Received: 25 February 2018 | Revised: 23 April 2018 | Accepted: 1 May 2018 DOI: 10.1111/zsc.12295

ORIGINAL ARTICLE

Evolution, biogeography and systematics of the western Palaearctic Zamenis ratsnakes

Daniele Salvi1,2 | Joana Mendes2,3 | Salvador Carranza3 | David James Harris2,4

1Department of Health, Life and Environmental Sciences, University of Abstract L'Aquila, L'Aquila, Italy The phylogenetic relationships between western Palaearctic Zamenis and Rhinechis 2CIBIO Research Centre in Biodiversity and ratsnakes have been troubled, with recent estimates based on the supermatrix ap- Genetic Resources, InBIO, Universidade do proach questioning their monophyly and providing contradictory results. In this Porto, Vairão, Vila do Conde, Portugal study, we generated a comprehensive molecular data set for Zamenis and closely re- 3Institute of Evolutionay Biology (CSIC‐ Universitat Pompeu Fabra), Barcelona, lated ratsnakes to assess their phylogenetic and systematic relationships and infer Spain their spatial and temporal modes of diversification. We obtained a fully resolved and 4 Departamento de Biologia, Faculdade de well‐supported phylogeny, which is consistent across markers, taxon‐sets and phylo- Ciências, Universidade do Porto, Porto, genetic methods. The close phylogenetic relationship between Rhinechis and Zamenis Portugal is well‐established. However, the early branching pattern within this clade, and the Correspondence position of R. scalaris and Z. hohenackeri, remains poorly supported. The Persian Daniele Salvi, Department of Health, Life and Environmental Sciences, University of ratsnake Z. persicus is sister to the Mediterranean species Z. situla, Z. longissimus L'Aquila, 67100 Coppito, L'Aquila, Italy. and Z. lineatus, of which Z. situla is sister to a clade containing the latter two species. Email: [email protected] These results are consistent with a recent phylogenomic study on ratsnakes based on Funding information hundreds of loci. Whereas, topological tests based on our data and evidence from Fundação para a Ciência e a Tecnologia, Grant/Award Number: IF/01627/2014 such phylogenomic study strongly rejected previous phylogenetic estimates based on and SFRH/BD/81528/2011; Ministero the supermatrix approach and demonstrate that these “mega phylogenies”, with hun- dell'Istruzione, dell'Università e dreds of taxa and high levels of missing data, have recovered inconsistent relation- della Ricerca; Consejo Superior de Investigaciones Científicas, Grant/Award ships with spurious nodal support. Biogeographical and molecular dating analyses Number: CGL2015‐70390 suggest an origin of the ancestor of Rhinechis and Zamenis in the Aegean region with early cladogenesis during the Late Miocene associated with the Aegean arch forma- tion and support a scenario of east‐to‐west diversification. Finally, while we have little morphological and phylogenetic evidence for the distinctiveness between Rhinechis and Zamenis, a classification of them in a single genus, and the designation of Zamenis scalaris (Schinz, 1822), reflects better their evolutionary relationships.

KEYWORDS Colubridae, fast radiation, multilocus phylogeny, Rhinechis, species tree, supermatrix

1 | INTRODUCTION sometimes considered as a tribe or as a distinct subfamily, Coronellini, January 1863 (Utiger, Schatti, & Helfenberger, Ratsnakes represent an evolutionary lineage of the subfam- 2005). Some 82 species and 17 genera of ratsnakes are cur- ily Colubrinae Oppel, 1811, within Colubridae Oppel, 1811, rently recognized (Uetz, Freed, & Hošek, 2018; not includ- ing Gonyosoma and Coelognathus, Chen, Mckelvy, et al.,

Daniele Salvi and Joana Mendes have contributed equally to this work. 2014), mainly distributed in the Holarctic region but also in

Zoologica Scripta. 2018;47:441–461. wileyonlinelibrary.com/journal/zsc © 2018 Royal Swedish Academy of Sciences | 441 442 | SALVI et al. the Oriental and northern portions of the Neotropical region. Rhinechis Michahelles, 1833 for the western Mediterranean Over the last decades, molecular studies have contributed to species R. scalaris (Schinz, 1822). Utiger et al. (2002), based a drastic reconsideration of ratsnakes’ systematics and evo- on mitochondrial DNA (mtDNA) sequence data, established lution. Until the end of the last century, most species distrib- the monophyletic genus Zamenis Wagler, 1830 including the uted in the Northern Hemisphere were included in a single Mediterranean species Z. longissimus Laurenti, 1768, Z. lin- genus (Elaphe) based on shared morphological features (see eatus Camerano, 1891 and Z. situla Linnaeus, 1758, and the Schulz, 1996). Extensive research based on DNA sequence species Z. persicus Werner, 1913 and Z. hohenackeri Strauch, data restricted the genus Elaphe Fitzinger, 1833 to a dozen 1873 from south‐western Asia (Figure 1). of Palaearctic species and provided strong phylogenetic ev- Phylogenetic estimates of the relationships between idence for a monophyletic origin of New World ratsnakes Zamenis and Rhinechis have been contradictory in differ- (Lampropeltini), which were assigned to distinct genera (e.g., ent studies (Figure 2). First molecular studies based mainly Burbrink & Lawson, 2007; Rodríguez‐Robles & De Jesús‐ on mtDNA showed R. scalaris either branching off early Escobar, 1999; Utiger et al., 2002, 2005). in a solitary phylogenetic lineage within the European Western Palaearctic ratsnakes belong to four distinct lin- ratsnakes (Lenk et al., 2001; Utiger et al., 2002) or sister to eages (Helfenberger, 2001; Lenk, Joger, & Wink, 2001; Schulz, the Zamenis clade (Burbrink & Lawson, 2007; Utiger et al., 1996; Utiger et al., 2002). These include representatives of 2002), although these phylogenetic relationships received the genera Coronella Laurenti, 1768, Elaphe and two other low statistical support. In contrast, recent studies based on recently designated genera. Helfenberger (2001), based on an- the supermatrix approach, with hundreds or thousands of atomical and allozyme data, resurrected the monotypic genus Colubridae or Squamata taxa, recovered R. scalaris deeply

FIGURE 1 Distribution of the ratsnakes genera Zamenis, Rhinechis and Coronella in the Palaearctic (based on IUCN, 2017; Salvi et al., 2017): (a) distribution of Zamenis species; (b) distribution of R. scalaris; (c) distribution of Coronella species [Colour figure can be viewed at wileyonlinelibrary.com] SALVI et al. | 443 (a) Lenk et al. 2001 (weightedMP mtDNAtree) (d) Pyronet al. 2011 (ML supermatrix tree)*

98 Zamenis longissimus 100 Rhinechis scalaris 97 Zamenislineatus 99 Zamenis situla 98 Zamenis situla 100 Zamenis longissimus 91 Zamenispersicus Zamenislineatus 97 82 Zamenis hohenackeri Zamenispersicus 100 Oreocryptophis porphyraceus 67 Zamenis hohenackeri 100 Elaphe spp. (3) 99 Rhinechis scalaris (e) Pyronet al. 2013 (ML supermatrix tree)

Zamenis hohenackeri 99 Zamenis persicus (b) Utigeret al. 2002 (weightedMP mtDNAtree) 100 Zamenis longissimus 64 100 Zamenis lineatus 97 Elaphe spp. (10) Rhinechisscalaris Lampropeltini (13) 100 Zamenis situla Oreocryptophis porphyraceus Rhinechis scalaris (f) 77 Figueroa et al. 2016 (ML supermatrix tree) Orthriophis spp. (4)

Zamenis hohenackeri Zamenis hohenackeri 47 89 Zamenis longissimus 99 Zamenis persicus Zamenis situla Rhinechis scalaris 75 93 Zamenis lineatus Zamenis situla Zamenis persicus 99 Zamenis lineatus 99 Oocatochus rufodorsatus Zamenis longissimus 58 Coronella spp. (2) 100 Ptyaskorros

(c) Burbrink & Lawson 2007 (ML mtDNA+cmos tree) (g) Zheng & Wiens 2016 (ML supermatrix tree)

Rhinechis scalaris Zamenis hohenackeri 91 Zamenis hohenackeri Zamenispersicus Zamenispersicus 38 100 Zamenislineatus 44 46 Zamenislineatus 100 Zamenis longissimus 100 Zamenis situla 100 Rhinechisscalaris Zamenis situla

