Chronic fracture and osteomyelitis in a large‐bodied ornithomimosaur with implications for the identification of unusual endosteal bone in the fossil record

Paleopathological diagnoses provide key information on the macroevolutionary origin of disease as well as behavioral and physiological inferences that are inaccessible via direct observation of extinct organisms. Here we describe the external gross morphology and internal architecture of a pathologic right second metatarsal (MMNS VP‐6332) of a large‐bodied ornithomimid (~432 kg) from the Santonian (Upper Cretaceous) Eutaw Formation in Mississippi, using a combination of X‐ray computed microtomography (microCT) and petrographic histological analyses. X‐ray microCT imaging and histopathologic features are consistent with multiple complete, oblique to comminuted, minimally displaced mid‐diaphyseal cortical fractures that produce a “butterfly” fragment fracture pattern, and secondary osteomyelitis with a bone fistula formation. We interpret this as evidence of blunt force trauma to the foot that could have resulted from intra‐ or interspecific competition or predator–prey interaction, and probably impaired the function of the metatarsal as a weight‐bearing element until the animal's death. Of particular interest is the apparent decoupling of endosteal and periosteal pathological bone deposition in MMNS VP‐6332, which produces transverse sections exhibiting homogenously thick endosteal pathological bone in the absence of localized periosteal reactive bone. These distribution and depositional patterns are used as criteria for ruling out a pathological origin in favor of a reproductive one for unusual endosteal bone in fossil specimens. On the basis of MMNS VP‐6332, we suggest caution in their use to substantiate a medullary bone identification in extinct archosaurians.

Here we report on a pathological second metatarsal of a newly described (Chinzorig et al., 2022), yet unnamed, ornithomimosaur dinosaur from the Santonian Eutaw Formation in Luxapallila Creek, Lowndes County, Mississippi, USA that exhibits abnormal bone growth along most of the diaphysis ( Figure 1). Historically, the study of ancient diseases was investigated primarily through gross morphological observations, which are of limited diagnostic utility. More recently, advances in bone microstructural diagnostic technologies, combining plain or polarized light microscopy with computed tomography, or synchrotron imaging, have become vital tools for investigating pathological processes and have opened a new era of paleopathological research (e.g., Anné et al., 2016;Hedrick et al., 2016;Molnar, 2001;Schultz, 2001). We use complimentary gross morphological observations, microCT imaging, and osteohistological data to diagnose the developmental timeline and etiology of this pathology, and to explore the correspondence between periosteal pathological bone visible in gross inspection and the extent of secondary endosteal bone deposition. The latter has important implications for the study of unusual endosteal tissues in fossil vertebrates, particularly the identification and evolution of reproductive (i.e., medullary) bone in bird-line archosaurs.

| Specimen
MMNS VP-6332 ( Figure 1a) is a nearly complete, isolated right second metatarsal of a theropod dinosaur recovered from a late Santonian interval in the lower Eutaw Formation (Cicimurri et al., 2014) of Lowndes County, Mississippi. The specimen is well-preserved, missing only the proximal articular surface and encased with epoxy to stabilize the poorly preserved region. To determine the taxonomic identity of the pathologic metatarsal MMNS VP-6332, Chinzorig et al. (2022) segmented the reactive pathological bone from the primary cortical bone and generated a model of the original morphology of the bone (Figure 1c), concluding that it was referable to an ornithomimosaurian theropod. Using a series of regressions, Chinzorig et al. (2022) estimated that MMNS VP-6332 stems from a large-bodied ornithomimosaur taxon broadly similar in overall size to Arkansaurus fridayi (380 kg), Beishanlong grandis (375 kg), and Gallimimus bullatus (400 kg) (Chinzorig et al., 2022, in review). All gross anatomical measurements are shown in Table 1. 2.2 | X-ray computed microtomography (microCT) MMNS VP-6332 was scanned with a Nikon XTH 225 ST μCT scanner at a voxel size of 87.5 μm and using 225 kV and 115 μA, and a 1 mm-thick copper filter at the Shared Materials Instrumentation Facility of the Duke University, Durham, NC, USA. Avizo 9 was used for the threedimensional reconstruction and segmentation of the specimen.
