The interplay of dispersal limitation, rivers, and ... · The interplay of dispersal limitation,...

The interplay of dispersal limitation, rivers, andhistorical events shapes the genetic structure of anAmazonian frog

ANTOINE FOUQUET1,3*, JEAN-BAPTISTE LEDOUX2, VINCENT DUBUT3,BRICE P. NOONAN4 and IVAN SCOTTI5

1Departamento de Zoologia, Instituto de Biociências, Universidade de São Paulo, Caixa Postal 11461,CEP 05422-970, São Paulo, Brazil2Institut de Ciencies del Mar, CSIC, Passeig Marítim de la Barceloneta 37-49, 08003 Barcelona,Spain3Université d’Aix-Marseille, CNRS, IRD, UMR 6116 – IMEP, Equipe Evolution GénomeEnvironnement, Centre Saint-Charles, Case 36, 3 Place Victor Hugo, 13331 Marseille Cedex 3,France4University of Mississippi, Department of Biology, Box 1848, MS 38677, USA5INRA, UMR ECOFOG, Campus agronomique, Avenue de France, BP 709, 97387 Kourou, FrenchGuiana, France

Received 3 October 2011; revised 25 December 2011; accepted for publication 25 December 2011bij_1871 356..373

Disentangling the impact of landscape features such as rivers and historical events on dispersal is a challengingbut necessary task to gain a comprehensive picture of the evolution of diverse biota such as that found inAmazonia. Adenomera andreae, a small, territorial, terrestrial frog species of the Amazonian forest represents agood model for such studies. We combined cytochrome b sequences with 12 microsatellites to investigate the geneticstructure at two contrasted spatial scales in French Guiana: along a ~6-km transect, to evaluate dispersal ability,and between paired bank populations along a ~65-km stretch of the Approuague river, to test the effect of riversas barriers to dispersal. We observed significant spatial genetic structure between individuals at a remarkablysmall geographical scale, and conclude that the species has a restricted dispersal ability that is probably tied toits life-history traits. Mitochondrial and microsatellite data also indicate a high level of differentiation amongpopulations on opposite banks of the river, and, in some cases, among populations on the same riverbank. Theseresults suggest that the observed population structure in A. andreae is the result of restricted dispersal abilitiescombined with the action of rivers and Quaternary population isolation. Given that Amazonia hosts a great portionof anurans, as well as other small vertebrates, that display life-history traits comparable with A. andreae, we arguethat our analyses provide new insights into the complex interactions among evolutionary processes shapingAmazonian biodiversity. © 2012 The Linnean Society of London, Biological Journal of the Linnean Society, 2012,106, 356–373.

ADDITIONAL KEYWORDS: Amazonia – amphibian – gene flow – microsatellites – mitochondrial DNA –spatial genetic structure.

INTRODUCTION

The tropical forests of South America host the great-est species richness on Earth for many groups oforganisms (Wilson, 1992; Gaston & Williams, 1996;

Myers et al., 2000). This enormous biodiversity hasintrigued naturalists for more than a century(Wallace, 1852; Bates, 1863). However, the questionof its origins is still much debated (Antonelliet al., 2010). Progress has been made to clarify thephylogenetic relationships among higher clades (e.g.for amphibians, see Grant et al., 2006; Heinicke,*Corresponding author. E-mail: [email protected]

Biological Journal of the Linnean Society, 2012, 106, 356–373. With 5 figures

bs_bs_banner

© 2012 The Linnean Society of London, Biological Journal of the Linnean Society, 2012, 106, 356–373356

Duellman & Hedges, 2007; Santos et al., 2009), butour understanding of evolutionary history and pro-cesses at lower taxonomic levels (i.e. from species topopulations) remains incomplete. For example, pat-terns of intraspecific diversity and the understandingof the relative influence of environmental factorsand life-history traits remain poorly understood (butsee Noonan & Wray, 2006; Fouquet et al., 2007a;Carnaval et al., 2009).

Dispersal ability directly impacts connectivityamong populations, and ultimately influences geneticstructure over time and space (Clobert et al., 2001).Species characteristics such as territoriality (Bogart,1991; Austin et al., 2003), body size (Lindsey, 1966;Losos, 2010), habitat (e.g. Crawford, Bermingham &Polania, 2007; Moussalli et al., 2009) and reproductivemode (Vences & Wake, 2007) have long been acknowl-edged to influence dispersal. On the other hand, popu-lation connectivity is also influenced by landscapefeatures that have long been recognized by biologistsfor their role in facilitating and preventing gene flow,and ultimately shaping biodiversity patterns (Wiens,2001). In Amazonia rivers represent the most obviouslandscape features that could prevent gene flow. Twomain hypotheses are based on the barrier effect of thelarge rivers of the region to explain patterns of diver-sity, the river-barrier hypothesis (Wallace, 1852) andthe refuge-river hypothesis (Haffer, 1993a, b). Empiri-cal support for these hypotheses comes from the obser-vation that the boundaries of closely related species orsubspecies often coincide with major Amazonian rivers[e.g. birds (Haffer, 1997); lizards (Avila-Pires, 1995);and frogs (Lampert et al., 2003; Noonan & Wray,2006)]. Nevertheless, contrasting evidence for theeffectiveness of rivers as barriers has been reported(Gascon, Lougheed & Bogart, 1998; Gascon et al.,2000; Patton, Da Silva & Malcolm, 1996; Lougheedet al., 1999). The permeability of rivers to gene flowcould thus depend on river course variation over time(Bates, Haffer & Grismer, 2004) and species’ ability todisperse through them, which itself may be ruled bylife-history traits such as body size and reproductivemode (e.g. Gascon, Lougheed & Bogart, 1998).

Amphibians are particularly diverse in Amazoniaand are considered valuable models to investigateprocesses shaping genetic structure (Zeisset &Beebee, 2008). In addition of being highly sensitive toenvironmental variation (Buckley & Jetz, 2007), theyare generally thought to have poor dispersal capabili-ties (Waldman & McKinnon, 1993; Blaustein, Wake &Sousa, 1994). Most evidence of the generally lowdispersal ability of anurans (reviewed by Smith &Green, 2005) comes from ecological and geneticstudies of temperate pond-breeding species that dem-onstrate high philopatry (Daugherty & Sheldon, 1982;Driscoll, 1997) and significant structure among spa-

tially close populations (García-París et al., 2000;Shaffer et al., 2000; Tallmon et al., 2000; Monsen &Blouin, 2003). Nonetheless, amphibian movementand levels of interpopulation gene flow can be exten-sive in some taxa (Breden, 1987; Berven & Grudzien,1990; Marsh & Trenham, 2001; Trenham, Koenig &Shaffer, 2001; Funk et al., 2005; Smith & Green,2005). For example, three temperate species that arespecialized in ephemeral breeding sites (Scaphiopuscouchii and Anaxyrus cognatus, Chan & Zamudio,2009; Epidalea calamita, Rowe, Beebee & Burke,2000) have been shown to be highly efficientdispersers possibly compensating for the stochasticityof breeding sites. These contrasting results confirmthat interpopulation connectivity largely depends onlife-history strategies and local environment (Berry,2001; Newman & Squire, 2001; Squire & Newman,2002; Lampert et al., 2003). Such drastically oppositestrategies may have dramatic influence on the largerscale pattern of biodiversity, as in Bufonids (VanBocxlaer et al., 2010).

Considering the contrasted environments (e.g.climate) and their differences in historical fluctua-tions, as well as different life-history strategies (e.g.modes of reproduction) and demography betweenmost temperate and tropical amphibians, one wouldexpect striking differences in population structure(e.g. Chek, Austin & Lougheed, 2003; Eckert, Samis &Lougheed, 2008). However, comparative material isscarce, as only a few studies (Leblois et al., 2000;Lampert et al., 2003; Elmer, Davila & Lougheed,2007a; Dixo et al., 2009; Robertson, Duryea &Zamudio, 2009) have focused on tropical amphibiansat a relatively fine scale. Additionally, these studieshave mostly targeted large species with free larvaldevelopment, and have largely ignored species thatlack a free-living larval stage (but see Elmer et al.,2007a), a condition absent in temperate anurans.

In this study we have evaluated the respectiveeffects of distance and river on the population struc-ture in a widespread Amazonian species: Adenomeraandreae (Müller, 1923). This species is small bodied,territorial, displays a terrestrial mode of development(lecithotrophic development, sensu Dubois, 2005),with small clutches. Therefore, levels of gene flow areexpected to be low and rivers to act as efficient bar-riers. Conversely, the continuity of the extensiveforest habitat, species abundance and the ability toreproduce several times each year should promotegene flow. Because these characteristics are actuallyshared by many co-occurring organisms (Duellman &Trueb, 1986), A. andreae is, in our opinion, a goodmodel to study the processes at stake in the evolutionof tropical terrestrial fauna (Fouquet et al., 2009).Previous phylogeographic work revealed at least sixhighly divergent, spatially exclusive mtDNA lineages

POPULATION GENETICS OF AN AMAZONIAN FROG 357


(12S + 16S; ~800 bp) of A. andreae in a small area ofthe eastern Guiana Shield (GS) (Fouquet et al., 2012).These data suggested that this genetic structureoriginated during the early Pleistocene from acomplex history of habitat fragmentation linked toclimatic fluctuations. Such evolutionary history iscommon among organisms occurring in the GS (inanurans, Noonan & Gaucher, 2005, 2006; Fouquetet al., 2007b, 2012; Lötters et al., 2010; in plants,Duputié et al., 2009; Boisselier-Dubayle et al., 2010).Yet none of the other species studied so far exhibitsuch a high degree of genetic subdivision and diver-sity within the region.

