Abstract: Remains of a peculiar traversodontid cynodont, Protuberum cabralensis gen. et sp. nov., are described herein. The material was collected from two outcrops representing the Therapsid Cenozone (Middle Triassic) of the Santa Maria Formation, and consists of a cranium with most of its dentition preserved and an associated postcranial skeleton. The upper postcanines have two sharp cusps that are connected by a medial crest on unworn postcanines. The specimens possess several autapomorphies, including: (1) presence of thickened bone on the dorsal surface of the skull; (2) thick dorsal ribs, with remarkable processes situated on their dorsal borders that decrease in size distally; and (3) an iliac blade with a series of rugosities along its dorsal border. The lumbar ribs bear overlapping costal plates and have distally projecting rib shafts that differ from the pattern observed in Thrinaxodon, Pascualgnathus and Cynognathus.
The traversodontids were diverse herbivores that can be characterized by their specialized dental morphology. The postcanines are transversely enlarged in occlusal view, providing the upper ones with a rectangular outline and the lower ones with a square outline. This structure allowed precise dental occlusion that resembles the pattern observed in most mammals in some respects. The external crest functioned to shear food items, while the internal basin facilitated crushing. This dental configuration of traversodontids represents an important evolutionary step, representing a great improvement in terms of food processing (Hopson 1984). Although traversodontids possess many derived (and mammal-like) features, details of their teeth indicate that this family is not close to mammalian origins (Hopson and Kitching 2001; Rowe 1988; Luo 1994; Luo and Wible 2005).
There are no specific postcranial characteristics that diagnose traversodontids that are not known in other non-mammalian cynodonts, but this may reflect the poor fossil record of their rather conservative postcranial skeletons. Rib specializations are known in some cynodonts (Brink 1955; Jenkins 1971) and some mammaliaforms (Ji et al. 2006), but these specialized features are not diagnostic of traversodontids and are also present in widely variable forms in galesaurids, cynognathids and diademodontids (Jenkins 1970). The presence of a series of processes along the ribs represents a remarkable feature of the new cynodont taxon described herein: much needed anatomical data on other parts of the cynodont postcranium are also provided by this new taxon.
The new taxon is based on specimens collected by Father Daniel Cargnin at two outcrops in the state of Rio Grande do Sul, southern Brazil (Text-fig. 1). Some peculiar ribs and vertebrae were found in the Municipality of Paraíso do Sul in 1977, in a taphocoenosis comprising specimens of several taxa, including cynodonts (Massetognathus ochagaviae and Probelesodon kitchingii) and dicynodonts (Dinodontosaurus sp.). In 1989, Father Cargnin collected part of an articulated skeleton with the associated skull from an outcrop located in the Municipality of Novo Cabrais. This specimen is housed at the Guido Borgomanero Museum in Mata (Rio Grande do Sul).
The specimen from Novo Cabrais includes articulated vertebrae and ribs. A skull and some additional thoracic ribs were associated with the specimen. The vertebrae are poorly preserved, but their general features can be described. The cranium is also poorly preserved, with the exception of the dentition. The specimen from Paraíso do Sul is better preserved but, consists of only a few isolated elements.
The holotype of Protuberum cabralensis (MGB 368/100) was collected from the ‘Sítio Cortado’ outcrop (S 29°44′54.4′′ W 53°01′49.4′′: Text-fig. 1), located in the municipal district of Novo Cabrais. The paratypes were collected from the ‘Rincão do Pinhal’ outcrop (S 29°43′14.8′′ W 53°13′46.8′′; Text-fig. 1) in the municipal district of Paraíso do Sul. Both outcrops are composed of thin layers (each only a few centimetres thick) of red massive mudstones that alternate with amalgamated lenticular bodies of fine sandstones. Two layers of carbonate concretions were described by Da Rosa et al. (2004) in the former outcrop. These lithologies are typical of the Alemoa Member of the Santa Maria Formation (Schultz et al. 2000).
The tetrapod fauna of the Santa Maria Formation has been divided into four biostratigraphic units (Text-fig. 2). The lowermost, which includes the two outcrops yeilding Protuberum material, represents the Ladinian aged Therapsid Cenozone (Rubert and Schultz 2004), and is dominated by the herbivorous dicynodont Dinodontosaurus, with some occurrences of the cynodonts Massetognathus and Probelesodon.
Institutional abbreviations. MCP, Museu de Ciências e Tecnologia da Pontifícia Universidade Católica, Porto Alegre; MGB, Museu Guido Borgomanero, Mata; UFRGS, Universidade Federal do Rio Grande do Sul, Porto Alegre.
Etymology. The generic name refers to the numerous protuberances present on the ribs and ilia.
Diagnosis. As for the type and only species.
Protuberum cabralensis sp. nov.
Etymology. In reference to the Municipality of Novo Cabrais where the holotype was collected.
Holotype. MGB 368/100, a skull with most of its dentition but lacking the lower jaw and an articulated series of 19 vertebrae (two cervicals?, nine thoracics, five lumbar and three sacral) including four articulated thoracics, 10 lumbar ribs (five from each side) and four sacral ribs (two from each side), a fragment of the left ilium and seven isolated thoracic ribs.
Paratypes. The paratypes are represented by (1) UFRGS PV 0981T, a proximal fragment of a right cervical rib; (2) UFRGS PV 0983T, an isolated vertebra; (3) UFRGS PV 0985T, an isolated vertebra; (4) UFRGS PV 0986T, an isolated vertebra; (5) UFRGS PV 1009T, a left cervical rib; (6) UFRGS PV 1010T, a left thoracic rib; and (7) UFRGS PV 1011T, a fragment of a thoracic rib.