FIGURE 2 Summary of the previous phylogenetic hypotheses on Zamenis and Rhinechis based on molecular data (mainly mitochondrial data) and different taxon‐sets: ratsnakes taxon‐sets (a–c), Serpentes taxon‐sets (d, f), and Squamata taxon‐sets (e, g). Branch lenghts in each tree are scaled as in the original study; trees are simplified by collapsing some clades (in brackets is reported the number of species included in each collapsed clade); bootstrap values as in the original studies are reported. (* the tree from Zaher et al., 2012 is identical to the tree of Pyron et al., 2011 as it is based on the same data‐matrix with sequences added for a single species) nested within Zamenis in a highly supported clade with re- ratsnake, Ptyas korros. This precipitate taxonomic change solved intraclade relationships (Figueroa, McKelvy, Grismer, is not justified in the light of the current phylogenetic insta- Bell, & Lailvaux, 2016; Pyron, Burbrink, & Wiens, 2013; bility of Rhinechis and Zamenis across available studies, and Pyron et al., 2011; Zheng & Wiens, 2016; Zaher et al., 2012; has been hardly ever accepted in recent literature (e.g., Chen, see Figure 2). These “mega‐phylogenies” questioned the Lemmon, Lemmon, Pyron, & Burbrink, 2017; Speybroeck, monophyly of Zamenis and its distinction from Rhinechis and Beukema, Bok, & Van Der Voort, 2016; Zheng & Wiens, prompted taxonomic instability of these taxa. Wallach’s cata- 2016; a Google Scholar search showed that the synonymiza- logue of world’s snakes (2014) synonymized Rhinechis with tion of Rhinechis with Zamenis has been adopted in only three Zamenis based on the re‐analysis of the Pyron et al. (2011) studies since 2014, compared to over 140 hits for the binomen supermatrix by Zaher et al. (2012), who however did not rec- Rhinechis scalaris). Furthermore, the relationships between ommend any taxonomic changes. Later, Figueroa et al. (2016) Zamenis species are unstable across phylogenetic studies, also suggested to synonymize Rhinechis with Zamenis, how- with each work presenting a different, well‐supported, topol- ever their phylogenetic results showed a paraphyletic Zamenis ogy [except the trees from Pyron et al. (2013) and Zheng and clade, which included Rhinechis but also the Indo‐Chinese Wiens (2016), because these used the same data set]. 444 | SALVI et al. Rapid evolution of Palaearctic ratsnakes, and especially 1843, Pantherophis Fitzinger, 1843, Pituophis Holbrook, within Zamenis, might be an additional explanation, besides 1842 and Oocatochus Helfenberger, 2001 to be used in the methodological artefacts, for the phylogenetic instability of phylogenetic analyses for ingroup and outgroup testing. this clade in previous molecular studies. Rapid radiations Information regarding the original specimens used in the generate a phylogenetic pattern with short deep branches analyses, sampling localities and GenBank accession num- combined with long terminal branches, which is notori- bers of new and published sequences is given in Table 1. ously difficult to resolve in phylogenetic analyses, especially Total genomic DNA was extracted from alcohol‐preserved when these are based on fast‐evolving characters such as muscle following the standard high‐salt protocol (Sambrook, mtDNA sequences (Cummins & McInerney, 2011; Hoelzer Fritsch, & Maniatis, 1989). We generated DNA sequences from & Meinick, 1994). Indeed, previous studies showed deep seven gene fragments, including two mitochondrial genes: cy- phylogenetic divisions, with associated low statistical sup- tochrome‐b (cytb) and NADH Dehydrogenase 4 plus flanking port between Palaearctic ratsnakes (Lenk et al., 2001; Utiger tRNAs Serine, Histidine and Leucine (nd4); and five nuclear et al., 2002), suggesting that most cladogenetic events—such genes: oocyte maturation factor mos (cmos), dynein axone- as those leading to Rhinechis and Zamenis—took place in a mal heavy chain 3 (dnah3), prolactin receptor (prlr), spectrin short time frame (Burbrink & Lawson, 2007). Rapid diversi- beta, non‐erythrocytic 1 intron 1 (sptbn1) and vimentin intron fication within Zamenis may have been associated with the 5 (vim). These mitochondrial markers have been successfully invasion of the western Palearctic from the east as suggested used in many intra‐ and interspecific studies on squamates (e.g., by the east‐west pattern of phyletic diversification, with east- Mendes, Harris, Carranza, & Salvi, 2017; Salvi, Harris, Bombi, ern species Z. hohenackeri and Z. persicus branching off ear- Carretero, & Bologna, 2010; Salvi, Schembri, Sciberras, & lier than Mediterranean species Z. situla, Z. longissimus and Harris, 2014; Sampaio, Harris, Perera, & Salvi, 2014), whereas Z. lineatus (Utiger et al., 2002). However, this biogeographi- fast‐evolving nuclear markers were chosen based on Townsend, cal hypothesis has not yet been formally tested. Alegre, Kelley, Wiens, and Reeder (2008), Chen, Jiang, et al. In this study, we generated a comprehensive molecular (2014), and Salvi et al. (2017). Target sequences were amplified phylogenetic framework for Zamenis and closely related through polymerase chain reaction (PCR) using primers and ratsnakes based on seven gene fragments and multiple indi- conditions reported in Supplementary Table S1. Purification viduals from each species. We tested the robustness of this and sequencing of PCR products were carried out by a commer- phylogeny to ingroup and outgroup choice, marker choice cial sequencing company (GENEWIZ: www.genewiz.com), (mitochondrial vs. nuclear) and missing data, phylogenetic using the same primers employed for amplification. methods (Maximum Likelihood vs. Bayesian Inference) and approach (gene trees and concatenation vs. species trees). Molecular dating and biogeographical analysis were used to 2.2 | Phylogenetic data sets further explore the temporal and spatial pattern of phyletic DNA sequences were edited in Geneious (Kearse et al., 2012) diversification. The main aims of this study are to infer the and heterozygous positions in the nuclear genes were coded evolutionary and biogeographical history of Zamenis and according to the IUPAC ambiguity codes. All sequences were Rhinechis ratsnakes and to assess their phylogenetic rela- aligned using the MUSCLE algorithm (Edgar, 2004) imple- tionships and taxonomy in the light of current controversial mented in Geneious with default parameters. Insertions or estimates based on mitochondrial and supermatrix data sets. deletions (indels) in the intronic fragments sptbn1 and vim were manually phased. Haplotype reconstruction of the nu- clear genes was performed in PHASE 2.1 (Stephens & Scheet, 2 | MATERIAL AND METHODS 2005; Stephens, Smith, & Donnelly, 2001) using the input files obtained in SeqPHASE (Flot, 2010; available at http:// 2.1 | Sample and sequence data collection seqphase.mpg.de/seqphase/). We ran PHASE three times for We collected, either from live individuals or museum spec- each nuclear gene to ensure consistency, applying a probabil- imens, muscle tissues from three samples of each Zamenis ity threshold of 0.7, 100 iterations and the remaining settings species, of R. scalaris and of Coronella austriaca Laurenti, as default. The possible occurrence of recombination events 1768 and C. girondica Daudin, 1803 and two samples of was assessed using the Pairwise Homoplasy Index (phi) test the species Hierophis viridiflavus Lacépède, 1789 and (Bruen, Philippe, & Bryant, 2006) implemented in SplitsTree4 Hemorrhois algirus January 1863. Sequences obtained (Huson & Bryant, 2006). We used MEGA 6 (Tamura, Stecher, from these specimens were complemented with GenBank Peterson, Filipski, & Kumar, 2013) to estimate the number of data including, whenever possible, samples from a close variable and parsimony informative sites in all genes. geographical location to our specimens. Additional se- Phylogenetic inference was performed by (a) single‐ quences were downloaded for Old and New World ratsnake locus network analyses, based on full‐length sequences of species of the genera Elaphe, Lampropeltis Fitzinger, individual nuclear gene alignments (phased data) and the SALVI et al. | 445 concatenated mtDNA (cytb+nd4) alignment; (b) multilocus included 10 random addition replicates and 1,000 nonpara- Maximum Likelihood (ML) and Bayesian Inference (BI) metric bootstrap replicates, applying the general time‐re- methods, based on a concatenated alignment of unphased versible model with a gamma model of rate heterogeneity nuclear genes (nucDNA) and with the mitochondrial and un- (GTRGAMMA), with individual gene partitions. phased nuclear genes combined (mt‐nucDNA); and (c) the Bayesian Inference (BI) analyses were performed in multilocus coalescent species tree approach based on single BEAST 1.8.0 (Drummond, Suchard, Xie, & Rambaut, 2012). gene data sets for mtDNA and phased nucDNA. Hierophis The input file was built in the BEAUTi utility with models viridiflavus and Hemorrhois algirus were used as outgroups and prior specifications applied as follows (otherwise by in ML analyses following Burbrink and Lawson (2007). default): each gene was used as a partition; nucleotide sub- To test for the effect of missing data in the phylogenetic stitution and relaxed uncorrelated lognormal clock models inference, we performed the phylogenetic analysis with two were unlinked across partitions, except for the clock mod- distinct data sets: (a) with all samples (“complete data set,” els of the mtDNA genes cytb and nd4, which were linked; with 20.7% of missing sequences), and (b) excluding samples the tree model was linked across all partitions; models of that missed data for more than two gene fragments (“reduced nucleotide substitution for each gene partition as selected data set,” with 9.8% of missing sequences; Table 1). by PartitionFinder (reported in Supplementary Table S2); In order to test for the effect of taxon‐set in the phylogenetic Yule process of speciation as tree prior, random starting tree, inference, and to compare our results with previous studies, alpha uniform prior (0, 10), ucld.mean gamma prior (1, 1); we performed multilocus phylogenetic analyses (ML and BI kappa operator (2.0). The xml file was manually edited to analyses based on concatenated loci and species tree) with two account for variability in heterozygous positions by defining different taxon‐sets: (a) the Zamenis taxon‐set, including spe- “ambiguities = true” for the nuclear partitions. BEAST was cies of the genera Zamenis, Rhinechis, Coronella and the out- run three times with 100 million generations each, sampling groups Hierophis viridiflavus and Hemorrhois algirus; and (b) every 10,000 generations. As the Yule process of speciation the Zamenis + ratsnakes taxon‐set adding GenBank data from prior assumes that each terminal taxon represents a distinct New and Old World species of the genera Elaphe, Lampropeltis, species (i.e., requires only one single sequence per species), Pantherophis, Pituophis and Oocatochus (Table 1). The two whereas our data set contains multiple samples per species, in taxon‐sets were used to verify if relationships between the order to inspect the sensitivity of our estimates to the choice western Palaearctic ratsnakes were consistent either excluding of tree prior, we performed an additional run applying the (Zamenis taxon‐set) or including (Zamenis + ratsnakes taxon‐ same settings described above but using only one specimen set) taxa from Asia and the New World. for each species and we obtained identical estimates (). Time‐calibrated species trees were estimated with *BEAST, an extension of the BEAST software imple- 2.3 | Phylogenetic analyses and divergence menting the multispecies coalescent model (Heled & time estimation Drummond, 2010). The ratsnakes fossil record is scarce and Phylogenetic networks were inferred with the Neighbor‐ unevenly distributed in the Palaearctic and offers little help Net algorithm (Bryant & Moulton, 2004) implemented in for calibrating the ratsnakes tree. Most importantly, the phy- SplitsTree4 (Huson & Bryant, 2006) and with the Median‐ logenetic position of extinct ratsnakes is very hard to define Joining algorithm (Bandelt, Forster, & Rohl, 1999) imple- because they show a mixture of characters found in diverse mented in the software Network 5.0.0.0. (available at: http:// extant species and some unique features. In order to assign www.fluxus-engineering.com/sharenet.htm), using in both them to the crown group of some of the extant genera, a de- cases default parameters. tailed phylogenetic analysis including all fossils and extant Partition schemes and models of nucleotide substitu- ratsnakes would be necessary (Massimo Delfino, personal tion for ML and BI phylogenetic analyses were defined by communication). For example, the fossils of the extinct PartitionFinder 2.1.0 (Lanfear, Frandsen, Wright, Senfeld, & species Elaphe praelongissima (Venczel, 1994) have been Calcott, 2016) with the following search parameters: linked tentatively interpreted as remains of the ancestor of Z. lon- branch length, BEAST models, BIC model selection, and gissimus or Z. situla and used in previous dating estimates user schemes with data blocks defined by genes as prelimi- (Lenk et al., 2001), but many features of Elaphe praelon- nary phylogenetic analyses implementing models partitioned gissima are shared with Old World ratsnakes which are dis- by codons positions (1st, 2nd, 3rd or 1st+2nd, 3rd) for the tantly related to Zamenis, such as Elaphe dione Pallas, 1773 protein‐coding genes (cytb, nd4, cmos, dnah3 and prlr) gave (see Venczel, 1994). The same applies for Elaphe algoren- similar results but slower convergence in BI analyses. sis (Szyndlar, 1985), which has been related to the ances- Maximum likelihood (ML) analyses were performed in tor of R. scalaris (Lenk et al., 2001; see also Helfenberger raxmlGUI 1.3 (Silvestro & Michalak, 2012), a graphical & Schulz, 2013). However, Elaphe algorensis displays front‐end for RAxML 7.4.2 (Stamatakis, 2006). Searches most morphological features attributed to both Elaphe and 446 | SALVI et al. 3 3 3 (Continues) MH291187 FJ627909 MH291191 MH291197 MH291189 MH291190 MH291185 MH291186 MH291193 MH291194 MH291195 MH291196 vim MH291184 MH291188 KM870881 - - MH291198 MH291183 - MH291177 MH291178 MH291179 MH291180 MH291181 MH291182 KM870880 - 3 3 3 3 3 3 3 3 MH291169 - - MH291176 MH291171 MH291172 MH291167 MH291168 MH291173 - MH291174 MH291175 sptbn1 MH291166 MH291170 KY495573 KY495577 KY495581 KY495588 KY495591 KY495590 - - KM870820 FJ627921 - - MH291165 - 3 MH291154 - MH291158 MH291163 MH291156 MH291157 MH291152 MH291153 MH291160 - MH291161 MH291162 prlr - MH291155 - MH291148 MH291149 - - MH291150 - - - - - MH291164 LN551943 MH291151 - - MH291129 MH291132 MH291127 MH291128 MH291125 MH291126 - - - MH291131 dnah3 MH291124 - - MH291121 - MH291122 ------MH291133 - MH291123 3 3 3 3 3 3 3 3 3 3 3 3 3 AY486935 AF471113 MH291098 MH291102 MH291103 MH291109 MH291100 MH291101 KY495516 KY495520 KY495524 KY495529 KY495526 KY495531 DQ902075 MH291096 DQ902083 MH291097 DQ902098 MH291105 AY486954 MH291106 MH291107 MH291108 AY376803 cmos MH291095 MH291099 MH291094 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 MH291120 MH291118 MH291112 JQ837597 KF639748 MH291116 EU022668 JX315468 MH291117 HQ392561 HQ392562 HQ392564 HQ392567 KF639746 KF639747 HQ392568 DQ902125 DQ902137 AY486930 JQ837595 MH291119 LN551980 MH291114 MH291115 cytb MH291111 MH291113 EU022663 MH291110 3 3 3 3 3 3 3 3 3 3 3 3 3 MH291147 DQ902320 - AY487037 MH291138 AY487065 AY487066 MH291146 KY495534 MH291136 DQ902297 DQ902303 MH291142 - LN552062 MH291137 MH291143 KY495545 MH291140 MH291141 nd4 MH291135 MH291139 MH291145 KY495542 KY495560 KY495570 KY495565 MH291134 Coordinates 32.221 / -4.676 - 31.208 / -7.860 32.262 / -5.146 41.333 / 41.640 - 41.350 / -7.783 42.316 / 0.733 46.779 / 11.644 40.836 / 17.080 - - 41.318 / -6.398 41.027 / -8.645 - 39.941 / 20.676 42.241 / 13.021 39.514 / -5.347 41.103 / -7.964 Geographic 36.647 / 53.298 39.484 / 43.987 41.027 / -8.645 42.458 / 13.929 40.519 / 16.063 37.337 / 13.424 39.189 / 16.199 36.565 / 53.058 Morocco Morocco Morocco Italy - Imellalen, Borçka, Turkey - Alvão, Portugal Sopeira, Portugal Bressanone, Italy Putignano, Italy - - Fermoselle, Spain Espinho, Portugal Oukaimeden, - Aristi, Greece Michelifene, Italy Guadalupe, Spain Resende, Portugal Sampling locality Neka, Iran A ǧ ri, Turkey Espinho, Portugal Penne, Italy Varco Sabino, Pietrapertosa, Italy Torre Salsa, Italy Potame, Italy Sari, Iran 2 2 2 2 2 2 GB3 GB4 DB1562 GB2 1591 GB5 DB2686 DB2689 DB1725 z106 SF01 SIT1 DB1525 EL304 GB1 DB22 DB38 DB41 Sample ZPF 1589 DB4002 z85 z93 z204 z68 z99 SARI DB4003 1 1 1 Code, location coordinates and GenBank accession numbers of the colubrid specimens used in this study Zamenis hohenackeri Hierophis viridiflavus Hemorrhois algirus Rhinechis scalaris Rhinechis scalaris Zamenis situla Zamenis hohenackeri Coronella austriaca Coronella girondica Coronella girondica Coronella girondica Zamenis longissimus Zamenis situla Rhinechis scalaris Zamenis situla Hemorrhois algirus Hierophis viridiflavus Zamenis persicus Species Zamenis persicus Zamenis hohenackeri Coronella austriaca Zamenis longissimus Zamenis longissimus Zamenis lineatus Zamenis lineatus Zamenis lineatus Zamenis persicus Coronella austriaca TABLE 1 SALVI et al. | 447 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 JX648636 MH291192 - - - - KF215529 KF215534 FJ627886 KF215552 FJ627882 - KF669227 KF669226 KM870878 FJ627888 KM870871 KF669217 KF669216 KF669221 KF669220 - FJ627902 FJ627897 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 - - FJ627930 KM655046 KF669171 KF669172 KM870839 - JX648617 KF669162 KF669161 KF669166 KF669165 KM870834 - FJ627938 - FJ627939 FJ627924 KF215129 KF215126 KF215135 KF215130 KF215163 3 3 3 3 3 KF215308 - KF215244 KF215240 MH291159 KF215239 KF215243 ------MH291130 ------3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 - - DQ902091 - - FJ627805 AF471140 KX694817 AY486953 KR814679 AF471151 FJ627803 MH291104 AY486955 JN799416 JN799415 DQ902087 EF076705 DQ902081 FJ627801 FJ627798 AF435015 FJ627790 KX694814 3 3 3 3 3 3 3 3 3 3 3 3 3 3 Sequences downloaded from GenBank. 3 3 3 3 3 3 3 3 3 AY739641 DQ902311 AY739644 KM655218 DQ902296 AY739642 KX694869 KR814695 LC105628 AY487067 JN799414 JN799413 DQ902305 AH015912 DQ902301 FJ627850 - - NC022146 JF308311 AF138766 AF138765 AY486929 LN551964 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 KF216143 KF728020 KX881114 DQ538344 DQ538343 KF216244 DQ902123 KF216289 KX835885 JQ798793 AF337064 - KR814726 NC030041 AF337065 MH291144 KF669244 KF669243 KF669248 KF669247 KF728019 KX694867 FJ627818 AY487062 Samples excluded in the reduced data set analyses; 2 ------41.883 / 15.433 ------Lesina, Italy ------GB21 GB10 GB26 GB27 GB28 GB22 GB16 GB25 GB18 GB17 GB24 GB7 GB8 GB9 GB23 DB2291 GB12 GB13 GB14 GB15 GB11 GB19 GB20 GB6 (Continued) triangulum calligaster obsoletus obsoletus triangulum rufodorsatus calligaster rufodorsatus This species is designated as Zamenis scalaris following this study; Elaphe quatuorlineata Lampropeltis Pituophis catenifer Lampropeltis ruthveni Lampropeltis Pantherophis Pantherophis Ptyas korros Ptyas mucosa Ptyas mucosa Lampropeltis Lampropeltis ruthveni Hierophis viridiflavus Elaphe carinata Elaphe carinata Elaphe taeniura Elaphe taeniura Oocatochus Elaphe quatuorlineata Pituophis deppei Pituophis deppei Lampropeltis Ptyas korros Oocatochus 1

