A new proteid salamander (Urodela, Proteidae) from the middle Miocene of Hambach (Germany) and implications for the evolution of the family

Members of the urodele family Proteidae currently account for eight extant species within two genera and at least four extinct species within three genera. The clade has a clear disjunct geographical range, with the extinct Paranecturus and the extant Necturus in North America and the extinct Mioproteus and the extant Proteus in Europe and Asia. A recent phylogenetic analysis supported a Eurasian clade including both fossil and living species found east of the Atlantic Ocean. However, the finding of a new proteid salamander, herein named Euronecturus grogu, from the Miocene of western Germany sheds new light on the evolution of this family, challenging the idea of all Eurasian members of the group deriving from a single lineage separated from the North American ones at least prior to the Oligocene. This new proteid taxon is based on five isolated atlases found in late Orleanian (MN 5) sediments in Hambach 6C, and displays features that are unknown in any other proteid, such as the presence of secondary dorsal crests, small and posteriorly‐directed postzygapophyses, and (in at least some specimens) a wide and deep ventral fossa between the anterior cotyles. A phylogenetic analysis recovered the new taxon in an early‐branching position within Proteidae, sister to all other proteids but the late Maastrichtian Paranecturus. It thus suggests the presence in Europe of a second proteid lineage, currently known only in the middle Miocene, that appears unrelated to the Mioproteus–Proteus clade.

P R O T E U S anguinus Laurenti, 1768 is the only living representative of Proteidae in Europe, showing extreme troglobiont adaptations and a distribution limited to the Dinaric Alps, in the Balkans Peninsula (Sket 1997). The other extant proteid genus, Necturus Rafinesque, 1819 (including seven extant species), is commonly found in the surface waters of a large part of the eastern United States and Canada (Vitt & Caldwell 2014). The fossil record testifies to the deep roots of the disjunct geographical range of this family and shows both the existence of various extinct genera and species and a wider past distribution. The oldest proteid remains are from the upper Maastrichtian of the Hell Creek Formation in the USA, where atlases and trunk vertebrae were attributed to the early-branching species Paranecturus garbanii DeMar, 2013, that shows affinities with Necturus. Rare fossils are also attributed directly to this latter genus, being limited to some vertebrae described as the extinct species Necturus krausei Naylor, 1978 from the upper Paleocene (Tiffanian) of Canada and remains of an unidentified species from the upper Pleistocene of Florida (Naylor 1978;Webb & Wilkins 1984). In Europe, the extant genus Proteus has a very scarce fossil record, limited only to an isolated parasphenoid from the Pleistocene of Germany attributed to the extinct species Proteus bavaricus Brunner, 1956. However, the validity of this species was questioned by different authors (e.g. Estes 1981;Bailon 1995). On the other hand, several Neogene European localities yielded remains attributed to the extinct genus Mioproteus Estes & Darevsky, 1977, that a recent phylogenetic analysis detected as the sister taxon of the extant Proteus, together forming therefore a monophyletic Eurasian group of proteids (Venczel & Codrea 2018). Mioproteus counts three species described up to now: M. caucasicus Estes & Darevsky, 1977, M. gardneri Venczel & Codrea, 2018, and M. wezei Młynarski et al., 1984. Mioproteus gardneri is the oldest species, being known from the lower Oligocene of Romania (Venczel & Codrea 2018), whereas remains attributed to M. caucasicus (or M. cf. caucasicus) are common in Miocene sites, mostly in Germany (Heizmann et al. 1980;Sach & Heizmann 2001;B€ ottcher et al. 2009;Ivanov & B€ ohme 2011), but also in Austria (Bachmayer & Szyndlar 1985) and Hungary (Bernor et al. 2004). Occurrences of this latter species reach western Asia, including the Caucasus (where it was first described) and Kazakhstan (Estes & Darevsky 1977;Malakhov 2003;Vasilyan et al. 2017). Published remains attributed to M. wezei are from the upper Pliocene of Poland, France, and Russia (Młynarski et al. 1984;Bailon 1995;Syromyatnikova et al. 2021) and from the lower Pleistocene of Moldova (Averianov 2001). Ro cek (2004) questioned the validity of this species, considering that the diagnostic features established by Młynarski et al. (1984) are in fact intraspecifically variable within Mioproteus. Syromyatnikova et al. (2021), on the other hand, recently supported its taxonomic status based on a large sample of vertebrae from Russia, even though acknowledging variation in most of the characters. Remains of unidentified Mioproteus have also been found in the upper Oligocene and Miocene of Kazakhstan (Malakhov 2003;Lopatin 1996Lopatin , 2004Vasilyan et al. 2017 We herein report on a new proteid from the middle Miocene of Hambach 6C, in western Germany, which gives new information on the past diversity of the European members of this clade.

