First African record of the Miocene Asian mouse-deer Siamotragulus (Mammalia, Ruminantia, Tragulidae): implications for the phylogeny and evolutionary history of the advanced selenodont tragulids

New remains of the small tragulid Dorcatherium songhorensis Whitworth, 1958 from the Early Miocene fossil site of Napak XXI (Uganda) include the first significant sample of postcranial bones from this species ever described. The limb bones of this tragulid are very similar to that described in the Miocene Asian long-legged tragulids of the genus Siamotragulus Thomas et al., 1990, a type previously unknown in the African Miocene. A cladistic analysis links D. songhorensis to a Siamotragulus clade as its basal offshoot, so we propose the name Siamotragulus songhorensis (Whitworth, 1958) for this species. Also, the Siamotragulus clade belongs to a monophyletic group that includes Afrotragulus Sánchez et al., 2010 and the extant Asian genera Moschiola and Tragulus. This inclusive clade is characterized by both a derived selenodont dentition and an advanced postcranial skeleton. Additionally Siamotragulus shows some cursorial refinements reflected in its postcranial skeleton including the pecoran-like metatarsals III–IV. Siamotragulus songhorensis shows that the genus Siamotragulus was not endemic to Asia as previously thought, and that a highly diverse guild of tragulids, including different members of the advanced selenodont clade, inhabited Africa as early as the Early Miocene (19–20 Ma).


Introduction
The Tragulidae (chevrotains and mouse-deer) are the most basal of extant Ruminantia, and the only living remnant of an ancient radiation of ruminants that produced successive sister groups to the Pecora (R€ ossner 2007; S anchez et al. 2010). They include the smallest living cetartiodactyls and survive today as Old World tropical relics: Moschiola in India and Sri Lanka, Tragulus in South-East Asia and the Philippines, and Hyemoschus in Africa from Sierra Leona to Uganda (Grubb 1993;Nowak 1999;Meijard & Groves 2004;Groves & Meijard 2005;R€ ossner 2007). The fossil record of Palaeogene tragulids is extremely poor (M etais et al. 2001;Tsubamoto et al. 2003;M etais & Vislobokova 2007). The Late Eocene Asian ruminant Archaeotragulus krabiensis M etais et al., 2001 is considered the most basal of tragulids, and also constitutes the only Palaeogene record of the family. This lack of Palaeogene remains opens a gap in the fossil record that extends up to the Early Miocene, when tragulids suddenly reappear with a high diversity in Africa, Asia and Europe (Whitworth 1958;Hamilton 1973;Mein 1989;Mein & Ginsburg 1997;Gentry et al. 1999;Ginsburg et al. 2001;Pickford 2001Pickford , 2002R€ ossner 2007;Quiralte et al. 2008;S anchez et al. 2010). Tragulids experienced a great evolutionary success during the Miocene, and where their habitat preferences were met they even successfully competed with pecoran ruminants (R€ ossner 2004).
Classically it has been thought that only the genus Dorcatherium inhabited Africa during the Miocene, but this scenario is in need of urgent revision (see e.g. R€ ossner 2007; S anchez et al. 2010). The description of the highly derived and tiny Afrotragulus S anchez et al., 2010 from the late Early Miocene of Kenya (probably present also in Southern Africa) showed the presence in Africa of a new type of tragulid different from Dorcatherium and increased the morphological diversity of African tragulids. Moreover, Afrotragulus demonstrated that the classical use of size as the main taxonomic criterion for the Tragulidae (see e.g. Arambourg & Piveteau 1929;Colbert 1935;Whitworth 1958;West 1980;Gaur 1992;Ginsburg et al. 2001;Pickford 2001Pickford , 2002Morales et al. 2003;Quiralte et al. 2008) was critically flawed, and that morphology is of the upmost importance for understanding tragulid systematics. In this work we present another step towards the systematic revision of the Miocene African tragulids.
