Since Busack 12, the faunal relationships between North Africa and Iberia have been the focus of numerous molecular phylogeographic studies, yet little attention has been drawn to the consequences that the last landbridge between Africa and Sicily may have had on Mediterranean biogeography. As was the Strait of Gibraltar, the Strait of Sicily (no universally accepted name in English; other names include Sicilian Strait, Sicilian Channel, Channel of Sicily, Pantelleria Channel) is purported to have formed at the end of the Messinian salinity crisis (5.3 Mya), at the Miocene/Pliocene boundary [e.g. 3456, F. Rögl pers. comm.]. The Messinian 4 was a geological period from 5.59 to 5.33 Mya during which the Mediterranean Sea was isolated from the Atlantic Ocean, resulting in a large decrease in the Mediterranean Sea level and the formation of landbridges between Africa, Europe and most Mediterranean islands. This included a well-documented landbridge between Africa and a landmass that later became part of Sicily, which may be the last terrestrial connection between the African mainland and the island. However, while it is known that Sicily and Tunisia are at present approximately 140 km apart, low Pleistocene sea levels of about -120 m 789 would have repeatedly drawn the north African paleo-coast and the Sicilian landmass closer than ~50 km (Figure 1). In addition, current shoals 1011 may be remnants of Pleistocene "stepping stone islands" that may have facilitated terrestrial animals, including humans 12, in overcoming the sea barrier. Therefore, the phylogenetic depths of sister relationships between Sicily and Africa may vary as they do for trans-Gibraltarian relationships 213. Observations of this pattern in widespread taxa might be explained by multiple invasions across the Strait of Sicily and/or additional potential invasions from the region that now forms the Italian (Apennine) Peninsula.

As expected for an island with a long Pliocene isolation 14, endemic (island specific) forms in several animal groups on Sicily have been revealed by molecular analyses (e.g.: hedgehogs, Erinaceus 15; shrews, Crocidura 16; mice, Apodemus 17; beetles, Pachydemidae 18; terrapins, Emys 19; skinks, Chalcides 20). Using these methods, some of these species have been shown to be sister taxa to populations from the Apennine Peninsula (e.g. Apodemus, Erinaceus) or to be of very recent Apennine peninsular (or Calabrian) origin (reviewed in 21). This pattern may be explained by "an intermittent filter barrier in the area of the Strait of Messina [which] probably controlled the processes and timing of the Late Middle Pleistocene-Late Glacial vertebrate faunal dispersal in Sicily" [Bonfiglio in: 22]. Many Pleistocene fossil megafaunal elements entered Sicily from peninsular Italy [Bonfiglio in: 22]. However, Bonfiglio 14 also hypothesized a Lower Pleistocene African origin of fossil elephants (Elephas falconeri) as a debatable alternative to derivation from European/Italian E. antiquus [Bonfiglio in: 22]. Other paleontologists have also discussed whether Sicily had early Pleistocene connections to Africa, as in an, in this respect, unresolved investigation of a Pleistocene amphisbaenid lizard from Sicily 23. All these authors 222324, however, refer to the lack of conclusive evidence for a Pleistocene landbridge. Recent molecular data have suggested phylogeographic links across the Strait of Sicily, either based on very little data, or, with some speculation on the dating of these divergences [2025; see Discussion].

Green toads are widespread in the Palearctic region where they have differentiated into several lineages 26. The occurrence of three bisexually reproducing ploidy levels 27 makes them a uniquely interesting vertebrate group. A recent survey of mtDNA variation characterized a deeply branched assemblage of at least twelve major haplotype groups 28. Other data on green toad variation from North Africa have been relatively scarce 2930. Research has been mainly restricted to faunistics 313233 or analyses of single populations (Egypt 34) without taxonomic or phylogeographic focus. Green toad biology and ecology are relatively well known from the Balearic Islands 3536, Corsica [37 incl. refs.], Sardinia and mainland Italy 38. In addition, morphometric analyses of green toads in Italy have been restricted to peninsular and Sardinian (plus Corsican) populations 3839 until morphometric data for a population from northwestern Sicily became available 40. Otherwise, regional research has been restricted to classical biogeography and faunistics including ecological and taxonomic remarks 414243 and studies on phenology 4044.

