A lower Pliocene Erymnochelyini turtle (Pleurodira, Podocnemididae) from the Democratic Republic of Congo

The presence of a pleurodiran turtle putatively attributable to Erymnochelys madagascariensis (i.e., the only representative of Erymnochelyini that is part of the current biodiversity) or to a closely related form was recognized, in the early 1990s, for the Lower Pliocene record of the Sinda‐Mohari region (Democratic Republic of Congo). The material attributable to it was restricted to postcranial remains. Although some elements of the anterior region of the skull were subsequently found, this taxon had not been analyzed in detail until now. Knowledge about Erymnochelyini has increased remarkably since the 1990s and, especially, during the 21st century. The taxon from the Democratic Republic of Congo is here analyzed in detail, not only considering current knowledge about the lineage but also incorporating new material into this study, highlighting several partial skulls. Thus, it corresponds to one of the members of this lineage represented by a greater number of skulls. It is attributed to a new taxon, Mokelemys mbembe gen. et sp. nov., being the only representative of Erymnochelyini currently known for the Pliocene record.


| INTRODUCTION
The pleurodiran turtle E. madagascariensis (Grandidier, 1867) (i.e., the "Madagascan big-headed turtle") is the only representative of Erymnochelyini (Pelomedusoides, Podocnemididae) that is part of current biodiversity (see Pérez-García, 2021a; and references therein). This critically endangered species is exclusive to western Madagascar (Rhodin et al., 2021) (Figure 1). Erymnochelys was traditionally considered for the generic attribution of most extinct representatives of the Erymnochelyini lineage. However, they were subsequently reattributed to new genera.
In the early 1990s, the remains of several groups of turtles were reported from various outcrops of the Sinda beds, at the Ituri Province, in the Democratic Republic of Congo. These fossils were exclusively postcranial elements, including several that were attributed to F I G U R E 1 Geographical (a) and stratigraphic (b) position of the type localities of all extinct representatives of Erymnochelyini (Pleurodira, Podocnemididae) previously described (1-6) and of Mokelemys mbembe gen. et sp. nov. (7), and region where the only extant representative of this lineage lives (8) Erymnochelys sp. (Yasui, Makinouchi, et al., 1992). Those sites are currently identified as being located in Early Pliocene levels (Werdelin, 2010). The remains attributed to the genus Erymnochelys (see figure 7 in Yasui, Makinouchi, et al., 1992; figures 1-2 and plates 1-3 in Hirayama, 1992), from the sites identified as 3, 7, 11 and 15 (see figure 2 in  and figure 3 in Makinouchi et al., 1992), were analyzed by Hirayama (1992). He supported that generical attribution, which was justified by the presence of a reduced gular scute, so that a medial contact of the extragulars was present. This character is not currently considered as exclusive to this genus, being identified as a synapomorphy for all members of Erymnochelyini (Pérez-García et al., 2021). In her compilation and update on the African turtle record, de Lapparent de Broin (2000a) alluded to that material as aff. Erymnochelys sp., without providing new data on the finds from the Sinda beds. Wood (2003) reported the presence of several partial skulls from that locality, F I G U R E 2 MRAC rt15_eg_01, holotype of Mokelemys mbembe gen. et sp. nov., from the early Zanclean (Early Pliocene) of the Sinda-Mohari region (southeastern Ituri Province, Western Rift Valley, Democratic Republic of Congo), in dorsal (a), anterior (b), ventral (c), left lateral (d), and right lateral (e) views. The dotted lines represent broken margins; the texture regions correspond to broken bones referred to as nearly indistinguishable, insofar as they had been preserved, from those of E. madagascariensis. Thus, Wood (2003) pointed out that, considering the information available in these cranial remains, and in the shell elements from that locality, the Erymnochelyini present in the Sinda beds could be a direct ancestor of the extant species, being radically different from the only extinct Neogene Erymnochelyini known by the skull (i.e., T. pattersoni). Gaffney et al. (2011) alluded to those undescribed cranial remains from the Sinda beds (pointing out that all of them corresponded exclusively to the anterior area of the skull, material corresponding to the cheek or the basicranium not being recognized), indicating that they were compatible with E. madagascariensis. No additional information on the material of this group of pleurodires from the Democratic Republic of Congo has so far been published. Since most of the genera currently recognized for the Erymnochelyini fossil record were defined after the publication of the Hirayama (1992) and de Lapparent de Broin (2000a) papers, the precise systematic attribution of these remains is currently uncertain, their attribution to the genus Erymnochelys having recently been preliminarily indicated, without being compatible with the recent taxonomic practice (see Pérez-García, 2021b, and references therein).
