Ceratosaur palaeobiology: new insights on evolution and ecology of the southern rulers

Ceratosaur theropods ruled the Southern Hemisphere until the end of the Late Cretaceous. However, their origin was earlier, during the Early Jurassic, a fact which allowed the group to reach great morphological diversity. The body plans of the two main branches (Noasauridae and new name Etrigansauria: Ceratosauridae + Abelisauridae) are quite different; nevertheless, they are sister taxa. Abelisaurids have lost the ability to grasp in the most derived taxa, but the reduced forelimb might have had some display function. The ontogenetic changes are well known in Limusaurus which lost all their teeth and probably changed the dietary preference at maturity. The results presented here suggest that abelisaurids had different soft tissues on the skull. These tissues might have been associated with evolution of a strong cervicocephalic complex and should have allowed derived taxa (e.g. Majungasaurus and Carnotaurus) to have low-displacement headbutting matches. The ability to live in different semi-arid environment plus high morphological disparity allowed the ceratosaurs to become an evolutionary success.


Results and Discussion
Phylogenetic relationships. Ceratosauria traditionally consists of Ceratosaurus and all taxa closer to it than to Neornithes 19 . However, taxonomy within Ceratosauria has been complicated. Abelisaurs were formally known as Abelisauroidea (=Ceratosauroidea), that comprises Carnotaurus, Noasaurus and all their most recent common ancestors and all descendants (see below for further discussion). Ceratosauroidea are included in the clade called Averostra which comprises the taxa related to Ceratosauria and all derived theropods 20 . Approximately 32 Ceratosauroidea genera are currently known with most of the taxa originating from the Late Cretaceous (Table S1).
SCIeNtIFIC RePoRTs | (2018) 8:9730 | DOI: 10.1038/s41598-018-28154-x ilium of Noasaurinae (subfamily included in Noasauridae) is as low as in Carnotaurinae (subfamily included in Abelisauridae) despite the fact that these two groups are not closely related. The skull of noasaurids are long and low compared to those of abelisaurids 17,33 . Interestingly, even among noasaurids the morphology of the skull varies substantially. The skull of Limusaurus becomes toothless through ontogeny, likely to meet a change in diet (see below) 14 , whereas the skull of Masiakasaurus knopfleri presents strong procumbent dentitions which probably indicate additional divergence from the typical theropod diet 35 . The forearms of noasaurids are poorly known, but as in other ceratosauroids the humerus, radius and ulna are more reduced distally than proximally suggesting that the reduction may have occurred in a modular fashion, from the distal to proximal across the phylogeny 12 . However, the humeri of noasaurids are slenderer than those of abelisaurids (Fig. 3A).
The body plan of etrigansaurians strongly differs from other theropods, and their morphology is more thoroughly known than that of noasaurids 1,4 . Whereas the noasaurids have long skulls, the etrigansaurians have strong and deep skulls, especially those of Brachyrostra which also showed encroachment of the postorbital into the orbit, just beneath the eye 22 . The skull of abelisaurids became shorter and more rugose in more derived taxa. Ceratosaurus, Eoabelisaurus, and possibly Genyodectes have longer skulls compared to those of abelisaurids. The skull's shortening and deepening started in abelisaurid basal forms, such as the Aptian-Albian Kryptops palaios and the Cenomanian Rugops primus, both from Niger 36,37 , and reached its extremity in the Carnotaurinae taxa. The skull of Carnotaurus is exaggeratedly short and deep compared with those other taxa of the same clade. The skull of Abelisaurus was largely reconstructed in the snout as well as in the posterior area 1,3,38 , and taphonomic  15 . The phylogenetic position of Chenanisaurus is from Longrich et al. 24  distortion has modified the proportions and several contacts between elements are missing such as the jugal articulations 3,38 . Therefore, as previously suggested 38 , Abelisaurus should have had a shorter skull than was previously reconstructed and frequently reproduced resembling those of Carnotaurinae (e.g. Majungasaurus) instead of Ceratosaurus (as suggested by Bonaparte and Novas 27 ).
