A new genus of Rhinocerotidae (Mammalia, Perissodactyla) from the Oligocene of Europe

A newly discovered, well-preserved skull and associated fragment of a juvenile mandible from the Early Oligocene locality of Poillat (Canton Jura, NW Switzerland), bearing close affinities with the rhinocerotid Protaceratherium albigense (Roman, 1912), are attributed to a new small-sized representative of early diverging Rhinocerotinae, Molassitherium delemontense gen. et sp. nov. Other specimens from Western Europe, formerly questionably referred to Epiaceratherium Abel, 1910, are assigned to this new genus. Comparison with the previously described Protaceratherium Abel, 1910 (including type material) and a phylogenetic analysis highlight the mismatch of Protaceratherium minutum (Cuvier, 1822) and Protaceratherium albigense (Roman, 1912). Given the topology of the most parsimonious tree, a basal split within Rhinocerotidae coincides with the well-supported divergence of the Elasmotheriinae and Rhinocerotinae clades. Relationships within Rhinocerotinae are [Epiaceratherium bolcense Abel, 1910 [Epiaceratherium magnum Uhlig, 1999 [Molassitherium gen. nov. [Mesaceratherium Heissig, 1969 [Pleuroceros Roger, 1898 [Protaceratherium minutum (Cuvier, 1822) [Plesiaceratherium mirallesi (Crusafont, Villalta and Truyols, 1955) [Aceratheriini, Rhinocerotini]]]]]]]]. The only paraphyletic genus in the analysis is Epiaceratherium, with the earliest Oligocene Epiaceratherium bolcense Abel, 1910 being sister taxon to an [Epiaceratherium magnum Uhlig, 1999, Rhinocerotinae] clade. In the single most parsimonious tree, Molassitherium gen. nov., included within the early diverging Rhinocerotinae, forms a clade encompassing Molassitherium delemontense gen. et sp. nov. and the type species Molassitherium albigense comb. nov. The range of Molassitherium delemontense gen. et sp. nov. is so far restricted to the late Early–early Late Oligocene interval in Western Europe (Germany, Switzerland, France; ‘late MP22’–MP25).


Introduction
established the genus Protaceratherium for the small slender rhinoceros P. minutum (Cuvier, 1822) from the Early Miocene of Europe, previously assigned to Diceratheriinae Dollo, 1885 under the name Diceratherium minutum by Osborn (1900). Roman (1912) described the species Acerotherium albigense, reassessed as P. albigense by von Breuning (1924), on the basis of an anterior part of an adult skull, with preserved left and right P1-M3, discovered in the molassic deposits of the early Late Oligocene of La Sauzière Saint-Jean (MP25-26; SW France). The affinities and the suprageneric assignment of Protaceratherium species have been discussed for a long time. Heissig (1969) considered them as Dicerorhininae Simpson, 1945, likewise Spillmann (1969, who even suggested Protaceratherium as junior synonym of Diceratherium Marsh, 1875 (Dicerorhininae), whereas Cerdeño (1995) * Corresponding author. Email: damien.becker@jura.ch proposed a synonymy with Plesiaceratherium Young, 1937 (Aceratheriinae Dollo, 1885). On the other hand, Heissig (1973) was indecisive between Caenopinae Cope, 1887 and Aceratheriinae but suggested in 1989 assignment to Menoceratini Prothero et al., 1986, which he considered as a tribe within Aceratheriinae. Indeed, many authors have regularly attributed the genus Protaceratherium to the subfamily Aceratheriinae (e.g. von Breuning 1924  We report here a recently discovered, well-preserved skull and associated fragment of a juvenile mandible attributed to a new small-sized representative of Rhinocerotinae very close to P. albigense, Molassitherium delemontense gen. et sp. nov., from Poillat, a new Early Oligocene vertebrate locality within the Delémont valley (Canton Jura, NW Switzerland; Fig. 1). We include this sample as a terminal taxon in a cladistic analysis in order to establish its phylogenetic relationships, notably with other European Oligocene and Miocene rhinocerotids.