To gain insight into the phylogeography of the region, we test a wide range of possible substitution rates in order to better date the African-Sicilian divergence of green toads: Based on only two haplotypes from Sicily, Stöck et al. 28 roughly calibrated their tree by "assuming that the last landbridge between North Africa and Sicily broke about 5.3 Mya." Alternatively, this divergence could be younger. Here, we examine genetic data from many individuals and many more localities in Tunisia, Italy and Sicily, to scrutinize the North Africa-Sicily divergence employing a Bayesian coalescent demographic reconstruction method (BEAST 1.4.6). Using two mitochondrial and two nuclear sequence markers and additional morphological, phenological, bioacoustic and biogeographic data, we characterize green toads from Sicily as a separate evolutionary lineage that is a sister taxon to African green toads, but substantially different from other western Mediterranean green toad forms. These phylogeographic relationships substantiate scarce knowledge on western Mediterranean terrestrial biogeography and may have implications for comparative research on the phylogeography of other terrestrial animals in the region.

We found genetic markers in green toads (Additional file 1) within our geographic scope (Figure 1) to indicate five major spatially structured lineages (Figure 2). (I) The first lineage was found on Corsica, Sardinia, the Balearic Islands, Apennine Peninsula and the northeastern extreme of Sicily. This clade (balearicus) was different from other Eurasian mainland green toads (lineages II and III: viridis and variabilis), whose ranges border northeast to the Po River drainage and belong to widespread monophyletic groups in Eurasia 28. (IV) Although geographically neighboring I, on most of Sicily and its surrounding islands another lineage (Bufo n. sp.) occurs that is substantially different from the first. (V) The entire North African range and the off-coast islands constitutes a fifth lineage (boulengeri). The details of the results defining each lineage are described below.

In particular, our analyses shows a deep sister relationship between African (V) and Sicilian (VI) groups and therefore a rarely studied phylogeographic connection across the Strait of Sicily.

Here we consider and name lineages that maintain their evolutionary integrity with respect to other lineages through both time and space and name them as species under the phylogenetic species concept.

Divergence time estimates for the principal mitochondrial (control region and 16S rRNA) clades recovered are shown in Table 3. We provide the 95% highest posterior density intervals (95% HPD, that is, the shortest intervals that include 95% of the posterior sampled values) as well as the mean of the sampled values. Time estimates are based on the reconstruction of the most common recent ancestor for the mtDNA control regions under the coalescent (see Material & Methods section for details). The 95% HPD, although overlapping, show consistent values within and between clades. Divergence between African and Sicilian haplotypes falls within the range of 0.6 and 3.5 My [mean 1.8 My, around the Pliocene/Pleistocene boundary (~1.8 My)] for the control region. Divergence time estimates were also derived using the 16S data (see Material & Methods for details). Results were similar to those obtained using the mtDNA control region (divergence between African and Sicilian haplotypes: 95% HPD from 0.164 to 3.603 My (mean: 2.051); split of balearicus/viridis clades: 95% HPD from 0.18 and 3.795 (mean: 2.123). All the dates estimated in our analysis seem to represent post-Messinian divergence events.

A discriminant analysis using all 20 characters demonstrated the distinctness of toads from Sicily by correct reclassification of 100% of individuals into the four groups from: (i) Sicily, (ii) N-Africa, (iii) Corsica and Sardinia, and (iv) the Apennine Peninsula. Although encoded as belonging to different geographic groups, the genetically more closely-related groups (iii) and (iv) of the B. balearicus clade showed fewer differences (Figure 4), making us confident of the power of the analyses' and this morphometric dataset. However, in order to test for potential over-parameterization of the analysis, we also reduced the number of morphometric characters to five, which again reclassified 100% of the individuals into their four respective groups.