GEOLOGICAL SETTING AND FAUNAL CONTENT
The large-scale Hambach lignite mine is situated in western Germany about 35 km west of Cologne (50°5 4 0 44″ N, 6°29 0 51″ E; Fig. 1). Geologically, the mine is situated on the tectonic Erft block within the Lower Rhine Embayment, a graben structure that has cut into the Palaeozoic Rhenish Massif since the Oligocene. The mine exposes the Miocene Ville Formation that contains the paralic Rhenish Main Seam, intercalated in a succession of interlocking beach sands of the transgressing North Sea and fluvial sediments of the Palaeo-Rhine and Meuse river system (Boenigk 2002;Sch€ afer et al. 2004). The material described here is derived from channel fill and floodplain deposits (horizon 6C according to the local lithostratigraphy) within the middle Miocene Frimmersdorf lignite seam (M€ ors et al. 2000;Sch€ afer et al. 2004).

Material
The new taxon is described based on material stored in the Steinmann Institute of the Rheinische Friedrich-Wilhelms-Universit€ at Bonn, Germany. Comparisons were made with representatives of all extant families of urodeles, based on both published literature (Estes 1981;Ratnikov & Litvinchuk 2009;Venczel & Codrea 2018; Vasilyan & Yanenko 2020) and personal observations. These latter include comparisons with specimens from the Hungarian Natural History Museum of Budapest, the Museum of Geology and Palaeontology of the University of Turin, the National Museum of Natural Sciences of Madrid and Paris, and from the Staatliche Naturwissenschaftliche Sammlungen Bayerns-Bayerische Staatssammlung f€ ur Pal€ aontologie und Geologie of Munich (see Appendix S1 for a complete list of the observed specimens). For families that were not represented in the above collections and for which no literature sources were available, specimens in online repositories (MorphoSource and DigiMorph) were used, including all available cryptobranchids, ambystomatids, dicamptodontids and rhyacotritonids (Appendix S1). If not differently stated, the anatomical terminology follows DeMar (2013).

Phylogenetic analysis
To evaluate the phylogenetic position of the new taxon, we conducted an analysis based on the matrix published by Venczel & Codrea (2018), which in turn was an update of the one originally created by DeMar (2013). LSID. urn:lsid:zoobank.org:act:0F121BFF-676B-44CC-8152-48F 595B261C1 Derivation of name. Euro in reference to its European distribution, and necturus to recognize its closer resemblance and probable phylogenetic affinity to forms such as Paranecturus and Necturus rather than to Proteus and Mioproteus. Gender is masculine.
Diagnosis. As for the type and only known species. †Euronecturus grogu sp. nov. Derivation of name. Specific epithet refers to Grogu, the Child in the popular Star Wars series 'The Mandalorian'. This recalls the fact that, like the new taxon herein described, Grogu is a member of an ancient lineage we know nothing or almost nothing about, which appears in an unexpected place at an unexpected time.
