New insights into the giant mustelids (Mammalia, Carnivora, Mustelidae) from Langebaanweg fossil site (West Coast Fossil Park, South Africa, early Pliocene)

Giant mustelids are a paraphyletic group of mustelids found in the Neogene of Eurasia, Africa and North America. Most are known largely from dental remains, with their postcranial skeleton mostly unknown. Here, we describe new craniodental and postcranial remains of the large lutrine Sivaonyx hendeyi and the leopard-size gulonine Plesiogulo aff. monspessulanus from the early Pliocene site Langebaanweg, South Africa. The new material of the endemic S. hendeyi, includes upper incisors and premolars, and fragmentary humerus, ulna and a complete astragalus. Its postcrania shares more traits with the living Aonyx capensis than the late Miocene Sivaonyx beyi from Chad. Sivaonyx hendeyi could therefore be tentatively interpreted as a relatively more aquatic taxon than the Chadian species, comparable to A. capensis. The new specimens of Plesiogulo comprise two edentulous maxillae, including one of a juvenile individual with incomplete decidual dentition, and a fragmentary forelimb of an adult individual. The new dental measurements point to this form being amongst the largest specimens of the genus. Both P3-4 differs from the very large species Plesiogulo botori from late Miocene of Kenya and Ethiopia. This confirms the existence of two distinct large species of Plesiogulo in Africa during the Mio/Pliocene, P. botori in the Late Miocene of Eastern Africa (6.1–5.5 Ma) and Plesiogulo aff. monspessulanus at the beginning of the Pliocene in southern Africa (5.2 Ma). Lastly, we report for the first time the presence of both Sivaonyx and Plesiogulo in MPPM and LQSM at Langebaanweg, suggesting that the differences observed from the locality may be produced by sedimentation or sampling biases instead of temporal replacement within the carnivoran guild.


Measurement
Sivaonyx hendeyi (Morales, Pickford & Soria, 2005) 1974 New material from Langebaanweg (LQSM and MPPM, see Table 6 Emended Diagnosis: Modified after . Sivaonyx of medium to large size. Robust P3 with distal accessory cusp. P4 with subquadrate outline. Paracone-metastyle compressed transversely, with a residual notch between them. Parastyle of medium size. Buccal cingulum strong joining the metastyle and parastyle. Protocone lingually projected from the paracone but joined to it by a crista oblique which joins the lingual crest of the paracone. Mesial valley present, but of modest dimensions. The hypocone is low and extensive, connecting the protocone and closing the tooth lingually. The median valley of the tooth is wide. p4 with very robust and high posterior cuspid located in a buccal position, with a small lingual platform. m1robust, with a crescentic-shape paraconid, mesiolingually located. Metaconid higher than the paraconid. Protoconulid very well developed. Talonid short and very wide, dominated by an extensive but relatively low hypoconid. Differential Diagnosis: Differs from S. bathygnatus in a larger size, and a more bunodont dentition. P4 with larger parastyle, and shorter paracone-metastyle edge, and less conical hypocone. Distal accessory cuspid of p4 more robust, with m1 less basined talonid and absence of hypoconulid; Differs from S. africanus in having a more developed p3, and more developed basal cingulum in p4-m1. m1 with lower height of trigonid and hypoconid, smaller protoconulid, and shallower talonid basin; Differs from S. beyi in having a more robust cingulum on p4. m1 more bunodont, with larger and more transversely orientated paraconid, more robust metaconid, and metaconid higher than paraconid. Smaller postcranial size, more robust femur comprising a thinner neck and a larger more proximally orientated head, less developed trochanters and less extended trochanteric fossa. Differs from S. ekecaman in a less robust P4 with smaller buccal cusps, and a less broad lingual platform in the distal part. m1 slenderer, with a more elongated paraconid, less conical metaconid, more robust hypoconid, a lesser development of the entoconid, entoconulid and a shallow talonid valley. Differs from S. soriae (=S. senutae) in a larger size and more robust cingulum, P4 without accessory cusp on the protocone, protocone more lingually projected, hypocone more mesiodistally extended, and less conical; m1 with paraconid more bucollingually wide, more robust metaconid, absent hypoconulid and shallower talonid basin. Differs from S. kamuhangirei, Enhydriodon dikikae and Djourabus dabba in smaller size. Differs from S. kamuhangirei in a higher hypoconid and a deeper talonid valley. Differs from E. dikikae in a slenderer mandibular corpus, P4 less robust, without accessory cusps on protocone, less conical protocone and hypocone. Shorter distal accessory cuspid of p4. m1 with lower trigonid cuspids, hypoconid more developed, and more reduced lingual cuspids on the talonid. Slenderer humerus and femur. Differs from D. dabba in a slenderer mandibular corpus, m1 more elongated and a less broad trigonid cuspids, paraconid with a lesser transverse orientation, and longer talonid with more robust cingulid and a non-reduced valley.
