A proboscidean from the late Oligocene of Eritrea, a “missing link” between early Elephantiformes and Elephantimorpha, and biogeographic implications

Systematic Paleontology.

Class Mammalia Linnaeus, 1758.

Order Proboscidea Illiger, 1811.

    Nonranked category Elephantida Tassy and Shoshani, 1997.

        Family incertae sedis.

            Eritreum melakeghebrekristosi gen. et sp. nov.

Holotype.

National Museum of Eritrea (NME) field number DOG87.1, left mandibular ramus with m2 and m3, right mandibular ramus with m2 and partial first lophid of m3, portions of mandibular symphysis (in two parts) with remnants of tusk alveoli (Figs. 2 and 3).

Etymology.

Eritreum from the name of the new nation of Eritrea in the Horn of Africa, and melakeghebrekristosi in honor of the farmer, Melake Ghebrekristos, who found the specimen and realized its importance.

Locality and Age.

Situated at the base of the Eritrean eastern escarpment, the locality, known as Wedeg Melatse Farm, has low relief (≈100 m) composed of tilted and faulted Oligocene sediments of the Dogali Formation (21, 22). Fossil fragments were found partly embedded in modern outwash clays after the sudden breach of a small irrigation dam in 1997. The edges of the bones, all from the same individual, are angular, suggesting that they did not travel far. The subdued topography near the dam limits the possible source of the fossil to a narrow range of beds. Field surveys reveal that the likely source was deposits close to the dam. We suspect that breaching of the dam excavated the fossils and carried them in a mud slurry 400 m downslope and 1.6 m below the dam. Observations indicate that a basaltic ridge near the discovery site stratigraphically overlies the presumed fossil-bearing strata (Fig. 1B). Thus, the stratigraphy of the area around the dam is the same as that of the bone site (Fig. 1). A basalt sample (DG-2) was collected from above the presumed fossil-bearing layer. Three ⁴⁰Ar/³⁹Ar step-heating experiments yielded nonlinear age spectra, with slight U-shaped patterns indicating excess argon. Isochron analyses (numbers AG-59, AG-68, and AG-69 at CICESE laboratories), which are independent of excess argon, yield a composite age of 26.8 ± 1.5 million years (Mya), providing a minimum age for the stratigraphically subjacent mandible. This age is compatible with an age of 28 Myr obtained by Drury et al. (22) on flood basalts that underlie the Dogali Formation. Furthermore, this date implies that unroofing of the Eritrean basement began in the mid- to late Oligocene.

Diagnosis.

This is a small-sized elephantimorph proboscidean. The molars are smaller than those of all other early elephantimorph proboscideans and within the size range for palaeomastodonts but with pretrite anterior and posterior central accessory conules throughout the crown (Figs. 2 and 3). The m2 is fully trilophodont. The m3 has four lophids and no postcingulid. The alveoli for lower tusks have a pyriform cross-section and are higher than they are wide. The horizontal molar displacement was hypothesized.

Description.

The material consists of mandibular corpora, with molars from both sides, as well as two parts of a mandibular symphysis, with alveoli for lower tusks. The molars include right and left m2 in occlusion and erupting m3s. The sections of the dentaries for premolars are not preserved. Alveoli for m1, however, are present on both sides.

In lateral view, the left corpus displays a retracted angulus mandibulae (reduced angular process of dentary), a trait known in Elephantida. The pyriform section of alveolus of incisor measures 3.5 cm dorsoventrally and 2.3 cm at its widest point. Based on these remains, we conservatively reconstruct this mandible to be ≈50 cm long, 30 cm wide, and 20 cm high, bearing tusks ≈20 cm long and, perhaps, to have slanted gently downwards anteriorly (Fig. 2).

The right and left m2 each have three complete lophids and a thick postcingulid. Each half-lophid has two conelets; the conelet closer to the median sulcus (adaxial conelet or mesoconelet) is smaller than the conelet that is farther from the midline (abaxial conelet). The postcingulid is constructed of two bulbous conelets, with the one on the pretrite side being larger than the one on the posttrite side and connected to the third lophid. In occlusal view, the enamel wear figures exhibit the cloverleaf trefoil pattern typical of advanced gomphotheres (Fig. 3) without cement on the crown. All three interlophids are blocked by anterior and posterior accessory central conules.

