A Geographic Mosaic of Evolutionary Lineages Within the Insular Endemic Newt Euproctus Montanus

Molecular Ecology (2013) 22, 143–156 doi: 10.1111/mec.12085

A geographic mosaic of evolutionary lineages within the insular endemic newt Euproctus montanus

ROBERTA BISCONTI,*1 DANIELE CANESTRELLI,*1 DANIELE SALVI† and GIUSEPPE NASCETTI* *Dipartimento di Scienze Ecologiche e Biologiche, Universita` della Tuscia, Viale dell’Universita` s.n.c., I-01100, Viterbo, Italy, †CIBIO, Centro de Investigaçaõ em Biodiversidade e Recursos Gene´ticos, Campus Agra´rio de Vairaõ, 4485-661 Vairaõ, Portugal

Abstract Islands are hotspots of biodiversity, with a disproportionately high fraction of endemic lineages, often of ancient origin. Nevertheless, intra-island phylogeographies are surprisingly scarce, leading to a scanty knowledge about the microevolutionary processes induced on island populations by Plio-Pleistocene climatic oscillations, and the manner in which these processes contributed to shape their current genetic diversity. We investigated the phylogeography, historical demography and species distribution models of the Corsican endemic newt Euproctus montanus (north-western Mediterranean). As for many island endemics, the continuous distribution of E. montanus throughout its range has hitherto been considered as evidence for a single large population, a belief that also guided the species’ categorization for conservation purposes. Instead, we found a geographic mosaic of ancient evolutionary lineages, with ﬁve main clades of likely Pliocene origin (2.6–5.8 My), all but one restricted to northern Corsica. Moreover, the copresence between main lineages in the same population was limited to a single case. As also suggested by growing literature on intra-island phylogeographic variation, it seems that the extensive use of simplifying assumption on the population structure and historical demography of island populations—both in theoretical and applicative studies—should be carefully reconsidered, a claim that is well exempliﬁed by the case presented here.

Keywords: Corsica, diversiﬁcation, Euproctus montanus, intra-island phylogeography, Island biogeography, Mediterranean basin Received 16 February 2012; revision received 10 September 2012; accepted 13 September 2012

Taberlet et al. 1998; Thompson 2005; Nieberding et al. Introduction 2006; Schmitt 2007; Stock et al. 2008; Papadopoulou In the last three decades, the temperate species of the et al. 2009; La´zaro et al. 2011; Stroscio et al. 2011). Nev- Mediterranean region have been the subject of one of ertheless, there is also increasing realization that there the most intensive phylogeographic surveys (Feliner are fundamental questions about this history still wait- 2011; Hewitt 2011a). As a consequence, there is now ing thorough answers, including how populations ample knowledge of key issues of the Plio-Pleistocene survived climatic oscillations within areas of long-term history of both mainland and island species, including persistence (either glacial refugia or islands), what the location of major glacial refugia, hotspots of genetic microevolutionary processes were induced by these diversity, postglacial re-colonization routes of northern oscillations, and how these processes have shaped the environments, patterns of island occupancy and patterns current genetic structure and diversity within these of subsequent gene exchange among islands and with areas (Hampe & Petit 2005; Gomez & Lunt 2007; the continent (e.g. Hewitt 1996, 2000, 2004, 2011a,b; Canestrelli et al. 2010; Feliner 2011). In fact, these areas have been crucial for the survival of species, of their Correspondence: Daniele Canestrelli, Fax: +039-761-357751; intraspeciﬁc diversity and thus of their evolutionary E-mail: [email protected] potential, and they are particularly threatened by 1These authors contributed equally to this work. the recent climate changes and other human-induced

© 2012 Blackwell Publishing Ltd 144 R. BISCONTI ET AL. environmental alterations (Hampe & Petit 2005; Arau´ jo islands (Cronk 1997), has not been a prominent feature et al. 2006; Kier et al. 2009). Therefore, a thorough of this island. Indeed, during the last glaciation, north- appreciation of how diversity has been moulded within ern Corsica was substantially invaded by polar air these areas is of primary interest under evolutionary, (Kuhlemann et al. 2008), and thus, climatic oscillations ecological and conservation perspectives (see Hampe & were more pronounced here than in neighbouring areas. Petit 2005; Canestrelli et al. 2010; Feliner 2011 for length- Moreover, extensive glaciers formed on the central ier discussions of these issues). mountain chain during Pleistocene glacial phases To move in this direction, islands appear particularly (Kuhlemann et al. 2005). Finally, as few phylogeographic appealing (see also Thorpe & Malhotra 1998; Emerson studies have been conducted thus far at the intra-island et al. 2006). First, islands have a longstanding and fruit- level on Corsican species (see Discussion), the genetic ful history as natural laboratories for the study of evo- imprints of Plio-Pleistocene climatic oscillations on this lutionary processes (e.g. Darwin 1859; Wallace 1880; island biota are still largely unknown. MacArthur & Wilson 1967; Grant 1998; Grant & Grant The Corsican endemic Euproctus montanus seems an 2007; Whittaker & Fernańdez-Palacios 2007). In fact, outstanding model organism to start looking at these several features of islands, particularly that they are imprints. It is a lung-less newt, widely and continu- discrete and isolated spatial entities, allow making a ously distributed on the island, particularly—albeit not number of simplifying assumptions that are useful for exclusively—at intermediate and high altitudes both experimental designs and data interpretation (e.g. (between 600 and 1500 m above sea level; Gasc et al. Frankham 1996a,b; Woolfit & Bromham 2005; Whittaker 1997). It has primarily aquatic habits, living within & Fernańdez-Palacios 2007; Vellend & Orrock 2010). mountain streams, brooks (which are abundant on the Second, both at the global level and in the Mediterra- island) or ponds during the aquatic period, or close to nean basin, islands are hotspots of biodiversity, with a them during the terrestrial period (Gasc et al. 1997). disproportionately high fraction of this diversity being Interestingly, as no study has yet investigated the popu- endemic, a feature that is consistent across taxa and lation structure of E. montanus, it has not been given a geographic regions and that makes islands a conserva- high conservation priority in the International Union tion priority (Thompson 2005; Whittaker & Fernańdez- for Conservation of Nature (IUCN) Red List in spite of Palacios 2007; Vogiatzakis et al. 2008; Kier et al. 2009; its restricted range, because ‘…although its Extent of Me´dail & Diadema 2009; Blondel et al. 2010). Third, in Occurrence might be <20 000 km2, it is common with a spite of their usefulness as natural laboratories and their presumed large population…’ (IUCN 2011). disproportionate importance as biodiversity hotspots Previous studies of the phylogenetic relationships both at global and at regional levels, <5% of the phylog- and divergence timing of E. montanus indicated that this eographic studies carried out so far (nearly 7100, species started diverging from the closely related Sardi- according to ISI WoK database) presented and discussed nian brook newt Euproctus platycephalus in the Miocene intra-island phylogeographic patterns (Canestrelli & (Caccone et al. 1994; Carranza & Amat 2005; Veith et al. Bisconti in prep.). 2004; see also below). In this study, we analyse the phy- If islands can be used as natural laboratories for the logeographic and historical demographic patterns of study of microevolutionary processes, then Corsica E. montanus. Our aim is to assess whether and how the Island is certainly a good candidate top-level laboratory Plio-Pleistocene climatic oscillations influenced the evo- within the Mediterranean basin. It is a hotspot of Medi- lutionary history of this island endemic species and, terranean biodiversity, with both flora and fauna that more specifically, what microevolutionary processes are rich of endemic species (Thompson 2005; Mouillot they primed, and how these processes contributed et al. 2008). It is the northernmost, the wettest and the shaping its current patterns of population genetic struc- most mountainous island of the Mediterranean basin, ture and diversity. Finally, we will also evaluate the with a central mountain chain having many summits conservation implications of our results. exceeding 2000 m in altitude. Consequently, it has a particularly complex array of landscapes and microcli- Materials and methods matic regions, spanning from Mediterranean climate at low altitudes, to temperate montane climate at interme- Sampling and laboratory procedures diate altitudes, to alpine climate at higher altitudes (see Mouillot et al. 2008 for an extensive description of the We sampled individuals of Euproctus montanus from 15 Corsican environments). Furthermore—and interest- localities spanning the whole species range. Geographic ingly—the marine buffering of climatic oscillations, references for sampling localities and sample sizes are which is thought to contribute to the high endemism shown in Table 1 and Fig. 1A. Tissue samples were richness and to the persistence of old lineages on collected as tail tips after anaesthesia in a 0.1% solution

