Cranial shape evolution of extant and fossil crocodile newts and its relation to reproduction and ecology

Abstract The diversity of the vertebrate cranial shape of phylogenetically related taxa allows conclusions on ecology and life history. As pleurodeline newts (the genera Echinotriton, Pleurodeles and Tylototriton) have polymorphic reproductive modes, they are highly suitable for following cranial shape evolution in relation to reproduction and environment. We investigated interspecific differences externally and differences in the cranial shape of pleurodeline newts via two‐dimensional geometric morphometrics. Our analyses also included the closely related but extinct genus Chelotriton to better follow the evolutionary history of cranial shape. Pleurodeles was morphologically distinct in relation to other phylogenetically basal salamanders. The subgenera within Tylototriton (Tylototriton and Yaotriton) were well separated in morphospace, whereas Echinotriton resembled the subgenus Yaotriton more than Tylototriton. Oviposition site choice correlated with phylogeny and morphology. Only the mating mode, with a random distribution along the phylogenetic tree, separated crocodile newts into two morphologically distinct groups. Extinct Chelotriton likely represented several species and were morphologically and ecologically more similar to Echinotriton and Yaotriton than to Tylototriton subgenera. Our data also provide the first comprehensive morphological support for the molecular phylogeny of pleurodeline newts.

ecology, e.g. the dentition is usually closely related to the diet (Hotton, 1955;Strait, 1993), and this relation is of special importance for the reconstruction of ancient lineages. As the skeleton usually represents the only remains in the fossil record, it is most promising to carry out comparative osteology that includes extant taxa to draw conclusions about the ecology of extinct vertebrates. Furthermore, by linking morphological traits of extant species to their environment, ecology and life history in a comparative way, it is possible to obtain insights into the evolutionary history of extinct taxa. For example, patterns of countershading in extant species allow conclusions to be made concerning predatorprey interactions or the habitat of dinosaurs (Brown et al., 2017;Smithwick et al., 2017).
True salamanders of the family Salamandridae evolved a variety of cranial shapes (Ivanović and Arntzen, 2017). The taxon of Pleurodelini, often referred to as 'primitive newts', represent a basal group of Salamandridae comprising three extant genera (Zhang et al., 2008;Veith et al., 2018). Whereas the three species of ribbed newts, genus Pleurodeles Michahelles, 1830, inhabit a restricted range in southwest Europe and North Africa along the Mediterranean, crocodile newts are much more diverse and are currently assigned to two genera. Echinotriton Nussbaum and Brodie, 1982 comprises three species, inhabiting the Ryu-Kyu archipelago, Japan and east China (Chang, 1932;Hou et al., 2014). Tylototriton Anderson, 1871 includes 25 species divided into two subgenera: Tylototriton Anderson, 1871and Yaotriton Dubois and Raffaëlli, 2009(Dubois and Raffaelli, 2009. Tylototriton is widely distributed from east Nepal to east and central China, southwards to Myanmar, central Vietnam, Laos and Thailand (Wang et al., 2018;Zaw et al., 2019). Apparently, crocodile newts have a quite conservative morphological evolution (Hernandez et al., 2018), leading to a high number of species mainly recognized by genetic studies in recent years.
Pleurodeline newts are polymorphic in their reproductive mode and mating strategy , including terrestrial and aquatic mating, as well as the choice of oviposition sites (Kuzmin et al., 1994;Ziegler et al., 2008;Igawa et al., 2013;Bernardes et al., 2017). Whereas some species use a ventral amplexus similar to terrestrial salamandridae, others perform a circular mating dance comparable to European newts (Dasgupta, 1994;Roy and Mushahidunnabi, 2001;Fleck, 2010a;2010b;Wang et al., 2017;Gong et al., 2018). Ribbed and crocodile newts occupy various habitats along the latitudinal and altitudinal gradient from tropical lowland rainforests to montane forests and grassy landscapes Hernandez et al., 2017;Hernandez et al., 2019). Their diverse ecology may result in indistinct morphological adaptations hard to access with traditional morphological approaches. Additionally, pleurodeline newts are represented by several fossil taxa. Three extinct species of Tylototriton were described from Germany (Noble, 1928;Herre, 1935;1949), today being recognized as members of other fossil newt genera (Estes, 1981;Nussbaum and Brodie, 1982;Böhme and Ilg, 2003). The most prominent one is the genus Chelotriton Pomel, 1853, currently consisting of four nominally described species (Goldfuss, 1831;Pomel, 1853;Westphal, 1980;Bailon, 1989) of which Chelotriton paradoxus is the best known. Chelotriton is known from Spain to east Europe from the Eocene to Miocene (about 50-11 mya).
Based on unique morphological characters, Chelotriton was assigned to the tribe Pleurodelini by various authors and is regarded as more closely related to crocodile newts, i.e. the genera Echinotriton and Tylototriton, than to Pleurodeles (Marjanović and Witzmann, 2015;Schoch et al., 2015). In recent years, several exceptionally well-preserved specimens of Chelotriton have been excavated from localities in southwest Germany (Figure 1; Roček and Wuttke, 2010;Schoch et al., 2015).
Via two-dimensional (2D) geometric morphometrics of external head and skull morphology, accessed via micro-computed tomography (µCT) scans, we investigated how cranial shape of ribbed and crocodile newts differs interspecifically. We tested how cranial morphology relates to selected ecological and reproductive traits. Fossil Chelotriton specimens from deposits of Randeck Maar and Enspel Crater Lake were included in the analyses of cranial morphology to obtain further hints on the relationship between extant and extinct taxa and to draw conclusions on the ecology of Chelotriton based on morphology-ecology correlations of extant taxa. The overall aim was to obtain novel insights into the evolution of cranial shape in relation to ecology of selected, phylogenetically basal salamandrids.

