Preclinical models of vertebral osteomyelitis and associated infections: Current models and recommendations for study design

Abstract Spine‐related infections, such as vertebral osteomyelitis, discitis, or spondylitis, are rare diseases that mostly affect adults, and are usually of hematogenous origin. The incidence of this condition has gradually risen in recent years because of increases in spine‐related surgery and hospital‐acquired infections, an aging population, and intravenous (IV) drug use. Spine infections are most commonly caused by Staphylococcus aureus, while other systemic infections such as tuberculosis and brucellosis can also cause spondylitis. Various animal models of vertebral osteomyelitis and associated infections have been investigated in mouse, rat, chicken, rabbit, dog, and sheep models by hematogenous and direct inoculation in surgery, each with their strengths and limitations. This review is the first of its kind to concisely analyze the various existing animal models used to reproduce clinically relevant models of infection. Spine‐related infection models must address the unique anatomy of the spine, the avascular nature of its structures and tissues and the consequences of tissue destruction such as spinal cord compression. Further investigation is necessary to elucidate the specific mechanisms of host‐microbe response to inform antimicrobial therapy and administration techniques in a technically demanding body cavity. Small‐animal models are not suitable for large instrumentation, and difficult IV access thwarts antibiotic administration. In contrast, large‐animal models can be implanted with clinically relevant instrumentation and are resilient to repeat procedures to study postoperative infection. A canine model of infection offers a unique opportunity to design and investigate antimicrobial treatments through recruitment a rich population of canine patients, presenting with a natural disease that is suitable for randomized trials.

use. Spine infections are most commonly caused by Staphylococcus aureus, while other systemic infections such as tuberculosis and brucellosis can also cause spondylitis. Various animal models of vertebral osteomyelitis and associated infections have been investigated in mouse, rat, chicken, rabbit, dog, and sheep models by hematogenous and direct inoculation in surgery, each with their strengths and limitations. This review is the first of its kind to concisely analyze the various existing animal models used to reproduce clinically relevant models of infection. Spine-related infection models must address the unique anatomy of the spine, the avascular nature of its structures and tissues and the consequences of tissue destruction such as spinal cord compression. Further investigation is necessary to elucidate the specific mechanisms of host-microbe response to inform antimicrobial therapy and administration techniques in a technically demanding body cavity. Small-animal models are not suitable for large instrumentation, and difficult IV access thwarts antibiotic administration. In contrast, large-animal models can be implanted with clinically relevant instrumentation and are resilient to repeat procedures to study postoperative infection. A canine model of infection offers a unique opportunity to design and investigate antimicrobial treatments through recruitment a rich population of canine patients, presenting with a natural disease that is suitable for randomized trials. Vertebral osteomyelitis, also called spondylodiscitis or spondylitis, is a rare disease that mostly affects adults, and is usually of hematogenous origin. Vertebral osteomyelitis is an infection of the vertebral bodies which may involve the adjacent intervertebral disc (IVD) resulting in an associated discitis. Vertebral osteomyelitis, accounting for about 5% of all osteomyelitis cases, has an incidence of approximately 2.4 cases per 100 000 population, and incidence increases from 1.7 per 100 000 in <59 years of age to 25.1 per 100 000 in >80 years of age. 1 Risk factors for developing an infection include intravenous (IV) drug use, bacterial endocarditis, intervertebral disc degeneration, Kieran Joyce and Daisuke Sakai shared first authorship. previous spinal surgery, diabetes mellitus, corticosteroid therapy, or other immunocompromising conditions. [2][3][4][5][6] Vertebral osteomyelitis most commonly occurs due to hematologically derived seeding, direct inoculation in spinal surgery, or from the invasion of infection from surrounding tissue. 7 The infection is commonly due to bacteria, but fungi and parasites have also been identified as causative agents. The most commonly implicated organism in spinerelated infections is Staphylococcus aureus (methicillin resistance becoming more prevalent), followed by Escherichia coli. 8 Osteomyelitis due to direct inoculation during spinal surgery, particularly after instrumentation, is most often caused by S. aureus and Propionibacterium acnes, a typically normal inhabitant of skin flora. 7,8 However, low-virulence microorganisms such as coagulase-negative S. aureus may induce hematogenous vertebral osteomyelitis, especially in the setting of a prolonged implantassociated bacteremia. 9 Most patients that develop hematogenous vertebral osteomyelitis have ongoing co-morbidities, such as diabetes mellitus, coronary artery disease, immunosuppression, cancer, or renal failure requiring hemodialysis. [10][11][12][13] Spinal infections have a variable presentation, and as such, vertebral osteomyelitis can be complicated by paravertebral, epidural, or psoas abscesses by direct seeding. 7 In a study reporting on the complicated presentation of vertebral osteomyelitis, an epidural abscess was reported in 17% of cases, paravertebral abscess in 26%, and intradiscal abscess in 5%. 7 Motor weakness or paralysis develops in approximately a quarter of patients, with an increased incidence of neuropathy occurring in patients with osteomyelitis of the cervical spine. Overall, neurological complications are common in vertebral osteomyelitis, where 38% of patients will develop neurological symptoms. 14