Paratypes. IPB-HaH 2120 and IPB-HaH 2150, two atlases ( Diagnosis. Euronecturus grogu is diagnosed by the following combination of features: (1) anterior cotyles dorsoventrally compressed and not confluent medially; (2) processus odontoideus with an unseparated articular surface; (3) presence of secondary dorsal crests; (4) postzygapophyses small and directed ventrolaterally. Our phylogenetic analysis recovers features 3 and 4 as possible unambiguous autapomorphies of the new taxon. The presence of a deep ventral fossa on the ventral surface of the atlas, between the anterior cotyles, is also known only in E. grogu, even though IPB-HaH 2165 clearly shows that it is not unequivocally present in all atlases referred to the taxon.
Description of the holotype. IPB-HaH 2121, a moderately small atlas (Fig. 2), is rather well preserved. It has strongly dorsoventrally compressed anterior cotyles, the articular surface of which is flat. The joints are separated by a rather narrow, short, and not gutter-shaped processus odontoideus, which has an unseparated and round articular surface. It is the only specimen referred to the new taxon preserving the dorsal part of the neural arch. The processus odontoideus is distinctly narrower than both the posterior cotyle and the neural canal. In anterior view, the neural canal has an asymmetrically-rhomboidal section. It is as wide as each anterior cotyle and twice as high as them. The neural canal clearly narrows posteriorly. Still in anterior view, the ventral edge of the processus odontoideus lays in a horizontal plane dorsal to the one containing the ventral edges of the anterior cotyles.
In dorsal view, the anterior margin of the neural arch is slightly concave. The dorsal surface carries a sharp and distinct longitudinal ridge in the middle (the neural crest; ncr in Fig. 2) and two distinctly-developed and parallel secondary dorsal crests (sdcr in Fig. 2), which are more robust and located laterally to the median one. The neural crest is low and narrow, enlarging only slightly at its posterior end. Contrariwise, the secondary crests are anteriorly enlarged and posteriorly narrower. None of these ridges touches the posterior margin, even if they get close to it. The posterior margin of the neural arch in dorsal view appears wavy and bears no notch; moreover, it extends slightly posterior to the end of the postzygapophyses with a short neural spine. The posterior cotyle is not visible in dorsal view.
In lateral view, the lateral walls of the neural arch show an anterior convexity, flanked ventrally by a small notch (incisura vertebralis cranialis; ivcr in Fig. 2) close to the anterior cotyles. The convexities partially cover the anterior cotyles in dorsal view. Still in lateral view, the dorsal surface of the neural arch is strongly anteriorly inclined. The posterior margin of the lateral walls is deeply concave and it does not reach the posterior edges F I G . 2 . Holotype atlas (IPB-HaH 2121) of Euronecturus grogu gen. et sp. nov. from Hambach 6C in: A, anterior; B, dorsal; C, right lateral; D, left lateral; E, posterior; F, ventral view. Abbreviations: acot, anterior cotyles; alpr, alar process; fsn, foramen for the first spinal nerve; ivcr, incisura vertebralis cranialis; lcr, lateral crest; ncr, neural crest; nsp, neural spine; odpr, processus odontoideus; sdcr, secondary dorsal crest; tub, tuberculum; vfo, ventral fossa. Scale bar represents 1 mm. of the cotyle. Only the left postzygapophysis is preserved, extending posteriorly beyond the cotyle for half of its total length. Dorsally to the latter, the posterior margin of the dorsal surface of the arch is slightly S-shaped in lateral view, with a posterior convexity starting from the posterior margin of the postzygapophyses and a concavity more dorsally. The short and thick neural spine (nsp in Fig. 2) is visible between the lateral flanges of the neural arch. The lateral crests (lcr in Fig. 2) of the atlas are well developed; they are sigmoid, being weakly inclined (20-30°) close to the anterior cotyles, highly inclined in the middle part (70-80°), and horizontal in the posterior trait, where they form the lateral margin of the postzygapophyses. The foramen for the first spinal nerve is present on the external surface of the lateral walls of the neural arch (fsn in Fig. 2).
The posterior cotyle is circular. Due to the accumulation of sediment inside it, it is not possible to state whether a notochordal foramen is present or not. The neural canal is roughly ogival in posterior view and as wide as the cotyle. The dorsal surface of the neural arch is rounded in posterior view, with an inverted U-shape and the thick neural spine quite evident in the middle. The preserved postzygapophysis does not project laterally in a significant way compared to the dorsal portion of the neural arch.