Using data from 607 bp of cytochrome b and 12microsatellite loci (Fouquet et al., 2009) for 221 indi-viduals sampled at two different spatial scales, weaddressed two points. We first determined ‘uncon-strained’ dispersal ability by examining fine-scale pat-terns along a ~6-km transect perpendicular to theApprouague River in the Nouragues reserve withindividuals sampled within the same phylogeographicunit. Next, we tested the effect of a river as a barrier,by examining population structure along a ~65-kmstretch (six main sampling points – three on eachbank) of the Approuague River (AR). This approach,combining different markers and two sampling scales,has allowed us to infer the interaction between effec-tive dispersal, rivers and evolutionary history toexplain observed patterns of biological diversity.

MATERIAL AND METHODSSAMPLING

Our study was designed to successively focus on asmall area of the AR basin. The AR extends ~270 km,flowing from the south-west to the north-east coast ofFrench Guiana. In March 2009, during the peak malecalling activity period, 109 individuals were sampledand geo-referenced at four locations along a ~6-kmtransect extending north in the Nourague Reserve onthe west bank of the river: (1) the forest edge of theisolated Inselberg at the heart of the reserve (NON-1); (2) near the Inselberg camp (NON-2); (3) theplateau (Balenfois) 3 km north of the river (NON-3);and (4) the riverside Parare camp on the Arataï River(NON-4) (Fig. 1), a tributary of the AR. To allowpaired bank comparisons, we also sampled five loca-tions with similar habitats during the same periodalong the river from headwater in the Arataï tribu-tary to 60 km from the mouth, corresponding to 112individuals. Thus, besides samples NON-1 to NON-4collected from the north (left) bank of the Arataï, wesampled: (1) the south (right) bank of this tributary(Parare East = NOS); (2) the two banks at Cisame,oriented north-east/south-west (CIE and CIW) at mid

river, around 50 km downstream from Nouragues;and (3) the two banks near Regina, also orientednorth-east/south-west (REE and REW), around 60 kmfrom the mouth of the Approuague, where the riverbanks transition to estuarine habitat. Overall 221individuals of A. andreae were collected with anaverage of 24.5 (SD 6.4) individuals per location(minimum at NOS, N = 13).

The reasons behind this sampling design aretwofold, both logistical and biological. The location ofinfrastructures along the Approuagues allowed thissampling design. It also makes biological sense giventhe downstream part of the Approuagues correspondsto the estuarine part of the river (large, with rela-tively stable flow and water level), whereas theupstream part corresponds to fast-flowing headwaters(reduced width, with strong fluctuations of flow andwater level). The intermediate location (Cisame) dis-plays intermediate topology.

DNA EXTRACTION AND PCR PROTOCOLS

Tissue was taken from thigh muscle or liver andpreserved in 95% ethanol. Total genomic DNA wasextracted using Puregen® Gentra™ DNA Tissue Kit(QIAGEN). A 607-bp cytochrome b (cyt b) fragmentwas amplified by standard PCR techniques forall individuals except six. We used new primers,LeptoCBF2, 5′-ATTGCMCAAATYGCYACAGG-3′,and LeptoCBR2, 5′-GTGAAGTTRTCYGGGTCYCC-3′,designed for the Leptodactylus clade (sensu Frostet al., 2006) with a 54 °C annealing temperature.We also used four additional Adenomera species(Adenomera heyeri, Adenomera hylaedactyla, Adeno-mera lutzi and an undescribed species from Peru)and Lithodytes lineatus as out-groups. Sequenc-ing reactions of the purified PCR products wereobtained in both directions using ABI Big Dye V3.1,purified with ethanol precipitation, and resolvedon an automated sequencer ABI 3130XL Genetic Ana-lyzer (Applied Biosystems) at Ecofog (INRA, Kourou,French Guiana).

The cyt b sequences were edited and aligned withSEQSCAPE 2.5 (Applied Biosystems) and submittedto GenBank (Appendix S1).

Nineteen microsatellite loci were amplified follow-ing Fouquet et al. (2009) for all individuals, exceptone. Visualization of amplicon size was performed andresolved on an automated sequencer ABI 3130XLGenetic Analyzer (Applied Biosystems). Microsatelliteallele sizes were calibrated using an internalGeneScan-500 LIZ® size standard. Genotypes wereobtained using GENEMAPPER 3.7 (Applied Biosys-tems). To ensure reliability, ~10% of the individualswere genotyped twice, in all cases yielding identicalresults.

358 A. FOUQUET ET AL.


Figure 1. A, maximum clade credibility tree inferred from Bayesian analysis in BEAST for the sampled Adenomeraandreae cytochome b haplotypes. The divergence times are indicated below nodes and correspond to the mean posteriorestimate of their age in millions of years. The grey bars indicate the 95% highest posterior density (HPD) interval for thedivergence time estimates. Posterior probabilities > 0.5 are labelled in bold above nodes and are *100 when < 1. Clades arecoded with successive labels A/B, ½, and a/b/c, as well as with different colours. These labels are used on the map to locateeach mitochondrial DNA (mtDNA) clade. B, the TCS statistical parsimony networks are illustrated in front of eachcorresponding clade (except for A2a, which holds only two haplotypes). The size of the circle is proportional to the frequencyof the haplotype. The colours correspond to frequency in the corresponding populations coded in the legend above.



DATA ANALYSES

Cytochrome bWe used BEAST 1.5.4 (Drummond & Rambaut,2007) with a strict clock of 1% substitutions per Myrper lineage (2% pairwise divergence) to reconstruct atime-calibrated phylogenetic tree. We applied a strictclock because of the limited size of the charactermatrix and lack of recent calibration points. Thechosen rate was adopted from previous studies oftropical anurans (e.g. Lougheed et al., 1999; Elmer,Davila & Lougheed, 2007b), and allows comparisonswith other studies. We used the software MRMOD-ELTEST 2.3 (Nylander, 2004) to select the substitu-tion models that best fit each codon position of cyt baccording to Akaike’s information criterion (Akaike,1981). Accordingly, a GTR + I + G partitioned modelfor each codon position was used, and the tree priorincluded the birth–death process, with a UPGMAstarting tree. We undertook a first run of 107 gen-erations sampled every 1000 generations withdefault values to bond the priors (Drummond &Rambaut, 2003), and then undertook a new run thatcomprised 2.107 generations. An adequate burn-inwas determined by examining a plot of the likelihoodscores of the heated chain for convergence on sta-tionarity in TRACER 1.5 (Rambaut & Drummond,2007).

To graphically represent the inter-relationshipsamong closely related haplotypes, and their frequen-cies, we used TCS 1.21 (Clement et al., 2000) to createstatistical parsimony networks (Templeton et al.,1992) with a connection limit set to 95%.

Microsatellite locus characteristics andHardy–Weinberg equilibriumWe used MICROCHECKER 2.2.3 (Van Oosterhoutet al., 2004) to test for large allele dropout and/orscoring errors caused by stutter peaks. Null-allelefrequencies were estimated for each sample and locusby the expectation–maximization algorithm of Demp-ster, Laird & Rubin (1977), implemented in FREENA(Chapuis & Estoup, 2007). All loci were polymorphicwith no evidence of large allele dropout or scoringerrors caused by stuttering. High null-allele frequen-cies (mean over samples > 0.15; data not shown) wereobtained for seven loci; these were excluded fromsubsequent analyses. The following analyses are thusbased on the remaining 12 loci.

The total number of alleles, size ranges, andobserved (Ho) and unbiased expected heterozygosity(He) (Nei, 1973) were calculated for each locus.The null hypothesis of independence between lociwas tested within each location and over all loca-tions using a permutation procedure (N = 1000) inGENETIX 4.05 (Belkhir et al., 1996–2004). Single

and multilocus estimators of FIS (Weir & Cockerham,1984) in each sample, and departures from panmixia,were tested with a score test for heterozygote defi-ciency and default parameters for the Markov Chain(MC) algorithm (Guo & Thompson, 1992; Raymond& Rousset, 1995) using GENEPOP 4.0 (Rousset,2008).

To further investigate the cause of observedHardy–Weinberg (HW) disequilibrium (see Results),the rxyIdentity relatedness coefficient was computed foreach pair of individuals within each sample usingIDENTIX 1.1 (Belkhir, Castric & Bonhomme, 2002).We then tested whether sample means and vari-ances differed significantly from their null expecta-tions under panmixia. The expected distribution wasobtained using 1000 permutations of the genotypes.Statistically significant higher means would suggestthat individuals are more inbred (i.e. biparentalinbreeding) than expected under random mating,whereas significantly higher variance would suggestthat multiple groups of related individuals had beensampled (i.e. family) (Belkhir et al., 2002).

Fine-scale spatial genetic structure anddemographic parametersThe occurrence of spatial genetic structure (SGS)was tested by applying spatial autocorrelationanalyses and permutation tests implemented inSPAGEDI 1.0 (Hardy & Vekemans, 2002). Wefocused on patterns among the populations NON-1to NON-4 because of their spatial proximity on a~6-km transect and the absence of landscape barri-ers. Following Vekemans & Hardy (2004), thegenetic distance between individuals was estimatedusing Nason’s estimator of kinship coefficient(Loiselle et al., 1995). Kinship values (Fij) were thenregressed on the natural logarithm of the distancebetween individuals [ln(dij)]. The significance of theslope of the regression (bLd) was tested using 1000permutations of the spatial position of individualsunder the hypothesis of no correlation betweengenetic and geographic distances. A correlogram wasused to visualize the shape of the SGS, designingdistance classes that maximized the sample size ofeach class. Kinship values were averaged over a setof different distance intervals d. The [F(d)] valuesobtained were plotted against geographic distances.For each distance interval, individual locations werepermuted (N = 1000) to test for the significance ofthe correlation coefficient.