Table 1 Geographic location of the 15 populations sampled of Euproctus montanus and number of individuals analysed for both mitochondrial DNA (mtDNA) and nuclear DNA (nuDNA) markers

nuDNA

Locality Latitude N Longitude E mtDNA GH NCX1

1 Naseo 41°34′ 9°05′ 11 7 11 2 Agnarone 41°40′ 9°11′ 15 10 10 3 Spartu 41°46′ 9°13′ 17 3 4 4 Col de Bavella 41°48′ 9°14′ 19 6 7 5 Scrivano 41°57′ 9°11′ 1 – 1 6 Col de Verde 42°02′ 9°11′ 18 6 6 7 Vizzavona 42°07′ 9°08′ 18 7 7 8 Gorges de la Restonica 42°16′ 9°05′ 13 5 4 9 Aitone 42°15′ 8°50′ 13 6 9 10 Parata 42°22′ 9°24′ 9411 11 San-Gavino-d’Ampugnan 42°24′ 9°25′ 13 7 9 12 Foreˆt Communale d’Asco 42°25′ 8°57′ 333 13 Campitello 42°31′ 9°20′ 21 10 10 14 Bocca Capanna 42°33′ 9°03′ 654 15 Vignale 42°57′ 9°23′ 16 6 8

100/100 EI (A) (B)E (C) 95/98

95/98 EII 1155

99/100 D I 1144 G 1133 86/97 1122 CI 1111 C 1100 100/100 9 8 91/99 CII H 7 90/100 B 6 5 F 86/95 AII 81/73 4 84/99

3 AIII A 97/99 2 100/100 1

78/81 AI

E. platycephalus

0.01

Fig. 1 (A) Geographic location of the 15 sampled populations of Euproctus montanus and frequency distribution of the main haplogroups, shown as pie diagrams. Populations are numbered as in Table 1. (B) Maximum-likelihood (ML) phylogenetic tree based on the TrN + Γ model of sequence evolution. Bootstrap supports > 70% are shown at the nodes for both ML and MP tree-building methods (ML/MP); terminal haplogroups were collapsed. (C) Phylogenetic networks based on the statistical parsimony procedure. Circle sizes are proportional to haplotype frequency; missing intermediate haplotypes are shown as open dots. of MS222 (3-aminobenzoic acid ethyl ester) and were performed using the standard cetyltrimethylammonium stored in 96% ethanol. All the individuals were then bromide (CTAB) protocol (Doyle & Doyle 1987). Two released in the collection place. DNA extractions were mitochondrial DNA (mtDNA) and two nuclear DNA

(nuDNA) fragments were amplified and sequenced. phylogenetic signal. Thus, subsequent analyses were The mtDNA fragments were one from the cytochrome conducted on the combined data set. The best-fit model b gene (cytb) and one from the cytochrome oxidase I gene of nucleotide substitution for our data set was selected (cox1). The nuclear ones were from the growth hormone using the Bayesian information criterion as imple- + 2+ gene (GH) and the Na –Ca exchanger gene (NCX1). mented in JMODELTEST 0.1.1 (Posada 2008). The following specific primers were designed and Maximum-likelihood (ML) and maximum parsimony used to carry out PCR and sequencing reaction: 494SAL- (MP) methods were used to investigate phylogenetic AMOD (CATCAAACATCTCCTACTGATGAAA) and relationships among mitochondrial haplotypes. ML Mvz16mod (AAATAGGAARTATCAYTCTGGTTTRAT) analysis was performed using PHYML 3.0 (Guindon et al. for the cytb fragment; VF1deupr (TTCTCYACRAATCA 2010). The tree topology was estimated using the YAAAGACATTGG) and VR1deupr (TATACTTCAGGG SPR&NNI option and the best-fit model suggested by TGRCCAAAAAATCA) for the cox1 fragment; NCXF1 JMODELTEST. MP analysis was conducted in PAUP, with all (ATGATTATAGAAACAGAAGGTGATA) and NCXR1 characters equally weighted and unordered. A heuristic (GACTTTTATGTTATGAGGTGAAC) for the NCX1 frag- search was performed, with tree-bisection–reconnection ment; GHFH2 (TCTGGTTCAAAAATGTGTGTCA) and branch swapping and ten rounds of random sequence GHRHDEG (CATGTTTGTACATGGRTAGGTGA) for addition. The robustness of both ML and MP tree topol- the GH fragment. ogies was assessed by the nonparametric bootstrap Amplifications were performed in a 25-ll volume method with 1000 replicates. 9 containing MgCl2 (2.5 mM), the reaction buffer (5 ; Pro- The statistical parsimony procedure implemented in mega), the four dNTPs (0.2 mM each), the two primers the software TCS 1.2.1 (Clement et al. 2000) was used to (0.2 lM each), the enzyme Taq polymerase (1 U; Pro- infer phylogenetic networks among the haplotypes both mega) and 2 lL of DNA template. PCR cycling started for the concatenated mtDNA data set and for each with a step at 94 °C for 4 min followed by 35 cycles at nuDNA data set (Templeton et al. 1992). 94 °C for 1 min, an annealing step at 48 °C(cytb), 53 °C Times to the most recent common ancestor (TMRCA) (cox1), 56 °C(NCX1)or62°C(GH) for 1 min, 72 °C for of the main mtDNA haplogroups and divergence time 1 min, and a single final step at 72 °C for 10 min. Puri- among them were estimated through the distance-based fication and sequencing of PCR products were con- least squares (LS) method described by Xia & Yang ducted by Macrogen Inc. (http://www.macrogen.com). (2011) and implemented in the software DAMBE (Xia & All sequences were deposited in GenBank (accession Xie 2001). The hypothesis that our sequences evolved in numbers: JX626403-JX626977). a clock-like manner was tested by means of a likelihood ratio test as implemented in DAMBE. This test did not reject the molecular clock hypothesis. The ML tree as Phylogenetic analyses and molecular dating previously estimated by PHYML was used to specify a The electrophoregrams were checked using CHROMAS tree topology, and the divergence between E. montanus 2.31 (Technelysium Ltd.), and consensus sequences and its sister species Euproctus platycephalus was used to were aligned with CLUSTALX (Thompson et al. 1997). Het- set a calibration point. Caccone et al. (1994) suggested erozygous nuclear sequences were phased using the that E. montanus and E. platycephalus began diverging PHASE method as implemented in DNASP 5 (Librado & after the separation between Corsica and Sardinia Rozas 2009). For each nuclear gene, the possible occur- islands was completed (9 My). However, Carranza & rence of recombination events was assessed using the Amat (2005) suggested that this divergence would pairwise homoplasy index (PHI statistic, Bruen et al. date back to the onset of the Messinian salinity crisis 2006) implemented in SPLITSTREE v.4.11 (Huson & Bryant (MSC), 5.5 My. The estimation by Carranza & Amat 2006). Downstream analyses were conducted using was calibrated using a hypothesized split time between phased nuclear data and indels treated as missing data. Pleurodeles waltl and Pleurodeles poiretii of 5.33 My, Nucleotide variation was assessed using MEGA 5 corresponding to the end of the MSC. Nevertheless, an (Tamura et al. 2011), whereas haplotype (h) and nucleo- explicit test of this hypothesis against its alternatives tide (p) diversity (Nei 1987) and the associated standard showed that the scenario that would best fit molecular, deviations were computed using ARLEQUIN 3.5.1.2 fossil and paleogeographic data would involve a (Excoffier et al. 2005). much older split among Pleurodeles lineages, approxi- A partition homogeneity test (Farris et al. 1994) with mately 14 My (Veith et al. 2004). This would suggest an 100 replicates was performed in PAUP* 4.0b10 (Swofford accordingly older split between E. montanus and 2003) to test the homogeneity of the phylogenetic signal E. platycephalus, approximately at the time when Sardi- between the two mtDNA fragments. This test did not nia and Corsica islands began separating, 15 My (see reject the null hypothesis of the homogeneity of the Caccone et al. 1994). Furthermore, the use of the MSC

