Gnathovorax cabreirai: a new early dinosaur and the origin and initial radiation of predatory dinosaurs

Phylogenetic analysis

In order to infer its phylogenetic affinities and potential implications for understanding early dinosaur evolution, the new taxon was scored in a modified version of the data matrix of Müller et al. (2018), which is an updated version of that of Cabreira et al. (2016). In addition to the new herrerasaurid, eight new operational taxonomic units (OTUs) were included. These new OTUs and their respective sources are: the lagerpetid Dromomeron gigas, following the scores of Martínez et al. (2016) and Müller, Langer & Dias-da Silva (2018a); the ornithischians Tianyulong confuciusi, Fruitadens haagarorum, and Echinodon becklesii, following scores of Agnolín & Rozadilla (2018); the saurischian Nhandumirim waldsangae, following the scores of Marsola et al. (2018); and the sauropodomorphs Bagualosaurus agudoensis, Unaysaurus tolentinoi, and Macrocollum itaquii, following the scores of Pretto, Langer & Schultz (2018) and Müller, Langer & Dias-da Silva (2018b). Finally, the modifications in scores of Pisanosaurus mertii and Silesaurus opolensis performed by Agnolín & Rozadilla (2018) were also applied. The final data matrix includes 259 characters and 52 OTUs.

The data matrix was the subject of an equally weighted parsimony analysis in TNT v. 1.1 Goloboff, Farris & Nixon (2008). Characters 3, 4, 6, 11, 36, 60, 62, 64, 83, 115, 123, 139, 147, 148, 157, 160, 171, 173, 175, 178, 179, 182, 195, 200, 201, 202, 205, 216, 222, 240, and 248 were treated as ordered. Euparkeria was used to root the most parsimonious trees (MPTs), which were recovered with a ‘Traditional search’ (random addition sequence + tree bisection reconnection) with 1,000 replicates of Wagner trees (with random seed = 0), tree bisection-reconnection and branch swapping (holding ten trees per replicate).

CT-scanning and three-dimensional reconstructions

The skull of CAPPA/UFSM 0009 was scanned with a Philips Brilliance 64-Slice CT Scanner (located at Santa Maria city), using 120 kV, 150.52 mAs. The analysis generated 1,002 slices with a 0.67 mm thickness, increment of 0.33 mm, and pixel size of 0.553 mm. Dragonfly 3.8 was used for rendering the three-dimensional volume of the skull. To access the endocranial information of CAPPA/UFSM 0009, its braincase was scanned with a µCT scan Skyscan™ 1173 at Laboratório de Sedimentologia e Petrologia of the Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Porto Alegre, Brazil, using 130 kV and 61 µA. The scan resulted in 2,631 tomographic slices, with a voxel size of 29.98 µm. Before segmenting the regions of interest, slices without information were deleted, and the remaining files were cropped and binned (Balanoff et al., 2016) using the software ImageJ (Abràmoff, Magalhães & Ram, 2004). After these procedures, the sequence of tomographic slices was limited to 1,200, with a voxel size of 59.96 µm. The slices were imported in Avizo™ v. 8.1, the internal cavities were manually segmented, and 3D-models were generated (.stl format). The resulting 3D-models were coloured using Design Spark Mechanical 2.0 (the 3D models are in Pacheco et al. (2019) and the Supplementary File).

Morphological disparity analysis

A morphological disparity analysis was performed in order to explore ecomorphological patterns among early dinosaurs and their relatives. We employed the dental characters of the phylogenetic taxon-character matrix (see Supplemental Information) and followed the parameters adopted by Müller & Garcia (2019), where an Euclidean distance matrix (EDMA) was calculated from the ecomorphological dataset using the software MATRIX (Wills, 1998). Then, a principal coordinate analysis (PCoA) was performed for the EDMA with the multivariate package GINKGO (Bouxin, 2005). The centroid of all operational taxonomic units (OTUs) was taken as the origin of multivariate axes, and the Cailliez method of negative eigenvalue correction was adopted. Finally, a bivariate graph with axes 1 and 2 of the PCoA was constructed using the software PAST (Hammer, Harper & Ryan, 2001).