We estimated the so-called neighbourhood size (Nb)as Nb = -(1 - FN)/bLd, where FN is the mean of Fij

between adjacent individuals i and j, and is approxi-mated by F(d) for the first distance interval (seeVekemans & Hardy, 2004). The 95% confidence inter-val of Nb was computed as (FN - 1)/(bLd + 2SEb) and



(FN - 1)/(bLd - 2SEb), with SEb, the standard error ofbLd, estimated using a jackknife procedure over loci,as implemented in SPAGEDI. We then used theobtained value of Nb to compute a moment estimateof the mean axial parent–offspring distance (sg) usingthe relationship linking the two parameters in atwo-dimensional model of isolation by distance (IBD):Nb = 4pDes2

g, where De is the effective density(Rousset, 1997, 2000; Hardy & Vekemans, 1999). Toestimate the effective density, we used individual-based georeferenced groups of adjacent calling malesthat were putatively sampled exhaustively on eachNouragues sample (each being ~1000 m2). We multi-plied it by two to include both sexes (which is a grossestimation for the effective reproductive proportion ofthe population, i.e. the ‘effective density’, assuming asex ratio of 50 : 50), and obtained a density estimateof 0.05 individuals m-2. Preliminary results frommitochondrial DNA (mtDNA) demonstrated the exist-ence of two shallow lineages in each NON locationand the occurrence of only one cluster and low FST

using microsatellites (see Results). We are thus con-fident in using these samples as a homogenous groupwith which neighborhood size can be estimated.

Clustering analysis, population structure andrelationships, and isolation by distance on a broadspatial scaleThe population structure was analysed usingthe Bayesian clustering method implemented inSTRUCTURE 2.2 (Pritchard, Stephens & Donnelly,2000; Falush, Stephens & Pritchard, 2003, 2007).This Bayesian approach was used to infer thenumber of genetic clusters K from the individualgenotype data set, without information on sampleboundaries, while optimizing the HW and linkageequilibrium within each cluster. We performed tenindependent runs for each K (from K = 1 to K = 9)using an admixture and the ‘correlated alleles fre-quencies’ model, and treating non-amplified geno-types as null alleles with the recessive alleles option(Falush et al., 2007). Each replicate was run for200 000 iterations following a burn-in of 50 000 onthe entire data set (N = 221). An estimate of thelogarithm of the likelihood of observing the dataunder each K [‘LnP(D)’] was computed for each run.These values were then plotted as a function of theputative number of clusters (K). We selected theK value using the method of Pritchard et al. (2000),i.e. by looking for the presence of a plateau and forthe K value that captured the major structure of thedata (Pritchard, Wen & Falush, 2007). We chose thisstandard approach over the alternative ‘ad hoc’approach of Evanno, Regnaut & Goudet (2005)because the improvement of the latter was ques-tioned by Waples & Gaggiotti (2006). CLUMPP 1.1

(Jakobsson & Rosenberg, 2007) and DISTRUCT1.1 (Rosenberg, 2004) were used, respectively, toaverage the assignment of each individual over thedifferent runs and for graphical display (no multi-modality was observed). Finally, to identify putativemigrants along with individuals with immigrantancestry in the last two generations (GENS-BACK = 2), we ran STRUCTURE again usingcluster assignment information and the K valueobtained at the first run as a prior under thePOPINFO option, with default parameters andother settings identical to the previous analyses. Foreach individual, we thus obtained the posteriorprobability (PP) of the individual originating fromthe population from which it was sampled.

We used a genetic distance method among ARpopulations to examine population affinities. Anallelic frequencies data file was constructed usingGENETIX 4.05 (Belkhir et al., 2004), and PHYLIP(Felsenstein, 2005) was used to compute the Cavalli-Sforza chord (Dc) distance matrix, and to then con-struct a neighbour-joining (NJ) phenogram (Saitou &Nei, 1987). Node robustness was tested by performing1000 bootstrap replicates over all loci.

We calculated pairwise qST, an estimator of FST

(Weir & Cockerham, 1984), for each population pairusing GENETIX 4.05. Statistical significance wastested with 1000 permutations. The maximum valueof genetic differentiation metrics, such as FST, isconstrained by the overall observed heterozygosity ofthe genetic markers used, such that the maximumpossible measure of differentiation decreases withincreasing heterozygosity (Hedrick, 1999, 2005). Con-sequently, to be comparable with other populationgenetics studies, we used F′ST (Meirmans, 2006),which is a standardized measure of FST. F′ST =FST/FST(max), following Meirmans (2006). PairwiseFST(max) values among populations were calculatedwith FSTAT (Goudet, 1995) after recoding the datausing RECODATA 0.1 (Meirmans, 2006).

We used a regression major axis (RMA) regressionof FST/(1 - FST) against the natural log of lineargeographic distance [ln(dist)], as recommended byRousset (1997) for a two-dimensional system, andimplemented in IBDWS 3.15 (Jensen, Bohonak &Kelley, 2005), to investigate the relationship betweengeographical and genetic distances. A Mantel test with1000 permutations was used to determine whether theslope of the regression was significantly greater thanzero. All populations were included in the Mantel test.We also performed additional univariate regressionsincluding only parts of the data in order to highlightthe observed discontinuities.

Significance levels were corrected using a falsediscovery rate (FDR) correction for multiple tests(Benjamini & Hochberg, 1995).



RESULTSMTDNA

Two major clades, estimated to have diverged~2.9 Mya (mean p distance = 5.5%), were identified.The three west-bank populations, REE, and half of theNOS individuals belong to clade A (Fig. 1A). Clade B ison the other hand composed only of two of the east-bank samples (CIE and the other half of the individu-als in NOS). More recent, albeit strong, subdivisionsoccur within the two major clades. Clade A is differen-tiated into clade A1, present in all west-bank popula-tions and half of the NOS individuals, and clade A2,present in REE on the east bank. These two clades areestimated to have diverged 1.6 Mya (mean p dis-tance = 2.2%). Clade A1 is further subdivided intothree closely related haplogroups: A1a, restricted toNON; A1b, found in NON, NOS, and CIW; and A1c,found in CIW and REW. Both clades A1a and A1boccurring in NOU populations are represented in eachof these four populations. All nodes are strongly sup-ported (PP � 0.95), except among subclades within A1.Clade B is subdivided into three lineages, two of whichoccur in CIE and the third one that occurs in NOS.

Interestingly, although downstream there is verystrong differentiation among populations on oppositesides of the river, upstream the NOS and NON popu-lations share two haplotypes of clade A1b. This cladeis, however, otherwise restricted to the west bank,and the rest of the individuals in the NOS populationharbour clade-B haplotypes.

MICROSATELLITE LOCI CHARACTERISTICS AND

HARDY–WEINBERG EQUILIBRIUM

The total numbers of alleles ranged from eight forAdan-18 to 44 for Adan-27, with a mean of 19.4 ± 10.1

(SD) alleles per locus (Table 1). Observed and unbi-ased heterozygosity varied, respectively, from 0.541for Adan-29 to 0.84 for Adan-12, and from 0.609 forAdan-29 to 0.939 for Adan-27 (Table 1). No significantlinkage disequilibrium among loci was detected con-sidering the global data set or each location indepen-dently after correction for false discovery rate. Withineach sample we observed significant heterozygotedeficiencies, with FIS values ranging from 0.18 in CIEto 0.6 in NOS and REW, with a mean value equal to0.10 (Table 1). Considering each locus separately, FIS

values ranged from -0.31 for Adan-40 for REW to0.64 for Adan-43 for CIE. Departures from panmixiavaried between loci and samples. Adan-27 showedsignificant heterozygote deficiencies in six of the ninepopulations, whereas panmixia could not be rejectedfor Adan-22, Adan-12, and Adan-18 in any population(Table S1). None of the nine populations displayed amean relatedness coefficient that was significantlydifferent from expected values with random mating(rxyidentity ranged from 0.271 in NOS to 0.355 in REW)(Table 2). However, four samples displayed variancevalues that were significantly different than expectedunder panmixia. Variance was greater than expectedin three populations (CIE, REE, and NON-2), andwas lower in one (NON-4) (Table 2).

FINE-SCALE GENETIC STRUCTURE

Considering the four NON samples as a whole (exceptthe two identified migrants via the POPINFO option,see below), the observed bLd value was significantlydifferent from zero (P < 0.001). The autocorrelogram(Fig. 2) confirms the decrease of the kinship coeffi-cient with increasing geographical distance. Correla-tion coefficients only significantly differed from 0 for

Nouragues populations IBD

-0,01

-0,005

0

0,005

0,01

0,015

30 110640

1,320

2,150 2,550

3,210

4,330

5,150

5,470m

geographical distance (d) (mean d for interval indicated in m)

*

*kinship

Figure 2. Autocorrelogram of the Nouragues (NON) individuals using ten distance classes. Vertical bars indicate thestandard errors of the observed values for each class. Asterisks indicate the distance classes for which correlationcoefficients were significantly different from 0. Dashed lines indicate the 95% confidence intervals of the permuted values.



Tab

le1.

Nu

mbe

rof

indi

vidu

als

per

popu

lati

on(N

),n

um

ber

ofal

lele

spe

rlo

cipe

rpo

pula

tion

(n),

and

valu

esof

FIS

per

locu

spe

rpo

pula

tion

(in

colu

mn

belo

wth

elo

cus

nam

es)

Pop

.N

nA

dan

-2n

Ada

n-2

2n

Ada

n-2

9n

Ada

n-3

0n

Ada

n-4

0n

Ada

n-4

2

NO

N-1

315

0.28

98

0.04

96

-0.0

9812

0.05

26

0.18

112

0.26

2N

ON

-225

30.

742

5-0

.042

30.

102

10-0

.019

50.

254

11-0

.087

NO

N-3

206

0.63

28

-0.0

215

0.20

78

0.11

96

0.26

89

0.09

5N

ON

-433

60.

048

100.

016

4-0

.017

8-0

.117

80.

078

120.

204

NO

S12

30.

570

4-0

.106

3-0

.132

9-0

.030

8-0

.052

80.

277

CIW

245

0.09

06

0.02

74

0.18

07

0.08

94

-0.2

0116

0.18

0C

IE24

70.

529

80.

046

50.

270

90.

331

50.

177

90.

221

RE

W22

40.

095

40.

193

30.

214

60.

110

4-0

.312

110.

072

RE

E30

40.

492

100.

033

50.

299

8-0

.014

60.

086

150.

192

Ove

rall

221

110.

093

180.

025

90.

112

190.

