Combat Lesion in Triceratops: Histological and Chemical Diagnosis

In the collective imagination derived from scientic and popular literature, Triceratops often faced each other in combat. Thus, from the second half of the twentieth century, these ceratopsids were described as pugnacious animals. This arises primarily from the interpretation of extracranial fenestrae in ceratopsids being the result of combat trauma. However, the diagnosis of the traumatic nature of these anatomical variants of their neck frill requires evidence of bone healing and remodelling by microscopy analysis. Here, we present the case of the Triceratops horridus known as Big John, which is one of the largest specimens discovered in the Hell Creek Formation (Upper Cretaceous; MT, USA). Its right squamosal bone shows an extrafenestra with irregular margins and signs of inammation. Microscopy analysis revealed newly formed and healing bone, with histological signs typical of the bone remodelling phase. Chemical analysis revealed sulphur that was derived from glucosaminoglycans and sulphated glycoproteins of the preosseous osteoid substance present in the healing phases of a bone trauma. Histological and microanalytical analyses conrm that the squamosal fenestra of Big John is the result of a traumatic event, which might indeed have occurred during combat with another Triceratops.


Abstract
In the collective imagination derived from scienti c and popular literature, Triceratops often faced each other in combat. Thus, from the second half of the twentieth century, these ceratopsids were described as pugnacious animals. This arises primarily from the interpretation of extracranial fenestrae in ceratopsids being the result of combat trauma. However, the diagnosis of the traumatic nature of these anatomical variants of their neck frill requires evidence of bone healing and remodelling by microscopy analysis. Here, we present the case of the Triceratops horridus known as Big John, which is one of the largest specimens discovered in the Hell Creek Formation (Upper Cretaceous; MT, USA). Its right squamosal bone shows an extrafenestra with irregular margins and signs of in ammation. Microscopy analysis revealed newly formed and healing bone, with histological signs typical of the bone remodelling phase. Chemical analysis revealed sulphur that was derived from glucosaminoglycans and sulphated glycoproteins of the preosseous osteoid substance present in the healing phases of a bone trauma. Histological and microanalytical analyses con rm that the squamosal fenestra of Big John is the result of a traumatic event, which might indeed have occurred during combat with another Triceratops.

Background
The Late Cretaceous dinosaur Triceratops was characterised by its large neck frill, which predominantly consisted of the hyperelongated parietal and squamosal bones. It also had two large supraorbital horns and a smaller nasal horn. This cranial ornamentation of Triceratops probably served several functions. The evolutionary predecessors of ceratopsids had a relatively developed frill, while the horns, if present, were reduced to simple protuberances. The primitive function of the frill, therefore, was that of visual display and/or species recognition 1 .
The development of the solid frill and the long supraorbital horns during the evolution of ceratopsids such as Triceratops suggests that their function was to protect the cervical region of the skull from blows in icted by individuals of the same species. Farke and colleagues stated that in Triceratops the horns not only had a visual function, but were also used in intraspeci c combat 2,3 . Therefore, the cranial ornaments in Triceratops were used not only as a display, but also as a means of attack and defence against conspeci cs 3 .
The specimen of Triceratops horridus known as Big John (due to its large size) was discovered in 2014 in the Upper Cretaceous Hell Creek Formation (MT, USA), from where numerous remains of Ceratopsidae (Chasmosaurinae) have been recovered 4,5,6 . Big John has a fenestra on the right squamosal that completely perforates the bone (Fig. 1a). The presence of squamosal fenestrae has been documented for the chasmosaurines, which are usually unilateral and with variable shapes, from oval to circular. Their aetiology stimulated a debate that generated two hypotheses. The rst hypothesis claims that squamosal fenestrae are of traumatic origin, probably as the result of intraspeci c ghting or attacks by predators, such as Tyrannosaurus 7,8 . The second hypothesis argues that the squamosal fenestrae are not pathological, but are the result of bone resorption related to aging, or of bone removal for bone that was no longer necessary from a biomechanical and structural point of view 9 .
Macroscopic evidence of associated traumatic injuries, such as bone calli, radial fracture lines and displacement of bone fragments, should contribute to our understanding of whether these fenestrae are indeed traumatic lesions or are the outcome of other pathologies (e.g., infectious diseases). Histological analysis of the affected areas is a fundamental tool for ascertaining the pathological nature of such lesions, to potentially exclude the fenestrae as taphonomic lesions or an anatomical variant linked to aging.
In modern mammals and reptiles, post-traumatic bone remodelling involves recruitment of osteogenic cells and activation of the molecular and chemical mechanisms that control the formation of soft calli that are rich in procollagen and proteoglycans, and their subsequent calci cation 10,11 . The composition and distribution of the chemical elements in the healing bone also changes during the healing process.
Consequently, the study of these factors can also provide useful information for diagnosis of the possible traumatic nature of these lesions.