Diagnosis. Protuberum is a large traversodontid in which the upper postcanines have two main cusps (one labial and one lingual) that are connected by a sharp transverse crest. As in Luangwa, Pascualgnathus, Scalenodon, Andescynodon and Traversodon, the postcanines of Protuberum lack the shouldering pattern (in which the mesial of one tooth ‘shoulders’ into the proper area of the preceeding tooth) observed in Massetognathus and Exaeretodon. The paracanine fossae are anteroposteriorly elongated and posteriorly placed in relation to the upper canine. This feature is also present in Exaeretodon and Scalenodontoides. The parietal crest is short, as in Scalenodontoides macrodontes (Gow and Hancox 1993). Protuberum has well-developed masseteric processes of the jugals, as in Exaeretodon. The paraoccipital process is bifurcated in Protuberum, as in tritylodontids, brasilodontids (Bonaparte et al. 2005) and some early mammals (Kermack et al. 1981; Crompton and Luo 1993), in contrast to the unbifurcated condition present in other traversodontids. Autapomorphies observed in the skull include: (1) incisive foramina totally enclosed by the maxillae; and (2) a bony thickening that forms wide crests on the dorsal surface of the skull. The postcranium of Protuberum is robust and the contacts between lumbar vertebrae, as well as between the lumbar vertebrae and ribs, are very strong. The lumbar ribs of Protuberum have costal plates that differ from those in Thrinaxodon, Pascualgnathus and Cynognathus, as rib shafts are situated distal to all of the costal plates in the new taxon. Other postcranial autapomorphies are: (1) all ribs show very pronounced processes on their dorsal border, the most proximal of these is generally the largest and the others become smaller distally; and (2) the iliac blade has a series of rugosities along its dorsal border.
The type specimen, MGB 368/100, includes an almost complete cranium (Text-figs 3–10), with the right quadrate and quadratojugal preserved in their original positions and with most of the upper dentition present; the right postorbital bar and the lower jaws were not preserved. The cranium is somewhat distorted, especially in its dorsal portion. The total cranial length (from the anterior border of the premaxilla to the condyles) is approximately 200 mm, while the parietal crest measures about 15 mm, representing only 7.5 per cent of the length of the cranium.
The cranium is heavily built, with thick bones in the preorbital region and in the lambdoid crests. The short parietal crest is high and the temporal openings are wide. The occiput is exposed in dorsal view with posteriorly projecting occipital condyles. The internarial process of the premaxilla is not present, so that the external naris is a single and confluent opening.
The canines are reduced in size, when compared to the proportions observed in Exaeretodon. The incisors (which are of similar size to the canines) and postcanines are slightly procumbent. In palatal view, the postcanine rows diverge posteriorly.
Facial region. The rostrum is short and wide. The anterodorsal surfaces of the premaxillae are very similar to those of Exaeretodon (Bonaparte 1962) as an ossified internarial process is absent in both taxa. In Protuberum, however, there is no relict of a conic appendix, in contrast to the condition in Exaeretodon (Bonaparte 1962). The anterior margin of the premaxillae is broad and the opening of the external naris is wide (Text-fig. 5). The maxillary process of the premaxilla (Text-fig. 6A) is reduced (it does not reach the nasals), as also occurs in Massetognathus (Romer 1967) and Exaeretodon.
The margins of the maxillae are unclear, especially on the dorsal part of the cranium. In lateral view (Text-fig. 6), a depression can be observed adjacent to the dorsal border of the maxilla. A wide foramen is present near the anterior margin of the maxilla, above the canine. The septomaxillary foramen, which should be immediately dorsal to this foramen, is not preserved. Another maxillary foramen opens lateral to the third postcanine. The facial process of the septomaxilla (Text-fig. 6A) is well developed in Protuberum. Its subtriangular shape is similar to that of Exaeretodon (Bonaparte 1962). The premaxillae participate in the formation of the posterior and ventral borders of the narial opening.
The margins of the nasals are not clear. Their middle portion is constricted at the level of the paracanine fossae, forming a bony thickening that is developed as an anteriorly directed ‘V’ shaped crest (Text-fig. 5). This crest continues posterolaterally along the prefrontals and postorbitals, and connects with the dorsal margin of the orbit. The crest is symmetrical and well-defined and is unlikely to be the result of post-mortem deformation.
Palate. Each premaxilla contains three incisors. The canine is placed near the anterior border of the maxilla and only a small space is present between the canine and the posteriormost incisor, as also occurs in Exaeretodon (Bonaparte 1962), Massetognathus (Romer 1967) and Probainognathus (Romer 1970). The premaxillae participate in the borders of the incisive foramina (Text-fig. 7), which are placed near the anteromedial border of the maxillae and are totally enclosed by them. This feature has not been observed in other traversodontids, but is present in Diademodon (Brink 1955). The paracanine fossae (Text-fig. 7) are positioned posterior to the canines (a feature observed in only two other traversodontids: Exaeretodon and Scalenodontoides; see Abdala and Ribeiro 2003) and are deep and anteroposteriorly elongate.
A marked ridge separates the dorsal depression of the maxillae from their ventral surface. This ridge corresponds to the upper limit of the area described as a maxillary bulge by von Huene (1935–1942). This maxillary bulge is quite well developed in Protuberum, as in other traversodontids, with an accentuated lateral ridge on the maxillae.
The posterior border of each maxilla reaches the subtemporal fossa, by means of a narrow process that intervenes between the anterior border of the jugal and the pterygoid, as in Exaeretodon. The process of the jugal that extends between the posterior border of the maxilla and the pterygoid in Exaeretodon (Bonaparte 1962) could not be observed in Protuberum. The jugal does, however, have a small medial process that is inclined towards the pterygoid (Text-fig. 7), but which does not contact it.
The ventral contact between the maxillae and the palatines begins at the posterior border of the third postcanine along a serrated suture and extends backwards, medial to the postcanine tooth row, until reaching the pterygoids posteriorly. The palatines extend for a considerably distance posteriorly, reaching the level of the masseteric processes of the jugals (Text-fig. 7). No palatal foramina could be observed.
A crest along the medial contact between the maxillae starts at a point level with the anterior border of the paracanine fossae and continues along the medial contact of the palatines to the posterior terminus of the secondary palate (Text-fig. 7). This crest is continuous with an ossified septum within the nasal cavity. This septum is probably formed from the fused vomers and is also observed in Traversodon (MR, pers. obs.) and Exaeretodon (Bonaparte 1962). The opening of the internal choanae is tall, as in Exaeretodon (Bonaparte 1962), and wide, as in Massetognathus (Romer 1967).
The pterygoids of Protuberum form long descending processes (Text-figs 5–7), resembling those in Exaeretodon (Bonaparte 1962), but they are more vertically directed and are proportionally wider than in the latter taxon. Posterolaterally, the pterygoids contact the epipterygoids.