F I G U R E 3 MRAC 13686a, partial skull of Mokelemys mbembe gen. et sp. nov., from the early Zanclean (Early Pliocene) of the Sinda-Mohari region (southeastern Ituri Province, Western Rift Valley, Democratic Republic of Congo), in dorsal (a), anterior (b), ventral (c), left lateral (d), and right lateral (e) views. The dotted lines represent broken margins; the texture regions correspond to broken bones Unpublished shell remains show that the fossils of Erymnochelyini are very abundant in the Sinda beds. In addition, it includes a set of 10 unpublished cranial remains, deposited in the Royal Museum for Central Africa (Tervuren, Belgium), which not only correspond to the 4 fragments of the anterior area of the skull reported by Wood (2003) and Gaffney et al. (2011), but also to other cranial regions, allowing to know the almost complete anatomy of the skull of the taxon represented there, as well as that of its partial lower jaw (Figures 2-6). Among the three species of the lineage of Erymnochelyini currently known by the skull (i.e., the European middle Eocene E. eremberti, the Egyptian early Oligocene S. fajumensis and the Kenyan Upper Miocene T. pattersoni) only S. fajumensis was known from several skulls. Therefore, no set of several skulls of a single Erymnochelyini species has been documented for the Neogene record so far. In fact, only a single skull was known for that Period (i.e., that of the partial skeleton corresponding to the holotype of T. pattersoni). The cranial remains of Erymnochelyini from the Democratic Republic of Congo are presented here, with the aim of analyzing if the taxon represented there corresponds to a new form or if it is attributable to a genus and/or species previously defined in another country, expanding its hitherto known paleobiogeographic and/or biostratigraphic distribution.

| Diagnosis
Member of Erymnochelyini characterized by the following exclusive combination of cranial characters: subtriangular interparietal scute, reaching at least the anterior level of the temporal emargination; absence of interorbital longitudinal depression; relatively well developed temporal emarginations, only generating the partial roofing of the fossae temporalis; dorsally directed orbits; absence of medial premaxillary hook; wider than tall nares; height of the suborbital space similar to that of the orbits; high cheek emarginations, reaching at least half the height of the orbits; greatly reduced to absent ventral midline depression on the premaxillae; single accessory ridge on the palatal surface of the maxilla; very broad biting surface at the mandibular symphysis; lateral surface of the dentaries partially exposed in dorsal view; welldeveloped medial accessory ridge on each triturating surface of the lower jaw. In addition, it shows the following exclusive combination of shell characters within Erymnochelyini: absence of a medial carapacial keel; six neurals; relatively long distal margins on the seventh costals, more than two times longer than those of the eighth ones; heptagonal first vertebral, with short latero-anterior margins, and wider than the nuchal; second vertebral approximately as wide as the third, these scutes lacking lateral rounded protrusions; first pair of marginals overlapping less than half the lateral nuchal margins; absence of extragular protrusions; relatively long gular, reaching the anterior area of the entoplastron.

| Description
The anterior view of the skull of the taxon studied here shows that the nares are wider than tall (Figures 2b,3b,and 4b). This species lacks a medial protrusion at the anterior end of the premaxillae. Therefore, a premaxillary hook is not present. A sulcus is not developed on the dorsal surface of the premaxillae.