© 2012 Blackwell Publishing Ltd PHYLOGEOGRAPHY OF EUPROCTUS MONTANUS 147 as a calibration point has been recently criticized on observed distribution and the one expected under a several grounds (Hewitt 2011b). Therefore, we per- sudden expansion model of population growth were formed TMRCA estimates setting the split between compared (Rogers & Harpending 1992; Excoffier 2004). E. montanus and E. platycephalus alternatively to 9 My The sum of square deviations among the observed and (hereon referred to as calibration I) and 15 My (hereon estimated mismatch distributions was used as a good- referred to as calibration II). Finally, the LS analyses in ness-of-fit statistic, and its significance was evaluated by DAMBE were run using the ‘softbound’ option and the means of 1000 parametric bootstrap replicates. From the ‘MLCompositeTN93’ genetic distance, as suggested by mismatch distribution analysis, we also estimated the Xia & Yang (2011), with 1000 bootstrap re-samplings to parameter τ, the mutational time since the expansion obtain standard deviations (SD) of the estimates. (τ = 2ut, where u is the mutation rate per sequence and per generation, and t is time in generations). Finally, for each mtDNA lineage, we also computed the R (Ramos- Population genetic structure and historical 2 Onsins & Rozas 2002) and the F (Fu 1997) statistics using demography S the software DNASP 5 (Librado & Rozas 2009). To investigate the geographic pattern of genetic differentiation among populations, we used a modified ver- Spatial distribution modelling sion of the genetic landscape shape (GLS) interpolation analysis implemented in AIS (Miller 2005). Modifications To predict the potential distribution of E. montanus included the use of the best-fit model of sequence evo- under both current and glacial bioclimatic conditions, lution for our mtDNA data and a two-dimensional (2D) species distribution models (SDMs) were generated representation of the GLS, instead of the less easily using MAXENT 3.3.3e (Phillips et al. 2006), a program for interpretable 3D version yielded by AIS. A population maximum entropy modelling of the geographic distri- network was built using the Delaunay triangulation by butions of species, based on presence-only data. A total means of QUANTUM GIS (Quantum GIS Development of 104 data points were used to build the models (see Team 2012). The net between-population genetic diver- Appendix S3, Supporting Information). These included gences based on the mtDNA data were computed with our own 15 collection sites, plus 89 sites drawn from MEGA 5 (Tamura et al. 2011), using the best-fit model of previous literature or from museum collections. We built sequence evolution as suggested by JMODELTEST. Diver- the models using the default parameters for convergence À gence values among population pairs connected with threshold (10 5) and number of iterations (500). To edges by the Delaunay triangulation were imported in ensure the consistency of the model predictions, 75% of QUANTUM GIS and georeferenced to the mid-point of the the localities were used to train the model and 25% were corresponding edges. Finally, an inverse-distance- used to test it. weighting interpolation procedure was performed to An SDM was generated under present-day biocli- obtain the 2D representation of the GLS. matic conditions and then projected into last glacial The amount of genetic variation that can be maximum (LGM) conditions. The bioclimatic layers accounted for by differences among groups of popula- were downloaded from the WorldClim database website tion, among populations within groups and within pop- (http://www.worldclim.org). For the LGM prediction, ulations was assessed by carrying out an analysis of we used data from both the Community Climate molecular variance (AMOVA) with ARLEQUIN 3.5.1.2 (Excof- System Model (CCSM) and the Model for Interdisciplin- fier et al. 2005). Groups of populations were defined ary Research on Climate (MIROC). Layers for the pres- according to the main phylogeographic discontinuities ent conditions are available at a resolution of 30 arc as suggested by previous phylogenetic analyses. The seconds, while those for the LGM conditions are avail- analysis was run based on the mtDNA data, using the able at a resolution of 2.5 arc minutes, which was too best-fit model of sequence evolution as suggested by coarse grained for the geographic scale relevant in this JMODELTEST. The statistical significance of the variance study. Thus, we re-scaled these layers at a resolution of components and fixation indices was tested using 1092 30 arc seconds, using the following procedure: (i) we permutations. downloaded the layers at a 2.5-arc-minute resolution Historical demographic trends were investigated also for the present conditions, (ii) the difference through a mismatch distribution analysis (Rogers & Har- between present and LGM conditions was computed pending 1992) with the software ARLEQUIN. This distribu- for each variable with DIVA-GIS 7.5 (Hijmans et al. 2001), tion is expected to be unimodal and bell-shaped in and (iii) the obtained differences were then subtracted populations that underwent a sudden demographic to the 30-arc-second resolution layers for the present expansion, whereas a prolonged demographic stability is conditions. This procedure has the dual advantage of expected to lead to a multimodal distribution. The yielding data at a spatial resolution relevant to the

© 2012 Blackwell Publishing Ltd 148 R. BISCONTI ET AL. geographic scale of this study and at the same resolu- model with a gamma distribution shape parameter of tion of current data (Ro¨dder et al. 2010). 0.13 as the best-fit one for our mtDNA data. The WorldClim database provides 19 bioclimatic vari- The ML and MP algorithms yielded phylogenetic ables. We built two series of models: one with the entire trees with identical topologies at the main nodes, with set of variables and the other with only variables that minor differences limited to terminal nodes. The MP were not strongly correlated with each other (Pearson tree length was 452 steps (consistency index = 0.68; correlation coefficient, r2 < 0.80), choosing among corre- retention index = 0.93). The log-likelihood score for the lated variables those that we considered more biologi- ML tree was -4421.33. The ML tree is shown in Fig. 1B. cally significant for E. montanus. These variables Five main clades were found (clades A, B, C, D and E), included the following: BIO2, mean diurnal range in whose geographic distribution is shown in Fig. 1A. temperature; BIO4, temperature seasonality (SD 9100); Four of these clades (clades B, C, D and E) were geo- BIO5, maximum temperature of warmest month; BIO12, graphically restricted to no more than two populations mean temperature of the coldest quarter; BIO14, precip- in the northern portion of the island, whereas a single itation of driest month; and BIO16, precipitation of the clade (clade A) was widespread throughout central and wettest quarter of the year. southern Corsican populations. The average genetic Warren & Seifert (2011) recently showed that the divergence (ML corrected) among these five clades default settings in MAXENT have inferior performances ranged between 0.049 (0.009 SD; clades B and A) and compared with tuned settings, particularly the regulari- 0.082 (0.014 SD; clades B and C; see Table S1 in Appendix zation multiplier beta (Warren & Seifert 2011). Therefore, S1, Supporting Information). Within clade A, three subc- as suggested by these authors, we built models with a lades were found, each supported by high bootstrap range of beta values (from 1 to 15) for both the whole set values. Throughout the south-central and southern of 19 variables and the reduced set of six variables. samples, subclade AI was either fixed (samples 1–5) or Finally, to select the model that best fit the data, we used the most abundant, while subclades AII and AIII were the Akaike Information Criterion (AIC) as implemented restricted to the central Corsican populations (samples in the software ENMTOOLS 1.3 (Warren et al. 2010). 6–9). The average genetic divergence (ML corrected) among these subclades ranged between 0.011 (0.003 SD; clades AII–AIII) and 0.015 (0.004 SD; clades AI–AIII). Results Also within clades C and E, two well-differentiated subc- For the 193 individuals analysed for the mtDNA varia- lades were observed, showing genetic divergences of tion, we obtained a fragment of 643 bp from both the 0.013 (0.004 SD) and 0.021 (0.005 SD), respectively. cytb gene and the cox1 gene (overall 1286 bp). In the The phylogenetic networks among haplotypes found combined data set, 211 variable positions were found, at the mtDNA are shown in Fig. 1C. Based on the 95% 172 being parsimony informative. No indels, stop criterion for a parsimonious connection, it was not pos- codons or nonsense codons were observed. A compre- sible to connect all the haplotypes into a single network. hensive number of 97 haplotypes were found. Haplotype Instead, six separate networks were built, correspond- and nucleotide diversity for the whole mtDNA data set ing to the four main clades A, B, C and D, and the two were 0.984 (±0.003 SD) and 0.0275 (±0.0133 SD), respec- subnetworks EI and EII. A single star-like structure was tively. The nuDNA data set included 452 bp of the GH observed, within the subclade AI. The phylogenetic net- gene (45 variable positions, 31 parsimony informative) works among haplotypes found at the nuDNA data sets and 490 bp of the NCX1 gene (15 variable positions, 12 are shown in Fig. 2. In both cases, all haplotypes were parsimony informative) sequenced across 85 and 104 connected into a single network. At the NCX1 fragment, individuals, respectively (see Table 1). Forty-six haplo- a clear phylogeographic structure was not observed. types were identified at the GH fragment and 24 at the Furthermore, the most common haplotype at level of NCX1. No recombination events were indicated by the the whole data set (n = 127) was found spanning the PHI tests carried out on the nuclear gene fragments whole species range. On the other hand, at the GH frag- (both P > 0.05). Haplotype and nucleotide diversity ment, some evidence of geographic association of hapl- were 0.627 (±0.038 SD) and 0.0027 (±0.0019 SD) for otypes was observed. Indeed, most of the individuals NCX1 fragment and 0.911 (±0.014 SD) and 0.0107 sampled among populations located in the central and (±0.058 SD) for the GH fragment, respectively. southern portion of the island (populations 1-9; purple in Fig. 2) carried the haplotype most common at level of the whole data set (n = 41) or haplotypes closely Phylogenetic analyses and molecular dating related to this, whereas all the individuals from the The Bayesian information criterion implemented in northernmost sample (15; orange in Fig. 2) carried a JMODELTEST indicated the TrN+Γ (Tamura & Nei 1993) group of slightly differentiated haplotypes.