Description SAM-PQL-72229: It is a caniniform fragment of an upper incisor, tentatively determined as a left I2 (Figs. 3A-3D), with a single cusp lingually curved and a small wear facet on its tip. It represents the left half of the tooth. Its enamel is thick and wrinkled. There is a flat cingulum starting in the buccal part and running into the distal area, which is proximodistally enlarged. There is a crista from this point to the tip.
SAM-PQL-69635: It is a fragment of an indeterminate tooth, interpreted as a right I3 (Figs. 3E-3I). It is the right side of the tooth. Its shape suggests a conical crown. It is tall with a single cusp, lingually curved. As in SAM-PQL-72229, it has a wrinkled and thick enamel throughout the crown. A larger and crowned cingulum is present. There is a small projection of the cingulum in the lingual part. The tip has greater wear than those of the specimen SAM-PQL-72229.
SAM-PQL-52861: Left P2 mesially broken (Figs. 3J-3L; Table 1). It is elongated, oval in occlusal view, and unicuspid. It has two roots, but only the distal one is preserved. Its cusp is located in the mesial portion of the tooth. It is slightly worn. The tooth is mesially wider. There are crenulations and roughnes on the whole crown. There are also a sharp mesial and distal cristaes. A distally tall and crowned cingulum is present.
SAM-PQL-50000C: Complete right P3 (Figs. 3M-3O; Table 1). It is robust and bunodont with a lingual bulge. The main cusp is mesially located. There are two distal accessory cusps, a small one located on the most lingual point of the bulge and a larger one positioned on the distal corner of the premolar. The buccal wall is convex. A strong cingulum surrounds the whole crown, which displays a course texture.
SAM-PQL-60416: Distal epiphysis of a right humerus (Figs. 3P-3T; Table 3). The proximal part is broken. The distal epiphysis is broad and craniocaudally compressed. The preserved portion of the lateral epicondylar crest is laterally projected. There is a large surface area for the attachment of the M. extensor carpi ulnaris and the M. extensor digitorum lateralis. The trochlea and capitulum are relatively long in the lateromedial axis. Both radial and coronoid fossae are well development over the capitulum and trochlea respectively. The medial epicondyle is distomedially enlarged, increasing the surface area for the attachment of the M. pronator teres and the M. flexor digitorum profundus. A large and proximocaudally oval entepicondylar foramen is present. The olecranon fossa is deep. The supratrochlear foramen is absent.
SAM-PQL-21264: Fragmentary left ulna including the distal half portion of the diaphysis and the distal epiphysis (Figs. 3U-3W; Table 3). The lateral side is damaged, being proximodistally crushed and laterally collapsed. There is calcrete with yellow-orange iron minerals on the surface of the ulna, particularly in the cranial and medial faces. The interosseus tubercle is not preserved. Caudally, the ulna is sigmoid, with the medially projected crest for the attachment of the M. pronator quadratus. The distal epiphysis is robust, with a large and rostrally projected articular circumference and a large and round styloid process.
SAM-PQL-50120: Fragment of the left proximal femur. It is proximodistally broken (Figs. 3X-3AA; Table 3), and neither the trochanteric fossa nor the greater trochanter is preserved. The head is complete. It is round and the fovea capitis is located medially. The neck is short. In the caudal side, a robust lesser trochanter is caudomedially projected.
SAM-PQL-72172: Left astragalus (Figs. 3BB-3GG; Table 3). The body is trapezoidal in shape. There is a broad, shallow astragalar trochlea, and a noticeable medioproximal projection of the astragalar tubercle (plantar tendinal groove) (Fig. 3BB). It is a relatively robust, deep groove for the tendons of the plantar flexor muscles, and extends to the ventral side of the bone (Figs. 3CC and 3DD). The medial tibial facet is located on the medial portion of the trochlea that merges into the plantar tendinal groove (Fig. 3EE). The lateral area of the trochlea has a drop-shape fibular facet, slightly concave, and dorsoventrally higher (Fig. 3FF). In ventral view, the ectal facet is wider than the sustentacular facet. Both facets are separated by a deep groove, in which there are several foramina. The neck is relatively short, mediolaterally wide, and distomedially orientated (Figs. 3BB and 3GG). The head is lateromedially broad, dorsoventrally compressed and the navicular articular surface is strongly convex. It is orientated parallel to the mediolateral axis, with the lateral border being slightly more dorsally elevated to the medial one (Fig. 3GG).