| MATERIAL S AND ME THODS
We have investigated the crania of 157 newt specimens covering 21 of 31 extant species (68%) which are currently ascribed to the Pleurodelini (Frost, 2018). With additional data from the literature, we covered information on up to 26 species (see below and Table 1). As populations of Echinotriton andersoni originating from the island of Okinawa and the Amami archipelago showed deep divergence and are under debate as to whether they should be recognized as distinct taxonomic units (Hayashi et al., 1992;Honda et al., 2012;Kurabayashi et al., 2012), we treated those separately in our analyses. To exclude additional variation due to sexual dimorphism in extant species, only male specimens were analysed, except for a photograph of the Echinotriton maxiquadratus holotype, which is a female. Additionally, eight well-preserved fossil specimens of the genus Chelotriton from deposits of the Randeck Maar, Baden-Wuerttemberg (17-15 Ma, mammal zone MN5, see Böhme, 2003;Rasser et al., 2013) and Enspel Crater lake,  Ma, mammal zone MP28, see Roček and Wuttke, 2010;Schindler and Wuttke, 2010), both in Germany, were included in the analysis (Figure 1; Schoch et al., 2015). Currently, specimens of Chelotriton from these deposits are tentatively associated with the type species C. paradoxus (Schoch et al., 2015).

| Landmark data acquisition
We investigated the cranial shape of pleurodeline salamanders by 2D geometric morphometric (GM) approaches. Two-dimensional analysis was preferred over three-dimensional analysis as the crania of fossils newts were too flat to apply 3D GM for comparison between extant and extinct samples (see below). First, we took Second, extant and fossil specimens were scanned via µCT to allow investigation of the cranial skeleton. Scans were carried out either with a Bruker SkyScan1272 or within the X-ray imaging laboratory at the Institute for Photon Science and Synchrotron Radiation, Karlsruhe Institute of Technology (KIT), employing a microfocus X-ray tube (XWT-225, X-RAY WorX) and a flat panel detector (XRD 1621 CN14 ES, Perkin Elmer) in combination with a custom-designed mechanical sample manipulator. For the datasets measured at KIT, OctOpus 8.6 (Inside Matters) was used to perform the tomographic reconstruction. Due to the time-consuming procedure of µCT scanning, only a subsample per species was scanned (Table 1, Supporting Information Table S1). In some species with larger distribution areas or presumably different morphologies among localities, additional specimens were analyzed via µCT. In total, 121 specimens including fossils were µCT-scanned (Table S1). The scan resolution for extant newts was either 20.1 (SkyScan) or 21.3 µm (KIT-scanner). Chelotriton specimens were scanned at 35.2 µm resolution. Three-dimensional reconstructions were processed in AmirA® 6.5 (Visualisation Science Group). Flattened and distorted during fossilization, Chelotriton specimens did not allow sufficient reconstruction in a three-dimensional space. Nevertheless, to reconstruct a morphology which is most likely to represent its original dorsal shape, retrodeformation by algorithmic symmetrization using the software IDAV LAndmArk EditOr v.3.7 (http://graph ics.idav.ucdav is.edu/resea rch/ EvoMorph) was performed to reduce asymmetrical distortion in the fossil crania of the Enspel specimens (Tallman et al., 2014).
Landmark configurations for retrodeformation were specifically adapted to each single fossil specimen, as deformation was different in each specimen. We also employed several retrodeformations with different sets of landmarks to receive results appearing as symmetrical as possible but simultaneously not diverging too much from the original shape (Supporting Information Figure S1).
The only sample from Randeck Maar appears to be symmetrical and was therefore not retrodeformed. Two-dimensional images of skulls were taken in dorsal view and, additionally, skull images of extant taxa were taken in the right lateral view ( Figure 2) to allow comparison with external morphology. In the following, 'head shape/morphology' will refer to external cranial morphology including soft tissue, 'skull shape/morphology' to the osteology, and F I G U R E 1 Well-preserved cranial skeletons of fossil Chelotriton specimens from the Enspel crater lake, Rhineland-Palatine (a,b) and the deposits of the Randeck Maar, Baden-Wuerttemberg (c), Germany a c b 'cranial shape/morphology' more generally to cranial morphology irrespective of the dataset analyzed herein.  Table S2). Landmark digitization was carried out by one author using tpsUtil and tpsDig (Rohlf, 2016a;2016b). Specimens were randomly shuffled. To test for accuracy of landmark placement, each landmark configuration was tested by digitizing one specimen five times and five other specimens of the same species. Procrustes distance to the mean shape of replicates and interindividuals were tested against each other to test whether intraindividual landmark placement was consistent in comparison with landmark placement between different individuals.

