Neochorlakkia myaingensis n . gen . , n . sp . , a new Dichobunidae ( Mammalia , Cetartiodactyla ) from the middle Eocene Pondaung Formation

The diversity of basal ungulates is significant in the late middle Eocene Pondaung Formation (Myan­ mar). However, the precise identification or attribution of some fossils is sometimes difficult because of the scarcity and poor preservation of the material. We describe here a new genus and species of a Dichobunidae from Paukkaung Kyitchaung 2 locality, that is morphologically and probably phylo­ genetically close to a taxon from the early­middle Eocene of Pakistan, and we discuss the attribution of dental material to Myanmarius chitseini Tsubamoto, Egi, Takai, Htike & Zin­Maung­Maung­Thein, 2013, an inderterminate artiodactyl from Pondaung. This study also confirms that dispersals were possible between the Indian subcontinent and Southeast Asia during the first part of the Eocene.


INTRODUCTION
The late middle Eocene Pondaung fauna is characterized by its rich diversity of perissodactyls, anthracothere artiodactyls, and primates. However, recent studies have revealed that a significant number of basal "ungulates" also diversified in this area and represents a large proportion of the mammalian community (Ducrocq et al. 2000;Métais 2006;Métais et al. 2007;Tsubamoto et al. 2012;Ducrocq et al. 2016;Ducrocq 2019), most of them being represented by fragmentary remains and thus of uncertain affinities. Particularly, Tsubamoto et al. (2005: fig. 2B, C) described two upper molars from Pondaung that they referred to an indeterminate artiodactyl. Theodor et al. (2007) later suggested that this material might be related to the Raoellidae, a family of Asian small Eocene artiodactyls closely related to the Cetacea. In their revision of the dental features of the family Raoellidae, Orliac & Ducrocq (2012) demonstrated that the molars described by Tsubamoto et al. (2005) was not part of the raoellid clade, but was probably more closely related to the Suoidea. Tsubamoto et al. (2013) then reexamined this material, creating the genus and spe cies Myanmarius chitseini Tsubamoto, Egi, Takai, Htike & ZinMaungMaungThein, 2013, and attributed two addi tional upper molars and a fragmentary and worn distal part of a lower molar to this taxon. According to them, Myanmarius Tsubamoto, Egi, Takai, Htike & ZinMaungMaungThein, 2013 was closely related to either the raoellids or the suoids depending on the inclusion of the fragmentary lower molar or not in their phylogenetic study. There is no strong evidence that the upper and lower molars described by Tsubamoto et al. (2005Tsubamoto et al. ( , 2013 belong to the same taxon because they were not found in association. Indeed, Tsubamoto et al. (2013) only considered matching size, bunodonty and the occurrence of a distobuccal wear facet on the worn m3 to associate this tooth with the upper molars and to attribute it to Myanmarius. We describe here a complete isolated m3 recently found in the locality of Pauk kaung Kyitchaung 2 ( Fig. 1) that is almost identical to the fragmentary m3 illustrated by Tsubamoto et al. (2013), and we attribute both of them to a new genus and species distinct from Myanmarius chitseini. This complete lower molar displays additional morphological information that will allow discussing its systematic relationships, and that also might help under standing the status of the material referred to Myanmarius. DiAgnosis. -Middlesized ungulate characterized by its very bunodont and short m3, lacking a paraconid and most of cristids. Differs from most of the primitive artiodactyls (Diacodexeidae, Homacodontidae, European Dichobunidae, Raoellidae) by its strongly bunodont m3 with poorly expressed crests, with a short trigonid lacking a paraconid, absence of a hypolophid, with a mesiodistal cristid obliqua and a reduced hypoconulid lobe. Differs from the Asian dichobunid Chorlakkia by its larger size, its hypoconid taller than the entoconid and hypoconulid, its entoconid in line with the hypoconid and by its hypoconulid lobe consisting in two small cusps.