Terminology

This study employs veterinarian anatomical terms. Therefore, ‘cranial’/‘rostral’ and ‘caudal’ are used rather than the traditional or “Romerian” terms ‘anterior’ and ‘posterior’.

Nomenclatural acts

The electronic version of this article in Portable Document Format (PDF) will represent a published work according to the International Commission on Zoological Nomenclature (ICZN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/. The LSID for this publication is: [urn:lsid:zoobank.org:pub:B555EB12-8CEA-4043-A652-4AFE6B02EC97]. The online version of this work is archived and available from the following digital repositories: PeerJ, PubMed Central and CLOCKSS.

Systematic paleontology

Holotype: CAPPA/UFSM 0009 (Centro de Apoio à Pesquisa Paleontológica da Quarta Colônia/Universidade Federal de Santa Maria), an almost complete and partially articulated skeleton (lacking only part of the left shoulder girdle and part of the left forelimb).

Locality and Horizon: Marchezan site (29°37′52″S; 53°27′02″W), municipality of São João do Polêsine, Rio Grande do Sul, Brazil (Fig. 1A); Santa Maria Formation, Candelária Sequence, Paraná Basin (sensu Horn et al., 2014); stratigraphically correlated beds from a nearby site were dated as mid-Carnian (ca 233.23 ± 0.73), Late Triassic (Langer, Ramezani & Da-Rosa, 2018).

Etymology: The generic name is derived from the Greek gnathos, jaw, and the Latin vorō (“devour”) + −āx (“inclined to”). The specific epithet honours Dr. Sérgio Furtado Cabreira, the palaeontologist that found the specimen described here.

Diagnosis: Gnathovorax cabreirai differs from all other known herrerasaurids based on a unique combination of character states (*local autapomorphy): three premaxillary teeth*; presence of an additonal fenestra (=oval fenestra) between the maxilla and premaxilla contact, which is located dorsal to the subnarial fenestra, though markedly smaller than the subnarial fenestra*; two well defined laminae in the antorbital fossa of the maxilla, with a depression between them*; promaxillary fenestra absent*; slender ventral ramus of the lacrimal extending caudally almost until the midpoint of the orbit*; supraoccipital trapezoidal in caudal view; anterior portion of the basioccipital articulates with the parabasisphenoid through a V-shaped suture*, transverse processes of the last dorsal vertebra lacks contact with the ilium; distal end of the scapula craniocaudally expanded relative to the more proximal shaft; distally extended ente- and ectepicondyle in the humerus; preacetabular ala of the ilium with a pointed cranial tip*; pubis with a sinuous lateral margin in cranial/caudal views and a ventrally oriented shaft in lateral view; no depressed surface (=bevel) on the craniomedial region of the distal end of the pubis; presence of an opening on the obturator plate of the ischium*; proximal portion of the femur with lacks a caudomedial tuber; tibia equals to 90% of the femoral length; and three phalanges in pedal digit V*.

Taphonomic remarks: The skeleton of Gnathovorax cabreirai was preserved almost entirely articulated (Fig. 1B), lying on its right side in a mudstone layer, ridden with mudcracks and invertebrate bioturbation. Most of the missing elements are from its left side (mostly forelimb elements and ribs), indicating that the skeleton rested on the substract for an unknown amount of time before its final burial. The carcass as a whole shows no sign of transport, apart from the aforementioned elements that were carried away. Remains of small prozostrodont cynodonts, including semi-articulated specimens (Pacheco et al., 2018 and undescribed material) overlap the herrerasaurid skeleton, suggesting that the cynodonts were also not substantially transported, and may have roamed the environment surrounding the carcass. The mudstone that preserved both herrerasaurid and cynodonts was covered by a sandstone layer that also infilled the underlying mudcracks, indicating a moderate energy sedimentary episode. This sandstone preserved rhynchosaur bones, including limb elements, ribs, vertebrae and cranial material of at least two individuals. Some of the rhynchosaur elements overlapped the skeleton of G. cabreirai, which was also partially buried by the sandstone layer. The rhynchosaur bones were clearly deposited after the death of the herrerasaurid, but their disarticulation suggests that the bones were exposed for an extended time in the active taphonomic zone (or that the disarticulation occurred quickly, which seems less plausible), being later incorporated into this taphocoenosis probably via hydraulic transport. Yet, the absence of abrasion, rounded edges, and hydraulic selection, suggests that the rhynchosaur bones were not transported for a long time/distance. Therefore, it is likely that these three taxa (dinosaur, rhynchosaurs, and cynodonts) shared the same habitat in life (Fig. 1C).