During the field campaigns of 2008À2013, fossils of a small tragulid were discovered in the Napak XXI fossil site (Uganda; Fig. 1). The lower molars of this form are virtually indistinguishable from those of Dorcatherium songhorensis Whitworth, 1958, a species previously cited from other Napak fossil sites (Pickford 2002). The type material of D. songhorensis from Songhor lacked any postcranial bones (Whitworth 1958), and although some postcranial fossil material from other sites in Napak and from the Sperrgebiet (Namibia) were previously described (Pickford 2002;Quiralte et al. 2008), it was fragmentary and lacked significantly different features from what was then known about the African forms. Contrary to this, the majority of fossils from Napak XXI are postcranial remains in a fairly good state of preservation, and some of them shed significant light on the nature of the appendicular skeleton of D. songhorensis: surprisingly, the limb bone morphology of this species corresponded to a type never described before in any African tragulid, but already known in the Asian Miocene long-legged genus Siamotragulus Thomas et al., 1990(see Thomas et al. 1990Ginsburg et al. 2001). The aim of this work is to describe these new findings of Dorcatherium songhorensis from Napak XXI and test the phylogenetic relationships of this species within the Tragulidae, reassess the diagnosis and definition of the genus Siamotragulus, and finally offer new information on the phylogeny and evolutionary history of the clade of selenodont derived tragulids in which Siamotragulus is included, with special remarks on the African Miocene.

The Napak XXI fossil sites
The locality of Napak XXI ( Fig. 1) was discovered in 2008 in the southern slopes of Akisim Mountain, a remnant of the Early Miocene Napak carbonatite-nephelinite volcano (Bishop 1958(Bishop , 1962(Bishop , 1964(Bishop , 1967(Bishop , 1968(Bishop , 1972Bishop & Whyte 1962;Bishop & Trendall 1967). The sediments, exposed at the top of a hillock, comprise red pedogenic clays with calcareous nodules developed on volcanic ash of the Napak Member. The associated fauna and flora is typical of the Napak area, with taxa such as the rodent Diamantomys luederitzi, the primate Micropithecus clarki and the ruminant Dorcatherium songhorensis . The locality is exceptional, however, in yielding an abundance of extremely well-preserved gastropods and plant seeds (Celtis rusingensis) which indicate a tropical woodland to forest environment. Among the gastropods the first shells of Koruella magnifica and Edentulina rusingensis from the Napak area were collected in association with specimens of Gulella, Haplonepion naggsi, Maizania, subulinids, achatinids (Tholachatina leakeyi, Burtoa nilotica), Trochozonites, and other lineages characteristic of tropical forest to woodland (Pickford 2009). It also yielded the first beetle fossil from Napak preserved in three dimensions. Vertebrates are rare at the site, but enough are present to establish the age as Early Miocene, older than Rusinga and approximately equivalent to Songhor, Kenya (Faunal Set I of Pickford 1981;Musalizi et al. 2009). Radio-isotopic analyses indicate an age for the deposits of between 19 and 20 Ma (Bishop et al. 1969;Werdelin 2010).

Material and methods
The tragulid fossils from Napak XXI described in this paper are curated by the Uganda Museum in Kampala (Uganda). This material probably belongs to a single individual since the fossils were found scattered over an area of 3 metres by 5 metres, there is no duplication of skeletal parts, and the preservation characteristics of all the bones are similar, with parts of the skeleton still connected to each other. It has been compared with some extant and extinct tragulid taxa (listed in Supplemental Table 1).