F I G U R E 1 Right second metatarsal (MMNS VP-6332). (a) Gross external morphology; (b) 3D reconstruction based on CT data; (c) 3D reconstruction, showing the original bone cortical surface (in blue) and the distribution of the periosteal reactive bone along the shaft (in green); (d) Sagittal-virtual sections demonstrate areas of fracture and distribution of pathologic bone deposition along the endosteal margin. (a, bI, cI, cIII, and dI) in cranial and (bII, cII, and dII) lateral views

| Osteohistological ground-section preparation
Histological processing of MMNS VP-6332 was performed at the Paleontology Research Laboratory, North Carolina Museum of Natural Sciences, Raleigh, NC, USA, using standard paleohistological techniques (Lamm, 2013). In order to assess the relative histovariability and extent of the pathological tissues, the sectioned metatarsal was prepared by cutting transversely through the callus at three different levels along the shaft. This sampling protocol produced a total of three cross-sections, labeled as proximal 1 (p1), distal 1 (d1), and distal 2 (d2) (Figure 1). Cross-section d1 was made at the level of the widest periosteal bone growth found on both gross and microCT examination of the pathologic metatarsal diaphysis. D2 was taken approximately 1 cm distal to d1. The most proximal cross-section p1 was taken from the diaphyseal region exhibiting no obvious external pathology just distal to the proximally preserved end of the metatarsal (Figure 1).
Transverse cuts of the whole bone were made using a Ryobi tile saw (model WS750L). The specimen was then stabilized using a paraloid B-72 (5%) on the cut surface of the pathologic bone. All samples were embedded in and mounted on plexiglass slides by using a clear epoxy resin (EPO-TEK 301). We used a diamond wafer blade mounted on a Buehler Isomet 1,000 low-speed precision saw for cutting the resin-embedded samples along the transversal plane to obtain bone sections prior to polishing. The polishing process (down to a thickness of approximately 50-60 μm) was done on a wheel grinder/ polisher (Hillquist cross-section machine and Buehler Metaserv-250 grinder and polishing machine) using a series of abrasive paper disks with increasing grit sizes (400, 800, 1,200, and a microcloth for polishing stage). All cross-sections were examined through both plane and cross-polarized light using a Nikon ECLIPSE Ci POL microscope. Photomicrographs were captured using both the Nikon ECLIPSE Ci POL with a Prior motorized stage and DS-Ri2 camera, and with a Keyence VHX-7000 digital microscope with an FI (VHX-7100) head and free-angle observation system (VHX-S770E) with a polarizer, and a custom-built lambda filter. The pictures were combined together in NIS-Elements D 5.41.01.64-bit microscope imaging software. All cross-sectional measurements are shown in Table 2.

| Fracture pattern terminology
Butterfly fracture -A classic pattern of bending bone failure that results from three-point bending (a specific type of loading failure in blunt force trauma cases) occurring when a long bone is impacted between the proximal and distal ends. The fracture is formed by an initial transverse fracture on one side of the bone and is followed by oblique fractures on the opposite side of the bone, resulting in the separation of a V-or Y-shaped wedge fragment (known as a butterfly fracture fragment) (Isa et al., 2021;Reber & Simmons, 2015). Two types of butterfly fractures are characterized as tension wedge and compression wedge fractures (Figure 5c).
Tension wedge fracture (i.e., "impact side butterfly fragment") -is created when the initial transverse fracture begins on the opposite side of the blunt force impact (or tension side of the bone), and is followed by oblique fractures that radiate towards the same side as the impact (Isa et al., 2021;Reber & Simmons, 2015) (Figure 5c).
Compression wedge fracture (i.e., "nonimpact side butterfly fragment")is created when the initial transverse fracture begins on the same side as the blunt force impact (or compressed side of the bone), and is followed by oblique fractures that radiate away from the impact side. This fracture pattern is essentially a 180 rotation of a tension wedge fragment (Isa et al., 2021;Reber & Simmons, 2015) (Figure 5c). Combined tension and compression wedge fractureoccurs in some cases of blunt force trauma when the impact occurs on the tension and compressed sides of the bone, thus resulting in the creation of both tension and compression wedge fragments (Isa et al., 2021;Reber & Simmons, 2015) (Figure 5c).

