The distribution and taxonomy of Lissotriton newts in Turkey (Amphibia, Salamandridae)

Abstract Two and perhaps three taxa of Lissotriton newt occur in Turkey. Their species status is controversial. The distribution of these taxa and the taxonomic status of each are reviewed and discussed. A database of 128 Turkish Lissotriton localities was compiled and species distribution models were constructed. We reiterate that the presence of Lissotriton (vulgaris) lantzi in Turkey is disputed and needs confirmation. The range of Lissotriton (vulgaris) kosswigi is restricted to north-western Anatolia – given the small global range of this Turkey endemic, a closer look at its conservation status is warranted. The distribution of Lissotriton vulgaris schmidtleri covers western Asiatic and European Turkey. The findings support an allopatric distribution of the Turkish Lissotriton species. We reflect on the biological significance of previously reported morphological intermediates between Lissotriton (vulgaris) kosswigi and Lissotriton vulgaris schmidtleri in the light of the recent proposal to recognize kosswigi at the species level. The available data are in line with species status for Lissotriton (vulgaris) lantzi and Lissotriton (vulgaris) kosswigi. Although Lissotriton vulgaris schmidtleri is a genetically diverged taxon as well, the extent of gene flow with parapatric European Lissotriton taxa is as yet unknown.

In Turkey, two and perhaps three Lissotriton occur ( Fig. 1; Schmidtler and Franzen 2004). The range of lantzi covers the Caucasus region and the taxon might occur in the extreme north-east of Turkey, near the border with Georgia Franzen 2004, Skorinov et al. 2014). The taxon kosswigi is restricted to north-western Anatolia (Schmidtler and Franzen 2004). The taxon schmidtleri was originally considered to be restricted to western Asiatic Turkey, but Raxworthy (1988Raxworthy ( , 1990 suggested it might extend into Europe. Genetic data have subsequently confirmed that this taxon's range encompasses European Turkey (Nadachowska and Babik 2009;Pabijan et al. 2014). The range of schmidtleri protrudes further into the Balkan Peninsula, but its range limit is as yet unclear; newts with mitochondrial DNA typical of schmidtleri have been recorded as far north-west as easternmost Greece and central Bulgaria (Pabijan et al. 2014). Previous records of vulgaris from Turkey reflect incomplete taxonomy and can be referred to the other taxa (cf. Dubois andRaffaëlli 2012, Olgun et al. 1999).
The Lissotriton vulgaris group comprises two main morphological types: one with a smooth crest and flappy feet and another with a ragged crest and limited fringing on the feet (Fig. 2). Distinguishing the taxa within the two main groups is less straightforward and this topic is beyond the scope of the present paper (we refer to Raxworthy (1990) and Schmidtler and Franzen (2004) for a detailed treatment). Relevant for the current paper is that kosswigi belongs to the 'smooth-crested with flappy feet' type and schmidtleri to the 'ragged-crested with limited feet-fringing' type and that morphological intergradation has been reported between these two taxa (e.g. Freytag 1955, 1957, Tabrizi 1980, Yılmaz 1983). In Fig. 2 typical males of kosswigi and schmidtleri are depicted. Next to the smooth crest and flappy feet, kosswigi possesses a tail filament and its crest starts at a more posterior position than in schmidtleri. Although lantzi belongs to the 'ragged-crested with limited feet-fringing' type as well, confusion with schmidtleri is ruled out based on geography.
An overview of the distribution of the Turkish Lissotriton taxa is provided by composing a database of localities and constructing species distribution models. The focus is mainly on the taxa kosswigi and schmidtleri and particularly the supposed genetic admixture between the two. Finally, we reflect on the as yet controversial proposal to treat the Turkish Lissotriton taxa as distinct species.