Most of the African giant bunodont otters are represented by very scarce and fragmentary remains, which make any new fossils significant in order to understand the role and lifestyle of these peculiar mustelids. The first remains of an extinct large otter in South Africa was reported by Stromer (1931), who erected Enhydriodon africanus on the basis of a fragmentary right M1 and a right hemimandible with p3-m1 from the early Pliocene deposit of Klein Zee, Namaqualand, South Africa. A second discovery of the same taxon was reported by Hendey (1974Hendey ( , 1978b in LBW, a similarly aged early Pliocene locality situated 500 km south of Klein Zee. He described new dental and postcranial material of E. africanus from the beds 3aN and 3aS of the MPPM, comprising two hemimandibles (SAM-PQL-9138 and SAM-PQL-50000A) and a left P4 (SAM-PQL-50000B), and tentatively assigned to this species a left femur (SAM-PQL-41523), two distal epiphysis of radii (SAM-PQL-50001A, B) and one astragalus (SAM-PQL-50117) (Hendey, 1978b). Morales, Pickford & Soria (2005) moved the LBW taxon to a new species, separate from Klein Zee, establishing Enhydriodon hendeyi, which included SAM-PQL-50000A (holotype), SAM-PQL-9138, SAM-PQL-50000B and SAM-PQL-41523. The astragalus SAM-PQL-50117 was re-interpreted as Orycteropus Geoffroy Saint Hilaire, 1796, a relative of the living aardvark (Pickford, 2005). The same year,  reassigned the species from LBW and Klein Zee to the genus Sivaonyx, while retaining the species names. Despite the similar age and size, it is widely accepted that both South African bunodont otters are different species (Pickford, 2007;De Bonis et al., 2008;Haile-Selassie, 2008;Peigné et al., 2008;Lewis, 2008;Haile-Selassie & Howell, 2009;Werdelin & Peigné, 2010;Geraads et al., 2011;Grohé et al., 2013). Sivaonyx hendeyi can be distinguished from S. africanus (Fig. 4) in several dental traits summarized in the differential diagnosis of this manuscript. In general terms, S. hendeyi possesses a more robust dentition, including more robust cingulids, a more developed p3 and a m1 with lower crown, smaller protoconulid and shallower talonid valley to that of S. africanus. The new dental material of S. hendeyi described, include new data on the upper incisor and upper premolars, previously unknown. All the new dentition is robust and have strong cingula (Fig. 3), which are in consonance with the known robust and bunodont dentition (Hendey, 1974(Hendey, , 1978b.  . The P3 (SAM-PQL-50000C) of S. hendeyi also has the typical bunodont otter's morphology, with a convex buccal wall, a lingual bulge and a robust cingulum surrounding the whole tooth, resembling those of E. dikikae and En. terraenovae; and unlike to those of S. ekecaman from Sagatia, Magabet Fm., Kenya which distal area is almost circular in occlusal outline and no distal accessory cusp is present (Morales, Pickford & Soria, 2005;. It is also dissimilar to Si. melilutra in having a wider P3 . The morphological differences of the dentition between S. hendeyi and the other African species of Sivaonyx (S. ekecaman, S. soriae, S. beyi, and S. kamuhangirei), as well as En. dikikae and D. dabba has been largely detailed (Werdelin, 2003b;Morales, Pickford & Soria, 2005;Pickford, 2007;Peigné et al., 2008;Geraads et al., 2011) and is summarized in the differential diagnosis section. Metrically, the dentition of S. hendeyi is comparable to those of S. africanus, S. ekecaman from the Late Miocene-Early Pliocene of Ethiopia and Kenya (Werdelin, 2003b;Morales, Pickford & Soria, 2005;Morales & (Tables 1 and 2). All of them are larger than the type species of the genus S. bathygnathus from Hasnot, Punjab, Pakistan (Late Miocene) (Pickford, 2007) and also larger than the Kenyan and Ethiopian S. soriae Haile-Selassie, 2008) from the Late Miocene, but smaller than Sivaonyx kamuhangirei  from Kazinga, Uganda, c.a., 3.5 Ma (Pickford, 2007).
Postcranial remains of bunodont otters are even scarcer than craniodental material. With the exception of S. beyi, S. hendeyi, represents the only species of Sivaonyx for which postcranial bones have been described . Additional large otters with postcranial bones recovered are E. dikikae, Enhydriodon sp., and Torulutra ougandensis Petter, Pickford & Howell, 1991 from Ethiopia and Uganda (Petter, 1994;Haile-Selassie, 2008;Geraads et al., 2011;Werdelin, Lewis & Haile-Selassie, 2014) as well as En. terraenovae from Florida (Lambert, 1997) and the late Pliocene Satherium piscinarium (Leidy, 1873) from the Hagerman fauna (Bjork, 1970), which was a large otter resembling the living South American Pteronura Gray, 1837, but with sharper dentition. The morphology of the fragmentary humerus SAM-PQL-60416 is similar to that of the S. beyi and the living A. capensis (Fig. 5). It shares with both lutrines the expanded medial epicondyle, which is more expanded in the Chadian one, and an enlarged lateral epicondylar crest. The dental proportions of S. hendeyi and S. beyi are analogous, but the postcranial skeleton of the former is larger (Figs. 5 and 6; Tables 2 and 3). The distal epiphysis of the humerus of T. ougandensis from Middle Awash and the North American En. terraenovae and Sa. piscinarium are similar, but smaller in size (Table 3). Moreover, the distal epiphysis SAM-PQL-60416 is rostrocaudally compressed, similar to A. capensis and En. terraenovae and unlike E. dikikae, which less rostrocaudal compression. The presence or lack of that trait in S. beyi was not described by Peigné et al. (2008). It is also differs from a distal epiphysis of a fragmentary humerus of a medium sized otter from Nkondo Uganda (5-4.5 Ma), determined as Enhydriodon sp. by Petter (1994, pl. 1, fig. 3-4). Later,  transferred the dentition associated with this material to S. kamuhangirei, but they did not mention the humerus. Based on the relatively smaller dimensions of that humerus, Peigné et al. (2008) noted that it may belong to a smaller-sized species of Sivaonyx or to a different genus. We agree with their suggestion, and since its relatively small size compared to specimen SAM-PQL-60416, we discard its designation to S. kamuhangirei or S. hendeyi, being closer in shape and size to the medium size T. ougandensis, also present in Nkondo (Petter, Pickford & Howell, 1991). The fragmentary ulna (SAM-PQL-21264) is essentially identical to those of S. beyi, and shares traits with the semifossorial mustelid M. capensis (living honey badger). The distal area of the diaphysis of both S. hendeyi and S. beyi has a distinct medially projected crest for the attachment of the M. pronator quadratus. The articular circumference is robust and the rostrally projected styloid process is large and round. The ulna of Sivaonyx spp., is distinguished from semiaquatic otters such as the extant A. capensis and the extinct En. terraenovae. The former has a reduced crest, and a reduced articular process (Lambert, 1997).