| Osteopathological analysis
Gross anatomical examination-The diaphysis of MMNS VP-6332 exhibits irregularly-shaped pathologic expansion of the normal diaphyseal circumference and periosteal reactive bone growth that extends from the proximal to distal regions (Figure 1a,b; Table 1). This periosteal reactive bone growth and expansion begins just distal to the preserved proximal aspect (a portion of the proximalmost end, including the articular aspect, is damaged) and terminates near the distal end. There is a predominant, and more severe, craniomedial distribution of periosteal reactive bone deposition along the proximal and middle diaphysis that transitions to a caudomedial and less severe, caudal to caudolateral distribution distally ( Figure 1c). Most of the total length of the pathologic diaphysis (from just proximal to the mid-diaphyseal region distally) is characterized by an undulated and rough periosteal surface (Figure 1b). Only the proximal third and the cranial and craniomedial aspects of the distal third of the pathologic diaphysis have a smooth periosteal surface covering the diaphyseal expansion.
Sagittal and cross-sectional views of the microCT imaging-Within the proximal region of the diaphysis, there is increased radiolucency (porosity) of the cortex, extending from the cortical-endosteal junction across approximately two-thirds of the total cortical width that is characterized by variably sized, frequently medium to large-sized, irregularly round to oval-shaped cavities ( Figure 2a). Within the region of the mid-diaphysis, there are multiple complete, oblique to comminuted fractures in the cranial and caudal cortices observed on sagittal and cross-sectional microCT imaging ( Figure 1dII). The oblique fracture lines join together between the proximal and distal fracture ends to create variably sized, irregular triangular-shaped fracture fragments with minimal displacement (Figure 1dII). Most of these fracture fragments have the long base of the wedge facing the periosteal surface, whereas the apices of the few remaining smaller fragments face the periosteal surface. The extensive periosteal reactive bone growth and cortical expansion predominantly extends across the craniolateral to the caudomedial cortex (Figures 3a and 4a), along approximately three-fourths of the diaphysis, beginning just proximal to the mid-diaphyseal region that becomes less severely affected distally (Figure 1d), and extends to the caudal and caudolateral cortex (Figures 3a and 4a). The periosteal reactive bone growth infills the spaces between the cortical fractures (Figures 3a,cIV and 4a). Cross-sectional microCT T A B L E 2 Select cross-sectional measurements of MMNS VP-6332 Elements with measurement explanation Lengths (in, mm) p1 d1 d2

Measurements of histologic thin-sections
The cortical bone width, cranially 11.0 12.1 11.9 The cortical bone width, caudally 8.5 5.6 4.5 The cortical bone width, laterally 7.6 5.1 5.5 The cortical bone width, medially 8.2 4.6 4.5 The width of the medullary cavity, craniocaudally The width, craniolaterally N/A 9.9 11.4 The width, caudomedially N/A 9.7 7.4 imaging demonstrates some of the cortical expansion is attributed to the infilling of bone periosteal reaction between the minimally displaced fracture ends, whereas other areas of expansion are due to multifocal to coalescing cavities with radiolucency similar to those previously described in the cortical-endosteal junction along the proximal region of the diaphysis (Figures 1d and 2a-4a). These radiolucent cavities are also present within multiple regions of the periosteal reactive bone (Figures 1d, 3a, and 4a). There is a clear demarcation between the cortical surface and periosteal reactive bone growth along with the majority of the affected diaphysis, however, some regions near the fracture sites exhibit loss of cortical-periosteal distinction due to obscurity by the adjacent reactive bone. The medullary cavity is filled with minimal to abundant amounts of endosteal reactive and/or pathologic bone deposition along the entire length of the metatarsus (Figure 1d). The most abundant endosteal reactive/pathologic bone deposition fills the middle to distal regions of the medullary cavity. As similarly observed in regions of the interface between the cortical-periosteal reactive bone, there are multiple regions along the endosteal surface that exhibit loss of endosteal-medullary cavity distinction due to the obscurity of the endosteal reactive/pathologic bone deposition and radiolucent cavities. On cross-sectional microCT imaging, there is a large focal region of the medial to caudomedial periosteal reactive bone growth that exhibits increased ill-defined radiolucency (disruption of the reactive bone and porosity) that extends across the full-width of the adjacent cortex to the endosteal surface, and into the medullary cavity (tunnel "bone fistula" formation) (Figure 4).

