Introduction

For decades, grass snakes (Natrix natrix sensu lato) were thought to be one of the most widely distributed Palearctic snake species, occurring from the North African Maghreb region through the Iberian Peninsula and most of Europe to Lake Baikal in Central Asia (Kabisch 1999). However, two recent investigations have shown that what was previously understood as one polytypic species was composed of three distinct species (Pokrant et al. 2016; Kindler et al. 2017), each with phylogeographic differentiation (Kindler et al. 2013). Ibero-Maghrebian grass snakes are now recognised as the distinct species Natrix astreptophora (Seoane, 1884) (Pokrant et al. 2016). Grass snakes in the remaining western part of the distribution range represent the western species N. helvetica (Lacepède, 1789), which hybridises with the eastern species N. natrix (Linnaeus, 1758) in a narrow belt in the Rhine region (Kindler et al. 2017). Natrix astreptophora differs from the other two grass snake species, N. helvetica and N. natrix, in skull morphology, coloration and pattern, in particular by its unique reddish iris. Moreover, the contact zone of N. astreptophora and N. helvetica is characterised by a parapatric distribution of the two taxa, with rare hybridization (Pokrant et al. 2016).

Two previous studies using mitochondrial DNA (mtDNA) found Iberian and French N. astreptophora deeply divergent from a single sample from Tunisia (Kindler et al. 2013; Pokrant et al. 2016). Natrix astreptophora has a wide relictual distribution in the Maghreb, with few isolated records from Morocco through Algeria to Tunisia (Fig. 1; Bons and Geniez 1996; Schleich et al. 1996). Many amphibian and reptile species, some small mammals and at least one land snail show a pronounced west-east differentiation in the Maghreb (reviews in Husemann et al. 2014; Stuckas et al. 2014; Nicolas et al. 2015), posing the question how Tunisian grass snakes are related to other North African populations. Until now, the rareness of North African grass snakes has prevented examining their differentiation, and populations from Morocco and Algeria remained completely unstudied. For this paper, we expanded upon previous sampling to include novel specimens from most of the North African distribution, including Algeria and Morocco. Using two mitochondrial DNA (mtDNA) fragments and 13 microsatellite loci, we examined the phylogeography of N. astreptophora across its entire range and compared it to other co-distributed taxa.

Fig. 1
figure 1

Sampling sites and mitochondrial identity of studied red-eyed grass snakes (n = 56). Olive green areas indicate distribution of Natrix astreptophora in Northern Africa according to Bons and Geniez (1996), Schleich et al. (1996) and Sindaco et al. (2013). Two questionable localities in North Africa are not shown (Atlantic coast of Morocco, Schleich et al. 1996; southern Algeria, Hecht 1930). Colours of sampling sites correspond to Figs. 2, 3, 4. Inset: N. astreptophora from Morocco; photo: Salvador Carranza

Full size image

Besides the scientific names, we use below the well-established vernacular name ‘barred grass snake’ for N. helvetica and refer to N. natrix as ‘common grass snake’ or ‘eastern grass snake’. For N. astreptophora, we introduce here the name ‘red-eyed grass snake’, acknowledging its unique and highly diagnostic iris coloration.