Ventrally, the cultriform process of the parasphenoid (Text-fig. 7) extends between the posterior ends of the pterygoids. At this level, the pterygoids form a pronounced crest continuous with the parasphenoid and the basipterygoid process of the basisphenoid (Text-fig. 7). These features are similar to those in Exaeretodon, but are not present in Massetognathus (MR, pers. obs. UFRGS PV 0968T). Protuberum lacks an interpterygoid vacuity (a feature that is variably present in Massetognathus: Romer 1967; Reichel and Schultz 2004). Anterolaterally, the pterygoids contact the maxillae, participating in the anterior border of the subtemporal fossa. In contrast to the condition in Massetognathus (Romer 1967) and Traversodon (Barberena 1981a), the pterygoids do not reach the jugals laterally.
Orbital region and zygomatic arch. The lacrimal forms most of the anterior border of the orbit and the anterior wall within the orbit (Text-fig. 6). The limits of this bone inside the orbits are not visible, but a clear serrated suture is present at the contact with the maxilla. This suture has a semicircular outline. The lacrimal foramen is not preserved. A foramen located in the internal wall of the orbit is probably the posterior opening of the infraorbital canal.
The limits of prefrontals and frontals are not clear (Text-fig. 5). The region occupied by the frontals is quite depressed because of the thickening of the lateral border of the prefrontal, which forms the dorsomedial rim of the orbit. The postorbital bar is preserved only on the left side of the skull. The dorsal portion of the postorbital process is divided by an ascending cuneiform projection of the jugal, as in Exaeretodon (Bonaparte 1962). However, the resulting ‘V’-shaped contact in Protuberum is observed on the anterior and posterior faces of the postorbital bar, differing from Exaeretodon, in which this contact is developed on the lateral and medial faces. In lateral view, the postorbitals extend posteriorly covering most of the anterolateral portion of the reduced parietal.
The parietals are fused dorsally to form the parietal crest (Text-fig. 5), which is remarkably short in Protuberum. The parietal foramen (Text-fig. 5) opens dorsally and slightly anteriorly. It is located at the level of the posterior margin of the postorbitals, which cover the anterior ends of the parietals.
The masseteric process of the jugal (Text-figs 6–7) can be observed on the anteroventral portion of the zygomatic arch. This process has also been described in Exaeretodon (Bonaparte 1962), Luangwa (Kemp 1980), Traversodon (Barberena 1981a), Pascualgnathus (Bonaparte 1966) and Santacruzodon (Abdala and Ribeiro 2002). In Protuberum it is ventrally directed, as in Exaeretodon. The lateral view of the zygomatic arch resembles that in Exaeretodon, although the dorsal prolongation of the jugals does not project as far posteriorly as in the latter taxon. The squamosals extend for a considerable distance anteriorly, forming a cuneiform process that divides the posterior portion of the jugals, and almost reaches the level of the postorbital bar, as in Exaeretodon (Bonaparte 1962) and Massetognathus (Romer 1967).
Lateral wall of the braincase. In dorsal view, the posterior portion of the squamosals resembles the condition in Exaeretodon. The groove that separates the neurocranium and the zygomatic arch is very deep (Text-fig. 5) and this contact is narrow, resembling that of Massetognathus (Romer 1967). In lateral view, the posterodorsal border of the portions of the squamosals that form the lambdoid crests is rectangular (in contrast to that in Massetognathus, in which this border is rounded) and extends ventrally, outlining the external auditory meatus (Text-fig. 6A), as occurs in Exaeretodon and Massetognathus.
The posteroventral articulation of the epipterygoid with the prootic is preserved on the right side of the braincase. On this border, a foramen is present for the exit of the maxillary (V2) and mandibular (V3) branches of the trigeminal nerve (Text-fig. 6B).
The pterygoparoccipital foramen appears to open laterally, but this region is not well preserved. A similar condition is observed in tritylodontids, in the tritheledontid cynodonts Riograndia guaibensis and Pachygenelus (Luo 1994), and in the early mammals Sinoconodon and Morganucodon (Wible and Hopson 1993), but not in other traversodontids, in which the pterygoparoccipital foramen is enclosed by the squamosal. For most eucynodonts (including other traversodontids), the laterally open pterygoparoccipital foramen is a derived feature (Luo 1994; Luo et al. 2002), but in mammaliaforms, this feature is variable and homoplastic. It is open in Sinoconodon and all morganucodontians (Wible and Hopson 1993; Luo 1994) but closed in Adelobasileus (Lucas and Luo 1993) and Hadrocodium (Luo et al. 2001).
The limits of the prootics are unclear (especially ventrally). Laterally the sutures with the parietal and epipterygoid indicate that this bone forms much of the sidewall of the braincase and that it extends further dorsally than in Exaeretodon (Bonaparte 1962). In lateral view, the anteroventral limits of the prootic and the epipterygoid are very clear, especially at the level of the trigeminal foramen. Posteroventrally, the prootic delimitates the pterygoparoccipital foramen and posterodorsally the posttemporal fossa (Text-fig. 6B). Both openings communicate through a deep groove, in which only the portion closest to the posttemporal fossa (Brink 1955) is enclosed in a canal. A similar feature is present in Massetognathus: in the latter taxon this short canal was described as a lateral flange vascular canal by Rougier et al. (1992). Conversely, in Exaeretodon the pterygoparoccipital foramen and the post-temporal fossa communicate through a closed canal (Bonaparte 1962).
Basicranium. In ventral view, the squamosal bears a groove for the external auditory meatus that rapidly decreases in diameter medially and curves anteromedially (Text-fig. 7), terminating behind the quadrate and the quadratojugal. The quadrate and quadratojugal articulate with grooves in the squamosal. These elements are tightly fused, as in Massetognathus (Luo and Crompton 1994). Medial to the groove for the quadrate, a crest of the squamosal encloses the quadrate into the concavity. The squamosal does not make contact with the quadrate ramus of the epipterygoid (Text-fig. 7). The dorsal plate of the quadrate (Text-fig. 9) of Protuberum is similar to that of Massetognathus, but it possesses a quadrangular anterior outline. The surface of the dorsal plate is slightly concave and the dorsal angle is partially covered by the squamosal, as in Massetognathus.
On the ventral area of both prootics, the cavum epiptericum (Text-fig. 7) is open, as in most non-mammalian cynodonts (Wible and Hopson 1993; Luo and Crompton 1994). In ventral view, the paraoccipital process (Text-fig. 7) is better preserved on the left side of the cranium. The process is bifurcated, with an anterior and a posterior region, which is not observed in other traversodontids. This is a derived feature, present in tritylodontids, brasilodontids (Bonaparte et al. 2005) and in early mammals such as Morganucodon (Kermack et al. 1981) and Sinoconodon (Crompton and Luo 1993). The posterior paraoccipital crest (Text-fig. 7) is very clear and more strongly developed in Protuberum than in Brasilodon (Bonaparte et al. 2005) and extends from the contact with the squamosal to the border of the jugular foramen.