Description
The tooth (MFPPK22019003) is a right m3 (L = 14.8 mm; W = 9.2 mm) without any contact facet on its distal wall. The lingual and buccal roots are fused and there is a fifth tiny root under the most posterior cusp that runs along the distal roots. The molar is bunodont with four main cusps and weakly expressed crests. The trigonid is somewhat taller than the talonid. The preprotocristid and premetacristid form a low and continuous blunt crest that mesially closes the trigonid.
There is no other crest in the trigonid part and only a slight furrow separates both bulbous mesial cusps. The distal wall of the trigonid is flat and mesially slanted. The lingual part of the talonid basin seems to be lined by a very slight and low postmetacristid that connects with the preentocristid, and its buccal part is edged by the mesially directed cristid obliqua (prehypocristid) that ends low on the distal wall of the protoconid. It is not possible to distinguish whether the lingual end of the transverse valley was open because this area is cracked, but the occurrence of a low postmetacristid and preentocristid suggests that they likely constitute a weak lin gual edge. There is no distinct crest (hypolophid) connecting the hypoconid and the smaller and lower entoconid, but a narrow mesiodistally oriented groove separates both cusps and curves buccally between the hypoconulid and the hypoconid. The short and slightly distally protruding hypoconulid lobe is composed of two small cusps separated by a very slight fur row, and a shallow slit can be observed between the mesial hypoconulid cusp and the entoconid, suggesting that both cusps might have been twinned. The hypoconulid is not con nected to the hypoconid. A narrow cingulid is present along the mesial face of the molar and a short one can be observed at the buccal end of the transverse valley between the proto conid and the hypoconid (Fig. 2AC, E).  Averianov, 1996) display less bunodont crowns, with a more obliquely oriented cristid obliqua, a distinct paraconid (Elaschitotherium), a hypolophid and a stronger hypoconulid lobe (Fig. 3E, F). Haqueina Dehm & OettingenSpielberg, 1958 from the middle Eocene of Pakistan is slightly smaller than Neochorlakkia n. gen., but it is also less bunodont with better expressed crests (postcristids on mesial cusps), more oblique cristid obliqua, with a discrete hypolophid connecting the hypoconid and the entoconid, and a narrower and more distally protruding hypoconulid (Fig. 3G). Pakibune Thewis sen, Gingerich & Russell, 1987 from the earlymiddle Eocene of Pakistan is known only from an isolated m3 that is much smaller and more elongated than the Pondaung molar, with less bunodont cusps, a paraconid, a more oblique cristid obliqua, an entoconid distinctly more distal than the hypoconid, and better developed hypoconulid and buccal cingulid (Fig. 3H). Although Wutuhyus Tong & Wang, 2006 from the early Eocene of eastern China shares bunodont cusps with the Pondaung form, it is also much smaller, with an enlarged trigonid that exhibits a prominent paraconid and a metaconid posterior to the protoconid, and an unreduced hypoconulid with two cusps. Tsubamoto et al. (2013) noted that the general morphology of the partial m3 NMMPKU 2000 (Fig. 2D) somewhat reminds that of some raoellid forms: its strong bunodonty, blunt crests, absence of paraconid, reduced hypoconulid, mesially oriented cristid obliqua and well developed talonid basin ("crushing basin") are features that can also be observed in raoellids. These are features that can also be observed in Neochorlakkia n. gen. (Figs 2A, E;  3A). However, the lack of crests that enclose the central basin that are characteristic of raoellids (Orliac & Ducrocq 2012), especially the hypolophid that borders the basin distally and the almost absent postcristids of the mesial cusps on both m3 points to different affinities (Fig. 3I). Interestingly enough, although the Dichobunidae Chorlakkia hassani from the earlymiddle Eocene of Pakistan is much smaller than the Pondaung specimen, the occlusal outline and morphology of the m3 of both taxa are strikingly similar. Indeed, the bunodont protoconid and metaconid not connected distally by crests, the lack of a paraconid, the deep talonid basin, the almost mesiodistally oriented cristid obliqua, the entoconid lingually protruding, and the reduced hypoconulid lobe are features shared by the Pakistani and Pondaung forms (Fig. 2E, F). However, MFP PK22019003 exhibits a hypoconid slightly taller than the entoconidhypoconulid complex, an entoconid in line with the hypoconid, an hypoconulid con sisting of two very small cusps and smaller than the entoconid (Fig. 3A, B). Vislobokova (2004) identified a new species of Chorlakkia (C. valerii Vislobokova, 2004) from the middle Eocene of Mongolia. This taxon displays several differences with C. hassani such as a lower molar crown more elongated, a trigonid much higher than the talonid, a hypoconulid lobe single cusped and much more developed distally, an entoconid more lingually protruding and in line with the hypoconid. It also differs from Neochorlakkia n. gen. by its much smaller size, more elongated crown with the trigonid much taller than the talonid, more oblique cristid obliqua, entoconid lower than hypoconid, presence of a posthypocristid, singlecusped hypoconulid separated from the entoconid by a depression, more elongated hypoconulid lobe, and weaker mesial cingulid. Nevertheless, the structural features that can be observed in the ungulate from Mongolia but not in Chorlakkia cast doubts on the generic attribution of C. valerii.