Regarding the basal abelisaurids, Kryptops was diagnosed based on a left maxilla, several partial vertebrae and ribs and an articulated pelvic girdle and sacrum 36 . However, as noted by Novas et al. 4 and Carrano et al. 39 , the pelvic gridle and sacrum of Kryptops were found "eroded and free of the rock some 15 meters distant" and have more shared features with tetanurans than abelisaurids. The vertebral non-sacral remains also share features with ceratosaurians as well as tetanurans 36 . The maxilla is also incomplete and with only a general diagnosis possible (e.g. external texture on the maxilla, which is composed of short linear grooves that are also shared with Majungasaurus and Rugops). The only autapomorphy is a secondary wall in the anteroventral corner of the antorbital fossa obscuring it and that has a scalloped and fluted dorsal margin 36 . Therefore, as the holotype of Kryptops is a miscellany of materials belonging to different groups with just one autapomorphy supporting the species, this taxon might have been considered as nomen dubium rather than a valid taxon. The postcranial skeleton probably has a phylogenetic relationship with carcharodontosaurids instead of abelisaurids as suggested by Novas et al. 4 and Carrano et al. 39 .
Abelisaurids has strongly reduced forearms without grasping ability 40 (Fig. 3B). According to Agnolin and Chiarelli 40 , abelisaurs probably also lacked forearm mobility. However, recent analyses on Majungasaurus musculature suggest that, although much reduced, abelisaurids did not lose full mobility of the forelimb, and may have used it for intraspecific display 41 . Some taxa such as Aucasaurus, Majungasaurus and Carnotaurus may have lost the ungual of the digits I and IV 31,40,42 whereas the ceratosaurid Eoabelisaurus has strongly reduced the manual unguals 12 . The digit IV is fused to the metacarpal in Majungasaurus and Aucasaurus precluding mobility. Extreme reduction also reduced autonomy of all digits due to the extreme reduction, although the hemispherical humeral head and distal radius and ulna suggests that the shoulder and the wrist had a large range of motion 38,41 . However, as pointed by Gianechini et al. 38 the range of motion of the humerus should have been higher in lateromedially (i.e. abduction-adduction) than in anteroposteriorly (i.e. flexion-extension) because the development of the dorsal and ventral rim of the glenoid fossa reduced anteroposteriorly movements. Also, is worth noting that the large scapulocoracoids and reduced forelimbs in ceratosauroids might be related to a close developmental association between scapular blade and the axial skeleton, holding the shoulder girdle to the axial skeleton and for mobility of the girdle and the ribcage 38,41,43 . Those muscles attached to the neck could have had an important role in feeding as in extant crocodiles (e.g. muscle levator scapulae which is an effective abductor of the neck and hence the head) 41,44 .
The hindlimbs of ceratosaurs are different in the two main branches. In noasaurids, the hindlimbs are more slender than the etrigansaurians; however this is due to the overall size of individuals of the groups 1 . Abelisaurids' hindlimbs and caudal vertebrae suggest that these taxa, specially the brachyrostrans, may have had powerful cursorial abilities. The tibia have well developed dorsal anterior projection (cnemial crest) onto which the main knee extensor muscles are inserted (i.e. iliotibiales) 45 . The large size of the cnemial crest and its dorsal inclination suggest that some ankle extensors and digital flexors muscles were large, increasing their force-producing capability. Additionally, the dorsal inclination of the transverse processes in the caudal vertebrae suggests that the muscle caudofemoralis longus, the main femur extensor, may have been larger than in other theropods contributing to the cursorial ability 10 . Also, the presence of accessory articulations in caudal vertebrae (hyposphene-hypantrum) apart of the inclined transverse processes, increases the tail rigidity 10,46 and may have enhanced overall speed and acceleration 10 . However, acceleration might have been more impressive than top speed. When preserved, feet of some abelisaurids are short (e.g. Majungasaurus 47 ), indicating low tangential velocity at the ankle. The type of Carnotaurus lacks feet and the distal portion of the epipodials, even though it is often reconstructed as having gracile legs and feet 17 .