Although the preservation of the fenestra ovalis is not good on either side of the braincase, their borders appear to have been formed by a thickened ring of bone, which is a primitive feature that is also observed in other traversodontids (Kemp 1980) and non-mammalian cynodonts with the exception of Brasilitherium (Bonaparte et al. 2003, 2005). This is otherwise a derived condition present in mammaliaforms (Luo et al. 1995). The jugular foramina are better preserved and each is clearly confluent with the fenestra rotunda.
The margins of the parasphenoid and basisphenoid are not clear. In ventral view, the crest formed by the cultriform process of the parasphenoid is very high, as observed in Exaeretodon (MR, pers. obs., UFRGS PV0715T). Its anterior portion (extending between the pterygoids) is much shorter than in Massetognathus (Romer 1967) and is more similar to that of Exaeretodon (Bonaparte 1962). Posteriorly, the basisphenoid contacts the basioccipital along a clear serrated suture (Text-fig. 7). Laterally, the basisphenoid wing extends towards the fenestra ovalis (contributing to its medial margin), but its contact with the prootic is indistinguishable. Anteriorly, the basipterygoid process of the basisphenoid forms the anterior margin of the ventral opening of the cavum epiptericum. The foramina for the internal carotid are absent in the basisphenoid, as also occurs in other traversodontids, Cynognathus and trytilodontids (Hopson and Barghusen 1986).
It is not possible to determine if the basioccipital extends posteriorly between the occipital condyles (Text-fig. 7), as it does in Massetognathus (Romer 1967). No hypoglossal foramina are evident in the specimen, so they were probably confluent with the jugular foramina (Text-fig. 7) as in most non-mammalian cynodonts, including traversodontids (Luo 1994; Luo et al. 2002). As in most non-mammalian cynodonts, the occipital condyles of Protuberum are mostly composed of the exoccipitals. The exoccipitals contact each other medially, but their articulation with the basioccipital is not clear. The odontoid notch could not be observed. The condyles project posteriorly, so that in ventral view they are aligned with the posterior border of the lambdoid crests, in contrast to the condition in Exaeretodon (Bonaparte 1962) and Massetognathus (Romer 1967). The condyles are posteriorly placed in relation to the quadrates, as also occurs in Massetognathus. The ventral parts of the condyles are very close to each other, to a greater extent than occurs in Exaeretodon frenguellii (Bonaparte 1962), but similar to their position in Exaeretodon riograndensis (Abdala et al. 2002). They are not as vertically inclined as in the former and their form is bulbous, which is the most primitive form among non-mammalian cynodonts, including traversodontids (Lucas and Luo 1993).
The dental formula of Protuberum consists of three incisors, one canine and seven postcanines. Only the second and third incisors are preserved on the left side and the second incisor on the right side. They are slightly procumbent, but not as much as in Exaeretodon (Bonaparte 1962). Each incisor is similar in size, but the second has a spatulate tip. It is possible to observe enamel layers only on the labial face of incisors, which is also the case in Exaeretodon (Chatterjee 1982; Abdala et al. 2002) and Luangwa (Kemp 1980).
The right canine is small and approximately the same size as the incisors: the canine is vertically oriented. As with the incisors, the enamel layer can only be observed on the labial surface. The paracanine fossa is placed in a posteromedial position in relation to the upper canine, as is also observed in Exaeretodon and Scalenodontoides. This fossa is very deep and anteroposteriorly elongate so that it almost fills the diastema that is present between the canine and postcanines. The size of the lower canine is unknown, but the anteroposterior length of the paracanine fossa suggests that movement of the lower jaw occurred in a posterior direction.
The postcanine rows diverge somewhat in their posterior half. The first four teeth are implanted at right angles to the medial plane. The posteriormost postcanines begin to diverge markedly, reaching an angle of 50 degrees relative to the medial plane. They increase in size posteriorly and are extremely worn, with the exception of the posteriormost one (Text-fig. 10), which shows no wear. The third and fourth left postcanines are free from their alveoli. One of these became attached to the palate during fossilization, while the other became associated with the right paracanine fossa. A single root is visible on the former and does not show any sign of division. Its length is the same as the width of the crown of the tooth, so that the root shows the shape of an inverted triangle in distal view. The empty alveoli corroborate this observation.
The unworn teeth (Text-fig. 10) are similar in morphology to the postcanines of Pascualgnathus. Two cusps can be observed, one lingual and one labial. The lingual cusp is slightly higher than the labial and a low transverse crest (that can be observed only on the left unworn tooth) connects the cusps. There is no cingulum. Wear increases from the back to the front along the tooth row, as also observed in Luangwa (Kemp 1980), Massetognathus pascuali (Romer 1967) and other traversodontids. On the fifth postcanine the transverse crest has been worn off, producing two well-marked crests: a low one at the anterior border of the tooth and a high one at the posterior border. These crests enclose a deep concavity, so that the crown surface forms an oval concave area that lacks internal features. The original difference in size between the lingual and labial cusps is accentuated in comparison to the most anterior postcanines.
On the first four postcanines wear has produced a flat occlusal surface, that is continuous anteroposteriorly. The lingual and labial crests tend to decrease in size and no longer differ in size from each other. The general morphology of the worn postcanines resembles that in Luangwa but is much simpler. The shouldering of postcanines (an advanced feature for traversodontids, which is observed in Exaeretodon and Massetognathus, for example) is absent in Protuberum, a characteristic it shares with Luangwa, Pascualgnathus, Scalenodon, Andescynodon and Traversodon. In occlusal view, the teeth are subrectangular in outline and are nearly twice as broad as they are long anteroposteriorly, as also occurs in Massetognathus pascuali.
The axial skeleton of Protuberum (Text-figs 11–14) is peculiar due to the presence of numerous processes on the ribs and neural spines with expanded distal ends that are similar to those observed by Sues (1985) in the thoracolumbar vertebrae of Kayentatherium. The axial skeleton of Protuberum also resembles that of Pascualgnathus (Bonaparte 1966), Thrinaxodon, Cynognathus and Diademodon (Jenkins 1971), in its development of costal plates.