DISCUSSION
Despite its heavy wear, the general structure and size of the broken m3 (NMMPKU 2000) described by Tsubamoto et al. (2013) as Myanmarius chitseini seem to match those of the molar described here except for its absence of a buccal cin gulid between the protoconid and the hypoconid that might be caused by wear, and its slightly less lingually protruding entoconid that causes the distal part of the crown to taper (Fig. 2D). We can thus confidently refer NMMPKU 2000 to Neochorlakkia myaingensis n. gen., n. sp. However, there is some doubt about the association of the upper molars attributed to M. chitseini with the lower ones. Indeed, most of the early and middle Eocene Asian Dichobunidae usually possess triangular and wider than long upper molars with three main cusps (paracone, metacone, protocone), distinct but small conules, sometimes a hypocone, and a protocone more lingually protruding than the distolingual cusp. The morphology of both m3 (MFPPK22019003 and NMMP KU 2000) being similar to that of the m3 of Chorlakkia hassani, it is expected that the corresponding upper molars might resemble those of the Pakistani genus. The upper teeth attributed to Chorlakkia by Thewissen et al. (1987) exhibit this primitive morphology very different from that of the upper molars referred to Myanmarius that are more quadratic, with a welldeveloped distolingual metaconule and no hypocone, and that would more likely correspond to more advanced ungulates. In addition, the size and peculiar morphology of the M3 referred to Myanmarius (NMMPKU 1742: position of the cusps on the buccolingual wide crown, presence of dis tinct crests on the paracone, metacone and protocone) do not seem to provide a proper occlusion pattern with the structure of the narrow and very bulbous m3 MFPPK22019003. Especially, Tsubamoto et al. (2013: 303) claimed that the distobuccal wear facet next to the hypoconulid of the m3 might correspond to the "small cusplike structure distal to the metaconemetaconule on the corresponding M3". However, the occlusal interpretation of the upper and lower last molars provided in Figure 4 illustrates that a proper occlusion was likely not possible because of the size and position of cusps and crests on antagonist molars, and it thus strongly suggests that they do not belong to the same taxon. Indeed, although the trigonid cusps leave enough surface for the protocone and paraconule of NMMPKU 1742, the paracone of this molar occludes against the cristid obliqua, and the accessory cusps on the distal margin of the M3 come into occlusal contact with the hypoconulid cusps of m3. Likewise, the M2 (NMMP KU 1765) attributed to Myanmarius that exhibits a structure similar to that of the M3 NMMPKU 1742 in terms of cusp and crest arrangement might also not be associated to the lower molars. Finally, the two other upper molars (NMMP KU 2208a and NMMPKU 2208b) tentatively attributed to Myanmarius (Tsubamoto et al. 2013) display structural differences when compared with NMMPKU 1765, which casts doubts on their generic assignement: their paraconule is more in line with the paracone and the protocone (more mesially situated in NMMPKU 1765), their lingual wall is inflated at the protocone level thus giving a more triangular outline to the crown (this region is more square in NMMP KU 1765), their protocone lacks a distobuccal crest, and their mesial cingulum is somewhat weaker. As a consequence, the material included in the hypodigm of Myanmarius chitseini should be restricted to the holotype (right M2, NMMP KU 1765) and possibly the right M3 (NMMPKU 1742), whereas the two upper molars (NMMPKU 2208a and

CONCLUSIONS
The systematic affinities of Neochorlakkia n. gen. can be chal lenging because only two lower molars are attributed to that genus so far. However, their peculiar morphology including the association of their occlusal outline, strong bunodonty, poorly expressed crests, lack of a paraconid and of a hypolophid, and mesiodistal cristid obliqua sets them apart from most of the primitive artiodactyls (Diacodexeidae, Homacodontidae, European Dichobunidae, Raoellidae). On the other hand, although they are more derived by some features (larger size, hypoconid slightly taller than the entoconid and in line with it, entoconid not twinned with the hypoconulid), the Pondaung teeth are much more similar to the m3 of the Dichobunidae Chorlakkia. Although morphological convergence cannot be excluded, these similarities, together with the chronological occurrence of both taxa, might reflect phylogenetic relation ships between Chorlakkia and Neochorlakkia n. gen. If such relationships were proved correct, this might thus reinforce the hypothesis that dispersal events were possible between the Indian Subcontinent and southeastern Asia until the middle Eocene, as already demonstrated for the Raoellidae by Orliac & Ducrocq (2012). No additional dental material can be referred to Neochorlakkia n. gen. at the moment, but a detailed resassessement of the Pondaung specimens described as "indeterminate artiodactyl" as well as further discoveries from ongoing work in the Paleogene of Myanmar might lead to a better knowledge of the dental morphology of this taxon and to more precisely identify its systematic affinities. CNRS UMR 7262, the University of Poitiers, and the Ministry of Culture of the Republic of the Union of Myanmar. We are indebted to the chairmen and villagers of Bahin and Paukkaung of Pondaung area for their help, kindness, and enthusiasm that greatly facilitated our fieldwork. We thank M. Rugbumrung for preparation and casting of the material and L. FoleyDucrocq who corrected the English. Thanks to G. Métais, an anonymous reviewer and the editor L. Rook whose comments significantly improved the earlier version of the manuscript.