Description: The skull of Gnathovorax cabreirai is almost entirely preserved (Fig. 2). Some of the bones from the left side are disarticulated, whereas the right side is better preserved and articulated. The body of the premaxilla is roughly quadrangular in lateral view and its caudodorsal process extends caudal of the caudal margin of the external naris, excluding the maxilla from its margin. The bone bears three tooth positions, whereas in Herrerasaurus ischigualasensis (Sereno & Novas, 1993) there are four. An oval fenestra is located between the premaxilla and maxilla, above the subnarial foramen and caudal to the external naris. That fenestra is smaller than the subnarial foramen, differing from H. ischigualasensis and possibly also Sanjuansaurus gordilloi (Alcober & Martinez, 2010). The maxilla presents a line of several foramina placed dorsal to the tooth row as in H. ischigualasensis and other saurischians. As in the aforementioned Argentinean herrerasaurids, the antorbital fossa is narrow, bordering the antorbital fenestra. Gnathovorax cabreirai lacks the promaxillary fenestra, which is seen in some other herrerasaurids, unaysaurid sauropodomorphs, Eodromaeus murphi, and most latter theropods (Sereno & Novas, 1993; Alcober & Martinez, 2010; Martinez et al., 2011; Müller, Langer & Dias-da Silva, 2018b). As in many other dinosaurs, an invagination is present between the maxillary ascending process and the lateral surface of the antorbital fossa. However, the new taxon uniquely shows a second invagination, forming two laminae in the antorbital fossa (Figs. 3A–3B). The ventral (=jugal) ramus of the lacrimal presents a slender caudal projection that extends along the ventral margin of the orbit (Fig. 1A), almost reaching its midpoint, as in Daemonosaurus chauliodus. This condition differs from that of H. ischigualastensis, in which the rostral lamina of the jugal is overlapped by a short ventral process of the lacrimal. As in many saurischians, the jugal has a prominent longitudinal ridge just dorsal to its ventral margin. In addition, distinct from early sauropodomorphs, the rostral tip of the jugal is dorsoventrally expanded. The infratemporal fenestra is trapezoidal in lateral view, with the dorsal half rostrocaudally shorter than the ventral, as found in H. ischigualastensis. The quadratojugal is partially fused to the quadrate. The ventral process of the squamosal is rostrocaudally broad, distinct from the strap-like shape of early sauropodomorphs, such as Eoraptor lunensis (Sereno, Martínez & Alcober, 2013), Buriolestes schultzi (Müller et al., 2018), and Saturnalia tupiniquim (Bronzati, Müller & Langer, 2019). On the dorsolateral surface of the bone, there is a marked slot for the postorbital, which is mediodorsally bounded by a ridge. The posterior process of the squamosal is tongue-like and proportionally longer than in H. ischigualastensis. The new taxon has a broad supratemporal fossa (Fig. 2C), which invades a significant portion of the caudal half of the frontal. Distinct from other non-herrerasaurid dinosaurs from Carnian (Cabreira et al., 2011; Martinez et al., 2011; Sereno, Martínez & Alcober, 2013), G. cabreirai lacks any palatal teeth.