In ventral view, the anterior cotyles are almost perpendicular to the processus odontoideus, forming an angle of slightly more than 180°. The ventral surface of the centrum displays a large fossa at the base of the processus odontoideus (vfo in Fig. 2). The rest of the surface does not bear significant foramina, but two symmetrical and slightly sunken areas on the sides. The ventral surface is marked laterally by well-developed alar processes ('alar-like processes' in DeMar 2013; alpr in Fig. 2), starting from the posterior edge of the posterior cotyle and defining the ventral margin of each processus lateralis (sensu Ratnikov & Litvinchuk 2009). Laterally, the latter processes have some more or less large foramina and a well-developed ridge running posterodorsally. In both ventral and lateral view, the alar processes show a tuberculum (tub in Fig. 2) not far from the anterior cotyles. In ventral view, the processus odontoideus is not much extended anteriorly. A short neck separates its articular surface from the rest of the atlas, including the anterior cotyles.
Description of the paratypes. Both paratypes are almost entirely missing the neural arch but the centrum is generally well preserved. IPB-HaH 2120 and IPB-HaH 2150 (Fig. 3A, B) show similar morphologies to the holotype, except for the anterior cotyles that are oval and less dorsoventrally compressed. IPB-HaH 2150 also displays a less-distinct neck compared to the holotype.
Description of the referred material. IPB-HaH 2119 (Fig. 3C) is also missing the neural arch. The preserved centrum is ruined in its left part, missing the left anterior cotyle and the left half of the articular surface of the processus odontoideus. The overall morphology is similar to the holotype, but the neck of the processus odontoideus is longer so that the articular surface of the processus is further away from the anterior cotyles. Moreover, the articular surface of the single preserved joint is slightly concave instead of flat. IPB-HaH 2165 (Fig. 3D) also has a morphology of the occipital area similar to the holotype, but ventrally the alar processes are less laterally extended, and they show no tuberculum close to the anterior cotyles. Moreover, in this specimen, the deep fossa at the base of the processus odontoideus on the ventral surface is absent, whereas two smaller foramina are present more posteriorly. As in IPB-HaH 2119, the neck is slightly longer than in the type material. Most of the posterior cotyle of IPB-HaH 2165 is broken off. . It cannot be referred to salamandrids due to not having a gutter-shaped processus odontoideus with separated articular surfaces and due to the neural arch extending posterior to the postzygapophyses. Amphiumids have a thicker neural crest and a larger size. As far as fossils urodeles are concerned, the Hambach taxon differs from the Mesozoic taxa of Europe (mentioned and described by Evans & McGowan 2002) in having flat anterior cotyles (except perhaps in IPB-HaH 2119) and a thick processus odontoideus with a well-defined neck. The main differences between the Hambach taxon and the extinct family Batrachosauroididae are the flat articular surface of the anterior cotyles and the more rounded and developed processus odontoideus (Naylor 1983;Evans & McGowan 2002). A single species belonging to this family, Palaeoproteus miocenicus Vasilyan & Yanenko, 2020 from the upper Miocene of Europe, has a very long processus odontoideus, which is however much longer than that of the new Hambach taxon. The latter differs from P. miocenicus also in the more dorsoventrally compressed anterior cotyles, the relatively longer centrum, and the shape of the processus odontoideus, that is lip-shaped and anteroventrally bent in P. miocenicus and rounded and knob-like in E. grogu. The general morphology of the atlases described herein closely resembles the North American taxon Paranecturus garbanii, from the upper Maastrichtian of Montana. However, atlases attributed to this species (including the holotype) show a neural canal that is partly flanked laterally by the anterior cotyles, rather than being completely above them as in the Hambach urodele. Also, the latter shows a deep ventral fossa at the base of the processus odontoideus and dorsoventrally compressed anterior cotyles, both features lacking in Paranecturus. Euronecturus grogu also differs from all known extant proteids (and extinct scapherpetontids; J. Gardner pers. comm. 2021) in the presence of secondary dorsal crests. However, this character is unknown for Mioproteus, Paranecturus, and most fossil urodeles, because the dorsal part of the neural arch is very rarely preserved.