We have demonstrated a sister relationship on the mitochondrial and nuclear levels between green toads from Sicily (B. siculus, see below) and those from the entire North African range (B. boulengeri). The phylogenetic depth of this divergence, which may range between 0.63 My and 3.5 My (mean 1.83 My) for the control region and from 0.164 to 3.603 My (mean 2.051) for the 16S rRNA, probably excludes human introduction and reveals an infrequently-studied phylogeographic relationship in terrestrial vertebrates. This evolutionary connection may be significant for the phylogeography of the Mediterranean and have implications for research on other terrestrial animals. While some biogeographic studies have suggested common ranges for terrestrial vertebrates (for example the "Siculo-Maltese-Maghrebin" range type of Turrisi and Vaccarro 42), and paleontologists discuss a possible early Pleistocene faunal exchange with North Africa (see introduction), so far, very few molecular studies have suggested any genetic relationships within terrestrial vertebrates across the Strait of Sicily. Two examples come from species (e.g. Discoglossus 25; Chalcides 4720), for which human introduction elsewhere at circum-Mediterranean sites has been demonstrated and for which African-Sicilian relationships have never been demonstrated using any nuclear sequence data. Despite some recent speculation (e.g. Crocidura sicula, potential Messinian origin from North Africa 16), a possible African origin of terrestrial Sicilian fauna has rarely been thoroughly tested with molecular methods, and never using any nuclear sequence, and thus supported by both mitochondrial and nuclear sequence data. To our knowledge, a genetic African-Sicilian link has only been inferred in two other terrestrial species, but even these have ambiguous information on timing and direction: (i) Phylogenetic inference based on allozymes regarding relationships among Chalcides lizards from Sicily, Africa, the Apennine Peninsula and Sardinia 47 showed Sicilian skinks most closely related to Italian mainland skinks (Calabria in the south to Liguria in the north), while lizards from Tunisia appeared genetically almost identical to Sardinian Chalcides, interpreted as a human introduction from Tunisia into Sardinia. Very recently, however, Giovannotti et al. 20 found two Sicilian Chalcides mtDNA haplotypes to have a sister relationship with Tunisian and Sardinian haplotypes (the latter data not shown in the Chalcides phylogeny) and interpreted this as a possible Lower Pleistocene (1.8 My) colonization event of Sicily (and Italy) from North Africa. (ii) Using allozymes and cytochrome b, Zangari et al. 25 assessed lineage relationships among discoglossid frogs from multiple locations in the western Mediterranean (including Sicily, Sardinia, Tunisia). These authors showed a deep divergence of some mtDNA haplotypes across the Strait of Sicily, but also some nearly identical haplotypes across Sicily, Tunisia and Algeria. While the authors commented imprecisely (i.e., they did not specify if the following referred to Gibraltar or Sicily or both) that the "spread of Discoglossus between Europe and Africa should have occurred at the end of the Messinian salinity crisis", they interpreted closely-related haplotypes across the Strait of Sicily as "little genetic differentiation detected among Algerian and Tunisian D. auritus with respect to Maltese and Sicilian D. pictus suggest [ing] a very recent isolation", without further interpreting or dating their results. The significance of land connections for faunal exchanges between Africa and Sicily is mildly challenged by the well-known salt tolerance of green toads. Gordon 48 experimentally demonstrated survival of B. viridis in seawater for several hours or days. Breeding and swimming in brackish (beach or desert) pools and waters 49 as well as estuaries has been observed 38. Therefore, green toads may have the potential for transmarine dispersal. The distance from Africa to Sicily (~140 km), however, can be considered a very strong barrier, even during low sea levels (>45 km). In addition, external fertilization (and absence of brood care) requires that several adults or larvae disperse in order to found new populations. This and the early divergence time (0.63 My < 1.83 My < 3.5 My) of African and Sicilian lineages make it more plausible that toads with a terrestrial life style dispersed via land connections than by rafting. This hypothesis was supported in the recent study of terrestrial Chalcides lizards (see above), which estimated a similar divergence between Tunisian and Sicilian clades (1.8 My 20) to the one we propose for green toads. However, several papers have suggested the ability of anurans to overcome large sea barriers 5051, especially in the tropics, where swimming islands may facilitate rafting. Therefore, comparative phylogeographic data from other terrestrial species in Sicily and North Africa would be necessary to test the feasibility of competing hypotheses.