The supraoccipital of G. cabreirai is trapezoidal in caudal view (Fig. 4A), differently from the triangular bone of H. ischigualastensis. The bone bears a tall nuchal crest that is lateromedially compressed at its caudal portion and lacks any dorsal excavation for the vena occipitalis externa. The distal portion of the paraoccipital process is dorsoventrally expanded and the entire process is proportionally longer than in H. ischigualastensis. The laterosphenoid has two notches at its ventromedial margin. The rostral one corresponds to the medial margin of the foramen for the trochlear nerve (IV), whereas the caudal one corresponds to the medial margin for the foramen for the oculomotor nerve (III). In addition, the ventral contact of the laterosphenoid with the prootic bounds the entrance for the trigeminal nerve (V). Caudal to the notch for the trigeminal nerve, the prootic also bears a pair of foramina for the facial nerve (VII) (Fig. 4C). The floor of the endocranial cavity is smooth, lacking a longitudinal crest (=eminentia medullaris). The occipital condyle of H. ischigualastensis (PVSJ 407) is far more gracile compared to G. cabreirai. The anterior portion of the basioccipital articulates with the parabasisphenoid through a V-shaped suture, contrasting with the U-shaped suture of H. ischigualastensis. The parabasisphenoid presents a weak recess on its ventral surface.

The hindbrain endocast (medulla oblongata and the cerebellum) shows a constriction at the level of the inner ear and, on its dorsal surface, there is a longitudinal eminence (Figs. 4E–4F). There is a well-developed Floccular Fossa Lobe of the cerebellum (FFL) (see Bronzati et al., 2017), which forms a caudolaterally oriented protuberance, passing through the anterior semicircular canal. In the inner ear, the anterior semicircular canal (ASC) is rounded and more developed than the posterior (PSC) and lateral (LSC) and taller than the PSC. In dorsal view, the angle between ASC and PSC is 78°; in lateral view, the angle between PSC and LSC is around 60°, and 84 between LSC and ASC in anterior view. In lateral view, the crus commune is oriented 70° in relation to LSC and slightly curved caudally. The cochlear canal is ventrally projected, tapering in that direction.

The dentary is a long and robust bone with several vascular foramina along its length, where the last one lies on a longitudinal groove (Fig. 3C). The dentary symphysis is short and limited to the rostral portion of the bone. Contrasting with sauropodomorphs (Nesbitt, 2011; Cabreira et al., 2016; Bronzati, Müller & Langer, 2019), the dorsal surface of the rostral tip of the dentary does not deflect ventrally. Moreover, there is no edentulous gap between the rostral extremity of the dentary and the first tooth. Apparently, G. cabreirai possesses an intramandibular sliding joint similar to that inferred for H. ischigualastensis and St. pricei, which encompasses the dentary/surangular contact, dorsal to the external mandibular fenestra, and the splenial/angular articulation, ventral to that aperture. However, as both mandibular rami are broken in the area of these articulations, it is difficult to properly access their anatomy.

The dental formula of G. cabreirai is three premaxillary, 19 maxillary, and about to 14 dentary teeth. All tooth crowns are blade like, caudally curved and labiolingually compressed. The premaxillay and dentary teeth lack serrations in their mesial margins. However, in the distal margin there are small serrations that form a right angle with the main axis of the tooth. In the maxillary teeth the serration occur in both margins.

