Taxonomic Revision of Therocephalians (Therapsida: Theriodontia) from the Lower Triassic of Antarctica

ABSTRACT We reevaluate the taxonomic status of therocephalian fossils recovered from the lower Fremouw Formation (Lower Triassic) of the central Transantarctic Mountains, Antarctica. The material, which includes mostly fragmentary juvenile specimens, is reidentified using an apomorphy-based approach. We recognize the presence of three higher-level taxa: Eutherocephalia, Akidnognathidae, and Baurioidea. The only genus-level identification is for a partial lower jaw and pterygoid tentatively attributed to the baurioid, Ericiolacerta parva. An indeterminate theriodont partial skull is reassigned to the therocephalian family Akidnognathidae. The holotypes of Pedaeosaurus parvus and Rhigosaurus glacialis are represented by indeterminate juvenile baurioids and, in the absence of clear autapomorphies, are considered nomina dubia. The results of the taxonomic revision indicate that the therocephalian fauna of Antarctica lacks endemic genera and thus corresponds to that of the Triassic Lystrosaurus Assemblage Zone fauna of South Africa's Karoo Basin. More generally, we consider the southern Gondwanan basins of South Africa and Antarctica to sample a broadly distributed Lower Triassic tetrapod fauna, although the latter basin documents the first occurrence of several taxa (e.g., Kombuisia, Palacrodon). More precise (i.e., species-level) identifications are needed to better constrain the biogeographic signal for therocephalians, but the presence of juveniles strongly suggests that this group of therapsids, like dicynodonts, were year-round high-latitude inhabitants during Early Triassic times.


INTRODUCTION GEOLOGICAL SETTING
Over 600 meters in thickness, the Fremouw Formation is a laterally extensive sequence of siliciclastic and volcaniclastic sediments that rests conformably on top of the Buckley Forma¬ tion, and ranges laterally from approximately 83° to 86° S latitude near the Shackleton and Beardmore glacier regions of the central Trans antarctic Mountains (Barrett et al., 1986).  (Hopson and Barghusen, 1986;van den Heever, 1994;Botha et al., 2007;Abdala, 2007;Huttenlocker, 2009;Huttenlocker et al., 2011). Numbers in parentheses indicate apomorphies observable in the studied specimens and listed in table 1.
Although most of the Fremouw Formation is considered Triassic in age and most of the Buckley is Permian, their contact is diachronous across the basin (McManus et al., 2002;Collinson et al., 2006). As a result, at some localities rocks below the first prominent sandstone (i.e., below the defined base of the Fremouw) can be Triassic.
Fremouw sediments were deposited in an alluvial setting dominated by coarse channel sands with subangular quartzose clasts indicative of a low-sinuosity (i.e., braided) system (Bar¬ rett et al., 1986;Isbell and MacDonald, 1991). The formation is informally subdivided into three members: lower, middle, and upper (Collinson and Elliot, 1984;Barrett et al., 1986). The lower member, ranging from 75 to 125 m in thickness, is well-exposed in the Shackleton Gla¬ cier region where it is characterized by the cyclic deposition of green mudstones, siltstones, and large channel sands (Collinson and Elliot, 1984). Vertebrate skeletons are often localized within ferruginous, olive-green siltstones exhibiting root traces and other evidences of soil formation (i.e., Dolores pedotype of Retallack et al., 1998Retallack et al., , 2007Retallack and Krull, 1999).