In some non-mammalian cynodonts (e.g. Thrinaxodon, Cynognathus and Diademodon), posterior ribs with costal plates delimit the intergirdle portion of the vertebral column into thoracic and lumbar regions (Brink 1955; Jenkins 1971). In other cynodonts costal plates are absent and it is not possible to recognise this division (e.g. Exaeretodon; Bonaparte 1963). There is also another pattern, which may be represented by Massetognathus (Jenkins 1970), in which specialized lumbar ribs occur; these are not in the form of costal plates but appear to be reduced versions of these plates, lending a ‘Y’-shape to the distal end of ribs. In Protuberum, the transition between the last thoracic and the first lumbar rib is based on the presence/absence of overlapping costal plates. On the lumbar ribs, the costal plates overlap with adjacent ribs anteriorly and posteriorly, with each posterior plate overlapping the anterior plate of the following rib.
The criteria proposed by Jenkins (1971) to distinguish between thoracic and lumbar regions in non-mammalian cynodonts were (1) thoracic ribs possess a rib shaft that extends distal to the costal plates and (2) the first lumbar rib can be defined as the first rib that bears a costal plate that contacts adjacent costal plates. In the case of Protuberum, all of the ribs possess a shaft distal to the costal plates. Therefore, the second criterion applies better in Protuberum, but in this taxon the definition is modified slightly so that the first lumbar rib is identified as that which bears a costal plate that contacts adjacent costal plates both anteriorly and posteriorly.
There are 19 vertebrae in the holotype of Protuberum, consisting of two cervicals, nine thoracics, five lumbars and three sacral vertebrae. All of the vertebrae are amphicoelous and robust with posteriorly inclined neural spines. Each neural spine bears posterolateral expansions at its dorsal end, so that in dorsal view the spines have a triangular outline. These expansions are more robust than those observed in Kayentatherium (Sues 1985).
The transverse processes of the cervical and the first thoracic vertebrae project posterolaterally, but from the third to the eighth thoracic vertebra these processes extend laterally. From the ninth thoracic (the last) to the third lumbar the transverse processes project anterolaterally, but on the two last lumbars (fourth and fifth), these processes project laterally again, as do those of the sacral vertebrae. Cervical and anterior thoracic centra are anteroposteriorly short, while the centra of the posterior thoracics, lumbars and sacral vertebrae increase in length and become more spool shaped. This trend is especially marked in the lumbar region where the centra have expanded anterior and posterior borders.
Cervical series. The two first vertebrae (Text-figs 11, 12A) of the holotype are considered to be cervicals, because they differ from the next vertebrae (here considered to be thoracic) in several aspects. The expansions at the extremity of the neural spine are much smaller in the cervicals than in the first thoracic, even though the expansions seen in the second cervical are larger than those present on the first cervical. In addition, the cervical neural arches are much narrower than those on the thoracic vertebrae and they lack the anterior lamina that is present in the thoracics (see description, below). Moreover, the proportions of the neural spines and vertebral centra are quite different, so that the neural arches of the cervicals are proportionately much higher than those of the thoracics, although the total height of the cervical and anterior thoracic vertebrae remains almost the same.
The cervical centra do not differ significantly from those of the first thoracics. They are short anteroposteriorly and bear a small laterally compressed ventral ridge. Cervical zygapophyses resemble those of the thoracic vertebrae: the prezygapophyses project beyond the anterior border of the vertebral centrum, while the postzygapophyses terminate above the posterior border of the centrum. A shallow anterior concavity is present on the dorsal border of the centrum, ventral to the prezygapophyses, which appears to act as an articular surface for a small tuberosity that extends posteriorly from the preceding centrum (Text-fig. 12A–B). This feature creates a very precise intervertebral articulation.
Thoracic series. The largest neural spines expansions are present in thoracic vertebrae 1–6 (Text-fig. 11), which might indicate that they were subjected to some kind of additional mechanical stress. The thoracic vertebrae (Text-fig. 12B) have anteroposteriorly elongated neural spines with an anterior lamina that projects into the posterior border of the anteceding vertebra. The proportions of the neural arches and centra differ from those observed in Exaeretodon (Bonaparte 1963) and Diademodon (Jenkins 1971): however, the thoracic neural spines of Protuberum are generally low and resemble the pattern observed in Massetognathus pascuali (Jenkins 1970). There are no clear sutures between the neural arches and centra (in contrast to Exaeretodon, in which those elements were not fused).
The anterior centra are anteroposteriorly short and compressed laterally, so that a low ventral ridge is present. These proportions change along the column and the posteriormost vertebrae have centra are at least 50 per cent longer. The length of the centra ranges from 18–30 mm. There is no evidence of intercentra. The zygapophyses have the same pattern as described above for the cervical vertebrae. Although it is hard to observe the articular surfaces of the pre- and postzygapophyses in the articulated vertebrae, it seems that these are almost vertical on the anterior thoracic vertebrae but are oriented more horizontally in the posterior vertebrae. The parapophyses are well developed on the dorsolateral aspect of centrum, adjacent to the articular surface. In Protuberum the facets for the thoracic rib heads are situated intervertebrally, as in Thrinaxodon and Cynognathus (Jenkins 1971) and in the anterior dorsal vertebrae of Massetognathus (Jenkins 1970). There are no indications of anapophyses in MGB 368/100, although poor preservation and the fact that most of the vertebrae are articulated in this specimen (with some matrix filling the spaces between them) could have obscured this feature. In the isolated vertebra UFRGS PV0985T (probably a thoracic), well-preserved anapophyses are present.
Lumbar series. The lumbar vertebrae (Text-fig. 12C) have short, robust and anteroposteriorly broad neural spines. No space is present between adjacent neural spines, due to poor preservation. The expansions at the distal extremities of the neural spines tend to decrease in size posteriorly.
The vertebral centra of the lumbar vertebrae are spool-shaped in ventral view and their lengths range from 27–37 mm. The anterior and posterior borders of the centra are expanded so that the contact area between centra of adjacent vertebrae is enlarged. The strength of their union can be confirmed by a postmortem inclination of the posterior part of the column that caused the breakage of the third, fourth and fifth lumbar vertebrae through the middle portion of their centra, but did not disarticulate them. The parapophyses are intervertebral, as in Massetognathus (Jenkins 1970). In Thrinaxodon, the last lumbar vertebrae have parapophyses located on the centra (Jenkins 1971). From the third lumbar vertebra and continuing posteriorly, synapophyses (fused parapophyses and diapophyses) occur in an intervertebral position in Protuberum, as also seen in the lumbars of Luangwa (Kemp 1980) and Thrinaxodon and the sacrals of Massetognathus. Articulations between lumbar pre- and postzygapophyses are unclear. The transverse processes are robust and anteroposteriorly broad.