Material
The referred type material is stored in collection PAL A16 of the Natural History Museum of Canton Jura in Porrentruy, Switzerland (Musée jurassien des sciences naturelles). Large mammal remains were quarried in 2007 at Poillat in the Delémont valley (Canton Jura, NW Switzerland), during construction of motorway A16 (Transjurane) and small mammal teeth were discovered by screening washing the deposits from the same fossiliferous level (c.350 kg).
The additional referred specimens of this study include dental remains from Offenheim (Germany), Kleinblauen (Switzerland) and Monclar-de-Quercy (France), attributed by Uhlig (1999) and Becker (2009) to Epiaceratherium aff. magnum, and also specimens from Habach 5, attributed by Göhlich (1992) to Epiaceratherium sp. and by Uhlig (1999) to cf. Epiaceratherium sp. The specimens from Kleinblauen and from Monclar-de-Quercy have been reviewed based on the direct observations of the specimens housed in the Naturhistorisches Museum Basel (Switzerland) and the Muséum d'Histoire naturelle de Toulouse (France), respectively. Data on specimens from Offenheim (stored in the Hessischen Landesmuseum, Darmstadt, Germany) and Habach 5 (stored in the Bayerische Staatssammlung für Paläontologie und Historische Geologie, Munich, Germany) are based on the work of Uhlig (1999).

Stratigraphical context
The mammal remains from Poillat were trapped in Rupelian sandy deposits corresponding to the transition between the brackish lower part and the continental upper part of the 'Molasse alsacienne' (USM: Lower Freshwater Molasse). The general stratigraphical context of this Jura Molasse Formation (termed NW Swiss Molasse Basin) was described in previous works (Picot. et al. 2008;Becker 2009). The succession consists of a lithofacies assemblage (tabular sandy beds with sigmoidal or planar crossstratifications, erosional sandy beds with low angle trough cross-stratifications or tough cross-stratifications, massive fines) typical of a coastal to alluvial floodplain controlled by a mouth complex of distributary channels, interdistributary bays and tidal bars, and by sandy channels and muddy floodplains. The new specimens reported in this paper originate from a sandy mud pebble channel.

Anatomical terminology and characters
Dental terminology follows Heissig (1969), Uhlig (1999) and Antoine (2002). Dental and osteological features described correspond basically to cladistic characters used and listed by Antoine (2002). Measurements were made according to Guérin (1980) and are given in mm.

Phylogenetic relationships
The dataset (character list, character states) derives from that of Antoine (2002Antoine ( , 2003 and Antoine et al. (2003bAntoine et al. ( , 2010. It was reduced to 214 morphological characters (36 cranial, eight mandibular, 85 dental and 85 postcranial), as 68 characters from the original matrices were phylogenetically uninformative for the present taxonomic sample and therefore were removed prior to the analysis. The character listing and the data matrix can be found in Appendices 1 and 2.
Character coding sources, through direct observation and/or the literature, are given in Online Supplementary Material. Thirty terminal taxa were included in the phylogenetic analysis. Three terminals were selected as outgroups: the extant tapirid Tapirus terrestris Linnaeus, 1758, the Eocene hyrachyid rhinocerotoid Hyrachyus eximius Leidy, 1871 and the Eocene stem rhinocerotid Trigonias osborni (Lucas, 1900) from North America.
The in-group sensu lato consists of both taxa of interest (in-group sensu stricto) and selected terminals forming a 'branching group', sensu Antoine (2002) and Orliac et al. (2010). The in-group sensu stricto includes the earliest European rhinocerotid Ronzotherium filholi (Osborn, 1900) (from the earliest Oligocene of Europe) and an exhaustive specific sampling for Pleuroceros Roger, 1898 (with P. pleuroceros (Duvernoy, 1853) and P. blanfordi (Lydekker, 1884) (Cuvier, 1822), from the Early Miocene of Western Europe, and P. albigense (Roman, 1912), from the 'middle' Oligocene of Europe, were also considered in the analysis in order to test the monophyly of the concerned genus, recently challenged in the phylogeny proposed by Antoine et al. (2010).