Fig. 2 Statistical parsimony networks based on the two nuclear gene fragments analysed (NCX1 and GH). Haplotypes are shown as pie diagrams, with slices proportional to haplotype frequencies at level of the whole data set. Populations are numbered as in Figure 1 and Table 1, and—for comparative purposes—coloured according to the geographic distribution of the main mitochondrial DNA (mtDNA) lineages as presented in Fig. 1 (the northernmost populations carrying the southern mtDNA clade A—populations 8 and 9– were given a distinct col- our tone).

E. platycephalus and H were very close in time to one another, indicat-

EII ing that the divergence between clades C, D, E and E B + A occurred within a very short time lapse, in the EI late Miocene according to calibration II or in the Plio- I D cene according to calibration I. Also of Pliocene origin G CII would have been the divergence between clades A and C B according to both calibrations I and II. Finally, accord- CI ing to both the alternative calibrations, the divergences H B within each of these clades (i.e. AI–AII–AIII, CI–CII and EI–EII) were dated to the Early Pleistocene, whereas F AIII most of the TMRCAs of the terminal haplogroups fell A AII well within the Middle Pleistocene. AI

EP MP LP calibration I Population genetic structure and historical 4.03.0 2.0 1.0 0.0

PL EP MP LP demography calibration II 6.05.0 4.0 3.0 2.0 1.0 0.0 The GLS interpolation analysis showing the geographic Fig. 3 Chronogram based on the ML phylogeny of the haplo- pattern of distribution of mtDNA genetic differentiation types found in Euproctus montanus, estimated using the least among populations is presented in Fig. 4. Two main square procedure of Xia & Yang (2011). Scale axes are in mil- areas are clearly present: one area of extensive homoge- lion years. Above the axes, the beginning of the main historical neity, encompassing the southern and central-western epochs mentioned in the main text is reported: PL, Pliocene; EP, Early Pleistocene; MP, Middle Pleistocene; LP, Late Pleisto- portions of the island, and one area of high differentia- cene. See Table S2 (Supporting Information) for estimated tion in the northern and central-eastern portions. Within values and relative standard deviations. the latter, spotted areas of low differentiation are also apparent. AMOVA analysis was performed by separating popula- Estimates of the TMRCAs for the main clades and tions according to the geographic distribution of the subclades are shown in Fig. 3 and presented in detail in nine terminal mtDNA haplogroups. As haplogroups CI Table S2 (see Appendix S1, Supporting Information). and CII showed an identical distribution, eight popula- As expected, the two alternative calibrations led to sub- tion groups were deﬁned: [1–6], [7, 9], [8], [10, 11], [12], stantially different estimates, although some common [13], [14] and [15]. With this grouping option, the AMOVA patterns were clearly apparent. TMRCAs for nodes I, G analysis indicated that 77.3% of the overall genetic

et al. 2005). Among the selected variables, the one giv- ing the highest percentage contribution to the model (57%) was BIO 14, that is, the precipitation during the driest month. The SDM under the present bioclimatic conditions (Fig. S1A in Appendix S1, Supporting Information) indicated high suitability for most of the island area (as expected on the basis of the known species distribution), with exceptions for mountain tops and coastal areas, especially southern and eastern lowlands. Projections of the model over LGM bioclimatic conditions were performed using both the MIROC and the CCSM databases. Nevertheless, several variables were outside the training range when the CCSM database was used, indicating poor predictive value of this pro- High jection (see Elith et al. 2010). Therefore, only the model built using the MIROC database was considered further Low (Fig. S1B in Appendix S1, Supporting Information). This model suggested that during the LGM, areas of very high suitability for the species were restricted to the Fig. 4 Genetic landscape shape showing genetic differentiation southern and eastern portions of the island. Southern among populations, based on an inverse-distance-weighted and eastern lowlands appeared more suitable for the interpolation of the estimated Nei’s (1987) net between-popula- species than under present-day conditions, while mod- tion genetic divergence (TrN+Γ corrected). Warmer colours erate-to-high suitability was still present in the northern represent higher levels of genetic differentiation. lowlands. The least suitable areas were those located at higher altitudes, throughout the central mountain chain. variation can be attributed to the among-group level of On the whole, although bioclimatic suitability appeared = variation (FCT 0.77), 4.5% to the among-population reduced during the LGM with respect to present condi- = within-group level (FSC 0.20) and 18.2% to the within- tions in the northern and western portions of the island, = population level (FST 0.82), all variance components in the southern and the eastern portions, wider areas of and fixation indices being highly statistically significant high suitability seem to have been present for the spe- (P < 0.001). cies during the LGM than at present. Mismatch distribution analyses and values of the demographic test statistics for the main terminal Discussion mtDNA haplogroups are presented in Fig. 5. We limited the analyses to those haplogroups having a The continuous geographic distribution of Euproctus sample size of no <10 individuals. The only haplogroups montanus populations throughout its restricted range for which all analyses converged in indicating a recent would have plausibly predicted a phylogeographic demographic expansion were AI, CII and D, whereas pattern without substantial discontinuities, suggestive for haplogroup B, all analyses rejected this scenario. of a single large population (e.g. IUCN 2011). Instead, Finally, a demographic growth was suggested for haplo- we found evidence of a geographic mosaic of deeply groups AII and AIII just by the test statistics R2 and FS, divergent and ancient evolutionary lineages, a pattern respectively. that also has major implications for the species’ conservation. In the subsequent sections, we will first discuss these issues, and then we will evaluate our data in the Spatial distribution modelling context of the intra-island patterns of genetic diversity, The model comparison conducted in ENMTOOLS indicated to highlight some key issues arising. that our selection of six bioclimatic variables fit the data better than the entire set of 19 variables and that the Evolutionary history of Euproctus montanus model receiving the lowest AIC was the one built using the value 3 for the regularization multiplier beta param- Five main mtDNA lineages (clades A, B, C, D and E) eter. This model yielded AUC scores of 0.73 and 0.67 were clearly apparent by phylogenetic analyses. Their for the training and the test data, respectively, indicat- genetic divergence (shown in Table S1, Supporting ing an ‘acceptable performance’ of the model (Arau´ jo Information) is conspicuous, largely exceeding the one

0.3 (Ai) (Aiii) 0.4 (Cii)

0.2 0.1 0.3

0.2 2 2 R2 R * = 0.0442 0.1 R = 0.0898 * = 0.0735 F * = –21.159 F * = –5.173 F * = –5.544 s s 0.1 s SSD = 0.0011 SSD* = 0.0391 SSD = 0.0371 tau = 4.74 tau = 4.75 tau = 1.81 0.0 0.0 0.0 01020 30 0 10 20 30 0 10 20 30