Sacrum. The sacral vertebrae (Text-fig. 12D) are identified on the basis of contact between the ribs and the internal surface of the iliac blade. The solid construction of the posterior part of the vertebral column continues in this region. The first three sacral vertebrae are preserved in the type specimen (Text-fig. 11), but the sacral region of Protuberum probably contained more vertebrae, as Thrinaxodon has five sacrals (Jenkins 1971), while Exaeretodon has seven (Bonaparte 1963).
The only preserved neural spine is that of the first sacral vertebra. It is shorter than those of the lumbar vertebrae and is less robust. The expansions at the distal extremity of the neural spine are smaller than those on the lumbar vertebrae. Sacral centra are spool-shaped (as in the lumbar vertebrae). The length of the first sacral vertebra is about 32 mm, the second has a severe fracture and dislocation of the centrum, making its measurement difficult, but the third is approximately 24 mm long, indicating a tendency towards shortening of the posterior sacral vertebrae. The breakage of the second sacral vertebra occurs in a similar way to that described for the third, fourth and fifth lumbar vertebrae (see above), with an intact intervertebral articulation indicating strong fusion between the sacrals as well as the lumbars.
Synapophyses are developed on the sacral vertebrae, but they are placed on the anterior border of each vertebra (instead of intervertebrally, in contrast to the lumbars). It is impossible to distinguish the pre- and postzygapophyses, due to the poor preservation and the strong intervertebral attachment. The transverse processes are anteroposteriorly narrower than those of the lumbar vertebrae.
The presacral ribs of Protuberum (Text-figs 11, 13) are unique among traversodontids, for having a series of distinctive knob-like processes along their dorsal borders. These processes decrease in size distally on each rib, so that the proximal processes (close to the tuberculum) are generally the largest. On some of thse processes, a dorsolateral concavity develops. The middle thoracic ribs are the longest and bear up to eight processes. The number of processes decreases in shorter ribs (i.e. the posterior thoracic, lumbar and cervical ribs), but the processes themselves do not change significantly in size wherever they are found along the column. Regular spacing separates each process and they form rows that are parasagittally aligned. Similar processes occur on the dorsal margin of the illium, with these processes in alignment with those on the ribs (see below). This condition is comparable to Thrinaxodon (Jenkins 1971), in which tubercles are present on the dorsomedial edge of costal plates and also form parasagittally aligned rows.
The surface of each process is smooth and they are composed primarily of compact (probably pachyostotic) bone, suggesting that they were not covered by cartilaginous tissue. All of the presacral ribs (including the cervical ribs) exhibit modest curvature in transverse section, so the trunk of Protuberum was probably wide and flat. The rib-vertebra contact is very strong on the posterior thoracic and lumbar regions, making it hard to distinguish the limits between the rib heads and their respective apophyses on the vertebrae. All of the ribs on the left side are fractured and somewhat displaced between the tuberculum and the first protuberance (this fracture has affected all of the ribs from the thoracic to sacral region) but their heads remain articulated with the vertebrae. Conversely, the anterior portion of the trunk in MGB 368/100 does not preserve any articulated ribs, indicating a weaker articulation of the ribs on the vertebrae in this region. The ventral and dorsal borders of ribs are convex to produce an ellipsoid transverse cross-section. A dichocephalous condition (Romer 1956, p. 277) is observed in most ribs, except for the last lumbar and sacrals, where the tuberculum and capitulum are fused.
Cervical ribs. Cervical ribs are rarely preserved in non-mammalian cynodonts, so few specimens are available for comparison. However, comparisons with Exaeretodon suggest that two ribs referred to Protuberum represent the cervical region and that they are probably from posterior cervicals. One of these ribs is well preserved (UFRGS PV 1009T; Text-fig. 13A) and lacks only the distal tip, while the other is represented by its proximal end only (UFRGS PV 0981T).
UFRGS-PV1009T has three processes on its dorsal margin, which decrease in size distally. These protuberances occur only on the proximal half of the rib. The first protuberance has a distinct morphology, being twice as elongate proximodistally as the other two. The rib has a small ridge on the posterior surface, which begins at the capitulum and ends 15 mm beneath the tuberculum. In the same region, but on the anterior surface of the rib, a shallow groove is present. The capitulum is very broad in UFRGS PV 1009T and presents a slightly convex, almost flat articular facet. UFRGS PV 0981T (a right proximal fragment) has a small capitulum, that is approximately the same size as the tuberculum. The larger size of the capitulum in the former is probably a consequence of preservational distortion, so it is most likely that the tuberculum and capitulum had similar sizes originally.
The curvature observed in UFRGS PV 1009T is not particularly strong: this contrast with the morphology of the first cervical ribs of Exaeretodon, which have a ‘horseshoe form’ (Bonaparte 1963, page 20).
Thoracic ribs. No anterior thoracic ribs are preserved, so it is not possible to observe the transition between the cervical and thoracic ribs. The processes of the thoracic ribs are characterized by a globular shape. The most proximal process sometimes displays a concavity dorsolaterally and is generally the largest and anteroposteriorly longest process on each rib, although the second process reaches a similar size in some ribs (e.g. in the last left thoracic rib the second process is longer than the first one). The number of processes present varies from eight (as seen on the seventh thoracic rib) to five (on the ninth thoracic rib).
An anterior crest arises in the proximal region of some thoracic ribs (Text-fig. 13B), also occurs in Massetognathus (Jenkins 1970) and Cynognathus (Jenkins 1971). This crest begins below the most proximal dorsal process. From the eighth thoracic rib onward, another crest on the posterior margin of the ribs can be observed. These crests do not overlap adjacent ribs. Both crests become slightly larger on the ninth (last) thoracic, so that the posterior crest of this rib overlaps the anterior costal plate of the first lumbar rib. The crests abruptly enlarge in the lumbar region (and are termed costal plates hereafter), whereas in Luangwa (Kemp 1980), Cynognathus and Diademodon (Jenkins 1971), these crests gradually increase in size along the thoracic and lumbar series. The posterior surface of the shaft is marked by a shallow groove (Text-fig. 13C), which is also observed on ribs bearing a posterior crest and lies directly beneath this structure. According to Jenkins (1971), this groove probably represents an intercostal neurovascular sulcus.