Materials And Methods
The histological sample of Triceratops horridus known as Big John is housed at Zoic S.r.l. (Trieste, Italy).
The microscopy and histological analyses were performed on a bone sample taken from the inferior Both the sample and the histological sections were examined under electronic scanning eld emission microscope (GeminiSEM 500; Zeiss) equipped with a microanalysis system (AZtecLIVE; EDS OXFORD) and a detector (Ultim Max 100). The analyses were carried out in 'analytical' mode (15kV; working distance, 8.5 mm; variable pressure); this allowed analysis of the samples directly, without any preliminary treatments with glues or surface metallisation.

Macroscopic And Microscopic Analyses Of The Squamosal Fenestra Of Big John
The fenestra on the right squamosal bone of Big John is shaped like a 'keyhole', with the widest part positioned caudally. The rostrocaudal length of the squamosal bone is 112 cm; the distance of the caudal border of the fenestra from the caudal apex of the squamosal is 75 cm; its rostrocaudal diameter and mediolateral width are 20 cm and 5 cm, respectively.
The lumen of this fenestra is partially lled with a ring of brownish material, which is particularly evident on the caudal and inferior edges (Fig. 1b). The bone surface around the fenestra is irregular and characterised by longitudinal cavitations and striations, accompanied by irregular thickening and plaques that resulted from periostitis; i.e., a non-speci c in ammatory process, which can be triggered by different stimuli (e.g., trauma, infection) 12 .
The histological analyses of a sample taken from the lower margin of the fenestra (Supplementary Data  Fig. 2). At high magni cation (400×), the border shows areas of bone resorption, the morphology and dimensions of which are compatible with Howship lacunae, which are seen as eroded depressions of the bone surface due to the activity of osteoclasts (Howship lacunae mean diameter, 19.38 ±2.25 μm) (Fig. 2b). Conversly, the bone tissue distant from the fenestra is characterised by its compact structure, with little vascularisation, and with the seondary osteons homogeneous in shape and size (Fig. 3c).
The dominant chemical elements in the bone sample seen under the electronic scanning eld emission microscope equipped with a microanalysis system were oxygen (O), calcium (Ca) and phosphorus (P) ( Supplementary Data Figs. 3, 4). The other elements seen were iron (Fe), sulphur (S), potassium (K), sodium (Na) and aluminium (Al), in smaller quantities and variously distributed, as well as traces of barium (Ba), strontium (Sr) and manganese (Mn). By focusing the microanalysis on the margin of the lesion, the presence of S was noted in the area that delimits the fenestra, with a distribution superimposable on that of Ca. The S is almost completely absent in the bone tissue distant from the lesion, where P, O and Na are the predominant elements (Fig. 4).