Posterior to the right m2 is a partial first lophid of m3. On the left side, the m3 was in the crypt or socket and was covered with bone and matrix. This tooth germ was excavated and left in the alveolus (Fig. 2, Table 1). The third left molar resembles the second molars just described, except that it is larger and has four complete lophids, with ecto- and entoflexus for each interlophid. The fourth lophid is made of two bulbous cones; the pretrite is slightly larger than the posttrite and is entirely disconnected from the third lophid. No postcingulum is observed posterior to the fourth lophid, but only an inflated rugosity of the enamel, so that this m3 displays an intermediate condition between Phiomia (three lophids and a small postcingulid) and gomphotheres (those with four lophids and a postcingulid). The enamel inflation, rather than being a true postcingulid, at the posterior end of m3 in Eritreum is morphologically more derived than the condition in the early mammutid Eozygodon, in which there is a very thin enamel inflation on the posterior face of the fourth lophid in m3, despite the fact that the molars of Eritreum are much smaller.

The lophids do not exhibit mammutid traits, such as zygodont or “yoke”-shaped lophids (23), nor do they have zygodont crests on their posttrite sides. The anterolingual cingulid is reduced, unlike Palaeomastodon and Phiomia, and similar to the condition in mammutids and gomphotheres, an elephantimorph trait (20).

Discussion.

Overall, the teeth of E. melakeghebrekristosi occupy an intermediate morphology between early elephantiform taxa (Palaeomastodon and Phiomia) and elephantimorph taxa, including mammutids and gomphotheres (incorporating Gomphotherium, family Gomphotheriidae, superfamily Gomphotherioidea). The teeth are larger than m2 and m3 of most palaeomastodonts and smaller than all mammutids and gomphotheres, even all diminutive “pygmy” gomphotheres from Ghaba, Oman, Gebel Zelten, Libya, and Siwa, Egypt (Fig. 4). Eritreum resembles advanced Gomphotherium (e.g., G. angustidens) in having a trefoil pattern on molars (trefoil wear patterns are present in a rudimentary manner in Phiomia and Palaeomastodon) and in lacking “zygodont crests.” The lower third molar of Eritreum lacks a postcingulid, unlike m3 in early gomphotheres. The reduced anterolingual cingulid in Eritreum is a derived feature shared with mammutids, gomphotheres, and elephantids.

Based on data from G. angustidens (24), the individual age, and the evolutionary stage of the Dogali specimen, the first lower molars would have been in situ when the animal died. These teeth (m1s) on both sides may have fallen out during burial. In the Dogali specimen, when the first and the second lower molars would have been simultaneously in occlusion, the third lower molar (m3) would be far at the posterior end of the lower jaw and inside the growing bone, and we conjecture that there would have physically been no space for this tooth to erupt and be in the same horizontal level of occlusion as m1 and m2. Fig. 2A′ depicts the m3 still in the crypt (socket), and the dorsal tips of its anterior unworn cusps are ≈1 cm below the occlusal surface of the worn cusps of m2. In addition, the roots in m3 have not yet developed, whereas they are well developed in the m2. We hypothesize that at a later stage of development, when the roots of m3 would have grown, the tooth would move upwards and forwards to be at the same horizontal level of occlusion of m2 (m1 would probably have fallen out by that time). From these anatomical features and the conjectured timing of dental eruption, we conclude that the mandible of E. melakeghebrekristosi may have exhibited horizontal tooth displacement (similar to slow movement on a conveyor belt), a trait of Elephantimorpha, including the living elephants (19, 20). Conversely, in primitive Elephantiformes such as Palaeomastodon and Phiomia, with complete dentition (p2-m3), when the state of wear is comparable to that seen on Dogali m2s, the m3 would have been entirely erupted and in occlusion. In mammutids, the sequence of eruption of m1, m2, and m3 is intermediate between Elephantiformes (Palaeomastodon and Phiomia) and Elephantida. Additional specimens of Eritreum or closely related taxa may help in evaluating whether the alternative hypothesis that Eritreum is a member of Elephantida is better supported than that Eritreum is a member of Elephantimorpha (Fig. 5).

In comparing data used to identify dental ages for extant African elephants and for fossil gomphotheres (24–26), we estimate that the Dogali individual was a subadult, close to the age of 26 at death. Using data from Andrews (2) and Roth (27), we extrapolate the shoulder height to have been ≈130 cm and the body mass ≈484 kg (Fig. 6). Extrapolation of the shoulder height of Eritreum was based on data for Palaeomastodon parvus from Andrews (2) on these skeletal elements: length of m2 (4.5 cm), length of mandibular fragment (36 cm), length of humerus (45.2 cm), and length of incomplete ulna (38 cm). The reconstructed jaw of E. melakeghebrekristosi (Fig. 2) was reduced to the same scale as P. parvus by using the size of m2 as the measurement criterion for both animals. The humerus and the ulna were then fit to the scaled P. parvus with the calculated shoulder height of 127 cm ± 5 cm. Using these extrapolations, we estimate the shoulder height of E. melakeghebrekristosi to have been between 122 and 132 cm. By using Roth's (27) 12 equations to calculate the body mass of Eritreum from the extrapolated shoulder height, a range from 309.47 to 434.39 kg (average of 371.51 kg) was obtained. Benedict's (28) data on body mass of living elephants indicate that elephants with the same shoulder height have different weights, with variation of 20–40%. Thus, if we add 74 kg (20% of 372 kg) or 149 kg (40% of 372 kg) to 372 kg, we obtain a range of 446–521 kg (average 484 kg) for the estimated body mass of E. melakeghebrekristosi. It is noted, however, that Eritreum was not a fully grown individual (≈26 years old), and it probably would have achieved a larger size.