| Hematogenous dissemination
Hematogenous dissemination is the most common cause of vertebral osteomyelitis. 15 Adult vertebral bone is highly vascularized with slow high-volume blood flow via the posterior spinal artery, making it susceptible to bacterial seeding. 16 Many patients with hematogenous pyogenic vertebral osteomyelitis are predisposed due to underlying conditions such as diabetes mellitus, heart disease and immunocompromising disorders. 8,13,17 The lumbar vertebral bodies are most often implicated, followed by thoracic and, less commonly, cervical vertebrae, while hematogenous sacral osteomyelitis is rare.
Noncontiguous epidural abscesses occur in approximately 10% of the cases that are complicated by abscess. 18

| Direct inoculation
Direct bacterial inoculation in spinal surgery and subsequent postoperative infection is a devastating complication, associated with increased morbidity and/or mortality. In vertebral osteomyelitis, management is further complicated by the avoidance of instrumentation explantation, which would destabilize the spine with potential neurologic compromise.
Patients that develop vertebral osteomyelitis require prolonged hospitalization, repeat surgeries for removal of instrumentation and/or debridement, and a long course of IV antibiotics, followed by oral antibiotics. Postoperative infections incur a heavy burden on healthcare systems estimated at one million excess inpatient days and 2.72 billion USD additional costs per year in the US alone. 19 Approximately 1% of the patients undergoing elective spine surgery without instrumentation are complicated by postoperative infection and incidence increases when the hardware is used, despite stringent aseptic surgical technique and prophylactic antibiotic protocols. 20 Implant-associated infection is complicated by biofilm formation, where bacteria readily adhere to implant surfaces, developing a biofilm layer over several days, reducing antibiotic susceptibility by 100 to 1000 times. 21

| Extension of primary infection site
Primary vertebral osteomyelitis can be complicated by an extension of the initial infection. Infection may extend posteriorly as an epidural abscess, subdural abscess, or even meningitis, and are more often associated with gram-positive bacterial infection than gram-negative bacterial infection. 16,22,23 Anterior or lateral extension of infection can lead to paravertebral, mediastinal, retroperitoneal, or psoas abscess. 3 Infection can occur in spinal elements other than the vertebral bodies, including the posterior spinous processes, the facet joints, and the pedicles. 24 Thoracic vertebral infections have even been recorded to extend into the pleural space to produce an empyema. 25

| Diagnosis
The diagnosis of vertebral osteomyelitis can be challenging, as infection may be insidious, often resulting in delayed identification of the condition and infecting organism. When clinical suspicion warrants investigation, diagnosis can be confirmed with the use of magnetic resonance (MR) imaging, microbiological cultures, and tissue biopsy examinations. Vertebral osteomyelitis is identified by high signal intensity on T2 weighted MR images. 26 Molecular diagnostics are not routinely used when investigating vertebral osteomyelitis; however, negative bacterial cultures spur the use of a panel polymerase-chainreaction (PCR) analysis to identify microbial DNA in biopsies. 27 This enables the detection of less common microorganisms, such as anaerobic bacteria, Brucella and Bartonella species. 27 Broad-range PCR is limited by reduced sensitivity and specificity, which dramatically decrease due to the probability of contamination and cannot provide an antibiotic resistance profile for the microorganisms. 27