Anatomical definitions
We use the terminology of Azanza (2000) for nomenclature of the dentition (English version in S anchez & Morales 2008). The Dorcatherium-fold is the fold that occurs on the linguodistal side of the metaconid. The Tragulus-fold is the fold situated on the distal side of the protoconid, usually linked to the post-protocristid (sometimes to the conid itself). Combined, both the Dorcatherium-fold and the Tragulus-fold form the so-called Mstructure (see e.g. Janis 1987;Geraads et al. 1987). The inter-lobular bridge is the rectilinear bridge of enamel that connects the anterior and posterior lobes in an Afrotragulus-type lower molar (Fig. 2;S anchez et al. 2010). The Zhailimeryx-fold is a fold of enamel that originates from the anterior part of the entoconid, the development of which is highly variable amongst tragulids (from well developed to absent). We define the Dorcatherium-platform as the mesial semicircular structure of the lower molars formed by a hyper-developed pre-protocristid that turns lingually to contact a very small pre-metacristid ( Fig. 2; see also S anchez et al. 2010; Morales et al. 2012). For nomenclature of the postcranial skeleton we follow Barone (1999).

Measurement abbreviations
The measurements taken on postcranial remains are defined in S anchez & . The detailed measurements of dental and postcranial material of the fossil tragulid from Napak XXI are presented in Supplemental Appendix 1.
Emended diagnosis. Tragulids with selenodont lower cheek teeth and postcranial skeleton characterized by the presence of fused, long and narrow metatarsals IIIÀIV that form a true, pecoran-like, cannon bone; presence of a dorsal constriction in the acetabular cavity that confers to it a 'three-lobed' morphology; elliptic and wide cavity located caudal to the acetabular notch in the pelvis; and short, deep and triangular distolateral furrow for the tendon of the fibularis longus muscle in the navicular-cuboid. Differing from Dorcatherium, Hyemoschus and Dorcabune in: presence of selenodont molars; weaker buccal rib on the metacone; presence of a well-developed pre-metacristid smaller than the pre-protocristid and absence of Dorcatherium-platform; fused metacarpals IIIÀIV; presence of a slight convexity over the palmar distal articular area in the humerus; presence of curved dorsal border of the medial epicondyle in the humerus. Differing from Tragulus, Moschiola and Afrotragulus in: lack of well-extended, Tragulus-like cristids; and lesser development of the pre-metacristid. Differing from Afrotragulus in: the lack in the lower molars of the interlobular bridge linking the lobes. Differing from Tragulus and Moschiola in: lingual and buccal distal cristids of the p4 originating from the central conid. Thomas et al., 1990 Emended diagnosis. Siamotragulus with poorly developed mesial cingulid; fused, long and narrow pecoran-like metatarsal IIIÀIV with square cross-section at mid-shaft; and metatarsals II and IV fused with the metatarsal IIIÀIV. Ginsburg et al., 2001 Emended diagnosis. Siamotragulus with relatively low cristids in lower dentition and navicular-cuboid not fused with the ectomesocuneiform.

Siamotragulus bugtiensis
Siamotragulus songhorensis (Whitworth, 1958) (   Morales et al. (2012). E, schematic depiction of the three types of mesial morphology in tragulid molars; from left to right: very large pre-protocristid which turns lingually and contacts a very small pre-metacristid, forming the Dorcatherium-platform; large rectilinear pre-protocristid that meets a smaller pre-metacristid, with no Dorcatherium-platform; straight pre-protocristid and premetacristid which are subequal in length, meeting parasagitally and forming a triangular mesial outline of the teeth; the centre morphology corresponds with Siamotragulus.    Occurrence. Late Early Miocene of Kenya and Uganda (Whitworth 1958;Pickford 2002;Musalizi et al. 2009); Faunal Set I of Pickford (1981).
Description. The lower molars of Siamotragulus songhorensis are very similar to those of Siamotragulus sanyathanai, with flat cuspids and well-developed cristids. The mesial closure of the trigonid is achieved by a long rectilinear pre-protocristid that meets a smaller pre-metacristid (with no Dorcatherium-platform present; see  well developed. Both the Dorcatherium-fold and the Tragulus-fold are robust, and the latter contacts the pre-hypocristid. Siamotragulus songhorensis lacks an inter-lobular bridge (see Fig. 2). The post-entocristid is virtually nonexistent. The post-hypocristid does not reach the lingual border of the tooth. The mesial cingulid is moderately developed, the buccal cingulids are weak, and the ectostylid is moderately well developed. The posterior cingulid is strongly developed, especially at the posthypocristid level. The only m3 recovered has only the distal part preserved. The pre-hypoconulidcristid is long and contacts the post-hypocristid. The post-hypoconulidcristid, albeit shorter, is still well developed. However it does not close the third lobe distally. There is a well-developed distolingual cingulid.