Materials and methods

A total of 56 Natrix astreptophora were studied (Table S1), corresponding to 11 new samples and 45 previously processed ones (Kindler et al. 2013; Pokrant et al. 2016), including GenBank data (Guicking et al. 2006). Our new samples came from Morocco (4), Algeria (2), Tunisia (3) and Spain (2). For these samples, the same two mtDNA fragments were sequenced as in Kindler et al. (2013) and Pokrant et al. (2016), i.e. the partial ND4 gene plus adjacent DNA coding for tRNAs (866 bp) and the cyt b gene (1117 bp). Laboratory procedures were the same as in Kindler et al. (2013). For phylogenetic analyses, the new sequences were concatenated and merged with previously published data for N. astreptophora plus three sequences each for the 14 terminal clades of N. helvetica and N. natrix as identified by Kindler et al. (2013). Sequences of N. maura, N. tessellata and Nerodia sipedon served as outgroups (Table S2). For phylogenetic analyses, the optimal partition scheme and best-fit evolutionary models were assessed using partitionfinder 1.1.1 (Lanfear et al. 2012), with linked branch lengths and the Bayesian Information Criterion (BIC) for model selection. Partitionfinder was run twice, evaluating either evolutionary models implemented in mrbayes 3.2.6 (Ronquist et al. 2012) or implemented in RAxML 8.2.4 (Stamatakis 2014). A user-specific search was run for the following partition schemes: (a) unpartitioned, (b) partitioned by gene with DNA coding for tRNAs merged in one partition, and (c) maximum partitioning, i.e., using each codon position of protein-coding genes and the merged tRNAs as a distinct partition, resulting in scheme (a) and the GTR+G+I model for both tree-building methods. Using mrbayes 3.2.6, two parallel runs were then computed, each with four chains. The chains ran for 10 million generations, with every 100th generation sampled. Convergence of runs was confirmed by the average standard deviation of split frequencies approaching zero. Using the sump command, stationarity was further verified by plotting the generation versus the log probability of the data, with a ‘white noise’ pattern indicating no tendency to increase or decrease over time. For the final 50% majority rule consensus tree, a burn-in of 2.5 million generations (25%) was used. In addition, Maximum Likelihood (ML) analyses were conducted using RAxML 8.2.4. Five independent ML searches were run with different starting conditions and the fast bootstrap algorithm to explore the robustness of the branching patterns by comparing the best trees. Then, 1000 nonparametric thorough bootstrap values were calculated and plotted against the best tree.

In addition, mtDNA sequences were examined using parsimony networks as implemented in tcs 1.21 (Clement et al. 2000), with gaps coded as fifth character state. Because for some samples only one mtDNA fragment was available, each fragment was analysed separately, as in Pokrant et al. (2016) and Kindler et al. (2017), due to the fact that tcs software cannot cope with missing data. Using the default 95% connection limit, unconnected haplotype clusters were obtained for the different genetic lineages, which is why the connection limit was arbitrarily set to 100 steps. Based on haplotypes of each mtDNA fragment (Table S3), uncorrected p distances (average) were computed using mega 7.0.21 (Kumar et al. 2016) and the pairwise deletion option for each of the three grass snake species and clades within each species.

Besides mtDNA, also 13 polymorphic microsatellite loci were studied. For 29 samples of European N. astreptophora, data were available from Pokrant et al. (2016). However, one sample (MTD T 13083) was excluded because it showed genetic impact of N. helvetica. Eleven new samples (Table S1) were genotyped according to the procedures of Pokrant et al. (2016) and data were examined using Principal Component Analyses (PCA) as implemented in the R package adegenet (Jombart 2008). Additional analyses using unsupervised Bayesian cluster analyses are described in the Supporting Information of this article. Pairwise F ST values and Analyses of Molecular Variance (AMOVAs) were calculated for each mtDNA fragment and for microsatellite data using arlequin 3.5.1.3 (Excoffier and Lischer 2010).

For mitochondrial clades, a dating analysis using a relaxed molecular clock was performed with beast 1.8.4 (Drummond et al. 2012) using the same fossil calibration and settings as in Fritz et al. (2012). A Natrix vertebra from Sardinia (Delfino et al. 2011) of known age (3.6 million years) was applied to constrain the divergence of the Corso-Sardinian clade (clade B in Fig. 2) from its sister clade. The lognormal prior for the TMRCA was set to mean = 0.4, log(stdev) = 1 and offset = 3.6. For finding the optimal partition scheme and model of sequence evolution for the dating analysis, partitionfinder was run again for the models available in beast using the same approach as described above, resulting in the selection of the GTR+G+I model for an unpartitioned data set. As in Fritz et al. (2012), the data set for the molecular clock was based on four mitochondrial genes (cyt b, ND1, ND2 and ND4 without tRNAs) consisting of one sequence of each mitochondrial clade of N. astreptophora, N. helvetica and N. natrix plus two sequences of N. maura and one sequence of N. tessellata (Table S4). For clades not included in the study by Fritz et al. (2012), only cyt b and ND4 data were available (clades A, 2, 5 and 6 of Kindler et al. 2013; western and eastern Maghrebian clades of this study).