Both preserved femora of S. hendeyi are of a similar size, while the proximal projection of the head of SAM-PQL-50120 is less than SAM-PQL-41523. The femora from LBW are smaller than those of S. beyi, and E. dikikae but larger than the extinct En. terraenovae, Sa. piscinarium and the living A. capensis ( Fig. 6; Table 3). Following the sample and methodology of Samuels, Meachen & Sakai (2013), we calculated the femoral robustness index (FRI) and femoral epicondylar index (FEI) of S. hendeyi (SAM-PQL-41523), and the other extinct bunodont otters analyzed (Table 3), obtaining for both indices the highest values of the whole sample. The FRI value of S. hendeyi is similar to those of the extinct S. beyi, En. Terraenovae, and Sa. piscinarium, analogous to the largest living otters (Enhydra Fleming, 1822 and Pteronura), and higher to those of the living marine otter [Lontra felina (Molina, 1782)], smooth-coated otter [Lutrogale perspicillata (Geoffroy Saint Hilaire, 1826)], North American river otter [Lontra canadensis (Schreber, 1777)] and A. capensis, suggesting a possible allometric factor. The FEI value of S. hendeyi is analogous to the living African clawless otter and the Asian small-clawed otter [Amblonyx cinereus (Illiger, 1815)]. Interestingly, the lowest value of the analyzed extinct otters corresponds with the largest one, E. dikikae, with similar values to the semifossorial musteloids [American badger Taxidea taxus (Schreber, 1777), and the striped skunk Mephitis mephitis Hutchinson, 2019; Parsi-Pour & Killbourne, 2020). A robust femur indicates better abilities to resist bending and shearing stress, accordingly, the FRI is high in living semiaquatic and semifossorial carnivorans (Samuels, Meachen & Sakai, 2013;Fabre et al., 2015;Botton-Divet et al., 2016;Parsi-Pour & Killbourne, 2020). Higher FEI values indicate relatively large area available for the origins and insertion of the muscles gastrocnemius and soleus, used in extension of the knee and plantar-flexion of the pes (Samuels, Meachen & Sakai, 2013), with the lutrines having the highest values of the carnivorans. In the context of lutrinae, Am. cinereus and A. capensis are less linked to water bodies than other river otters and are interpreted by some authors as the least aquatic living otters (Perrin & Carugati, 2000a, 2000bAngelici et al., 2005;Kruuk, 2006;Lewis, 2008).
The astragalus (SAM-PQL-72172) of S. hendeyi is comparable in general morphology to those of S. beyi, Enhydriodon sp. (Werdelin, Lewis & Haile-Selassie, 2014), and A. capensis (Figs. 6H-6K). However, it differs from the large Enhydriodon sp., in having a relatively smaller head and thinner neck, and a bigger distal projection of the astragalar tubercle. It is not known if that projection is also present in S. beyi because the preserved one lacks the astragalus tubercle . While SAM-PQL-72172 is very similar to the astragalus of A. capensis, it also displays some differences, such as a shallower and mediolaterally wider trochlea, and a more medioproximal projected astragalar tubercle, including a more robust and deeper groove for the tendons of the plantar flexor muscles (Figs. 6H-6K). The relatively larger trochlea of S. hendeyi implies a relatively lateromedially wider distal tibia epiphysis, to that of living A. capensis. Based on the new astragalus (SAM-PQL-72172) (Figs. 3BB-3GG; Table 3), and its similarities with the astragal of A. capensis (Fig. 6), the astragalus SAM-PQL-50117 was correctly re-interpreted by Pickford (2005) as not belonging to an otter. The fragmentary radii (SAM-PQL-50001A and B) (Fig. 7) assigned by Hendey (1978b) to the bunodont otter from LBW, does not belong to S. hendeyi. A re-examination showed that SAM-PQL-50001A and B have a sharper and more distally projected styloid process, and a deeper, and enlarger groove for the tendon of the M. abductor digiti I longus to that of the living otter A. capensis (SAM-ZM-41474) (Figs. 7A-7J). Medially, a relatively large proximodistal crest-like projection is present, contrary to the round and reduced one present in SAM-ZM-41474. In distal view, both distal epiphyses are craniocaudally wide. However, SAM-PQL-50001A and B are distinguished to SAM-ZM-41474 in the smaller lateral groove for the tendon of the common digital extensor. All these traits suggest SAM-PQL-50001A and B cannot be assigned to S. hendeyi or other known mustelid from LWB on the basis of its morphological differences and measurements (Hendey, 1978b; Tables 4 and 5 Table 4).