In the postcranial skeleton the scapula (Fig. 4A, B) has a subcircular and not very deep glenoid cavity. The supraglenoid tubercle and the coracoid process form a high and well-developed structure that attains an inverted 'L' shape, forming a strong canal for the supraspinatus muscle. A similar canal is present in the African chevrotain but not in Tragulus.
There is a single, almost complete, humerus (Fig. 5CÀE). The tricipital line is very narrow and well marked. It contacts the distal border of the convexity of the major tubercle, reaching the area for the infraspinatus muscle. Both the deltoid tuberosity and the tuberosity for the teres minor muscle are poorly developed. Overall, the entire bone is as slender as in Tragulus. In Siamotragulus sanyathanai the tuberosity for the teres minor is marked by a superficial rounded concavity, so the condition in the smaller sized S. songhorensis and Tragulus is probably due to allometry. The palmar border of the medial epicondyle shows a slight convexity over the distal articular area, very similar to the condition in Tragulus. However, Dorcatherium and Hyemoschus have a straight border. On the other hand, the dorsal border is clearly curved in S. songhorensis and also in Tragulus, and straight or almost straight in Hyemoschus and Dorcatherium. The coronoid pit is triangular and deep, comparatively wider than that of Tragulus. Also, the capitulum and the trochlea have more similar proximodistal lengths than in Tragulus, resulting in a more rectangular shape of the distal articulation of the humerus.
The radius (Fig. 5A, B) is long and slender, very similar to that of Tragulus and Siamotragulus sanyathanai, contrasting with the short and wider radius of Hyemoschus. As in S. sanyathanai and Hyemoschus the medial proximal facet for the humeral trochlea is dorsopalmarly wide, with divergent proximal and palmar borders, different from the rectangular facet of Tragulus. The attachment area for the biceps brachii muscle is more developed than in Tragulus. The palmar surface is distinctively flattish, as in S. sanyathanai. The distal articular area is not well preserved.
The scaphoid has a dorsally wide proximal facet, with a very clear lateral projection similar to that of Hyemoschus. The dorsomedial border is rectilinear, contrasting with the more bulging border present in Hyemoschus and Tragulus. As in Hyemoschus the centromedial apophysis is weak, differing from the well-developed apophysis present in Tragulus.
Two right coxal fragments preserve the acetabular cavity and its surroundings (Fig. 4CÀE). The acetabular cavity is subcircular with a dorsal constriction. Together with the caudal acetabular notch the constriction breaks the subcircular profile into a three-lobed profile. This is also the case with Siamotragulus sanyathanai, but not with Tragulus and Hyemoschus, in which the acetabular cavity lacks the dorsal constriction and is perfectly circular in shape. The acetabular notch is wide and triangular. The major sciatic notch is triangular and weak, very similar to that of S. sanyathanai and Tragulus. The cavity located caudal to the acetabular notch is elliptic and relatively wider (Fig. 4E) than that of Tragulus and Hyemoschus, which show a narrow and slit-like cavity. As observed in Tragulus, the crest for the psoas minor in the inner surface of the ilium is very weak.
The femur of Siamotragulus songhorensis (Fig. 5F, G) is nearly morphologically identical to that of S. sanyathanai. The crest of the greater trochanter is well marked, and runs almost vertically. The caput femoris is transversely elongated, with a triangular and upwards-oriented fovea capitis, corresponding to K€ ohler's type-A (see K€ ohler 1993). The trochanteric pit is wide and triangular. Hyemoschus shows a similar morphology, but apparently Tragulus has a narrower pit. The distal part of the femur is not preserved.