The branching group includes (1) type species or well-represented species of type genera of suprageneric groups recognized within Rhinocerotidae; and (2) early representatives of these suprageneric groups, in order to branch the taxa of interest within Rhinocerotidae, to define their generic and suprageneric affinities, and to avoid long-branch attraction artefacts due to parallelism (e.g. late representatives of Elasmotheriinae versus Rhinocerotinae; Antoine 2002 (Osborn, 1900), from the late Early Miocene of Europe; Aceratheriini: Aceratherium incisivum Kaup, 1832, Alicornops simorrense (Lartet, 1851) and Hoploaceratherium tetradactylum (Lartet, 1851), from the middle and/or late Miocene of Europe; the hornless rhino Plesiaceratherium mirallesi (Crusafont et al., 1955), from the late Early Miocene of Europe).  Occurrence. Late Early Oligocene ('late MP22'-MP25) of Western Europe (Germany, Switzerland and France).

Description
Skull. The well-preserved skull (MJSN POI007-245) belongs to a small-sized adult rhinoceros. It lacks the basioccipital, the nasal tip, as well as the premaxilla and the anterior dentition. The nasal bones have a forked tip in dorsal view but they do not display any lateral apophysis on their ventral edge in lateral view. The foramen infraorbitalis is above P3. The nasal notch is wide, deep and U-shaped, reaching the P3/4 limit. The anterior border of the orbit is above the anterior part of M2. There is neither septum ossification nor lateral projection of the orbit. The nasal/lacrymal suture is long. The jugal/squamosal suture is smooth. A weak lacrimal process is present and the postorbital process is absent. The anterior base of the zygomatic process of the maxilla is low, beginning less than one centimetre above the neck of M2. The zygomatic arch is high (nearly reaching the level of the cranial roof) and fairly developed. It forms a thin sigmoid strip, without postorbital process. The dorsal profile of the skull is flat. The sphenorbitale and rotundum foramens are not observable and the area between the temporal and nuchal crests is depressed. The external auditory pseudomeatus is open ventrally. The occipital side is inclined up and backwards with a very acute angle and a developed nuchal tubercle. The posterior margin of the pterygoid is nearly horizontal. The nasal bones are totally separate from one another by a shallow median groove. They are straight, short and triangular, and lacking any vascular print or domed structure indicating the presence of nasal horn(s), although the tip is lacking. Also, despite the lack of the premaxilla, it can be assumed that the skull was dolicocephalic (maximum zygomatic width/nasal-occipital length ratio < 0.50). There is no evidence for any frontal horn. The frontal bones are wide with respect to the zygomatic bones (zygomatic width/frontal width ratio = 1.34). The fronto-parietal crests are sharp and salient. They are joined (constricted) in their posterior halves, only separated by a strongly constricted groove (c.2 mm wide), and then slightly separated, forming a weak dome just prior the occipital crest. The latter is strongly concave, deeply forked and narrow (c.106 mm).
In palatine view, the anterior end of the zygomatic process of the maxilla progressively diverges from the curvature of the tooth row and distally becomes parallel to the skull axis. The palate is quite wide. The palatine fossa reaches mid-length of the M2 and the vomer is acute. The glenoid cavity (fossa mandibularis) is flat and forms a  smooth semi-cylinder in lateral view. The articular tubercle of the squamosal is high with a concave transverse profile. The foramen postglenoideum is distant from the postglenoid process. The latter is robust and its articular surface is angular in cross-section. The post-tympanic and paraoccipital processes are well developed. There is no posterior groove of the processus zygomatic.
Mandible. The corpus mandibulae of the fragmentary juvenile specimen MJSN POI007-268 bears a lingual groove and seems to have a straight ventral profile.

Dentition.