| Osteohistological analysis
Although the fracture can be identified by microCT digital sections, the degree of remodeling and the stages of fracture repair are best determined histologically. The three cross-sections from the proximal (p1) and middiaphyseal (d1 and d2) regions of MMNS VP-6332 capture various histologic features associated with the development of the chronic disease process in this study (Figures 2-4). Within the p1 cross-section, there is no overt external pathology observed microscopically along the periosteal surface (Figure 2b-cI). The outer one-third of the cortex is composed of fibrolamellar bone tissue with predominant longitudinal and reticular, and a minor component of sub-plexiform vascularization pattern ( Figure 2CI). The remaining inner two-thirds of the cortex are composed of multiple generations of secondary osteons (Haversian bone), consistent with extensive remodeling (Figure 2cI-cII,IV). The large medullary cavity is lined by multiple layers of avascular endosteal lamellar bone (ELB). The ELB is mostly wellpreserved ( Figure 2CII); however, there are multiple regions of the ELB that are disrupted and/or obliterated by the deposition of endosteal pathologic bone (Figure 2cIII-IV). This endosteal pathologic bone is characterized by disorganized, anastomosing trabeculae of woven bone containing numerous, haphazardly arranged, relatively large osteocyte lacunae (Figure 2cII-IV). Frequently, medium to extremely large-sized resorption cavities further disrupts the endosteal reactive/pathologic bone (Figure 2cI-cIV). These resorption cavities extend outward into the middle of the cortex, similarly disrupting the Haversian bone (Figure 2cII-IV). The extensive periosteal reactive bone growth observed grossly and on microCT imaging in the middiaphyseal regions is characterized histologically in the d1 and d2 cross-sections by woven bone trabeculae containing large osteocyte lacunae (Figure 3cII). However, these anastomosing trabeculae exhibit a radially deposited outgrowth that is oriented perpendicularly to the periosteum (Figure 3cI). The woven bone is more abundantly deposited across the craniolateral to the caudomedial region of the cortex, whereas it is minimally to mildly deposited across the remaining lateral to caudal F I G U R E 2 CT and osteohistological overview of the proximalmost p1 cross-section of MMNS VP-6332. A full-width cross-section based on (a) CT data and (b) osteohistological data; (c) Close-up images of (b) aspects of the cortex in d1 and d2 cross-sections (Figures 3b and 4b). The normal cortical bone in the d1 and d2 cross-sections is composed of fibrolamellar bone tissue with a vascularization pattern that transitions from a reticular in the deep cortex, to a combined plexiform and laminar, into primarily plexiform vascularization across the deep, middle, and superficial cortex (Figure 3cIV). Unlike the abundance of Haversian bone tissue observed in the p1 cross-section, there are clusters of secondary osteons in the d1 and d2 cross-sections that are mostly confined to regions of remodeling in the lateral, caudolateral, and caudomedial cortex (Figure 3b, shaded dark gray areas). Single, isolated secondary osteons are occasionally present throughout the remaining cortices (Figure 3cIV). The mid-diaphyseal crosssections (d1 and d2) capture a full-width, complete, minimally displaced mid-cortical fracture (Figures 3b and 4b). In the d1 cross-section, additional smaller, complete, oblique, minimally displaced fractures radiate from the larger midcortical fracture to the periosteal surface, creating multiple irregular triangular-or wedge-shaped fracture fragments (Figure 3b,cIV). The aforementioned reactive woven bone that regionally covers the periosteum diffusely infills the spaces between these multiple cortical fractures (Figures 3b,cII,cIV and 4b). There is more extensive disruption of the ELB in the d1 and d2 cross-sections where most of the ELB lining the large medullary cavity is obliterated (Figures 3b,  cIV and 4b-cI), and only remnants remain along the caudolateral region in the d1 cross-section. This destruction is due to an increased amount of endosteal reactive pathologic bone deposition that partially (d1) to almost completely (d2) fills the medullary space, as well as larger resorption cavities that efface the normal bone tissue along the corticalendosteal junction (Figures 3b,cIII and 4b-cI). Resorption cavities are additionally present multifocally throughout the cortex, and the mid-cortical and periosteal woven bone tissue (Figures 3b and 4b). Within the d2 cross-section, the periosteal woven bone near the medial to the caudomedial cortex is disrupted by coalescing voids and remnant periosteal woven bone fragments that are lined by undulating and scalloped edges (Figure 4cI). Some of the periosteal woven bone fragments appear to be depleted or devoid of large (c) Close-up images of (b), showing a process of osteonecrosis and osteolysis with a full-thickness cortical extension to the endosteum creating a bone fistula. Note that: Green dashed line refers to the region of osteonecrosis and osteolysis, and red star refers to a portal from the bone fistula for secondary osteomyelitis osteocyte lacunae. A fairly wide tunnel (termed a bone fistula) that is also lined by undulating and scalloped edges extends from the cortical-periosteal surface, across the fullwidth of the cortex, through the endosteal surface, and into the medullary cavity. The coalescing voids within the periosteal woven bone communicate with both the most superficial periosteal surface as well as the medullary cavity, creating an abnormal connection between the outer bone surface and inner medullary cavity (Figure 4b-cI).