Living lutrinae have a large shape diversity, which may be related to the large evolutionary and ecological variation of this group (Botton-Divet et al., 2016). While all species are semi-aquatic, the amount of time they spend in water differs with their type of habitat, swim, and food (Kruuk, 2006;Botton-Divet et al., 2016). The fragmentary nature of the known postcranial skeleton of S. hendeyi makes it difficult to suggest paleobiological interpretations about its locomotion or lifestyle. Previous interpretations were based entirely on its femur. The first one was made by Lewis (2008), who, based on the complete femur SAM-PQL-41523, inferred that this taxon could be a locomotor generalist in comparison with some other living and extinct Mio-Pliocene otters, but she did not reject the possibility of being semiaquatic, like extant river otters, or occasionally aquatic, like many non-lutrine mustelids. In the same year, Peigné et al. (2008) analyzed a relatively complete skeleton of S. beyi including the femur. He found that the femur of S. beyi was slenderer than that of S. hendeyi, and suggested that the Chadian specie was a terrestrial predator with poorly developed aquatic adaptations. The expanded sample of postcranial material of S. hendeyi, shows that this extinct otter shares more traits with A. capensis -for example, rostrocaudal compression of the distal epiphysis of the humerus, a more robust diaphysis of the femur, which suggests a relatively reduced total length (higher FRI), and a similar astragalus-than with S. beyi, indicating that S. hendeyi could be interpreted as a relatively more aquatic taxon than the former. Moreover, discussion of the paleoecology of extinct carnivores must take into account that all the categories described in the life history of the living ones such as cursorial, semifossorial, semiaquatic, arboreal, or terrestrial-generalist showing a complex continuum of behaviors none of them mutually exclusives (see Samuels, Meachen & Sakai, 2013;Fabre et al., 2015). Both semiaquatic and semifossorial carnivorans, have several shared anatomical traits, such as the rostrocaudal compression of the distal epiphysis of the humerus, relatively long olecranon process of the ulna and a medial projection of the diaphysis of the ulna for the attachment of the M. pronator quadratus, and an enlarged femoral epicondyles (Samuels, Meachen & Sakai, 2013;Fabre et al., 2015;Botton-Divet et al., 2016;Kilbourne & Hutchinson, 2019;Parsi-Pour & Killbourne, 2020;A. Valenciano, 2020, personal observations). We can preliminarily infer the locomotion or lifestyles of these extinct bunodont otters, based on the overall morphology of their skeletons. Enhydriodon dikikae and S. beyi can be interpreted as a more generalized terrestrial mustelid (Lewis, 2008;Peigné et al., 2008;Geraads et al., 2011) than S. hendeyi. Although more information about other bones are needed to assess the locomotion of S. hendeyi, we hypothesize it may have had a relatively more semiaquatic locomotion closer to that of the living A. capensis or Am. cinereus, but without excluding some digging capability. Similarly, S. hendeyi and S. beyi seems to have a lower association with water bodies than the North American En. terraenovae and Sa. piscinarium whose bones point to a more aquatic locomotion, similar to the living Lutra and Pteronura respectively (Bjork, 1970, Lambert, 1997.
SAM-PQL-47086: edentulous left maxillary of a juvenile specimen, with an abraded surface. The buccal roots of the DP3 are present (Figs. 8C and 8D). It also preserves the lingual roots of the DP4. The lingual area is mesiodistally enlarged, with an incisures in the middle of the base of the lingual platform (Fig. 8D). It also has the distal root for the P3. It is not clear if the roots of the definitive P4 are present. Due to ontogenetic stage of SAM-PQL-47086, the definitive M1 was located deeper on the maxilla, which is broken in the occlusal plane, and allows the observation of the M1 alveolus. This alveolus differs from SAM-PQL-40117 (Fig. 8D). It is mesiodistally enlarged and shows a clear concavity in its middle point (Fig. 8D), reflecting the shape of the definitive M1, which is characteristic in Plesiogulo. It has a well-developed lateral epicondylar crest. The distal epiphysis is rectangular in distal view. The olecranon fossa is deep and proximodistally high. Medially, the supracondyloid process, is broken, so it only preserves half of the entepicondylar foramen. The medial epicondyle is abraded, but distally it is projected in caudal direction (Figs. 9D and 9E).

