2 minus49215 minus16256 Central

11 limpo Campo Limpo de Goiaacutes GO 4 minus49082 minus16292 Central

12 piren Pirenoacutepolis GO 5 minus48964 minus15850 Central

13 domin Satildeo Domingos GO 3 minus46326 minus13395 Central

14 caval Cavalcante GO 3 minus47432 minus13812 Central

15 teres Teresina de Goiaacutes GO 7 minus47262 minus13875 Central

16 parai Alto Paraiacuteso de Goiaacutes GO 4 minus47506 minus14138 Central

17 alian Satildeo Joatildeo dAlianccedila GO 13 minus47508 minus14649 Central

18 brasi Brasiacutelia DF 1 minus47857 minus15677 Central

19 desco Santo Antocircnio do Descoberto GO 3 minus48272 minus16097 Central

20 luzia Luziacircnia GO 6 minus48176 minus16299 Central

21 migue Satildeo Miguel do Passa Quatro GO 1 minus48546 minus17042 Central

22 pires Pires do Rio GO 1 minus48258 minus17237 Central

23 catal Catalatildeo GO 9 minus47728 minus17923 Southeast

24 unaii Unaiacute MG 9 minus47259 minus16408 Southeast

25 parac Paracatu MG 5 minus46824 minus17149 Southeast

26 gauch Chapada Gauacutecha MG 7 minus45543 minus15364 Southeast

27 santa Santa Feacute de Minas MG 11 minus45414 minus16673 Southeast

28 janua Januaacuteria MG 10 minus44240 minus15109 Southeast

29 pinhe Joatildeo Pinheiro MG 10 minus46185 minus17677 Southeast

30 pocoe Claro dos Poccedilotildees MG 5 minus44140 minus17104 Southeast

31 crist Cristaacutelia MG 7 minus42873 minus16635 Southeast

32 riach Santana do Riacho MG 3 minus43721 minus19168 Southeast

33 hiz Belo Hizonte MG 1 minus43912 minus19946 Southeast

34 curve Curvelo MG 1 minus44606 minus19152 Southeast

35 furna Satildeo Joseacute da Barra MG 2 minus46327 minus20687 Southeast

36 roque Satildeo Roque de Minas MG 3 minus46622 minus20262 Southeast

37 araxa Araxaacute MG 1 minus46941 minus19594 Southeast

38 perdi Perdizes MG 8 minus47140 minus19209 Southeast

39 sacra Sacramento MG 2 minus47293 minus19875 Southeast

40 pedre Pedregulho SP 1 minus47463 minus20244 Southeast

41 simao Satildeo Simatildeo SP 1 minus47607 minus21497 Southeast

42 srita Santa Rita do Passa Quatro SP 1 minus47591 minus21726 Southeast

43 claro Rio Claro SP 1 minus47699 minus22313 Southeast

44 cosmo Cosmoacutepolis SP 1 minus47219 minus22646 Southeast

45 botuc Botucatu SP 1 minus48445 minus22886 Southeast

46 bauru Bauru SP 1 minus49072 minus22243 Southeast

47 assis Assis SP 3 minus50375 minus22598 Southeast


consensus sequences harboured more than 10 heterozygous sites and any locus containing gt20 SNPs in either the forward or reverse reads Our post-filtering loci consisted of at least 229 bp after removing barcodes enzyme cut-sites DBR and after con-catenating both paired-end reads with gaps allowed Finally to explore the effects of missing data (eg due to allele dropout or low coverage for some loci) five different data sets with differing levels of locus coverage across samples were produced in pyRAD We generated matrices of loci shared among at least 40 50 60 70 and 80 of all H lundii individuals In addition we gen-erated SNP matrices that included outgroups for phylogenetic analyses and removed outgroups from data used in population genetic analyses

23emsp|emspPopulation structure and SNP phylogeography

To infer population genetic structure in Hypsiboas lundii we used structure v234 (Pritchard Stephens amp Donnelly 2000) a Bayesian clustering method in which we implemented an admixture model with correlated allele frequency among populations (Falush Stephens amp Pritchard 2003) For this analysis we used matrices of unlinked SNPs (one per locus) not including the outgroup individuals Because preliminary runs with different levels of missing data had congruent results we decided to use a genotype matrix with up to 50 missing data at any site as a good trade-off between maximizing

SNP number and minimizing low-information sites Conservative ap-proaches including very low levels of missing data can adversely impact downstream phylogenetic analyses by dramatically reducing the number of informative SNPs while excluding fast evolving sites (Huang amp Knowles 2016) We performed 10 runs for each K popula-tions ranging from 1 to 8 using a MCMC run of 400000 steps after a burn-in of 100000 steps Following the authors recommendation for SNP data sets we inferred lambda from runs with K = 1 and fixed it to the inferred value of 028 for all subsequent runs We used the r package parallelstructure (Besnier amp Glover 2013) to parallelize runs on a computer cluster The appropriate number of genetic clus-ters was evaluated by examining the mean and variance of the log-likelihood for each K and implementing the ∆K statistic of Evanno Regnaut and Goudet (2005) in structure harvester (Earl amp vonHoldt 2011) Graphs were produced summarizing all runs across K values in the clumpak server (Kopelman Mayzel Jakobsson Rosenberg amp Mayrose 2015) We also used fineradstructure (Malinsky Trucchi Lawson amp Falush 2018) to visualize the identified clusters in a heat map of coancestry matrix among all individuals using the SNP matrix with 50 missing data and the RAxML tree (see below)

We used two approaches to infer phylogeographic structure from our SNP data set First to infer the relationship among all individuals including the outgroup we used a maximum-likelihood approach in raxml v8 (Stamatakis 2014) using the conditional likelihood method (Lewis 2001) as an acquisition bias correction for dealing only with

F I G U R E 1 emsp Distribution of Hypsiboas lundii sampled localities in the Cerrado savanna identified by different symbols that correspond to the four genetic clusters found in this study Numbers correspond to each of the 47 localities in Table 1 Blue lines represent main rivers and white dotted lines represent state boundaries Cerrado limits are depicted either as a red line in the large detailed map or as a grey area in the small South America map highlighting the study region [Colour figure can be viewed at wileyonlinelibrarycom]ndash55 ndash50 ndash45 ndash40

ndash25

ndash20

ndash15

ndash10

0

500

1000

1500

2000

17

4

37

4746 45

18

7 9

23

14

43

44

31

34

19

13

35

526

81

33

2

28

11 2021 25

16

40

38

10

29

12

22 30

32

3

3639

27

6

4142

15

24

0 200 400

km

Elevation (m)

WestCentral WestCentralSoutheast


variable sites We used the K80 model of nucleotide substitution with rate heterogeneity ASC_GTRCAT for all five concatenated SNP data sets with different levels of missing data The data sets including all SNPs in all loci were pre-processed using the phrynomics r package (Leacheacute Banbury Felsenstein Nieto-Montes de Oca amp Stamatakis 2015) to filter out non-binary sites in the alignment Branch support was estimated using bootstrap with an automatic stopping criterion once enough replicates had been generated (Pattengale Alipour Bininda-Emonds Moret amp Stamatakis 2010) Then for a species tree analysis we used svdquartets v10 (Chifman amp Kubatko 2014) a coalescent method for unlinked SNP data that evaluates quartets of taxa implemented in paup v40a147 (Swofford 2002) We used 10 million random quartets (135 of all possible quartets) with 100 bootstraps setting Hypsiboas pardalis as the outgroup

24emsp|emspMitochondrial DNA sequencing and analyses

To confirm the phylogeographic structure for Hypsiboas lundii and check the congruence between nuclear and mitochondrial DNA we also sequenced the ND2 mitochondrial gene (1032 bp) for a subset of 63 individuals (Supporting Information Table S1) encompassing all 47 localities (1ndash2 indlocality) in addition to two samples of the outgroup H pardalis We amplified this region with TaKaRa Ex Taqreg polymerase kit (Clontech laboratories Inc CA) using the primers ND2F (5prime AGG ACC TCC TTG ATA GGG AG 3prime) and ND2R (5prime TGC TTA GGG CTT TGA AGG CYC 3prime) and the following thermocycler conditions initial denaturation at 94degC for 2 min followed by 35 cy-cles (30 s denaturation at 94degC 30 s annealing at 48degC 1 min exten-sion at 72degC) and a final extension of 7 min at 72degC PCR products were purified using Genelute PCR clean-up (Sigma-Aldrich MO) and sequenced at both directions at Macrogen Inc (Seoul South Korea) We used geneiousreg 717 (Kearse et al 2012) to edit and assemble individual sequences and performed a multiple alignment across all individuals using the muscle plug-in (Edgar 2004) GenBank ac-cession numbers for mtDNA are provided in Table S1 (Supporting information)

We used partitionfinder v 111 (Lanfear Calcott Ho amp Guindon 2012) to determine (under the BIC) the following best-fit partition-ing scheme and substitution models using linked branch lengths (a) HKY+I for ND2 position 1 (b) HKY for ND2 position 2 + tRNAMet (c) TrN+G for ND2 position 3 To generate a mitochondrial phylogeo-graphic hypothesis for Hypsiboas lundii and estimate divergence times we used beast v175 (Drummond Suchard Xie amp Rambaut 2012) with the partition scheme above We used the uncorrelated lognormal relaxed clock with the published ND2 substitution rate of 13 per lineage per million year obtained from another frog species (Macey et al 2001) We ran BEAST under the coalescent prior of constant population size for 107 generations sampled at every 103 generations Convergence and stationarity (ESS gt200) were assessed in tracer v16 with a burn-in of 20 and the maximum clade credibility tree was summarized using treeannotator We also used beast v175 to re-construct demographic history in H lundii using Bayesian skyline plots (Drummond Rambaut Shapiro amp Pybus 2005) either by including all

populations or clades that were congruent with recognized genetic clusters (see Section 3) using the same settings as above

25emsp|emspSpecies distribution models (present and past projections)

To predict areas of historical persistence of the species during pe-riods of climate change we modelled the distribution of Hypsiboas lundii at four time periods the present the mid-Holocene (6 kya) the Last Glacial MaximummdashLGM (21 kya) and the Last InterglacialmdashLIG (120 kya) We downloaded current and past bioclimatic vari-ables from the worldclim 14 database (Hijmans Cameron Parra Jones amp Jarvis 2005) at a spatial resolution of 5 min The LGM and mid-Holocene climatic data were derived from three General Circulation Models (CCSM4 MIROC-ESM MPI-ESM-P) available for both periods in the WorldClim database and the LIG data were obtained from Otto-Bliesner Marshall Overpeck Miller and Hu (2006) We included in our model bioclimatic variables that were not highly correlated with each other (r lt 09) and predicted to be relevant to the ecological requirements of Hypsiboas lundii which is a treefrog with wide distribution in a highly seasonal savanna and an extended reproductive season encompassing both rainy and dry seasons (Mazzarelli 2015) Our seven selected biocli-matic predictors included mean diurnal temperature range (BIO 2) temperature seasonality (BIO 4) mean temperature of warmest quarter (BIO 10) mean temperature of coldest quarter (BIO 11) precipitation of the wettest month (BIO 13) precipitation of the driest month (BIO 14) and precipitation seasonality (BIO 15) To ensure that climatic conditions in our past projections did not de-part greatly from present-day climate used to train the model we ran a multivariate environmental similarity surface (MESS) anal-ysis (Elith Kearney amp Phillips 2010) using the dismo R package (Hijmans Phillips amp Leathwick 2017) on each General Circulation Model (GCM) for each time period