a TABLE 1 448 | SALVI et al. TABLE 2 Results of topological tests Test Constraint References AU SH comparing the topology of our phylogenetic A ((((Z. lin, Z. lon) (Z. sit, R. sca)) Pyron et al. (2011, 2013), 0.025 0.041 tree (((((Z. lin, Z. lon) Z. sit) Z. per) Z. hoh) Z. per) Z. hoh) Zheng & Wiens (2016) R. sca) with estimates from previous studies B ((((Ptyas korros, Z. lon) Z. lin) Figueroa et al. (2016) 7e‐0.005 0 Z. sit) R. sca) (Z. per, Z. hoh) C (((((Z. lin, Z. lon) Z. sit) Z. per) 0.117 0.161 R. sca) Z. hoh) D (((((Z. lin, Z. lon) Z. sit) R. sca) 0.197 0.207 Z. per) Z. hoh) E (((((Z. lin, Z. lon) Z. sit) Z. hoh) 0.06 0.097 R. sca) Z. per) F (((((Z. lin, Z. lon) Z. sit) R. sca) 0.217 0.228 Z. hoh) Z. per) aNote. Test A and Test B used topological constrains based on highly supported relationships recovered by previ- ous studies (see Figure 2): Test A—Rhinechis nested within Zamenis and sister to Z. situla as recovered by Pyron et al. (2011, 2013) and Zheng and Wiens (2016); Test B—Rhinechis and Ptyas korros nested within Zamenis and Ptyas korros sister to Z. longissimus as recovered by Figueroa et al. (2016). Tests C to F used topo- logical constrains based on different phyletic order between R. scalaris, Z. hohenackeri and Z. persicus in order to compare our best tree with alternative topological hypotheses of relationships between Rhinechis and Zamenis. The constrained relationships and p‐value results of Approximately Unbiased (AU) and Shimodaira‐Hasegawa (SH) tests are reported; significant p‐values are in bold (Z. lin: Zamenis lineatus, Z. lon: Zamenis longissimus, Z. sit: Zamenis situla, Z. hoh: Zamenis hohenackeri, Z. per: Zamenis persicus, R. sca: Rhinechis scalaris).

Coluber Linnaeus, 1758 (Szyndlar, 1985), which makes the (Guicking & Lawson, 2006; Nagy, Lawson, Joger, & Wink, assignment of Elaphe algorensis as an ancestor to R. sca- 2004; Pyron & Burbrink, 2009; Ruane, Torres‐Carvajal, laris unreliable. Therefore, due to the lack of internal cali- & Burbrink, 2015). We implemented these rate priors in bration points in Zamenis and the general difficulty in the *BEAST by defining a lognormal distribution for the mi- attribution of fossils to stem groups of extant ratsnake gen- tochondrial ucld.mean parameter, with mean 0.0134 and era, we recurred to literature for rates of evolution estimated standard deviation 0.0035 for Cal I and mean 0.00815 and for the genes cytb and nd4 in colubrids. The review of avail- standard deviation 0.0010 for Cal II. Model and prior spec- able literature indicated that rates of evolution estimated for ification were as follows (otherwise by default): the nucle- colubrids fall under two classes: fast (~1.3% substitutions otide substitution models were unlinked across loci and per million years, s/my) and medium (~0.8% s/my) rates. implemented for each gene as selected by PartitionFinder Therefore, we performed time estimation analyses with two (Supplementary Table S2), relaxed uncorrelated lognormal different calibration rates. In calibration I (Cal I) we used a clock models and tree models were unlinked, with the ex- fast rate of evolution of 1.34% s/my (with 95% CI = 0.99– ception of the clock and tree models of the mitochondrial 1.70%) as estimated by Daza, Smith, Páez, and Parkinson genes cytb and nd4 which were linked; Yule process of spe- (2009) and similar to rates applied in previous papers on ciation as tree prior, random starting tree; alpha uniform (0, Serpentes (e.g., Carranza, Arnold, & Pleguezuelos, 2006; 10); ucld.mean of nuclear genes Gamma (1, 1); operator Lenk et al., 2001; Mezzasalma et al., 2015). In calibration II kappa (2.0). *BEAST was run three times with 500 million (Cal II) we used a medium rate of evolution of 0.815% s/my generations, sampling every 50,000 generations. as implemented by Bryson, de Oca, and Velasco (2008) and All BEAST and *BEAST runs were performed on CIPRES similar to other studies on Serpentes based on New World Science Gateway 3.3. (available at: http://www.phylo.org/). snake fossils (Myers et al., 2013; Ruane, Bryson, Pyron, & We used Tracer 1.6 (Rambaut, Suchard, & Drummond, 2014) Burbrink, 2014). The standard deviation for this rate was not to check runs for convergence (burn‐in = 10%) and to ensure available in Bryson et al. (2008), thus we decided to apply a that all effective sample size (ESS) parameters were higher range from 0.7 to 0.9%, which also includes estimated rates than 200, as recommended in the software’s manual. Tree of evolution close to 0.8% which were applied previously files were combined with LogCombiner and TreeAnnotator

FIGURE 3 Neighbor‐Net haplotype networks for Zamenis, Rhinechis, Coronella, Hierophis and Hemorrhois species inferred from single‐ locus DNA sequences data: mitochondrial (concatenated cytb and nd4) and individual nuclear loci (cmos, dnah3, prlr, sptbn1 and vim). Circles represent different haplotypes. Values next to lines represent bootstrap support [Colour figure can be viewed at wileyonlinelibrary.com] SALVI et al. | 449 mtDNA cmos

dnah3 prlr

sptbn1 vim

Z. lineatus Z. longissimus Z. situla Z. persicus Z. hohenackeri R. scalaris C. austriaca C. girondica H. viridiflavus H. algirus 450 | SALVI et al. (a) (b)

(c)

FIGURE 4 Phylogenetic relationships between ratsnakes based on concatenated mitochondrial (cytb, nd4) and nuclear (cmos, dnah3, prlr, sptbn1 and vim) DNA sequences: (a) Bayesian tree of the Zamenis taxon‐set; (b) Bayesian tree of the Zamenis + ratsnakes taxon‐set. (c) Species tree of the Zamenis taxon‐set with the time axis, in million years, for the calibrations I (Cal I) and II (Cal II) represented below the tree; the age of the nodes and the associated 95% Highest Posterior Density intervals are presented in . Values above nodes represent Bayesian posterior probabilities (BPP) ≥ 0.9; values below nodes represent maximum likelihood bootstrap support (BS) values ≥70. Node support for intraspecific clades is represented by black circles: BPP ≥ 0.98 (upper half) and BS ≥ 95 (bottom half)

(both included in the BEAST package) was used to calculate 2.4 Topology tests the maximum clade credibility tree (MCC) summarizing the | posterior distribution of tree topologies and branch lengths. We performed topological tests in order to compare our phy- All trees were visualized with FigTree 1.4 (available at http:// logenetic tree with estimates from previous studies, which tree.bio.ed.ac.uk/software/figtree/). inferred the following phylogenetic relationships with the

FIGURE 5 Biogeographical reconstruction of the evolutionary history of Zamenis and Rhinechis according to rase and RASP analyses based on the species tree applying the time calibration I. (a) biogeographical reconstruction according to the rase analysis at eight evenly spaced time slices from 8 to 1 Ma. The 5%, 10%, and 15% highest posterior density is plotted for each extant branch. Top‐right: posterior distributions of the dispersal rate parameters (σ2) for latitude, in blue, and longitude, in pink, represented in degrees/Kyrs. (b) Biogeographical reconstruction according to the RASP analyses based on four areas (A, B, C, D as represented in the top‐left box): coloured pie charts in correspondence of the nodes represent results from ancestral reconstruction conducted with S‐DIVA (top) and BBM (bottom); * represent areas estimated by BBM alone [Colour figure can be viewed at wileyonlinelibrary.com] SALVI et al. | 451