0.4 (Aii) 0.2 (B) (D) 0.3 0.2

0.2 0.1 2 R2* = 0.1024 R = 0.1248 0.1 R2* = 0.0874 F F * = –1.454 F 0.1 s* = –2.288 s s* = –3.486 SSD = 0.0870 SSD* = 0.0634 SSD = 0.0178 tau = 4.158 tau = 10.73 tau = 3.64 0.0 0.0 0.0 01020 30 0 10 20 30 0 10 20 30

Fig. 5 Mismatch distributions and values of the demographic test statistics for six terminal haplogroups identified with phylogenetic > analyses. Only haplogroups having sample size 9 were considered. r: raggedness statistic (Harpending 1994). FS: Fu’s FS statistic * < (Fu 1997). R2: Ramos-Onsins & Rozas’ (2002) R2 statistic. P 0.05. Dotted line: observed mismatch distribution; continuous line: mismatch distribution expected under a pure demographic expansion model. observed among several species pairs of European the shrubby plant Cystus creticus showed main phylog- newts, such as Calotriton asper–Calotriton arnoldii eographic breaks among divergent lineages in northern (Carranza & Amat 2005), Triturus marmoratus–Triturus Corsica (Falchi et al. 2009; Gentile et al. 2010; Ketmaier pygmaeus (Carranza & Amat 2005) and Triturus carnifex– et al. 2010). Triturus macedonicus (Arntzen et al. 2007). Furthermore, The overall amount of diversity, the pattern of diver- the geographic scale at which these divergences were gence and the phylogeographic structure found at the observed is surprisingly small. Four lineages were mtDNA are not paralleled by the nuDNA data. Indeed, restricted to one or two populations each, all located in no phylogeographic structure was apparent at the the northern portion of the Corsica Island, whereas the NCX1 fragment, while the GH fragment showed no fifth lineage (A) was the only one extending its distribu- clear phylogenetic structure, but some evidence of geo- tion also to the central and southern portions of graphic association among scarcely differentiated haplo- the island. Furthermore, even within lineage A, two types. Discordance among mitochondrial and nuclear subclades were geographically restricted to northern sequence data is increasingly emerging, as more case Corsica, whereas only subclade AI was found in the studies become available using nuclear sequences data southern portion of the island. Thus, almost the entire to investigate patterns of variation among divergent genetic differentiation observed within E. montanus was populations or closely related species (e.g. Monsen & found restricted to northern Corsica (see Fig. 4). Blouin 2003; Gonçalves et al. 2007; Pinho et al. 2007; Irrespective of the calibration used to date the diver- Brunes et al. 2010; Rato et al. 2010; Guo et al. 2011; gence events, as shown in Fig. 3 and Table S2 in Prado et al. 2012). A plausible explanation for these Appendix S1 (Supporting Information), the three basal discordances is the generally slower evolutionary rates, splits (leading to lineages A + B, C, D and E) are prob- coalescence and lineage sorting of nuclear than mito- ably to have occurred at the same time or at least chondrial genes, leading the former to fail diverging within a very short time lapse, between 3.6 and 5.8 My and differentiating (Zhang & Hewitt 2003; Brito & depending on the calibration used, whereas the split Edwards 2009). Alternative explanations could be a between lineages A and B probably occurred some- stronger female phylopatry (Prugnolle & Meeus 2002) what later (2.6–4.1 My). Interestingly, an estimate of or stochastic (rather than historical) processes acting on the divergence time very close to those we found the mtDNA and generating spurious phylogeographic among the four basal lineages of E. montanus was signals at this marker (Irwin 2002; Kuo & Avise 2005). obtained for two lineages of the lizard Archeolacerta Nevertheless, we consider both these alternatives much bedriagae, which are parapatric in northern Corsica less plausible than the former in the present case. Two (3.7–5.9 My; Salvi et al. 2010), and also the land snail issues make a strong female phylopatry an improbable Solatopupa guidoni, the isopod Helleria brevicornis and explanation of the observed pattern. First, we should

© 2012 Blackwell Publishing Ltd 152 R. BISCONTI ET AL. assume a substantial lack of effective female dispersal divergences is considered (Edwards & Beerli 2000; Hein for not <3 My, which does not fit with the current con- et al. 2005). tinuous distribution of breeding sites in the area (imply- A close link between moisture regimes and the demo- ing a continuous distribution of females). Second, we graphic history of E. montanus would also comfortably found no evidence in the literature available to date explain the time estimates of subsequent splits and supporting such a pronounced female phylopatry in TMRCAs within each main clade to the Early and E. montanus or closely related species. On the other Middle Pleistocene, respectively, when cycles of humid- hand, migrant females with different mtDNA are less dry climate would have followed the initiation of the probable of transmitting the mtDNA copy to the recep- glacial–interglacial oscillations (see Thompson 2005 for tor populations because of drift on a small effective an account of climatic oscillations and the consequent population size when the receptor populations are vegetational changes). The SDM indicated that habitat already at carrying capacity. The chances that under the suitability decreased in northern but not in southern same scenario nuDNA from migrants become part of Corsica during phases of increased drought, as were the gene pool are higher. Nevertheless, this process the Pleistocene glacial phases, and recent data from the should have been extremely pervasive among several study by Kuhlemann et al. (2008) suggest that while pairs of populations and on the same large timescale, northern Corsica was substantially invaded by polar air which makes it improbable as well. Two main lines of during the LGM, milder climatic conditions were pres- evidence converge in indicating that the observed pat- ent more to the south. Therefore, genetic data, SDM tern is not spurious. First, stochastic processes are not analysis and paleoclimatic reconstructions concordantly expected to yield a pattern of multiple divergences that support the possible occurrence of a wider refugium to are both spatially and temporally clustered, as we the south than to the north. On the other hand, the found in northern Corsica and for the TMRCAs of the long-term persistence in northern Corsica of E. mont- main lineages and of several subclades (Avise 2008). anus populations is testified just by the occurrence in Second, stochastic processes are not expected to yield this area of most of the lineages found. Thus, it can be patterns of concordance (either spatial or temporal) argued that during these phases, E. montanus popula- among multiple codistributed taxa, as is the case here tions did not disappear from this area, but underwent (see Gentile et al. 2010; Ketmaier et al. 2010; Salvi et al. fragmentations and demographic contractions into 2010). Therefore, a failure of nuDNA sequence markers small areas of more suitable microclimatic conditions. to differentiate and diverge appears the most plausible As E. montanus is an essentially montane species living explanation for the observed pattern, although the use of close to streams and brooks, and considering the topo- more and more variable loci (e.g. microsatellites) will be graphic features of Corsica, we speculate that such needed to carry out an in-depth exploration of this issue. ‘microrefugia’ (see Rull 2009; Mosblech et al. 2011 and As pointed out in previous studies (Salvi et al. 2010), references therein) could have been provided by some there are no obvious geographic barriers to dispersal in mountain river valleys, which are so abundant on the northern Corsica that could be indicated as plausible island, have been already reported for several continen- causal factors for the observed divergences, and there tal species and are more likely to present significant are no evidence that they could have been active in the decoupling between local and regional climate trends past. Nevertheless, according to our modelling of the (Dobrowski 2011). Finally, according to the SDMs, bioclimatic niche, the precipitation of the driest month during the LGM, suitable areas in southern Corsica would account for most of the pattern of distribution of were even wider than at present, suggesting that in this E. montanus (57%). Interestingly, as suggested by Suc area, E. montanus populations could have expanded (1984) on the basis of palynologic and macroflora rather than contracted during the LGM, as already analyses, during the Pliocene and Early Pleistocene, the suggested for several montane species (Hewitt 2011a). paleoenvironmental evolution of north-western Mediter- In this respect, it is also worth nothing that among ranean was characterized by a progressive decrease in the subclades showing concordant signs of recent moisture, with two main events: the installation of sum- demographic expansion (clades AI, CII and D) across mer dryness, favouring forest clearing, approximately multiple tests (FS, R2, mismatch distribution analysis, 3.2 My ago, and the first dry period with development see Fig. 5), the higher value of the parameter τ (i.e. the of stepping associations, at approximately 2.3 My. time since the expansion in generations, scaled by the These estimates would fit particularly well with our mutation rate) was obtained for clade AI (and very divergence estimates for the four main lineages and for similar values were obtained for AII and AIII, also A vs. B (see calibration I in Table S2 in Appendix S1, showing some evidence of recent expansion), suggesting Supporting Information), particularly when the that its expansion phase could in fact have pre-dated expected time lag between genetic and population those of the northern subclades.