SYSTEMATIC FRAMEWORK
Colbert and Kitching's (1981) recognition of two endemic therocephalian genera from the lower Fremouw Formation of Antarctica was a significant contribution to the knowledge of therocephalian biogeography and diversity at high paleolatitudes. Nevertheless, their account was hindered by a combination of the imperfect state of preservation of the referred material and by a limited understanding of therocephalian systematics and character polarity in the group. In particular, the therocephalian family "Scaloposauridae" is diagnosed largely by juve¬ nile characters and has been argued to represent a wastebasket taxon likely containing juvenile therocephalians belonging to other known families (Kemp, 1982(Kemp, , 1986Hopson and Barghusen, 1986;Huttenlocker, 2009). As such, we will use the term in quotes throughout this paper To reassess Colbert and Kitching's (1981) findings, we follow an apomorphy-based approach that is built on recent analyses of therocephalian morphology and systematics. We identified characteristics of systematic value on all the relevant Antarctic material and assessed their level of phylogenetic generality (table 1; see Nesbitt and Stocker, 2008). The cladogram in figure 2 represents a consensus topology derived from the work of Hopson and Barghusen (1986), van den Heever (1994), Botha et al. (2007), Abdala (2007), Huttenlocker (2009), andHuttenlocker et al. (2011), although Therocephalia is considered monophyletic, as traditionally held (contra Botha et al., 2007, andAbdala, 2007).

SYSTEMATIC PALEONTOLOGY
Therapsida Broom, 1905Eutheriodontia Hopson and Barghusen, 1986Therocephalia Broom, 1903Eutherocephalia Hopson and Barghusen, 1986 Definition: All scylacosaurian therocephalians (sensu van den Heever, 1994) sharing a more recent common ancestry with Theriognathus microps and Akidnognathus parvus than with Scylacosaurus sclateri. Diagnosis: Postfrontal absent (but polymorphic in Hofmeyria); dorsal surface of the paroccipital process deeply hollowed in the floor of the posttemporal fenestra; dentary broadly overlaps surangular; lateral mandibular fenestra present between dentary and anterodorsal portion of angular; anterior dentition bears smooth cutting edges that lack serrations; incisors, precanines, and canines commonly possess longitudinal facets or grooves; well-developed obturator foramen bordered by pubis and ischium (rather than a small foramen confined to the pubis); femur with long, slender diaphysis that is round in cross section (rather than oval). Remarks: AMNH FARB 9550 was originally assigned to Ericiolacerta parva by Colbert and Kitching (1981) due to the size of the specimen and its collection near AMNH FARB 9542 (see below). The material consists of a posterior skeleton with ribs, vertebrae, and an articulated pelvic girdle and hind limbs ( fig. 3). There are apparently five thoracic ribs preserved in slight disarticula¬ tion, followed by as many as five (revised from three) lumbar vertebrae bearing laterally the impressions of their associated, fused lumbar ribs, then as many as three or four broken sacral vertebrae. There are no centra associated with the last two lumbar vertebrae, indicating that they were not fused to the neural arches (a likely indicator of the specimens immaturity). The number of caudal vertebrae cannot be determined. The sacral count was limited to three by Colbert and Kitching (1981) based on the apparent presence of three sacral vertebrae in Ericiolacerta, and in other therapsids in general. However, the number of sacral vertebrae varies from three to four in therocephalians Rubidge, 2007,2009). AMNH FARB 9550 exhibits a few apomorphies that are consistent with an assignment to Eutherocephalia. The most notable apomorphies are the broad, flat puboischiatic plate ( Few characters are available in the postcrania of AMNH FARB 9550 to merit a more spe¬ cific taxonomic assignment than Eutherocephalia. For example, Colbert and Kitching (1981: figs. 6, 7) correctly noted the absence of a tuber calcis on the posterior side of the calcaneum.
The tuber calcis was originally regarded by the authors as an autapomorphy of Ericiolacerta (based on the holotype; Watson, 1931), which incidentally confounded the assignment of AMNH FARB 9550, as it lacks this feature. However, the tuber calcis is present in other baurioids as well, including regisaurids (Kemp, 1978(Kemp, , 1986 and bauriids (Schaeffer, 1941;King, 1996), but is apparently absent in nonbaurioids where this region is preserved (e.g., Mirotenthes, lycosuchids, scylacosaurids; Fourie and Rubidge, 2009). The combination of a broader phylogenetic distribution than previously supposed (i.e., across Baurioidea), coupled with its absence in AMNH FARB 9550, weakens the utility of this character in assigning this specimen to Ericiolacerta or to any other baurioid genus.