| Osteopathological analysis
Our X-ray microCT and analyses of ground sections provide complementary information about the pathological process affecting metatarsal MMNS VP-6332 beyond what is observable via gross examination. Throughout p1, d1, and d2, there is increased radiolucency on microCT imaging and voided spaces on histology, consistent with increased resorptive activity within the cortex, as well as the endosteal and periosteal bone (Figs. 2CI-CII, CIV, and 3CIII-IV). In addition, sagittal and cross-sectional microCT imaging and histopathologic examination confirm the presence of multiple complete, minimally displaced, oblique to comminuted fractures extending across the cranial and caudal aspects of the mid-diaphyseal cortex in MMNS VP-6332 (Figures 1 and 3-4). Within MMNS VP-6332, these oblique fracture lines join together between the proximal and distal fracture ends to create variably sized wedge-or triangular-shaped fracture fragments known as "butterfly" fragments (Figs. 1DI-DII, 3B, 3CIV, and 5D-E) (Isa et al., 2021;Psihogios, 1995;Reber & Simmons, 2015).
A butterfly fracture is the most common fracture pattern associated with blunt force trauma to a long bone, and is considered a classic pattern of failure in a bending bone (Figure 5c-d) (Isa et al., 2021;Reber & Simmons, 2015). This bending failure is produced by the combination of compressive and tensile forces along the diaphysis of a long bone that creates an initial fracture on the nonimpact (tension) side, followed by oblique fractures on the impact (compressed) side, to produce the butterfly fragments (Figure 5c). The butterfly fracture pattern along the cranial and caudal aspects of the diaphysis of MMNS VP-6332 observed in this study is consistent with the fact that this ornithomimid dinosaur suffered a blunt force traumatic impact to its right second metatarsal (Figs. 1DII and 5D-E). However, the pattern and orientation of these butterfly fragments can also provide important information on the source of the trauma.
When the origin of the fracture (apex of the wedge) starts on the opposite side of the impact, or the tension side of the bone, the butterfly fragment is known as an "impact side butterfly" or "tension wedge" fragment (Cohen et al., 2016;Psihogios, 1995;L'Abbé et al., 2019;Reber & Simmons, 2015). Whereas a "nonimpact side butterfly" or "compression wedge" fragment is characterized by a 180 reversed orientation of a tension wedge fragment with the start (apex) of the fracture occurring on the same side as the impact, or the compressed side of the bone (Figure 5c). The majority of the butterfly fragments observable within MMNS VP-6332 have their widest base facing periosteally and their apices facing towards the endosteum. A smaller percentage of cortical bone fragments in MMNS VP-6332 exhibit a 180 reversed orientation (apices facing the periosteum and base along the endosteum) (Figs. 1DII and 5D-E). Although (to our knowledge) no studies have been published on the biomechanics of fracture patterns due to formation, infilling of callus bone around fracture fragments, and soft tissue injury leading to infection; (g) Soft tissue infection leading to bone fistula formation, secondary osteomyelitis, and endosteal pathologic bone deposition blunt force impact in living archosaurs (crocodilians and birds), there are numerous studies using mammalian models to provide supporting and reproducible evidence in the interpretation of blunt force trauma-induced fracture patterns and these patterns should be taxonomically transferrable (e.g., Cohen et al., 2016;Isa et al., 2021;Psihogios, 1995;L'Abbé et al., 2019;Passalacqua & Fenton, 2012;Reber & Simmons, 2015;Wheatley, 2008). Reber and Simmons (2015) demonstrated that the anterior (cranial) and posterior (caudal) aspects of long bone diaphysis were 1.7 and 5.7 times more likely to result in an impact side (tension wedge) butterfly fragment compared to a nonimpact side (compression wedge) butterfly fragment. It is unclear whether the butterfly fragments observed in this metatarsal are true tension wedge versus compression wedge fragments as the cause of this blunt force trauma and the impact side are unknown. However, based on the pattern and distribution of the butterfly fracture fragments in MMNS VP-6332, coupled with the higher propensity for the development of impact side (tension wedge) butterfly fragments along the caudal shaft of a long bone, it is likely that the caudal aspect of the diaphysis in this study was the site of the primary impact.