The heads on the middle thoracic ribs consist of a short tuberculum and an elongate and robust capitulum, as occurs in some thoracic ribs of Exaeretodon (Bonaparte 1963). A small ridge is present between the tuberculum and capitulum of each thoracic rib. The heads of the posterior thoracic ribs are comparable to those of many non-mammalian cynodonts (e.g. Cynognathus, Diademodon, Luangwa, Massetognathus and Exaeretodon), with an enlargement of the tuberculum and a shortening of the capitulum; consequently, these structures converge in the posterior region of the axial skeleton. The distal portion of each thoracic rib decreases in size gradually, but the tips terminate abruptly, presenting a peg-like morphology: in contrasts with Exaeretodon in which the distal ends of ribs taper gradually.
Lumbar ribs. All ten lumbar ribs are articulated with the five lumbar vertebrae in MGB 368/100 (Text-fig. 11). The costal plates of the lumbar ribs are lanceolate in shape, and lack the subrectangular outline described in Thrinaxodon, Diademodon and Cynognathus (Jenkins 1971) and Pascualgnathus (Bonaparte 1966). An anteriorly directed process arises from the anterior border of the last costal plate in ventral view.
The general morphology of the costal plates in Protuberum differs significantly from that of other non-mammalian cynodonts, in which the costal plates of the lumbar ribs are characterized by a reduction of the distal portion of the shaft of the ribs, so that just the plate remains. In Protuberum the shafts of the ribs do not exhibit such severe reduction, and a portion of the shaft continues distal to the costal plates. The dorsal processes of the lumbar ribs are similar to these observed on the thoracic ribs. They differ in number, however, with four on each lumbar rib, except for the last rib, which is short and bears only one process.
The lumbar ribs are fused to the vertebrae. The capitular processes become progressively shorter toward the sacrum and shift dorsally to be in closer proximity to the tuberculi. All capituli articulate intervertebrally. Starting from the third lumbar rib, the capitular and tubercular facets are essentially confluent. The last lumbar rib has a fused capitulum and tuberculum.
The last lumbar rib shows a marked decrease in length. Its posterior border has a broad contact with the anterior portion of the iliac blade, which is facilitated by the lateral curvature of the anterior portion of the iliac blade. The distal ends of the lumbar ribs terminate abruptly, with a sub-quadrangular outline in anterior view, and a small process is placed just dorsal to the distal tip of the rib.
Sacral ribs. Only two pairs of sacral ribs are preserved in the type specimen (Text-fig. 11). They are poorly preserved, but some of their features can still be described. The sacral ribs lack dorsal processes. Their general morphology resembles that of the sixth and seventh sacral ribs of Exaeretodon (Bonaparte 1963). The first sacral rib is the most robust, especially distally where it contacts the iliac blade. The second rib has a smaller distal expansion and a distal fragment of the third left sacral rib, which is still attached to the iliac blade, is even more slender.
As in Thrinaxodon (Jenkins 1971) and the last lumbar rib of Protuberum, the capituli and tuberculi are fused. The shaft is short and posterolaterally oriented and the distal ends are expanded anteroposteriorly and strongly fused to the iliac blade.
A fragmentary left ilium (Text-figs 11, 14) is present in the type specimen, but its acetabular and ventral portions are not preserved. It has a robust and elongate blade, with a lanceolate anterior edge, which possesses rounded processes along its dorsal border that are similar to those of the presacral ribs. These processes are aligned with those on the ribs and form part of the same parasagittal series (see above). The blade is laterally concave, especially in its anterior half, where it curves laterally: the anterior part of the ilium extends parallel to the last lumbar rib at an angle of approximately 80 degrees to vertebral column.
One of the most striking features of Protuberum is the short parietal crest, which represents only 7.5 per cent of skull length. In Exaeretodon (Bonaparte 1962), the parietal crest represents about 30 per cent of skull length, while in Luangwa sudamericana this figure is 28 per cent (Abdala and Teixeira 2004: MCP 3284) and is 24 per cent in Massetognathus (UFRGS PV 0968T). The thickened skull roof is quite remarkable: its function is unknown, but it is possible that it was associated with defense or burrowing.
The incisors of Protuberum have differential thicknesses of enamel on their labial and lingual surfaces. Kemp (1980) described a similar condition in Luangwa and observed that this feature ensured that the ridges on the teeth remained sharp throughout life. This self-sharpening tooth could have been employed in gripping, cutting and tearing off food. Moreover, the wide concave areas formed on the crowns could be employed for plant crushing. The postcanines of Protuberum are deeply worn (except for the posteriormost teeth) as also occurs in other traversodontids, such as Massetognathus (Crompton 1972a), Dadadon (Goswami et al. 2005) and Luangwa (Kemp 1980). The latter possesses a pattern of tooth wear that seems applicable to Protuberum (Text-fig. 15). The mechanism proposed for the formation of the broad but poorly matching concave wear facets present on the upper and lower postcanines of Luangwa was a form of food-to-tooth contact analogous with mammalian puncture-crushing (Crompton and Hiiemae 1969), rather than abrasion caused by occlusion. The wear pattern observed in these traversodontids supports the hypothesis of a posterodorsally directed power stroke of the mandible (Crompton 1972a), which is reinforced in Protuberum by the presence of anteroposteriorly elongate paracanine fossae and the probable mobility of the quadrate.
Luo and Crompton (1994) and Crompton (1972b) also interpreted proposed a posteriorly directed movement of the lower jaw of power stroke in Massetognathus, based on the morphology of the quadrate. The quadrate trochlea in Protuberum is not as well pronounced as in Massetognathus, but its orientation is similar and consistent with the same type of movement. An anteroposterior movement was also suggested for the quadrate of Diademodon (Brink 1955), which is closely related to traversodontids (Abdala and Ribeiro 2003). The quadrate in Protuberum is not rigidly attached to the squamosal and may, therefore, have had some freedom of movement as also suggested for Exaeretodon (Allin 1975).