040

100.

072

240.

160

Ho

0.55

10.

736

0.54

10.

666

0.61

00.

726

He

0.63

80.

825

0.73

70.

751

0.67

80.

900

Pop

.N

nA

dan

-12

nA

dan

-15

nA

dan

-18

nA

dan

-1n

Ada

n-2

7n

Ada

n-4

3F

ISm

ult

iloc

us

NO

N-1

3114

0.13

76

0.09

58

0.18

611

-0.1

0420

0.20

58

0.03

90.

079

NO

N-2

2512

-0.0

276

0.09

76

0.22

48

-0.0

2621

0.31

811

0.01

30.

081

NO

N-3

2016

-0.0

245

-0.1

1112

-0.0

889

0.14

917

0.37

09

0.05

50.

095

NO

N-4

3314

0.03

85

0.21

09

0.13

113

0.14

121

0.29

311

-0.0

190.

079

NO

S12

150.

290

4-0

.048

110.

048

90.

517

13-0

.086

6-0

.163

0.06

3C

IW24

150.

053

50.

203

70.

131

50.

127

240.

051

60.

084

0.11

3C

IE24

130.

037

40.

151

80.

049

70.

039

170.

199

60.

642

0.17

8R

EW

2213

0.03

95

-0.2

005

0.20

04

0.15

219

-0.0

137

0.23

20.

066

RE

E30

160.

039

40.

201

100.

038

40.

298

140.

160

80.

058

0.13

4O

vera

ll22

129

0.05

58

0.09

221

0.10

522

0.10

544

0.18

418

0.11

20.

100

Ho

0.84

00.

573

0.58

50.

548

0.78

20.

701

He

0.93

30.

719

0.70

20.

730

0.94

50.

855

FIS

valu

essi

gnifi

can

tly

depa

rtin

gfr

omth

ose

expe

cted

un

der

Har

dy–W

ein

berg

equ

ilib

riu

mar

ein

dica

tin

gin

bold

.O

bser

ved

and

expe

cted

het

eroz

ygos

ity

(Ho

and

He)

valu

espe

rlo

cus

are

also

indi

cate

dat

the

bott

omof

the

tabl

e.



the first (mean = 30 m; 0–60 m) and the eighth dis-tance classes (mean = 4.3 km; 4.2–4.4 km). The neigh-bourhood size (Nb) equalled 462 individuals (95% CI215–3164), leading to a sg value equal to 27.1 m(95% CI 18.5–78.9 m) when considering an approxi-mate estimated density of 0.05 individuals m-2.

CLUSTERING ANALYSIS AND RELATIONSHIP

BETWEEN SAMPLES

Bayesian assignment analyses identified four clus-ters (K = 4) (Fig. 3A). One cluster (A) corresponds toNON. Downstream populations on each side of theAR make up two further clusters: cluster B (CIWand REW) and cluster C (CIE and REE), respec-tively. The fourth cluster (D) corresponds to NOS.STRUCTURE analysis failed to detect significantsubdivision among the four NON sampling siteswhen analysed separately (not shown). Very limiteddispersal among populations was inferred based onthe assignment tests using the POPINFO option inSTRUCTURE. Only two individuals out of 221(0.9%; identified with asterisks in Fig. 3B) wereassigned to a different population than the one theyhad been sampled from. One individual from NON-4was identified as a migrant (PP < 0.01) from NOS.The other individual from NON-1 displayed only aPP = 0.25 of being from that population, suggestingthat it could have had recent immigrant ancestry.

The NJ tree based on Dc, calculated among ARpopulations, provides complementary information(Fig. 4). The downstream east-bank populations(CIE + REE) and west-bank populations (CIW +REW) formed clusters with high bootstrap (BP)values (91.8 and 91.9%, respectively) that togetherform a downstream cluster (BP = 92.6%) divergent

from the cluster formed by the NON populations(BP = 92.6%). The NOS population lies apart from theother populations.

POPULATION STRUCTURE AND IBD

Most pairwise FST values were high and significant,with maximum values involving NOS (maximumFST = 0.17, NOS versus REW; maximum F′ST = 0.58,NOS versus CIE) (Table 3). Among NON populations,FST values were generally not significant, suggestinggenetic homogeneity; only NON-2 was significantlydistinct from the other NON samples, but withlow FST (0.0072–0.0154) and F′ST (0.027–0.061)values.

A significant correlation was detected betweengenetic [FST/(1 - FST)] and geographical distances[ln(dist)] (Z = 4.51, r = 0.40, P = 0.014; Fig. 5). WhenNOS was removed, we observed a strong increase inthe fit of the correlation among the remainingsamples (including transriverine comparisons)(Z = 3.04, r = 0.55, P = 0.008, RMA slope = 0.069,95% CI 0.053–0.091). When using simple correlationtrends over data partitions, NOS versus other popu-lations and transriverine comparisons clearly lieoutside the general correlation trend and the com-parisons involving NON populations (Fig. 5). Suchresults are only suggestive, and must be interpretedwith caution given the small number of populationsinvolved.

DISCUSSION

Our analyses revealed deep genetic structure amongpopulations of A. andreae sampled along the AR. Mic-rosatellite markers as well as mtDNA distinguish

Table 2. Number of individuals per population (N) and values of observed and expected heterozygosity (Ho and He), andtheir SDs in brackets, relatedness coefficient, and variance

Population N Ho He rxyidentity VARrxy

NON-1 31 0.675 (0.121) 0.721 (0.129) 0.306 0.0106NON-2 25 0.663 (0.168) 0.716 (0.122) 0.319 0.0130NON-3 20 0.704 (0.128) 0.745 (0.097) 0.275 0.0091NON-4 33 0.700 (0.094) 0.736 (0.105) 0.307 0.0083NOS 12 0.701 (0.229) 0.714 (0.198) 0.271 0.0085CIW 24 0.625 (0.158) 0.689 (0.157) 0.278 0.0086CIE 24 0.590 (0.197) 0.701 (0.159) 0.301 0.0123REW 22 0.616 (0.208) 0.644 (0.177) 0.355 0.0089REE 30 0.621 (0.186) 0.701 (0.170) 0.307 0.0123Overall 221 0.655 0.784

Variance values significantly greater than expected are indicated in bold.



Figure 3. Bayesian assignment analysis from STRUCTURE. A, plot of the logarithm of the likelihood of each replicatefrom K = 1 to K = 10. B, individual assignment probabilities with K = 4. Populations are indicated below and clusters arenamed from A to D. Asterisks indicate significance for individuals identified as migrant with the POPINFO option.C, maps display the distribution of the different clusters along the Approuagues River with each population-associatedgraph representing the proportion of probabilities of assignment in each cluster: 1, on the Approuagues River; 2, on theNouragues transect.

Table 3. FST and F′ST values

FST/F′ST NON-1 NON-2 NON-3 NON-4 NOS CIW CIE REW REE

NON-1 – 0.0274 0.0223 0.0209 0.4410 0.2173 0.3287 0.2463 0.2901NON-2 0.0072** – 0.0550 0.0605 0.4233 0.2313 0.3842 0.2852 0.3643NON-3 0.0055 0.0136*** – -0.0013 0.4330 0.2832 0.3344 0.3317 0.3026NON-4 0.0053 0.0154*** -0.0003 – 0.3851 0.2522 0.3008 0.2385 0.2597NOS 0.1137*** 0.1095*** 0.1038*** 0.0957*** – 0.4605 0.5767 0.5639 0.5349CIW 0.0604*** 0.0646*** 0.0745*** 0.0679*** 0.127*** – 0.3625 0.2089 0.3561CIE 0.0893*** 0.1048*** 0.0856*** 0.0791*** 0.1545*** 0.1045*** – 0.4052 0.1866REW 0.0737*** 0.086*** 0.0949*** 0.0692*** 0.1699*** 0.0663*** 0.1259*** – 0.3351REE 0.0798*** 0.1007*** 0.0788*** 0.0693*** 0.1457*** 0.1038*** 0.0532*** 0.1049*** –

Significance: **P < 0.01; ***P < 0.001.Non-significant values are set in bold.



populations upstream from those downstream, as wellas those on opposite banks of the AR. We find a nearlycomplete absence of transriverine gene flow and astrong effect of geographic distance on genetic struc-ture at small scale among Nouragues populations.These results suggest that the observed populationstructure in A. andreae results from restricted dis-persal abilities combined with the action of rivers andQuaternary population isolation.

FINE-SCALE GENETIC STRUCTURE AND

HARDY–WEINBERG DISEQUILIBRIUM

Spatial genetic structure was detected on a fine scale(Fig. 2) between individuals separated by a fewhundred metres, unambiguously supporting thehypothesis of low dispersal in this species. We char-acterized this dispersal by estimating the mean axialparent–offspring distance: sg = 27.1 m (95% CI 18.5–

0.01

REE

CIE

REW

CIW

NOS

NON-3

NON-1 NON-4

NON-2

926

926 918

919

458513

B West

C East

DNouragues

South

A Nouragues North

Figure 4. Unrooted neighbour-joining phylogram, based on the Cavalli-Sforza chord distance for the nine ApprouaguesRiver populations. Bootstrap values from 1000 replicates are indicated.

–0.10

–0.06

–0.02

0.02

0.06

0.10

0.14

0.18

0.22

0.26

0.30

Gen

etic

Dis

tanc

e F

st

Mantel test with all comparisonsZ = 4.51, r = 0.40 p = 0.014

Cross-river comparisons East (C) VS West (B)

Nouragues North (A) VS West (B) + East (C)

Within West (B) and East (C)

NOS vs others

0.2 1.4 2.6 3.8 5.0–1.0

Within Nouragues North (A)

R²= 0.82 R²= 0.82

R²= 0.81

Ln Geographical distance

+ +

R²= 0.22

1.

2.

3.