SAM
SAM-PQL-L3440: fragmentary, abraded right distal part of a radius, including the distal epiphysis (Figs. 9F-9J). The diaphysis is craniocaudally curved similar to the radius of Plesiogulo (SAM-PQL-40042). In cranial view, a crest is present exceeding half the diaphysis (Fig. 9F). It can be interpreted as a scar for the most distal attachment of the M. supinator. Laterally on the radius, an edge with a rough surface from the proximal-most part to the lateral border is present, occupying half of the preserved diaphysis (Fig. 9G). It is interpreted as the most distal part of the interosseous border. The ulnar notch is oval and craniocaudally elongated. The knob shaped medial styloid process is the insertion for the M. brachioradialis (Fig. 9I). The styloid process is pointed. In cranial view, there are three grooves on the cranial border of the distal epiphysis (Fig. 9H). The medial one is for the tendon of the M. abductor digiti I. A lateromedially enlarged one in the middle of the epiphysis, over the cranial border. It is relatively deeper and represent the groove for the tendon of the M. extensor carpi radialis. A smaller lateral groove exists for the tendon of the M. extensor digitorum comunis. The distal epiphysis is craniocaudally wide (Fig. 9J).
SAM-PQL-6414: fragmentary right proximal epiphysis of an ulna (Figs. 9K-9N). It is abraded and has longitudinal cracks along the main axis of the bone similar to the humerus (SAM-PQL-6246). The olecranon process is missing and the bone is broken proximally at the trochlear notch area, and distally at the most proximal part of the interosseous border. The overall shape and size of the preserved ulna is similar to the one of Plesiogulo (SAM-PQL-40042), especially the medial coronoid process and the insertion of the M. brachialis, located below this process, which is proximodistally enlarged (Figs. 9L and  9M). There is a very noticeable roughness for the interosseous border (Fig. 9K) in the lateral side of the bone.
The systematic position of Plesiogulo within mustelidae and the living wolverines (Gulo) is complicated by several shared common craniodental traits and the long divergence time among these genera. Some researchers supported an ancestor-descendant relationship between Plesiogulo and Gulo (Viret, 1939;Kurtén, 1970;Kurtén & Anderson, 1980), whereas others considered that Plesiogulo is included in an extinct lineage without living descendants (Zdansky, 1924;Hendey, 1978b;Harrison, 1981;Xiaofeng & Haipo, 1987;Alcalá, Montoya & Morales, 1994;Sotnikova, 1995;Montoya, Morales & Abella, 2011;Samuels, Bredehoeft & Wallace, 2018). Valenciano et al. (2017bValenciano et al. ( , 2020) support Plesiogulo and the early Miocene Iberictis forming a sister group of Gulo based on morphological traits and cladistic analysis; considering these three taxa as member of the tribe Gulonini. Molecular analyses agree with the existence of a "Gulo lineage" in which there is a close relationship between Gulo, Martes and Pekania among other taxa, with a divergence time for Gulo around 7.6-5.5 Ma (Koepfli et al., 2008;Li et al., 2014;Malyarchuk, Derenko & Denisova, 2015;Law, Slater & Mehta, 2018). It does not exclude a priori that extinct taxa like Plesiogulo may belong to this clade. Nevertheless, Samuels, Bredehoeft & Wallace (2018) reinforced a sister group relationship between Gulo and Pekania through a new species of Gulo from the early Pliocene (4.9-4.5 Ma) of Tennessee (USA), considering Plesiogulo to be convergent with Gulo. Thus, morphologically an Early-Late Miocene clade of mustelids comprising Plesiogulo is well documented , and the systematics position of Gulo is debatable, depending of the divergence time of Gulo-Martes-Pekania and the absence of a direct systematic relationship between Gulo sudorus Samuels, Bredehoeft & Wallace, 2018 and Plesiogulo.