Some degree of variability is identified when the width and the length of the anterior area of some of the skulls are observed in dorsal view (Figures 2a, 3a, 4a, and 5a). In particular, the interorbital region is noticeably shorter and wider in some specimens (see Figure 3a) than in others (see Figure 2a). The anterior snout is short, representing less than half the length of the orbits. The width of the orbitonarial bar is slightly less than the diameter of the orbits. The taxon lacks nasals. The prefrontals show a long medial contact, hindering contact between the aperture narium externa and the frontals. No interorbital groove is developed at the medial contact between the prefrontals. The prefrontals are slightly shorter than the frontals. The contact between these pairs of bones is substraight in some specimens (Figure 4a), but shows a well-developed medial protrusion in others (Figure 5a), intermediate stages also being recognized (Figures 2a and 3a). The parietals do not contact the orbital margins, being relatively far from that structure. The maxillae define a well-developed margin at the ventral area of the orbit rims. The postorbitals, in contact with the orbital margins, are longer than wide (Figures 2a, 3a, 4a, 5a, and 6a). The contact between the parietals and the quadratojugals is relatively long (Figure 6a). The parietals do not contact the squamosals because well-developed upper temporal emarginations are present. It shows a concave margin. The otic chambers are partially covered because of the partial roofing of the fossae temporalis. A well-developed interparietal scute is recognized. It is subtriangular, and longer than wide. Its anterior margin overlaps the posterior region of the frontals. Posteriorly, it reaches, and most probably exceeds, the position of the anterior margin of the temporal emarginations. A dorsal contact between the supraoccipitals and the quadrates is absent (Figure 5e,h).
In lateral view, a poorly developed pinched snout is recognized, showing some degree of variability (see Figures 2d,e,3d,e,4d,e,and 5b). The height of the suborbital spaces is similar to those of the orbits. The foramen interorbitale is relatively high. The jugals reach the orbital margins. These bones do not contact the parietals. The taxon studied here lacks contact between the maxillae and the quadratojugals. In this sense, it shows a high cheek emargination (i.e., the lower temporal emargination), reaching at least half the height of the orbits (Figures 3d,e and 4d), and perhaps surpassing the most dorsal margin of the orbits (Figure 6c). The jugals shows a large exposure on the cheek emargination. The incisura columellae auris encloses both the stapes and the eustachian tube.
This taxon lacks contact between both the exoccipitals and quadrates and the basioccipitals and opisthotics (Figure 5g,j).
The ventral view of the skull shows that the labial ridge is relatively high and narrow (Figures 2c,3c,4c,and 5c,k). The premaxillae contact the apertura narium interna. The taxon lacks a vomer. The triturating surfaces are subtriangular, being slightly wider posteriorly than anteriorly. They are relatively wide, but they are well separated from each other, so that they do not constitute a secondary palate. A single maxillary ridge is recognized in the medial area of the palatal surface of each maxilla. However, these bones lack pits. The jugals are not exposed on the triturating surfaces. By contrast, the palatines show a relatively large contribution to these surfaces. A relatively long medial contact is present between both pterygoids (Figures 2c, 3c, 5d,f,i, and 6b). Laterally, these bones show trochlear processes. The taxon shows large anterior openings of the cavum pterygoidei, partly covered by the pterygoid wings, anteriorly delimited by completely developed pterygoid flanges and posteriorly by well-defined ridges of the quadrate medial processes. Trigeminal ridges are absent in the ventral surface of the skull. A long contact is recognized between the quadrates and the basisphenoid. The basisphenoid is pentagonal, and wider than long. A relatively long contact is also present between the quadrates and the basioccipital due to the absence of ventral processes of the exoccipitals. The condylus mandibularis of the quadrates are interpreted in a position distinctly anterior to that corresponding to the suture between the basisphenoid and the basioccipital (see Figure 6b). The basioccipital tubera are laterally located.