Figure 5. Plot of FST among the nine Approuagues River populations versus ln(geographical distance), with clusterassignment indicated in brackets (A–D). The result of the Mantel tests using 1000 permutations on all pairwisecomparisons is indicated (black continuous line). To illustrate the discontinuities among populations we also indicated thecorrelation trends (dashed lines) for: (1) NOS versus other populations (orange dashed line); (2) cross-river comparisons(red dashed line); and (3) NON populations versus downstream populations (green dashed line).



78.9 m). To our knowledge, the individual-based SGSof Amazonian amphibians has only been demon-strated for the frog Pristimantis kichwarum (Elmeret al., 2007a), a directly developing species of similarbody size in which patch size (comparable with meanaxial parent–offspring distance; for a discussion, seeFenster, Vekemans & Hardy (2003) was estimated at260 m. Cabe et al. (2007) observed an even smallerpatch size in the direct developing North Americansalamander Plethodon cinereus (3.53 m), and Driscoll(1998) reported genetic neighbourhoods as small as30 m in a direct developing Australian frog of thegenus Geocrinia. Compared with most studies focus-ing on vertebrate species our values are very low (e.g.in Martes americana sg = 3.8 km, Broquet et al., 2006;in Ptermys volens sg = 2.9 km, Selonen, Hanski &Painter, 2010; but in Microtus arvalis sg = 88 m,Gauffre et al., 2008 ). Given male territoriality andparental investment (Kokubum & Giaretta, 2005;Almeida & Angulo, 2006; Menin, Almeida &Kokubum, 2009), we may expect a female-biased dis-persal in this species (Greenwood, 1980). As wemostly sampled calling males we cannot rule out thishypothesis. Nevertheless, the particularly strongstructure seen in maternally transmitted mtDNA(Fig. 1) suggests that females disperse very little.Such limited dispersal for this species is probablylinked with life-history traits such as terrestrialdevelopment, territoriality, and small body size.

We observed a general heterozygote deficiencywithin each population, challenging the hypothesis ofpanmixia in these samples. Heterozygote deficiencycould result from null alleles, inbreeding, theWahlund effect, or from the ‘family’ Wahlund effect(i.e. non-random sampling from a limited number offamilies; Pudovkin, Zaykin & Hedgecock, 1996). Dis-entangling these factors is notoriously difficult (e.g.Castric et al., 2002). Although loci with high null-allele frequencies were discarded, the influence of nullalleles cannot be completely ruled out. However, thelow rates of null alleles computed with FREENA(mean over loci and populations < 0.04) for the 12remaining loci, and the fact that heterozygote defi-ciency occurs for most loci and populations, suggeststhat this deficiency is probably inherent to the speciesbiology. Although the results of the relatedness coef-ficient analysis allow us to reject the occurrence ofbiparental inbreeding in the samples, they confirmthe presence of several small groups of related indi-viduals in three localities. One explanation for theobserved heterozygote deficiencies could thus be aWahlund effect resulting from the presence of alimited number of ‘families’ or kin groups in oursamples. This result is consistent with the observedSGS that could result from the formation of a localpedigree structure (Wright, 1943). Interestingly, SGS

and heterozygote deficiencies on a similar spatialscale were also reported in Pristimantis kichwarum,but they were not associated with significant variancein relatedness values, leading Elmer et al. (2007a) toreject the occurrence of family structure, but they didnot provide a clear alternative explanation for thedeviation from panmixia. Pristimantis kichwarumbelongs to Terrarana, a clade of direct-developingfrogs for which territoriality has also been observed(Wells, 2007). Explosive breeders, on the other handgenerally exhibit a pattern of panmixia (Chan &Zamudio, 2009; Knopp & Merilä, 2009). Despite theobvious need for additional studies to further explorethis phenomenon in A. andreae, we suggest that theterritoriality and restricted dispersal abilities in thisspecies could result in the association of related indi-viduals in small, spatially structured groups that islikely to explain the deviation from panmixia.

The consequences of low dispersal ability may scaleup to explain patterns at larger geographical scales.The IBD observed among populations is consistentwith the restricted dispersal ability of A. andreae.Isolation by distance between amphibian populationshas been reported in many species (Vences & Wake,2007). In tropical species IBD was observed at com-parable spatial scales in Engystomops pustulosus(Lampert et al., 2003) and Eleutherodactylus coqui(Peacock et al., 2009). Although dispersal is verylimited in A. andreae, the degree of differentiation(F′ST; Table 3) observed among the four NON sites(~6 km) is not markedly different from that of mosttemperate anurans studied at similar spatial scales(e.g. Rana arvalis, Vos et al., 2001; Rana luteiventris,Funk et al., 2005; for a review, see Chan & Zamudio,2009). Measures of population differentiation such asFST are linked to gene flow, but also to other demo-graphic parameters such as effective population size(see Hedrick, 2005; Meirmans & Hedrick, 2011). Ourfocal species remains poorly known in terms of repro-duction biology (number of reproductions per year perindividual, survival, longevity, etc.), but we hypoth-esize that the effective population size is probablyvery high, and that it may compensate for the effectof low dispersal on FST compared with temperatespecies.

Rivers act as barriersAs found in other amphibians, our results also indi-cate that, in addition to life-history traits and dis-tance, rivers influence population structure by actingas barriers to gene flow (e.g. Hedgecock, 1978;Lampert et al., 2003; Noonan & Wray, 2006; Funket al., 2007). The AR appears to represent a consid-erable barrier to gene flow, especially in its lowerreaches, where it attains its greatest breadth (~300 min Régina versus ~100 m in Cisame). In the present



case, the downstream populations of A. andreae oneach side of the river (CIW and REW versus CIE andREE) are highly differentiated from their neighbour-ing populations on the opposite bank. The case of thetwo Cisame populations is compelling: they do notdisplay any allele sharing, but are only ~100 m apartand harbour mtDNA lineages that are highly diver-gent (~2.9 Mya). Moreover, the plot of FST/(1 - FST)versus ln(geographical distance) confirms a dis-continuity in transriverine comparisons (Fig. 5).Upstream, where the Arataï tributary is narrower(~30 m wide, and with the presence of islands) andprobably shallower, it appears to be a less effectivebarrier to migration, as migrants were detectedbi-directionally between NOS and NON with bothmitochondrial and nuclear markers.

Rivers are prominent landscape features in Amazo-nia, and their role in shaping the genetic structureand distribution of species has long been recognizedby biologists (Wallace, 1852). However, the impor-tance of rivers in shaping patterns of biodiversitymay depend on taxon-specific life-history traits (e.g.Gascon et al., 1998, 2000; Lougheed et al., 1999), orriver-specific features (Bates et al., 2004). The largerivers of the Amazon are known to be able to changecourse (Lundberg et al., 1998). However, this may beless common for smaller, more channelled, and stableclear-water rivers (Bates et al., 2004) of the Brazilianand Guianan shields (e.g. the AR). In the current caseof a small-bodied, terrestrial, and endotrophic species,the impact of rivers on the genetic structure is clearlyimportant. The generalization of this result to otherspecies along the Approuague and other riversremains to be investigated, as the impact of rivers onAmazonian biodiversity dynamics may be essential.

HISTORICAL PROCESSES AND

EVOLUTIONARY SCENARIO

Is dispersal so restricted by distance and rivers in thisparticular species that it could enhance the effect ofhistorical fragmentation, and be responsible for such apatchy phylogeographic pattern? The small body size,territoriality, mode of development (terrestrial, nolarval dispersal), and small clutch size of A. andreaeprobably act negatively on the levels of gene flow, andrivers act as efficient barriers. Conversely, the conti-nuity of the extensive forest habitat, putative largepopulation, and the ability to reproduce severaltimes each year (implicit from Kokubum & Giaretta,2005; Kokubum & de Sousa, 2008; Menin et al., 2009;A. Fouquet, pers. observ.) should promote gene flow.

The cumulative effects of limited dispersal andriver barriers clearly have a major impact on theobserved population structure. However, even suchlimited dispersal ability becomes significant over long

temporal scales (a simple projection of the mean axialparent–offspring distance of ~30 m, with a 1-yeargeneration time for 100 000 years leads to 2700 km).Moreover, given that cross-river gene flow does occurupstream, it is unlikely that only limited dispersaland rivers could have generated the overall geneticstructure seen in mtDNA and microsatellites. More-over, a strong genetic structure is also observed inproximate locations on each side of the river. Featuresof this structure at the scale of the Guiana Shield arerecurrent among co-occurring frog species (Noonan &Gaucher, 2005, 2006; Fouquet et al., 2012), and prob-ably have a concomitant origin in the early Pleis-tocene (Fouquet et al., 2012), which matches the basalsplit estimated herein at 2.9 Mya. This pattern hasalso been found in other organisms in this region (e.g.Duputié et al., 2009), indicating an origin that isprobably coincident with the drastic environmentalchanges that arose during Quaternary climaticfluctuations (e.g. Mayle et al., 2004; Colinvaux, DeOliveira & Bush, 2000). The exact nature of theseenvironmental changes, like the turnover of savan-nahs or dry forest versus evergreen forest, or justshifts in the composition of communities, is subject todebate, but the accumulation of evidence from phylo-geographic studies leaves little doubt that they havedrastically shaped the genetic structure of most GSorganisms.

Therefore, we argue that the effects of a particu-larly low dispersal ability and rivers overprinted thepre-existing structure in A. andreae, and produced afar more pronounced pattern than in any otherspecies studied to date in this area. In this view,rivers may have acted, and still do, as areas of sec-ondary contact, limiting the already low dispersalrates and thus preserving and subdividing the geneticstructure generated by past habitat disruption, ratherthan causing the initial isolation and divergence.Given that frog species with small body sizes anddirect or terrestrial development represent a largeproportion of the Neotropical amphibians, a signifi-cant portion of the amphibian species are likely toexhibit similar patterns over small spatial scales.