The current novel material described represents the first new specimens of this taxon from the locality since the Hendey's work in the 1970's ( Figs. 8 and 9). The new dental measurements of the P3 and P4 based on the alveoli and SAM-PQL-40042, point to this form falling among the largest specimens of the genus ( Fig. 10; Tables 1 and 2). The morphology and proportions of the M1 of this mustelid from LBW is unknown. However, we can infer it based on the alveolus of SAM-PQL-47086. Although it belongs to a juvenile individual, it is noted that it was not erupted, and the preserved part of the alveolus in which the lingual platform was placed, represents a relatively accurate dimension for the maximum length of the M1. It also suggests that the lingual platform has an inflexion in the middle of the crown (Fig. 8D), being a distinctive trait for the genus. The inferred dimension of SAM-PQL-47086 are close to the M1 of P. monspessulanus from Las Casiones (Alcalá, Montoya & Morales, 1994) (Fig. 10; Table 2). SAM-PQL-47086 also preserved part of the upper deciduous dentition. The only previous deciduous dentition described in Plesiogulo is the lower one of the North American P. marshalli (Harrison, 1981), therefore it is impossible to make a direct comparison. Interestingly, the DP4 of SAM-PQL-47086 is similar to the definitive M1, which indicates the possession of an enlarged lingual platform, while the contrary pattern occurs in the living gulonini G. gulo, which DP4 and M1 are reduced (Figs. 8C-8E). These differences in the M1 was interpreted by Valenciano et al. (2020) as alternative strategies to crush bones; food processing in Gulo is more focused on the postcanine dentition and in the carnassials (P4-m1) and in Plesiogulo on the most distal dentition comprising the carnassials and the M1. In general terms, the proportions of the dentition of the South African Plesiogulo are close to the holotype of P. monspessulanus and the one from Las Casiones ( Fig. 10; Tables 1 and 2). The m1 metaconid is absent in the holotype but it is present in the specimens from Las Casiones, Venta del Moro and Yushé (Viret, 1939;Teilhard de Chardin, 1945;Alcalá, Montoya & Morales, 1994;Montoya, Morales & Abella, 2011). The presence or absence of this trait in the one from LBW is not clear. The whole lingual part of the m1 of SAM-PQL-21570 is missing, and in SAM-PQL-28394 is quite worn. The worn area where the metaconid may be in the tooth, is widen at its base, suggesting it was really developed. This feature is variable in this genus and is not very useful for taxonomic analysis. However, the loss of the m1 metaconid in several caniform carnivorans such as canids, temnocyonine amphicyonids and mustelids has been interpreted as a derived trait ( Van Valkenburgh, 1991;Hunt, 2011;Valenciano et al., 2016Valenciano et al., , 2017bValenciano et al., , 2019, and in this case the presence of the metaconid in the oldest specimens from Las Casiones, Venta del Moro and the early Pliocene of Yushe indicate a more primitive stage for this trait. The classical material of Plesiogulo from LBW was found in LQSM (Hendey, 1978b;Werdelin, 2006), with the exception of SAM-PQL-40042, which according to Hendey (1978b) could come either from LQSM or from the lowermost levels of bed 3aS from MPPM. The new material confirms the presence of Plesiogulo in both LQSM and MPPM at LBW (Table 6).  Morales, 1984) from Venta del Moro (Late Miocene, Spain). (C) M1. Estimation of the paratype of P. botori (M1 ADD-VP-1/10) based on picture provided in Haile-Selassie, Hlusko & Howell (2004). Sources: (Zdansky, 1924;Teilhard de Chardin, 1945;Kurtén, 1970;Hendey, 1978b;Harrison, 1981;Morales, 1984;Alcalá, Montoya & Morales, 1994;Sotnikova, 1995;Haile-Selassie, Hlusko & Howell, 2004;Koufos, 2006;Montoya, Morales & Abella, 2011;Samuels, Bredehoeft & Wallace, 2018;Grohé, in press; and this work).
Full-size  DOI: 10.7717/peerj.9221/ fig-10 An additional very large species of Plesiogulo is present in Eastern Africa in deposits dated to 5.5-6.0 Ma (Haile-Selassie, Hlusko . Plesiogulo botori occurs in the localities of Narok (type locality), Lemudong'o Fm., Kenya and in Adu Dora, Middle Awash, Afar Depression, Ethiopia (Haile-Selassie, Hlusko & Howell, 2004). It represents the largest species of the genus (Figs. 10 and 11). However, only the upper dentition is known. Additionally, more material of Plesiogulo was described in contemporaneous sediments of Kenya in the Lukeino Fm., aged from the late Miocene, 6.1-5.7 Ma (Morales, Pickford & Soria, 2005). They assigned to P. praecocidens a right complete P4 from the locality of Cheboit and a fragmented M1 from the locality of Kapcheberek (Morales, Pickford & Soria, 2005). Recently, after a revaluation of these teeth, Morales, Pickford & Valenciano (2016) excluded the P4 for the genus and re-assigned the fragmentary M1 to P. botori. The P3 and P4 of the maxilla of SAM-PQL-40042 are slender and smaller to those of P. botori (Fig. 11C). The P4 protocone is partially broken in the specimen SAM-PQL-40042, however the overall morphology and its proportions can be inferred and it is distinguishable to those of P. botori (Fig. 11). Based on the above, the taxonomy of the large Plesiogulo from LBW is complex. The largest species from Eurasia (P. monspessulanus) and Africa (P. botori) are represented by incomplete holotypes; consisting of lower or upper dentition respectively. When Hendey described the material from LBW in 1978, he assigned it to P. monspessulanus (Hendey, 1978b) as only the Euroasiatic species was known. Since then, new large Plesiogulo described from the late Miocene of the Iberian Peninsula have been assigned to P. monspessulanus, adding more