Blunt force trauma in dinosaurs could have resulted from a variety of events. However, dinosaurs likely suffered from blunt force traumas resembling those most commonly reported in reptiles and birds presented to rehabilitation facilities (e.g., intra-or interspecific competition, predator-prey interaction, aggressive mating behavior, or other traumas) (Grosso, 2019;Mitchell, 2002). A resultant crushing injury from one of these blunt force trauma examples could have created forces strong enough to produce multiple tension wedge fractures as well as a soft tissue injury within the caudomedial region of the bone.
The butterfly fracture fragments in this study are encased within more abundant periosteal and a lesser amount of endosteal reactive woven bone that is consistent with chronic, stable fracture callus formation (Figures 3b-cI and 4b). This particular reactive woven bone is a rapidly deposited tissue formed during fracture healing, and is characterized by highly vascular, perpendicularly oriented, interconnecting trabeculae containing numerous, disorganized, fairly large osteocyte lacunae (Figure 3cI-cII) (Bigham-Sadegh & Oryan, 2015;Shapiro & Wu, 2019). Additionally, the periosteal reactive callus bone diffusely infills the spaces between the minimally displaced fracture ends, completely encircling the butterfly fragments (Figures 3b,cII,cIV and 4b). The periosteal callus deposition is concentrated more abundantly along the cranial and medial aspects of the affected cortex, which corresponds to the location of the movable butterfly fragments within the cranial (Figures 1dII, 3b, and 4b) and caudal cortex (Figure 1dII). This pattern and distribution of fracture callus deposition extending across a specific region of an affected bone is a common sequela observed in repair and healing bone fractures, and is associated with the body's attempt to stabilize those regions within the fracture exhibiting the most instability and loss of structural integrity (Mitchell, 2002;DiGeronimo & Brandão, 2019;Grosso, 2019;Sabater, 2019). Although bone healing varies between mammals, reptiles, and birds, the basic osteological or orthopedic principles to promote proper anatomical alignment and rigid stability of the fracture ends, minimize soft tissue disruption, and conserve blood supply to the injured bone are still imperative for successful fracture repair and healing across all species (Mitchell, 2002;DiGeronimo & Brandão, 2019;Grosso, 2019;Sabater, 2019). Information published on fracture healing in living reptiles and birds is limited, and controlled experiments to study the mechanisms involved in fracture repair processes are extremely scarce. In general, both living birds and reptiles undergo the same three phases of long bone healing (i.e., reactive, reparative, and remodeling phases), that can often occur concurrently within the same fracture site, and the reparative phase includes the formation of a soft and hard callus as normally observed in mammals (Bigham-Sadegh & Oryan, 2015;DiGeronimo & Brandão, 2019;Grosso, 2019;Sabater, 2019). A number of factors can impede and/or delay the normal stages of fracture healing; however, even in uncomplicated case reports, the average time for a bone to heal completely in reptiles is substantially longer (6-30 months) when compared to birds (2-4 weeks) and mammals (4-8 weeks). Dinosaurs are interpreted as endothermic animals with metabolic rates intermediate between extant ectotherms (reptiles) and extant endotherms (birds and mammals) (e.g., O'Connor & Dodson, 1999;Reid, 2012). Therefore, although it is unclear how long it would have taken this particular ornithomimid to completely heal this fractured metatarsal, we can reasonably infer an estimate within this range. Regardless, the fracture callus, captured in the d1 and d2 cross-sections histologically, indicates that this individual was well into the reparative, and beginning stages of the remodeling phases of fracture healing at the time of death. This is evidenced by a fully developed stable hard callus, yet limited remodeling of the reactive periosteal tissue.
Additional evidence captured both on microCT imaging and histology indicates a delay in the normal healing process of this fracture that was most likely caused by chronic osteomyelitis (an infection inside the bone). This lesion is evident in the d1 and d2 cross-sections where the ELB is extensively disrupted by increased amounts of endosteal pathologic bone deposition that partially or almost completely fills the medullary cavity, or is obliterated by large resorption cavities (Figures 3b,cIII and 4b).