Etrigansaurian soft tissue. The etrigansaurians also are well known by their rugosities and projections from the skull elements 3 . Carcharodontosaurid theropods have rugosities in lateral skull bones as well, but the morphology is different 48 and leads to misinterpretations of the group 49 . Although abelisaurids have strong rugose skulls, the textures are variable throughout the skull 48 . The texturization of the skull happened independently from the projections. For example, the skull of Ceratosaurus is diagnosed by having a rounded midline horn core on the fused nasals 3 and horn cores forming a dorsal crest on the lacrimals 50 , although the skull is otherwise smooth 48 . On the other hand, the skull of Skorpiovenator bustingorryi is strongly texturized but without any projections 22 . The skull roof in abelisaurids is thick but this feature varies among the species 48 . Both majungasaurini Majungasaurus and Rajasaurus normandensis have a single medial horn formed by the frontal and frontal/nasal, respectively 28,48 ,  whereas the brachyrostran Carnotaurus has two frontal horns laterally oriented 17 , Aucasaurus has the lateral margins of frontal elevated in the orbital region, and Viavenator exxoni has almost flattened frontals 51 . The flattened frontals of Ekrixinatosaurus novasi 52 and probably of Skorpiovenator suggest the basal position of these two taxa in relation to Furileusaura as proposed by Filippi et al. 15 .
The rugosities in abelisaurids resulted from a mineralization processes with specializations in the overlying dermis, such that the mineralized tissue includes the irregular surface texture representing mineralization of the bone's periosteum, overlying dermal fibers or combination of the two, characterizing the metaplastic ossification 48 . The sculpture of lateral bones (e.g. maxilla, jugal, quadratojugal, dentary) presents a higher percentage of tangential vascular canals and grooves, whereas the dorsal roofing elements (e.g. frontal, dorsal postorbital and lacrimal, nasal, nasal process of the premaxilla) tend to have more projecting, tuberculate and/or cauliflower-like texture that combine with the vascular canals and grooves (Figs 4 and 5A,B) 48 . Sampson and Witmer 48 have suggested that abelisaurids might have had more robust skulls than other theropods due to the high skull's mineralization. Following the results of Hieronymus et al. 53 for inference of soft tissues in Centrosaurine and Carr et al. 54 for Tyrannosauridae, it is possible to assess the superficial cranial soft tissues of abelisaurids. These tissues show a hierarchy of textures which became more complex towards the phylogeny.
The basal abelisaurid Rugops has the dorsal surface of nasals with a row of seven pits, visible sutures between then and hummocky rugose surface which is also present in the dorsal surface of frontal, prefrontal lacrimal and maxilla (Figs 4A and 5C). These features are correlated with overlying scales as observed in living crocodiles and reptiles 53 . On the other hand, the anterior-most snout has a different texture compared to other categories of soft tissue. The nasal articulation processes of premaxilla and the anterior processes of nasal, show a papillate texture indicating the presence of armour-like dermis as suggested by Hieronymus et al. 53 . The presence of these tissues suggests that Rugops had, at least two categories of tissues covering the surface of the skull. Interestingly, the type of Rugops could be a subadult individual due to its small size, incomplete fusion between the nasals and the presence of the fenestra between the prefrontal, frontal, postorbital and lacrimal 3 . As the rugosities tend to increase during ontogeny 18 , the armour-like dermis could reach a larger surface if Rugops grew up and developed more papillate texture.
Abelisaurus, as other abelisaurids, have a lateral cranium surface (e.g. maxilla) with dense tangentially arranged grooves suggesting it was covered by large scales or scutes, as suggested by Sampson and Witmer 48 and Hieronymus et al. 53 (Figs 4B and 5D). However, the nasal of Abelisaurus differs from that of Rugops being extremely rugose with bones lobules across its surface. This texture is associated with armour-like dermis 53 , as seen in the anterior snout of Rugops (Fig. 5C) The dorsal surface of carnotaurine skulls (nasal, frontal, dorsal lacrimal and dorsal postorbital) have coarse pitting and grooving on bone surfaces suggesting that these were covered by cornified tissue, being an osteological correlate with the cornified cover seen on muskoxen, centrosaurine dinosaurs 53 , and tyrannosaurids 54 (Figs 4C  and 5A,B). However, it is improbable that abelisaurids had projections higher than the frontal horns. This category of tissue increased the toughness of the head roof, which also might have had an important ecological function as discussed below.