Only the distal part of the diastema (3 cm long) is preserved on skull MJSN POI007-245; it shows neither canine nor incisor alveolus. The premolar series is long when compared to the molar series (LP3-4/LM1-3 ratio = 0.54). The P1-3 and M1 of skull MJSN POI007-245 are much worn, precluding detailed observation. The upper cheek teeth (except P1) are characterized by an internal wall strongly inclined labially. The dental structures are simple and there are no secondary enamel folds or cement on the crowns. The enamel is thin and wrinkled. The crowns are low (brachydont teeth) and the roots are long, distinct and divergent. The crochet and the medifossette are always absent and the postfossette narrow. The paracone fold is constant and thick on P2-M3, vanishing before the neck and thus not visible on very worn teeth. The parastyle is sagittally oriented, more developed on upper molars than on premolars. The metacone fold is weakly developed on P2-4, absent on M1-2 and fairly distinct on M3. The mesostyle is smooth on P2-4 and very faint on M1-2. There is a very thin continuous labial cingulum on P2-4 of MJSN POI007-245, running all along the cervix. This labial cingulum tends to be reduced on M1-2 and it is restricted to a strong distolabial spur on M3. It is reduced in specimens from Offenheim, Grafenmühle 11 and Monclarde-Quercy; by contrast, it is rather developed in specimens from Kleinblauen and Habach 5. The lingual cingulum is always present and strong: it is continuous on P2 (weaker under the protocone and hypocone), continuous to reduced under the protocone on P3-4 and reduced under the protocone and the hypocone on M1-3 (restricted to an enamel bridge at the lingual opening of the median valley).
P1 is two-rooted and trapezoidal in occlusal view (mesially tapered and approximately the same length as distal width), with a rectilinear lingual side and a rounded mesiolabial side. It is much narrower than P2 and bears a lingual groove on the protoloph and a thin lingual cingulum in its mesial half. The protocone and the hypocone on P2 are joined by a lingual bridge (semi-molariform pattern sensu Heissig 1969). The protocone is less developed than the hypocone. The protoloph is thin but continuous and widely connected with the ectoloph, and the metaloph is transverse. P3-4 display a lingual wall marked by a smooth vertical groove (P3 being semi-to submolariform, P4 subto premolariform sensu Heissig 1969) and taper distally, especially P4, with a transverse metaloph shorter than the protoloph. Additionally, P3 and P4 bear a smooth crista and a weak anterior groove on the protocone (particularly visible on P3-4 NMB KB7/1-7/2 from Kleinblauen) and the P3 NMB KB7/1 from Kleinblauen possesses an antecrochet. There is no pseudometaloph on P3 (sensu Antoine 2002).
The upper molars have a median valley with a labial pit. M2 is larger than M1. Both the metastyle and the metaloph are long and the posterior part of the ectoloph is concave on M1-2 (no metacone fold). The metaloph is constricted on M1-2: there is a mesiolingual groove on the hypocone of M1-2. There is a strong constriction of the protocone and a strongly developed anterochet on M1-3. The postfossette is deeper than the distal cingulum. M3 has an ectometaloph (resulting from the fusion of the ectoloph and the metaloph), a quadrangular occlusal outline, a strong bump-shaped posterior cingulum (metastyle artefact), a faint metacone fold and a smooth posterior groove on the lower part of the crown. The protoloph of M3 is transverse and straight and the protocone is trefoil-shaped.
The lower milk teeth do not exhibit any constriction of the metaconid. The d3 shows a strong protoconid fold and a forked paralophid. The mesial branch of the latter seems  . 2010). Also, the referred upper cheek teeth differ from those of Protaceratherium minutum by showing a labial cingulum present on the upper premolars, no crochet on the upper premolars, a protoloph joined to the ectoloph on P2, an antecrochet always present on the upper molars, no metacone fold but a long metaloph on M1-2, a mesostyle on M2, a constricted protocone on M3 and a posterior groove on the ectometaloph of M3. Additionally, the skull from Poillat displays both a narrow zygomatic arch with a low anterior base of the zygomatic process distally becoming parallel to the skull axis and a wide and deep U-shaped nasal notch, which are reminiscent of the North American early diverging rhinocerotid Trigonias Lucas, 1900. However, the latter differs in its concave dorsal profile, elevation of the occipital crest, a somewhat concave occipital crest and a more advanced molarization of the upper premolars (Wood 1932;Prothero 2005).