Divergence time estimates based on molecular data normally rely on external calibration points from the fossil record or from well-known paleogeographic events. Because neither fossil nor paleogeographic calibration points were available, we calculated divergence times using an uncorrelated relaxed molecular clock and a range of substitution rates for the two mitochondrial fragments. We acknowledge a potential limitation, that is, the use of only mtDNA to determine divergence times, which may lead to overestimation of the splitting dates. This overestimation may occur because the most recent common ancestor (MRCA) of the haplotypes (their coalescent) does not necessarily correspond to the real temporal split of the populations but may precede the actual divergence of the populations 52. We consider substitution rates between 1% and 3% per lineage per My to be reasonable values for amphibian mtDNA (for references see Materials and Methods) although they have not been empirically assessed in green toads. When regions with highly repetitive motifs are absent, as in green toads, the control region tends to show higher rates of substitution than the rest of the mitochondrial molecule 5354. Thus, the substitution rates selected here are conservative, higher than the rate for the rest of the mitochondrial molecule, but lower than the fastest observed rates for the control region in other taxa. Higher rates would imply younger dates for the splitting of the African and Sicilian lineages. This point is illustrated by our temporal estimates from the mitochondrial 16S gene which we obtained applying a slower substitution rate, which ranged from 0.03–1% per lineage per My (see refs. in the Discussion). The two markers, with different substitution rates, yield similar 95% HPD time intervals.

Although the values reported should be regarded with the caveats mentioned, it is highly improbable that the vicariant event that separated the African and Sicilian green toads took place during the Messinian (5.3 Mya), that is, earlier than the range of dates estimated by this method. Instead, it is very likely that the Africa-Sicily divergence is post-Messinian. In order to validate the hypothesis proposed here and to test competing paleobiogeographic scenarios, a formal comparative phylogeographic study including more genetic markers and other terrestrial species in Sicily and North Africa would be necessary. Several new comparative phylogeographic approaches have been proposed 555657, each requiring the inclusion of more data for a more reliable statistical inference to be obtained.

The second major result of our study is that green toads from most of the Apennine Peninsula, Corsica, Sardinia and the Balearic Islands represent a separate taxon, which is different from other Eurasian green toads on the mitochondrial and nuclear levels. Three B. balearicus subclades emerged from our mtDNA control region data; the subclade on Corsica and Sardinia exhibited basal haplotypes and the most substantial structure. These observations suggest speciation of B. balearicus on these islands around 0.9 Mya to 1.8 Mya (Table 3). The earliest green toad fossils in Italy are known from the Late Miocene from the northeast (Ravenna Province) and from the Pliocene in the southeast (Gargano region) of the Apennine Peninsula 58. They have also been identified from the Pleistocene of Corsica and Sardinia 59. Corsica and Sardinia had landbridges to (or were separated only by narrow marine straits from) the Apennine Peninsula during several geological periods (~18 Mya; 9 Mya; 5.3 Mya 60). Although a narrow strait may have mostly separated Corsica/Sardinia from Tuscany 61, paleontological 6263 and phylogeographic data (Discoglossus 25, isopods 64) suggest limited early Pleistocene faunal (perhaps oversea) exchange between these islands and the mainland.

This same signature is suggested by our data. The balearicus clade widespread on the Apennine Peninsula displayed a mismatch distribution visually, but not statistically fitting that expected for a population that underwent a sudden expansion (Figure 3). Similarly, the coalescent simulation program Fluctuate estimated a growth parameter ten times higher for the peninsular group than for the island balearicus clade (Table 2), suggesting recent expansion, perhaps after its arrival on the mainland; however, a no growth scenario could not be rejected statistically. Tajima's D did not support a demographic expansion in either of the balearicus groups separately, or combined. The northeastern range limit of B. balearicus appears to be the Po River drainage, which may also be the southwestern boundary for B. viridis. Recent mapping of green toad distributions in Italy 38, however, shows an apparently continuous range from the northeast (probable haplotype group of B. v. viridis) to the southwest (haplotype group of B. balearicus) across the Po drainage. It is likely that the groups are in contact in that area; discerning contacts and possible hybridization dynamics represent challenging future research topics. Indeed, the Po drainage seems to be a biogeographic border and/or contact zone between variously related taxa of amphibians and reptiles (for overview 65: Rana latastii/R. italica; Hyla arborea/H. intermedia 66; Bombina variegata/B. pachypus 67; Rana lessonae/R. bergeri68) or, it is considered a "source of genetic variability" 69.