Akidnognathidae Nopsca, 1928 Definition: All eutherocephalians sharing a more recent common ancestry with Akidnognathus parvus than with either Bauria cynops or Theriognathus microps. Diagnosis: Enlarged, anteriorly oriented external nares; pronounced facial exposure of the septomaxilla, broadly overlapping the premaxilla; short but deep maxilla; median frontonasal ridge (as in whaitsiids); anterior portion of vomer expanded, underlapping the ventral surface of the premaxilla; alveolar margin of maxillary slightly convex laterally in palatal view, rather than straight or concave; fossa for lower canine partially roofed by premaxilla and maxilla; spatulate incisors with mesiolingual and distolingual crests and concave lingual surface; ptery¬ goid teeth absent. , partial skull and lower jaw. Numbers refer to apomorphies listed in table 1. Abbreviations: il, exposed root of first lower incisor; i4, fourth lower incisor. Scale bar = 5 mm.
Remarks: The tooth formula of AMNH FARB 9527 as preserved was correctly described by Colbert and Kitching (1981), although they suggested the possibility of additional marginal teeth not preserved in the specimen. We recognize at most six upper postcanines present in the specimen. An additional unprepared jaw, which was not discussed by Colbert and Kitching (1981), remains in the matrix and also preserves dentition. A shallow lateral dentary groove is present in the left lower jaw ( fig. 4; table 1: 1.2), consistent with its identification as a therocephalian. Other features visible in the specimen are consistent with an assignment to the Eutherocephalia, in particular space for four lower incisors (as opposed to three in more basal therocephalians) and teeth lacking serrations.
Among eutherocephalians, the specimen may be most precisely referred to Akidnognathi¬ dae. Unlike hofmeyriids, whaitsiids, and baurioids, the specimen exhibits a robust dentary with a relatively enlarged lower canine for its small size and a distinct mental protuberance (table 1: 4.3) as in akidnognathids (but also present in Middle Permian lycosuchids and scylacosaurids). The maxilla further resembles that of akidnognathids with its deep facial exposure (table 1: 4.1) and slightly convex alveolar margin when viewed ventrolaterally (table 1: 4.2). At least one upper precanine maxillary tooth is preserved within the precanine diastema, anterior to the enlarged canine. The postcanines are few in number when compared to "scaloposaurids"  Baurioidea Broom, 1911 Definition: All eutherocephalians sharing a more recent common ancestry with Bauria cynops than with either Akidnognathus parvus or Theriognathus microps.
Diagnosis: Long, low maxilla with height less than 40% its length; rostrum appears low relative to raised orbits and antorbital buttress; anterior border of orbit located on or behind anteroposterior midpoint of skull (as in scylacosaurids and some akidnognathids); squamosal and paroccipital process of opisthotic form a distinct, posteriorly projecting "mastoid process" NO. 3738 (as in some akidnognathids); palatal process of maxilla bears well-developed crista choanalis, contacting or nearly contacting the vomer medially, and extending posteriorly onto the pala¬ tine; interpterygoid vacuity of adults large and somewhat heart shaped; dentary is long, slender, and relatively straight with smooth ventral edge; upper incisors numerous, greater than five (except in Bauria). Remarks: Colbert and Kitching (1981) erected Pedaeosaurus parvus on the basis of a unique combination of plesiomorphic and derived characters in AMNH FARB 9548, which they designated as the holotype. Specifically, the plesiomorphic retention of a pineal foramen was considered diagnostic for the new genus, as a pineal foramen is apparently absent in Ericiolacerta and was interpreted as absent in Scaloposaurus by the authors. However, study of the holotype of Scaloposaurus constrictus (NHMUK R1723) and additional specimens reveals that Scaloposaurus also exhibits a pineal foramen, as do more basal baurioids (e.g., Ictidosuchoides, Ictidosuchops). Assignment of AMNH FARB 9548 to "Scaloposauridae" was based on its rela¬ tively small size (skull length approximately 28 mm), slender dentary with a low coronoid process, and a wide intertemporal region (Colbert and Kitching, 1981). However, these features are broadly distributed in most juvenile eutheriodonts regardless of their taxonomic affinities (Hopson and Barghusen, 1986;Kemp, 1986).