| Clinical management
At present, there are no data from clinical trials to inform specific antimicrobial regimens for vertebral osteomyelitis and associated infections, nor are there guidelines on the duration of antibiotic therapy.
The choice and duration of therapy cited in case reports may be associated with the extent of infection or with patient-specific considerations, offering little value to prescribing guidelines. Given the increasing incidence of spine-related infections and the significant morbidity and mortality of this condition, further, preclinical research and clinical trials are needed to elucidate the variable onset and progression of this debilitating complication. 28 It is challenging to design and undertake clinical trials for this complicated disease process, considering the low incidence and high heterogeneity of induction and presentation of infection with various implicated organisms.
This paper provides an overview of the characterized preclinical models of vertebral osteomyelitis, highlighting the strengths and weaknesses of each model and suitability for controlled trials of treatment strategies. Since the first documented model of vertebral osteomyelitis was developed in chickens in 1971 by Wise et al, significant work progress has been made in developing a standardized model to replicate human disease as models for antimicrobial therapies and surgical management of vertebral osteomyelitis. 29 In the present review, the authors critically analyze characterized spine-related infections in each relevant animal model to evaluate reproducibility, clinical relevance and representation of natural disease. The authors also make a case for the use of veterinary patients (specifically dogs) presenting to clinic with natural disease as a suitable cohort for animal trials to test the efficacy of antibiotics and surgical treatments.

| MODELS OF SPINAL INFECTIONS
While the human disease is referred to as vertebral osteomyelitis (infection of the vertebral bone) or discitis (infection of the IVD) or spondylodiscitis (a combination of both), animal models replicating these conditions use highly variable language and definitions surrounding the type of infection ( Figure 1). Thus, the authors have attempted to discuss these models as uniformly as possible for comparability. To summarize the additional terminology used below; implant-associated spondylitis includes implantation of an inoculated foreign body, disseminated infection describes the spread of an infective organism throughout the body, abscesses are complications of localized disease and/or disseminated infection, and acute pyogenic spondylodiscitis is a subset of spondylodiscitis with the production of pus.
Various animal models of vertebral osteomyelitis and associated infections have been investigated in recent decades. Mouse, rat, chicken, rabbit, dog, and sheep models are summarized in Table 1.
Models have been sub-classified for ease of summation. Small-animal models, such as mouse and rat, and large-animal models, including dogs and sheep, have been discussed to evaluate the method of infection induction used, the following characteristics of infection that were generated and investigations performed to assess the response to infection. Rabbits have been most extensively studied and thus have been discussed separately. Case-reports of veterinary studies have also been included to highlight the subset of studies that investigate naturally occurring disease in animals.  • Macro: All animals that developed paralysis were found to have abscesses within the spinal column along the length of the spine.

| Rodent models
• Histo: Inflammatory cell infiltration and vertebrae destruction were observed.  • Clin: The degree of the disease process was more advanced in the S. aureus group and less severe in the Pseudomonas group.