The only tibia (Fig. 6A, B) lacks the proximal region above the mid-part of the tibial crest. The tibia of Siamotragulus songhorensis is long and slender, very similar to that of S. sanyathanai. The medial malleolus is short but surpasses the border of the distal articulation. In Siamotragulus sanyathanai it is quadrangular and shorter, and does not surpass the border of the articulation. The fibular notch is wide with well-marked ridges. The notch of origin of the long collateral medial ligament is also well marked and delimited by sharp ridges. Contrary to the case of Siamotragulus sanyathanai and S. bugtiensis the malleolar bone is fused to the tibia.
The calcaneum (Fig. 6A, B, H) has pentagonal-shaped tuber calcanei, with two well-developed plantar tubercles for the insertion of the gastrocnemius tendon. The two plantar ridges of the corpus are proximally well marked, forming a distinct canal that runs down to the middle of the shaft. The dorsal and plantar borders of the corpus are not parallel. The sustentaculum tali is dorsally less developed than in Siamotragulus sanyathanai. Also, the plantar border of the sustentaculum is concave whereas it is straight in S. sanyathanai. The dorsal surface of the sustentaculum is also concave, contrasting with the flattish and dorsally inclined surface in S. sanyathanai.
The astragalus (Fig. 6A, B) is narrow and slender, with an elliptical proximoplantar facet for the calcaneum.
The only recovered navicular-cuboid is fully articulated with the metatarsal IIIÀIV and the rest of the tarsus (Fig. 6A, B). The navicular-cuboid is fused with the ectomesocuneiform. As in Siamotragulus sanyathanai the laterodistal furrow for the tendon of the fibularis longus muscle is narrow and triangular and does not contact the lateral articular facet for the calcaneum. Contrary to this condition, the navicular-cuboid of Tragulus and Hyemoschus has a true deep canal that opens in front of the lateral articular facet for the calcaneum, contacting it, and Dorcatherium has a parallel-sided shallow canal that contacts the lateral side of the aforementioned facet. The plantodistal lateral process is short and blunt, similar to that of S. sanyathanai, not surpassing the distal border of the bone. In contrast, Tragulus has a pointed and long process that clearly surpasses that border. As in Tragulus and S. sanyathanai there is a strong (usually Y-shaped) attachment area for the plantodistal ligament on the plantomedial surface of the bone. However, Hyemoschus lacks that structure or it is extremely faint.
As occurs in Siamotragulus sanyathanai and contrary to Dorcatherium, the central metatarsals of S. songhorensis are completely fused (Fig. 6A, B, DÀG), forming a true, pecoran-like, cannon bone. However, as in the extant Tragulus, the cross section at mid-shaft is elliptical, dorsoventrally flattened, instead of the square cross section seen in S. sanyathanai. The plantar surface of the metatarsal IIIÀIV, where the interosseum muscles passes, is narrow and V-shaped whereas it is wider and flattish in S. sanyathanai and Tragulus. Also, in contrast to S. sanyathanai and Tragulus, the lateral metatarsals II and V are not fused to the metatarsal IIIÀIV. The metatarsal sulcus is distally closed. The distal articular keels are plantarly very well developed. The specimen NAPXXI 10a'08-1 has a still-attached proximal fragment of lateral metatarsal showing its unfused condition (Fig. 6E). Also, the distal articular fragments of lateral metatarsals (Fig. 6C) show a well-developed articular area with a strong (albeit diminutive) distal keel relatively better developed than in Tragulus. The metatarsal IIIÀIV from NAP XXI 10c'08 shows a very conspicuous carnivore bite mark on the medial side of the proximal articulation.