| Geometric morphometrics
Two-dimensional geometric morphometrics analysis was performed in R version 3.5.3 (R Development Core Team, 2019) using the packages geomorph v.3.1.1 and RRPP v. 0.4.1 (Collyer and Adams, 2018;Adams et al., 2019). The procedure of analysis was equal for each dataset. Missing landmarks were estimated by applying thin plate spline approach using the function 'estimate.missing', as complete landmark configurations are needed for subsequent procedures. The estimation of missing landmarks was done separately for extant and extinct members in order not to mix up shape variation. In extant specimens in total, three landmarks in two specimens were missing only, whereas in fossils, 105 of 672 landmarks were missing due to locally unsuitable preservation. Three fossils had preserved the entire landmark configuration. Two fossil specimens accumulated most of the missing landmarks, one having 50 and 44, respectively, missing landmarks comprised mainly of semilandmarks. A generalized Procrustes alignment (GPA) was employed with the function 'gpagen' to remove variation due to location, rotation and scale of the samples (Rohlf and Slice, 1990). Simultaneously, semilandmarks were slid by minimizing bending energy (Bookstein, 1997a;Perez et al., 2006). This resulted in a new dataset of so-called Procrustes coordinates of each landmark and centroid size (CS) for each sample. Centroid size is a measure of scale in geometric morphometrics being independent of shape and is calculated as the square root of the summed squared distances of each landmark from the centroid (Bookstein, 1997b;Zelditch et al., 2012). Ivanović and Arntzen (2017) showed that the allometric shape component explains a relatively low amount of shape variation within Salamandridae and even less within pleurodeline newts. Thus, we removed allometry, which is beyond the scope in this study, from the datasets to emphasize other potential sources of variation. Allometry-free shapes were generated by transforming the residuals from multivariate regression of shape to log(CS) using the generic function 'procD.lm' and applying these to the mean shape values. Allometry-free shapes were used to explore cranial shape. First, a principal component analysis (PCA) was performed on the covariance matrix of the Procrustes shape coordinates with the function 'plotTangentSpace'. To test the effect of species and genus on cranial shape and log(CS), we performed a Procrustes analysis of variance (ANOVA) using the function 'procD.lm'. A pairwise comparison of species and genera was carried out post hoc to clarify which species and genera were different from each other. Species with only one sample were excluded from post hoc testing. Alpha level for multiple testing was adjusted via Bonferroni correction.
For further analysis, species means were calculated from Procrustes coordinates and for log(CS). Again, a PCA was conducted on the species' mean shapes. Visualization including phylogeny TA B L E 1 Sample sizes per species of pleurodeline salamandrids for 2D geometric morphometrics analyses of cranial morphology