The horns of Carnotaurus and Ceratosaurus would have been more extended than the preserved fossil and covered with cornified sheath, indicated by neurovascular grooves, depressed lip and less rugosity than the other bones surfaces as suggested by the results of Hieronymus et al. 53 (Fig. 5E and F). Although the horn cores of Carnotaurus are more rugose than those of Ceratosaurus, ventral to the depressed lip the frontals are markedly lesser rugose. The single horn of Majungasaurus and Rajasaurus do not have the depressed lip seen in Carnotaurus and Ceratosaurus, suggesting that they were covered by cornified tissue without dorsal extension.
The only preserved soft tissues so far belongs to Carnotaurus and correspond to the anterior cervical region associated with cervical ribs, the shoulder region, thorax and tail 17 . The skin impressions present conical protuberances and there is no evidence for filaments or feathers (Fig. 3E). So far, the tubular filaments and feathers are only known in tetanuran theropods 55,56 .
Regarding the bone histology, some analyses also shed some light to the development of ceratosaurs as well as palaeoenvironment 14,[57][58][59] . For example, the robustness of Masiakasaurus, once believed as different morphs (robust and gracile) 60 , might be considered to be developmental feature instead of dimorphism 57 , as also shown in allometric analyses 1 . Additionally, the slow growth of the same species can be related to the low resources of Maevarano Formation 57,61 . The ontogenetic series of Ceratosaurus is still unclear. Madsen and Welles 50 described two different species of Ceratosaurus (C. magnicornis and C. dentisulcatus) based on cranial and post-cranial associated elements. Nevertheless, Rauhut 64 suggests that the diagnosis of these species are subjective and there might have been just one species of Ceratosaurus in Morrison Formation. Carrano and Sampson 3 , following Rauhut 64 , also argued that these two species have size-based diagnosis suggesting that they might be different ontogenetic specimens from Ceratosaurus. Although there are other materials attributed to Ceratosaurus 3 , no study was conducted to discuss the ontogenetic traits so far.
The series of Limusaurus shows at least 78 ontogenetic modifications through the growth from the analyses of 19 specimens 14 . Delcourt 30 reported the loss of teeth in mature individuals, while most juveniles had toothed jaws, the skull also becomes longer through ontogeny. In a parallel and broader study, Wang et al. 14 also reported several changes including the formation of a beak after birth. The amount of modifications in Limusaurus ontogeny and the presence of gastroliths in the abdominal region also suggest that this species change ontogenetically dietary preferences from omnivory to herbivory 14,32 .
The ontogeny of Majungasaurus was assessed by Ratsimbaholison et al. 18 using mainly landmark-based approaches in the skull and in some isolated cranial elements (premaxilla, maxilla, lacrimal, postorbital, jugal, quadrate, dentary and surangular). The authors suggested that the ontogenetic changes include: the skull becomes deeper, the orbit becomes smaller, the sutures among the bones become more complex, and the texture of lateral bones increase 18 . In this study, the postcranial elements were not assessed.
Histological analyses suggest that Masiakasaurus 57 and small abelisaurid theropods 58 had a cyclical growth strategy as well as slowdown growing. However, in larger taxa, such as Aucasaurus, the growth rate tend to be higher than in smaller forms 58 .
Apart from these studies, some inferences about ontogenetic stages were made based on fusion of bones. For example the types of Xenotarsosaurus bonapartei, Eoabelisaurus, and Aucasaurus are considered mature individuals because they have a fused tibia and astragalus 12,42 (Fig. 3C and D), whereas the type of Rugops has been suggested as being an immature individua based on the fusion of the cranial elements 3 (see above). The type of Pycnonemosaurus nevesi, despite being considered the largest abelisaurid so far 1 , is considered a subadult specimen 13 based on the presence of caudal vertebrae with unfused arches and centra as well as tibia. However, determining the maturity of a specimen based only on the fusion of arches with centrum is not safe because these elements are size-independent 65 . Ceratosaur behaviour. Ceratosaur behaviour can be inferred from several studies on anatomy 4,40,48 and biomechanics 8,9,66 . Also, the new information on soft tissue presented here (see above), suggest a behavioural pattern in abelisaurids as discussed below.