The first and second phalanges of Siamotragulus songhorensis (Fig. 5HÀM) are relatively slender albeit not as much as those of Tragulus. Despite being larger, their areas for ligamentary insertion in the diaphysis are far less marked than those of the extant mouse deer, and in this regard they are more similar to those of Hyemoschus. The second phalanx has strong proximoplantar attachments for the collateral ligaments. The third phalanx (Fig. 5N, O) lacks a plantar platform, and the articular facet occupies the entire proximal surface. Whitworth (1958) pointed out that the lower cheek teeth of this species were 'less bunoid' than those of D. chapuissi and D. pigotti. Actually not only are the cusps of Siamotragulus songhorensis flat instead of rounded, especially in their inner wall, but also the teeth are mesially closed by means of an enlarged and straight pre-metacristid that meets a non-curved pre-protocristid as in selenodont tragulids such as Afrotragulus, Tragulus, Moschiola and the other two species of Siamotragulus, lacking the plesiomorphic Dorcatherium-platform typical of such forms as Dorcatherium, Dorcabune and Archaeotragulus. These features are present both in the type series from Songhor (Western Kenya ;Whitworth 1958) and in the material from Napak (Pickford 2002; this paper). Quiralte et al. (2008) assigned to Dorcatherium songhorensis some small tragulid material from the sites of Langental and Grillental (Early Miocene, Sperrgebiet, Namibia). Some features of this form (e.g. lack of Dorcatherium-platform and a proximal metatarsal IIIÀIV fragment that could be of the derived type) are Siamotragulus-like. Thus the material from the Sperrgebiet should be re-examined keeping in mind its possible Siamotragulus nature. Also, a further re-analysis of the D. songhorensis material (both the type from Songhor and the new material from Uganda) allowed us to reject a previous characterization of the species as a bunoselonodont tragulid (see S anchez et al. 2010). This material highlights the difficulties in dealing with the waste-basket taxon that the genus Dorcatherium has become, the correct characterization of which only will be achieved through the redescription of the tremendous variability of forms that have been included into it.

Cladistic analysis
We performed a cladistic analysis at the species-level to explore the phylogenetic relationships of Dorcatherium songhorensis using the TNT software (Goloboff et al. 2008). We chose Zhailimeryx jingweni as the outgroup since this taxon was previously used successfully to root a tragulid cladistic analysis (M etais et al. 2001). The morphological characters found in the tragulid from Napak XXI indicated that this form does not belong to Dorcatherium (see further discussion). As we intended to test the hypothesis that 'Dorcatherium' songhorensis falls outside of the genus Dorcatherium we accordingly selected as the ingroup several taxa useful for performing that test: primitive tragulid forms such as Archaeotragulus krabiensis (M etais et al. 2001) and Dorcabune anthracotherioides Pilgrim, 1910, the bunodont Dorcatherium crassum and the buno-selenodont Dorcatherium naui (thus covering the dental morphological extremes included within Dorcatherium and also including postcranial data), the extant Hyemoschus aquaticus, Moschiola memmina and Tragulus javanicus, the two described species of Siamotragulus (Thomas et al. 1990;Ginsburg et al. 2001) First African record of the Miocene Asian mouse-deer Siamotragulus and the two described species of Afrotragulus (S anchez et al. 2010). For this analysis we combined the morphological data from the type series of Dorcatherium songhorensis (upper and lower molars) and the new material from Napak XXI D. songhorensis (lower molars and postcranial skeleton) into a single operational taxonomic unit (OTU). We did this after running a test-analysis with the two datasets of D. songhorensis (type and Napak XXI) as separate OTUs and checking that both terminals grouped together (see Fig. 7). The data matrix included 53 characters picked from the skull (six), upper dentition (seven), lower dentition (22) and postcranial skeleton (18), thus resulting in the largest morphological dataset used in a tragulid cladistic analysis so far. The data matrix and the list of characters and their descriptions are included as Supplemental Appendices 2 and 3 respectively. We performed a run using a traditional search with 1000 replicates with TBR that recovered one most parsimonious tree (MPT) of 88 steps (CI D 0.795; RI D .798) in which Dorcatherium songhorensis (Siamotragulus songhorensis hereafter) appears as the most basal of a Siamotragulus clade, which is part of a more inclusive clade of derived tragulids that contains the extant Asian forms (Fig. 7). The character6 state distribution for the discussed internal nodes as well as the autoapomorphies for each Siamotragulus species are presented in Supplemental Table 2.