(a)

a e

b f

c g

d h

(b) 452 | SALVI et al. maximum statistical support: (a) Rhinechis nested within converged to the posterior distribution with the coda package Zamenis and sister to Z. situla, Test A (Pyron et al., 2011, (Plummer, Best, Cowles, & Vines, 2006). We applied evenly 2013; Zheng & Wiens, 2016); (b) Rhinechis and Ptyas korros spaced time slices to infer the location of the ancestral nodes. nested within Zamenis and Ptyas korros sister to Z. longis- The rase and coda packages were run in R (R Core Team, simus, Test B (Figueroa et al., 2016). For Test B, we down- 2017) using the interface RStudio 1.0.153. loaded sequences of P. korros and P. mucosa Linnaeus, 1758 RASP analyses were performed in RASP 3.0 (Yu et al., available for the same genes applied in this study (Table 1). 2015) using the statistical dispersal‐vicariance analysis (S‐ These sequences were added to our Zamenis taxon‐set and DIVA; Yu, Harris, & He, 2010) and the Bayesian binary used to infer a ML mt‐nucDNA tree, which was compared MCMC (BBM; Yu, Harris, & Xingjin, 2011). The distri- with the ML tree enforcing the topology obtained by Figueroa bution range of Zamenis and Rhinechis was divided into et al. (2016). four main ranges: South‐western Asia (Caspian, Caucasus Additionally, given the low support for the relationships and Anatolia), the Balkan Peninsula, the Italian Peninsula, between R. scalaris, Z. hohenackeri and Z. persicus, and and western Europe, and each species was attributed to one between them and the remaining Zamenis species, we also or more areas where it is present (see Figure 1 for species tested for significant differences between our best tree and distribution). Besides the four‐area analyses, we performed four alternative topological hypotheses with varying phyletic additional runs with three or five areas and different areas order between R. scalaris, Z. hohenackeri and Z. persicus assemblages and we obtained consistent results. We used (Tests C‐F; see Table 2 for the alternative topologies tested). all the post‐burn‐in trees and the MCC tree resultant from The topological constrains (Table 2) were built in Mesquite the species tree run as input. S‐DIVA was run using 1,000 3.2 (Maddison & Maddison, 2003) and the per‐site log likeli- trees randomly sampled from the input trees and the BBM hoods were estimated in raxmlGUI 1.3. The constrained trees analyses were conducted with the Jukes‐Cantor model, site were compared with our best ML tree using the Shimodaira‐ variation set to equal and with two simultaneous runs with 5 Hasegawa (SH) and the approximately unbiased (AU) tests million generations, sampling each 100th generation. (Shimodaira, 2002; Shimodaira & Hasegawa, 1999; re- spectively), as implemented in Consel 0.2 (Shimodaira & Hasegawa, 2001) to determine if the constrained topology 3 | RESULTS could be rejected at the p < 0.05 level. We generated a total of 105 new sequences which were de- posited in GenBank (Table 1). Length of individual genes and 2.5 | Biogeographical reconstruction the number of variable sites are presented in Supplementary Biogeographical analyses were performed with the programs rase Table S2. Sequence alignments of the protein‐coding genes (Quintero, Keil, Jetz, & Crawford, 2015) and RASP (Yu, Harris, cytb, nd4, cmos and dnah3 did not require gaps; the alignment Blair, & He, 2015). To perform the biogeographical analyses, we of prlr required an insertion of 18 base pairs. The translation built a species tree including only Zamenis and Rhinechis species into amino acids of these genes did not contain stop codons. (i.e., without outgroups), applying the same procedure and set- The phi test did not find statistically significant evidence for tings as described above. Results from biogeographical analyses recombination in any gene fragment (p > 0.05). are not affected by the calibration used to build the species tree, given that species trees recovered with either Cal I or II had iden- tical topologies and branch lengths and rase time slices “cut” the 3.1 | Phylogenetic relationships and time of tree at the same points in both time‐trees. divergence estimates The rase method allows using the whole distribution Haplotype network reconstructions for both nuclear and range of species (instead of single localities) without an a‐ mitochondrial loci recovered equivalent relationships priori definition of areas to infer the geographical location between Neighbor‐Net and Median‐Joining analyses of the ancestors in a Bayesian framework. The geographical (Figure 3 and , respectively). The least variable genes were distributions of these species were extracted from the IUCN the nuclear cmos and prlr, which showed the lowest num- Red List of Threatened Species (IUCN, 2017; Figure 1); we ber of haplotypes and mutational steps between the ingroup assumed no correlation between dispersal rates in longitude species. The most variable loci were mtDNA and the nu- and latitude. We ran rase for 10,000 iterations, discarded the clear sptbn1 and vim. Coronella species are distinct from first 1,000 as burn‐in, logged every 10th iteration and ob- Rhinechis and Zamenis in the mtDNA and fast‐evolving tained the posterior distributions of ancestral nodes and mi- nuclear genes sptbn1 and vim, whereas these genera are not gration rates (σ2). We plotted the trace to evaluate the MCMC well sorted at slower‐evolving nuclear genes. Rhinechis is results, estimated the mean and posterior densities for each of always closely related with, or nested deep within Zamenis the estimated parameters and confirmed that the algorithm species. Zamenis species are closely related among them SALVI et al. | 453 but form a paraphyletic assemblage in the haplotype net- a sister relationship between Z. longissimus and Ptyas korros works of most of the loci analysed (mtDNA, cmos, prlr, with R. scalaris nested within Zamenis (Test B), were rejected sptbn1 and vim). There was no haplotype sharing between by the SH and AU tests (Table 2). On the contrary, varying species, except in the nuclear gene vim, where Z. situla and the branching order between R. scalaris, Z. hohenackeri and Z. lineatus share a single haplotype. Z. persicus resulted in topologies which were not statistically Phylogenetic relationships recovered by multilocus rejected by the SH and AU tests (Tests C – F; Table 2). analyses were consistent between methods (ML and BI; Figure 4a,b), data sets (concatenated nucDNA and mt‐nu- cDNA data sets; Supplementary Figure S3 and Figure 4a, 3.3 | Biogeographical reconstruction respectively) and approaches (based on the concatenation The results from the biogeographical reconstruction with and the species tree approaches; Figure 4a,c) and with any rase showed a remarkable longitudinal migration throughout of the taxon‐sets used (the Zamenis taxon‐set, Figure 4, and the evolutionary history of Zamenis and Rhinechis, whereas the Zamenis + ratsnakes, Figure 4b and and Supplementary the latitudinal migration was relatively low (Figure 5a). 2 Figures S6 and S7). Moreover, results from the phyloge- Indeed, the estimated longitudinal dispersal rate, x was high netic analyses based on the Zamenis taxon‐set either with (mean: 47.8604 degrees2/Myr, 95% HPD: 29.5124–77.4254) 2 the complete data set (Figure 4a) or the reduced data set and the latitudinal dispersal rate, y was comparatively low (Supplementary Figures S3–S5) were identical at supported (mean: 0.4667 degrees2/Myr, 95% HPD: 0.3199–0.6882). nodes, suggesting that adding taxa with a higher proportion According to the rase results, the location of the ancestor of missing data did not affect phylogenetic estimates. of Rhinechis and Zamenis was distributed in an area pres- The genus Coronella is monophyletic (Bayesian Posterior ently covering the Aegean region in the southern Balkan Probabilities, BPP = 1, Bootstrap Support, BS > 90; Peninsula (Figure 5a), at a latitudinal mean (latM) of 39.3020 Figure 4) and sister to the clade formed by Rhinechis and and a longitudinal mean (lonM) of 22.2972 (95% HPD val- Zamenis species. The latter is well supported in all the phy- ues of latitude and longitude for the ancestral locations are logenetic analyses (BPP > 0.98, BS > 90) and indicates presented in Supplementary Table S5). The ancestor of the a topology with R. scalaris sister to all Zamenis species. lineage leading to R. scalaris underwent a westward migra- However, the Zamenis clade is not supported in any phylo- tion (Figure 5b,c; purple), while the ancestor of all the other genetic analyses (BPP < 0.90, BS < 70), with all the trees species of Zamenis migrated eastwards (Figure 5b). The split showing a short internode between R. scalaris and the clade between the ancestor of Z. hohenackeri (Figure 5c; yellow) formed by all the other Zamenis species. Zamenis persicus, and the ancestor of Z. persicus, Z. situla, Z. longissimus and Z. situla, Z. longissimus and Z. lineatus form a relatively Z. lineatus (Figure 5c; dark green) occurred in the current well‐supported clade (BPP > 0.94, BS > 76). Within this Anatolia (latM: 39.0857, lonM: 28.5675), like the subsequent clade, Z. persicus is recovered as sister to the remaining split between Z. persicus (Figure 5e, pink) and the ancestor species (BPP > 0.93, BS > 98), and Z. longissimus and of Z. situla, Z. longissimus and Z. lineatus (latM: 38.6477, Z. lineatus form a highly supported clade in all the analyses lonM: 33.3818). The divergence of Z. situla (Figure 5g, (BPP > 0.91, BS > 91). blue) occurred in present western Anatolia (latM: 39.6099, Divergence time estimates based on Cal I and Cal II are lonM: 25.2771) and the divergence between Z. longissimus reported in the species tree (Figure 4c and Supplementary (Figure 5g, red) and Z. lineatus (Figure 5g, green) occurred Figure S7) and in Supplementary Tables S3 and S4 (along in an area presently including the south‐western Balkan with 95% High Posterior Density intervals, 95% HPD). The Peninsula and the westernmost part of Anatolia (latM: Time to the Most Recent Common Ancestor (TMRCA) be- 39.5779, lonM: 23.1189). By the time of the divergence tween Rhinechis and Zamenis is estimated in the Middle between Z. lineatus and Z. longissimus the species Z. ho- Miocene according to Cal II [~13 Million years ago (Ma)] henackeri (Figure 5g, yellow) and Z. persicus (Figure 5g, or in the Late Miocene according to Cal I (~7 Ma). The split pink) would had migrated towards the eastern Anatolia and of Z. hohenackeri is estimated about a million year later Caucasus regions (latM: 38.9985, lonM: 35.1886, and latM: according to both calibrations. The following cladogenetic 37.6885, lonM: 41.6369, respectively), while R. scalaris events leading to the remaining Zamenis species are esti- (Figure 5g, purple) had migrated west towards the Iberian mated during the Late Miocene (~9–5 Ma) or in the Pliocene Peninsula (latM: 39.9817, lonM: 2.8080). (~5–3 Ma) according to Cal II and I, respectively. The results from the RASP analyses suggest an eastern origin of the Zamenis clade (Figure 5b). The basal nodes within the clade leading to the split of the Z. hohenackeri and 3.2 | Topology tests Z. persicus lineages were estimated either in south‐western The phylogenetic hypotheses from previous studies of a sister Asia, as inferred the BMM method (>90%), or shared by the relationship between R. scalaris and Z. situla (Test A), and of south‐western Asia, Balkans and the Italian Peninsula, as 454 | SALVI et al. inferred the by S‐DIVA method (98%). The ancestral area of with the MP method. Indeed, when using the Minimum the lineage including Z. situla, Z. longissimus and Z. lineatus Evolution method Utiger et