Fig. 2
figure 2

Mitochondrial phylogeny of all three grass snake species inferred from Maximum Likelihood using 1984-bp mtDNA (ND4+tRNAs, cyt b). Terminal clades collapsed to cartoons. Outgroups (Natrix maura, N. tessellata, Nerodia sipedon) removed for clarity. Numbers along nodes indicate branch support under Maximum Likelihood (1000 bootstrap replicates) and Bayesian Inference (posterior probabilities). Asterisks indicate maximum support under both tree-building methods. For Natrix helvetica and N. natrix, clade names correspond to Kindler et al. (2013). Inset: European N. astreptophora (near Nohèdes, south-western France); photo: Philippe Geniez

Full size image

Results

Under both tree-building methods, sequences from Natrix astreptophora corresponded to three well-supported clades that were placed in a well-supported more inclusive clade (Fig. 2). The three clades matched with sequences from (i) the Iberian Peninsula and France, (ii) the western Maghreb region (Morocco), and (iii) the eastern Maghreb region (Algeria, Tunisia). Clades (ii) and (iii) were, with weak support, together the sister group of (i). The relationships and internal branching patterns of the other two grass snake species (N. helvetica, N. natrix) conformed to expectations from previous studies (Guicking et al. 2006; Fritz et al. 2012; Kindler et al. 2013).

Using the two individual mtDNA fragments of N. astreptophora for parsimony networks, each of the three clades corresponded to a distinct haplotype cluster (Fig. 3). For both mtDNA fragments, the European clade had the highest number of haplotypes, which is likely to be related to the large sample size. For the mtDNA sequences comprised of the partial ND4 gene and adjacent DNA coding for tRNAs, there were 13 haplotypes for the European cluster, three haplotypes for the Moroccan cluster and four haplotypes for the Algerian-Tunisian cluster. The European cluster differed from the Moroccan cluster by a minimum of 26 mutation steps and from the Algerian-Tunisian cluster by 21 steps. The two North African clusters differed by 29 steps from one another. Within the European cluster, a maximum of six steps occurred and within the Moroccan cluster a maximum of five steps. Sequences from the eastern Maghreb differed by a maximum of four steps. For the cyt b gene, the European cluster was comprised of 20 haplotypes, the Moroccan cluster contained three haplotypes, and the cluster with sequences from Algeria and Tunisia had five haplotypes. The European and Moroccan clusters differed by 44 mutation steps; the European and Algerian-Tunisian clusters, by 49 steps, and the two Maghrebian clusters, by 39 steps. Within the European cluster, up to nine steps occurred; within each Maghrebian cluster, up to five steps.

Fig. 3
figure 3

Parsimony networks of mtDNA sequences of Natrix astreptophora. Symbol size corresponds to haplotype frequency; lines connecting haplotypes represent one mutation step, if not otherwise indicated. Small black circles are missing node haplotypes. Haplotype colours correspond to lineages: European lineage in orange, western Maghrebian lineage in green and eastern Maghrebian lineage in brown. Haplotype names in blue. For European Nucleotide Archive (ENA) accession numbers, see Table S3. Differences in mutation counts compared to Pokrant et al. (2016) are due to longer DNA sequences

Full size image

Uncorrected p distances (averages) between the three grass snake species ranged for the haplotypes corresponding to the partial ND4 gene and adjacent mtDNA from 5.49 to 5.67%; within-species divergences ranged from 1.73 to 3.39% (Table S5). For the cyt b gene, the values between species were 6.32–7.08%, and within species, 1.90–3.59% (Table S6). Regarding the three clades within N. astreptophora, divergences between 2.61 and 3.43% were observed for the ND4 gene plus adjacent mtDNA, and for the cyt b gene, divergences between 3.69 and 4.38% (Tables S7, S8). These values fall into the upper ranges among the clades of the barred grass snake (N. helvetica: 0.63–6.01% for ND4+tRNAs, 0.45–5.22% for cyt b) and the common grass snake (N. natrix: 0.59–5.16% for ND4+tRNAs, 1.37–5.74% for cyt b; Tables S7, S8).