Discussion
We do not intend to reconstruct a complete phylogeny for Tragulidae, so we are going to discuss the region of our MPT that is important for describing the phylogenetic relationships of the Siamotragulus clade amongst tragulids and also the character6 state distribution within that clade. In past works we linked the presence of flat cuspids in the Tragulidae with the acquisition of the derived mesial closure of the lower molars, describing these two states as the combined morphological signature of the selenodont forms (S anchez et al. 2010). However, we did not discuss in detail the case of the extant African chevrotain Hyemoschus due to the lack of a proper phylogeny. The African chevrotain is a particular case in which the presence of a derived mesial closure of the lower cheek teeth is linked with bunoid cuspids. This mixture of features could reflect two phylogenetic scenarios. First, a single origin of the derived mesial closing of the teeth, and therefore this state would exist previous to the development of full selenodonty and flat cuspids, or alternatively, the appearance of the derived mesial closing of the teeth was achieved in parallel at least twice during the evolutionary history of the Tragulidae. Our present MPT clearly supports the former hypothesis. Therefore, if the derived mesial closure of the lower cheek teeth was achieved once (Hyemoschus + Node A), then the acquisition of the full set of features that characterize the selenodont tragulids (Node A) was a two-step evolutionary scenario. More taxa have to be added to the analysis to fully confirm this hypothesis (work in progress by the authors).
The node A (Siamotragulus plus the clade composed by Afrotragulus and the extant Asian species) is characterized by the presence of derived cheek teeth with reduced buccal structures in the upper molars and the presence of flat cuspids in the lower molars, morphological innovations that become extreme in the case of Afrotragulus. Also this clade features the acquisition of a new type of appendicular skeleton characterized by its slenderness and the fusion of the central metapodials. The central metatarsals in particular are fused together forming a true metatarsal IIIÀIV cannon bone, contrary to Hyemoschus and Dorcatherium in which the metatarsals III and IV are fused but still individualized. Of all the taxa belonging to this clade only the appendicular skeleton of Afrotragulus remains unknown, so future discoveries will test the evolutionary scenario proposed here. Both the longer legs and the fusion of metapodials into a single cannon-bone are characteristics usually linked with enhanced cursorial abilities (see e.g. Kardong 2009). Apart from that, whether or not these important morphological novelties are linked to changes in the common ancestor of clade-A with respect to habitat exploitation (e.g. extant Asian species are known to enter and live in drier and more open areas than the humid tropical deep-forest African chevrotain; Wilson & Mittermeier 2011) or maybe modifications in the aquatic escape behaviour typical of some tragulids still has to be explored.
We define Siamotragulus as the clade comprising S. songhorensis, S. bugtiensis and S. sanyathanai, their more recent common ancestor and all of its descendants. The Siamotragulus clade (node B) is entirely diagnosed by postcranial characters. The morphology and orientation of the furrow for the terminal tendon of the fibularis longus muscle, short and with a divergent outwards angle relative to the lateral plane of the navicular-cuboid, is notable. This configuration probably indicates that the origin area of this muscle is relatively more laterally placed than in the other tragulid forms, in which the tendon runs very close to the lateral side of the navicular-cuboid and hence the orientation of the muscle crossing over the tibia is more medial. One of the functions of the fibularis longus muscle is the rotation of the foot (Barone 1999), so if the entire muscle is more laterally placed its foot rotational capabilities would become restricted without affecting its main function as a tarsal extensor. This synapomorphic modification of Siamotragulus could be interpreted as a running improvement achieved by this genus by restricting the movement of the autopodium to a (mainly) foreand-aft (parasagittal) action.