tionary processes (see for example Grant 1998; Whittaker Implications for conservation & Fernańdez-Palacios 2007). Furthermore, this predic- The overall pattern of genetic variation that we have tion has also been commonly used in assessments of found in E. montanus has at least two main implications biodiversity conservation priorities (see for instance the for the species’ conservation. First, northern Corsica has other two Corsican endemic amphibians, Salamandra emerged as the core area for its conservation, as clearly corsica and Discoglossus montalenti; IUCN 2011). As indicated by the occurrence in this area of most of the discussed above, the pattern of genetic diversity found variation found in this species and of all the most dee- among Euproctus montanus populations do not fit this ply divergent lineages. Second, and most importantly, prediction at all. This pattern appears somewhat E. montanus can no longer be considered as if it was extreme—particularly for a vertebrate species—and formed by a single population ranging the whole island unlikely to be a general rule; however, according to the (IUCN 2011), as our data show that it is strongly frag- limited but growing literature about intra-island mented into several reciprocally monophyletic lineages phylogeographic variation of island endemics, similar of ancient origin. patterns of strong intra-island differentiations and Surprisingly, the only evidence of syntopy among ancient lineages are increasingly emerging in many main clades was observed within sample 14 for clades world regions for a wide range of taxonomic groups EI and B. As discussed above, a strong female phylopa- (see for example Thorpe & Malhotra 1998; Holland & try seems rather improbable explanation for the Hadfield 2002; Emerson et al. 2006; Villacorta et al. 2008; observed pattern. A more plausible explanation could Wallis & Trewick 2009; Bisconti et al. 2011). Furthermore, be some form of intrinsic limitation to gene exchange these patterns parallel recent findings concerning other among distinct lineages, affecting either the sole mtDNA areas of long-term species persistence, as are glacial or the whole genome. However, the exact nature of this refugia on mainlands (see for example Canestrelli et al. possible limitation will need to be investigated and can- 2006, 2010; Go´mez & Lunt 2007). On the whole, in the not be further speculated here. Indeed, two features of light of these findings, an expectation of panmixia in our data do not allow solving this issue: first, in the species with a restricted range, especially island endem- light of the continuous geographic distribution, contact ics, could in fact be unwarranted or at least far from populations among main lineages should exist, and being of universal value. Consequently, the extensive use thus, the study of their genetic structure will be manda- of several simplifying assumptions usually made on tory; second, being haploid, the mtDNA is inappropriate island populations, including range-wide panmixia, an to reveal the occurrence and extent of admixture among effective population size correlated with island size, and lineages, while our nuDNA data did not provided suffi- long-term stability, deserve to be reconsidered. cient resolution, and thus, the use of highly variable nuclear markers will be mandatory as well. Nonetheless, Acknowledgements even if future studies based on nuclear markers would show a more admixed pattern of population genetic We are grateful to Alessandra Perilli for her help with our laboratory work, Godfrey Hewitt and Daniele Porretta for their structure, as pointed out by Avise (2008), the pattern kind discussions and suggestions during the preparation of shown by the maternally inherited mtDNA would in this manuscript. Newts were captured under the permit from any case be evidence of demographic independence of the Direction re´gionale de l’environnement, de l’ameńagement each evolutionary lineage, which would thus deserve et du logement (DREAL) de Corse to DS. DS was supported specific conservation efforts (Moritz 1994). by FCT postdoctoral grants SFRH/BPD/66592/2009.

References Perspectives and conclusions Arau´ jo MB, Pearson RG, Thuiller W, Erhard M (2005) Valida- Species or populations with a continuous geographic tion of species–climate impact models under climate change. distribution across restricted ranges have long been Global Change Biology, 11, 1504–1513. thought as single homogeneous entities, a prediction Arau´ jo MB, Thuiller W, Pearson RG (2006) Climate warming particularly pervasive in studies involving island en- and the decline of amphibians and reptiles in Europe. Journal demics (e.g. Frankham 1996a,b). Notably, it is a neces- of Biogeography, 33, 1712–1728. sary premise to consider island size as a correlate of Arntzen JW, Themudo GE, Wielstra B (2007) The phylogeny of effective population size, for instance in comparisons of crested newts (Triturus cristatus superspecies): nuclear and mitochondrial genetic characters suggest a hard polytomy, in diversity patterns among islands of different sizes and line with the paleogeography of the centre of origin. Contri- with the continent (Frankham 1996a,b; Barton 1998; butions to Zoology, 76, 261–277. Woolﬁt & Bromham 2005). Such comparisons have been Avise JC (2008) Phylogeography: retrospect and prospect. Jour- extensively used in the study of ecological and evolu- nal of Biogeography, 36,3–15.

Barton NH (1998) Natural selection and random genetic drift Excoffier L (2004) Patterns of DNA sequence diversity and as causes of evolution on islands. In: Evolution on Islands (ed. genetic structure after a range expansion: lessons from the Grant PR), pp. 102–123. Oxford University Press, New York, infinite-island model. Molecular Ecology, 13, 853–864. NY. Excoffier L, Laval G, Schneider S (2005) ARLEQUIN (version Bisconti R, Canestrelli D, Nascetti G (2011) Multiple lines of 3.0): An integrated software package for population genetic evidence for demographic and range expansion of a temper- data analysis. Evolutionary Bioinformatics Online, 1,47–50. ate species (Hyla sarda) during the last glaciation. Molecular Falchi A, Paolini J, Desjobert JM, Melis A, Costa J, Varesi L Ecology, 20, 5313–5327. (2009) Phylogeography of Cistus creticus L. on Corsica and Blondel J, Aronson J, Bodiou JY, Boeuf G (2010) The Mediterra- Sardinia inferred by the TRNL-F and RPL32-TRNL nean Region: Biological Diversity in Space and Time, 2nd edn. sequences of cpDNA. Molecular Phylogenetics and Evolution, Oxford University Press Inc., New York. 52, 538–543. Brito PH, Edwards SV (2009) Multilocus phylogeography and Farris JS, Kallersjo M, Kluge AG, Bult C (1994) Testing signifi- phylogenetics using sequence-based markers. Genetica, 135, cance of incongruence. Cladistics, 10, 315–319. 439–455. Feliner GN (2011) Southern European glacial refugia: a tale of Bruen T, Phillipe H, Bryant D (2006) A quick and robust statis- tales. Taxon, 60, 365–372. tical test to detect the presence of recombination. Genetics, Frankham R (1996a) Do island populations have less genetic 172, 2665–2681. variation than mainland populations?. Heredity, 78, 311–327. Brunes TO, Sequeira F, Haddad CFB, Alexandrino J (2010) Frankham R (1996b) Relationship of genetic variation to popu- Gene and species trees of a Neotropical group of treefrogs: lation size in wildlife. Conservation Biology, 66, 1500–1508. genetic diversification in the Brazilian Atlantic forest and Fu YX (1997) Statistical tests of neutrality of mutations against theorigin of a polyploid species. Molecular Phylogenetics and population growth, hitchhiking and background selection. Evolution, 57, 1120–1133. Genetics, 147, 915–925. Caccone A, Milinkovitch MC, Sbordoni V, Powell JR (1994) Gasc JP, Cabela A, Crnobrnja-Isailovic J et al. (1997) Atlas of Molecular biogeography: Using the Corsica-Sardinia micro- Amphibians and Reptiles in Europe. Collection Patrimoines Nat- plate disjunction to calibrate mitochondrial rDNA evolution- urels, 29, Societas Europaea Herpetologica, Museúm ary rates in mountain newts (Euproctus). Journal of National d’Histoire Naturelle & Service du Petrimone Natu- Evolutionary Biology, 7, 227–245. rel, Paris, France, 496 pp. Canestrelli D, Cimmaruta R, Costantini V, Nascetti G (2006) Gentile G, Campanaro A, Carosi M, Sbordoni V, Argano R Genetic diversity and phylogeography of the Apennine yel- (2010) Phylogeography of Helleria brevicornis Ebner 1868 low-bellied toad Bombina pachypus, with implications for con- (Crustacea, Oniscidea): Old and recent differentiations of an servation. Molecular Ecology, 15, 3741–3754. ancient lineage. Molecular Phylogenetics and Evolution, 54, 640– Canestrelli D, Aloise G, Cecchetti S, Nascetti G (2010) Birth of 646. a hotspot of intraspecific genetic diversity: notes from the Go´mez A, Lunt DH (2007) Refugia within refugia: patterns of underground. Molecular Ecology, 19, 5432–5451. phylogeographic concordance in the Iberian Peninsula. In: Carranza S, Amat F (2005) Taxonomy, biogeography and evo- Phylogeography in Southern European Refugia: Evolutionary Per- lution of Euproctus (Amphibia: Salamandridae), with the spectives on the Origins and Conservation of European Biodiver- resurrection of the genus Calotriton and the description of a sity (eds Weiss S, Ferrand N), pp. 155–188. Springer Verlag, new endemic species from the Iberian Penisula. Zoological Dordrecht, the Netherlands. Journal of the Linnean Society, 145, 555–582. Gonçalves H, Martıńez-Solano I, Ferrand N, Garcıá-Parı´sM Clement M, Posada D, Crandall KA (2000) TCS: a computer (2007) Conflicting phylogenetic signal of nuclear vs mitochon- program to estimate gene genealogies. Molecular Ecology, 9, drial DNA markers in midwife toads (Anura, Discoglossidae, 1657–1660. Alytes): deep coalescence or ancestral hybridization? Molecular Cronk QCB (1997) Islands: stability, diversity, conservation. Phylogenetics and Evolution, 44, 494–500. Biodiversity and Conservation, 6, 477–493. Grant PR (1998) Evolution on Islands. Oxford University Press, Darwin C (1859) On the Origin of Species. John Murray, New York, NY. London. Grant PR, Grant BR (2007) How and Why Species Multiply: The Dobrowski SZ (2011) A climatic basis for microrefugia: the Radiation of Darwin’s Finches. Princeton University Press, influence of terrain on climate. Global Change Biology, 17, Princeton, New Jersey. 1022–1035. Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Doyle JJ, Doyle JL (1987) A rapid DNA isolation procedure for Gascuel O (2010) New algorithms and methods to estimate small quantities of fresh leaf tissue. Phytochemistry Bulletin, maximum-likelihood phylogenies: assessing the performance 19,11–15. of PhyML 3.0. Systematic Biology, 59, 307–321. Edwards SV, Beerli P (2000) Perspective: gene divergence, pop- Guo P, Liu Q, Li C et al. (2011) Molecular phylogeography of ulation divergence, and the variance in coalescence time in Jerdon’spitviper (Protobothrops jerdonii): importance of phylogeographic studies. Evolution, 54, 1839–1854. the uplift of the Tibetan plateau. Journal of Biogeogrphy, 38, Elith J, Kearney M, Steven P (2010) The art of modelling range- 2326–2336. shifting species. Methods in Ecology and Evolution, 1, 330–342. Hampe A, Petit RJ (2005) Conserving biodiversity under climate Emerson BC, Forgie SU, Goodacre S, Oromı` P (2006) Testing change: the rear edge matters. Ecology Letters, 8,461–467. phylogeographic predictions on an active volcanic island: Harpending HC (1994) Signature of ancient population growth in Brachyderes rugatus (Coleoptera: Curculionidae) on La Palma a low-resolution mitochondrial DNA mismatch distribution. (Canary Islands). Molecular Ecology, 15, 449–458. Human Biology, 66,591–600.