Tylototriton (Tylototriton) panwaensis 3 3
Tylototrition (Tylototriton) podichthys 3 3 Tylototrition (Tylototriton) shanjing 13 9 Tylototrition (Tylototriton) shanorum 6* 4 was performed using the function 'plotGMPhyloMorphoSpace', creating a plot of principal components for a set of Procrustes coordinates. Internal nodes were calculated by the squared-changed parsimony method (Rohlf, 2002;Klingenberg and Gidaszewski, 2010). We tested whether cranial shape was affected by phylogeny using the function 'physignal' on different taxonomic levels Finally, we tested whether cranial morphology correlates with ecological and reproductive biology via phylogenetic ANOVA (Procrustes ANOVA and regression models in a phylogenetic context assuming the Brownian motion model of evolution) using the function 'procD.pgls' (Adams, 2014b). We collected available data in the literature on the following traits: mating mode (amplexus, mating dance), mating habitat (terrestrial, aquatic) and oviposition site (terrestrial, aquatic) ( Figure 3). Further, species distribution area was assigned to one of the following main biomes: tropical, subtropical, temperate, Mediterranean according to Kottek et al. (2006) and Woodward et al. (2004) (Figure 3). Significance testing was performed by permutation procedures with 10,000 iterations implemented in the RRPP package (Collyer and Adams, 2018;Adams et al., 2019).

| Lateral head morphology
In terms of head morphology, principal component ( In the analysis on species means, PC1 explained 60.3% and PC2  Figure 2 were similar, as described above, but less pronounced (Figure 4b).
The subgenera Yaotriton and Tylototriton were morphologically well separated in the morphospace along the second PC axis. Echinotriton and Pleurodeles morphologically resembled Yaotriton more than

Tylototriton.
In skull morphology, PC1 explained 31.0% and PC2 22.6% of the observed variance. Pleurodeles occupied a separate area in the morphospace along the PC2 axis, whereas Echinotriton and Tylototriton overlapped ( Figure 4c). Positive PC1 scores were associated with a slender snout tip, a ventrally moved maxillary tip, a lower dorsolateral ridge and a shorter but higher occiput, which is tilted forward.
Negative PC1 scores were associated with a bulkier snout tip, dorsally moved maxillary tip and dorsolateral ridge, and a posterior elongated occiput which is orientated almost perpendicularly. Positive values on the PC2 axis were associated with an uplift and shortening of the maxillary, an elongated snout tip, a flatter skull roof slightly elongated posteriorly and a backward tilted occiput with posteriorly moved occipital condyles.
F I G U R E 5 PCA plots of GPA-aligned, allometry-free shapes of cranial morphology in dorsal view of pleurodeline newts excluding Chelotriton. Black wireframe corresponds to the mean shape, red wireframe represents the shape at the extreme value of the respective PC axes.

| Dorsal head morphology
In dorsal view, PCA on head morphology explained 25.6% on PC1 and 15.3% on PC2 of the observed variance. Although Pleurodeles occupied the morphospace at the end of PC2, it still largely overlapped with the morphospace of Tylototriton subgenera (Figure 5a).
Pleurodeles was best separated from crocodile newts along a gradient of PC2 and PC4 (10.1%, Supporting Information Figure   S2). Tylototriton subgenera overlapped with both Yaotriton and Yaotriton more than species of subgenus Tylototriton. The morphological changes along PC axes were similar to those described before.

| Morphology of fossil Chelotriton
When including fossil crocodile newts, PC1 explained 23.6% and

| Phylogeny, ecology and shape
In all datasets, a phylogenetic signal was present (Table 2. The influence of phylogeny was still strong within the genus Tylototriton, whereas within its subgenera it was only detectable among subgenus Tylototriton, and only in dorsal morphology. Shape differed interspecifically and generically in all datasets (see Table 3). Post-hoc testing revealed that all pleurodeline newt genera were morphologically distinct in cranial shape (Supporting Information Tables S2, S6, S14 and S18). Only in lateral skull view was no difference among  Tables S5 and S13). Yaotriton subgenera showed only little interspecific differences in various datasets, whereas divergence within Tylototriton subgenera was more marked (Tables S1, S5, S9 and S13).   Only mating mode was associated with the cranial shape in three of four datasets accounting for phylogeny (Table 4). In lateral morphology, species using an amplexus exhibited a smaller eye diameter, the eye being also slightly posteriorly shifted. The cranial roof was flatter in those species, whereas the posterior end of the dorsolateral bony ridges was elevated (Figure 7a,b). Species using a mating dance for copulation in general exhibited a bulkier cranium. The eye diameter was enlarged and anteriorly shifted. The cranial roof was elevated and the dorsolateral bony ridges were inclined. Furthermore, the occiput was shorter in those species. In dorsal view, main shape differences comprised more slender dorsolateral bony ridges and posteriorly moved pterygoid tips, occiput and midcranial suture among frontals and parietals in amplectant species (Figure 7c,d). In dancing species, the opposite shape changes were observed.