al. (2002) found Rhinechis sister was either shared by the Balkans and the Italian Peninsula to Zamenis (not shown in Figure 2). The phylogeny by Lenk or these areas plus south‐western Asia according to S‐DIVA et al. (2001) recovered R. scalaris early branching as sister to and BBM. The ancestor of Z. lineatus and Z. longissimus was all ingroup taxa analysed (Elaphe and Zamenis) and a sup- located in the Italian Peninsula according to S‐DIVA (100%), ported Zamenis clade sister to Oreocryptophis porphyraceus whereas it is highly uncertain in the BBM inference. Finally, Cantor, 1839 (formerly Elaphe porphyracea). it remains undetermined the ancestral area for the root node However, these results are clearly due to limited taxon Zamenis + Rhinechis, with S‐DIVA inferring a shared origin sampling and the unfortunate use of Lampropeltis triangulum between all areas (100%) while BBM estimated with equal as an outgroup for Palaearctic ratsnakes. The addition of New probability either the easternmost area (south‐western Asia) and Old World ratsnakes to the phylogenetic analyses clearly or the westernmost area (western Europe) as probable ances- showed that Lampropeltis triangulum and other Lampropeltini tral areas. are nested within the clade including Palaearctic ratsnakes (Figure 4b and Supplementary Figures S6 and S7), as de- scribed in further studies (Burbrink & Lawson, 2007; Chen 4 | DISCUSSION et al., 2017). As to the high nodal support recovered for the tree by Lenk et al. (2001), and specifically for the Zamenis In this study, we obtained a fully resolved and well‐sup- clade, these authors used MP bootstrapping using character ported phylogeny of western Palaearctic ratsnakes, which is weights derived from the successive weighting approach. consistent across markers, taxon‐sets, phylogenetic methods This procedure inflates bootstrap values as it is likely that the and incorporates coalescent models accounting for gene tree/ support values will tend to increase when incongruent char- species tree conflicts (Figure 4 and Supplementary Figures acters are down‐weighted and characters supporting that spe- S3–S7). Smooth snakes of the genus Coronella form a sis- cific topology are over‐weighted (Cummins & McInerney, ter clade to Rhinechis and Zamenis ratsnakes (Figure 4). 2011). By adding an appropriate outgroup (Hemorrhois algi- The close phylogenetic relationships between Rhinechis and rus and Hierophis viridiflavus instead of Lampropeltis trian- Zamenis is well‐established. However, the phyletic order gulum) to the data set of Lenk et al. (2001), and repeating the between the early branches within this clade, R. scalaris bootstrap analysis using an unweighted matrix, we recovered and Z. hohenackeri, remains poorly supported. The phylo- a tree which is congruent with results from this study and genetic relationships between Z. persicus, Z. situla, Z. lon- other previous studies: that is Rhinechis is sister to Zamenis, gissimus and Z. lineatus are fully resolved, with the Persian with the clade Zamenis not supported, and Oreocryptophis ratsnake Z. persicus sister to the Mediterranean species porphyraceus is sister to Elaphe. Z. situla, Z. longissimus and Z. lineatus, and among them, More difficult to conciliate are the discrepancies between Z. situla branching off earlier than the split between the two the phylogenetic relationships obtained in our study and Aesculapian snakes Z. longissimus and Z. lineatus, which early studies discussed above (Burbrink & Lawson, 2007; form a highly supported clade. Lenk et al., 2001; Utiger et al., 2002; Figures 2a–c and 4) with those recovered by the analyses of supermatrices with hundreds of squamate taxa (Figueroa et al., 2016; Pyron 4.1 | Resolving phylogenetic controversies Zamenis Rhinechis et al., 2011, 2013; Zaher et al., 2012; Zheng & Wiens, 2016; concerning and Figure 2d–f). All these latter mega‐phylogenies recovered The phylogenetic relationships between Zamenis and Rhinechis deeply nested within Zamenis, with maximum sup- Rhinechis inferred in this study are in line with those es- port (99–100), while showing conflicting results on the rela- timated by early studies mainly based on mtDNA data tionships between Zamenis species (Figure 2d–g). Moreover, (Burbrink & Lawson, 2007; Lenk et al., 2001; Utiger et al., the phylogeny by Figueroa et al. (2016) shows a paraphyletic 2002; Figure 2a–c), with a few discrepancies which are easy Zamenis clade, which included not only Rhinechis but also to reconcile. Our topology is fully congruent with the tree of the Indo‐Chinese ratsnake Ptyas korros (Figure 2f). These Burbrink and Lawson (2007) based on data from mtDNA and phylogenetic hypotheses are strongly rejected by topological the nuclear cmos gene, although in this latter study the nodal tests based on our data (Table 2). On the other hand, it is support was low and the Zamenis taxon‐set was not com- unlikely that incongruences between early phylogenetic esti- plete. The close relationship between Rhinechis and Zamenis mates (Figure 2a–c) and recent mega‐phylogenies based on was not retrieved in the Maximum Parsimony (MP) trees by supermatrices (Figure 2d–g) can be explained by differences Lenk et al. (2001) and by Utiger et al. (2002) (Figure 2). The in molecular data between studies. Indeed, although super- different results from these latter studies are probably due matrices are built with dozens of mitochondrial and nuclear to the use of small mitochondrial fragments in combination markers—up to 52 genes in Zheng and Wiens (2016)—in SALVI et al. | 455 each data set the representation of these genes in Zamenis The biogeographical reconstruction from rase attributes the and Rhinechis is limited to sequences of the same four‐five origin of the ancestor of Rhinechis and Zamenis to the Aegean mtDNA genes (differently distributed across studies) and region (Figure 5a). The Aegean area has a complex geologi- the slow evolving nuclear gene cmos generated by those cal history since the late Tertiary (summarized in Poulakakis early molecular studies. A compelling evidence that the et al., 2015). From a paleogeographic point of view, one of the nodes estimated by the supermatrix approach are incorrect major geologic events affecting this area was the formation comes from a recent phylogenomic study on ratsnakes that of the mid‐Aegean trench in the middle Miocene, from 12 to was published at the time of completion of our study (Chen 9 Ma, which caused the west‐east split of the former Aegean et al., 2017). The phylogenomic tree of ratsnakes by Chen landmass (Creutzburg, 1963; Dermitzakis & Papanikolaou, and colleagues based on over 300 loci shows identical phy- 1981). The formation of the mid‐Aegean trench acted as a logenetic relationships between Zamenis and Rhinechis as vicariant agent for the separation of lineages present in the inferred in our study, with R. scalaris and Z. hohenackeri as area at the time, including vertebrates and invertebrates genera early branches of the clade and sister to the remaining four (see Poulakakis et al., 2015 for a review on phylogeograph- Zamenis species, whose relationships are well resolved and ical studies on the Aegean region). Likewise, in a scenario highly supported (see Figure 4 and Chen et al., 2017). The of Aegean origin of Zamenis and Rhinechis, the divergence bad performance of the supermatrix approach in estimating between these genera could be the result of the vicariance relationships within terminal groups was documented also mediated by the formation of the mid‐Aegean trench and the for lacertid lizards (Mendes, Harris, Carranza, & Salvi, east‐west division of the region. This is in agreement with es- 2016; see also Mendes, Salvi, Harris, Els, & Carranza, 2018 timated time of divergence between these genera (about 7 or for discussion on the overestimates of divergence times by 13 Ma according to Cal I or Cal II respectively; Figure 4c). the supermatrix approach). This is probably due to the fact According to the biogeographical reconstruction, these that the analysis of supermatrices with thousands of termi- early cladogenetic events were followed by the eastward dis- nals require high‐speed approximations of tree topology persal of the ancestors of Z. hohenackeri and Z. persicus and searches, substitution models and nodal support as well as by the westward dispersal of the ancestors of Rhinechis and outgroups that are excessively distant from the tips of the remaining Zamenis species during Late Miocene (Figure 5a). tree (Mendes et al., 2016). Overall these findings indicate Land connections were available at the time for these dis- that while the supermatrix approach with hundreds of taxa persal events, both towards western Europe and towards and high levels of missing data is prone to recover wrong Anatolia, due to the existing land connections between this relationships and spurious support, a set of fast‐evolving nu- region and the eastern Aegean (Çaǧatay et al., 2006; Elmas, clear markers such as the one used in this study is enough 2003; Melinte‐Dobrinescu et al., 2009). The Anatolian and to recover the same relationships as a species tree based on Caucasus regions have been affected by the orogenic activity hundreds of loci. leading to the formation of the Anatolian mountain chains (including the Anatolian Diagonal, Taurus and Black Sea Mountains) as well as the Central Anatolian Plateau and the 4.2 | Evolutionary Central Anatolian lake system, and the uplift of the Greater history and biogeography of Zamenis and Rhinechis and Lesser Caucasus. The formation of these mountain chains, ratsnakes in combination with past habitat changes produced by cli- The ancestral area inferred by biogeographical analyses sug- matic oscillations between dryer and wetter conditions, have gests an eastern origin of Zamenis ratsnakes, with the ances- created significant barriers to gene flow (Eronen et al., 2009; tral areas of the basal nodes inferred in Anatolia and Balkan Oberprieler, 2005; Popov et al., 2006; Rögl, 1998). Therefore, peninsulas or south‐western Asia and western areas inferred it is plausible that the climatic and geological changes occur- for the following nodes (Figure 5). Such a scenario of east‐ ring in this area have been responsible for the divergence of to‐west diversification is consistent with the phyletic pattern Z. hohenackeri and Z. persicus, as suggested for many genera of diversification of Zamenis, with early diverging species of vertebrates (e.g., Kapli et al., 2013; Kornilios et al., 2011; having an eastern distribution (Z. hohenackeri and Z. persi- Skourtanioti et al., 2016; Weisrock, Macey, Ugurtas, Larson, cus) and species with a western distribution splitting more & Papenfuss, 2001) and invertebrates (e.g., Micó, Sanmartín, recently (Z. situla, Z. longissimus and Z. lineatus). An east- & Galante, 2009) in previous studies. ern origin of Rhinechis and Zamenis, and of the Palaearctic The divergence of the western Zamenis species occurred, ratsnakes in general, followed by dispersal towards western according to the ancestral reconstruction analyses, around Europe in the case of Rhinechis and Zamenis, had been sug- western Anatolia and southern Balkans for Z. situla and in gested by previous molecular studies on ratsnakes (Burbrink the Italian Peninsula between Z. longissimus and Z. lineatus & Lawson, 2007; Helfenberger & Schulz, 2013; Lenk et al., (Figure 5). The location of these splits implicates a dispersal 2001; Utiger et al., 2002). event of the ancestor of these three species from Anatolia/ 456 | SALVI et al. Caucasus to the Balkans and western Europe. Such dispersal of Z. situla from continental