The type species Siamotragulus sanyathanai is the most autapomorphic of the Siamotragulus clade. This species is also the most recent member of the clade (Middle Miocene; Thomas et al. 1990;Chavasseau et al. 2009;Coster et al. 2010). Its main morphological innovations lie in the acquisition of still more pecoran-like, very long metatarsals IIIÀIV with a square cross section at mid-shaft and fused lateral metatarsals. As Thomas et al. (1990) pointed out, this type of highly derived metatarsal IIIÀIV is not found in any other Neogene or recent tragulid. It is worth noting that the branch support for the node C (S. bugtiensis + S. sanyathanai) is low. However, we decided against collapsing it until more taxa are added in a future work, and so we can show the synapomorphies that link Siamotragulus sanyathanai and S. bugtiensis in this hypothesis of relationship: presence of derived cheek teeth with extended post-hypocristid and weak upper buccal ribs and sustentaculum tali in the calcaneum with rectilinear plantar border.
The genus Siamotragulus was previously recorded only in Asia; however, S. songhorensis extends the palaeobiogeographical distribution of the genus to Africa. The existence of an African Siamotragulus rejects the previous hypothesis about the Asian endemism of this genus (see e.g. R€ ossner 2007; S anchez et al. 2010) and also shows that both lineages of clade-A tragulids (represented by Siamotragulus and Afrotragulus respectively) were already present in the African Early Miocene (Fig. 8). Moreover, S. songhorensis is the most ancient of the three Siamotragulus species, recorded at the 19À20 Ma mark (a bit older than S. bugtiensis; see Ginsburg et al. 2000), thus adding complexity to the already complex Miocene biogeographical distribution of the Tragulidae. As far as we know, clade-A tragulids were not present in Europe; however, we must confirm this in future studies. As we pointed out in the case of Afrotragulus (S anchez et al. 2010) the fact that First African record of the Miocene Asian mouse-deer Siamotragulus some of the earliest records of the Miocene Tragulidae comprise such a diversity of members of such a derived clade strongly supports the idea of an unknown large-scale radiation event prior to the Early Miocene. This radiation of tragulids led to the situation we know from the late Early Miocene deposits, with a high diversity of tragulids that was poised to spread throughout the Old World.

Conclusions
Significant remains of the African tragulid 'Dorcatherium' songhorensis Whitworth, 1958 were discovered in the Early Miocene fossil site of Napak XXI (Uganda), including the first-known postcranial skeleton of the species. The derived limb bones (mainly the long cannon-bone metatarsals IIIÀIV) belong to a type that had never been described before in any African tragulid. Our phylogenetic analysis links 'Dorcatherium' songhorensis with the Asian long-legged genus Siamotragulus, hence we rename this form as Siamotragulus songhorensis (Whitworth 1958). The postcranial fossils from Napak XXI allow us to re-diagnose the genus Siamotragulus adding some interesting characters to the original diagnosis. Siamotragulus is defined as the clade comprising S. songhorensis, S. sanyathanai and S. bugtiensis, their more recent common ancestor and all of its descendants. Siamotragulus songhorensis branches off as the most basal of the Siamotragulus clade, and helps to root and diagnose a large clade of tragulids (in which the extant Asian tragulids Tragulus and Moschiola are included) characterized by their advanced long slender limbs and derived selenodont dentition. Additionally to the long limbs, the genus Siamotragulus probably developed cursorial refinements such as the loss of rotational capabilities in the hind leg autopodium. The existence of an African Siamotragulus allows us to reject the hypothesis that regarded the genus as an Asian endemic taxon. The clade of derived tragulids that includes Siamotragulus, Afrotragulus, Moschiola and Tragulus contains some of the oldest known forms that are recorded from the Early Miocene African deposits (19À20 Ma).
the Spanish Government MINECO. We thank the Willi Hennig Society for access to TNT software.

Supplemental material
Supplemental material for this article can be accessed here: http://dx.doi.org/10.1080/14772019.2014.930526