We used random forest as the modelling algorithm (Breiman 2001) in the randomforest r package (Liaw amp Wiener 2002) with a classification model using 500 trees and randomly sampling two variables at each split Occurrence data (109 points) included the GPS records of our field collections and the geo-referenced records from Brazilian herpetological collections for which we could per-sonally verify the species identification Because our data lack true absence points pseudo-absence points were randomly generated in a reduced spatial extent of 4deg around the outmost presence points (10degS to 27degS 58degW to 38degW) We followed Barbet-Massin Jiguet Albert and Thuiller (2012) and used the same number of pseudo-absence points as presence points averaging model predictions from 10 runs with different sets of random points and using a buffer excluding the immediate area around each presence point We used a buffer of 40 km which we considered as more appropriate for our reduced spatial extent

To evaluate model performance we used a fourfold spatial-block cross-validation method in which presence and pseudo-absence data points are split into four spatial-block partitions with similar number


of points (Roberts et al 2017) Each spatial bock is withheld once to test the model trained with the other three spatial blocks Spatial-block partitions can be used to evaluate how models can transfer in environmental spaces not used to calibrate the model We also used random partitions in a fourfold cross-validation to compare AUC val-ues obtained from spatial blocks We used the area under the curve (AUC) of the receiver operating characteristic (ROC) plot to evaluate commission errors (specificity) and omission errors (sensitivity) After ensuring good model performance and transferability using the cur-rent climate each of the 10 model iterations (each with a different set of pseudo-absence points) was then projected onto each GCM at each of the three past climates (Mid-Holocene LGM and LIG) Presenceabsence maps are the final output of Random Forests classification models however we first averaged the raw probability projections of the 10 iterations for each GCM at each past climate to account for uncertainty in the pseudo-absence points and then we averaged across the three GCMs for Mid-Holocene and LGM to account for uncertainty in the climate model for these time periods We used the same threshold of the Random Forest classifier to generate the binary classification of suitableunsuitable habitats for the averaged model Finally to evaluate range stability through time the binary presenceabsence map outputs for all four time periods were overlain and checked for areas of overlap among all time periods generating a cli-matic stability surface with values ranging from 0 (not present in any of the four time periods) to 4 (present during all four periods)

26emsp|emspTesting the effect of historical climate change on gene flow and population structure

To test the hypothesis that present genetic structure of Hypsiboas lundii was shaped primarily by range shifts caused by climate changes in the late Quaternary we examined whether our cli-matic stability surface could explain population differentiation due to resistance to gene flow We used the stability surface as a conductance map in circuitscape v40 to generate pairwise resist-ance distances among populations Therefore areas with lower predicted stability through time have lower conductance values and higher resistance to gene flow We refer to this model as isolation by instability (IBI) The resistance distances are calcu-lated using a connectivity model based on circuit theory using an eight-neighbour cell scheme to estimate an average resistance value among populations in a heterogeneous landscape (McRae Dickson Keitt amp Shah 2008) Pairwise FST which can be pre-cisely estimated for samples sizes as small as four individuals when using large number (gt1000) of SNPs (Willing Dreyer amp Oosterhout 2012) was used as a proxy of gene flow among pop-ulations We calculated pairwise FST (Weir amp Cockerham 1984) in vcftools (Danecek et al 2011) for the 23 populations with at least four individuals

First we evaluated a simple null model of isolation by dis-tance (IBD) (Wright 1943) using a matrix regression between the genetic and the Euclidean geographic distances matrix Given the

strong effect of IBD (see Section 3) we then tested our model of isolation by instability which is a variation of the isolation by resistance model (McRae 2006) including historical climate using a multiple matrix regression approach (with both geogra-phy and climatic instability as independent predictors) to eval-uate the combined and isolated effect of each predictor on the genetic differentiation of populations We incorporated slope as a measure of landscape topographic complexity in our null model of isolation by distance (as in Zellmer and Knowles (2009)) We calculated slope (in degrees) using the elevation raster from WorldClim at the same resolution of the remaining bioclimatic variables (5 min) and estimated resistance distances in circuits-cape 40 Resistance distance increases linearly with the log transformation of the Euclidean distance in homogeneous two-dimensional habitats (McRae et al 2008) an important property that allowed us to use it as our new null model for gene flow (a topographically corrected IBD) We performed the multiple ma-trix regression as outlined in Legendre Lapointe and Casgrain (1994) and implemented by Wang (2013) using the MRM func-tion in the ecodist r package (Goslee amp Urban 2007) with 1000 permutations We transformed the FST data to the linear approx-imation (FST1minusFST) proposed by Rousset (1997) to circumvent the possibility of non-linear relationships between the genetic and the other distances Completely removing the confounding effects of geographic distance in a distance-based landscape ge-netic analysis without any bias can be an extremely hard task (Guillot amp Rousset 2013 Kierepka amp Latch 2015 Legendre amp Fortin 2010) Therefore to evaluate the additional contribution of including climatic instability in our topographically corrected IBD model we partitioned the variance explained (R2) by each predictor separately in the complete model (including the contri-bution of both predictors)

3emsp |emspRESULTS

31emsp|emspddRADseq processing

Our genetic sampling scheme with ddRADseq aimed to generate ~2 million sequencing reads per individual (targeting 100000 loci with a mean read depth of 20times) After demultiplexing PCR dupli-cate removal and filtering out low-quality reads we obtained an average of ~16 million reads per sample (Supporting Information Table S2) Our pyRAD pipeline generated 33597 loci on average per sample with average depth coverage of 18times for all loci (Supporting Information Table S3) of those 27046 loci per sample on average passed the paralog filters (Supporting Information Table S4) After alignment of these loci across samples our final data matrices varied between 14397 and 169 loci depending on the amount of missing data allowed (Supporting Information Table S5) As expected includ-ing the outgroup in the assembly impacted the paralog filtering and caused the final matrices to contain slightly fewer loci (Supporting Information Table S5)



Our structure analyses recovered four main genetic clusters for Hypsiboas lundii arranged from west to east as follows cluster West comprising localities 1ndash3 from Mato Grosso state cluster Central-West comprising localities 4ndash7 from Mato Grosso and western Goiaacutes state cluster Central including localities 8ndash22 from Goiaacutes state and cluster Southeast including localities 23ndash47 mostly from Minas Gerais and Satildeo Paulo states (Figures 1 and 2) We observed an appro-priate level of convergence across replicate runs given the small vari-ance around the mean log-likelihood across K (Figure 2) The mean log-likelihood increased steadily until reaching a plateau at K = 4 and the ∆K statistic of Evanno et al (2005) also peaked at this number of clusters (Figure 2) indicating that larger K values did not substantially improve the model fit The coancestry matrix among all individuals produced in fineRADstructure showed higher coancestry values among individuals of the same population and the same genetic clus-ters identified in structure (Supporting Information Figure S2)

The topologies generated in RAxML for SNP matrices with dif-ferent levels of missing data were generally similar recovering the same geographic clades and the four genetic clusters with sup-port gt 95 (Figure 3a) The only exception was the topology gener-ated with up to 20 missing data (including only SNPs shared across at least 80 of the individuals) which did not recover the Central cluster as a clade and also depicted a slightly different relation-ship among the clusters with decreased bootstrap support overall (Supporting Information Figures S3ndashS7) This data set included only 1823 total SNPs compared to 148665 SNPs in the largest data set (60) which likely decreased the power to detect phylogenetic sig-nal Thus using settings that severely limit the amount of missing data in final SNP data sets (particularly when using a de novo assem-bly with no reference genome) might pose a challenge to recovering

enough SNPs with strong phylogenetic signal for reliable inferences The population tree generated in SVDquartets also recovered the same four geographic clusters with high support (Figure 3b)

Despite the congruence between phylogenetic trees using the concatenated SNPs (RAxML) or coalescent model (SVDquartets) we detected a few noteworthy differences between the phyloge-netic trees and the structure plot First structure assigned individual s from locality 13 (Satildeo Domingos) to the Central cluster but it was placed as the sister of the Central and Southeast clades in the SNP phylogenies Second locality 4 (Barra do Garccedilas) assigned to the Central-West cluster by structure and the mitochondrial tree was more closely related to the West clade in the SNP phylogenies

33emsp|emspMitochondrial phylogeography

We recovered strong geographic structure in the ND2 gene for Hypsiboas lundii portraying the same west-to-east geographic pat-tern as the SNP-based tree analyses The only disagreements be-tween mitochondrial and SNP phylogenies involved the relationship among localities 4 6 and 7 in the Central-West cluster (Alto Garccedilas Serranoacutepolis Caiapocircnia) and locality 13 (Satildeo Domingos) in the Central cluster The Central-West cluster was not monophyletic in any of the phylogenies but the mitochondrial tree depicted even more incongruences (Serranoacutepolis related to the Central clade and Caiapocircnia related to the Southeastern clade) However in agree-ment with the structure assignment Satildeo Domingos was part of the Central clade in the mitochondrial tree (Figure 3c) a relationship that was not revealed by the SNP phylogenies More importantly the mtDNA data allowed us to predict divergence times for the major clades in our tree The most recent common ancestor (MRCA) of H lundii was estimated to be 21 Mya (95 HPD 162ndash275 Mya) therefore the diversification in this species is very recent most

F I G U R E 2 emsp Results of structure analyses Individual assignment plots for K = 2 3 4 Mean log probability of the data for different K values (1ndash8) and plot of ∆K statistic of Evanno et al (2005) showing best K = 4 [Colour figure can be viewed at wileyonlinelibrarycom]

K = 4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

West Central West

Central Southeast

K = 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

K = 21 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

87654321K

ndash240000

ndash220000

ndash200000

ndash180000

ndash160000

ndash140000

Mea

n Ln

pro

b o

f dat

a

L(K) (mean + ndash SD)

765432K

400

300

200

100

0

Del

ta K

Delta K = mean (|Lrdquo(K)|) sd(L(K))