Osteomyelitis is one of the most common complications in fracture healing in wild reptiles and birds that have sustained traumatic fractures and soft tissue injuries (Craig et al., 2016;Grosso, 2019;Mitchell, 2002). Opportunistic pathogens-most frequently bacteria from the external environment and/or skin-can contaminate penetrating wounds and/or open comminuted fractures and lead to localized infections. Contamination of a fracture site via an abnormal pathway between the inside of the bone and the surface of the body is known as a bone fistula (Craig et al., 2016). In this study, microCT imaging and histology of the d2 cross-section reveal a fairly large region of the medial to caudomedial periosteal fracture callus that is disrupted by coalescing void spaces with remnants of fragmented callus bone that are depleted or devoid of osteocyte lacunae (Figure 4). These histologic features are most consistent with bone death (osteonecrosis), likely occurring in the region adjacent to the soft tissue infection. A wide tunnel creating an abnormal connection between the inner medullary cavity and the outer periosteal bone surface is observed extending across the full-width of the cortex (Figure 4b-cI). Both the periosteal callus bone remnants and the cortical tunnel are lined by undulating and scalloped edges, features most consistent with bone resorption (osteolysis).
Based on the overall patterns and distribution of these lesions, we suggest the primary etiology was blunt force trauma to the right second metatarsal (MMNS VP-6332), resulting in a butterfly fracture pattern with multiple tension wedge fragments, and soft tissue injury. As commonly reported in living wild birds and reptiles with contaminated wounds near an open fracture site, the injury in this study developed a subsequent secondary bacterial infection in the soft tissue adjacent to the fracture that eventually progressed to bone destruction (osteolysis) and death (osteonecrosis) of the periosteal fracture callus and adjacent cortex, creating an abnormal communication between the inside of the bone and outer wound surface (bone fistula formation). This resulted in a prolonged bacterial infection that spread to the inside of the bone (chronic osteomyelitis), culminating in the deposition of endosteal pathologic bone tissue. This is the first report of chronic osteomyelitis and bone fistula formation secondary to a traumatic fracture in ornithomimid dinosaurs (Molnar, 2001;Moodie, 1918Moodie, , 1921Rothschild et al., 2001;Tanke & Rothschild, 2002). Although we cannot definitively determine whether the chronic osteomyelitis and bone fistula were primary contributing factors leading to this ornithomimid's death, we can conclude that this animal did survive long enough past the initial traumatic fracture to develop a chronic callus and infected wound severe enough to destroy and invade the bone via a bone fistula and cause chronic osteomyelitis in the medullary cavity.
Overall, few pathologies have been described for ornithomimosaurian dinosaurs. Those that have are almost exclusively limited to fractures of the hind limb elements. These few examples include a comminuted impact fracture (the first case reported in a dinosaur) in the fourth metatarsal of an indeterminate ornithomimid (SMP VP-971) from the Upper Cretaceous Kirtland Formation of New Mexico; a possible stress fracture on the second metatarsal of the ornithomimid (LACM 42895) from the Upper Cretaceous Hell Creek Formation of northeastern Montana; gout on a pedal phalanx of an ornithomimid from the Horseshoe Canyon Formation of Alberta; and a traumatic fracture associated with tendon/ ligament avulsion, green-stick fracture, and possible gout on the pedal elements of the recently described eastern North American ornithomimosaur taxon A. fridayi (Bell et al., 2011;Molnar, 2001;Rothschild & Lambert, 2019;Sullivan et al., 2000;Tanke & Rothschild, 2002). Our identification of a traumatic impact fracture with subsequent chronic osteomyelitis in an ornithomimosaurian foot fits this pattern and the preponderance of these injuries and their skeletal localization may ultimately offer important insight into aspects of ornithomimosaurian paleobiology and behavior. A full survey of ornithomimosaurs for paleopathologies would be a good line of future research.

