We report here significant new material belonging to the oldest and most primitive known Proboscidean, Phosphatherium escuilliei Gheerbrant, Sudre & Cappetta, 1996, from the early Eocene of the Ouled Abdoun phosphatic basin, Morocco. This material permits the first reconstruction of the skull and most of upper and lower dentition of Phosphatherium escuilliei. The species, which is one of the oldest and most primitive known representatives of modern orders of ungulates, becomes one of the best known among them. Its dentition shows a noticeable dental variability, which is interpreted, at least provisionally, to be intraspecific. In addition to the individual variation, some features of Phosphatherium escuilliei such as the maxillary (infraorbital area) show a noticeable variability which is related to sexual dimorphism, as was also reported in Numidotherium. The skull of Phosphatherium escuilliei is very primitive in many respects. It is long with an elongated facial part and a narrow rostrum. The toothrow does not extend posteriorly beyond the middle of the skull. The nasals are long and located anteriorly (i.e. nasal fossa not retracted). There is no contact between the premaxilla and frontal. There is a strong postorbital constriction and a distinct postorbital process on the frontal. The zygomatic arches are noticeably expanded laterally. The sagittal and nuchal crests are strong. The alisphenoid canal is present. The external auditory meatus is open ventrally. The periotic has a primitive therian construction with the presence of a round window and a cochlear canaliculus. The hypoglossal foramen is present. The ethmoid foramen is in anterior position. The occipital is flat and lacks a median crest. The braincase is strongly compressed laterally. Some primitive features of the dentition are also noticeable: C1/c1 and P1 are retained, diastemata are small (upper) or absent (lower), lower central incisor is only moderately enlarged, lower premolars are simplified, and a dilambdodont ectoloph linked to the mesostyle is present. In the lower jaw, the symphysis is unfused, the horizontal ramus is narrow and the condyle is low. However, Phosphatherium escuilliei displays several strikingly advanced features, especially proboscidean and tethytherian features. P. escuilliei is actually one of those exceptionally known fossils that illustrates the initial radiation of modern placentals, at an early stage close to the generalized eutherian “condylarth” pattern that typifies the early Paleogene. P. escuilliei is one of the few (well) known fossil taxa that documents the emergence of a modern order of mammals at this stage. A cladistic study of 129 features of Phosphatherium escuilliei within an extended systematic framework, including most lophodont ungulate lineages and their presumed “condylarth” relatives (e.g., Phenacodontidae, Ocepeia), provides new insight into the intra-ordinal and supra-ordinal phylogenetic relationships of the Proboscidea. It supports the monophyly of Proboscidea and the inclusion of Phosphatherium within the order. The most significant Proboscidean synapomorphies found in Phosphatherium are: 1) the well developed zygomatic process of the maxillary which contributes significantly to the ventral border of the orbit and to the zygomatic arch; 2) the relatively large size of the pars mastoidea of the periotic; and 3) the hypoconulid in a labial position (a state unique to Proboscidea). Loss of i3 and (d)p1 are also potentially important synapomorphies of Proboscidea, but they are uncertain in Phosphatherium (homology of teeth) and Daouitherium (anterior teeth not known). Other Proboscidean synapomorphies are notable, though homoplastic: 1) frontal in contact with the squamosal (parallelism with Perissodactyla); 2) optic foramen in posterior position (parallelism with Embrithopoda); 3) true lophodont molars (reversal or convergence within advanced Proboscideans); and 4) enlarged i1 (reversal in Moeritherium, Deinotheres and Elephantiformes). The hypothesis of a true lophodont ancestral morphotype of the Proboscidea is especially supported by Phosphatherium and Daouitherium. No autapomorphies of Phosphatherium are identified in the cladograms. However, some of its features are notable among Proboscidea, even if they are interpreted as homoplastic or as more inclusive basal features: the cristid obliqua in labial position on the molars (convergence with Moeritherium and the Elephantiformes), and the p2 uniradicular (convergence with Numidotherium savagei). Within the order Proboscidea, the relationships of lophodont taxa – i.e. the basal radiation of the order – are still poorly resolved. The cladogram resulting from successive weighting shows a sequential arrangement of the lophodont taxa in accordance with their geological age. This argue against the concept of a monophyletic lophodont group of Proboscideans such as Plesielephantiformes or extended Barytherioidea. However, this sequential arrangement of lophodont taxa is weakly supported, mostly by a few homoplasies, and the unweighted analysis gives an alternative clade (Numidotherium (Barytherium, Daouitherium)) in three of the six resulting trees. However, Daouitherium is clearly more primitive than Numidotherium and Barytherium, and Numidotherium and Barytherium are sister groups (= clade Barytherioidea) according to the successive weighting analysis. Higher in the Proboscidean tree, successive weighting supports a clade of advanced Proboscideans (Moeritherium (Deinotheriidae, Elephantiformes)), and a sister-group relationship between Deinotheriidae and Elephantiformes is well supported by both unweighted and successively weighted analyses. Phosphatherium and Daouitherium, that are of the same earliest Eocene age, belong to two lineages which are distinct at the family level. Therefore, Phosphatherium is included here in its own family Phosphatheriidae n. fam. However, the original derived features of Phosphatherium are weak: in this regard it is the best known structural ancestor of the Proboscidean order. Khamsaconus might be related to the family Phosphatheriidae n. fam. according to our analysis, but the genus remains too poorly known for any definite statement. With respect to the supra-ordinal relationships of Proboscidea (i.e. relationships within Taxeopoda and Altungulata), our analysis results in a single tree that is nearly entirely resolved, except for the relative position of Radinskya and Perissodactyla. Sirenians are the sister group of Proboscideans, and they belong to a well supported node (e.g., molars with transverse lophs, occurrence of a submaxillary fossa). Our analysis supports the tethytherian relationships of north-Tethyan taxa such as Desmostylia, Anthracobunidae, and Minchenella. However, these taxa are more distant successive lateralbranches of stem tethytheres with respect to Proboscidea and Sirenia. As a result, the Anthracobunidae are removed from the order Proboscidea that appears to be an endemic strictly African clade. The mutual arrangement of these early tethytheres remains, however, weakly supported because they are poorly known (few features, mostly homoplastic). As a whole, tethytherian features are homoplastic and/or of ambiguous distribution (e.g., anterior orbits, robust zygomatic process of the squamosal, infraorbital foramen located close to the orbit, and presence of the postentoconule and the postentoconulid). Our analysis supports a relationship of Phenacolophus with the embrithopods. However it suggests, as an original and well supported node, that Embrithopoda (+Phenacolophus) is the sister-group of the whole Tethytheria (Minchenella (Anthracobunidae (Desmostylia (Sirenia, Proboscidea)))) and not of the Proboscidea. The analysis also supports the controversial clade Paenungulata (i.e. Hyracoidea as sister-group to Tethytheria and Embrithopoda). However, this is still based on relatively few features, some of them needing to be checked in Phosphatherium (amastoidy, large upper incisors). The clades Altungulata (= Pantomesaxonia) and Taxeopoda are accepted, but their monophyly is not really tested here because our analysis does not include key taxa included in the competing hypotheses of Afrotheria and Africana (e.g., Macroscelidea). Moreover, the comparison of the features of Phosphatherium, Khamsaconus, Seggeurius, and primitive embrithopods raises the question of a possible convergence of the lophodont pattern in Perissodactyla and Paenungulata, and indeed the question of the monophyly of Altungulata. Significant elements consistent with such a possible convergence are the bunodont, weakly lophodont (lopho-selenodont) molars of early hyraxes, the convergent lophodont-like (hyperdilambdodont) pattern of embrithopod upper molars, the absence of conules and the simplified premolars of early hyraxes and proboscideans, and the questionable homology of some of the lophs of perissodactyls (see features 79 and 111). Finally, it should be noted that our analysis excludes Ocepeia from the Altungulata and Taxeopoda (hypolophid stated convergent). Several of the ambiguities that remain about the intra- and supra-ordinal relationships of the Proboscidea result from gaps in our knowledge of fossil taxa. New data on Daouitherium would be especially helpful to clarify relationships of early lophodont proboscideans. A better knowledge of Anthracobunidae (e.g., skull anatomy) would also contribute to improving resolution of the relationships of primitive Tethytheres. The question of the origin of Paenungulata and their relationships (e.g., Altungulata?) needs further data on early Hyracoidea and Embrithopoda (Palaeoamasiidae). As for Phosphatherium, its postcranial skeleton remains unknown. It is especially important for the question of locomotion, which is presumed to be pivotal in Tethytheria which are hypothesized to have had a semi-aquatic ancestral morphotype. Finally, it should be stressed that Phosphatherium is the most primitive known Proboscidean, especially in skull morphology. Such a plesiomorphic skull anatomy contrasts strongly with the advanced truly lophodont molars that are indistinguishable from those of Eocene and Oligocene taxa such as Numidotherium and Barytherium. This suggests a precocious trophic adaptation (follivorous diet) within a favorable African paleoecosystem at least as early as the Paleocene-Eocene transition. The ecological niche involved seems to have been stable and important (from an evolutionary perspective) for a long time in the endemic Arabo-African province.
Mammalia, Ungulata, Proboscidea, Africa, Morocco, Ouled Abdoun Basin, early Eocene, phylogeny