AMNH FARB 9548 ( fig. 5A) is dorsoventrally flattened such that the lower jaw is displaced onto the palate. In ventral view, a broad palatal process of the right maxilla is preserved over¬ lapping the anteriorly displaced right palatine (both were interpreted as "palatal plates of the maxillae or ... premaxillae" by Colbert and Kitching, 1981: 15). A sharp crista choanalis pre¬ served on the palatine would have continued anteriorly onto the ventrally broad maxilla, resembling the condition in some baurioids (table 1: 5.2). The specific arrangement of these elements around the vomer, however, cannot be determined. The dorsal aspect demonstrates the absence of a postfrontal, as in baurioids and all other eutherocephalians (table 1: 3.1; Col¬ bert and Kitching, 1981: fig. 10). The long, straight dentary, having a loosely articulated sym¬ physis, is also typical of baurioids ( fig. 5A; table 1: 5.4). Although the morphology of the crowns is not preserved, the alveoli and partially preserved roots of at least nine maxillary teeth (the canines modest and nondistinct) and at least six postcanine dentary teeth are present in the specimen (fig. 5A). The high tooth count further supports an assignment to Baurioidea (table 1: 5.5). Although a single lower canine is observable, the anterior lower incisors could not be distinguished in the specimen. Moreover, no autapomorphic features could be distin¬ guished in the associated postcrania.  Colbert and Kitching (1981) designated as the holotype of Rhigosaurus glacialis, preserves a partial skull roof, missing basicranial and palatal elements, with a twisted rostrum and anterior portion of the left dentary. The snout is torqued and the right side of the rostrum missing, so that the left maxilla and associated dentary are visible in medial view. Although Colbert and Kitching (1981) attempted a detailed description of the skull, many of the diagnostic areas were either damaged or missing. For example, the authors wrote, in error, that the "flat parietals, lacking a pineal opening" revealed affinities to the Scaloposauridae (Colbert and Kitching, 1981: 16). Notwithstanding the contradictory pres¬ ence of a pineal foramen in Scaloposaurus noted above, this region in the specimen displays a cranial fontanelle between the parietals (probably owing to its juvenile status) and is damaged in the area where the pineal foramen would normally be located. The anterior portion of the dentary is long and straight (table 1: 5.4) with a large canine alveolus and as many as seven poorly preserved lower postcanine alveoli. Colbert and Kitching (1981) also reported the pres¬ ence of several lower incisors, yet none is preserved in the specimen or figured in their publica¬ tion. In general, the teeth and the deformed anterior snout are poorly preserved, showing only a few alveoli and making impossible the determination of a complete tooth count. An enlarged upper canine is present, with as many as four teeth preserved anterior to it (although it is impossible to distinguish whether they reside in the maxilla or premaxilla). There are at least seven upper postcanines (table 1: 5.5), as noted by Colbert and Kitching (1981). In summary, the lack of autapomorphies suggests that AMNH FARB 9525 is best regarded as an indetermi¬ nate baurioid therocephalian.
Ericiolacerta parva Watson, 1931 Holotype: CAMZM T 369, skull, lower jaw and nearly complete skeleton; Harrismith, These paired processes border an enlarged, heart-shaped interpterygoid vacuity typical of most baurioids and juvenile eutherocephalians in general. The stout transverse flange, preserved only on the right side, borders a very large suborbital vacuity approximating the condition in the types of Ericiolacerta and Scaloposaurus (Mendrez, 1975;Mendrez-Carroll, 1979). Colbert and Kitching (1981) noted the lateral extent of the transverse flange is unexpectedly short in AMNH FARB 9542, as it does not reach beyond that of the quadrate ramus. However, the lateral extent of the quadrate ramus is not easily distinguishable in the type of Ericiolacerta. In Scaloposaurus, the transverse flange only extends laterally to the lateral limits of the quadrate ramus.