PHYLOGENETIC ANALYSIS
Our phylogenetic analysis resulted in a single tree of 59 steps (Fig. 4). Consistency (CI) and retention (RI) indexes are 0.593 and 0.676, respectively. The analysis was able to recover a monophyletic Proteidae, with P. garbanii as the most early branching taxon. Euronecturus grogu is then sister to a clade including the extant Necturus maculosus (Rafinesque, 1818) and all other European proteids. Mioproteus species form a monophyletic group, which is sister to P. anguinus. Bremer support is, however, low for all nodes (Fig. 4).

DISCUSSION
Although most of the extant species of Proteidae are currently widespread in the eastern part of North America, the fossil record of this group there is limited, with only three occurrences reported from the area (Fig. 5A). In particular, North America hosts the oldest proteid record: Paranecturus garbanii from the upper Maastrichtian of the Hell Creek Formation (DeMar 2013). In this context, the finding of E. grogu in the lower middle Miocene of Hambach 6C is of exceptional relevance because of its clear resemblance to this Cretaceous, North American relative. In fact, E. grogu and P. garbanii share a similar morphology of the atlas, especially in the well-developed processus odontoideus with an unseparated articular surface and a short neck, as well as in the flat anterior cotyles. A comparison between E. grogu and N. krausei from the upper Paleocene of Canada to evaluate possible affinities was not possible, because the latter is only represented by trunk vertebrae. As a matter of fact, a detailed scrutiny of the amphibian fossils from Hambach 6C looking for skeletal elements other than atlases with a proteidlike morphology that could be attributed to E. grogu was unsuccessful. Several extant taxa and groups currently show similarities between North America and Europe (e.g. proteids themselves, or the plethodontid genera Hydromantes and Speleomantes; see Carranza et al. 2007) or showed them in the past (e.g. the extant plant genus Lycopus; see Martinetto et al. 2021).
However, the new species from Hambach is particularly significant for the evolution of the family Proteidae when considering the result of our phylogenetic analysis (Fig. 4), revealing its early-branching position. This finding has notable biogeographical implications, especially considering the more derived state of the extant N. maculosus. According to our analysis, the other European proteids (the extant Proteus and the extinct Mioproteus) are, therefore, more closely related to the American Necturus than to Euronecturus. It is interesting to note that proteids were generally widespread during the Neogene in Eurasia (Fig. 5B) (Estes 1981;Venczel & Codrea 2018), including the evaluation of a possible synonymy with one of the species of either. Mioproteus shares a very close morphological similarity with the extant Proteus and they form a monophyletic group in our analysis. Thus, they are most likely to have been derived from a single lineage of proteids widespread in Eurasia since at least the late Palaeogene. Whether this lineage entered the Palaearctic through a land bridge connecting it with the Nearctic (Beringia, the De Geer route, or the Thulean route; see Brikiatis 2014, for an extensive review on the topic) sometime before the Oligocene (when M. gardneri was already present in Europe; Venczel & Codrea 2018) or originated in Eurasia from a more ancient stock separated from its North American relatives when the two continents split, is difficult to ascertain at the moment, in particular due to the very scarce record of fossil Palaeogene proteids and especially Asiatic ones.