F I G U R E 3 emsp Phylogenetic trees with colours corresponding to genetic clusters in Figure 2 (K = 4) Node support gt95 shown as black dots and gt70 as grey dots (a) Maximum-likelihood tree of all individuals inferred in raxml using SNP matrix with up to 50 missing data Clades are summarized by populations (whenever monophyletic) and drawn as triangle cartoons scaled to the number of terminal branches Branch lengths are scaled to nucleotide substitutions but the outgroup branch was shortened to facilitate visualization (b) Cladogram of populations using SNP matrix with 50 missing data inferred in SVDquartets (c) ND2 mitochondrial tree inferred in beast Branch lengths are scaled to the time axis at the bottom and the 95 CI of divergence times shown as grey bars at the nodes [Colour figure can be viewed at wileyonlinelibrarycom]

alian05

alian07

altog02

araxa01assis01

bauru01

botuc01

brasi04

caiap02caiap06

campe01

catal01

catal10

caval02

caval04

claro01

cosmo01

crist03

curve02

desco01

domin01domin09

furna01

garca02

gauch02gauch04

goias02goias10

guima01guima11

horiz03

jacia02jacia09

janua01janua04

limpo01

luzia01

luzia07

parda01

migue01

parac02

parai06

parda02

pedre02

perdi01perdi04

petro02

pinhe02

piren01

pires02

pocoe02

riach02riach05

rondo01

roque01roque06

sacra01sacra02

santa01

serra01

simao01srita01

teres03

unaii02unaii09

Pleistocene2 0 Ma8

(c)

MtDNA

8 Mya

21 Mya

13 Mya

08 Mya

09 Mya

06 Mya

07 Mya

05 Mya

H pardalis

simao (41)

caval (14)

goias (8)

pocoe (30)

garca (5)domin (13)

migue (21)

desco (19)

guima (1)

caiap (7)

brasi (18)

catal (23)

roque (36)

parai (16)

pinhe (29)

cosmo (44)

sacra (39)

luzia (20)

l impo (11)

petro (10)piren (12)

claro (43)

curve (34)

bauru (46)

pedre (40)

teres (15)

altog (4)

riach (32)

furna (35)

jacia (2)

perdi (38)

botuc (45)

assis (47)

pires (22)

gauch (26)

araxa (37)

horiz (33)

crist (31)

serra (6)

campe (9)

santa (27)

janua (28)

unaii (24)

rondo (3)

parac (25)

alian (17)

srita (42)

(b)

West

Central

West

Central

Southeast

SNPs

002

H pardalis

1

3

4

2

6

7

513

17

1415

1618

19202122

12

91011

8

27

2425

26

28

23

30

31

32333435

36

29

3738

39 4043

414244454647

(a)

SNPs


likely in the Pleistocene (lt 26 Mya) Yet our divergence time es-timates show that all four recognized geographic clusters were already established before the LIG (120 kya) and could have been impacted in a similar way by past climatic oscillations Interestingly the West clade of H lundii sister group to all other clades is the geographically most distant clade from the outgroup sister species H pardalis an endemic Atlantic Forest species Despite the recent origin of H lundii the MRCA between H lundii and H pardalis was estimated to be 80 My old

A Bayesian skyline plot including all populations suggests that the species experienced a strong increase in population size at around 100 kya coinciding with the last glaciation cycle of the Pleistocene followed by a very recent decline (Figure 4a) However this recent decline could be an artefact of the presence of population struc-ture when using all populations in the analysis (Heller Chikhi amp Siegismund 2013) The Bayesian skyline plots for the Central and Southeast lineages confirmed the increasing trend showing popu-lation expansion for each lineage separately (Figure 4bc) However the slow steady increase in population size in the Central lineage (Figure 4b) contrasts with more recent and rapid increase in the Southeast lineage (Figure 4c) Bayesian inference of population size could not be performed for the West and the Central-West lineages due to their small sample sizes

34emsp|emspDistribution shifts during glaciations (present and past predictions)

Most of the predicted distribution of Hypsiboas lundii for the cur-rent climate was within the present Cerrado extent however the model over-predicted its eastern distribution by including part of the Atlantic Forest The inclusion of several presence points in the con-tact zone between the Cerrado and the Atlantic Forest may explain this over-prediction The model using current climate performed well predicting the distribution of Hypsiboas lundii with an average AUC of 088 using random folds cross-validation and 078 using spatial-block folds cross-validation (Supporting Information Table S6) indi-cating also good transferability properties of the model The most

important climatic predictors in the model were the precipitation of wettest month (BIO 13) and the mean temperature of warmest quarter (BIO 10) (Supporting Information Table S6) The MESS re-sults (Supporting Information Figure S1) showed that only two main areas far from the predicted range of H lundii (one in southern Brazil and one in the coast of Brazil close to Salvador Bahia) have con-siderable non-analogous climate Therefore for most of the study region past projections did not extrapolate much outside of the pre-sent climate extent used to build the model Based on all time-period projections H lundii experienced a substantial range fluctuation across the late Quaternary with the models departing considerably from each other The range predicted from the mid-Holocene (6 kya) was the most similar to the current range while predictions from the LGM and LIG greatly departed from each other and from the current range (Figure 5 Supporting Information Figures S8 and S9) The larg-est predicted range occurred during the LGM including many sites not currently occupied by the species (Figure 5) This indicates an increase in suitable habitat during the dry and cold LGM In contrast the smallest predicted range occurred in the LIG indicating that this time period had a stronger effect constraining the distribution of H lundii compared to the glacial period of the LGM After obtaining a stability surface of all time periods we identified four main areas where the species has been continuously present since 120k years ago (areas in red in Figure 6) These four main areas of higher climatic stability largely overlay with the distribution of the four genetic clus-ters identified (Figure 6)

35emsp|emspTesting the effect of historical climate change on population structure

A multiple matrix regression between genetic and geographic resist-ance distances (including slope) among populations was significant (R2 = 0162 F = 4847 p = 0003) corroborating the effect of IBD on the genetic structure of this species (Figure 6) Because the climatic instability resistance is also strongly correlated with the geographic resistance matrix (R2 = 0248 F = 8258 p = 0001) we used the residuals of this correlation in a further combined analysis to add

F I G U R E 4 emsp Bayesian skyline plots (BSP) showing variation in effective population sizes through time (a) For all populations (b) For the Central lineage (c) For the Southeast lineage The dark thicker line represents the median population size and the thinner lines the 95 higher posterior probabilities [Colour figure can be viewed at wileyonlinelibrarycom]

0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)


the effects of climatic instability Thus a multiple matrix regression model combining both geographic resistance and climatic instability resistance was used to proper assess isolation by instability while controlling for the effect of geographic distance and topographic

complexity in H lundii genetic differentiation in the landscape The combined regression model was significant and explained more ge-netic variation than the models of each component alone (R2 = 0541 F = 14737 p = 0001) with both variables contributing significantly

F I G U R E 5 emsp Species distribution models of Hypsiboas lundii using random forest under the current climate and projected to three different times in the past the mid-Holocene the Last Glacial Maximum (LGM) and the Last Interglacial Maxima (LIG) The LGM and mid-Holocene projections consist of an ensemble of three different General Circulation Models (CCSM4 MIROC-ESM MPI-ESM-P) For each time period the presence probability is shown on a scale from 0 to 1

minus55 minus50 minus45 minus40

minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10

MidminusHolocene (minus6 ky)

02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10

F I G U R E 6 emsp Historical climatic suitability surface generated by overlaying the four species distribution models in Figure 5 ranging from 0 (not present in any of the four time periods) to 4 (present during all four periods) Higher values mean higher historical stability and population persistence over time Populations are represented by symbols corresponding to their genetic clusters assignment Note the overlap of the four main stable areas in red with the distribution of the four genetic clusters Plots on the right show correlations between genetic distance and climatic stability resistance (isolation by instability) and between geographic and genetic distance (isolation by distance) and between climatic stability resistance and genetic distance (isolation by instability)00 01 02 03 04 05 06 07

00

02

04

06

08

10

Isolation by instability

Climatic stability resistance

Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10

Isolation by distance

Geographic distance (Km)

Gen

etic

dis

tanc

e (F

st)

ndash55 ndash50 ndash45 ndash40

ndash25

ndash20

ndash15

ndash10

Historical climatic suitability surface

0

1

2

3

4

0 300 600

km


to the final combined model (climatic instability β = 958 p = 0001 geographic resistance β = 088 p = 0001) The variance partitioning analysis indicated that climatic instability explained 70 of the vari-ance in the combined model followed by geographic resistance (in-cluding slope) explaining 30 (Supporting Information Figure S10)

4emsp |emspDISCUSSION

We recovered a strong spatial genetic structure in Hypsiboas lundii a treefrog endemic to Cerrado with four well-delimited genetic clus-ters which also formed well-supported clades except for the Central-West cluster Genome-wide SNP data and mitochondrial DNA both support this genetic structure with great level of consistency among markers demonstrating a well-established structure across the land-scape Using a multiple matrix regression approach we found support for the role of recent past climatic changes structuring populations in space by negatively affecting population persistence in climatically unstable areas Our study is not the first to detect a significant role of past climatic fluctuations structuring populations in the Cerrado region Diniz-Filho et al (2015) found genetic differentiation (FST) and low levels of heterozygosity significantly associated with shifts in cli-matic suitability since the Last Glacial Maximum (LGM) for Eugenia dysenterica an endemic Cerrado tree Studies in other tropical or sub-tropical areas have also recently found a significant association between genetic structure and climatic stable areas over the last glaciations (Bell et al 2010 Carnaval amp Bates 2007 Devitt Devitt Hollingsworth McGuire amp Moritz 2013 Gugger Ikegami amp Sork 2013 He et al 2013 Ortego et al 2014) Therefore isolation by instability (IBI) over the late Quaternary might be prevalent in places mildly affected by glaciations promoting population differentiation

Our IBI model was supported even when considering the geo-graphic distance among populations In fact it explained more of the genetic variation in the landscape than geography alone (Supporting Information Figure S10) Yet isolation by distance (IBD) still has a strong effect on population differentiation which was somewhat ex-pected given the widespread distribution of this treefrog the usual philopatry exhibited by anurans (Wells 2007) and the ubiquity of IBD among organisms (Meirmans 2012) A pattern of IBD can in many instances produce misleading hierarchical genetic clusters with the identification of delimited clusters even when none is existent (Meirmans 2012) especially when sampling is biased across the land-scape (Schwartz amp McKelvey 2009) Our study avoided this poten-tial bias by comprehensively and systematically sampling populations across the entire distribution and we found strong support for four genetic clusters using multiple methods including Bayesian clustering and phylogenetic analyses Nonetheless structure is often able to de-tect clinal variation whenever it is present (Chen Durand Forbes amp Franccedilois 2007) strongly suggesting that the clusters identified here are not the spurious result of clinal gradients For example the incon-gruent phylogenetic placement of populations in the Central-West was recognized by structure suggesting an admixed ancestry coefficient for some individuals at localities 4 6 and 7 in this cluster (Figure 2)