Gnathovorax cabreirai preserves the complete cervical-dorsal series, with nine partially articulated cervical and 16 articulated dorsal vertebrae. Unlike most sauropodomorphs (Müller, Langer & Dias-da Silva, 2018a; Müller, Langer & Dias-da Silva, 2018b), the neck of the new taxon is short, with vertebral centra no more than 2.5 times longer than tall (Fig. 5A). There is a longitudinal keel on the ventral surface of the centrum of each cervical vertebra, as also seen in other herrerasaurids. Gnathovorax cabreirai resembles San. gordilloi regarding the presence of a long and caudolaterally directed transverse process in the caudal cervical vertebrae, which are considerably shorter in H. ischigualastensis and St. pricei. In the first dorsal vertebra, the ventral keel is markedly reduced if compared with that of the first cervical element. Also, the dorsal transverse processes are more elongated and laterally oriented, differing from those of the last cervical element, which are short and caudoventrally oriented. In addition, the first dorsal vertebra is 30% shorter than the last cervical one, as in San. gordilloi. The neural spines of the dorsal vertebrae are “H” shaped in cross-section, resembling the typical herrerasaurid condition. Unlike San. gordilloi, the new species lacks any cranial or caudal processes in the distal end of the neural spine. Nonetheless, some of the dorsal vertebrae show slight lateral expansions, forming a lateral expansion at the dorsal margin. The sacrum is composed of the two primordial sacral vertebrae, but a dorsal element seems to be positioned between the ilia. Its transverse processes are slender and project craniolaterally, almost reaching, but not contacting the ilia. The proximal (=cranial) caudal vertebrae (Fig. 5B) have tall centra (taller than the neural arches) and zygapophyses located far from the transverse process. The transverse processes are semi-circular in section, as in H. ischigualastensis and St. pricei, and not flattened as in San. gordilloi. As in other herrerasaurids, the neural spines of those vertebrae project vertically (Novas, 1993; Bittencourt & Kellner, 2009; Alcober & Martinez, 2010). In the distal caudal vertebrae, as in H. ischigualastensis, the prezygapophyses occupy a lower position, and gradually become longer, overlapping the distal half of the preceding vertebra.

Unlike H. ischigualastensis and San. gordilloi, the scapula and coracoid (Fig. 6A) are not co-ossified in G. cabreirai. In addition, the glenoid is formed almost equally by both bones, differing from other herrerasaurids, in which the coracoid forms most of the glenoid area (Sereno, 1993; Alcober & Martinez, 2010). The distal end of the scapula expands craniocaudally, differing from San. gordilloi and St. pricei in which that part of the bone retains the same width of the midshaft. As in San. gordilloi, the scapular shaft is curved caudally, differing from the nearly straight blade of H. ischigualastensis. The coracoid is plate-shaped and broader craniocaudally than dorsoventrally, resembling that of H. ischigualastensis and San. gordilloi. The humeral shaft is nearly straight in cranial view, but sinuous in lateral view (Figs. 6B–6C). As in H. ischigualastensis, a prominent finger-shaped medial tuberosity is present in its proximal end, separated from the head by a deep trough. At the distal end, both ulnar and radial condyles are well developed, with prominent medial and lateral epicondyles. The ulna bears a well-developed olecranon process (Fig. 6D), as in H. ischigualastensis and Saturnalia tupiniquim. Although such a large process is also seen in other dinosaurs, it is unusual in early members of the group. The medial surface of the proximal end of the ulna is convex, resembling the condition in H. ischigualastensis, whereas it is slightly concave in San. gordilloi. The ulna-ulnare articulation has a strong concave-convex arrangement, similar to that of H. ischigualastensis, and the centrale is preserved distal to the radiale, in a configuration seemingly unique for herrerasaurids. In other dinosaurs, the centrale is located distal to either the ulnare or the intermedium (Sereno, 1993). The manus bears five digits and is elongated, more than 50% the length of the humerus. Digit and metacarpal III are longer than the others (Fig. 6E). As in H. ischigualastensis, manual digit IV is reduced relative to first three digits (digit and metacarpal IV measure 40 mm in proximodistal length, contrasting with 160 mm for digit and metacarpal III). There is no preserved evidence of digit V or its corresponding metacarpal.