Material and methods
The distribution of Lissotriton in Turkey has been reviewed and a database compiled of localities based on: 1) the collection of the Zoology Laboratory of the Department of Biology at Science and Arts Faculty, Adnan Menderes University, 2) extensive personal field observations, and 3) a review of the literature (Bozkurt et al. in press, Çevik et al. 1997, Çiçek and Ayaz 2011, Demirsoy 1996, Eiselt 1966, Freytag 1955, Freytag 1957, Mulder 1995, Olgun et al. 1999, Raxworthy 1988, Schmidtler and Schmidtler 1967, Skorinov et al. 2014, Sparreboom and Arntzen 1987, Tabrizi 1980, Taşkın and Olgun 2003, Yılmaz 1983, 1989. In this paper we particularly focused on kosswigi, this being the rarest and most restricted taxon globally. The aim was not to be exhaustive for schmidtleri, which is common were not included, and widely distributed in western Turkey. Localities within one kilometre of one another and in such cases the locality with the most accurate information available was chosen. We particularly focused on records of presumed transitional forms between kosswigi and schmidtleri reported in the literature, considering their relevance in the taxonomic treatment of the different Lissotriton taxa occurring in Turkey. For a species distribution modelling exercise for lantzi (and a comprehensive overview of the distribution of this taxon outside of Turkey) we refer to Skorinov et al. (2014). Species distribution models were constructed for kosswigi and schmidtleri using Maxent 3.3.3k (Phillips et al. 2006). For climate layers bioclimatic variables were used, at 2.5 arcminute resolution (c. 5 × 5 km) available from the WorldClim database 1.4 (Hijmans et al. 2005; http://www.worldclim.org). We trimmed these layers to an extent that broadly encompasses the distribution of the genus Lissotriton: the area between -15 and 65 degrees longitude and between 30 and 75 degrees latitude. Following Guisan and Thuiller (2005) and Peterson (2011) a subset considered to reflect physiological limitations of the study species (in this case seasonality) was selected while showing little multicollinearity (a Pearson's correlation of r < 0.7): bio10 = mean temperature of warmest quarter, bio11 = mean temperature of coldest quarter, bio15 = precipitation seasonality, bio16 = precipitation of wettest quarter, and bio17 = precipitation of driest quarter. To determine whether our species distribution model performs better than random expectation, we tested its AUC value against a null model based on 99 models for random localities (see Raes and ter Steege 2007 for details). Random point data were created with ENMTools 1.3 (Warren et al. 2010). To more thoroughly cover the range of environmental conditions experienced by schmidtleri the only four confirmed populations from outside the Turkish range (noted on Fig. 1; details in Pabijan et al. 2014) were included.

Results
A database of 128 distribution records of Turkish Lissotriton newts (49 kosswigi, 78 schmidtleri and one lantzi) is provided in Suppl. material 1. Fig. 1 shows these records The inset shows the rough outlines of the ranges of lantzi (in green), kosswigi (in red) and schmidtleri (in blue). Type localities are marked with a black star. The blue hatched area reflects the unclear range of schmidtleri outside of Turkey (see discussion), with four confirmed records denoted with white dots. The green hatched area reflects the potential occurrence of lantzi in the extreme northeast of Turkey, with a black dot depicting the single historical record for Turkey (see discussion); question marks denote historical records in Armenia and Azerbaijan. The cut-out shows Turkish localities for kosswigi and schmidtleri as black dots. Localities supposedly showing intergradation between kosswigi and schmidtleri are marked with a white star. The arrow highlights a poorly documented locality attributed to kosswigi (see discussion). Details on Turkish localities are provided in Suppl. material 1.
plotted on a map. The map also shows the type localities of kosswigi and schmidtleri, as well as populations reported to contain morphological intermediates between the two taxa. Fig. 3 shows the species distribution models for kosswigi and schmidtleri. The AUC values of these models (0.991 for both kosswigi and schmidtleri) rank above the 99 AUC values based on random points, meaning our species distribution models perform significantly better than random expectation (P < 0.05).