The incongruent relationship of some Central-Western popu-lations in the mitochondrial and SNP phylogenies (Figure 3) may have resulted from either some level of admixture across the Central-West and the West clusters or from the maintenance of shared ancestral polymorphism (Muir amp Schloumltterer 2004) This pattern was only observed in populations of the Central-West cluster Biogeographic discordance between the mitochondria and nucleus is widespread among animals and most commonly results from geographic isolation with varied levels of mitochondrial in-trogression after secondary contact (Toews amp Brelsford 2012) Mito-nuclear conflicts have been reported for several other frog species that show a geographic pattern consistent with recent or ancestral mitochondrial introgression (Toews amp Brelsford 2012 and references therein Bryson amp Smith 2014) This discordance might be facilitated by the smaller effective population size of the haploid maternally inherited mitochondria for which even low levels of gene flow might be sufficient to establish mitochondrial introgression more rapidly than nuclear markers producing the biogeographic conflict (Chan amp Levin 2005)

Yet the high general congruence of our reduced genomic data set and a mitochondrial gene indicates that the pervasive effects of gene flow in the mitochondria are not observed among any other ge-netic clusters Even when considering the relative recent divergence time of all genetic clusters (06ndash21 Mya) the lineage sorting pat-tern shown in the mitochondrial ND2 gene and the genomic SNPs are mostly congruent (Figure 3) Although RADseq provides an un-precedented opportunity to sample thousands of unlinked markers throughout the genome representing a far more powerful genetic sampling method that can illuminate on processes of gene flow and hybridization (Eaton Hipp Gonzaacutelez-Rodriacuteguez amp Cavender-Bares 2015) the usefulness of mtDNA to find genetic clusters in the land-scape cannot be fully disregarded (Hung et al 2016)

Hypsiboas lundii experienced a drastic distributional shift throughout the late Quaternary (Figure 5) Our projection of a widespread range during the LGM was not expected given that the Cerrado distribution is believed to have contracted during this time (Werneck Nogueira et al 2012b) Species distribution models are affected by many sources of uncertainty that when applied to an entire ecoregion such as the Cerrado savanna (Werneck Nogueira et al 2012b) may fail to predict species-specific climatic responses Moreover it is unlikely that an entire biome will respond as a sin-gle unit (Collevatti et al 2012) This is especially relevant for the highly heterogeneous Cerrado landscape with different vegeta-tion formations ranging from open grasslands to gallery forests Therefore different responses to climate change are expected for species inhabiting contrasting habitats with different environmental requirements Hypsiboas lundii mostly occurs in gallery forests the most mesic habitat of the Cerrado savanna and its response might be only representative of organisms in this environment Ledo and Colli (2017) for example showed that LGM favoured the expansion of forest habitats allowing a connection between the Atlantic Forest and the Amazon through a southeast-northwest route The gallery forests in the Cerrado might have followed this pattern given that


our LGM model was also more widespread towards the western Cerrado Further studies modelling the response of Cerrado organ-isms occupying contrasting habitats are warranted to assess the con-gruence of species range predictions under shifting climates

Our Bayesian skyline plots indicated a pronounced population expansion between the LIG (120 kya) and the LGM (21 kya) pro-viding independent evidence for range expansion of H lundii in the LGM Interestingly a phylogeographic study of H albopunctatus a related Cerrado treefrog associated with open habitats also found evidence for population expansion during the Pleistocene (at 89 kya) in the same regions identified herein the Central and the Southeast (Prado et al 2012) Therefore Cerrado treefrogs might be respond-ing in a similar fashion to fluctuations in climate regardless of the associated habitat type (eg gallery forest or typical savanna vege-tation sensu stricto)

Although the cold and dry LGM was initially proposed to have restricted the distribution of forest species in South America a recent study has challenged this view Leite et al (2016) showed the predicted distribution for several Atlantic Forest species to be displaced north (warmer) towards lower elevation and less topo-graphically complex areas allowing a more widespread distribution than the present or the LIG Our results support this trend also for Cerrado species at least the ones inhabiting mesic habitats of gal-lery forests During the LGM the distribution of H lundii expanded northwest (Figure 5) towards a more topographically flat landscape In contrast the warm and dry LIG was the period predicted to have most strongly constrained the distribution of this treefrog However the complete geographical separation of the Central-West from the remaining clusters by unsuitable conditions did not happen until the mid-Holocene which may have contributed to the conflicting pat-tern found in the Central-West cluster

The geographic structure of Hypsiboas lundii does not seem congruent with any current discrete barrier to gene flow such as topographic barriers (large rivers or river basins) that separate the geographic clusters In fact when compared to other South American ecoregions such as the Amazon (with its massive river barriers) the Atlantic Forest (with a steep latitudinal climatic gra-dient) or the Andes (with great topographic complexity and ele-vation range) the central-southern part of Cerrado seems less impacted by barriers to dispersal Nevertheless other studies in-vestigating the population structure of plants frogs and lizards in the Cerrado have found a similar pattern of strong genetic differ-entiation in the west-east or northwest-southeast axis (Barbosa et al 2015 Guarnizo et al 2016 Prado et al 2012 Ramos Lemos-Filho amp Lovato 2008 Ramos et al 2007 Santos et al 2014) without recognizing any conspicuous geographical barrier separating them The west-east pattern suggests that historical climatic suitability might have forced distribution shifts along this main axis which were important for delimiting the genetic struc-ture for many species Overall this pattern supports our interpre-tation that distribution shifts during climatic oscillations helped fuel recent genetic differentiation in this savanna that has not been overwritten by contemporary gene flow Interestingly the

four genetic clusters found in this study are also highly congru-ent with four of the recognized biogeographic units for endemic anuran and squamate species in the Cerrado (Azevedo Valdujo amp de C Nogueira C 2016) showing that areas of stable climate can shape not only genetic structure in this savanna but also endemic diversity in congruent areas

5emsp |emspCONCLUSION

The impact of Quaternary climatic fluctuations in tropical regions has been a controversial issue over the last decades Our study shows that climate has been unstable over the late Quaternary in tropical South America adding to the large body of evidence showing climatic instability in tropical regions around the globe As demonstrated here for a treefrog species this instability pro-moted range shifts of animals and plants negatively impacting gene flow among populations separated by areas not continuously suitable over time thus promoting their differentiation across the landscape We refer to this model as ldquoisolation by instabilityrdquo We found a significant pattern of isolation by distance but isolation by instability was still significant even after controlling for the effect of topographic corrected geographic distance among populations In other words population structure in the treefrog Hypsiboas lun‐dii was determined by the added contribution of climatic instabil-ity over the late Quaternary and the geographic distance among its populations

The Cerrado savanna is a global hotspot of biodiversity in South America but it is only recently that the evolutionary and ecologi-cal drivers of diversification in this savanna have been investigated (Simon et al 2009 Werneck 2011) Here we demonstrate that past climatic fluctuations were important in driving intraspecific genetic structure in this habitat as has already been demonstrated for the Atlantic Forest (Cabanne et al 2016 Carnaval et al 2009) Genetic differentiation in Hypsiboas lundii was positively correlated with the climatic instability resistance among populations a metric based on the persistence of populations across time using past predictions of species distribution models over the late Quaternary Understanding how species in biodiverse regions responded to past climate changes can be relevant considering the current scenario of climate change and global warming The conditions under which populations re-sponded to fluctuations in climate and habitat instability might give us perspective on how species will cope with the imminent climate warming

ACKNOWLEDG EMENTS

We thank the Brazilian herpetological collections CFBH CHUNB MTR MZUSP PUC-MG UFMT and ZUFG for tissue loans that al-lowed us to include more locality sites in our study We thank Scott Hunicke-Smith and Jessica Podnar at the Genomic Sequencing and Analysis Facility (GSAF-UT) Dan Bolnick and Tom Juenger at UT for their invaluable help during development of the laboratory protocol


Dennis Willie and especially Eric Ortego at If Center for Computational Biology and Bioinformatics (CCBB-UT) provided extensive support in bioinformatics We thank Eric Barbalho Tamara Vieira Iacutesis Arantes Mario Van Gastel and students in the Rodrigues and the Colli lab for assistance in field expeditions and collections We are also grateful to Paula Valdujo for sharing collection data Reuber Brandatildeo for advice during planning of field expeditions and the Brazilian agency ICMBio for collection permits (SISBIO3794-13324 and SISBIO30325-1 to MMV) MMV acknowledges funding from the National Science Foundation (DDIG award DEB 1311517) and fellowships from CAPESFulbright the Zoology Scholarship Endowment for Excellence and the Graduate Deans Prestigious Fellowship from UT MTR thanks FAPESP (Fundaccedilatildeo de Amparo agrave Pesquisa do Estado de Satildeo Paulo) and CNPq for support GRC thanks Coordenaccedilatildeo de Aperfeiccediloamento de Pessoal de Niacutevel Superior (CAPES) Conselho Nacional de Desenvolvimento Cientiacutefico e Tecnoloacutegico (CNPq) Fundaccedilatildeo de Apoio agrave Pesquisa do Distrito Federal (FAPDF) and the United States Agency for International Development (USAID) Partnerships for Enhanced Engagement in Research (PEER) program under cooperative agree-ment AID-OAA-A-11-00012 for financial support

AUTHOR CONTRIBUTIONS

MMV GRC and DCC designed the study MMV GRC and MTR con-ducted field collections MMV and JNW performed the lab work MMV and EMO analysed the data MMV JNW and EMO interpreted the results MMV wrote the manuscript with contributions from GRC JNW EMO MTR and DCC All authors reviewed and contrib-uted to the final version of this manuscript

DATA ACCE SSIBILIT Y

Demultiplexed ddRAD sequencing reads are available at the NCBI Short Read Archive (httpswwwncbinlmnihgovsraPRJNA521095) ND2 mitochondrial sequences are available on GenBank (Accession numbers MF279001ndashMF279043) Presence points used for the species distribution models are available in Dryad (httpsdoiorg105061dryad17b7mt1) as well as all matri-ces used in the multiple matrix regression analyses genetic distance climatic stability resistance geographic resistance (including slope) and geographic distance (in Km)

ORCID

Mariana M Vasconcellos httpsorcidorg0000-0002-3175-2649

Guarino R Colli httpsorcidorg0000-0002-2628-5652

Jesse N Weber httpsorcidorg0000-0003-4839-6684

Edgardo M Ortiz httpsorcidorg0000-0001-8052-1671

Miguel T Rodrigues httpsorcidorg0000-0003-3958-9919

David C Cannatella httpsorcidorg0000-0001-8675-0520

R E FE R E N C E S

Azevedo J A R Valdujo P H amp Nogueira C (2016) Biogeography of anurans and squamates in the Cerrado hotspot Coincident endemism patterns in the richest and most impacted savanna on the globe Journal of Biogeography 43 2454ndash2464 httpsdoiorg101111jbi12803

Barbet-Massin M Jiguet F Albert C H amp Thuiller W (2012) Selecting pseudo-absences for species distribution models How where and how many Methods in Ecology and Evolution 3 327ndash338 httpsdoiorg101111j2041-210X201100172x

Barbosa A C O F Collevatti R G Chaves L J Guedes L B S Diniz-Filho J A F amp Telles M P C (2015) Range-wide genetic differ-entiation of Eugenia dysenterica (Myrtaceae) populations in Brazilian Cerrado Biochemical Systematics and Ecology 59 288ndash296 httpsdoiorg101016jbse201502004