The phylogenetic distance between the Mioproteus-Proteus lineage and Euronecturus suggests that the latter is not related to the same colonization or speciation event as the former. Moreover, atlases of Euronecturus are the only remains that can be unambiguously attributed to a proteid salamander in Hambach 6C, thus indicating that this genus did not co-occur with any other member of the clade in the only locality where it is currently known. Again, Euronecturus could be a relict member of another lineage whose roots have to be sought in the Mesozoic or Palaeogene, when North America and Eurasia still presented shared faunal elements (as is the case for lizards, for example; Rage 2013). Alternatively, it could have originated from another wave of proteids entering Eurasia from North America sometime before the middle Miocene. If the colonization option is true for at least one of the two lineages, the west Eurasian distribution of both extant and extinct taxa in this part of the world, as well as the absence of fossil proteids in eastern Asia, may suggest that these urodeles did not reach the Palaearctic through Beringia. De Geer and Thulean routes may be more plausible. These two routes connected Greenland with west Asia and Fennoscandia, and Great Britain, respectively. However, these considerations should be treated only as speculative at the moment, given how poor our current knowledge of extinct proteids is.
In any case, Euronecturus has not been detected in any locality other than Hambach up to now, contrary to Mioproteus, whose fossil record is rich and widespread. This disparity could reflect an effective unevenness of the past distribution of the two genera, with Mioproteus having a wider geographical range. However, other options should also be considered, including taphonomic or geological biases, such as living habitats with different fossilization chances or different geographical ranges (i.e. Euronecturus could have lived in areas of Europe with scarce outcrops from the corresponding age). Be that as it may, more fossil proteids are needed, especially from the Palaeogene, to better understand the dynamics of the group and the origin of the different subclades and lineages.  1977, Heizmann et al. 1980, Młynarski et al. 1984, Bachmayer & Szyndlar 1985, Bailon 1995, Lopatin 1996, Averianov 2001, Sach & Heizmann 2001, M€ ors 2002, Malakhov 2003, Bernor et al. 2004, B€ ottcher et al. 2009, Prieto et al. 2009, Ivanov & B€ ohme 2011, Vasileiadou et al. 2017, Vasilyan et al. 2017, Ivanov et al. 2018, Venczel & Codrea 2018, B€ ohme et al. 2019, Syromyatnikova et al. 2021. Given the uncertain nature of Orthophyia Meyer, 1845 (see main text) this specimen is here indicated by 'Proteidae indet.'

CONCLUSION
The new proteid Euronecturus grogu is herein described based on five atlases from the German middle Miocene locality of Hambach 6C. Atlases belonging to this species are diagnosed by the combination of: anterior cotyles dorsoventrally compressed and not confluent medially; processus odontoideus with an unseparated articular surface; presence of secondary dorsal crests; and postzygapophyses small and directed ventrolaterally. Our phylogenetic analysis recovers the previously-known European clade of proteids (including the genera Mioproteus and Proteus) as the sister group to the extant American genus Necturus, whereas the new Hambach taxon has a more early-branching position within the clade Proteidae. The closest resemblance of the atlases of E. grogu is with the oldest known proteid Paranecturus garbanii from the upper Maastrichtian of North America. The phylogenetic distance between the Mioproteus-Proteus clade and Euronecturus, as well as the similarity between the latter and Paranecturus, suggest that two different proteid lineages lived in Europe during the Cenozoic. One of these two lineages (the Euronecturus one) is currently known only from the few fossils from Hambach 6C, and therefore future finds, especially from the Palaeogene, are required to disentangle the evolution and biogeographical origin of this enigmatic group of urodeles.
Acknowledgements. We gratefully acknowledge the support of the mining company RWE POWER AG (Cologne) during fieldwork at the Hambach locality. S. Bailon, J. V€ or€ os, and S. Schweiger kindly offered assistance, during our visits to the collections of Paris, Budapest and Vienna respectively. M. Calvo and O. Rauhut gave access to the collections of Madrid and Munich respectively. Special thanks go to J. Gardner and M. Delfino who gave valuable suggestions and critical comments on the material, as well as to Arnau Bolet, Antonio Borrani, and Emanuel Tschopp who helped with the phylogenetic analysis. We further thank the teams of Morpho-Source and DigiMorph, for their work, as well as the editor Philip Mannion, the technical editor Sally Thomas, and the two reviewers M arton Venczel and Davit Vasilyan, for useful comments that helped improve an earlier version of the manuscript.
LM received support from Fondi di Ateneo dell'Universit a di Torino