According to an AMOVA for the mtDNA fragment containing ND4+tRNAs, 95.58% of the molecular variance occurred between and 4.42% within the three mitochondrial lineages of N. astreptophora. Pairwise F ST values for the three lineages ranged from 0.93 to 0.96 (Table S9). For cyt b sequences, 94.15% of the molecular variance occurred between and 5.85% within the mitochondrial lineages. Pairwise F ST values varied from 0.93 to 0.94 (Table S10).

The microsatellite loci of N. astreptophora turned out to be highly polymorphic (Table 1). The PCA using these data (Fig. 4) confirmed the distinctness of red-eyed grass snakes representing the three mitochondrial lineages and found three different clusters without any overlap. This was also in line with the unsupervised Bayesian clustering approach (Supporting Information). An AMOVA for microsatellite data revealed that 33.71% of the molecular variance occurred between and 66.29% within the three clusters. Pairwise F ST values varied from 0.25 to 0.45 (Table S11).

Table 1 Repeat motifs, allele size ranges and number of alleles of microsatellite loci. For primer sequences, annealing temperatures and multiplex-sets, see Pokrant et al. (2016)
Full size table
Fig. 4
figure 4

Principal Component Analysis (PCA) for microsatellite data. Samples are coloured according to mitochondrial lineages. The oval outlines represent 95% confidential intervals. For axes 1–2 (left), the x axis explains 15.4% and the y axis 11.6% of variance. For axes 1–3 (right), the y axis explains 7.9% of variance. Non-overlapping groups denote significantly different clusters

Full size image

The dating analysis using mtDNA yielded for N. helvetica similar results as in Fritz et al. (2012). The only exception was clade B, which was inferred to be slightly younger (0.5 million years; Table S12). However, the divergence times of and the branching patterns within N. astreptophora and N. natrix were generally older (0.5–1.89 million years). According to our new estimates, N. astreptophora diverged 10.61 million years ago (mya) and its European clade branched off 5.44 mya, and the western and eastern Maghrebian clades split 4.63 mya (Fig. 5).

Fig. 5
figure 5

Estimated divergence times of grass snakes and their 95% HPD intervals (blue bars). Outgroups removed for clarity. The red arrow highlights the placement of the fossil (3.6 mya; Delfino et al. 2011) used for calibrating the respective node

Full size image

Discussion

Our new data clearly demonstrate that the red-eyed grass snake (Natrix astreptophora) shows pronounced phylogeographic structuring, with one genetic lineage each in the Iberian Peninsula plus adjacent France as well as in the western Maghreb (Morocco) and the eastern Maghreb (Algeria, Tunisia). The distinctness of these three lineages is concordantly confirmed by mitochondrial and microsatellite data. Moreover, with respect to mitochondrial haplotypes, the two North African lineages are remarkably variable, despite small sample sizes. This is especially obvious for ND4+tRNAs (Fig. 3, top). Among the 13 haplotypes from Europe, a maximum of six mutations occurs. Samples from the western and the eastern Maghreb differ by five and four mutations, respectively, i.e. have a similar degree of variation, although only three haplotypes were found in Morocco and four haplotypes in Algeria and Tunisia. This suggests that the North African populations might be genetically more diverse than the European ones.