Bell R C Parra J L Tonione M Hoskin C J Mackenzie J B Williams S E amp Moritz C (2010) Patterns of persistence and isolation indicate resilience to climate change in montane rain-forest lizards Molecular Ecology 19 2531ndash2544 httpsdoiorg101111j1365-294X201004676x

Besnier F amp Glover K A (2013) ParallelStructure A R package to distribute parallel runs of the population genetics program STRUCTURE on multi-core computers PLoS One 8 e70651 httpsdoiorg101371journalpone0070651

Bonaccorso E Koch I amp Peterson A T (2006) Pleistocene fragmenta-tion of Amazon species ranges Diversity and Distributions 12 157ndash164 httpsdoiorg101111j1366-9516200500212x

Breiman L (2001) Random forests Machine Learning 45 5ndash32Bryson R W Smith B T Nieto-Montes de Oca A Garciacutea-Vaacutezquez U O

amp Riddle B R (2014) The role of mitochondrial introgression in illumi-nating the evolutionary history of Nearctic treefrogs Zoological Journal of the Linnean Society 172 103ndash116 httpsdoiorg101111zoj12169

Bush M B (1994) Amazonian speciation A necessarily complex model Journal of Biogeography 21 5ndash17 httpsdoiorg1023072845600

Cabanne G S Calderoacuten L Trujillo Arias N Flores P Pessoa R dHorta F M amp Miyaki C Y (2016) Effects of Pleistocene cli-mate changes on species ranges and evolutionary processes in the Neotropical Atlantic Forest Biological Journal of the Linnean Society 119 856ndash872 httpsdoiorg101111bij12844

Carnaval A C amp Bates J M (2007) Amphibian DNA shows marked genetic structure and tracks Pleistocene climate change in northeastern Brazil Evolution 61 2942ndash2957 httpsdoiorg101111j1558-5646200700241x

Carnaval A C Hickerson M J Haddad C F B Rodrigues M T amp Moritz C (2009) Stability predicts genetic diversity in the Brazilian Atlantic Forest hotspot Science 323 785ndash789 httpsdoiorg101126science1166955

Carnaval A C Waltari E Rodrigues M T Rosauer D VanDerWal J Damasceno R hellip Moritz C (2014) Prediction of phylogeographic en-demism in an environmentally complex biome Proceedings of the Royal Society B‐Biological Sciences 281 20141461 httpsdoiorg101098rspb20141461

Chan K M A amp Levin S A (2005) Leaky prezygotic isolation and porous genomes Rapid introgression of maternally inherited DNA Evolution 59 720ndash729 httpsdoiorg101111j0014-38202005tb01748x

Chen C Durand E Forbes F amp Franccedilois O (2007) Bayesian cluster-ing algorithms ascertaining spatial population structure A new com-puter program and a comparison study Molecular Ecology Notes 7 747ndash756 httpsdoiorg101111j1471-8286200701769x

Chifman J amp Kubatko L (2014) Quartet inference from SNP data under the coalescent model Bioinformatics 30 3317ndash3324 httpsdoiorg101093bioinformaticsbtu530

Colinvaux P A De Oliveira P E amp Bush M B (2000) Amazonian and neotropical plant communities on glacial time-scales The failure of


the aridity and refuge hypotheses Quaternary Science Reviews 19 141ndash169 httpsdoiorg101016S0277-3791(99)00059-1

Collevatti R G Terribile L C de Oliveira G Lima-Ribeiro M S Nabout J C Rangel T F amp Diniz-Filho J A F (2012) Drawbacks to palaeodis-tribution modelling The case of South American seasonally dry forests Journal of Biogeography 40 345ndash358 httpsdoiorg101111jbi12005

Danecek P Auton A Abecasis G Albers C A Banks E DePristo M A hellip Durbin R (2011) The variant call format and VCFtools Bioinformatics 27 2156ndash2158 httpsdoiorg101093bioinformaticsbtr330

Davis M B amp Shaw R G (2001) Range shifts and adaptive responses to Quaternary climate change Science 292 673ndash679 httpsdoiorg101126science2925517673

Deutsch C A Tewksbury J J Huey R B Sheldon K S Ghalambor C K Haak D C amp Martin P R (2008) Impacts of climate warming on terrestrial ectotherms across latitude Proceedings of the National Academy of Sciences of the United States of America 105 6668ndash6672 httpsdoiorg101073pnas0709472105

Devitt T J Devitt S E C Hollingsworth B D McGuire J A amp Moritz C (2013) Montane refugia predict population genetic struc-ture in the Large-blotched Ensatina salamander Molecular Ecology 22 1650ndash1665 httpsdoiorg101111mec12196

Diniz-Filho J A F Barbosa A C O F Collevatti R G Chaves L J Terribile L C Lima-Ribeiro M S amp Telles M P C (2015) Spatial autocorrelation analysis and ecological niche modelling allows infer-ence of range dynamics driving the population genetic structure of a Neotropical savanna tree Journal of Biogeography 43 167ndash177 httpsdoiorg101111jbi12622

Domingos F M C B Bosque R J Cassimiro J Colli G R Rodrigues M T Santos M G amp Beheregaray L B (2014) Out of the deep Cryptic speciation in a Neotropical gecko (Squamata Phyllodactylidae) revealed by species delimitation methods Molecular Phylogenetics and Evolution 80 113ndash124 httpsdoiorg101016jympev201407022

Drummond A J Rambaut A Shapiro B amp Pybus O G (2005) Bayesian coalescent inference of past population dynamics from mo-lecular sequences Molecular Biology and Evolution 22 1185ndash1192 httpsdoiorg101093molbevmsi103

Drummond A J Suchard M A Xie D amp Rambaut A (2012) Bayesian phylogenetics with BEAUti and the BEAST 17 Molecular Biology and Evolution 29 1969ndash1973 httpsdoiorg101093molbevmss075

Dyer R J Nason J D amp Garrick R C (2010) Landscape mod-elling of gene flow Improved power using conditional ge-netic distance derived from the topology of population networks Molecular Ecology 19 3746ndash3759 httpsdoiorg101111j1365-294X201004748x

Earl D A amp vonHoldt B M (2011) STRUCTURE HARVESTER A web-site and program for visualizing STRUCTURE output and implement-ing the Evanno method Conservation Genetics Resources 4 359ndash361 httpsdoiorg101007s12686-011-9548-7

Eaton D A R (2014) PyRAD Assembly of de novo RADseq loci for phylogenetic analyses Bioinformatics 30 1844ndash1849 httpsdoiorg101093bioinformaticsbtu121

Eaton D A R Hipp A L Gonzaacutelez-Rodriacuteguez A amp Cavender-Bares J (2015) Historical introgression among the American live oaks and the comparative nature of tests for introgression Evolution 69 2587ndash2601 httpsdoiorg101111evo12758

Edgar R C (2004) MUSCLE Multiple sequence alignment with high ac-curacy and high throughput Nucleic Acids Research 32 1792ndash1797 httpsdoiorg101093nargkh340

Ehlers J Gibbard P L amp Hugues P D (2011) Quaternary glaciations ndash Extent and chronology a closer look (1st ed) Amsterdam The Netherlands Elsevier

Elith J Kearney M amp Phillips S (2010) The art of modelling range-shifting species Methods in Ecology and Evolution 1 330ndash342 httpsdoiorg101111j2041-210X201000036x

Epps C W amp Keyghobadi N (2015) Landscape genetics in a chang-ing world Disentangling historical and contemporary influences and inferring change Molecular Ecology 24 6021ndash6040 httpsdoiorg101111mec13454

Evanno G Regnaut S amp Goudet J (2005) Detecting the num-ber of clusters of individuals using the software STRUCTURE A simulation study Molecular Ecology 14 2611ndash2620 httpsdoiorg101111j1365-294X200502553x

Falush D Stephens M amp Pritchard J K (2003) Inference of popula-tion structure using multilocus genotype data Linked loci and cor-related allele frequencies Genetics 164 1567ndash1587

Goslee S C amp Urban D L (2007) The ecodist package for dissimilar-ity-based analysis of ecological data Journal of Statistical Software 22 1ndash19 httpsdoiorg1018637jssv022i07

Graham C H Moritz C amp Williams S E (2006) Habitat history im-proves prediction of biodiversity in rainforest fauna Proceedings of the National Academy of Sciences of the United States of America 103 632ndash636 httpsdoiorg101073pnas0505754103

Guarnizo C E Werneck F P Giugliano L G Santos M G Fenker J Sousa L hellip Colli G R (2016) Cryptic lineages and diversification of an endemic anole lizard (Squamata Dactyloidae) of the Cerrado hotspot Molecular Phylogenetics and Evolution 94 279ndash289 httpsdoiorg101016jympev201509005

Gugger P F Ikegami M amp Sork V L (2013) Influence of late Quaternary climate change on present patterns of genetic variation in valley oak Quercus lobata Neacutee Molecular Ecology 22 3598ndash3612 httpsdoiorg101111mec12317

Guillot G amp Rousset F (2013) Dismantling the Mantel tests Methods in Ecology and Evolution 4 336ndash344 httpsdoiorg1011112041-210x12018

Haffer J (1969) Speciation in amazonian forest birds Science 165 131ndash137 httpsdoiorg101126science1653889131

He Q Edwards D L amp Knowles L L (2013) Integrative testing of how environments from the past to the present shape genetic structure across landscapes Evolution 67 3386ndash3402 httpsdoiorg101111evo12159

Heller R Chikhi L amp Siegismund H R (2013) The confounding ef-fect of population structure on Bayesian skyline plot inferences of demographic history PLoS One 8 e62992 httpsdoiorg101371journalpone0062992

Hewitt G M (2000) The genetic legacy of the Quaternary ice ages Nature 405 907ndash913 httpsdoiorg10103835016000

Hewitt G M (2004) Genetic consequences of climatic oscillations in the Quaternary Philosophical Transactions of the Royal Society B Biological Sciences 359 183ndash195 httpsdoiorg101098rstb20031388

Hijmans R J Cameron S E Parra J L Jones P G amp Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas International Journal of Climatology 25 1965ndash1978 httpsdoiorg101002joc1276

Hijmans R J Phillips S J amp Leathwick J R (2017) dismo Species dis‐tribution modeling R package version 11-4 Retrieved from httpscranr-projectorgpackage=dismo

Huang H amp Knowles L L (2016) Unforeseen consequences of excluding missing data from next-generation sequences Simulation study of RAD sequences Systematic Biology 65 357ndash365 httpsdoiorg101093sysbiosyu046

Hung C-M Drovetski S V amp Zink R M (2016) Matching loci sur-veyed to questions asked in phylogeography Proceedings of the Royal Society of London Series B‐Biological Sciences 283(1826) 20152340ndash20152348 httpsdoiorg101098rspb20152340

Kearse M Moir R Wilson A Stones-Havas S Cheung M Sturrock S hellip Drummond A (2012) Geneious basic An integrated and ex-tendable desktop software platform for the organization and anal-ysis of sequence data Bioinformatics 28 1647ndash1649 httpsdoiorg101093bioinformaticsbts199