S. aureus
Direct inoculation-Inoculation of vertebral body using gauze Radiography, histological evaluation • Histo: Acute inflammation started within 1 or 2 weeks and subsided by 5 or 6 weeks.
• Macro: In 55% of the dogs, the inflammation was confined within the vertebral body, in 10% it invaded into the IVD, and in 35% inflammation invaded into the anterior longitudinal ligament.

| Rabbit models
To date, rabbit models of spine infections have been most intensely investigated over any other animal models. Studies have explicitly included models of lumbar vertebral osteomyelitis, 37 In 1998, Guiboux et al described the first implant-associated spine infection model in rabbits. 46 Since it is of particular interest to produce infection associated with, and complicated by, in situ instrumentation, a previous model of discitis was combined with an instrumentation technique to create a postoperative instrumentationassociated infection model. 46 In the nonantibiotic treated group, the swab and tissue cultures grew S. aureus in all five rabbits. Rabbits that received instrumentation and first-generation cephalosporin, cefazolin before surgery did not grow S. aureus. This study showed that prophy- Progressive neurological deficits were observed in 90% of animals, characterized by lower extremity weakness, sphincter dysfunction, and abnormalities of gait. Spinal abscess confirmed in 95% of cases. 43 This model presents a challenging infection for antibiotic therapies, considering the reduced antibiotic penetrance into abscesses. 50 Infectious discitis has been induced by intradiscal inoculation using S. aureus (10 4 -10 6 CFU). 40,41 Discitis was confirmed in the inoculated levels by different intensity on T1-weighted images of the IVD and hyper-intensity of the adjacent vertebrae on T2-weighted images. 41 The infected IVD presented with the destruction of nucleus pulposus tissue and necrosis with associated inflammatory and fibroblastic cell infiltration. 41 A further study found that while vancomycin reduced the overall bioburden within a contaminated surgical site of posterolateral fusion, the addition of the vancomycin to the demineralized bone matrix reduced the fusion capability of the demineralized bone graft. 51 Fusion rates were restored however with an ileal crest graft.

Spondylodiscitis has been induced by a Brucella melitensis
(3 × 10 8 CFU) inoculated gelatin sponge, co-implanted with a Kirschner wire (K-wire) insertion into L6. 39 Spondylodiscitis was observed on MRI, characterized by T2 hyper intensity, regional inflammation involving the vertebra and diffuse marrow edema with paraspinal abscess. Infiltration of inflammatory cells was observed on histology, predominantly consisting of lymphocytes and monocytes.
Rabbit models of infection have been developed to replicate hematogenous seeding, direct inoculation and instrumentation-associated infection. Despite their popularity due to sufficient vertebral size to sustain instrumentation and ease of handle ability, questions remain over their susceptibility to infection and are limited by their intolerance to repeat procedures.

| Ovine models
Large-animal models have advantages over small-animal models in the study of spine disease as they have more relevant anatomy, allow for easy IV access for antibiotic administration and support clinically relevant instrumentation. 52

| Canine models
Canine models of spine infection were first described in 1991 in Japan, induced by S. aureus, Pseudomonas, and E. coli intradiscal inoculation. 57 Results demonstrated a more advanced disease with increased tissue destruction in S. aureus infection. 57 More sophisticated models of spine infection have since been developed to produce a complicated acute pyogenic spondylodiscitis in canine models. 58

| Case-reports of veterinary patients with natural disease
Veterinary reports may be useful to understand common causative organisms in animals and natural disease presentation and progression. Several case reports exist of spinal infections described in animals. These reports may be useful to determine the fidelity of artificial models to environmentally induced disease. Vertebral osteomyelitis, complicated and uncomplicated, has been well described in dogs. 61,62 Vertebral osteomyelitis in dogs is most commonly caused by Scedosporium apiospermum infection, an eutrophic filamentous fungus, has been recorded in a canine case report of osteomyelitis and discospondylitis. 68 No randomized clinical trials have investigated antibiotic regimens in dog patients, and clinical management has mostly followed practices similar to human case reports. [69][70][71][72] Nonhuman primates have also been studied, as posterior paralysis and spinal osteo-

| LIMITATIONS OF CURRENT MODELS
The animal models described in Table 1 have been optimized to be reproducible when sufficient CFU count has been used in inoculation.
Early studies by Guiboux