The partial lower jaw identified for the taxon described here shows a relatively broad biting surface at the mandibular symphysis ( Figure 6f). However, its triturating surface is relatively narrow, and it is recognized as narrower posteriorly than anteriorly. In fact, the posterolateral surface of the dentary is exposed in dorsal view and not only in lateral view. A well-developed accessory ridge is present on the triturating surface, being parallel to the lingual and labial ridges, and located closest to the first one. The lingual ridges of the dentaries are recognized as lower than the labial ones (Figure 6d,e).

| Taxonomy
The cranial remains from the Lower Pliocene record of the Sinda-Mohari region (Democratic Republic of Congo) analyzed here correspond to Pleurodira considering the presence of trochlear processes on the lateral region of the pterygoids. These processus trochlearis pterygoidei are well-developed, as occurs in Pelomedusoides, a lineage to which they can be attributed considering, among other characters, the relative strong posterior emargination and the absence of nasals and splenials (de Lapparent de Broin, 2000b). They are recognized as belonging to Podocnemididae due to the presence of a fully developed and medially extensive cavum pterygoidei, with a completely developed pterygoid flange, and partly covered by the pterygoid wings; incisura columellae auris enclosing the stapes and the eustachian tube; and absence of exoccipitalquadrate contact (Gaffney et al., 2011;Pérez-García et al., 2017). Characters as the absence of a deep fossa precolumellaris and the development of a large anterior opening of the cavum pterygoidei, with a wide dorsal opening in the sulcus cavernosus, allow their attribution to Erymnochelyinae (de Lapparent de Broin, 2000b;Pérez-García & de Lapparent de Broin, 2015).
All Erymnochelyinae cranial remains identified in the Lower Pliocene levels of the Sinda beds are compatible with a single form. Some of these cranial remains were preliminarily considered as compatible with the Malagasy extant species E. madagascariensis (Gaffney et al., 2011;Wood, 2003). The skull of a single member of the Erymnochelyinae was so far known from the Upper Miocene fossil record: that of the Kenyan Erymnochelyini T. pattersoni. It was considered as not compatible with the taxon from the Sinda beds by Wood (2003), which was also identified as a member of Erymnochelyini, sharing with the representatives of this lineage a synapomorphy corresponding to the presence of a reduced gular scute, allowing the medial contact of the extragulars. Taking all this into account, the cranial characters recognized as different between E. madagascariensis and T. pattersoni are analyzed in the material studied here. Wood (2003) indicated several differences between the skull of the members of Erymnochelyini T. pattersoni and E. madagascariensis, the taxon from the Democratic Republic of Congo shares with the former but not with the extant species the following characters: high cheek emarginations; very broad biting surface at the mandibular symphysis, the masticating troughs becoming progressively narrower toward the rear region, so that part of the lateral face of the dentaries are exposed in dorsal view; and well-developed medial accessory ridge on each triturating surface of the lower jaw. Therefore, the taxon to which the cranial remains of the Sinda beds differs from E. madagascariensis. The presence of high cheek emarginations is shared with the Egyptian early Oligocene S. fajumensis but not with the European middle Eocene E. eremberti. However, the characters corresponding to the dorsal surface of the lower jaw of neither of these two taxa are known (see Gaffney et al., 2011;Pérez-García, 2021a;Pérez-García et al., 2017).