Gregarious behaviour is difficult to deduce; however small species found associated in the same assemblage localities, such as Masiakasaurus 33 and Limusaurus 14 , suggest that they might have lived together. In the case of Majungasaurus, several specimens were found associated, but some materials (ribs, chevron, neural spines, transverse processes and neural arches) have teeth marks made by its conspecifics suggesting that this species had cannibalistic behaviour 61 . This behaviour can be explained by the resource scarcity in the Maevarano Formation during the Late Cretaceous that was semi-arid 61 .
Going through the new information of soft tissues of abelisaurids shown here (above), it is possible to infer that this clade might have had some intraspecific headbutting matches behaviour at least in carnotaurine taxa (as suggested for Carnotaurus 8 and Majungasaurus 67 ). The presence of cornified cover on the skull, that was inferred for Carnotaurus and Majungasaurus, has been related to headbutting behaviour in extant taxa (e.g. Ovibos moschatus, Syncerus caffer and Buceros vigil) as well as extinct (e.g. Pachyrhinosaurus, Achelousaurus horneri 53 and Stegoceras validum 68 ). Nevertheless, differing from those that engage in violent headbutting and have deep cancellous bone 68 (which carnotaurine lack), the carnotaurine might have used the head in low-motion headbutting and shoving matches at low speeds (as marine iguana Amblyrhynchus cristatus 69 ) or engaged giraffe-like strikes to each other's neck and flanks 67 . The giraffe-like strikes have been proposed for Majungasaurus 67 due to the presence of tall, rugose nasals, struts within sinuses and a unicorn-like projection of the frontals 48,67 , although stresses. Also, the mechanical analyses of Carnotaurus skull performed by Mazzetta et al. 9 support the low-motion headbutting in this taxa. Furthermore, the presence of well-developed occipital region (e.g. nuchal crest) 48 associated with large epipophysis and neural spines in the cervical vertebra increasing the neck musculature 70,71 strongly suggest that the cervicocephalic complex (head and neck) withstood high stress. Indeed, the well-developed epipophyses indicate a good leverage for intervertebral dorsiflexion by the muscle tranversospinalis cervicis and the origin of a strong muscle complexus, a head dorsoflexior 72 . As similar features on neck and skull are spread throughout the carnotaurine abelisaurs, all the taxa belonging to this clade may have had similar behaviour in territoriality or mating matches for instance. It is worth noting that cranio-facial biting was reported for non-avian theropods [73][74][75] . This behaviour could have had several possible reasons, including territoriality, courtship/mating, play, predation/cannibalism, intrapack dominance and subadult dispersal 74 . In the case of carnotaurine, the headbutting and/or giraffe-like strikes could also have been added to the behavioural repertoire for any reasons above.
The low-motion headbutting behaviour also may have been present or began in more basal taxa such as Rugops and Abelisaurus in parallel with the development of scales and armour-like dermis on the dorsal cranium (e.g. nasal). For example, the dorsal surface of marine iguana skull has hummocky rugosities 53 as in Rugops, suggesting that this structure associated with armour-like dermis might have allowed the abelisaurid a similar behaviour (i.e. low-motion headbutting).This hypothesis of low-motion headbutting developing through the phylogeny in abelisaurids can be tested if a species with similar skull showed Rugops hummocky rugosities plus well-developed cervical epipophyses and neural spine and if it was found in Early Cretaceous beds (e.g. Aptian). If the headbutting was not developed in this taxon, certainly the development of armour-like dermis and later cornified cover on the skull in more derived abelisaurids might have allowed for this behaviour. It is worth noting that the giraffe-like strikes seem to be more complex than the iguana-like low-motion headbutting because the first requires more complex development of the skull, as seen in Majungasaurus 67 , than in Rugops. Therefore, carnotaurine could potentially have adopted both combat styles. The possibilities of these behaviours in abelisaurids are testable with quantitative biomechanical methods 8,9,67 and could be assessed in the future.