Hein J, Schierup M, Wiuf C (2005) Gene Genealogies, Variation Monsen KJ, Blouin MS (2003) Genetic structure in a montaner- and Evolution: A Primer in Coalescent Theory. Oxford Univer- anid frog: restricted gene flow and nuclear-mitochondrial sity Press, Oxford, UK. discordance. Molecular Ecology, 12, 3275–3286. Hewitt GM (1996) Some genetic consequences of ice ages and Moritz C (1994) Defining “Evolutionarily Significant Units” for their role in divergence and speciation. Biological Journal of conservation. Trends in Ecology and Evolution, 9, 373–375. the Linnean Society, 58, 247–276. Mosblech NAS, Bush MB, Van Woesik R (2011) On metapopu- Hewitt GM (2000) The genetic legacy of the Quaternary ice lations and microrefugia: palaeoecological insights. Journal of ages. Nature, 405, 907–913. Biogeography, 38, 419–429. Hewitt GM (2004) Genetic consequences of climatic oscillations Mouillot F, Paradis G, Andrei-Ruiz MC, Quilichini A (2008) in the Quaternary. Philosophical Transactions of the Royal Soci- Corsica. In: Mediterranean Island Landscapes: Natural and Cul- ety of London. Series B, Biological Sciences, 359, 183–195. tural Approaches (eds Vogiatzakis IN, Pungetti G, Mannion Hewitt GM (2011a) Quaternary phylogeography: the roots to AM), pp. 220–244. Springer Publishing, New York, NY. hybrid zones. Genetica, 139, 617–638. Nei M (1987) Molecular Evolutionary Genetics. Columbia Univer- Hewitt GM (2011b) Mediterranean Peninsulas-the evolution of sity Press, New York, NY. hotspots. In: Biodiversity Hotspots (eds Zachos FE, Habel JC), Nieberding C, Morand S, Libois R, Michaux JR (2006) Parasites pp. 123–147. Springer, Amsterdam, the Netherlands. and the island syndrome: the colonization of the western Hijmans RJ, Guarino L, Cruz M, Rojas E (2001) Computer tools Mediterranean islands by Heligmosomoides polygyrus for spatial analysis of plant genetic resources data: 1. diva- (Dujardin, 1845). Journal of Biogeography, 33, 1212–1222. gis. Plant Genetic Resources Newsletter, 127,15–19. Papadopoulou A, Anastasiou I, Keskin B, Vogler AP (2009) Holland BS, Hadfield MG (2002) Islands within an island: phyloge- Comparative phylogeography of tenebrionid beetles in the ography and conservation genetics of the endangered Hawaiian Aegean archipelago: the effect of dispersal ability and habitat tree snail Achatinella mustelina. Molecular Ecology, 11,365–376. preference. Molecular Ecology, 18, 2503–2517. Huson DH, Bryant D (2006) Application of phylogenetic net- Phillips SJ, Anderson RP, Schapire RE (2006) Maximum works in evolutionary studies. Molecular Biology and Evolu- entropy modelling of species’ geographic distributions. Eco- tion, 23, 254–267. logical Modelling, 190, 231–259. Irwin DE (2002) Phylogeographic breaks without geographic Pinho C, Harris JD, Ferrand N (2007) Comparing patterns of barriers to gene flow. Evolution, 56, 2383–2394. nuclear and mitochondrial divergence in a cryptic species IUCN (2011) IUCN Red List of Threatened Species. Version 2011.2. complex: the case of Iberian and North Africanwall lizards IUCN, http://www.iucnredlist.org. (Podarcis, Lacertidae). Biological Journal of the Linnean Society, Ketmaier V, Manganelli G, Tiedemann R, Giusti F (2010) Peri- 91, 121–133. Tyrrhenian phylogeography in the land snail Solatopupa gui- Posada D (2008) jModelTest: Phylogenetic model averaging. doni (Pulmonata). Malacologia, 52,81–96. Molecular Biology and Evolution, 25, 1253–1256. Kier G, Kreft H, Lee TM et al. (2009) A global assessment of ende- Prado CPA, Haddad CFB, Zamudio KR (2012) Cryptic lineages mism and species richness across island and mainland regions. and Pleistocene population expansion in a Brazilian Cerrado Proceedings of the National Academy of Sciences, 106,9322–9327. frog. Molecular Ecology, 21, 921–941. Kuhlemann J, Frisch W, Sze´kely B, Dunkl I, Ka´zme´r M (2005) Prugnolle F, Meeus T (2002) Inferring sex-biased dispersal Wu¨rmian maximum glaciation in Corsica. Austrian Journal of from population genetic tools: a review. Heredity, 88, Earth Sciences, 97,68–81. 161–165. Kuhlemann J, Rohling EJ, Krumrei I, Kubik P, Ivy-Ochs S, Quantum GIS Development Team (2012) Quantum GIS Geo- Kucera M (2008) Regional synthesis of Mediterranean atmo- graphic Information System. Open Source Geospatial Foundation spheric circulation during the Last Glacial Maximum. Science, Project. Available from http://qgis.osgeo.org. 321, 1338–1340. Ramos-Onsins SE, Rozas J (2002) Statistical properties of new Kuo CH, Avise JC (2005) Phylogeographic breaks in low-dis- neutrality tests against population growth. Molecular Biology persal species: the emergence of concordance across gene and Evolution, 19, 2092–2100. trees. Genetica, 124, 179–186. Rato C, Carranza S, Perera A, Carretero M, Harris DJ (2010) La´zaro EM, Harrath AH, Stocchino GA, Pala M, Bagun˜ J, RRiu- Conflicting patterns of nucleotide diversity between mtDNA tort M (2011) Schmidtea mediterranea phylogeography: an and nDNA in the Moorish gecko, Tarentola mauritanica. old species surviving on a few Mediterranean islands?. BMC Molecular Phylogenetics and Evolution, 56, 962–971. Evolutionary Biology, 11, 274. Ro¨dder D, Schlu¨ ter A, Lo¨tters S (2010) Is the ‘lost world’ lost? Librado P, Rozas J (2009) DnaSP v5: A software for compre- High endemism of amphibian and reptiles on South hensive analysis of DNA polymorphism data. Bioinformatics, American Tepuı´s in a changing climate. In: Relict Species (eds 25, 1451–1452. Habel JC, Assmann T), pp 401–416. Springer-Verlag, Berlin. MacArthur RH, Wilson EO (1967) The Theory of Island Biogeog- Rogers AR, Harpending HC (1992) Population growth makes raphy. Princeton University Press, Princeton, New Jersey, waves in the distribution of pairwise. Molecular Biology and USA, 203 pp. Evolution, 9, 552–569. Me´dail F, Diadema K (2009) Glacial refugia influence plant Rull V (2009) Microrefugia. Journal of Biogeography, 36, 481–484. diversity patterns in the Mediterranean basin. Journal of Salvi D, James Harris D, Bombi P, Carretero MA, Bologna MA Biogeography, 36, 1333–1345. (2010) Mitochondrial phylogeography of the Bedriaga’s rock Miller MP (2005) Alleles In Space (AIS): Computer software for lizard, Archaeolacerta bedriagae (Reptilia: Lacertidae) endemic the joint analysis of interindividual spatial and genetic infor- to Corsica and Sardinia. Molecular Phylogenetics and Evolution, mation. Journal of Heredity, 96, 722–724. 56, 690–697.