Fossils demonstrate the presence of green toads on the Balearic Islands from the Upper Pleistocene of Mallorca 70, but early human introduction from Corsica was proposed as the source of green toads on these islands 35. Our data show toads from Mallorca and Menorca to be nested within the mitochondrial clades from the Apennine Peninsula and Corsica/Sardinia, and thus corroborate serum albumin data on toads from Mallorca and Corsica that prove them more closely related to each other than to green toads from Palestine, Africa, and Greece 35. We cannot locate the exact geographic origin of Balearic green toads with our current data; faster evolving nuclear markers and denser sampling in regions of possible origin are required. Our data also corroborate provisional allozyme data (Lattes, A. 1997, Abstr. 3rd World Congr. Herpetol., Prague: p. 123) that associated toads from Corsica, Sardinia and NW-Italy (all from the B. balearicus range) with a ~0.25 Nei's distance from toads from Vienna based on a UPGMA dendrogram (B. viridis; 71 quoting Lattes' abstr.).

Although Calabria consisted of islands until the Pleistocene 72, episodic Upper Pleistocene faunal exchange across the Strait of Messina is well documented by the fossil record 72. While few amphibians crossed the Strait of Messina 67, B. balearicus haplotypes have been detected in northeastern Sicily (loc. 20).

Our phylogeographic findings agree with paleogeographic and fossil data 72, which suggest a long Plio-Pleistocene isolation of Sicily. After their "out of Africa" origin, green toads apparently spread across most of Sicily, where they are well known from the Pleistocene fossil record 22. So far, no phylogeographic signature of the Pliocene subdivision of Sicily into two islands was detected. The effective population size of mtDNA in Bufo siculus n.sp., as inferred from θ (Fluctuate) is small (Table 2).

The distribution analysis indicated an observed distribution in the B. siculus clade that was unimodal and visually congruent with the distribution of expected values for a sudden population expansion, although not quite statistically significant (Figure 3f, p = 0.077). The distribution for this clade is extremely compressed to the left, indicating very low numbers of mismatches and low variance. This profile may indicate an even more recent expansion in this group than in the mainland B. balearicus group. The Fluctuate analysis also did not statistically permit rejection of a no-growth scenario for this clade, and Tajima's D did not fit the expectation for an expanded population.

However, the high similarity among mtDNA haplotypes across the entire island is consistent with the known high mobility of green toads, exemplified by fast post-Pleistocene re-colonization of northern Europe by two haplotype groups 28. The current range of B. siculus appears to follow a pattern known in another Sicilian endemic, the lizard Podarcis wagleriana, that is widespread across Sicily but absent in the very northeast 4273.

Our data reveal taxonomic identity (B. siculus) of toads on Favignana Island (loc. 25) and Ustica Island (loc. 24); green toads on the Aeolian Islands may be members of the geographically proximate B. balearicus (see Additional file 3 for further comments on Circum-Sicilian islands). Detection of Calabrian B. balearicus haplotypes in northeastern Sicily (loc. 20) suggests their relatively recent invasion; either during Pleistocene sea level lows 72, by rafting across the narrow Strait of Messina, or possibly by human introduction. It remains unknown whether both taxa (B. balearicus and B. siculus) occur in sympatry in northern Sicily as range maps suggest 42, and whether they would hybridize, given the considerable mtDNA and breeding-phenology differences. Bioacoustic data in the potential contact zone will also help to illuminate this question.

Mitochondrial and nuclear, morphometric and other preliminary biological data show that green toads from most of Sicily represent a separate lineage. Sicilian green toads have been physically separated by the Strait of Sicily from their closest African relatives for long evolutionary periods. In order to acknowledge these facts, and to raise the potential conservation status of this form, which represents an island endemic, we hereby describe it as a new species.

Genomic DNA was extracted from frozen or ethanol preserved blood, liver, muscle tissue, tail tips (tadpoles) and muscle of vouchers from scientific collections using the Qiagen DNeasy™ kit. We either amplified most of the mitochondrial control region (or "d-loop", ~860 bps 28) or a shorter fragment (577 bps) of the control region using the primer pairs CytbA-L/ControlK-H (PCR: 95°C, 3 min, denaturation; cycle [94°C, 45 s, denaturation; 55°C, 45 s, annealing; 72°C, 60 s, extension] 35 times; 72°C, 5 min, final extension 28).

In those individuals from which we sampled the shorter control region fragment, an additional 611 bps of mitochondrial 16S rRNA were amplified using the primer pairs 16Sar-L/16Sbr-H (PCR: 95°C, 3 min, denaturation; cycle [94°C, 45 s, denaturation; 55°C, 45 s, annealing; 72°C, 60 s, extension] 35 times; 72°C, 5 min, final extension 80).