Considering the characters discussed above, the taxon from the Sinda beds could be recognized as a form closer to T. pattersoni than to E. madagascariensis. However, Wood (2003) included in the diagnosis of T. pattersoni several characteristics not shared with the material analyzed here. Thus, that species was defined by a trapeziform interparietal scute, being short and, therefore, allowing a long midline contact between the parietal scutes; and by the presence of higher than broad nares. Wood (2003) characterized the states for these characters in E. madagascariensis as opposed to those in T. pattersoni: subtriangular and long interparietal scute, generating a short midline contact between the parietal scutes; and wider than tall nares. As in E. madagascariensis, the taxon from the Sinda-Mohari region shows wider than tall nares (Figures 2b, 3b, 4b); and a subtriangular and long interparietal scute. Although the only known skull for E. eremberti is dorsoventrally flattened, and that it lacks the posterior area of the cranial roof, the preserved region of the interparietal scute shows that this subtriangular element was notably longer than that in T. pattersoni, and that its nares were probably wider than high (Pérez-García et al., 2017). A similar condition is observed for the nares of S. fajumensis, the length of its interparietal scute being variable, not reaching the anterior margin of the temporal emargination in some individuals (see figure 46 in Gaffney et al., 2011), but slightly surpassing it in others (see figures 44-45 in Gaffney et al., 2011). Wood (2003) characterized the skull T. pattersoni as relatively narrow and elongate, that of E. madagascariensis being moderately broad and squat. This character is difficult to evaluate in the material analyzed here considering both that no complete skull is preserved and that, as indicated (see Section 2.6), there is a certain range of variability in the cranial morphology and in the ratio between the width and the length in these skulls. Despite their deformation, the two so far figured skulls of the only extinct taxon of Erymnochelyini that was hitherto known from various cranial remains (i.e., S. fajumensis) appear to show certain differences in the general morphology when they are compared (see figures 44-46 in Gaffney et al., 2011). However, the anterior region of none of the skulls of the taxon analyzed here is as narrow as that of the skull of T. pattersoni. Also considering the skull morphology, E. eremberti was defined as sharing exclusively with T. pattersoni within Erymnochelyini the presence of a short anterior snout, less than half the length of the orbits (in contrast to those of E. madagascariensis and S. fajumensis) (Pérez-García et al., 2017). This condition is not shared with the taxon from the Sinda-Mohari region either (see Figures 2 and 3).
The study of the only known skull of T. pattersoni (i.e., that of its holotype) allowed Wood (2003) to diagnose this species as having a prominent triturating ridge on the palatal surface of the maxilla, absent in E. madagascariensis. In fact, two accessory ridges are present in the only known skull of this taxon, this character having been incorporated by Gaffney et al. (2011) in the amended diagnosis of the species, the most prominent being the most distal ridge (see figures 4.5 and 4.8 in Wood, 2003). The personal observation of the skulls of several individuals of this extant species allows me to suggest that the development of accessory ridges on the triturating surface is relatively variable, but that, when it is present, it is a single structure (located in a position equivalent to the most distal ridge of T. pattersoni), being poorly developed (e.g., it is almost absent in MNHN.ZA. AC 1946-71, showing slightly more development in MNHN.ZA.AC 1932-578). In this sense, the presence of an accessory ridge on the triturating surface was previously coded as variable (i.e., absent or present) for E. madagascariensis in some data matrices (see Cadena, 2015); although this structure was erroneously coded as absent for T. pattersoni. The taxon from Sinda-Mogari shows some degree of variability for this character. It shows only a single well-developed accessory ridge (a barely developed ridge, situated in a region equivalent to that of the notably more developed medial ridge of T. pattersoni, can only be recognized for MRAC rt15_eg_01; Figure 2c). That maxillary ridge is never as prominent as that known for T. pattersoni, but it varies from being as poorly developed as in some specimens of E. madagascariensis (Figures 4c and 5c), to showing an intermediate development between the distal ridges of both species (Figures 2c and 3c). Since the only known skull for E. eremberti retains its lower jaw in articulation with it, this character is unknown for the species (see Pérez-García et al., 2017). A poorly developed to absent single accessory ridge is identified for S. fajumensis, this condition being similar to that observed for E. madagascariensis (see Gaffney et al., 2011).
After the publication of the paper where T. pattersoni was established (see Wood, 2003), some authors reported certain additional differences between the skull of that species and that of the extant E. madagascariensis, or between those of these species and those of other representatives of Erymnochelyini. Some of them correspond to characters located on elements preserved in the cranial material from the Democratic Republic of Congo studied here. The orbits of T. pattersoni were characterized by Gaffney et al. (2011) as facing laterally, which is shared with E. madagascariensis. However, they are more dorsally directed in the specimens analyzed here, which is shared with the so far published skulls of S. fajumensis and probably also with that of E. eremberti (see Pérez-García et al., 2017).