Schmitt T (2007) Molecular biogeography of Europe: Pleisto- Wallis GP, Trewick SA (2009) New Zealand phylogeogra- cene cycles and postglacial trends. Frontiers in Zoology, 4, 11. phy: evolution on a small continent. Molecular Ecology, 18, Stock M, Dubey S, Klutsch C, Litvinchuk SN, Scheidt U, Perrin 3548–3580. N (2008) Mitochondrial and nuclear phylogeny of circum- Warren DL, Seifert SN (2011) Environmental niche modeling in Mediterranean tree frogs from the Hyla arborea group. Molec- Maxent: the importance of model complexity and the perfor- ular Phylogenetics and Evolution, 49, 1019–1024. mance of model selection criteria. Ecological Applications, 21, Stroscio S, Baviera C, Frati F, Lo Paro G, Nardi F (2011) Coloniza- 335–342. tion of the Aeolian Islands by Pimelia rugulosa rugulosa Ger- Warren DL, Glor RE, Turelli M (2010) ENMTools: a toolbox for mar, 1824 (Coleoptera: Tenebrionidae) inferred from the comparative studies of environmental niche models. Ecogra- genetic structure of populations: geological and environmental phy, 33, 607–611. relations. Biological Journal of the Linnean Society, 104,29–37. Whittaker RJ, Fernańdez-Palacios JM (2007) Island Biogeography: Suc JP (1984) Origin and evolution of the Mediterranean vege- Ecology, Evolution, and Conservation. Oxford University Press, tation and climate in Europe. Nature, 307, 429–432. Oxford, UK. Swofford DL (2003) PAUP*. Phylogenetic Analysis Using Parsi- Woolfit M, Bromham L (2005) Population size and molecular mony (*and Other Methods). Version 4. Sinauer Associates, evolution on islands. Proceedings of the Royal Society of London Sunderland, Massachusetts. Biological Sciences, 272, 2277–2282. Taberlet P, Fumagalli L, Wust-Saucy A-G, Cosson JF (1998) Xia X, Xie Z (2001) DAMBE: software package for data analysis Comparative phylogeography and postglacial colonization in molecular biology and evolution. Journal of Heredity, 92, routes in Europe. Molecular Ecology, 7, 453–464. 371–373. Tamura K, Nei M (1993) Estimation of the number of nucleo- Xia X, Yang Q (2011) A distance-based least-square method for tide substitutions in the control region of mitochondrial dating speciation events. Molecular Phylogenetics and Evolu- DNA in humans and chimpanzees. Molecular Biology and tion, 59, 342–353. Evolution, 10, 512–526. Zhang D, Hewitt GM (2003) Nuclear DNA analyses in genetic Templeton AR, Crandall KA, Sing CF (1992) A cladistic analy- studies of populations: practice, problems and prospects. sis of phenotypic associations with haplotypes inferred from Molecular Ecology, 12, 563–584. restriction endonuclease mapping and DNA sequence data. III. Cladogram estimation. Genetics, 132, 619–633. Thompson JD (2005) Plant Evolution in the Mediterranean. R.B. is a postdoctoral researcher, interested in evolution, eco- Oxford University Press, New York, NY. logy and conservation of insular species. This study is a part Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins of her PhD programme. D.C. is an assistant professor inter- DG (1997) The ClustalX windows interface: flexible strategies ested in the evolutionary and ecological processes underlying for multiple sequence alignment aided by quality analysis the geographic patterns of biodiversity, and in conservation tools. Nucleic Acids Research, 25, 4876–4882. genetics. D.S. is a postdoctoral researcher at CIBIO, working on Thorpe RS, Malhotra A (1998) Molecular and morphological biogeographic and evolutionary patterns in endemic reptiles evolution within small islands. In: Evolution on Islands (ed. and amphibians from Corsica and Sardinia and other Mediter- Grant PR), pp. 67–82. Oxford University Press, New York, ranean islands. G.N. is a full professor of Ecology at the Tuscia NY. University. His interests encompass co-evolution, speciation, Tamura K, Peterson D, Peterson N, Stecher G, Nei M, Kumar S phylogeography and conservation genetics. (2011) MEGA5: Molecular Evolutionary Genetics Analysis using Maximum Likelihood, Evolutionary Distance, and Maximum Parsimony Methods. Molecular Biology and Evolu- tion, 28, 2731–2739. Veith M, Mayer C, Samraoui B, Barroso DD, Bogaerts S (2004) Data accessibility From Europe to Africa and vice versa: evidence for multiple intercontinental dispersal in ribbed salamanders (genus Pleu- DNA sequences: GenBank accessions for each sample rodeles). Journal of Biogeography, 31, 159–171. examined are given in the online Supporting informa- Vellend M, Orrock JL (2010) Ecological and genetic models of tion. Data points used for the SDM analysis are given diversity. In: The Theory of Island Biogeography Revisited (eds in the online Supporting information. Losos JB, Ricklefs R), pp. 439–461. Princeton University Press, Princeton, New Jersey and Oxford. Villacorta C, Jaume D, Oromı` P, Juan C (2008) Under the vol- Supporting information cano: phylogeography and evolution of the cave-dwelling Palmorchestia hypogaea (Amphipoda, Crustacea) at La Palma Additional supporting information may be found in the online ver- (Canary Islands). BMC Biology, 6, 7. doi:10.1186/1741-7007-6-7. sion of this article. Vogiatzakis IN, Pungetti G, Mannion AM(2008) Mediterranean Appendix S1 Supplementary Tables and Figures. Island Landscapes: Natural and Cultural Approaches. Springer Publishing, New York, NY. Appendix S2 GenBank accession numbers for each sample and Wallace AR (1880) Island Life, or the Phenomena and Causes of gene fragment examined. Insular Faunas and Floras, Including a Revision and Attempted Solution of the Problem of Geological Climates. Macmillan & Appendix S3 Data points used for the SDM analysis. Co., London & New York.