In representatives from all major mitochondrial clades, we sequenced two nuclear markers. To amplify a fragment of ~880 bps of RAG-1 (recombination activating gene), we used the primers MartFL1 and AmpR1 81 in a touch down PCR (95°C, 4 min, denaturation; first cycle [95°C, 30s; decreasing annealing temperature from 60°C to 45°C of -1°C per cycle, 30 s; 72°C, 1:30 min] 15 times; followed by a second cycle [95°C, 30s; 45°C, 30 s; 72°C, 1:00 min] 20 times; 72°C, 10 min). We also applied primers developed by Friesen et al. 82 for birds to amplify an intron of alpha-tropomyosine, situated between the exons 5 and 6, for the first time (to our knowledge) to anuran amphibians. PCR conditions for Friesen's primers proposed for "frogs" were adapted as follows: 95°C, 1:30 min; cycle [94°C, 30 s; 55.9°, 30 s; 72°C, 45 s] 30 times; 72°C, 5 min. All PCR products were sequenced directly and apparent heterozygote genotypes of tropomyosine were also cloned using the pGEM^®-T vector system (Promega). PCR product concentrations were quantified (NanoDrop^® ND-1000 spectrometer) and adjusted to 25 ng/μl. We mixed 1.5 μl template, 0.075 μl of vector (50 ng/μl), 2.5 μl 2× ligation buffer, 0.5 μl T4 ligase, and 0.425 μl water and ligated overnight (10°C). Transformations (2.5 μl ligation plus 12–25 μl competent cells) were recovered in SOC for 1 h 30 min; 80–100 μl of cell suspension was applied to small agar plates. After incubation (18 h, 37°C), at least eight white colonies were amplified with vector primers M13forw./M13rev. Nested vector primers T7 and Sp6 (Promega) were used as sequencing primers.

All PCR-products and clones were sequenced in both directions and visualized on an ABI 3730 sequencer. Sequences were aligned using Sequencher, v. 4.1.2 and edited using MacClade 4.06.

For each sequence fragment, the best fitting model of sequence evolution was selected using MrModeltest 83. Phylogeny was inferred for each locus separately, using the program MrBayes (v. 3.0b4 84), running four chains for 5 or 10 million generations, with tree sampling every 1000 generations. The "burnin"-value was selected by visualizing the log likelihoods associated with the posterior distribution of trees in the program Tracer. All trees generated before the log likelihood curve flattened out were discarded. From the two different control region fragments, we assembled an overlapping alignment of sequences comprising 541 characters from all 148 individuals included in this study; the best fitting model of evolution was HKY+G (AIC; 10 M generations; burnin = 1500, the same settings were used for the 16S rRNA sequences). For RAG-1 fragments (860 bp, model: GTR+G; AIC, burnin = 1000) and the alpha-tropomyosine intron (612 bp; model: HKY+G; AIC; because of the occurrence of insertions and deletions, all missing and ambiguous data were excluded and only 405 bp were analyzed; 5 M generations; burnin = 1000).

Because amplification or alignment of markers was not equally possible for identical outgroup taxa from the available material, more than one species had to serve as outgroup. In previous analyses [28 and unpublished data], we demonstrated that all used taxa (Bufo surdus, B. raddei, B. calamita, B. bufo) represent suitable outgroup species.

To evaluate the distinctiveness of genetic groups of West-Mediterranean green toads of the B. viridis subgroup [sensu 28], and of further geographic subdivisions within each genetic group, we calculated pairwise F_ST values between all groups using Arlequin 2.0 85 based on the mitochondrial control region.

We assessed the possibility of population expansion for green toad groupings assembled into different geographical populations. We applied three different tests, each with different strengths, to the mitochondrial control region dataset (846 or 541 bp), to detect evidence of recent expansion. First, we applied Fluctuate 86, a maximum-likelihood estimator of the parameters θ and g (θ = 2Ne/μ; g = exponential population growth rate parameter). The exponential growth parameter (g) was used to estimate the size of the population at time in the past from N_t = θ^e-(gμ)t where N_t is the effective population size at time t in the past 86. Using this equation, t was estimated by substituting N_t with N_t/N_{t = 0} = 0.1 (μ is DNA substitution rate per site per generation, N_t is the female effective population size at time t). Repeated analyses to ensure stability of estimates were run as described by Stöck et al. 28. Growth was inferred using logarithmic likelihood ratio tests with one degree of freedom 87.