Distribution
The taxon lantzi is widely distributed in the Caucasus region Franzen 2004, Skorinov et al. 2014). A species distribution modelling exercise (Skorinov et al. 2014) revealed that suitable environmental conditions protrude into the extreme northeast of Turkey, near the border with Georgia. However, the continued occurrence of lantzi in Turkey needs confirmation; there is only a single record, dating from the beginning of the twentieth century Franzen 2004, Skorinov et al. 2014). Intriguingly, there are also old reports of the Crested newt Triturus karelinii (Strauch 1870) from this part of Turkey (Wielstra et al. 2010). Just as lantzi, T. karelinii is widely distributed in the Caucasus and, although its occurrence in Turkey is suggested by species distribution modelling (Wielstra et al. 2013c), its actual presence requires further scrutiny. In any case, lantzi is allopatric from the other Turkish taxa: Lissotriton newts are absent from north-east Anatolia (Schmidtler and Franzen 2004; Fig. 1).
The distribution of the Turkey endemic kosswigi is restricted to north-western Anatolia (our exhaustive survey revealed 49 localities; Fig. 1). The species distribution model suggests that suitable environmental conditions extend further to the east along most of the Turkish Black Sea coast (Fig. 3). However, this area appears to be devoid of Lissotriton newts (Fig. 1). Over-prediction is a well-known problem in species distribution modelling (Elith et al. 2011). This could suggest that the climate layers used to create the species distribution model do not properly reflect the factors limiting the distribution of the species, but it could also suggest that not all suitable area could be colonized due to dispersal constraints.
The taxon schmidtleri occurs in the west of Asiatic Turkey and is now known to extend into Europe, across the marine corridor connecting the Aegean and Black Seas Babik 2009, Pabijan et al. 2014). The permeability of this apparent barrier can be ascribed to sea level fluctuations related to glacial cycles and the disjunct distribution pattern of schmidtleri is mirrored by the co-distributed crested newt species T. ivanbureschi Arntzen & Wielstra, 2013 in Wielstra et al. (2013a) (Wielstra and Arntzen 2012). Although the Balkan range of schmidtleri outside of Turkey is poorly understood, the taxon appears to occur well into Bulgaria (Pabijan et al. 2014). This is in conflict with the species distribution model (Fig. 3). It could be that mitochondrial DNA does not properly reflect the range of schmidtleri and overestimates its occurrence in Bulgaria. However, we consider it more likely that, due to the lack of confirmed schmidtleri localities from Bulgaria (whereas the taxon might well be abundant there), the species distribution model underestimates the environmental space inhabited by schmidtleri.
The taxa kosswigi and schmidtleri currently appear allopatric. We have particularly surveyed the area for Lissotriton (pers. obs.) and no localities are known between the schmidtleri locality Gemlik (Olgun et al. 1999; locality 62 in Suppl. material 1) and kosswigi locality Yalova (Demirsoy 1996; locality 18 in Suppl. material 1). The Yalova locality lacks documentation and needs confirmation (note that the locality was not included in Schmidtler and Franzen 2004) and it is suggested that there is probably a larger distribution gap, with the next closest kosswigi locality from the perspective of schmidtleri being Kocaeli (museum record; locality 3 in Suppl. material 1). This apparent distribution gap disagrees with the species distribution models, which suggest suitable environmental conditions for both kosswigi and schmidtleri occur south of the Marmara Sea (Fig. 3).
Based on introgression of schmidtleri mitochondrial DNA into kosswigi (very similar to mitochondrial DNA found in schmidtleri today) it has been hypothesized that Shown (not to scale) are a typical kosswigi male (A) and a typical schmidtleri male (B). Notice the shape of the doral fin (smooth in kosswigi and ragged in schmidtleri), the position where the dorsal fin starts (approximately above the forelimbs in kosswigi and at the back of the head in schmidtleri), the presence of a thread-like tail filament (found in kosswigi but not in schmidtleri) and the extensiveness of the fringing on the feet (with kosswigi having much more flappy feet than schmidtleri). kosswigi displaced schmidtleri on the Istanbul Peninsula as the waterway between the Black and Marmara Seas rerouted within the last 10,000 years Babik 2009, Wielstra et al. 2013b). Similarly, an as yet undescribed Triturus species was proposed to have displaced T. ivanbureschi in this region (Wielstra et al. 2013a(Wielstra et al. , 2013b. The species distribution models suggest suitable environmental conditions here for both kosswigi and schmidtleri and hence do not provide further insight into how kosswigi was able to locally outcompete schmidtleri (Fig. 3).