The referred specimens share with Epiaceratherium the following features: P2-M3 with convergent lingual and labial walls, upper premolars with a metacone fold always developed and a lingual cingulum strongly developed, upper molars with a strong paracone fold and a restricted lingual cingulum at the lingual opening of the median valley, and an M2 larger than M1 (Uhlig 1999). However, Epiaceratherium displays a wide postfossette on P2-4 and a metacone fold and the posterior part of the ectoloph straight on M1-2. Epiaceratherium bolcense differs by being smaller, by having a proportionally narrower nasal notch, a posterior margin of the pterygoid nearly horizontal, less molarized upper premolars, the presence of a metaconid constriction on lower milk teeth, and also by lacking any metaloph constriction on P2-4, and any protocone constriction and antecrochet on upper molars. On the other hand, the referred specimens are of similar size and have numerous similarities with those of E. magnum, such as a strong lingual cingulum elevated under the main cusps and a distinct mesostyle on P2-4 (Uhlig 1999;Becker 2009). However, the absence of cement on adult cheek teeth, the slightly more advanced molarization of the upper premolars, the more developed paracone fold on P2-M3, the anterior protocone groove and the lingual wall marked by a smooth groove on P3-4, the absence of crista on upper molars, the constricted protocone, the strongly marked antecrochet the concavity of the posterior part of the ectoloph on M1-2, the quadrangular M3 with distinct ectoloph and metaloph, a metastyle artefact, a faint metacone fold, a constricted protocone and a smooth posterior groove on the lower part of the crown, as well as a protoconid fold on d3, make them distinct from E. magnum. On the other hand, most of these characters are also described on the upper cheek teeth specimens from Offenheim (Germany), Kleinblauen (Switzerland) and Monclar-de-Quercy (France) attributed to E. aff. magnum by Uhlig (1999) and Becker (2009), and partly on the specimens from Habach 5 (Germany) attributed to cf. Epiaceratherium sp. by Uhlig (1999). Only the P2 from Weissenburg 16 (Germany, MP21?; Uhlig 1999) attributed by Uhlig (1999) to cf. Epiaceratherium sp. differs in being more primitive (submolariform) and smaller, and by having a straight ectoloph profile. This specimen, despite the presence of a labial cingulum, seems to be referable to E. bolcense (Dal Piaz 1930;Uhlig, 1999).
Finally, most of the characters observed on all referred specimens, such as an occipital side of the skull inclined up and backwards, a posterior margin of the pterygoid nearly horizontal, short nasal bones, a forked occipital crest, an acute vomer, a labial cingulum usually present on the upper premolars, as well as a strongly marked antecrochet, a crochet always absent, a constriction of the protocone present, and a posterior part of the ectoloph concave on upper molars, a weakly developed mesostyle on M2, and a quadrangular M3, point to strong similarities with Protaceratherium albigense (Spillmann 1969; Lihoreau et al. 2009). However, the specimens from Poillat and the additional referred material can be distinguished from the latter by having smaller dimensions (especially the M2), a high anterior base of the zygomatic process distally widely diverging, a narrower nasal notch and no labial pit of the median valley on the upper molars (Roman 1912;Hugueney & Guérin 1981;Uhlig 1999;Lihoreau et al. 2009), as well as additional features discussed in the phylogenetic analysis section below.
Regarding the lower cheek teeth from Offenheim, attributed by Uhlig (1999) to E. aff. magnum, they display a trigonid with an acute dihedron and a posterior valley lingually open on p2 similar to P. albigense but differ in being larger and lacking any labial and lingual cingula. As described above, the smaller dimensions and the reduction of the cingula can also be observed on the upper cheek teeth to distinguish P. albigense from the referred material. The association of these lower cheek teeth from Offenheim with the upper ones as proposed by Uhlig (1999) is probably correct, leading us to list them in the additional material referred to Molassitherium delemontense gen. et sp. nov. However, as the specimens are scarce and their specific assignment is insufficiently constrained, we have excluded them from the phylogenetic analysis.
As a result, we favour assignment of the specimens from Poillat and the additional referred material from Germany, Switzerland and France, attibuted to Epiaceratherium aff. magnum and to cf. Epiaceratherium sp. by Uhlig (1999) and Becker (2009) Tables 1 and 2 for comparisons of cranial dental measurements).