Kierepka E M amp Latch E K (2015) Performance of partial statistics in individual-based landscape genetics Molecular Ecology Resources 15 512ndash525 httpsdoiorg1011111755-099812332

Klink C A amp Machado R B (2005) Conservation of the Brazilian Cerrado Conservation Biology 19 707ndash713 httpsdoiorg101111j1523-1739200500702x

Kopelman N M Mayzel J Jakobsson M Rosenberg N A amp Mayrose I (2015) CLUMPAK A program for identifying clustering modes and packaging population structure inferences across K Molecular Ecology Resources 15 1179ndash1191 httpsdoiorg1011111755-099812387

Lanfear R Calcott B Ho S Y W amp Guindon S (2012) PartitionFinder Combined selection of partitioning schemes and substitution models for phylogenetic analyses Molecular Biology and Evolution 29 1695ndash1701 httpsdoiorg101093molbevmss020

Leacheacute A D Banbury B L Felsenstein J Nieto-Montes de Oca A amp Stamatakis A (2015) Short tree long tree right tree wrong tree New acquisition bias corrections for inferring SNP phylogenies Systematic Biology 64 1032ndash1047 httpsdoiorg101093sysbiosyv053

Ledo R M D amp Colli G R (2017) The historical connections between the Amazon and the Atlantic Forest revisited Journal of Biogeography 312 673ndash613 httpsdoiorg101111jbi13049

Legendre P amp Fortin M J (2010) Comparison of the Mantel test and al-ternative approaches for detecting complex multivariate relationships in the spatial analysis of genetic data Molecular Ecology Resources 10 831ndash844 httpsdoiorg101111j1755-0998201002866x

Legendre P Lapointe F J amp Casgrain P (1994) Modeling brain evolu-tion from behavior A permutational regression approach Evolution 48 1487ndash1499 httpsdoiorg101111j1558-56461994tb02191x

Leite Y L R Costa L P Loss A C Rocha R G Batalha-Filho H Bastos A C hellip Pardini R (2016) Neotropical forest expansion during the last glacial period challenges refuge hypothesis Proceedings of the National Academy of Sciences of the United States of America 113 1008ndash1013 httpsdoiorg101073pnas1513062113

Lewis P O (2001) A likelihood approach to estimating phylogeny from discrete morphological character data Systematic Biology 50 913ndash925 httpsdoiorg101080106351501753462876

Liaw A amp Wiener M (2002) Classification and regression by random-Forest R News 2 18ndash22

Macey J R Strasburg J L Brisson J A Vredenburg V T Jennings M amp Larson A (2001) Molecular phylogenetics of western North American frogs of the Rana boylii species group Molecular Phylogenetics and Evolution 19 131ndash143 httpsdoiorg101006mpev20000908

Malinsky M Trucchi E Lawson D J amp Falush D (2018) RADpainter and fineRADstructure Population inference from RADseq data Molecular Biology and Evolution 35 1284ndash1290 httpsdoiorg101093molbevmsy023

Mayle F E Beerling D J Gosling W D amp Bush M B (2004) Responses of Amazonian ecosystems to climatic and atmospheric carbon dioxide changes since the last glacial maximum Philosophical Transactions of the Royal Society B Biological Sciences 359 499ndash514 httpsdoiorg101098rstb20031434

Mazzarelli C C M (2015) Biologia reprodutiva de Hypsiboas lundii (Anura Hylidae) em um fragmento de Cerrado no sudoeste de Minas Gerais Unpublished Masterrsquos thesis Universidade Estadual Paulista - Rio Claro

McRae B H (2006) Isolation by resistance Evolution 60 1551ndash1561 httpsdoiorg101111j0014-38202006tb00500x

McRae B H Dickson B G Keitt T H amp Shah V B (2008) Using cir-cuit theory to model connectivity in ecology evolution and conser-vation Ecology 89 2712ndash2724 httpsdoiorg10189007-18611

Meirmans P G (2012) The trouble with isolation by distance Molecular Ecology 21 2839ndash2846 httpsdoiorg101111j1365-294X201205578x

Muir G amp Schloumltterer C (2004) Evidence for shared ancestral polymor-phism rather than recurrent gene flow at microsatellite loci differ-entiating two hybridizing oaks (Quercus spp) Molecular Ecology 14 549ndash561 httpsdoiorg101111j1365-294X200402418x

Myers N Mittermeier R A Mittermeier C G da Fonseca G A B amp Kent J (2000) Biodiversity hotspots for conservation priorities Nature 403 853ndash858 httpsdoiorg10103835002501

Nascimento F F Lazar A Menezes A N Durans A M Moreira J C Salazar-Bravo J hellip Bonvicino C R (2013) The role of historical barriers in the diversification processes in open vegetation formations during the MiocenePliocene using an ancient rodent lineage as a model PLoS One 8 e61924 httpsdoiorg101371journalpone0061924

Nelson B W Ferreira C A C da Silva M F amp Kawasaki M L (1990) Endemism centres refugia and botanical collection density in Brazilian Amazonia Nature 345 714ndash716 httpsdoiorg1010383457 14a0

Nogueira C Ribeiro S Costa G C amp Colli G R (2011) Vicariance and endemism in a Neotropical savanna hotspot Distribution patterns of Cerrado squamate reptiles Journal of Biogeography 38 1907ndash1922 httpsdoiorg101111j1365-2699201102538x

Novaes R M L de Lemos J P Ribeiro R A amp Lovato M B (2010) Phylogeography of Plathymenia reticulata (Leguminosae) reveals pat-terns of recent range expansion towards northeastern Brazil and south-ern cerrados in eastern Tropical South America Molecular Ecology 19 985ndash998 httpsdoiorg101111j1365-294X201004530x

Ortego J Gugger P F amp Sork V L (2014) Climatically stable land-scapes predict patterns of genetic structure and admixture in the Californian canyon live oak Journal of Biogeography 42 328ndash338 httpsdoiorg101111jbi12419

Otto-Bliesner B L Marshall S J Overpeck J T Miller G H amp Hu A (2006) Simulating Arctic climate warmth and icefield retreat in the last interglaciation Science 311 1751ndash1753 httpsdoiorg101126science1120808

Pattengale N D Alipour M Bininda-Emonds O R P Moret B M E amp Stamatakis A (2010) How many bootstrap replicates are nec-essary Journal of Computational Biology 17 337ndash354 httpsdoiorg101089cmb20090179

Peterson B K Weber J N Kay E H Fisher H S amp Hoekstra H E (2012) Double digest RADseq An inexpensive method for de novo SNP discovery and genotyping in model and non-model species PLoS One 7 e37135 httpsdoiorg101371journalpone0037135

Prado C P A Haddad C F B amp Zamudio K R (2012) Cryptic lineages and Pleistocene population expansion in a Brazilian Cerrado frog Molecular Ecology 21 921ndash941 httpsdoiorg101111j1365-294X201105409x

Pritchard J K Stephens M amp Donnelly P (2000) Inference of pop-ulation structure using multilocus genotype data Genetics 155 945ndash959

Puritz J B Matz M V Toonen R J Weber J N Bolnick D I amp Bird C E (2014) Demystifying the RAD fad Molecular Ecology 23 5937ndash5942 httpsdoiorg101111mec12965

Ramos A C S Lemos-Filho J P amp Lovato M B (2008) Phylogeographical structure of the Neotropical forest tree Hymenaea courbaril (Leguminosae Caesalpinioideae) and its relation-ship with the vicariant Hymenaea stigonocarpa from Cerrado Journal of Heredity 100 206ndash216 httpsdoiorg101093jheredesn092

Ramos A C S Lemos-Filho J P Ribeiro R A Santos F R amp Lovato M B (2007) Phylogeography of the tree Hymenaea stigonocarpa (Fabaceae Caesalpinioideae) and the influence of Quaternary cli-mate changes in the Brazilian Cerrado Annals of Botany 100 1219ndash1228 httpsdoiorg101093aobmcm221

Renaud G Stenzel U Maricic T Wiebe V amp Kelso J (2015) deML Robust demultiplexing of Illumina sequences using a likelihood-based approach Bioinformatics 31 770ndash772 httpsdoiorg101093bioinformaticsbtu719

Roberts D R Bahn V Ciuti S Boyce M S Elith J Guillera-Arroita G hellip Dormann C F (2017) Cross-validation strategies for data with temporal spatial hierarchical or phylogenetic structure Ecography 40 913ndash929 httpsdoiorg101111ecog02881


Rousset F (1997) Genetic differentiation and estimation of gene flow from F-statistics under isolation by distance Genetics 145 1219ndash1228

Rull V (2008) Speciation timing and neotropical biodiversity The Tertiary-Quaternary debate in the light of molecular phyloge-netic evidence Molecular Ecology 17 2722ndash2729 httpsdoiorg101111j1365-294X200803789x

Saia S E M G Pessenda L C R Gouveia S E M Aravena R amp Bendassolli J A (2008) Last glacial maximum (LGM) vegetation changes in the Atlantic Forest southeastern Brazil Quaternary International 184 195ndash201 httpsdoiorg101016jquaint200706029

Santos M G Nogueira C Giugliano L G amp Colli G R (2014) Landscape evolution and phylogeography of Micrablepharus at‐ticolus (Squamata Gymnophthalmidae) an endemic lizard of the Brazilian Cerrado Journal of Biogeography 41 1506ndash1519 httpsdoiorg101111jbi12291

Schwartz M K amp McKelvey K S (2009) Why sampling scheme mat-ters The effect of sampling scheme on landscape genetic results Conservation Genetics 10 441ndash452 httpsdoiorg101007s10592-008-9622-1

Schweyen H Rozenberg A amp Leese F (2014) Detection and re-moval of PCR duplicates in population genomic ddRAD studies by addition of a degenerate base region (DBR) in sequencing adapt-ers The Biological Bulletin 227 146ndash160 httpsdoiorg101086BBLv227n2p146

Silva J M C amp Bates J M (2002) Biogeographic patterns and conservation in the South American Cerrado A tropi-cal savanna hotspot BioScience 52 225ndash233 httpsdoiorg1016410006-3568(2002)052[0225BPACIT]20CO2

Simon M F Grether R de Queiroz L P Skema C Pennington R T amp Hughes C E (2009) Recent assembly of the Cerrado a neo-tropical plant diversity hotspot by in situ evolution of adaptations to fire Proceedings of the National Academy of Sciences of the United States of America 106 20359ndash20364 httpsdoiorg101073pnas0903410106

Stamatakis A (2014) RAxML version 8 A tool for phylogenetic analy-sis and post-analysis of large phylogenies Bioinformatics 30 1312ndash1313 httpsdoiorg101093bioinformaticsbtu033

Swofford D L (2002) PAUP Phylogenetic Analysis using Parsimony (and other methods) Version 4 Sunderland MA Sinauer Associates