| Implications for the identification of endosteal bone tissues
The presence of enigmatic endosteal tissue coupled with the absence of external pathology is one of several criteria considered useful for confidently identifying reproductivelyinduced medullary bone in fossil archosaurs (Cerda et al., 2014;Chinsamy et al., 2013;Hübner, 2012;O'Connor et al., 2018). However, in a recent review, Canoville et al. (2020) suggested caution in applying too much weight to the absence or presence of reactive periosteal tissue as a means to discriminate medullary bone from pathological endosteal tissue. They first noted that, theoretically, a female exhibiting a skeletal pathology could also undergo a normal reproductive cycle and hence simultaneously exhibit medullary bone and pathological tissues. Such a specimen was actually observed (see the pathological ulna of Colinus virginianus [TM M-M 6536] in Canoville et al., 2020: Figure 4a-d). Therefore, the presence of one tissue type (reproductive or pathological) cannot categorically refute the presence of the other. In the specific case of TM M-M 6536, the presence of periosteal pathological tissues does not contradict the identification of endosteal tissue as medullary bone in some skeletal elements. Likewise, Canoville et al. (2020) surmised that the absence of grossly observable external pathological indicators, either localized or across the entire bony element, cannot be used to refute a pathological identification for unusual endosteal bone. As mentioned by Chinsamy and Tumarkin-Deratzian (2009) and Cerda et al. (2014), and although less common than occurrences with extensive periosteal bone proliferation, instances of avian osteopetrosis in extant dinosaurs (living birds) are described where endosteal pathological bone is deposited with only negligible corresponding periosteal deposition visible only on radiographic examination ("type C" of Holmes [Holmes, 1961]). A similar distribution was observed within the fossilized femur of a sauropodomorph dinosaur by Cerda et al. (2014), who describe it as a probable case of "avian" osteopetrosis despite the apparent absence of periosteal reactive bone, and who also raised the issue of preservational artifacts as an explanation for the lack of periosteal pathological bone.
Moreover, as recently demonstrated by Canoville et al. (2020) after analyzing a series of 20 extant avian skeletal pathologies with different etiologies, the distribution of endosteal and periosteal pathological tissues across the shaft of a single long bone are often decoupled in extant birds, and endosteal deposition of pathological bone can extend proximal or distal to the localization of periosteal reactive bone and/or abnormalities within the cortex. Hence, when only a highly localized region of a pathological element is examined, endosteal deposition of pathological bone might be observed without associated, grossly observable periosteal reactive bone and inaccurately interpreted as representing reproductively-induced medullary bone.
The decoupling of endosteal and periosteal pathological bone along the shaft becomes an enhanced problem of concern for fossil specimens that are often fragmentary or whose external surface has been weathered. The majority of non-avian specimens of extinct species published as containing medullary bone because they exhibited an unusual endosteal tissue are either described solely from incomplete skeletal elements (e.g., Hübner, 2012;Skutschas et al., 2017) and/or the histological examination of the unusual endosteal tissue was restricted to one cross-sectional level of the shaft due to the consumptive nature of osteohistological sampling (e.g., Chinsamy & Tumarkin-Deratzian, 2009;Cerda et al., 2014;Redelstorff et al., 2014;Jentgen-Ceschino et al., 2020), leaving the intra-elemental distribution pattern of these tissues unclear. MMNS VP-6332 is an interesting example of this decoupling and the first definitive example that we are aware of in a fossil specimen (with the possible exception of Cerda et al., 2014). In the case of MMNS VP-6332, endosteal bone deposition extends further proximally than the external fracture callus due to osteomyelitis (Figure 1dI-II). The cross-section taken near the proximal end of the metatarsal (p1) exhibits significant amounts of endosteal pathological tissue similar microstructurally to medullary bone (e.g., trabecular architecture, well-vascularized, woven bone matrix with large osteocyte lacunae), yet is free from external pathological bone deposition, and impacts within the primary cortex are limited to resorption and remodeling (Figure 2cII). Thus, if only the proximal portion of this element was preserved, this tissue would satisfy this particular criterion for the identification of medullary bone. The etiology of this specimen (fracture and secondary infection) is expected to have been commonplace in extinct archosaurs, as opposed to rarely observed, and may therefore be much more common than currently realized.
Finally, Canoville et al (2021) noted that in modern birds, medullary bone is homogeneously deposited and typically lines the entire peri-endosteal margin of the medullary cavity, whereas pathological endosteal tissues are typically of non-uniform thickness and usually more localized in their distributions in transverse section. They used this observation to further refute a reproductive nature of endosteal tissues in some fossil specimens, for which the given tissues were restricted to a small subset of the medullary region. Here we note that the pathological endosteal tissue in section p1 of MMNS VP-6332 was homogeneously deposited around the endosteal margin (visible in transverse section), a pattern consistent with the relatively uniform deposition of medullary bone. As such, we suggest that whereas a high degree of localization around the endosteal margin in transverse section is reasonable evidence against an identification of medullary bone, endosteal tissue of uniform thickness in transverse section cannot be used as a confident criterion for discriminating between medullary bone and pathological endosteal bone.