The height of the suborbital space was recognized as different when the skulls of the representatives of Erymnochelyini are compared (Pérez-García et al., 2017). Thus, E. eremberti was characterized by having a high suborbital space, slightly higher than that in E. madagascariensis, but significantly higher than those in S. fajumensis and T. pattersoni, in which the orbits are higher than the suborbital space. The condition observed in the taxon analyzed here, with a height of the suborbital space similar to that of the orbit, is similar to that recognized for E. madagascariensis.
The presence of a pinched snout (generating a concave outline near the premaxilla-maxilla contact instead of a convex or straight rostral outline) has recently been considered in the encoding of T. pattersoni, being defined as absent for E. madagascariensis and S. fajumensis (see Joyce & Bandyopadhyay, 2020 and references therein). A pinched snout is present in the taxon from the Democratic Republic of Congo, being poorly developed, but showing some degree of variability (see Figures 2d and  5b). A more pronounced pinched snout than that recognized for the skulls analyzed here can be identified in some specimens of E. madagascariensis (e.g., MNHN.ZA. AC 1946-71 andMNHN.ZA.AC 1932-578), the development of this structure being variable in the species (see figure 6 in de Lapparent de Broin & Werner, 1998), but always being present. In fact, it is also present in S. fajumensis and E. eremberti, so its presence or absence cannot be a character to differentiate the representatives of Erymnochelyini from each other. Given the variability identified in some species (i.e., E. madagascariensis and the taxon studied here), and that this region has only been characterized based on a single skull in others (i.e., T. pattersoni, S. fajumensis and E. eremberti), their degree of development cannot currently be identified as valid to characterize each species.
The dorsal temporal emargination has been recognized as interspecifically variable when members of Erymnochelyini are compared (Cadena, 2015). Thus, it has been characterized as reduced in the case of E. madagascariensis, implying the almost complete roofing of the fossa temporalis (the posterior margins of the parietals and squamosals being slightly convex), but not in T. pattersoni and S. fajumensis, in which it is more developed, only generating a partial roofing of the fossa temporalis. The development of the temporal emargination is greater in S. fajumensis than in T. pattersoni, a larger portion of the otic chamber being dorsally exposed, and showing concave instead of substraight posterior margins of the parietals and squamosals. Although this region is not completely preserved for the remains of the Sinda beds taxon (see Figure 6a), it shows a greater development than in E. madagascariensis, its anterior region showing a morphology intermediate between that in T. pattersoni and S. fajumensis. This region is unknown for E. eremberti.
The taxon from the Sinda beds shares with T. pattersoni the presence of a short to almost absent ventral midline depression on the premaxillae (see Cadena, 2015;Gaffney et al., 2006), a much longer concavity being developed on the triturating surface of the premaxillae of E. madagascariensis and S. fajumensis. This character is unknown for E. eremberti.
As indicated above, Wood (2003) characterized T. pattersoni by the presence of high cheek emarginations. This character was more detailed defined in the amended diagnosis of T. pattersoni proposed by Gaffney et al. (2011), indicating that a cheek emargination is present but that it does not reach the dorsal edge of the orbits. The only member of Erymnochelyini in which this structure is totally absent is E. eremberti. It is very poorly developed in E. madagascariensis, not reaching the ventral edge of the orbits. The development of the cheek emargination described for T. pattersoni, surpassing the ventral edge of the orbits but not reaching the dorsal one, is shared with S. fajumensis. However, this structure reaches half the height of the orbits in both T. pattersoni and in the taxon described here, but not in S. fajumensis.
The prefrontals-frontals contact of all members of Erymnochelyini except S. fajumensis is characterized as transverse to slightly sinuous, lacking a medial protrusion (Pérez-García et al., 2017). However, variability in this character is identified for the taxon from the Sinda beds (see Section 2.6).
The presence of prefrontals shorter than the frontals was recognized to characterize E. eremberti and S. fajumensis, this condition being shared with the taxon from the Sinda beds. That character state contrasts with the condition in E. madagascariensis and T. pattersoni, which show prefrontals larger than the frontals (Pérez-García et al., 2017).