| Morphology and phylogeny
We investigated the cranial morphology of ribbed and crocodile newts in an integrative approach including their external head and skull morphology. To shed more light on the evolutionary history of pleurodeline newts, well-preserved fossil specimens of the closely related genus Chelotriton were included in our multivariate analyses of cranial morphology in relation to selected reproductive traits and distribution across biomes.
General shape changes in external and osteological morphology were similar. Correlations of soft and hard tissue morphometrics were shown already in another basal salamandrid salamander (Pogoda and Kupfer, 2018). However, osteology provides many more possibilities for placing precise landmarks, likely representing a better basis for evolutionary research. Ribbed newts were well separated from crocodile newts in various morphospaces previously confirmed by cranial three-dimensional morphometric analysis (Ivanović and Arntzen, 2017). Cranial shape differentiation coincides with the spatial distribution patterns, Mediterranean Pleurodeles being differentiated from the Asian Echinotriton and Tylototriton.
Nevertheless, mean shapes of the latter genera were different in all except but one of the datasets. To increase support for these results, it is suggested to add more specimens of E. chinhaiensis and E. maxiquadratus. Dubois and Raffaelli (2009)  Echinotriton resembles more the species of Yaotriton than Tylototriton subgenera, including a generally shared appearance of Yaotriton and Echinotriton: the latter exhibit only few orange highlighted body structures, e.g. tail edges, digits and parotoid tips, whereas species of Tylototriton subgenera are often more colorful (Nussbaum and Brodie, 1982;Hernandez, 2016 correlate with a peculiar defence behavior, the 'unkenreflex', which is known only from Echinotriton and Yaotriton (Brodie et al., 1984;Sparreboom et al., 2001;Gong and Mu, 2008). Other antipredator postures were described for T. (T.) verrucosus (Brodie et al., 1984).
However, to draw final conclusions about phylogenetic relationships on a larger scale, more data would be required.
The phylogenetic signal is strongest among Tylototriton, whereas within its subgenera we detected only an influence of phylogeny within subgenus Tylototriton. In the phylomorphospace,  (Zhang et al., 2007;Zhao et al., 2012).
Although the genetic divergence is quite low among the two taxa  (Table S9)  a Note that Chelotriton-shape per se is not included in the models of LH traits, as no information on these is available for this genus. The models instead concern data processing (GPA alignment) of the remaining shapes prior analysis.
components, Randeck-Chelotriton arrived closer to Tylototriton subgenera than to Pleurodeles (not shown), indicative of the newt fossil remains of the two deposits belonging to different species (already assumed by Schoch et al., 2015). Obvious morphological differences are visible among Chelotriton specimens, e.g. a shorter maxilla and missing quadrate spines in the specimen from Randeck crater lake (Schoch et al., 2015). Analysis of different skull datasets, both including and excluding Chelotriton, showed weak differences in the morphospace of extant relatives and also revealed that Chelotriton does not add much additional variation to the morphospace. Thus, the analyses of fossil newts likely allow some general conclusions to be drawn. Nevertheless, one must always keep in mind that taphonomic processes could severely alter the shape of fossil crania. Most notably, the dorsolateral ridges are displaced distally, presumably altered by taphonomy. Further, morphological traits for landmark acquisition are frequently altered or destroyed, and estimating those by algorithms never can reproduce the full truth. Chelotriton specimens cluster well together in the morphospace and specimens with estimated missing landmarks are not clustered in a specific region of it. Hence, we assume that the estimation of the missing landmarks has not led to significant alteration of the data. More detailed morphological comparisons among Chelotriton specimens remain to be made, and the four currently known species within the genus await validation (Marjanović and Witzmann, 2015). Nevertheless, we have shown that Chelotriton represents the largest bodied members within pleurodeline newts, although extant members deviate only little in their cranial size. Echinotriton andersoni from the Amami archipelago is thought to be smaller in body size than the populations from Okinawa (Utsunomiya et al., 1978;Hernandez, 2016). Our study does not support the same pattern in cranial size, although this might be due to the small sample size of Amami specimens.