Greece, Aegean islands and events have been described for plants (Lo Presti & Oberprieler, Turkey revealed the existence of two distinct clades that 2009), warblers (Blondel, Catzeflis, & Perret, 1996), bush diverged in the Pleistocene: one from Crete, Thera and the crickets (Chobanov, Kaya, Grzywacz, Warchalowska‐ Peloponnese and the other from Turkey, northern continental Sliwa, & Ciplak, 2017; Kaya & Çiplak, 2017) and lizards Greece and the eastern Aegean island Samos (Kyriazi et al., (Ahmadzadeh et al., 2013). The connection between Anatolia 2013). It remains unclear whether the distribution of Z. sit- and the Balkan has been intermittent during the last millions ula in the Aegean area is a result of over‐sea dispersal from of years, particularly at the Bosphorus—Marmara Sea— mainland Greece or if it was mediated by humans (Kyriazi Dardanelles. Land bridges between these landmasses have et al., 2013; Poulakakis et al., 2015). Phenotypic and genetic been established during the Messinian Salinity Crisis, and re- data on Z. lineatus suggested reduced diversity and indicated peatedly during the Quaternary (Çaǧatay et al., 2006; Elmas, a wide area of introgression with Z. longissimus at the eastern 2003; Gökaşan et al., 1997; Melinte‐Dobrinescu et al., 2009). contact zone between these species (Salvi et al., 2017). The The divergence time estimation for the split of Z. situla sup- pattern of introgression seems asymmetric, with a prevalent ports a dispersal of the ancestor of Z. situla, Z. longissimus introgression of Z. lineatus traits into a Z. longissimus back- and Z. lineatus towards the west either during the Pliocene ground. Whether such a pattern resulted from a scenario of (according to Cal I) or during the late Miocene (according to demographic and geographic range expansion of one species Cal II), followed by the split between Z. lineatus and Z. lon- into the range of the other is a question that remains unan- gissimus in the Italian Peninsula (Figures 4c and 5; Table 2). swered pending further studies. The climatic changes associated to the glacial/intergla- cial cycles during the Pleistocene have played an important Zamenis Rhinechis role in shaping the current distribution of extant Rhinechis 4.3 | Systematics of and : and Zamenis species and of their genetic diversity, as indi- one or two genera? cated by recent studies. Over 35 fossil records in the Iberian From a phenotypic point of view, R. scalaris shares several mor- Peninsula assigned to Rhinechis support the establishment phological features with Zamenis species (Schulz, 1996). The of this species in its current distribution since the Pliocene head scalation pattern is particularly similar between R. sca- (from 2.9 Ma until the Upper Pleistocene; Database of laris and Z. situla/Z. longissimus (Schulz, 1996), except as re- Vertebrates: fossil fishes, amphibians, reptiles and birds, gards the rostral shield. In R. scalaris the rostral shield forms available at: http://www.wahre-staerke.com/). The low ge- an acute angle posteriorly, clearly pointing behind and wedged netic diversity observed in R. scalaris across its range likely between the internasal scales, which represents a unique trait reflects a drastic demographic and range contraction during in ratsnakes (Boulenger, 1894). To our knowledge, this is the the last glacial period, followed by a recent expansion across only external feature that differentiated the genera Rhinechis the Iberian Peninsula (Nulchis, Biaggini, Carretero, & Harris, and Zamenis (Boulenger, 1894; Venchi & Sindaco, 2006). 2008; Silva‐Rocha, Salvi, Sillero, Mateo, & Carretero, 2015). In contrast, Helfenberger (2001) pointed out the phyloge- Range contraction and divergence during the last glacial pe- netic distinctiveness of R. scalaris based on the analysis of the riods were suggested also for Z. hohenackeri, Z. longissimus anatomical, osteological and allozyme variation of Palaearctic and Z. situla. The former presents a fragmented distribution ratsnakes. However, the characters analysed in this study show with three genetic lineages probably associated with distinct little phylogenetic value and high level of homoplasy, and the glacial refugia: one from the Caucasus and north‐eastern uniqueness of R. scalaris is not apparent in any of these data Anatolia, another from southern Turkey and the last from the sets. Each derived visceral and osteological feature listed for southern Amanos and Lebanon Mountains (Jandzik, Avci, & R. scalaris (the cranial reduction of the hypapophyses, a cra- Gvoždík, 2013). According to Musilová, Zavadil, Marková, nially shifted heart, long right and rudimentary left lung, a and Kotlík (2010), the current distribution of Z. longissi- tracheal opening which ends far caudally and large overlap of mus is the result of the expansion of two main lineages from lung and liver) is shared with one or more species of the genera their respective western and eastern refugia after the last analysed, Oocatochus, Coronella, Elaphe and Coelognathus glacial maximum. Holocene fossils in Denmark, Germany Fitzinger, 1843. Phenograms based on soft anatomical char- and Poland suggest a wider range of the species in northern acters (position and length of visceral organs) and vertebral areas where the species is no longer present (e.g., Böhme, characters revealed conflicting patterns across characters and 2000; Gomille, 2002; Helfenberger & Schulz, 2013; Ljungar, sexes (diffuse homoplasy, sensu Lee, 2000), whereas one‐ 1995; Szyndlar, 1984). Similarly, the range of Z. situla ex- third of the analysed taxa lacks private alleles at the allozyme tended further north than presently, with fossils attributed loci analysed, suggesting the occurrence of homoplastic al- to the lower Pleistocene in Austria and the Czech Republic, leles (distinct alleles with equal electrophoretic mobility). where the species became extinct likely during glacial pe- Many phenograms based on visceral and vertebrae features riods. Molecular phylogenetic analyses including specimens show similarity between R. scalaris and Z. hohenackeri or SALVI et al. | 457 Z. situla (e.g., figs 14–17 in Helfenberger, 2001), while the ACKNOWLEDGMENTS one based on allozyme data shows a close relationship be- We are indebted with Philippe Geniez (CEFE, Centre tween the genera Rhinechis, Coronella and Zamenis (fig. 19 d’Écologie Fonctionnelle et Évolutive, Montpellier, in Helfenberger, 2001), which is consistent with molecular France) and Faraham Ahmadzadeh (Shahid Beheshti phylogenetic studies (Burbrink & Lawson, 2007; Chen et al., University, Tehran, Iran) for donating tissue samples of 2017; this study). However, it is problematic drawing “phylo- Middle‐East species and with Notker Helfenberger and genetic” inferences from the analysis of characters such those Juan M. Pleguezuelos for constructive discussion. We analysed by Helfenberger (2001) because of the lack of a sta- thank Àlex Cortada for the help with the biogeographi- tistical or genealogical framework to understand the evolu- cal reconstruction analyses. DS is currently supported by tionary relationships between character states or alleles. the program “Rita Levi Montalcini” for the recruitment Phylogenetic reconstruction based on molecular data pro- of young researchers at the University of L’Aquila. JM vide compelling evidence for a close relationship between and DJH are supported by the Fundação para a Ciência e Rhinechis and Zamenis. These taxa form a well‐supported a Tecnologia (FCT, Portugal): JM, doctoral grant SFRH/ clade in all previous phylogenetic assessments based on dif- BD/81528/2011; DJH IF contract 01627/2014. SC is sup- ferent molecular markers, taxon‐set, phylogenetic methods ported by Grant CGL2015‐70390 from the Ministerio de (Burbrink & Lawson, 2007; Pyron et al., 2011, 2013; Zheng Economía y Competitividad, Spain (co‐funded by Fondos & Wiens, 2016; and also Lenk et al., 2001; or see section were FEDER—EU). Lenk et al., 2001 is discussed). On the other hand, while in many phylogenies R. scalaris is sister to the clade including all Zamenis species, there is no statistical support for the mono- ORCID phyly of Zamenis. It is unlikely that this phylogenetic uncer- Daniele Salvi http://orcid.org/0000-0002-3804-2690 tainty stems from a lack of sufficient data in previous studies and in this study. Indeed, even a recent phylogenomic study employing as much as 304 nuclear loci failed to solve the rela- REFERENCES tionships between R. scalaris, Z. hohenackeri and the remain- Ahmadzadeh, F., Flecks, M., Rödder, D., Böhme, W., Ilgaz, Ç., Harris, ing Zamenis species (Chen et al., 2017). We argue that this D. J., … Carretero, M. A. (2013). Multiple dispersal out of phylogenetic uncertainty reflects an evolutionary scenario in Anatolia: Biogeography and evolution of oriental green liz- which early cladogenetic events within the Rhinechis‐Zamenis ards. Biological Journal of the Linnean Society, 110, 398–408. clade took place at the same time or in a short time frame so https://doi.org/10.1111/bij.12129 that it is hard disentangling which (if any) lineage branched Bandelt, H.‐J., Forster, P., & Rohl, A. (1999). Median‐Joining networks off earlier between R. scalaris, Z. hohenackeri and possibly for inferring intraspecific phylogenies. Molecular Biology and Evolution 16 Z. persicus (Figure 4, Table 2, see also Chen et al., 2017). , , 37–48. https://doi.org/10.1093/oxfordjournals. molbev.a026036 As a consequence, while the monophyletic clade of Zamenis Blondel, J., Catzeflis, F., & Perret, P. (1996). Molecular phylogeny and received little support in all phylogenetic analyses, the clade the historical biogeography of the warblers of the genus Sylvia including Rhinechis and Zamenis received maximum support. (Aves). Journal of Evolutionary Biology, 9, 871–891. https:// In conclusion, while we have little morphological and doi.org/10.1046/j.1420-9101.1996.9060871.x phylogenetic evidence for the distinctiveness of Rhinechis Böhme, G. (2000). Fossile amphibien und reptilien im quartär thürin- relative to Zamenis, a classification considering Rhinechis gens. Veröffentlichungen – Naturkundemuseum Erfurt, 19, scalaris as part of Zamenis would better reflect the evolu- 79–97. Catalogue of the Snakes in the British Museum tionary relationships of this group. Such a classification is Boulenger, G. A. (1894). (Natural History). Volume II., Containing the Conclusion of based on a straightforward definition of a clade, and allows the Colubridæ Aglyphæ. London: British Museum of Natural avoiding the use of monotypic genera, which are question- History. able, especially when compelling evidence for merging them Bruen, T. C., Philippe, H., & Bryant, D. (2006). A simple and robust sta- with other genera is available (e.g., Farris, 1976; Platnick, tistical test for detecting the presence of recombination. Genetics, 1976; Vences, Guayasamin, Miralles, & De La Riva, 2013; 172, 2665–2681. https://doi.org/10.1534/genetics.105.048975 Wiley, 1979). Phylogenetic analyses of R. scalaris show no Bryant, D., & Moulton, V. (2004). Neighbor‐Net: An agglomera- signs of phylogeographical structure across its distribution tive method for the construction of phylogenetic networks. Molecular Biology and Evolution 21 range, further confirming the monotypic status of Rhinechis , , 255–265. https://doi. org/10.1093/molbev/msh018 (Nulchis et al., 2008). The genus Zamenis Wagler, 1830 has Bryson, R. W., de Oca, A. N.