The presence or absence of a medial protrusion at the anterior end of the premaxillae, generating a premaxillary hook, has not been considered until now for the characterization of the different species of Erymnochelyini. It is present in the extant E. madagascariensis, being also well-developed in E. eremberti and in T. pattersoni. The absence of this structure in the taxon from the Sinda beds is exclusively shared with S. fajumensis.
The characters discussed here allow me to justify that the taxon of Erymnochelyini represented by cranial remains in the Sinda beds is not compatible with the extant E. madagascariensis. Thus, it shows more dorsally directed orbits, broader temporal emarginations, much higher cheek emarginations, prefrontals shorter than the frontals, absence of medial premaxillary hook, almost absent ventral midline depression on the premaxillae, a well-developed medial accessory ridge on each triturating surface of the lower jaw, and a very broad biting surface at the mandibular symphysis. It clearly differs from the Erymnochelyini members of the Paleogene record known from the skull. Thus, the presence of cheek emarginations and the absence of both a short anterior snout and a medial premaxillary hook represent notable differences with E. eremberti. It differs from S. fajumensis in characters such as a higher height of the suborbital space, similar to that of the orbits; very high cheek emarginations, reaching at least half of the orbits height; absence of the interorbital longitudinal depression only identified for some representatives of this species within Erymnochelyini; and a barely developed ventral midline depression on the premaxillae. Several characters show that, as it was preliminarily interpreted by Wood (2003) and Gaffney et al. (2011), the Lower Pliocene taxon of the Democratic Republic of Congo differs from the Cenozoic taxon of Kenya, T. pattersoni: wider anterior region of the skull; longer snout; prefrontals shorter than the frontals; more dorsally directed orbits; wider than tall nares; absence of medial premaxillary hook; height of the suborbital space similar to that of the orbits; absence of two welldeveloped accessory ridges on the palatal surface of each maxilla; and subtriangular and relatively long interparietal scute.
In addition to the aforementioned taxa, only another member of Erymnochelyini was named for the Neogene record: the Kenyan Upper Miocene K. williamsi. This taxon was only identified through postcranial remains, corresponding to an almost complete shell as well as several isolated plates (see Wood, 1983Wood, , 2003. The previously figured remains from the Sinda beds attributable to shells and plates of Erymnochelyini (see figure 7 in Yasui, Makinouchi, et al., 1992; figures 1-2 and plates 1-2 in Hirayama, 1992), and other unpublished shell remains deposited in the MRAC and compatible with the same taxon, show that it is not compatible with K. williamsi. Thus, they lack several of the characters that allow this taxon to be characterized, such as: medial carapacial keel; medial contact of the posterior costals restricted to last two pairs; nuchal wider than vertebral 1; pentagonal vertebral 1; first pair of marginals overlapping half of the lateroanterior nuchal margins (shared with S. fajumensis); straight anterior plastral margin, perpendicular to the axial axis (extragular protrusions being also identified in some specimens of S. fajumensis); and short gular scute (see Gaffney et al., 2011;Pérez-García et al., 2017;Wood, 1983Wood, , 2003. Therefore, the member of Erymnochelyini identified in the Sinda beds is here recognized as attributable to a new taxon, Mokelemys mbembe gen. et sp. nov.