CONCLUSION

On a broader scale, clade-specific life-history traits, inconjunction with dispersal ability, may shape pat-terns of diversification (Losos, 2010). The tremendousdiversification of direct-developing anuran lineageshas been interpreted as potentially resulting fromtheir mode of reproduction, fostering high familymortality, and consequently diversification (Dubois,2005), a hypothesis contested by Vences & Wake(2007). An alternative view is that the consequencesof these modes of reproduction on genetic struc-



ture and diversification are linked to associatedlife-history traits. In other words, we argue thatdirect-developing/endotrophic amphibians, like mostAdenomera spp., are likely to have low dispersalcapabilities driven by a combination of life-historystrategy traits (small size, no larval dispersal, terri-toriality, etc.) that promote small-scale differentiationand ultimately speciation. Moreover, limited ecologi-cal tolerance to climatic disturbance promotes theisolation of small demes (Moussalli et al., 2009), andthis, combined with low dispersal ability, wouldpromote genetic differentiation and at least partlyexplain the observation of highly divergent lineagesin close proximity and even sympatry.

As suggested by Buckley (2009), the explanatorypower of phylogeographic inference would be greatlyenhanced by incorporating information from differentlevels, such as population genetics and molecularecology. Understanding the combined effects of lowgene flow and rivers helps distinguish the print ofhistorical events at larger scale in A. andreae. Thecase presented here provides insight into the com-plexity of interactions among mechanisms structuringAmazonian biodiversity.

ACKNOWLEDGEMENTS

This work has been supported by a grant to A.F. fromthe Centre National de la Recherche Scientifique‘Amazonie’ programme. Data used in this work weremostly produced through the molecular genetic analy-sis technical facilities of the ‘Institut National dela Recherche Agronomique’ Kourou. We are alsoindebted to the ‘Office National des Forêts’ and toMaël Dewynter in particular for allowing and helpingwith this research. Rémi Chappaz and André Gillesalso contributed to the completion of this project. Wethank Philippe Gaucher and Michel Blanc for helpingwith fieldwork organization, as well as PhilippeGilabert for hosting A.F. in camp CISAME. Alec Baxt,Partick Chatelet, Dirk Schmeller, Fabrice Hibert,Aurélie Condevaux, Max Ringler, Eva Ringler, andStéphane Icho also assisted with fieldwork. Labora-tory work benefited from the experience of ValérieTroispoux and Eliane Louisianna. We are also thank-ful to the four anonymous reviewers, whose remarksgreatly improved the article.

REFERENCES

Akaike H. 1981. A new look at the statistical–model identi-fication. Engineering, Technology and Applied Sciences 19:716–723.

Almeida AP, Angulo A. 2006. A new species of Leptodactylus(Anura: Leptodactylidae) from the state of Espírito Santo,

Brazil, with remarks on the systematics of associated popu-lations. Zootaxa 1334: 1–25.

Antonelli A, Quijada–Masareñas A, Crawford AJ, BatesAJ, Velazco JM, Wüster W. 2010. Molecular studies andphylogeography of Amazonian tetrapods and their relationto geological and climatic models. In: Hoorn C, WesselinghFP, eds. Amazonia, Landscape and Species Evolution, 1stedn. Oxford, UK: Blackwell Publishing Ltd., 386–404.

Austin JD, Davila JA, Lougheed SC, Boag PT. 2003.Genetic evidence for female–biased dispersal in the bullfrog,Rana catesbeiana (Ranidae). Molecular Ecology 12: 3165–3172.

Avila-Pires TCS. 1995. Lizards of brazilian Amazonia (Rep-tilia: Squamata). Zoologische verhandelingen 299: 1–706.

Bates HW. 1863–1864. The naturalist on the river Amazon, 2Vol. London: John Murray.

Bates JM, Haffer J, Grismer E. 2004. Avian mitochondrialDNA sequence divergence across a headwater stream ofthe Rio Tapajos, a major Amazonian river. Journal ofOrnithology 145: 199–205.

Belkhir K, Borsa P, Chikhi L, Raufaste N, Bonhomme F.1996–2004. GENETIX 4.05, logiciel sous Windows TM pourla génétique des populations, Laboratoire Génome, Popula-tions, Interactions, CNRS UMR 5000. Montpellier, France:Université de Montpellier II.

Belkhir K, Castric V, Bonhomme F. 2002. IDENTIX, asoftware to test for relatedness in a population using per-mutation methods. Molecular Ecology Notes 2: 611–614.

Benjamini Y, Hochberg Y. 1995. Controlling the false dis-covery rate: a practical and powerful approach to multipletesting. Journal of the Royal Statistical Society. Series B 57:289–300.

Berry O. 2001. Genetic evidence for wide dispersal by the sandfrog, Heleioporus psammophilus (Anura: Myobatrachidae),in Western Australia. Journal of Herpetology 35: 136–141.

Berven KA, Grudzien TA. 1990. Dispersal in the wood frog(Rana sylvatica) – Implications for genetic population–structure. Evolution 44: 2047–2056.

Blaustein AR, Wake DB, Sousa WP. 1994. Amphibiandeclines – Jjudging stability, persistence, and susceptibilityof populations to local and global extinctions. ConservationBiology 8: 60–71.

Bogart JP. 1991. The influence of life history on karyotypicevolution in frogs. In: Green DM, Sessions SK, eds. Amphib-ian Cytogenetics and Evolution. San Diego, CA: AcademicPress, 233–358.

Boisselier-Dubayle MC, Leblois R, Samadi S, Lambour-dière J, Khodabux MI, Sarthou C. 2010. Populationstructure of the Bromeliaceae Pitcairnia geyskesii from theFrench Guianan inselbergs: possible implications on pasthistory of rainforests. Ecography. 33: 175–184.

Breden F. 1987. The effect of post–metamorphic dispersal onthe population genetic structure of Fowler’s toad, Bufowoodhousei fowleri. Copeia 1987: 386–395.

Broquet T, Johnson CA, Petit E, Thompson I, Burel F,Fryxell J. 2006. Dispersal and genetic structure in theAmerican marten, Martes americana. Molecular Ecology 15:1689–1697.



Buckley D. 2009. Toward an organismal, integrative, anditerative phylogeography. BioEssays 31: 784–793.

Buckley LB, Jetz W. 2007. Environmental and historicalconstraints on global patterns of amphibian richness. Pro-ceedings of the Royal Society B-Biological Sciences 274:1167–1173.

Cabe PR, Page RB, Hanlon TJ, Aldrich ME, Marsh DM.2007. Fine-scale population differentiation and gene flow ina terrestrial salamander (Plethodon cinereus) living in con-tinuous habitat. Heredity 98: 53–60.

Carnaval AC, Hickerson MJ, Haddad CFB, RodriguesMT, Moritz C. 2009. Stability predicts genetic diversity inthe brazilian atlantic forest hotspot. Science 323: 785–789.

Castric V, Belkhir K, Bernatchez L, Bonhomme F. 2002.Heterozygote deficiencies in small lacustrine populations ofbrook charr Salvelinus fontinalis Mitchill (Pisces, Salmo-nidae): a test of alternative hypotheses. Heredity 89: 27–35.

Chan LM, Zamudio KR. 2009. Population differentiation oftemperate amphibians in unpredictable environments.Molecular Ecology 18: 3185–3200.

Chapuis MP, Estoup A. 2007. Microsatellite null alleles andestimation of population differentiation. Molecular Biologyand Evolution 24: 621–631.

Chek AA, Austin JD, Lougheed SC. 2003. Why is there atropical-temperate disparity in the genetic diversity andtaxonomy of species? : Evolutionary Ecology Research 5:69–77.

Clement M, Posada D, Crandall KA. 2000. TCS: a com-puter program to estimate gene genealogies. MolecularEcology 9: 1657–1659.

Clobert J, Danchin E, Dhondt AA, Nichols JD. 2001.Dispersal. New York: Oxford University Press.

Colinvaux PA, De Oliveira PE, Bush MB. 2000. Amazo-nian and Neotropical plant communities on glacial time-scales: The failure of the aridity and refuge hypotheses.Quaternary Science Review 19: 141–169.

Crawford AJ, Bermingham E, Polania C. 2007. The role oftropical dry forest as a long–term barrier to dispersal: acomparative phylogeographical analysis of dry forest toler-ant and intolerant frogs. Molecular Ecology 16: 4789–4807.

Daugherty CH, Sheldon AL. 1982. Age-specific movementpatterns of the frog Ascaphus truei. Herpetologica 38: 468–474.

Dempster AP, Laird NM, Rubin DB. 1977. Maximumlikelihood from incomplete data via the EM algorithm.Journal of the Royal Statistical Society Series B 34: 1–38.

Dixo M, Metzger JP, Morgante JS, Zamudio KR. 2009.Habitat fragmentation reduces genetic diversity and con-nectivity among toad populations in the brazilian atlanticcoastal forest. Biological Conservation 142: 1560–1569.

Driscoll DA. 1997. Mobility and metapopulation structure ofGeocrinia alba and Geocrinia vitellina, two endangered frogspecies from southwestern Australia. Australian Journal ofEcology 22: 185–195.

Driscoll DA. 1998. Genetic structure, metapopulation pro-cesses and evolution influence the conservation strategiesfor two endangered frog species. Biological Conservation 83:43–54.

Drummond AJ, Rambaut A. 2007. BEAST: Bayesian evo-lutionary analysis by sampling trees. BMC EvolutionaryBiology 7: 214.

Dubois A. 2005. Developmental pathway, speciation andsupraspecific taxonomy in amphibians 1. Why are there somany frog species in Sri Lanka? Alytes 22: 19–37.

Duellman WE, Trueb L. 1986. Biology of amphibians. NewYork: McGraw Hill.

Duputié A, Deletre M, De Granville JJ, Mckey D. 2009.Population genetics of Manihot esculenta ssp flabellifoliagives insight into past distribution of xeric vegetation in apostulated forest refugium area in northern Amazonia.Molecular Ecology 18: 2897–2907.

Eckert CG, Samis KE, Lougheed SC. 2008. Genetic varia-tion across species’ geographical ranges: the central-marginal hypothesis and beyond. Molecular Ecology 17:1170–1188.