‐M., & Velasco, J. R. (2008). Phylogenetic Rhinechis priority over Michahelles, 1833. Therefore, based position of Porthidium hespere (Viperidae: Crotalinae) and on the evidence presented in this study, we consider Coluber phylogeography of arid‐adapted Hognosed Pitvipers based scalaris Schinz, 1822 as part of the genus Zamenis with the on mitochondrial DNA. Copeia, 2008, 172–178. https://doi. binomen Zamenis scalaris (Schinz, 1822). org/10.1643/CH-07-043 458 | SALVI et al. Burbrink, F. T., & Lawson, R. (2007). How and when did Old World the region. International Journal of Earth Sciences, 92, 380– ratsnakes disperse into the New World? Molecular Phylogenetics 396. https://doi.org/10.1007/s00531-003-0322-2 and Evolution, 43, 173–189. https://doi.org/10.1016/j. Eronen, J. T., Ataabadia, M. M., Micheelsb, A., Karme, A., Bernor, ympev.2006.09.009 R. L., & Fortelius, M. (2009). Distribution history and cli- Çaǧatay, M. N., Görür, N., Flecker, R., Sakinç, M., Tünoǧlu, C., Ellam, matic controls of the Late Miocene Pikermian chronofauna. R., … Dikbaş, A. (2006). Paratethyan‐Mediterranean connec- Proceedings of the National Academy of Sciences of the United tivity in the Sea of Marmara region (NW Turkey) during the States of America, 106, 11867–11871. https://doi.org/10.1073/ Messinian. Sedimentary Geology, 188–189, 171–187. https:// pnas.0902598106 doi.org/10.1016/j.sedgeo.2006.03.004 Farris, J. S. (1976). Phylogenetic classification of fossils with re- Carranza, S., Arnold, E. N., & Pleguezuelos, J. M. (2006). Phylogeny, cent species. Systematic Zoology, 25, 271–282. https://doi. biogeography, and evolution of two Mediterranean snakes, org/10.2307/2412495 Malpolon monspessulanus and Hemorrhois hippocrepis Figueroa, A., McKelvy, A. D., Grismer, L. L., Bell, C. D., & Lailvaux, (Squamata, Colubridae), using mtDNA sequences. Molecular S. P. (2016). A Species‐level phylogeny of extant snakes with Phylogenetics and Evolution, 40, 532–546. https://doi. description of a new colubrid subfamily and genus. PLoS One, org/10.1016/j.ympev.2006.03.028 11, e0161070. https://doi.org/10.1371/journal.pone.0161070 Chen, X., Jiang, K., Guo, P., Huang, S., Rao, D., Ding, L., … Burbrink, Flot, J.‐F. (2010). SEQPHASE: A web tool for interconverting F. T. (2014). Assessing species boundaries and the phylogenetic PHASE input/output files and FASTA sequence alignments. position of the rare Szechwan ratsnake, Euprepiophis perla- Molecular Ecology Resources, 10, 162–166. https://doi. ceus (Serpentes: Colubridae), using coalescent‐based methods. org/10.1111/j.1755-0998.2009.02732.x Molecular Phylogenetics and Evolution, 70, 130–136. Gökaşan, E., Demirbag, E., Oktay, F. Y., Ecevitoglu, B., Şimşek, Chen, X., Lemmon, A. R., Lemmon, E. M., Pyron, R. A., & Burbrink, F. M., & Yüce, H. (1997). On the origin of the Bosphorus. T. (2017). Using phylogenomics to understand the link between Marine Geology, 140, 183–199. https://doi.org/10.1016/ biogeographic origins and regional diversification in ratsnakes. s0025-3227(97)00022-4 Molecular Phylogenetics and Evolution, 111, 206–218. https:// Gomille, A. (2002). Die Äskulapnatter Elaphe longissima – Verbreitung doi.org/10.1016/j.ympev.2017.03.017 und Lebensweise in Mitteleuropa. Frankfurt am Main: Edition Chen, X., Mckelvy, A. D., Grismer, L. L., Matsui, M., Nishikaw, K., Chimaira. & Burbrink, F. T. (2014). The phylogenetic position and tax- Guicking, D., & Lawson, R. (2006). Evolution and phylogeny onomic status of the rainbow tree snake Gonyophis margari- of the genus Natrix (Serpentes: Colubridae). Biological tatus (Peters, 1871) (Squamata: Colubridae). Zootaxa, 3881, Journal of the Linnean Society, 87, 127–143. https://doi. 532–548. org/10.1111/j.1095-8312.2006.00561.x Chobanov, D. P., Kaya, S., Grzywacz, B., Warchalowska‐Sliwa, E., & Heled, J., & Drummond, A. J. (2010). Bayesian inference of species Ciplak, B. (2017). The Anatolio‐Balkan phylogeographic fault: trees from multilocus data. Molecular Biology and Evolution, A snapshot from the genus Isophya (Orthoptera, Tettigoniidae). 27, 570–580. https://doi.org/10.1093/molbev/msp274 Zoologica Scripta, 46, 165–179. https://doi.org/10.1111/ Helfenberger, N. (2001). Phylogenetic relationships of Old World zsc.12194 ratsnakes based on visceral organ topography, osteology, Creutzburg, N. (1963). Paleogeographic evolution of Crete from and allozyme variation. Russian Journal of Herpetology, 8, Miocene till our days. Cretan Annals, 15–16, 336–342. 1–62. Cummins, C. A., & McInerney, J. O. (2011). A method for inferring Helfenberger, N., & Schulz, K.‐D. (2013). A checklist of the extinct taxa the rate of evolution of homologous characters that can poten- allied to the Old World ratsnakes, with remarks on the phylog- tially improve phylogenetic inference, resolve deep divergence eny and radiation of Elaphe, Rhinechis and Zamenis. In K.‐D. and correct systematic biases. Systematic Biology, 60, 833–844. Schultz (Ed.), Old World ratsnakes – A collection of papers (pp. https://doi.org/10.1093/sysbio/syr064 423–427). Berg SG: Bushmaster Publications. Daza, J. M., Smith, E. N., Páez, V. P., & Parkinson, C. L. (2009). Hoelzer, G. A., & Meinick, D. J. (1994). Patterns of speciation and lim- Complex evolution in the Neotropics: The origin and diver- its to phylogenetic resolution. Trends in Ecology & Evolution, sification of the widespread genus Leptodeira (Serpentes: 9, 104–107. Colubridae). Molecular Phylogenetics and Evolution, 53, 653– Huson, D. H., & Bryant, D. (2006). Application of phylogenetic net- 667. https://doi.org/10.1016/j.ympev.2009.07.022 works in evolutionary studies. Molecular Biology and Evolution, Dermitzakis, D. M., & Papanikolaou, D. J. (1981). Paleogeography and 23, 254–267. https://doi.org/10.1093/molbev/msj030 geodynamics of the Aegean region during the Neogene. Annales IUCN (2017). The IUCN Red List of Threatened Species. Version Geologique Des Pays Helleenique, 30, 245–289. 2017‐3. http://www.iucnredlist.org Drummond, A. J., Suchard, M. A., Xie, D., & Rambaut, A. (2012). Jandzik, D., Avci, A., & Gvoždík, V. (2013). Incongruence between Bayesian phylogenetics with BEAUti and the BEAST 1.7. taxonomy and genetics: Three divergent lineages within two Molecular Biology and Evolution, 29, 1969–1973. https://doi. subspecies of the rare Transcaucasian rat snake (Zamenis org/10.1093/molbev/mss075 hohenackeri). Amphibia Reptilia, 34, 579–584. https://doi. Edgar, R. C. (2004). MUSCLE: Multiple sequence alignment with high org/10.1163/15685381-00002911 accuracy and high throughput. Nucleic Acids Research, 32, Kapli, P., Botoni, D., Ilgaz, C., Kumlutaş, Y., Avcı, A., Rastegar‐ 1792–1797. https://doi.org/10.1093/nar/gkh340 Pouyani, N., … Poulakakis, N. (2013). Molecular phylog- Elmas, A. (2003). Late Cenozoic tectonics and stratigraphy and north- eny and historical biogeography of the Anatolian lizard western Anatolia: The effects of the North Anatolian Fault to Apathya (Squamata, Lacertidae). Molecular Phylogenetics SALVI et al. | 459 and Evolution, 66, 992–1001. https://doi.org/10.1016/j. Evolutionary history of Psammodromus lizards in space and ympev.2012.12.002 time. Journal of Biogeography, 44, 2877–2890. https://doi. Kaya, S., & Çiplak, B. (2017). Phylogeography and taxonomy of the org/10.1111/jbi.13100 Psorodonotus caucasicus (Orthoptera, Tettigoniidae) group: Mendes, J., Salvi, D., Harris, D. J., Els, J., & Carranza, S. (2018). Independent double invasion of the Balkans from the Caucasus. Hidden in the Arabian mountains: Multilocus phylogeny reveals Systematic Entomology, 42, 118–133. https://doi.org/10.1111/ cryptic diversity in the endemic Omanosaura lizards. Journal syen.12197 of Zoological Systematics and Evolutionary Research, in press. Kearse, M., Moir, R., Wilson, A., Stones‐Havas, S., Cheung, M., https://doi.org/10.1111/jzs.12210 Sturrock, S., … Drummond, A. (2012). Geneious Basic: An Mezzasalma, M., Dall'Asta, A., Loy, A., Cheylan, M., Lymberakis, integrated and extendable desktop software platform for the or- P., Zuffi, M. A. L., … Guarino, F. M. (2015). A sisters’ story: ganization and analysis of sequence data. Bioinformatics, 28, Comparative phylogeography and taxonomy of Hierophis viri- 1647–1649. https://doi.org/10.1093/bioinformatics/bts199 diflavus and H. gemonensis (Serpentes, Colubridae). Zoologica Kornilios, P., Ilgaz, Ç., Kumlutaş, Y., Giokas, S., Fraguedakis‐Tsolis, Scripta, 44, 495–508. https://doi.org/10.1111/zsc.12115 S., & Chondropoulos, B. (2011). The role of Anatolian Micó, E., Sanmartín, I., & Galante, E. (2009). Mediterranean diversifi- refugia in herpetofaunal diversity: An mtDNA analy- cation of the grass‐feeding Anisopliina beetles (Scarabaeidae, sis of Typhlops vermicularis Merrem, 1820 (Squamata, Rutelinae, Anomalini) as inferred by bootstrap‐averaged disper- Typhlopidae). Amphibia‐Reptilia, 32, 351–363. https://doi. sal‐vicariance analysis. Journal of Biogeography, 36, 546–560. org/10.1163/017353711X579858 https://doi.org/10.1111/j.1365-2699.2008.02010.x Kyriazi, P., Kornilios, P., Nagy, Z. T., Poulakakis, N., Kumlutaş, Y., Musilová, R., Zavadil, V., Marková, S., & Kotlík, P. (2010). Relics of the Ilgaz, Ç., … Lymberakis, P. (2013). Comparative phylo- Europe's warm past: Phylogeography of the Aesculapian snake. geography reveals distinct colonization patterns of Cretan Molecular Phylogenetics and Evolution, 57, 1245–1252. https:// snakes. Journal of Biogeography, 40, 1143–1155. https://doi. doi.org/10.1016/j.ympev.2010.09.017 org/10.1111/jbi.12057 Myers, E. A., Rodríguez‐Robles, J. A., Denardo, D. F., Staub, R. E., Lanfear, R., Frandsen, P. B., Wright, A. M., Senfeld, T., & Calcott, Stropoli, A., Ruane, S., & Burbrink, F. T. (2013). Multilocus B. (2016). PartitionFinder 2: New methods for selecting par- phylogeographic assessment of the California Mountain titioned models of evolution for molecular and morphological Kingsnake (Lampropeltis zonata) suggests alternative pat- phylogenetic analyses. Molecular Biology and Evolution, 34, terns of diversification for the California Floristic Province. 772–773. https://doi.org/10.1093/molbev/msw260 Molecular Ecology, 22, 5418–5429. https://doi.org/10.1111/ Lee, M. S. (2000). Soft anatomy, diffuse homoplasy, and the relation- mec.12478 ships of lizards and snakes. Zoologica Scripta, 29, 101–130. Nagy, Z. T., Lawson, R., Joger, U., & Wink, M. (2004). Molecular https://doi.org/10.1046/j.1463-6409.2000.00035.x systematics of racers, whipsnakes and relatives (Reptilia: Lenk, P., Joger, U., & Wink, M. (2001). Phylogenetic relationships Colubridae) using mitochondrial and nuclear markers. Journal among European rat snakes of the genus Elaphe Fitzinger based of Zoological Systematics and Evolutionary Research, 42, 223– on mitochondrial DNA sequence comparisons. Amphibia‐ 233. https://doi.org/10.1111/j.1439-0469.2004.00249.x Reptilia, 22, 329–339. Nulchis, V., Biaggini, M., Carretero, M. A., & Harris, D. J. (2008). Ljungar, L. (1995). First subfossil find of the Aesculapian snake, Elaphe Unexpectedly low mitochondrial DNA variation within the longissima (Laur.) from a Mesolithic settlement in Denmark. ladder snake Rhinechis scalaris. North‐Western Journal of Amphibia‐Reptilia, 16, 93–94. Zoology, 4, 119–124. Lo Presti, R. M., & Oberprieler, C. (2009). Evolutionary history, Oberprieler, C. (2005). Temporal and spatial diversification of Circum‐ biogeography and eco‐climatological differentiation of the Mediterranean Compositae‐Anthemideae. Taxon, 54, 951–966. genus Anthemis L. (Compositae, Anthemideae) in the circum‐ Platnick, N. I. (1976). Are monotypic genera possible? Systematic Mediterranean area. Journal of Biogeography, 36, 1313–1332. Zoology, 25, 198–199. https://doi.org/10.2307/2412749 https://doi.org/10.1111/j.1365-2699.2009.02121.x Plummer, M., Best, N., Cowles, K., & Vines, K. (2006). CODA: Maddison, W. P., & Maddison, D. R. (2003). Mesquite: a modular sys- Convergence diagnosis and output analysis for MCMC. R News, tem for evolutionary analysis. Version 3.2. http://mesquiteproj- 6, 7–11. https://doi.org/10.1159/000323281 ect.org Popov, S. V., Shcherba, I. G., Ilyina, L. B., Nevesskaya, L. A., Melinte‐Dobrinescu, M. C., Suc, J.