Genetic admixture
In light of the current allopatric distribution pattern of kosswigi and schmidtleri, previous reports of transitional forms are curious. Following up on a possible intermediate specimen from Sapanca, Eiselt (1966) could only confirm the presence of pure kosswigi there. Freytag (1955) indicated that in a Lissotriton population from Kanlıca (locality 37 in Suppl. material 1), on the eastern side of the Bosphorus, some males showed characteristics of schmidtleri, namely the dorsal fin being ragged and starting at the back of the head and the lack of a tail filament (cf. Fig. 2). Tabrizi (1980) studied a larger sample of newts from populations throughout the range of kosswigi. He found that four out of 70 newts in Kanlıca showed a schmidtleri-like, relatively anterior starting position of the dorsal fins; all other newts were classified as typical kosswigi. Considering the biogeographical scenario outlined above, a relict schmidtleri population in the process of being replaced by kosswigi via genetic swamping is a possibility. A study on historical gene flow between the two taxa unfortunately did not include samples from the potentially admixed populations, but did suggest ancient gene flow from schmidtleri into kosswigi (Nadachowska and Babik 2009).
Furthermore, Freytag (1957) mentioned that in a Lissotriton population from Baltalimanı (locality 104 in Suppl. material 1), on the western side of the Bosphorus, some males shared similarities with kosswigi, in terms of possessing tail filaments and smooth dorsal fins that started relatively posteriorly (Fig. 2). Yılmaz (1983) studied a larger sample encompassing more populations from European Turkey. He noted newts with kosswigi characteristics at Habibler and Küçükçekmece (localities 121 and 125 in Suppl. material 1). Out of 80 studied newts, 20 had dorsal fins that began at the forelimbs rather than at the back of the head, 41 had smooth dorsal fins and 37 had tail filaments to varying degree (17 with 0-2 mm, 10 with 2-4 mm, 5 with 4-6 mm, and 5 with over 6 mm). Schmidtler and Franzen (2004) state that in schmidtleri males can show kosswigi-like characteristics, but do not provide further details. The presence of kosswigi west of the Bosphorus would not make sense in light of the biogeographical scenario outlined above, unless it could be proven that the Bosphorus on initial formation had a more westward position or formed only after kosswigi reached European Turkey. The rerouting of the marine connection between the Marmara and Black Seas is not yet fully understood and a matter of debate in the paleogeological literature (e.g. , Yaltırak et al. 2012. We suggest that historical biogeographical patterns such as shown by Lissotriton (and Triturus) newts might assist paleogeological reconstruction.
Genetic data from the potentially admixed kosswigi and schmidtleri populations are as yet lacking, but would provide more insight in the matter. However, considering the expansion of the Istanbul agglomeration it should be taken into account that these populations might well have gone extinct. We conclude that potential kosswigischmidtleri admixture represents, at most, the remnants of a former contact zone. The main ranges of the two taxa are currently isolated in the region by the Bosphorus and hence the influence of potentially admixed populations on the genetic integrity of the two taxa can be expected to be negligible. In this light we make some remarks on the not (yet) generally accepted treatment of the Turkish Lissotriton taxa as distinct species (Dubois andRaffaëlli 2009, Frost 2014). . Species distribution models for two Turkish Lissotriton taxa. Shown are species distribution models for kosswigi (A) and schmidtleri (B). The maps depict predicted suitability, which ranges from 0 to 1, in ten equal intervals, with higher values expressed by warmer colours.

Taxonomy
Following the taxonomy of Babik et al. (2005), the Lissotriton vulgaris group consists of seven taxa, namely ampelensis, graecus, kosswigi, lantzi, meridionalis, schmidtleri and the nominal vulgaris. Four of these taxa, graecus, kosswigi, lantzi and meridionalis, are sometimes regarded as specifically distinct (Dubois andRaffaëlli 2009, Frost 2014). The split of graecus and meridionalis has been criticised (Speybroeck et al. 2010) as a misinterpretation of the phylogenetic position of the congener L. montandoni which, due to mitochondrial DNA introgression, is nested within the L. vulgaris group from the perspective of mitochondrial DNA (Babik et al. 2005, Zieliński et al. 2013. However, the taxa kosswigi and lantzi are genuinely genetically diverged for mitochondrial DNA (Babik et al. 2005).
Within the Lissotriton vulgaris group mitochondrial DNA suggests a basal split between lantzi and the rest (Babik et al. 2005). Although the distinction of lantzi from the perspective of the nuclear genome has as yet not been determined, the divergence in the mitochondrial genome and the at least currently disjunct distribution support a scenario of long-term disrupted gene flow with other Lissotriton newts.
The next split in the Lissotriton vulgaris group is between kosswigi and the remaining taxa (Babik et al. 2005). The distinction of kosswigi from its geographical neighbour schmidtleri has been supported in a study exploring gene flow based on eight nuclear DNA markers (Nadachowska and Babik 2009). Given that kosswigi is genetically distinct and currently allopatric from other Lissotriton taxa, its treatment at the species level seems justified. From the conservation perspective it is important whether this geographically restricted, Turkish endemic is treated as a 'unique species' or 'merely a subspecies'.
Although schmidtleri represents a distinct mitochondrial DNA clade as well, it is genetically nested within the European Lissotriton taxa (Babik et al. 2005, Pabijan et al. in prep.). The phylogeography of Lissotriton on the Balkan Peninsula is highly complex, with morphologically distinct subspecies being highly polyphyletic from the mitochondrial DNA perspective (Babik et al. 2005, Pabijan et al. in prep.). Furthermore, no doubt in part because of its turbulent taxonomical history, the morphological distinctiveness of schmidtleri is not well understood (Schmidtler and Franzen 2004). Hence, we refrain from making further comments on the taxonomic status of schmidtleri and rather await further research on nuclear gene flow between schmidtleri and the other Lissotriton taxa on the Balkan Peninsula.