The dentary ( fig. 7) is a slender element with a smooth, but slightly rounded ventral margin (lacking a sharp angular process). Along with the dentition, this element provides the most compelling evidence, albeit modest, for an assignment to Ericiolacerta parva (Colbert and Kitching, 1981). Although the crown of the first incisor is missing, its root is preserved in its alveolus ( fig. 7) and demonstrates that it was large and procumbent, as in the type (Watson, 1931: figs. 4-6). There is likewise no indication of an enlarged lower canine. Rather, the anterior incisor is immediately followed by a smaller alveolus for another small, somewhat procumbent tooth, then a series of at least six irregularly spaced teeth. The dental patterns, including irregu¬ lar spacing of the individual teeth, as well as an anterior accessory cusp and strong wear facet on the distal edge of the last dentary tooth (pc 6?), are consistent with what was described in Ericiolacerta by Crompton (1962), who reconstructed the tooth-wear and replacement patterns. Colbert and Kitching (1981) interpreted space for additional teeth between these, accommo¬ dating nine small dentary teeth behind the anterior incisiforms, with the posteriormost denti¬ tion bearing wear facets and accessory cusps, as in the type (Watson, 1931;Crompton, 1962). NO Usage of the names Pedaeosaurus and Rhigosaurus should therefore be discontinued and the genera regarded as nomina dubia. Colbert and Kitching (1981: 12) diagnosed Pedaeosaurus parvus as a small "scaloposaurid" in which the "characters of the skull and dentition are in general those seen in £n'do/acerto," yet differs from Ericiolacerta on the basis of a "well-defined pineal foramen." It is unclear why the presence of a pineal foramen influenced their establishment of the new taxon, as this feature (and thereby the diagnosis) is consistent with a possible affinity to Scaloposaurus. Nevertheless, such synonymy cannot be confirmed given the poor state of preservation anteriorly in the palate, a region that is most useful in diagnosing Ericiolacerta and Scaloposaurus (based on whether the maxillae share a suture at the midline in the former, or suture to the vomer in the latter).
Likewise, diagnosis of Rhigosaurus glacialis by Colbert and Kitching (1981: 16) was: "large orbits and a broad temporal region; frontals between the orbits are broad, parietals broad and flat; no pineal foramen; temporal fenestra broad and postorbital bar probably incomplete; jugal- Lacking definitive endemic therocephalians, the lower Fremouw Formation appears to host a fauna not substantially different from its Karoo Basin counterpart (Sidor et al., 2008a).
Indeed, we suggest that therocephalians had a wide biogeographic distribution across highlatitude basins during Early Triassic times. However, the collection of more complete material is required to better resolve the taxonomic identities of the akidnognathid and baurioid thero¬ cephalians recognized from the lower Fremouw Formation. Given the small sample size avail¬ able, future collecting has the potential to reveal rare endemic elements among the lower Fremouw fauna, if present. Additional therocephalian material (preferably of adult specimens) is needed to better resolve the taxonomic identities of the Antarctic therocephalians, and fur¬ ther constrain their biogeographic distributions and biostratigraphic correlations with the LAZ of South Africa. Collinson and Hammer (2007) suggested that climatic amelioration allowed some earliest Triassic tetrapods to disperse into higher paleolatitudes (~41°-65° S latitude) along a continu¬ ous foreland basin that stretched between the African and Antarctic components of southern Pangaea. Our recognition of an akidnognathid suggests that the lower Fremouw fauna of the Shackleton Glacier region may represent the earliest portion of the Triassic, as akidnognathids are not represented in the postextinction recovery fauna recorded in the Katberg Formation of the Karoo Basin (Smith and Botha, 2005;Botha and Smith, 2006). Alternatively, it is possible that the lower Fremouw fauna represents a later part of the LAZ and that akidnognathids persisted in Antarctica longer than at lower latitudes. Unfortunately, the fossil assemblage at Sentinel Hill, where AMNH FARB 9257 was collected, includes records with only limited biostratigraphic utility (.Lystrosaurus curvatus, Prolacerta broomi, and one indeterminate temnospondyl fossil).