‐P., Clauzon, G., Popescu, S.‐M., Paramonova, N. P., Khondkarian, S. O., & Magyar, I. (2006). Armijo, R., Meyer, B., … Çakir, Z. (2009). The messinian sa- Late Miocene to pliocene palaeogeography of the paratethys linity crisis in the dardanelles region: Chronostratigraphic con- and its relation to the mediterranean. Palaeogeography, straints. Palaeogeography, Palaeoclimatology, Palaeoecology, Palaeoclimatology, Palaeoecology, 238, 91–106. https://doi. 278, 24–39. https://doi.org/10.1016/j.palaeo.2009.04.009 org/10.1016/j.palaeo.2006.03.020 Mendes, J., Harris, D. J., Carranza, S., & Salvi, D. (2016). Evaluating Poulakakis, N., Kapli, P., Lymberakis, P., Trichas, A., Vardinoyiannis, the phylogenetic signal limit from mitogenomes, slow evolving K., Sfenthourakis, S., & Mylonas, M. (2015). A review of phy- nuclear genes, and the concatenation approach. New insights logeographic analyses of animal taxa from the Aegean and into the Lacertini radiation using fast evolving nuclear genes surrounding regions. Journal of Zoological Systematics and and species trees. Molecular Phylogenetics and Evolution, 100, Evolutionary Research, 53, 18–32. https://doi.org/10.1111/ 254–267. https://doi.org/10.1016/j.ympev.2016.04.016 jzs.12071 Mendes, J., Harris, D. J., Carranza, S., & Salvi, D. (2017). Pyron, R. A., & Burbrink, F. T. (2009). Neogene diversification and Biogeographical crossroad across the Pillars of Hercules: taxonomic stability in the snake tribe Lampropeltini (Serpentes: 460 | SALVI et al. Colubridae). Molecular Phylogenetics and Evolution, 52, 524– genealogies and species tree. Journal of Zoological Systematics 529. https://doi.org/10.1016/j.ympev.2009.02.008 and Evolutionary Research, 53, 45–54. https://doi.org/10.1111/ Pyron, R. A., Burbrink, F. T., Colli, G. R., de Oca, A. N. M., Vitt, L. jzs.12075 J., Kuczynski, C. A., & Wiens, J. J. (2011). The phylogeny of Schulz, K.‐D. (1996). A monograph of the colubrid snakes of the genus advanced snakes (Colubroidea), with discovery of a new sub- Elaphe Fitzinger. Würselen: Koeltz Scientific Books. family and comparison of support methods for likelihood trees. Shimodaira, H. (2002). An approximately unbiased test of phylogenetic Molecular Phylogenetics and Evolution, 58, 329–342. https:// tree selection. Systematic Biology, 51, 492–508. https://doi. doi.org/10.1016/j.ympev.2010.11.006 org/10.1080/10635150290069913 Pyron, R., Burbrink, F. T., & Wiens, J. J. (2013). A phylogeny and re- Shimodaira, H., & Hasegawa, M. (1999). Multiple comparisons of vised classification of Squamata, including 4161 species of liz- log‐likelihoods with applications to phylogenetic inference. ards and snakes. BMC Evolutionary Biology, 13, 93. https://doi. Molecular Biology and Evolution, 16, 1114–1116. org/10.1186/1471-2148-13-93 Shimodaira, H., & Hasegawa, M. (2001). CONSEL: For as- Quintero, I., Keil, P., Jetz, W., & Crawford, F. W. (2015). Historical sessing the confidence of phylogenetic tree selection. biogeography using species geographical ranges. Systematic Bioinformatics, 17, 1246–1247. https://doi.org/10.1093/ Biology, 64, 1059–1073. https://doi.org/10.1093/sysbio/syv057 bioinformatics/17.12.1246 R Core Team (2017). R: A language and environment for statisti- Silva‐Rocha, I., Salvi, D., Sillero, N., Mateo, J. A., & Carretero, M. A. cal computing. Vienna, Austria: R Foundation for Statistical (2015). Snakes on the Balearic Islands: An Invasion tale with Computing. https://www.r-project.org/ implications for native biodiversity conservation. PLoS One, 10, Rambaut, A., Suchard, M. A., & Drummond, A. J. (2014). Tracer v1.6. e0121026. https://doi.org/10.1371/journal.pone.0121026 http://tree.bio.ed.ac.uk/software/tracer/ Silvestro, D., & Michalak, I. (2012). raxmlGUI: A graphical front‐end Rodríguez‐Robles, J. A., & De Jesús‐Escobar, J. M. (1999). Molecular for RAxML. Organisms Diversity & Evolution, 12, 335–337. systematics of New World lampropeltinine snakes (Colubridae): https://doi.org/10.1007/s13127-011-0056-0 Implications for biogeography and evolution of food habits. Skourtanioti, E., Kapli, P., Ilgaz, Ç., Kumlutaş, Y., Avcı, A., Biological Journal of the Linnean Society, 68, 355–385. https:// Ahmadzadeh, F., … Poulakakis, N. (2016). A reinvestigation doi.org/10.1006/bijl.1999.0320 of phylogeny and divergence times of the Ablepharus kitaibelii Rögl, F. (1998). Palaeogeographic considerations for Mediterranean species complex (Sauria, Scincidae) based on mtDNA and and Paratethys seaways (Oligocene to Miocene). Annalen Des nuDNA genes. Molecular Phylogenetics and Evolution, 103, Naturhistorischen Museums in Wien, 99, 279–310. 199–214. https://doi.org/10.1016/j.ympev.2016.07.005 Ruane, S., Bryson, R. W., Pyron, R. A., & Burbrink, F. T. (2014). Speybroeck, J., Beukema, W., Bok, B., & Van Der Voort, J. (2016). Field Coalescent species delimitation in Milksnakes (Genus guide to the amphibians and reptiles of Britain and Europe. Lampropeltis) and impacts on phylogenetic comparative anal- London, UK: Bloomsbury Publishing. yses. Systematic Biology, 63, 231–250. https://doi.org/10.1093/ Stamatakis, A. (2006). RAxML‐VI‐HPC: Maximum likelihood‐based sysbio/syt099 phylogenetic analyses with thousands of taxa and mixed mod- Ruane, S., Torres‐Carvajal, O., & Burbrink, F. T. (2015). Independent els. Bioinformatics, 22, 2688–2690. https://doi.org/10.1093/ demographic responses to climate change among temperate and bioinformatics/btl446 tropical milksnakes (Colubridae: Genus Lampropeltis). PLoS Stephens, M., & Scheet, P. (2005). Accounting for decay of linkage One, 10, 1–17. https://doi.org/10.1371/journal.pone.0128543 disequilibrium in haplotype inference and missing‐data impu- Salvi, D., Harris, D. J., Bombi, P., Carretero, M. A., & Bologna, M. tation. American Journal of Human Genetics, 76, 449–462. A. (2010). Mitochondrial phylogeography of the Bedriaga's https://doi.org/10.1086/428594 rock lizard, Archaeolacerta bedriagae (Reptilia: Lacertidae) Stephens, M., Smith, N. J., & Donnelly, P. (2001). A new statistical endemic to Corsica and Sardinia. Molecular Phylogenetics method for haplotype reconstruction from population data. and Evolution, 56, 690–697. https://doi.org/10.1016/j. American Journal of Human Genetics, 68, 978–989. https://doi. ympev.2010.03.017 org/10.1086/319501 Salvi, D., Lucente, D., Mendes, J., Liuzzi, C., Harris, D. J., & Bologna, Szyndlar, Z. (1984). Fossil snakes from Poland. Acta Zoologica M. A. (2017). Diversity and distribution of the Italian Cracoviensia, 28, 1–156. Aesculapian snake Zamenis lineatus: A phylogeographic assess- Szyndlar, Z. (1985). Ophidian fauna (Reptilia, Serpentes) from the up- ment with implications for conservation. Journal of Zoological permost Miocene of Algora (Spain). Estudios Geológicos, 41, Systematics and Evolutionary Research, 55, 222–237. https:// 447–465. doi.org/10.1111/jzs.12167 Tamura, K., Stecher, G., Peterson, D., Filipski, A., & Kumar, S. (2013). Salvi, D., Schembri, P. J., Sciberras, A., & Harris, D. J. (2014). MEGA6: Molecular evolutionary genetics analysis version 6.0. Evolutionary history of the Maltese wall lizard Podarcis fil- Molecular Biology and Evolution, 30, 2725–2729. https://doi. folensis: Insights on the “Expansion‐Contraction” model of org/10.1093/molbev/mst197 Pleistocene biogeography. Molecular Ecology, 23, 1167–1187. Townsend, T. M., Alegre, R. E., Kelley, S. T., Wiens, J. J., & Reeder, T. https://doi.org/10.1111/mec.12668 W. (2008). Rapid development of multiple nuclear loci for phy- Sambrook, J., Fritsch, E. F., & Maniatis, T. (1989). Molecular cloning: logenetic analysis using genomic resources: An example from A laboratory manual. New York: Cold Spring Harbor Press. squamate reptiles. Molecular Phylogenetics and Evolution, 47, Sampaio, F. L., Harris, D. J., Perera, A., & Salvi, D. (2014). Phylogenetic 129–142. and diversity patterns of Blanus worm lizards (Squamata: Uetz, P., Freed, P., & Hošek, J. (2018). The Reptile Database. Retrieved Amphisbaenia): Insights from mitochondrial and nuclear gene from http://www.reptile-database.org. accessed February, 2018. SALVI et al. | 461 Utiger, U., Helfenberger, N., Schätti, B., Schmidt, C., Ruf, M., & Yu, Y., Harris, A. J., Blair, C., & He, X. (2015). RASP (Reconstruct Ziswiler, V. (2002). Molecular systematics and phylogeny of Ancestral State in Phylogenies): A tool for historical biogeogra- Old World and New World ratsnakes, Elaphe Auct., and related phy. Molecular Phylogenetics and Evolution, 87, 46–49. https:// genera (Reptilia, Squamata, Colubridae). Russian Journal of doi.org/10.1016/j.ympev.2015.03.008 Herpetology, 9, 105–124. Yu, Y., Harris, A. J., & He, X. (2010). S‐DIVA (Statistical Dispersal‐ Utiger, U., Schatti, B., & Helfenberger, N. (2005). The oriental colu- Vicariance Analysis): A tool for inferring biogeographic his- brine genus Coelognathus Fitzinger, 1843 and classification of tories. Molecular Phylogenetics and Evolution, 56, 848–850. old and new world racers and ratsnakes (Reptilia, Squamata, https://doi.org/10.1016/j.ympev.2010.04.011 Colubridae, Colubrinae). Russian Journal of Herpetology, 12, Yu, Y., Harris, A. J., & Xingjin, H. (2011). RASP version 2.1b. http:// 39–60. mnh.scu.edu.cn/soft/blog/RASP Vences, M., Guayasamin, J. M., Miralles, A., & De La Riva, I. (2013). Zaher, H., Grazziotin, F. G., Graboski, R., Fuentes, R. G., Sánchez‐ To name or not to name: Criteria to promote economy of change Martinez, P., Montingelli, G. G., … Murphy, R. W. (2012). in Linnaean classification schemes. Zootaxa, 3636, 201–244. Phylogenetic relationships of the genus Sibynophis (Serpentes: https://doi.org/10.11646/zootaxa.3636.2.1 Colubroidea). Papéis Avulsos de Zoologia, 52, 141–149. Venchi, A., & Sindaco, R. (2006). Annotated checklist of the reptiles of Zheng, Y., & Wiens, J. J. (2016). Combining phylogenomic and super- the Mediterranean countries, with keys to species identification. matrix approaches, and a time‐calibrated phylogeny for squa- Part 2 – Snakes (Reptilia, Serpentes). Annali Del Museo Civico mate reptiles (lizards and snakes) based on 52 genes and 4162 Di Storia Naturale “G.Doria”, XCVIII, 259–364. species. Molecular Phylogenetics and Evolution, 94, 537–547. Venczel, M. (1994). Late Miocene snakes from Polgárdi (Hungary). https://doi.org/10.1016/j.ympev.2015.10.009 Acta Zoologica. Cracoviensia, 37, 1–29. Wallach, V., Williams, K. L., & Boundy, J. (2014). Snakes of the world: A catalogue of living and extinct species (pp. 1237). Boca Raton, FL: Taylor and Francis, CRC Press. SUPPORTING INFORMATION Weisrock, D. W., Macey, J. R., Ugurtas, I. H., Larson, A., & Papenfuss, Additional supporting information may be found online in T. J. (2001). Molecular phylogenetics and historical biogeog- raphy among salamandrids of the “true” salamander clade: the Supporting Information section at the end of the article. Rapid branching of numerous highly divergent lineages in Mertensiella luschani associated with the rise of Anatolia. Molecular Phylogenetics and Evolution, 18, 434–448. https:// How to cite this article: Salvi D, Mendes J, Carranza S, doi.org/10.1006/mpev.2000.0905 Harris DJ. Evolution, biogeography and systematics of Wiley, E. O. (1979). An annotated Linnaean hierarchy, with comments the western Palaearctic Zamenis ratsnakes. Zool Scr. on natural taxa and competing systems. Systematic Zoology, 28, 2018;47:441–461. https://doi.org/10.1111/zsc.12295 308–337. https://doi.org/10.2307/2412585