Numerous taxa in North Africa display a phylogeographic pattern resembling that of N. astreptophora, with different lineages in the western and eastern Maghreb, in many cases with the Moroccan Moulouya River as a divide. Moroccan populations west of the Moulouya River are often closely related to or indistinguishable from Iberian conspecifics. However, in some instances, Morocco can also harbour one or more clades being distinct from those in the Iberian Peninsula and the eastern Maghreb (Husemann et al. 2014; Stuckas et al. 2014; Martínez-Freiría et al. 2017). Such endemic lineages may occur in Morocco alone or in addition to the ‘Iberian’ ones. Examples for the general west-east paradigm in the Maghreb are the following snake species: Coronella girondica (Santos et al. 2012), Macroprotodon spp. (Carranza et al. 2004), Malpolon spp. (Carranza et al. 2006), Natrix maura (Barata et al. 2008; Guicking et al. 2008), the Vipera latastei complex (Velo-Antón et al. 2012) and Daboia mauritanica (Martínez-Freiría et al. 2017). Shared genetic lineages between the Iberian Peninsula and Morocco occur in Macroprotodon brevis (Carranza et al. 2004) and Malpolon monspessulanus (Carranza et al. 2006). Other taxa are reviewed in Barata et al. (2008), Husemann et al. (2014) and Stuckas et al. (2014). The lack of genetic differentiation between Moroccan and Iberian conspecifics is explained either by natural transoceanic dispersal across the Strait of Gibraltar or translocation by humans (Veith et al. 2004; Fritz et al. 2006; Recuero et al. 2007; Velo-Antón et al. 2015; Veríssimo et al. 2016). Since grass snakes are excellent swimmers (Kabisch 1999), the differentiation of Iberian and Moroccan N. astreptophora was not necessarily expected. Today, the Strait of Gibraltar is only 15 km wide and during glacial low sea level stands it was further reduced to 5 km (Zazo 1999); distances that should be easily bridged by semiaquatic species. Accordingly, both freshwater turtle species with similar distribution patterns as N. astreptophora have mitochondrial clades and haplotypes occurring on both sides of the Strait of Gibraltar (Emys orbicularis: Stuckas et al. 2014; Velo-Antón et al. 2015; Mauremys leprosa: Fritz et al. 2006; Veríssimo et al. 2016). However, as in the red-eyed grass snake, the Strait of Gibraltar constitutes a significant phylogeographic barrier for the semiaquatic viperine snake N. maura (Barata et al. 2008; Guicking et al. 2008).

Moroccan N. astreptophora are neither closely related to nor undifferentiated from Iberian conspecifics, but represent a genetic lineage being deeply divergent from both European and eastern Maghrebian N. astreptophora. Their genetic divergences fall into the upper ranges observed between the genetic lineages of N. helvetica and N. natrix. Our divergence time estimates suggest that the clades from the western and eastern Maghreb split 4.63 mya and that the European clade is slightly older, with 5.44 mya (Fig. 5). These dates resemble the inferred age of the Messinian Salinity Crisis (MSC), which commenced ca. 7.25 mya with the gradual restriction of the Atlantic-Mediterranean connection, leading to the complete isolation of the Mediterranean Basin and a huge sea level fall or even complete drying-out of the Mediterranean Basin. It was reflooded in a catastrophic event at end of the Messinian, 5.33 mya (Roveri et al. 2014), and it seems likely that, as a corollary of the MSC, massive environmental changes triggered the radiation of the extant lineages of N. astreptophora. Yet, the reestablishment of a sea strait between the Iberian Peninsula and North Africa was not necessarily the only responsible vicariance event. During the Late Miocene and Early Pliocene, severe climatic fluctuations caused large-scale environmental shifts (Roveri et al. 2014) that probably contributed to repeated range contractions and expansions. It seems likely that these events played a major role in shaping the current genetic structure of many Mediterranean animal and plant species. Using mitochondrial and nuclear markers, Guicking et al. (2008) found in the viperine snake (N. maura) a similar phylogeographic pattern as in N. astreptophora. These authors, however, concluded that the reflooding of the Mediterranean at the end of the MSC caused the genetic divergence of European and North African viperine snakes.

In North Africa, and in Morocco in particular, N. astreptophora is extremely rare and confined to relictual, humid and relatively cool habitats (Bons and Geniez 1996), which are threatened by destruction. With the discovery of two endemic genetic lineages in North Africa, it is clear that their conservation needs more efforts than before. Furthermore, the presented genetic data indicate that a full taxonomic revision of N. astreptophora is needed, which should include a morphological comparison of all three genetic lineages.