Elmer KR, Davila JA, Lougheed SC. 2007a. Applying newinter–individual approaches to assess fine–scale populationgenetic diversity in a Neotropical frog, Eleutherodactylusockendeni. Heredity 99: 506–515.

Elmer KR, Davila JA, Lougheed SC. 2007b. Cryptic diver-sity and deep divergence in an upper Amazonian leaflitterfrog, Eleutherodactylus ockendeni. BMC EvolutionaryBiology 7: 247.

Evanno G, Regnaut S, Goudet J. 2005. Detecting thenumber of clusters of individuals using the software STRUC-TURE: a simulation study. Molecular Ecology 14: 2611–2620.

Falush D, Stephens M, Pritchard JK. 2003. Inference ofpopulation structure using multilocus genotype data: linkedloci and correlated allele frequencies. Genetics 164: 1567–1587.

Falush D, Stephens M, Pritchard JK. 2007. Inference ofpopulation structure using multilocus genotype data: domi-nant markers and null alleles. Molecular Ecology Notes 7:574–578.

Felsenstein J. 2005. PHYLIP (Phylogeny Inference Package),version 3.6. Distributed by the author. Seattle, WA: Depart-ment of Genome Sciences, University of Washington.

Fenster CB, Vekemans X, Hardy OJ. 2003. Quantifyinggene flow from spatial genetic structure data in a metapo-pulation of Chamaecrista fasciculate (Leguminosae). Evolu-tion 57: 995–1007.

Fouquet A, Dubut V, Hataway RA, Scotti–Saintagne C,Scotti I, Noonan BP. 2009. Isolation and characterisationof 19 microsatellite loci from the Amazonian frog Adenomeraandreae (Amphibia: Anura: Leptodactylidae). ConservationGenetics Resources 1: 217–220.

Fouquet A, Gilles A, Vences M, Marty C, Blanc M,Gemmell NJ. 2007a. Underestimation of species richnessin neotropical frogs revealed by mtDNA analyses. PLoSONE 2: e1109.

Fouquet A, Noonan BP, Rodrigues MT, Pech N, Gilles A,Gemmell NJ. 2012. Multiple quaternary refugia in theeastern Guiana Shield revealed by comparative phylogeog-raphy of 12 frog species. Systematic Biology. doi: 10.1093/sysbio/syr130

Fouquet A, Vences M, Salducci MD, Meyer A, Marty C,



Blanc M, Gilles A. 2007b. Revealing cryptic diversityusing molecular phylogenetics and phylogeography in frogsof the Scinax ruber and Rhinella margaritifera speciesgroups. Molecular Phylogenetics and Evolution 43: 567–582.

Frost DR, Grant T, Faivovich J, Bain RH, Haas A,Haddad CFB, de Sá RO, Channing A, Wilkinson M,Donnellan SC, Raxworthy CJ, Campbell JA, BlottoBL, Moler P, Drewes RC, Nussbaum RA, Lynch JD,Green DM, Wheeler WC. 2006. The amphibian tree of life.Bulletin of the American Museum of Natural History 297:1–370.

Funk WC, Blouin MS, Corn PS, Maxell BA, Pilliod DS,Amish S, Allendorf FW. 2005. Population structure ofColumbia spotted frogs (Rana luteiventris) is stronglyaffected by the landscape. Molecular Ecology 14: 483–496.

Funk WC, Caldwell JP, Peden CE, Padial JM, De la RivaI, Cannatella DC. 2007. Tests of biogeographic hypothesesfor diversification in the Amazonian forest frog, Physalae-mus petersi. Molecular Phylogenetics and Evolution 44:825–837.

García-París M, Good DA, Parra-Olea G, Wake DB. 2000.Biodiversity of Costa Rican salamanders: Implications ofhigh levels of genetic differentiation and phylogeographicstructure for species formation. Proceedings of the NationalAcademy of Sciences of the United States of America 97:1640–1647.

Gascon C, Lougheed SC, Bogart JP. 1998. Patterns ofgenetic population differentiation in four species of Amazo-nian frogs: A test of the riverine barrier hypothesis. Biotro-pica 30: 104–119.

Gascon C, Malcolm JR, Patton JL, da Silva MNF, BogartJP, Lougheed SC, Peres CA, Neckel S, Boag PT. 2000.Riverine barriers and the geographic distribution of Ama-zonian species. Proceedings of the National Academy ofSciences of the United States of America 97: 13672–13677.

Gaston KJ, Williams PH. 1996. Spatial patterns in taxo-nomic diversity. In: Gaston KJ, ed. Biodiversity: A biology ofnumbers and difference. Mississauga, Ontario, Canada:Blackwell Science, 202–229.

Gauffre B, Estoup A, Bretagnolle V, Cosson JF. 2008.Spatial genetic structure of a small rodent in a heteroge-neous landscape. Molecular Ecolog 17: 4619–4629.

Goudet J. 1995. FSTAT (vers. 1.2): a computer program tocalculate F–statistics. Journal of Heredity 86: 485–486.

Grant T, Frost DR, Caldwell JP, Gagliardo R, HaddadCFB, Kok PJR, Means DB, Noonan BP, Schargel WE,Wheeler WC. 2006. Phylogenetic systematics of dart–poison frogs and their relatives (Amphibia: Athesphatanura:Dendrobatidae). Bulletin of the American Museum ofNatural History 299: 1–262.

Greenwood PJ. 1980. Mating systems, philopatry and dis-persal in birds and mammals. Animal Behavior 28: 1140–1162.

Guo SW, Thompson EA. 1992. Performing the exact test ofHardy–Weinberg proportion for multiple alleles. Biometrics48: 367–372.

Haffer J. 1993a. On the river effect in some forest birds ofsouthern Amazonia. Boletim do Museu Paraense EmilioGoeldi, Zoologia 8: 217–245.

Haffer J. 1993b. Time’s cycle and time’s arrow in the historyof Amazonia. Biogeographica 69: 15–45.

Haffer J. 1997. Contact zones between birds of southern.Amazonia. Ornithological Monographs 48: 281–305.

Hardy OJ, Vekemans X. 1999. Isolation by distance in acontinuous population: reconciliation between spatial auto-correlation analysis and population genetics models. Hered-ity 83: 145–154.

Hardy OJ, Vekemans X. 2002. SPAGeDi: a versatile com-puter program to analyse spatial genetic structure at theindividual or population levels. Molecular Ecology Notes 2:618–620.

Hedgecock D. 1978. Population subdivision and geneticdivergence in the Red–Bellied Newt, Taricha rivularis. Evo-lution 32: 271–286.

Hedrick PW. 1999. Perspective: highly variable loci and theirinterpretation in evolution and conservation. Evolution 53:313–318.

Hedrick PW. 2005. A standardized genetic differentiationmeasure. Evolution 59: 1633–1638.

Heinicke MP, Duellman WE, Hedges SB. 2007. Majorcaribbean and central american frog faunas originated byancient oceanic dispersal. Proceedings of the NationalAcademy of Sciences of the United States of America 104:10092–10097.

Jakobsson M, Rosenberg NA. 2007. CLUMPP: a clustermatching and permutation program for dealing with labelswitching and multimodality in analysis of population struc-ture. Bioinformatics 23: 1801–1806.

Jensen JL, Bohonak AJ, Kelley ST. 2005. Isolation bydistance, web service. BMC Genetics 6: 13; v.3.15. Availableat: http://ibdws.sdsu.edu/

Knopp T, Merilä J. 2009. Microsatellite variation and popu-lation structure of the moor frog (Rana arvalis) in Scandi-navia. Molecular Ecology 18: 2996–3005.

Kokubum MNDC, Giaretta AA. 2005. Reproductive ecologyand behaviour of a species of Adenomera (Anura, Leptodac-tylidae) with endotrophic tadpoles: Systematic implications.Journal of Natural History 39: 1745–1758.

Kokubum MNDC, de Sousa MB. 2008. Reproductiveecology of Leptodactylus aff hylaedactylus (Anura, Leptodac-tylidae) from an open area in northern Brazil. South Ameri-can Journal of Herpetology 3: 15–21.

Lampert KP, Rand AS, Mueller UG, Ryan MJ. 2003.Fine–scale genetic pattern and evidence for sex–biased dis-persal in the tungara frog, Physalaemus pustulosus. Molecu-lar Ecology 12: 3325–3334.

Leblois R, Rousset F, Tikel D, Moritz C, Estoup A. 2000.Absence of evidence for isolation by distance in an expand-ing cane toad (Bufo marinus) population: an individual-based analysis of microsatellite genotypes. MolecularEcology 9: 1905–1909.

Lindsey CC. 1966. Body size of poikilotherm vertebrates atdifferent latitudes. Evolution 20: 456–465.

Loiselle BA, Sork VL, Nason J, Graham C. 1995. Spatial



genetic structure of a tropical understory shrub, Psychotriaofficinalis (Rubiaceae). American Journal of Botany 82:1420–1425.

Losos JB. 2010. Adaptive radiation, ecological opportunity,and evolutionary determinism. American Naturalist 175:623–639.

Lötters S, van der Meijden A, Rödder D, Köster TE,Kraus T, La Marca E, Haddad CFB, Veith M. 2010.Reinforcing and expanding the predictions of the distur-bance vicariance hypothesis in Amazonian harlequin frogs:a molecular phylogenetic and climate envelope modellingapproach. Biodiversity Conservation 19: 2125–2146.

Lougheed SC, Gascon C, Jones DA, Bogart JP, Boag PT.1999. Ridges and rivers: a test of competing hypothesesof Amazonian diversification using a dart–poison frog(Epipedobates femoralis). Proceedings of the Royal Societyof London Series B–Biological Sciences 266: 1829–1835.

Lundberg JG, Marshall LG, Guerrero J, Horton B, Mala-barba MCSL, Wesselingh F. 1998. The stage for neotro-pical fish diversification: a history of South American rivers.In: Malabarba MCSL, Reis RE, Vari RP, Lucena ZM,Lucena CAS, eds. Phylogeny and Classification of Neotropi-cal Fishes. Porto Alegre, Brazil: Edipucrs, 13–48.