| Animal species
Rabbit models of osteomyelitis and spondylitis have been popular due to their predisposition to infection over other animals; however, this susceptibility should be considered for cross species comparison. Rabbits have large enough vertebrae for models that use instrumentation to investigate biofilm formation and penetrance, while being less costly than dogs. While rats are even less expensive than rabbits, their smaller spines are not suitable large instrumentation and difficult IV access impedes antibiotic administration. Furthermore, animal models smaller than rabbits make it challenging to investigate complications such as neurological compromise due to abscess or vertebral instability because of their small spines. Mouse and chicken models may be good candidates when investigating hematogenous osteomyelitis as vertebral complications have been characterized . [29][30][31] Mice can be genetically modified to study immunomodulation and its role in developing an infection. A relevant, validated mouse model would be of great benefit in this field, to reliably investigate potential therapies and new devices in a preclinical setting. Promising therapies may progress to testing in larger animal models for further validation and eventually in clinical trials. More often, canine patients are presenting to veterinary clinics with complicated disease and compounding morbidity. 74 Canine patients receive chemotherapy and undergo complex surgeries. 74 Often, aged animals are subject to similar medical interventions as humans. These canine patients are highly valuable as they replicate human-like disease. While presenting data of high clinical value, a study involving these veterinary patients would incur highcosts with logistical and ethical challenges.

| Bacterial species and sufficient inoculum
Current animal models of spinal infections have primarily focused on postoperative infection with S. aureus as the inoculating microorganism, given its high incidence in humans. An overview of existing animal models in each given species is outlined in Figure 1. Future models must be tunable to evaluate many causative microorganisms, whether bacterial or fungal given the variety of causative agents listed in numerous case reports. Table 1

| Efficacy of antibiotic therapy
In general, antibiotics should be withheld until the infectious microorganism has been identified, which is most often the case, provided that the patient was not previously treated with antibiotics before culture samples were taken. 16 No randomized controlled trials have been performed to study the efficacy of antimicrobial therapy in vertebral osteomyelitis, and recommendation online of therapy is mainly derived from observational studies. A retrospective study of 120 participants with clinically diagnosed vertebral osteomyelitis of various microbial origin was treated with appropriate IV regimens for 32 days on average. An infection clearance rate of 91% at six months. 79 A meta-analysis investigating antibiotic therapy for the treatment of varying presentations of osteomyelitis produced an average eradication rate was 79% after 1 year across 22 studies. 80 Differences in antibiotic therapy did not significantly affect the outcomes, except in implant-associated infection where rifampin was superior. 81 Controlled trials do not yet suggest the optimal duration of therapy, and antibiotic regimen recommendations range from 4 to 6 weeks, 79 up to 3 months. 82 Patients with persistent abscesses and retained instrumentation often require prolonged antibiotic regimens. 81,83 Studies of antibiotic activity examine many facets of use including The same group furthered this study by examining antibiotic efficacy in spinal instrumentation-associated infection model using inoculated surgical wire implantation around facet joints. 46 In a model that otherwise produced an established S. aureus induced infection, prophylactic antibiotics effectively inhibited infection. 46 Similarly, a rabbit model of inoculated

| Surgical management
Surgical intervention is rarely performed, though it may be appropriate in spine-related infections, indicated by (a) neurologic deficits. (d) Recurrence of disease despite appropriate antimicrobial therapy. 15,89 No randomized trials are evaluating surgical management of vertebral osteomyelitis. 15 It is common practice to administer an additional six-week "tail" of oral antibiotics in recurrent or chronic infection, although there is little evidence to guide management. When spinal instrumentation is required for stabilization, timely implantation may be safe in the setting of appropriate selection and duration of antimicrobial therapy. 90 Surgical outcomes for patients with vertebral osteomyelitis are highly variable, with one-quarter reporting residual pain and a similar proportion requiring repeat procedures. 91 Anterior approaches of debridement and strut grafting with del-

| Designing a randomized control trial
Small animal models, such as mouse and rat models, use a more significant number of animals to offer statistical power and genetic diversity. The spine of the rat, in particular, is sufficient in size to be used in studies that require drilling and fixation techniques using implants.
Access to the spine is easily achieved, especially in models of isolated discitis where the tail may be used for inoculation. However