Results of character analysis [taxon n = 40; character n = 125; run in PAUP (29) and MacClade (30)] are depicted in Fig. 5. E. melakeghebrekristosi can be equally placed as the earliest offshoot of Elephantimorpha or of Elephantida (additional data and descriptions are available in Appendices A–D, which are published as supporting information on the PNAS web site). Both clades of Elephantida and Elephantimorpha have bootstrap values >70%, which is indicative of reasonably good branch support (31). There are two good characters in support of Eritreum to join Elephantimorpha and only one good character in support of Eritreum to join Elephantida. In addition, the bootstrap value for Elephantimorpha with Eritreum is higher (91%) compared with the bootstrap value for Elephantida (78%). For these reasons, we favored the alternative that Eritreum is a member of the clade Elephantimorpha rather than of Elephantida.

The synapomorphies that Eritreum shares with Elephantimorpha include hypothesized horizontal tooth displacement (character 67 in Appendices B and C; this is a good character with consistency index, CI = 1.0, see Appendix D) and reduced cingulum on the anterolingual side of lower molars (character 46; CI = 1.0). The hypothesized horizontal tooth displacement of Eritreum would put this genus within the Elephantimorpha clade. A good synapomorphy that Eritreum shares with Elephantida is the reduced angular process of dentary (character 96, CI = 1.0). Two other characters that might be used as possible synapomorphies for Elephantida, Eritreum included, are the complete and well delineated cloverleaf trefoil pattern on the pretrite side of the lower molars and lack of zygodont posttrite crest (characters 55 and 66, respectively); these characters are weak for this clade (details in Appendix A). Additional synapomorphies for these and other nodes in Fig. 5 are found in Shoshani (19) and Tassy (20) and in Appendix D.

The cladistic position of Eritreum in Fig. 5 is intermediate between Phiomia and Palaeomastodon on one hand, and between Mammutida and Elephantida on the other. The morphological (especially the horizontal tooth displacement) and temporal (late Oligocene) characters of Eritreum are well suited to fill this missing link gap, a hypothesis that was predicted ≈84 years ago (32).

Taxa depicted in Fig. 5 and Table 2 within the order Proboscidea and above the superfamily category level (Elephantiformes, Elephantimorpha, Mammutida, and Elephantida) are not assigned formal ranks or categories (following the International Code of Zoological Nomenclature (33), ICZN, see below), because to give them ranks would imply that we are certain of their phylogenetic position (11). For the same reason we did not assign a family rank to E. melakeghebrekristosi. Following the ICZN (Recommendation 29a and Article 36a), taxonomists have to abide by the rules for the Family Group, but it is not required to follow any rules for ranks above the Family Group; a Family Group includes categories of Superfamily, Family, Subfamily, Genus, and Tribe.

The hypothesis that Phiomia is the sister taxon to Elephantida and Palaeomastodon is the sister taxon to Mammutida (18) increases the tree length by 10 steps. Employing the parsimony principle, this hypothesis is rejected in favor of the more recent hypothesis (19, 20) that Phiomia and Palaeomastodon are a sequence of sister taxa to Elephantimorpha as depicted in Fig. 5 and Table 2.

When viewed in a paleobiogeographical context, E. melakeghebrekristosi provides additional evidence that East Africa preserves major, previously unknown aspects of proboscidean differentiation during the transition from Paleogene to Neogene (late Oligocene to early Miocene), preceding the arrival of proboscidean immigrants from Eurasia (7). The proximity of Dogali, Eritrea, and Chilga, Ethiopia (only 200–250 km apart) (7,8) in this region lends credence to this hypothesis. Data presented here provide additional evidence for an African contribution to major events in proboscidean evolution and that East Africa in particular was an important Afro-Arabian setting of proboscidean evolution during the basal radiation of Elephantiformes (7, 8, 10, 17).

Supplementary Material