| Morphology and ecology
Data on the general ecology are still scarce for crocodile newts (see also Kieren et al., 2018), although several studies deal with the reproductive ecology of Echinotriton (Utsunomiya et al., 1978;Xie et al., 2000;Sparreboom et al., 2001;Utsunomiya and Matsui, 2002;Igawa et al., 2013). Only little information is available for Tylototriton, often only from anecdotal observations (e.g. Gong and Mu, 2008;Phimmachak et al., 2015b). Although the mating mode and habitat of E. maxiquadratus are unknown, we could infer from phylogeny that its mating is terrestrial. Although various crocodile newt species are kept as pets for quite a long time, the origin and species affiliation of captive newts is often uncertain (Mudrack, 1972;Fleck, 2010a;2010b). Various observations of the mating mode or habitat in T. (T.) verrucosus in captivity are available (Rehberg, 1986;Sparreboom, 1999;Jungnickel, 2007 Echinotriton and Yaotriton is due to ecology or is constrained by phylogeny. In the latter case, Tylototriton subgenera would have reevolved aquatic oviposition, as Pleurodeles as the stem group also deposits clutches in water (Figure 3). Only the mating mode is correlated with cranial shape, simultaneously accounting for phylogeny, the different character states being randomly distributed within Chelotriton as a whole rather represents a subtropical to tropical distributed genus (Böhme, 2003;Uhl and Herrmann, 2010) as are the known distribution ranges of most extant crocodile newts. The relative long timespan between the two deposits investigated here, different climates and morphological disparity support the idea that different species were involved herein and in other European deposits. Westphal (1980) noticed that Chelotriton is rarely found in lake deposits and argued that Chelotriton was more terrestrial than other urodeles such as Tylototriton. This would coincide with our findings that Enspel-Chelotriton more closely resembles Echinotriton and Yaotriton, which mate and lay their clutches terrestrially and spend less time in aquatic habitats. A more extended study including more material of Chelotriton from other deposits would be helpful to resolve intergeneric relationships of fossil in relation to extant newt species.

| CON CLUS IONS
The different datasets (external morphology and osteology) were mostly congruent in their results, osteological datasets leading to a better separation of taxonomic units such as in Pleurodeles. With skulls, there are more possibilities to place accurate landmarks along bones and their sutures and the repeatability is higher compared with landmarks placed on soft tissues.
Cranial morphology of crocodile newts provides a congruent phylogenetic signal separating the subgenera. As we had no access to specimens of T. pseudoverrucosus, we could not draw a conclusion about the morphological distinctness and validity of Liangshantriton and rather follow the opinion placing T. taliangensis in Tylototriton subgenera. Within subgenera, phylogeny plays a minor role in the evolution of cranial shape in crocodile newts. Among reproductive traits, oviposition site was evidently correlated to phylogeny. This also supports the morphological similarity of Echinotriton and subgenus Yaotriton, both of which deposit clutches on land. It is not apparent in this case whether cranial shape represents an adaptation to ecology or rather is constrained by phylogeny. Mating mode was the only trait associated with cranial shape, simultaneously correcting for phylogeny. Climate zone had no effect on cranial shape of pleurodeline newts, confirming their quite conservative morphology (Hernandez et al., 2018).
Fossil remains were partly distorted and not fully preserved, so that retrodeformation was applied and some landmarks were virtually reconstructed. Nevertheless, the analysis of data with and without fossils revealed a similar amount of morphological variation, promoting cranial shape conservatism in crocodile newts.
But fossil Chelotriton showed a larger disparity in cranial shape and size in comparison with extant species, underpinning that Chelotriton represents a separate lineage of pleurodeline newts rather than a grade towards extant species groups (Schoch et al., 2015). Chelotriton from the deposits of Enspel and Randeck probably represent different species and cannot be assigned to C. paradoxus simultaneously. As skull morphology of fossil Chelotriton from Enspel closely resembles that of Echinotriton and Yaotriton, we conclude a more terrestrial ecology of the fossil pleurodeline newt. Further studies on the ecology of crocodile newts are urgently needed for two reasons: to better understand how ecology affects evolution of morphology and for conservation purposes, as crocodile newts are highly threatened by various factors driving them close to extinction in the near future (Rowley et al., 2010;et al.2016).

ACK N OWLED G EM ENTS
Collection-based research of PP was partly funded by the being a source of good scientific advice. We thank Andrea Villa and two anonymous reviewers for their comments and corrections, which significantly improving the manuscript. This work represents a contribution towards a PhD degree of (P.P.) at the