Biomechanical studies on the skull of abelisaurids have suggested that they had cranial mechanical advantage similar to allosaurs (e.g. Allosaurus fragilis and Carcharodontosaurus saharicus) 66 and similar bite force (e.g. Carnotaurus: 3,341 Newtons 9 ; Allosaurus: 3,573 Newtons 76 ). These results mean that these two groups had high efficient mechanical advantage, but a bite force not as strong as that of Tyrannosaurus 9,66 .
According to the analyses of Therrien et al. 77 , carnotaurines (e.g. Majungasaurus and Carnotaurus) might have been ambush predators attacking large prey. Additionally, Sampson and Witmer 48 have suggested that Majungasaurus, and possibly other carnotaurines, were "adapted for a mode of predation that entailed relatively few, penetrating bites accompanied by powerful neck retraction, as well as bite-and-hold behaviour". This predatory behaviour is consistent with results on skull biomechanics 9,66 as well as neck analyses 69,70 .
The development of advantageous features (e.g. large muscles for cursorial abilities) 10 plus the increase the body size towards the phylogeny 1 granted abelisaurids the opportunity to succeed the carcharodontosaurids as main predators in the Southern Hemisphere after their extinction in Turonian 49,78 . Interestingly, these two groups share dentary 22,49 and skull advantage mechanics 66 that might have helped the extinction of carcharodontosaurids through ecological interactions 1 when this group was becoming rare in the Cenomanian, possibly due to climate changes (i.e. changing in the mean temperatures and floral compositions) 79 . Therefore, it is reasonable to suggest that the latest abelisaurids (carnotaurine) were tyrannoasaurid counterparts since the former were dominant in Southern Hemisphere 3 and the latter in Northern Hemisphere 2 .
SCIeNtIFIC RePoRTs | (2018) 8:9730 | DOI:10.1038/s41598-018-28154-x Ceratosaur biogeography. The new phylogenetic analyses presented by Wang et al. 14 suggest that Ceratosauroidea was present in North America (Ceratosaurus) and Asia (Limusaurus, also suggested by Rauhut and Carrano 24 ), instead just in South America, Europe, Africa, India and Madagascar 4,5 . However, Ceratosauroidea originated in Africa 29 and the taxonomic diversity spread during the Middle Jurassic to North America, Europe, Asia, Africa, South America and Madagascar (Fig. 1). Australia and Antarctica do not have ceratosaur remains so far 4 , nevertheless it is possible that this group was present there and future discoveries can change this scenario.
The division of the mains branches of Ceratosauroidea (Noasauridae and Etrigansauria) happened in the Early Jurassic 14,29 just after the origin of this group. The latest ceratosaurs, from the Aptian 36 , were restricted to Southern Hemisphere and Europe 5 . However, during the Barremian to Santonian Gondwana remained isolated from Laurasia when the fauna could acquire a wide geographic distribution across the southern landmass; relating to Europe in Campanian-Maastrichtian rather than Asiamerica 4,80 . The presence of the European majungasaurini Arcovenator escotae corroborates this biogeographic hypothesis 5 whereas the European noasaurid Genusaurus sisteronis from Aptian 14,81 would have to be considered a relic from the early origin of noasaurids.
It seems the abelisaurids body size increases along the phylogeny 1 ; however, the new phylogenetic analyses presented by Wang et al. 14 suggest a large abelisaur (i.e. Abelisaurus) in the base of the clade. Also, there is a new evidence that abelisaurids reached medium/large sizes (between 5.6 and 7.6 m long, based on a partial tibia) from Berriasian-Valanginian of South America 82 . Nevertheless, the largest species were restricted to South America and Africa so far 1,23,83 . This is because insular environments, such as Late Cretaceous of Europe 5 and Madagascar, supports smaller fauna than continental landmass. Finally, the ability to live in semi-arid palaeoenvironment with low resources, such as those of Majungasaurus and Pycnonemosaurus 61,84 , and the high disparity of the group facilitated the evolutionary success of ceratosaurs during this time (Fig. 6).