The ilium (Fig. 7A) is proportionally craniocaudally short compared to the condition of other coeval non-herrerasaurid dinosaurs. In dorsal view, the cranial and caudal portions of the iliac blade are thickened, whereas the midportion of the blade is lateromedially thin, showing a concavity that faces laterally. The preacetabular process tapers cranially forming a sharp point, differing from the rounded shape observed in H. ischigualastensis and St. pricei. Though cranially projected, the preacetabular area does not surpass the cranialmost extent of the pubic peduncle. The postacetabular process is short, bearing a horizontal furrow in its ventral margin, which is similar to the condition of H. ischigualastensis, neither presenting the deep pockets seen in neotheropods. Indeed, this furrow is reduced in G. cabreirai, even if compared with those of other herrerasaurids. The stout pubic peduncle is cranioventrally directed and extends beyond the preacetabular process. The acetabulum is rounded with a laterally projected supracetabular crest in its craniodorsal margin and a well-developed antitrochanter. The shaft of the pubis is strongly bent caudally, resulting in a vertical orientation (Fig. 7A), distinct from San. gordilloi. The lateral surface of the pubis is sinuous as in H. ischigualastensis and San. gordilloi. Conversely, the pubic shaft of St. pricei is straight in cranial view. Distally, the pubis has a craniocaudal expansion as in other herrerasaurids, although G. cabreirai lacks the distal bevel present in St. pricei. This is because the caudal part of the pubic foot is craniocaudally expanded in the new species as in H. ischigualastensis and San. gordilloi. In St. pricei the medial portion of the distal end is proximally deflected, forming the proximal wall of the bevelled area (Bittencourt & Kellner, 2009). The ischium of G. cabreirai bears a large foramen in the obturator plate (Fig. 7A). This is a unique feature of the new species among herrerasaurids. However, a similar trait is observed in the lagerpetid Ixalerpeton polesinensis (Cabreira et al., 2016). Similar to St. pricei, G. cabreirai has a shallow excavation distal to the caudoproximal process of the ischium (caudal ischial excavation).

The femur (Fig. 7C) of the new species is sigmoid in lateral and cranial views and differs from that of other herrerasaurids in the absence of the caudomedial tuber on the proximal end (Fig. 7D). The craniomedial tuber is smooth as in St. pricei and less pronounced than in H. ischigualastensis. The proximal outline of the femur is kidney-shaped, with a shallow groove on the proximal articular surface. The proximal tip of the cranial trochanter does not form a wing-like projection and the distal portion is connected to a well-developed trochanteric shelf. The dorsolateral trochanter is rounded and extends along the anterior surface of the proximal portion of the bone. The fourth trochanter is aliform and subrectangular in lateral/medial view. It projects caudally and is located proximal to the midpoint of the femur, similar to that of H. ischigualastensis. The tibia (Fig. 7E) is 90% of the length of the femur, as in H. ischigualastensis and San. gordilloi, whereas in St. pricei the tibia is longer than the femur. Distal to the cnemial crest, the cranial margin of the tibial shaft is concave in lateral view, as in H. ischigualastensis, but unlike the straight margin of San. gordilloi and St. pricei. The proximal end of the fibula is slightly flattened lateromedially and the distal end is rounded, with caudomedial and caudolateral prominences, as in H. ischigualastensis. The pes (Fig. 7F) of the new species has three phalanges in digit V, a condition that differs from the single phalanx seen in the pedal digit V of H. ischigualastensis.

Phylogenetic position: A phylogenetic analysis recovered 90 most parsimonious trees (MPTs) of 891 steps each, with a consistency index of 0.334 and a retention index of 0.663. In all the MPTs, G. cabreirai nests within Herrerasauridae (Fig. 8A), closer to the Argentinean taxa H. ischigualastensis and San. gordilloi than to the Brazilian St. pricei. This less inclusive clade is supported by the distal portion of the neural spine of the trunk vertebrae with a lateromedial expansion [92 (0 → 1)], a subtriangular outline of the distal end of the ischium [169 (1 → 2)], and the femur longer than the tibia [198 (0 → 1)]. Herrerasauridae nests external to the Theropoda/Sauropodomorpha dichotomy, as is also the case of some other saurischians (e.g., Daemonosaurus chauliodus; Eodromaeus murphi; Tawa hallae).