Marsh DM, Trenham PC. 2001. Metapopulation dynamicsand amphibian conservation. Conservation Biology 15:40–49.

Mayle FE, Beerling DJ, Gosling WD, Bush MB. 2004.Responses of Amazonian ecosystems to climatic and atmo-spheric carbon dioxide changes since the last glacialmaximum. Proceedings of the Royal Society B-BiologicalSciences 359: 499–514.

Meirmans PG. 2006. Using the AMOVA framework toestimate a standardized genetic differentiation measure.Evolution 60: 2399–2402.

Meirmans PG, Hedrick PW. 2011. Assessing populationstructure: FST and related measures. Molecular EcologyResources 11: 5–18.

Menin M, Almeida AP, Kokubum MNC. 2009. Reproduc-tive aspects of Leptodactytus hylaedactylus (Anura: Lepto-dactylidae), a member of the Leptodactylus marmoratusspecies group, with a description of tadpoles and calls.Journal of Natural History 43: 2257–2270.

Monsen KJ, Blouin MS. 2003. Genetic structure in amontane ranid frog: restricted gene flow and nuclear–mitochondrial discordance. Molecular Ecology 12: 3275–3286.

Moussalli A, Moritz C, Williams SE, Carnaval AC. 2009.Variable responses of skinks to a common history ofrainforest fluctuation: concordance between phylogeographyand palaeo–distribution models. Molecular Ecology 18: 483–499.

Müller L. 1923. Neue oder seltene Reptilien und Batrachierder zoologischen Sammlung des bayerischen Staates. Zoolo-gischer Anzeiger 57: 38–42.

Myers N, Mittermeier RA, Mittermeier CG, Da FonsecaGAB, Kent J. 2000. Biodiversity hotspots for conservationpriorities. Nature 403: 853–858.

Nei M. 1973. Analysis of gene diversity in subdivided popu-lations. Proceedings of the National Academy of Sciences ofthe United States of America 70: 3321–3323.

Newman RA, Squire T. 2001. Microsatellite variation andfine–scale population structure in the wood frog (Rana syl-vatica). Molecular Ecology 10: 1087–1100.

Noonan BP, Gaucher P. 2005. Phylogeography and demog-raphy of Guianan harlequin toads (Atelopus): diversificationwithin a refuge. Molecular Ecology 14: 3017–3031.

Noonan BP, Gaucher P. 2006. Refugial isolation and sec-ondary contact in the dyeing poison frog Dendrobates tinc-torius. Molecular Ecology 15: 4425–4435.

Noonan BP, Wray KP. 2006. Neotropical diversification: theeffects of a complex history on diversity within the poisonfrog genus Dendrobates. Journal of Biogeography 33: 1007–1020.

Nylander JA. 2004. Mrmodeltest 2.3. Program distributed bythe author. Uppsala, Sweden: Evolutionary Biology Centre,Uppsala University.

Patton JL, Da Silva MNF, Malcolm JR. 1996. Hierarchicalgenetic structure and gene flow in three sympatric species ofAmazonian rodents. Molecular Ecology 5: 229–238.

Peacock MM, Beard KH, O’Neill EM, Kirchoff VS, PetersMB. 2009. Strong founder effects and low genetic diversityin introduced populations of Coqui frogs. Molecular Ecology18: 3603–3615.

Pritchard JK, Stephens M, Donnelly P. 2000. Inference ofpopulation structure using multilocus genotype data. Genet-ics 155: 945–959.

Pritchard JK, Wen X, Falush D. 2007. Documentation forthe STRUCTURE software, Version 2. Chicago. Available at:http://pritch.bds.uchicago.edu

Pudovkin AI, Zaykin D, Hedgecock D. 1996. On the poten-tial for estimating effective number of breeders fromheterozygote–excess in progeny. Genetics 144: 383–387.

Rambaut A, Drummond AJ. 2007. Tracer v1.4, Availablefrom http://beast.bio.ed.ac.uk/Tracer

Raymond M, Rousset F. 1995. GENEPOP (version 1.2):population genetics software for exact tests and ecumeni-cism. Heredity 86: 248–249.

Robertson JM, Duryea MC, Zamudio KR. 2009. Discor-dant patterns of evolutionary differentiation in two Neotro-pical treefrogs. Molecular Ecology 18: 1375–1395.

Rosenberg NA. 2004. DISTRUCT: a program for thegraphical display of population structure. Molecular EcologyNotes 4: 137–138.

Rousset F. 1997. Genetic differentiation and estimation ofgene flow from F–statistics under isolation by distance.Genetics 145: 1219–1228.

Rousset F. 2000. Genetic differentiation between individuals.Journal of Evolutionary Biology 13: 58–62.

Rousset F. 2008. Genepop’007: a complete reimplementationof the Genepop software for Windows and Linux. MolecularEcology Resources 8: 103–106.

Rowe G, Beebee TJC, Burke T. 2000. A microsatelliteanalysis of natterjack toad, Bufo calamita, metapopulations.Oikos 88: 641–651.

Saitou N, Nei M. 1987. The neighbor–joining method – a new



method for reconstructing phylogenetic trees. MolecularBiology and Evolution 4: 406–425.

Santos JC, Coloma LA, Summers K, Caldwell JP, Ree R,Cannatella DC. 2009. Amazonian amphibian diversity isprimarily derived from late Miocene andean lineages. PLoSBiology 7: 448–461.

Selonen V, Hanski IK, Painter JN. 2010. Gene flow andnatal dispersal in the Siberian flying squirrel based on directand indirect data. Conservation Genetics 11: 1257–1264.

Shaffer HB, Fellers G, Magee A, Voss R. 2000. The genet-ics of amphibian declines: population substructure andmolecular differentiation in the Yosemite toad, Bufo canorus(Anura, Bufonidae) based on single–strand conformationpolymorphism analysis (SSCP) and mitochondrial DNAsequence data. Molecular Ecology 9: 245–257.

Smith MA, Green DM. 2005. Dispersal and the metapopu-lation paradigm in amphibian ecology and conservation: areall amphibian populations metapopulations? Ecography 28:110–128.

Squire T, Newman RA. 2002. Genetic population structureof the wood frog (Rana sylvatica) in a northern woodland.Herpetologica 58: 119–130.

Tallmon DA, Funk WC, Dunlap WW, Allendorf FW. 2000.Genetic differentiation among long–toed salamander(Ambystoma macrodactylum) populations. Copeia 2000:27–35.

Templeton AR, Crandall KA, Sing CF. 1992. A Cladistic-Analysis of Phenotypic Associations with HaplotypesInferred from Restriction Endonuclease Mapping and DNA-Sequence Data. III. Cladogram Estimation. Genetics 132:619–633.

Trenham PC, Koenig WD, Shaffer HB. 2001. Spatiallyautocorrelated demography and interpond dispersal in thesalamander Ambystoma californiense. Ecology 82: 3519–3530.

van Bocxlaer I, Loader SP, Roelants K, Biju SD,Menegon M, Bossuyt F. 2010. Gradual adaptation towarda range–expansion phenotype initiated the global radiationof toads. Science 327: 679–682.

Van Oosterhout C, Hutchinson WF, Wills DPM, Shipley

P. 2004. MICRO–CHECKER: software for identifying andcorrecting genotyping errors in microsatellite data. Molecu-lar Ecology Notes 4: 535–538.

Vekemans X, Hardy OJ. 2004. New insights from fine-scalespatial genetic structure analyses in plant populations.Molecular Ecology 13: 921–935.

Vences M, Wake D. 2007. Speciation, species boundaries andphylogeography of amphibians. In: Heatwole H, Tyler M,eds. Amphibian biology, Vol. 6. Chipping Norton, Australia:Surrey Beatty and Sons, 2613–2669.

Vos CC, Antonisse–De Jong AG, Goedhart PW, SmuldersMJM. 2001. Genetic similarity as a measure for connectiv-ity between fragmented moor frog (Rana arvalis) popula-tions. Heredity 86: 598–608.

Waldman B, McKinnon JS. 1993. Inbreeding and outbreed-ing in fishes, amphibians, and reptiles. In: Thornhill NW,ed. The natural history of inbreeding and outbreeding: theo-retical and empirical perspectives. Chicago, IL: University ofChicago Press, 250–282.

Wallace AR. 1852. The monkeys of the Amazon. Proceedingsof the Zoological Society of London 20: 107–110.

Waples RS, Gaggiotti O. 2006. What is a population? Anempirical evaluation of some genetic methods for identifyingthe number of gene pools and their degree of connectivity.Molecular Ecology 15: 1419–1439.

Weir BS, Cockerham CC. 1984. Estimating F–statistics forthe analysis of population structure. Evolution 38: 1358–1370.

Wells KD. 2007. The ecology and behavior of Amphibians.Chicago, IL: The University of Chicago Press.

Wiens JA. 2001. The landscape context of dispersal. In:Clobert J, Danchin E, Dhondt AA, Nichols JD, eds. Dis-persal. Oxford, UK: Oxford University Press, 96–109.

Wilson EO. 1992. The Diversity of Life. Cambridge, MA:Belknap Press of Harvard University.

Wright S. 1943. Isolation by distance. Genetics 28: 114–138.

Zeisset I, Beebee T. 2008. Amphibian phylogeography: amodel for understanding historical aspects of species distri-butions. Heredity 101: 109–119.

ARCHIVED DATA

Data deposited at Dryad: doi:10.5061/dryad.7fb195gg

SUPPORTING INFORMATION

Additional supporting information may be found in the online version of this article:

Appendix S1. Samples information including for each specimen: field number, population acronym, Cyt bhaplotype number, geographical coordinates and sequence accession number.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materialssupplied by the authors. Any queries (other than missing material) should be directed to the correspondingauthor for the article.



The interplay of dispersal limitation, rivers, and ... · The interplay of dispersal limitation,...

Documents

Transcript of The interplay of dispersal limitation, rivers, and ... · The interplay of dispersal limitation,...