| Implications
Considering the shells of the Erymnochelyini from the Sinda beds figured by Yasui, Makinouchi, et al., 1992 (see figure 7) and Hirayama, 1992 (figures 1 and 2; plates 1 and 2), as well as some partial shells and isolated plates deposited in the MRAC, Wood (2003) indicated that the shell of this taxon and that of E. madagascariensis are virtually identical. However, the presence of a relatively long gular in the taxon studied here, which reaches the entoplastron and overlaps the anterior region of this plate, contrasts with the remarkably reduced scute that characterizes E. madagascariensis (see Pérez-García et al., 2017; and references therein). The characters discussed as different from those in K. williamsi and E. madagascariensis, as well as others, allow the recognition of a combination of characters unique to the Mokelemys mbembe shell, including: absence of the very short distal margins on the seventh costals (i.e., more than two times shorter than those of the eighth ones) that characterizes T. pattersoni; absence of the more than six neurals in some species of the genus Eocenochelus; absence of the rounded protrusions in the second and third vertebrals of A. aegyptiaca, anterior to the inter-pleural sulci, the lateral margin of the vertebrals of that species being markedly sinuous, and the second vertebral being significantly wider than the third (see Pérez-García, 2021a; and references therein).
Unpublished shell remains from the Early Pliocene of Kenya (from two members of the Nachukui Formation in Lothagam, the Apak and the Kaiyumung members; and from the Kanapoi Formation in Kanapoi. For the age see Bobe et al., 2020;Stewart & Rufolo, 2020;van Bocxlaer, 2020;Bobe & Wood, 2021) were cited by Wood (2003) as attributable to T. pattersoni. He characterized the species by a skeleton from the Late Miocene (for the age see Geraads, 2019;Bobe & Wood, 2021) of lower member of the Nawata Formation, in Lothagam. Therefore, Wood (2003) proposed, based on the preliminary analysis of shell remains, that the species could be recorded in Kenya from sites separated by a time differential of as much as three to five million years. However, he stated that the shell of T. pattersoni was virtually indistinguishable from that of E. madagascariensis. In fact, in the subsequently published amended diagnosis of the species (see Gaffney et al., 2011), no shell character was also recognized as different between T. pattersoni and E. madagascariensis. Considering the other species of Erymnochelyini known in Kenya, Wood (2003) indicated that K. williamsi is not only present in Late Miocene record (in the lower member of the Nawata Formation, where it was defined), but also in the Early Pliocene. The Pliocene identification of the species is exclusively based on the find of an unpublished neural showing a keel in the Apak Member. Unpublished shells of Erymnochelyini from the Early Pliocene of Kenya (from the Kanapoi Formation, deposited in the National Museums of Kenya, in Nairobi), which do not share several diagnostic characters of this species, including its autapomorphies, also show keeled neurals. Therefore, the putative presence of both T. pattersoni and K. williamsi in the Pliocene record cannot currently be supported. In this context, the taxon defined here, Mokelemys mbembe, is the only representative of Erymnochelyini currently known for the Pliocene record.

| CONCLUSIONS
The fossil turtle fauna of the Democratic Republic of Congo is poorly known. However, several lineages can be identified there, generally lacking precise specific and even generic determinations, but including members of both Cryptodira and Pleurodira. In the early 1990s, the presence of a pleurodiran turtle that could be closely related to the extant E. madagascariensis (i.e., the only representative of the Podocnemididae Erymnochelyini lineage that is part of the current biodiversity) was identified in Lower Pliocene outcrops of the Sinda-Mohari region (Ituri Province, Democratic Republic of Congo). It was restricted to a few postcranial elements, including two relatively complete shells.
Knowledge about the lineage of Erymnochelyini has increased remarkably since the 1990s and, especially, during the 21st century. However, the taxon represented in the Democratic Republic of Congo has not been analyzed in detail until now. Thus, although some authors reported the presence of some cranial remains of Erymnochelyini from Sinda-Mohari, these were restricted to fragments of the anterior region of the skull, which remained unpublished until now.
The study of several partial skulls of the Erymnochelyini from Sinda-Mohari is performed here, so that the almost complete external cranial anatomy of the form represented there is characterized. Contrary to the proposal so far suggested, it is not compatible with E. madagascariensis. In fact, the taxon from the Democratic Republic of Congo cannot be assigned to any hitherto defined form. Therefore, it corresponds to a new form of Erymnochelyini, Mokelemys mbembe gen. et sp. nov. It is the only representative of that lineage recognized for the Pliocene record worldwide, being the first turtle species named based on fossils from the Democratic Republic of Congo.