Second, DnaSP version 4.0 88 was used to calculate and show in graphic form the distributions of observed and expected pairwise nucleotide site differences, also called mismatch distributions, between all individuals within each group, and the respective expected values for growing populations 89. The model of sudden expansion describes an initial population at equilibrium, with the expected pairwise differences, θ₀, (θ = 2N_eμ) and assumes rapid population growth, resulting in θ₁ (Theta final). Tau, τ, is the time of the growth measured in units of substitutional time (τ = 2μt; t is the time in generations, μ the substitution rate per locus and per generation 89). Graphically, the mismatch distribution of a recently expanded population is unimodal and smooth; the wave-shaped curve is centered nearer the y-axis the more recent the expansion, moving away as the number of mismatches increases 89. By letting θ₁ become very large, θ₀ and τ are estimated from the data 90. We set θ₁ to 1,000,000. We assessed the deviation of the observed distribution from the expected under a model of sudden expansion by comparing the raggedness statistics of the observed distribution with a simulated distribution to determine the probability that the raggedness of the observed distribution could have arisen by chance. Third, we estimated Tajima's D 91 in DnaSP for each grouping. Tajima's test of selective neutrality compares two θ estimators and its significance is evaluated by comparison of the test statistic (D) with values randomly generated under "neutrality." Significant values indicate the population has deviated from neutrality, or that another demographic force has caused the deviation from expectation, such as population expansion (significant negative D).

Divergence times among the main mitochondrial lineages were estimated using a Bayesian-coalescence approach, as implemented in BEAST 1.4.6 92939495. In analysis of the control region, we used a matrix of 60 individuals and 752 bp. We started the search with an UPGMA tree, constraining the clade B. boulengeri-B. siculus from Sicily to be monophyletic. Because we were analyzing a species-level phylogeny, we used a Yule tree prior, which assumes a constant speciation rate per lineage. We applied an uncorrelated relaxed molecular clock, with the substitution rate of the branch lengths being sampled from a prior normal distribution with a mean value of 0.02 and a standard deviation of 0.007 94. The search was conducted for a range of substitution rates that varied from 1% to 3% per million years. We ran four independent analyses for 20 × 10⁶generations. We checked for convergence and stationarity of the different analyses in Tracer 1.4 and combined the results in the BEAST module LogCombiner 1.4.4 (after removing the first 2 × 10⁶ generations from each analysis as "burnin").

In analysis of the 16S rRNA, we modified the search strategy to overcome the slow rate of convergence and stationarity of the MCMC chains. We analyzed a matrix of 80 individuals and 512 bp. We first constructed a UPGMA tree with maximum likelihood distances (model selected in Mr.ModelTest v.2: TrN + I), which was specified as the starting tree in Beast. Previous studies have used values of 0.33% for the 16S rRNA or 0.7% more generally for a variety of different regions of the mitochondrial genome [e.g., 7879]. We specified the prior for the mean substitution rate as a normal distribution, with a mean of 0.008 and standard deviation of 0.003. This normal distribution thus covered the relevant range from 0.3% to 1% substitutions per site per My. We conducted two independent runs of 100 × 10⁶ generations and combined the results with LogCombiner 1.4.4, after discarding the first 50 × 10⁶ generations from the two runs as burnin. Results were checked using the program Tracer1.4.

We measured most (20 out of 22) of the morphometric traits described by Castellano and Giacoma 39. As in that paper, we only included male toads: 17 from North Africa (from Morocco, Algeria, Tunisia, Egypt; collection ZFMK), 21 males of the new species from Sicily and the population means published by Castellano and Giacoma 39 on 118 male green toads from six populations from the Apennine Peninsula and NW-Italy [Vado L. (N = 18), Pellice (N = 45), Isolabella (N = 13), Maremma (N = 21), Calopezzati (N = 11), Leverano (N = 10)] and 56 male toads from the islands of Corsica [Cirindinu (N = 18)] and Sardinia [Barratz (N = 19), Portoscuso (N = 19)]. In the discriminant analyses, we initially included all 20 variables. However, as this number of variables probably represents an over-parameterization of a discriminant analysis with this number of specimens, we later reduced this number to five variables to test for its effect. For statistical analyses, the program SPSS 11.0 for Windows was used.