Thomaz A T Malabarba L R Bonatto S L amp Knowles L L (2015) Testing the effect of palaeodrainages versus habitat stability on genetic divergence in riverine systems Study of a Neotropical fish of the Brazilian coastal Atlantic Forest Journal of Biogeography 42 2389ndash2401 httpsdoiorg101111jbi12597

Toews D P L amp Brelsford A (2012) The biogeography of mitochon-drial and nuclear discordance in animals Molecular Ecology 21 3907ndash3930 httpsdoiorg101111j1365-294X201205664x

Valdujo P H Silvano D L Colli G R amp Martins M (2012) Anuran species composition and distribution patterns in Brazilian Cerrado a Neotropical hotspot South American Journal of Herpetology 7 63ndash78 httpsdoiorg1029940570070209

Van der Hammen T (1974) The Pleistocene changes of vegetation and climate in tropical South America Journal of Biogeography 1 3ndash26 httpsdoiorg1023073038066

Vuilleumier B S (1971) Pleistocene changes in the fauna and flora of South America Science 173 771ndash780 httpsdoiorg101126science1733999771

Wang I J (2013) Examining the full effects of landscape heterogeneity on spatial genetic variation A multiple matrix regression approach for quantifying geographic and ecological isolation Evolution 67 3403ndash3411 httpsdoiorg101111evo12134

Weir B S amp Cockerham C C (1984) Estimating F-statistics for the analysis of population structure Evolution 38 1358ndash1370 httpsdoiorg1023072408641

Wells K D (2007) The ecology and behavior of Amphibians Chicago IL The University of Chicago Press

Werneck F P (2011) The diversification of eastern South American open vegetation biomes Historical biogeography and perspec-tives Quaternary Science Reviews 30 1630ndash1648 httpsdoiorg101016jquascirev201103009

Werneck F P Gamble T Colli G R Rodrigues M T amp Sites J W Jr (2012a) Deep diversification and long-term persistence in the South American dry diagonal Integrating continent-wide phylogeogra-phy and distribution modeling of geckos Evolution 66 3014ndash3034 httpsdoiorg101111j1558-5646201201682x

Werneck F P Nogueira C Colli G R Sites J W Jr amp Costa G C (2012b) Climatic stability in the Brazilian Cerrado Implications for biogeo-graphical connections of South American savannas species richness and conservation in a biodiversity hotspot Journal of Biogeography 39 1695ndash1706 httpsdoiorg101111j1365-2699201202715x

Willing E-M Dreyer C amp van Oosterhout C (2012) Estimates of genetic differentiation measured by FST do not necessarily require large sample sizes when using many SNP markers PLoS One 7 e42649 httpsdoiorg101371journalpone0042649

Willis K J amp Whittaker R J (2000) Perspectives paleoecology The refugial debate Science 287 1406ndash1407 httpsdoiorg101126science28754571406

Wright S (1943) Isolation by distance Genetics 28 114ndash138Zellmer A J amp Knowles L L (2009) Disentangling the effects of

historic vs contemporary landscape structure on population ge-netic divergence Molecular Ecology 18 3593ndash3602 httpsdoiorg101111j1365-294X200904305x

SUPPORTING INFORMATION

Additional supporting information may be found online in the Supporting Information section at the end of the article

How to cite this article Vasconcellos MM Colli GR Weber JN Ortiz EM Rodrigues MT Cannatella DC Isolation by instability Historical climate change shapes population structure and genomic divergence of treefrogs in the Neotropical Cerrado savanna Mol Ecol 2019281748ndash1764 httpsdoiorg101111mec15045


1emsp |emspINTRODUC TION

The climatic fluctuations of Pleistocene glacial cycles have greatly impacted the distribution of organisms imprinting on their popu-lations a genetic signature of such changes (Davis amp Shaw 2001 Hewitt 2000) Even though the genetic consequences of the ice ages are well documented for many organisms in the Northern Hemisphere (Hewitt 2000 2004) less evidence exists for the Southern Hemisphere (but see Cabanne et al 2016) possibly due to the reduced extent of glaciations in this region Although gla-ciers covered only small portions of South America most notably in the extreme southern region and in the Andes (Ehlers Gibbard amp Hugues 2011 Vuilleumier 1971) cycles of contraction and ex-pansion of forests and savannas during glacial and interglacial pe-riods are thought to have occurred (Van der Hammen 1974) and to have significantly affected vegetation structure and distribu-tion of organisms throughout the continent (Bonaccorso Koch amp Peterson 2006 Saia Pessenda Gouveia Aravena amp Bendassolli 2008)

The effect of Pleistocene climatic fluctuations on the South American biota has long been a controversial topic especially since it was first hypothesized to explain the large number of spe-cies in the Amazon (Bush 1994 Haffer 1969) The forest refugia hypothesis (Haffer 1969) posits that glacial periods promoted fragmentation of forest into isolated patches serving as faunal refuges which then spurred allopatric speciation Given that addi-tional studies did not support fragmentation of the Amazon forest (Colinvaux Oliveira amp Bush 2000) elevated speciation rates in the Pleistocene (Rull 2008) or refugia locations (Nelson Ferreira Silva amp Kawasaki 1990) the impact of Pleistocene climatic fluc-tuations was generally dismissed for Neotropical regions (Willis amp Whittaker 2000) Nonetheless climatic changes have been since then invoked to explain the marked genetic structure in Neotropical species in the Brazilian Atlantic Forest (Cabanne et al 2016 Carnaval Hickerson Haddad Rodrigues amp Moritz 2009) and the Cerrado (Diniz-Filho et al 2015 Prado Haddad amp Zamudio 2012) for example Historically stable habitats since the Pleistocene also seem to be an important predictor of endemic diversity (Carnaval et al 2014) species richness (Graham Moritz amp Williams 2006 Werneck Nogueira Colli Sites amp Costa 2012b) and genetic struc-ture in several other tropical regions (Bell et al 2010 Carnaval amp Bates 2007 He Edwards amp Knowles 2013) suggesting that range shifts associated with climatic cycles must have occurred in the tropics Moreover due to their narrow thermal tolerance tropical ectotherm species are expected to have been particularly affected by temperature shifts during climatic cycles (Deutsch et al 2008)

Fluctuation in species ranges associated with Pleistocene cli-matic cycles might strongly affect the configuration of popula-tions in space limiting gene flow and affecting genetic structure Although the current spatial genetic structure of populations can reflect both historical and contemporary ecological factors his-torical factors such as paleoclimate dynamics or barriers to gene flow that are not currently present (eg paleodrainages) seem to

have a long and lasting contribution to the current pattern of ge-netic differentiation (Epps amp Keyghobadi 2015) In many cases historical factors can better explain genetic variation across the landscape than current climate or recent barriers to gene flow (He et al 2013 Nascimento et al 2013 Ortego Gugger amp Sork 2014 Thomaz Malabarba Bonatto amp Knowles 2015) Therefore current low levels of admixture might not overwrite the genetic signatures of demographic and distributional changes due to past environmental fluctuations which have to be carefully considered whenever inferring current ecological correlates of gene flow (Dyer Nason amp Garrick 2010)

Understanding the factors limiting gene flow and structuring populations in space can also provide important insights about driv-ers of diversification and their relative contribution to high biodi-versity regions In particular the Cerrado which is the largest and most species-rich tropical savanna and a global hotspot of biodi-versity (Klink amp Machado 2005 Myers Mittermeier Mittermeier Fonseca amp Kent 2000) is remarkable for high endemicity levels reported for plants (44 Klink amp Machado 2005) frogs (53 Valdujo Silvano Colli amp Martins 2012) and squamate reptiles (39 Nogueira Ribeiro Costa amp Colli 2011) Yet the evolutionary and ecological processes that led to this high species diversity are still not well understood (Silva amp Bates 2002 Werneck 2011) This sa-vanna spreads over the central part of an open dry diagonal in South America separating the Amazon and the Atlantic forest and is com-posed of a mosaic of heterogeneous vegetation ranging from grass-land to riverine gallery forest While adaptation to fire seems to be an important and recent promoter of plant diversification in Cerrado (Simon et al 2009) geological events in particular the uplift of the Central Brazilian Shield are regarded as more important for diver-sification in vertebrates (Domingos et al 2014 Santos Nogueira Giugliano amp Colli 2014 Werneck Gamble Colli Rodrigues amp Sites 2012a) Even though some studies have associated intraspecific levels of differentiation with past climatic changes in the Cerrado (Diniz-Filho et al 2015 Novaes Lemos Ribeiro amp Lovato 2010 Prado et al 2012 Ramos Lemos-Filho Ribeiro Santos amp Lovato 2007) only one study (of an endemic tree) has explicitly tested for and found significant association between climate change and ge-netic structure (Diniz-Filho et al 2015)

Paleodistribution models for the Cerrado savanna predict a dy-namic history of vegetation fluctuation over the late Quaternary in-cluding both the Pleistocene and the Holocene (Werneck Nogueira et al 2012b) These models indicate that climate conditions during the Last Interglacial (LIG ~120 k years ago) which was 05ndash2degC warmer and slightly drier than present in the Cerrado were some-what favourable to the occurrence of Cerrado vegetation They also indicate a contraction in the extent of the Cerrado during the Last Glacial Maximum (LGM ~21 k years ago) which was 1ndash35degC colder and drier than the present More recently during the mid-Holocene (~6 k years ago) the climate and extent of the Cerrado vegetation did not substantially differ from present conditions (Werneck Nogueira et al 2012b) Although this response departs from the expectation of savanna expansion towards adjacent forested regions during the





































































ndash25

ndash20

ndash15

ndash10

0

500

1000

1500

2000

17

4

37

4746 45

18

7 9

23

14

43

44

31

34

19

13

35

526

81

33

2

28

11 2021 25

16

40

38

10

29

12

22 30

32

3

3639

27

6

4142

15

24

0 200 400

km

Elevation (m)


















3emsp |emspRESULTS












K = 4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

West Central West

Central Southeast

K = 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

K = 21 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

87654321K

ndash240000

ndash220000

ndash200000

ndash180000

ndash160000

ndash140000

Mea

n Ln

pro

b o

f dat

a


765432K

400

300

200

100

0

Del

ta K




alian05

alian07

altog02

araxa01assis01

bauru01

botuc01

brasi04

caiap02caiap06

campe01

catal01

catal10

caval02

caval04

claro01

cosmo01

crist03

curve02

desco01

domin01domin09

furna01

garca02

gauch02gauch04

goias02goias10

guima01guima11

horiz03

jacia02jacia09

janua01janua04

limpo01

luzia01

luzia07

parda01

migue01

parac02

parai06

parda02

pedre02

perdi01perdi04

petro02

pinhe02

piren01

pires02

pocoe02

riach02riach05

rondo01

roque01roque06

sacra01sacra02

santa01

serra01

simao01srita01

teres03

unaii02unaii09

Pleistocene2 0 Ma8

(c)