variability to the taxon (Alcalá, Montoya & Morales, 1994;Montoya, Morales & Abella, 2011), although with certain doubt for the remains from Venta del Moro (Montoya, Morales & Abella, 2011). These findings are however still insufficient to confidently assign the LBW material to P. monspessulanus. In the absence of complete diagnostic dentition such as M1 and a better-preserved P4 and m1, to make a direct comparison with the holotypes of P. botori and P. monspessulanus, and based on the similarities with the known lower dentition of the later, we refer the LBW sample to Plesiogulo aff. The postcrania of Plesiogulo spp., are poorly known with the exception of the fragmentary forelimbs of P. marshalli and P. lindsayi described by Harrison (1981), and the abundant but fragmentary skeleton of P. aff. monspessulanus from LBW (Hendey, 1978b) (Table 5). The new forelimb remains complement the previous one of SAM-PQL-40042. These fragmentary humerus, ulna and radius, have a relatively equivalent size and preservation, which suggest they may belong to the same individual ( Fig. 9; Table 5). They represent a smaller individual than SAM-PQL-40042, which also possesses the largest dentition of Plesiogulo from LBW. Based on the preservation and that they are from MPPM, they do not belong to the individuals SAM-PQL-28394 or SAM-PQL-21570, which came from LQSM. The overall morphology of the forelimb of P. aff. monspessulanus is robust. The humeri SAM-PQL-6246 and SAM-PQL-40042 are larger than the one of P. marshalli (Table 5), having a relatively better developed lateral epicondylar crest and a more enlarged medial epicondyle to those of the North American species. However, these differences can be interpreted as allometric differences because of the larger size of the South African taxon, instead of biomechanical implications. The ulnae SAM-PQL-6414 and SAM-PQL-40042 are quite similar to that of P. marshalli and P. lindsayi, but larger (Table 5). We calculated the olecranon length index (OLI) and ulnar robustness (URI) for the ulnae of P. aff monspessulans SAM-PQL-40042, P. marshalli and G. gulo following Samuels, Meachen & Sakai (2013) ( Table 5). The OLI indicates the relative mechanical advantage of the M. triceps brachii used in the elbow extension (Samuels, Meachen & Sakai, 2013). The value of SAM-PQL-40042 is higher than P. marshallli, and equivalent to the scansorial Chinese ferret-badger Melogale moschata (Gray, 1831) and close to the semifossorial honey badger Mellivora capensis and the generalist Melursus ursinus (Shaw, 1791). The URI indicates the degree of robustness of the ulna and its ability to resist bending and shearing stresses and relative area available for the origin and insertion of forearm and manus flexors, pronators and supinators (Samuels, Meachen & Sakai, 2013). Plesiogulo aff. monspessulanus from LBW have a very robust ulna with the highest value for URI, only comparable with the saber-tooth felid Smilodon fatalis (Leidy, 1868), which was classified as terrestrial by Samuels, Meachen & Sakai (2013).
Based on these indices P. aff monspessulans possesses a very robust ulna with a regular olecranon development. Hendey (1978b) described the skeleton of P. aff. monspessulanus from LBW as similar to G. gulo, but with the limb bones, tarsals and metapodials less elongated and more stoutly proportioned, which reflect the stoutness of this species and its heavier musculature. The curvature of the ulna SAM-PQL-40042 and the radii SAM-PQL-40042 and SAM-PQL-L3440 point toward this trend. The possession of a very robust forelimb in P. aff monspessulans could suggest that this species was an ambush predator, in a similar mode to that of the North American Sm. fatalis (Meachen-Samuels & Van Valkenburgh, 2010, Lewis, 2018, which would surprise and subdue their preys rather than chase them down.
The co-occurrence of three mustelids in LBW can be explained by dietary resource/ ecological niche partitioning, in which none of them seem overlap. It has been suggested that Plesiogulo was an inhabitant of open, and grassy plains (Kurtén, 1970;Harrison, 1981), with the role of an ambush predator, with durophagous abilities . Werdelin (2015) stated that we can only speculate about the habits and diet of extinct bunodont otters, because they differ from any living relatives, highlighting that they appear to have been somewhat more terrestrial than living otters, but they are always found in association with large bodies of water. Even though the preserved craniomandibular remains of S. hendeyi are not complete, following the ecomorphological analysis of extant small carnivorans of Friscia, Van Valkenburgh & Biknevicius (2007), which include the extant otters Enhydra and Amblonyx, which feed on mollusks and crustaceans, it can be inferred that S. hendeyi possesses a comparable dentition, including larger molar grinding areas, larger post-canine dentitions, and wider fourth premolars, typical of omnivores/hard-object feeders. Also, a diet based on relatively hard items, as in the living Aonyx capensis has been previously suggested for other African bunodont otters (Sivaonyx spp., and Enhydriodon spp.) hypothesizing that they fed on armored catfishes, mollusks, and other armored preys (Pickford, 2007;Peigné et al., 2008;Geraads et al., 2011;Werdelin, 2015;Werdelin & Lewis, 2017). Sivaonyx hendeyi could have had a similar role to that of the living A. capensis, having a locomotion relatively similar to it while less semiaquatic. Its bunodont and robust dentition suggests an even more durophagous diet to those of A. capensis; and M. benfieldi can be interpreted as a small-medium opportunistic carnivoran analogous to the living honey badger M. capensis.