Limited morphological diversity is apparent in the postcranial skeletons of Triassic non-mammalian cynodonts. However, rib morphology does appear to be useful for distinguishing traversodontid taxa. Pascualgnathus polanskii, the earliest traversodontid (Puesto Viejo Formation, Late Olenekian–Early Anisian; Bonaparte 1966) for which ribs are known, has a costal morphology similar to that of Diademodon (Scythian–early Anisian; Kitching 1995), Luangwa (Anisian; Brink 1963) and Cynognathus (Scythian–early Anisian; Kitching 1995). Massetognathus pascuali (Chañares Formation, Ladinian; Jenkins 1970), which is temporally intermediate between P. polanskii and Exaeretodon, has rib specializations that are restricted to the lumbar region, while Exaeretodon (Ischigualasto Formation, Carnian; Bonaparte 1963) lacks rib specializations. This reversal to a non-specialized condition was interpreted as a derived feature by Jenkins (1970). Protuberum (Ladinian), however, shows unique and distinctive rib morphology: while other traversodontids, such as Exaeretodon, tended to simplify rib structure, Protuberum developed more elaborate ribs.
The presence of costal plates of Protuberum allows recognition of separate thoracic and lumbar regions, as in Diademodon, Thrinaxodon and Cynognathus (Jenkins 1971). This distinction points to a possible division into thoracic and abdominal cavities, as suggested by Brink (1955). In the case of Protuberum, the presence of both dorsal processes and costal plates gives the lumbar region an exceptionally robust appearence. This, in combination with the simple curvature of the ribs, indicates that Protuberum had a distinctive trunk morphology. It is possible that this robust and specialized postcranial structure may have provided protection against predators, a conclusion supported by some other features of the skeleton, such as the thickened skull roof (see above).
The accentuated curvature of the anterior portion of the iliac blade provided a wide surface for muscle insertion in that region. Therefore, the hindlimb and associated musculature may have been very strong, especially the m. iliofemoralis and m. iliotibialis, both of which originate on the iliac blade. Indeed, strong, robust limbs would have been necessary to support the robust axial skeleton, and could have been employed in burrowing or digging for food.
A phylogenetic analysis to investigate the relationships of Protuberum to other cynodonts was conducted using the 28 craniodental characters proposed by Abdala and Ribeiro (2003). Our data matrix incorporates 16 taxa, including Diademodon, Trirachodon and 14 traversodontids. Analyses were performed using the beta version of PAUP 4.0 (Swofford 1998) with Diademodon as the outgroup. The branch-and-bound search method was applied and produced a single most parsimonious tree (tree length = 58, Consistency Index = 0.569, Rescaled Consistency Index = 0.418).
It is important to be cautious with the resulting cladogram (Text-fig. 16), because the data matrix used by Abdala and Ribeiro (2003) focused on dental characters (21 characters, with the remaining seven relating to the rest of the cranium and mandible) and excluded postcranial information. In addition, as the mandible of Protuberum is currently unknown, our analysis is missing data for eight lower dentition/mandibular characters. However, this preliminary analysis places Protuberum in a clade with Gomphodontosuchus, Exaeretodon, Scalenodontoides hirschsoni and Menadon. This clade shares characters such as large incisors (acquired convergently in S. hirschsoni), the presence of three upper incisors (acquired convergently in S. hirschsoni and Pascualgnathus), paracanine fossae positioned posterior to the upper canine, absence of the internarial bar and the presence of a well developed posterior extension of the jugal above the squamosal in the zygomatic arch (this condition is unknown in Gomphodontosuchus).
Protuberum cabralensis is a new traversodontid cynodont that can be diagnosed on the basis of several distinctive features, particularly the presence of prominent processes on the cervical, thoracic and lumbar ribs, similar protuberances on the anterodorsal margin of the ilium and in its unique dental morphology. Protuberum possesses a combination of characters thought to be derived (e.g. the lack of the internarial process of the premaxilla, bifurcation of the paroccipital process and the lateral opening of the pterygoparoccipital foramen) and primitive (e.g. morphology of the iliac blade and presence of rib specializations) among traversodontids: more work is needed to establish the phylogenetic position of this taxon within an expanded analysis of traversodontid relationships.
Acknowledgements. We would like to thank C. Salla Bortolaz (technician at MGB) for completion of the arduous and skillful preparation of the type specimen and for assistance on beginning this work. Additional thanks to A. Battaglin Cafaro (Secretary of Tourism and Culture in the Municipality of Mata) for lending us the type specimen. Special thanks to J. F. Bonaparte, E. Snively, R. C. Fox, T. Kemp and Z.-X. Luo for valuable comments on earlier versions of the manuscript. D. Larson helped with the phylogenetic analysis. L. Morato and T. Veiga de Oliveira also contributed important ideas. Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) provided financial support.
Morphological characters used in this paper are as follows. Characters are derived from Abdala and Ribeiro (2003). ‘0’ represents the primitive state.
1Number of upper incisors: four (0); three (1).
2Incisor size: small (0), large (1).
3Diastema between upper incisors and canine: present (0), absent (1).
4Upper canine size: large (0), reduced (1).
5Lower canine size: large (0), reduced (1).
6Position of paracanine fossae in relation to the upper canine: anteromedial (0), medial (1), posteromedial (2).
7Overall morphology of the upper postcanines: ovoid-ellipsoid (0), rectangular-trapezoidal (1).
8Shouldering in upper postcanines: absent (0), present (1).
9Posteromedial inclination of the last upper postcanines: absent or small (0), oblique (1).
10Transverse crest of upper postcanines: central (0), anterior (1), posterior (2).
11Number of cusps in the transverse crest of upper postcanines: two (0), three (1).
12Central cusp of upper transverse crest: midway between buccal and lingual cusps (0), closer to the lingual cusp (1).
13Posterior cingulum in upper postcanines: present (0), absent (1).
14External cingulum in the anterior portion of the upper postcanines: absent (0), present (1).
15Anterolingual cusp in upper postcanines: absent (0), present (1).
16Number of cusps in the sectorial border of the upper postcanines: three (0), one (1), two (2).
17Overall morphology of the lower postcanines: ovoid-ellipsoid (0), quadrangular (1).
18Transverse crest in lower postcanines: central (0), anterior (1).
19Number of cusps in the transverse crest of the lower postcanines: two (0), three (1).
20Size of the anterior cusps in the lower postcanines: labial lower than lingual (0), labial higher than lingual (1).
21Cingulum in front of the transverse crest in the lower postcanines: absent (0), present (1).