MtDNA

8 Mya

21 Mya

13 Mya

08 Mya

09 Mya

06 Mya

07 Mya

05 Mya

H pardalis

simao (41)

caval (14)

goias (8)

pocoe (30)

garca (5)domin (13)

migue (21)

desco (19)

guima (1)

caiap (7)

brasi (18)

catal (23)

roque (36)

parai (16)

pinhe (29)

cosmo (44)

sacra (39)

luzia (20)

l impo (11)


claro (43)

curve (34)

bauru (46)

pedre (40)

teres (15)

altog (4)

riach (32)

furna (35)

jacia (2)

perdi (38)

botuc (45)

assis (47)

pires (22)

gauch (26)

araxa (37)

horiz (33)

crist (31)

serra (6)

campe (9)

santa (27)

janua (28)

unaii (24)

rondo (3)

parac (25)

alian (17)

srita (42)

(b)

West

Central

West

Central

Southeast

SNPs

002

H pardalis

1

3

4

2

6

7

513

17

1415

1618

19202122

12

91011

8

27

2425

26

28

23

30

31

32333435

36

29

3738

39 4043

414244454647

(a)

SNPs










0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)






minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S



























































































































































































ndash25

ndash20

ndash15

ndash10

0

500

1000

1500

2000

17

4

37

4746 45

18

7 9

23

14

43

44

31

34

19

13

35

526

81

33

2

28

11 2021 25

16

40

38

10

29

12

22 30

32

3

3639

27

6

4142

15

24

0 200 400

km

Elevation (m)


















3emsp |emspRESULTS












K = 4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

West Central West

Central Southeast

K = 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

K = 21 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

87654321K

ndash240000

ndash220000

ndash200000

ndash180000

ndash160000

ndash140000

Mea

n Ln

pro

b o

f dat

a


765432K

400

300

200

100

0

Del

ta K




alian05

alian07

altog02

araxa01assis01

bauru01

botuc01

brasi04

caiap02caiap06

campe01

catal01

catal10

caval02

caval04

claro01

cosmo01

crist03

curve02

desco01

domin01domin09

furna01

garca02

gauch02gauch04

goias02goias10

guima01guima11

horiz03

jacia02jacia09

janua01janua04

limpo01

luzia01

luzia07

parda01

migue01

parac02

parai06

parda02

pedre02

perdi01perdi04

petro02

pinhe02

piren01

pires02

pocoe02

riach02riach05

rondo01

roque01roque06

sacra01sacra02

santa01

serra01

simao01srita01

teres03

unaii02unaii09

Pleistocene2 0 Ma8

(c)

MtDNA

8 Mya

21 Mya

13 Mya

08 Mya

09 Mya

06 Mya

07 Mya

05 Mya

H pardalis

simao (41)

caval (14)

goias (8)

pocoe (30)

garca (5)domin (13)

migue (21)

desco (19)

guima (1)

caiap (7)

brasi (18)

catal (23)

roque (36)

parai (16)

pinhe (29)

cosmo (44)

sacra (39)

luzia (20)

l impo (11)


claro (43)

curve (34)

bauru (46)

pedre (40)

teres (15)

altog (4)

riach (32)

furna (35)

jacia (2)

perdi (38)

botuc (45)

assis (47)

pires (22)

gauch (26)

araxa (37)

horiz (33)

crist (31)

serra (6)

campe (9)

santa (27)

janua (28)

unaii (24)

rondo (3)

parac (25)

alian (17)

srita (42)

(b)

West

Central

West

Central

Southeast

SNPs

002

H pardalis

1

3

4

2

6

7

513

17

1415

1618

19202122

12

91011

8

27

2425

26

28

23

30

31

32333435

36

29

3738

39 4043

414244454647

(a)

SNPs










0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)






minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S







































































































































3emsp |emspRESULTS












K = 4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

West Central West

Central Southeast

K = 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

K = 21 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

87654321K

ndash240000

ndash220000

ndash200000

ndash180000

ndash160000

ndash140000

Mea

n Ln

pro

b o

f dat

a


765432K

400

300

200

100

0

Del

ta K




alian05

alian07

altog02

araxa01assis01

bauru01

botuc01

brasi04

caiap02caiap06

campe01

catal01

catal10

caval02

caval04

claro01

cosmo01

crist03

curve02

desco01

domin01domin09

furna01

garca02

gauch02gauch04

goias02goias10

guima01guima11

horiz03

jacia02jacia09

janua01janua04

limpo01

luzia01

luzia07

parda01

migue01

parac02

parai06

parda02

pedre02

perdi01perdi04

petro02

pinhe02

piren01

pires02

pocoe02

riach02riach05

rondo01

roque01roque06

sacra01sacra02

santa01

serra01

simao01srita01

teres03

unaii02unaii09

Pleistocene2 0 Ma8

(c)

MtDNA

8 Mya

21 Mya

13 Mya

08 Mya

09 Mya

06 Mya

07 Mya

05 Mya

H pardalis

simao (41)

caval (14)

goias (8)

pocoe (30)

garca (5)domin (13)

migue (21)

desco (19)

guima (1)

caiap (7)

brasi (18)

catal (23)

roque (36)

parai (16)

pinhe (29)

cosmo (44)

sacra (39)

luzia (20)

l impo (11)


claro (43)

curve (34)

bauru (46)

pedre (40)

teres (15)

altog (4)

riach (32)

furna (35)

jacia (2)

perdi (38)

botuc (45)

assis (47)

pires (22)

gauch (26)

araxa (37)

horiz (33)

crist (31)

serra (6)

campe (9)

santa (27)

janua (28)

unaii (24)

rondo (3)

parac (25)

alian (17)

srita (42)

(b)

West

Central

West

Central

Southeast

SNPs

002

H pardalis

1

3

4

2

6

7

513

17

1415

1618

19202122

12

91011

8

27

2425

26

28

23

30

31

32333435

36

29

3738

39 4043

414244454647

(a)

SNPs










0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)






minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S
































































































































K = 4

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

West Central West

Central Southeast

K = 3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

K = 21 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 35 36 37 38 39 40 46 47ndashndash

87654321K

ndash240000

ndash220000

ndash200000

ndash180000

ndash160000

ndash140000

Mea

n Ln

pro

b o

f dat

a


765432K

400

300

200

100

0

Del

ta K




alian05

alian07

altog02

araxa01assis01

bauru01

botuc01

brasi04

caiap02caiap06

campe01

catal01

catal10

caval02

caval04

claro01

cosmo01

crist03

curve02

desco01

domin01domin09

furna01

garca02

gauch02gauch04

goias02goias10

guima01guima11

horiz03

jacia02jacia09

janua01janua04

limpo01

luzia01

luzia07

parda01

migue01

parac02

parai06

parda02

pedre02

perdi01perdi04

petro02

pinhe02

piren01

pires02

pocoe02

riach02riach05

rondo01

roque01roque06

sacra01sacra02

santa01

serra01

simao01srita01

teres03

unaii02unaii09

Pleistocene2 0 Ma8

(c)

MtDNA

8 Mya

21 Mya

13 Mya

08 Mya

09 Mya

06 Mya

07 Mya

05 Mya

H pardalis

simao (41)

caval (14)

goias (8)

pocoe (30)

garca (5)domin (13)

migue (21)

desco (19)

guima (1)

caiap (7)

brasi (18)

catal (23)

roque (36)

parai (16)

pinhe (29)

cosmo (44)

sacra (39)

luzia (20)

l impo (11)


claro (43)

curve (34)

bauru (46)

pedre (40)

teres (15)

altog (4)

riach (32)

furna (35)

jacia (2)

perdi (38)

botuc (45)

assis (47)

pires (22)

gauch (26)

araxa (37)

horiz (33)

crist (31)

serra (6)

campe (9)

santa (27)

janua (28)

unaii (24)

rondo (3)

parac (25)

alian (17)

srita (42)

(b)

West

Central

West

Central

Southeast

SNPs

002

H pardalis

1

3

4

2

6

7

513

17

1415

1618

19202122

12

91011

8

27

2425

26

28

23

30

31

32333435

36

29

3738

39 4043

414244454647

(a)

SNPs










0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)






minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S

























































































































alian05

alian07

altog02

araxa01assis01

bauru01

botuc01

brasi04

caiap02caiap06

campe01

catal01

catal10

caval02

caval04

claro01

cosmo01

crist03

curve02

desco01

domin01domin09

furna01

garca02

gauch02gauch04

goias02goias10

guima01guima11

horiz03

jacia02jacia09

janua01janua04

limpo01

luzia01

luzia07

parda01

migue01

parac02

parai06

parda02

pedre02

perdi01perdi04

petro02

pinhe02

piren01

pires02

pocoe02

riach02riach05

rondo01

roque01roque06

sacra01sacra02

santa01

serra01

simao01srita01

teres03

unaii02unaii09

Pleistocene2 0 Ma8

(c)

MtDNA

8 Mya

21 Mya

13 Mya

08 Mya

09 Mya

06 Mya

07 Mya

05 Mya

H pardalis

simao (41)

caval (14)

goias (8)

pocoe (30)

garca (5)domin (13)

migue (21)

desco (19)

guima (1)

caiap (7)

brasi (18)

catal (23)

roque (36)

parai (16)

pinhe (29)

cosmo (44)

sacra (39)

luzia (20)

l impo (11)


claro (43)

curve (34)

bauru (46)

pedre (40)

teres (15)

altog (4)

riach (32)

furna (35)

jacia (2)

perdi (38)

botuc (45)

assis (47)

pires (22)

gauch (26)

araxa (37)

horiz (33)

crist (31)

serra (6)

campe (9)

santa (27)

janua (28)

unaii (24)

rondo (3)

parac (25)

alian (17)

srita (42)

(b)

West

Central

West

Central

Southeast

SNPs

002

H pardalis

1

3

4

2

6

7

513

17

1415

1618

19202122

12

91011

8

27

2425

26

28

23

30

31

32333435

36

29

3738

39 4043

414244454647

(a)

SNPs










0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)






minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S
































































































































0 250000 500000 750000 1000000 1250000

Time (years ago)

1E8

1E7

1E6

1E5

Time (years ago)0 100000 200000 300000 400000

1E5

1E6

1E7

1E8(b)

Time (years ago)0 50000 100000 150000 200000

1E4

1E5

1E6

1E7

1E8(a) (c)






minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S




























































































































minus25

minus20

minus15

minus10

Current


minus25

minus20

minus15

minus10


02

04


minus25

minus20

minus15

minus10

LGM (minus21 ky)

02

04

06

08


minus25

minus20

minus15

minus10

LIG (minus120 ky)

02

04

06

08

00

10

00

10

02

04

06

08

00

10

06

08

00

10


00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)

020

040

060

080

010

0012

0014

00

00

02

04

06

08

10



Gen

etic

dis

tanc

e (F

st)


ndash25

ndash20

ndash15

ndash10


0

1

2

3

4

0 300 600

km


















ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S








































































































































ACKNOWLEDG EMENTS








ORCID







R E FE R E N C E S





























































































































ORCID







R E FE R E N C E S























































































































isolation by instability: historical climate change shapes

Documents