Phylogenetic relationships
Only one most parsimonious tree (1117 steps; Consistency Index (CI) = 0.26; Retention Index (RI) = 0.48) was  obtained by using the 'mh * bb * ' command of Hennig86, 1.5 (Farris 1988) and the heuristic search of PAUP 4.0v10 (unweighted parsimony; branchswapping TBR, 1000 replications with random taxa addition, 100 treeholds by replication; Swofford, 2002). This tree is shown in Fig. 6. Branch support, assessed by calculating the Bremer indices (Bremer 1994), is indicated below the branches in Fig. 6 (italicized), while the number of unambiguous synapomorphies (detailed in Table 3) appears above the branches, both left of the corresponding node. Nodes discussed in the text are designated by a letter, right of each node in the same figure (Fig. 6). Suprageneric relationships within Rhinocerotidae are consistent with other recent phylogenies, such as those proposed by Antoine et al. (2010; based on a similar taxonomic sample) and, to a lesser extent, Antoine (2002) and Antoine et al. (2003a). The early rhinocerotid Trigonias osborni is remote from other Rhinocerotidae (Fig. 6). A basal split within Rhinocerotidae coincides with the well-supported divergence of the Elasmotheriinae and Rhinocerotinae clades (Fig. 6,   of key early taxa referred either to Plesiaceratherium, Teleoceratina or Rhinocerotina. Based on data from Heissig (1969Heissig ( , 1989Heissig ( , 1999, Antoine et al. (1997Antoine et al. ( , 2003bAntoine et al. ( , 2006Antoine et al. ( , 2010Antoine et al. ( , 2011, Uhlig (1999) Fig. 6, node I). Plesiaceratherium mirallesi shares 11 synapomorphies with the clade Aceratheriini + Rhinocerotini (Bremer Index = 4; Fig. 6, node J), such as a low zygomatic arch (RI = 0.70), a triangular M3 in occlusal view (RI = 1), lower cheek teeth with a rounded trigonid (RI = 0.75) and kidney-like condylar facets on the atlas (RI = 1). The clade Aceratheriini + Rhinocerotini is well supported, with eight unambiguous synapomorphies and a Bremer Index > 5 (Fig. 6, Fig. 6). Detailing the distribution of synapomorphies within Aceratheriini and Rhinocerotini is beyond the scope of the present article. Ghost lineages are inferred within Rhinocerotinae (Fig. 7), due to the absence in the present taxonomic sampling of early terminals such as the earliest representative of Plesiaceratherium (P. naricum, earliest Miocene of Pakistan; Antoine et al. 2010), the teleoceratine Diaceratherium massiliae (MP26, early Late Oligocene; Ménouret & Guérin 2009) or the late Miocene twohorned rhinocerotines Stephanorhinus pikermiensis and Ceratotherium neumayri (e.g. Heissig 1999; Antoine & Saraç 2005). Including these taxa has no consequence on the topology of the parsimonious tree but lowers the Consistency Index.

Conclusion
Given the topology of the most parsimonious tree and the strong support of all nodes (24 synapomorphies; 2 ≤ Bremer Indices ≤ 4), the referral of 'Acerotherium albigense Roman, 1912' to the genus Protaceratherium Abel, 1910, can be discounted. On the other hand, the small hornless rhinocerotid from the 'middle' Oligocene of Europe forms a well-supported clade with the Delémont rhinocerotid, described here, which leads us to propose a new monophyletic genus, Molassitherium gen. nov., encompassing the two taxa under the names M. albigense (Roman, 1912) comb. nov. and M. delemontense sp. nov. Molassitherium gen. nov. is clearly distinct from coeval but less derived rhinocerotids such as Ronzotherium filholi, Epiaceratherium bolcense and E. magnum, from the more derived (and younger) representatives of Mesaceratherium, Pleuroceros and Plesiaceratherium, as well as from Protaceratherium minutum (the type species of Protaceratherium).
Also, this work highlights the mistaken identifications for a decade of 'Epiaceratherium magnum Uhlig, 1999', because the genus Epiaceratherium obviously appears as paraphyletic in the cladogram. Following the principle of priority, this implies that Epiaceratherium can be considered as a monospecific genus for the type species E. bolcense Abel, 1910, whereas 'Epiaceratherium magnum Uhlig, 1999' should be assigned to a new genus incertae sedis.