Discussion
Macroscopy analysies of the bone surface that delimits the fenestra on the right sqamosal bone of Big John show alterations that are consequent to in ammatory processes 12 . Histological and chemical analyses of the bone sample from the margin of the fenestra demonstrate that it consists of newly formed bone tissue that appears porous and disorganised, with numerous vascular canals and osteocytic lacunae (Fig. 3a), where layers of bone matrix that were undergoing mineralisation tend to ll the opening ( Fig. 3b). At higher magni cation, the margin of the bone tissue facing the lumen of the fenestra shows resorption areas, with morphology and dimensions that correspond to the Howship lacunae that are characteristic of the remodelling of human bone tissue (Fig. 2b) 13 . All of these histological characteristics are compatible with previously metabolically active and remodelling bone.
The speci c cellular and molecular processes that occur during bone remodelling in modern reptiles and dinosaurs are not known in detail. Therefore, the reference interpretative models here are the physiology and biochemistry of healing bone tissue in mammals (including humans). Bone remodelling begins with a phase of bone resorption by osteoclasts. Through a combination of acid dissolution of the mineralised matrix and digestion of the organic matrix by proteolytic enzymes, small areas of bone resorption are formed (the so-called Howship lacunae); these are concave in shape and a few microns in diameter 12,13 .
The bone resorption is followed by a phase of apposition of osteoid substance by the osteoblasts, which undergoes mineralisation through precipitation of calcium phosphate.
The main constituents of the mineralised bone matrix in new bone are Ca, O and P (which are typically present in calcium hydroxyapatite), and also S (Fig. 4). The phases of bone resorption and neoformation are separated by a transition phase, during which macrophage-like mononuclear cells deposit an initial layer of bone matrix that forms the cementing line that delimits the osteons 14 . This cementing line is rich in sulphated proteins 15 , which explains the presence of S among the elements that constitute bone during the mineralisation phase.
The distribution of S in repaired bone tissue might also be linked to glucosaminoglycans and glycoproteins in the osteoid substance, which is the preosseous substance found in ossi cation centres.
Glucosaminoglycans contain sulphate and are essential for osteogenesis 16,17,18 . Baylink and colleagues demonstrated that the concentration of glucosaminoglycans and S are elevated in the osteoid substance, and then are reduced during the mineralisation process, as a result of enzymatic digestion 10 . The results of the studies carried out by Baylink were con rmed by Takagi and colleagues, who showed a reduction in S in areas undergoing mineralisation and in the calci ed matrix 19 . The removal of glycosaminoglycans and their sulphate compounds might be the prerequisite to the start of the calci cation process of the osteoid substance. The SEM images in the present study showed the mineralisation front of the bone tissue in the remodelling phase, whereby the distribution map of the elements highlighted the presence of S in the areas that had still been occupied by the osteoid substance that had not yet calci ed (Fig. 4).
The bone that was more distant from the margins of the lesion was mainly made up of bone matrix with few vascular channels and reduced numbers of osteocytic lacunae. These are histological characteristics that are compatible with compact bone tissue that has completed or almost completed the remodelling phase. In this region, moreover, there were lower amounts of S, which con rms the hypothesis of a gradual reduction in S during deposition of hydroxyapatite in the osteoid substance.
The edge of the lesion thus appeared to be made up of repaired bone tissue that had been undergoing mineralisation, as seen for both the light and dark areas of the histological preparations. Conversely, the innermost region that was more distant from the fenestra showed well mineralised bone or bone that was in an advanced state of mineralisation.
The results of the histological and chemical analyses show that the bone that circumscribed and partially lled the lesion had been made up of metabolically active and remodelling tissue. The fenestra is therefore of traumatic origin, and at the time of the death of Big John, the lesion was still healing.
The presence of newly formed bone and the histological characteristics of bone remodelling that was in progress excludes the possibilities that the lesion occurred post mortem (i.e., that it is of a taphonomic nature) and that the fenestra represented an anatomical variant. This conclusion is supported by the lack of regular, sharp and blunt margins for the lesion. Furthermore, although congenital squamosal fenestrae have been described for the subfamily of Chamsmosaurinae, they have not been found in Triceratops or Anchiceratops to date 9 . Instead, the lesion was possibly caused by the horn of another Triceratops. The size and shape of the caudal region of the fenestra coincide with those of the apex of the supraorbital horns. Furthermore, the triceratops used their horns in intraspeci c combat 3 .
Farke hypothesised three models of engagement in the intraspeci c combat of Triceratops 2 . The "single horn contact position" de nes the contact that occurs between only one postorbital horn of each combatant, where lesions of the squamosal bone are in icted by the nasal horn. In the "oblique horn locking position", the skulls come into contact when they are slightly rotated with respect to each other, and contact occurs between both postorbital horns of both individuals. This model predicts lesions in icted by the nasal horn in the rostral portion of both the parietals and squamosals. The third model of engagement is de ned as "full horn locking", which consists of the collision between the skulls when tilted at about 45° with respect to the horizontal plane, and involves contact between both postorbital horns of the opponents. This combat mode will result in injuries in the region of the temporal fenestra. The location of the lesion analysed here on the squamosal bone of Big John does not respect these predictive models of Farke for horn use in Triceratops 2 . It appears likely that the wound was instead in icted from behind Big John (Supplementary Data Fig. 5), whereby the rival's horn would have penetrated the frill and then slipped towards the rostrum, giving this lesion the shape of a keyhole. In describing his models on the trauma in icted by the horns in the ghting between Triceratops, Farke did not exclude that there might also be other engagement dynamics besides those he hypothesised 2 .
Big John appears to have survived this trauma for some time. Two cases have been described in the literature where adult specimens of Centrosaurus survived extensive traumatic lesions to the skull 9 . However, it remains di cult to establish how long Big John survived following this trauma. Although the bone remodelling process can already begin in the third to fourth weeks after the trauma 11 , the healing times vary according to the species and to the extent of the trauma. Considering the healing times of traumatic injuries in modern reptiles, along with the size of the traumatic injury and the amount of bone repair, it is likely that the death of Big John occurred at least 6 months after this traumatic event 20,21 . This study con rms the existence of intraspeci c ghting in Triceratops. Furthermore, although the physiological and cellular mechanisms underlying the healing process in dinosaurs are still largely unknown, it would appear to be similar to those described in humans and mammals. Further histological and microanalytical investigations on fossil remains with traumatic lesions might shed light on the bone physiopathology of these reptiles. Figure 1 The skull of Big John during the restoration. (a) The fenestra analysed here is indicated by the white circle. (b) Detail of the fenestra. The region where the sample to be analysed was taken is visible on the lower margin.