Rabbit model of Staphylococcus aureus implant-associated spinal infection

Post-surgical implant-associated spinal infection (IASI) is one of the most common causes of morbidity following spinal surgery and can potentially lead to neurological sequelae and disability. The reported infection rate following spinal surgery varies between 0.5% and 18.8% and is dependent mostly on patient population and the procedure performed (Beiner et al., 2003; Chahoud et al., 2014; Smith et al., 2011). Patients with a history of trauma, diabetes, advanced age or neuromuscular scoliosis are the groups with the highest risk (Aleissa et al., 2011; McDermott et al., 2012; Smith et al., 2011; Sponseller et al., 2013, 2000). Removal of the implanted materials is often not feasible due to the risk of spine destabilization, necessitating repeated surgical interventions and prolonged antibiotic therapy. Still, there is risk for treatment failure and recurrence in the months or years following infection. In a series of 102 IASI patients, treatment failure occurred in 36%, with a median time to failure of 113 days (Cho et al., 2018). Another study followed 253 patients with vertebral osteomyelitis for a median of 6.5 years and relapse occurred in 14% (McHenry et al., 2002). In some rare cases, recurrence may even occur decades following the initial infection (Mansour et al., 1979). Staphylococcus aureus is the most common bacterial pathogen recovered from IASI (Jiménez-Mejías et al., 1999; Pull ter Gunne et al., 2010), and the emergence of methicillin-resistant S. aureus (MRSA) poses an increased risk of treatment failure (Cho et al., 2018). Biofilm formation is thought to reduce antibiotic and immune cell penetration, leading to persistent and difficult-to-treat infections (Cho et al., 2018; Sampedro et al., 2010). The diagnosis of IASI is also difficult to establish, but is suggested by the presence of a sinus tract and persistent or late postoperative wound drainage or dehiscence (Chahoud et al., 2014; Jiménez-Mejías et al., 1999). Non-invasive diagnosis of IASI is challenging, and evaluation of progression as well as response to treatment relies on non-specific inflammatory markers in serum (McDermott et al., 2012). ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) positron emission tomography (PET) is a complementary tool often used to image IASI and might also be beneficial to evaluate disease progression (Gemmel et al., 2010).

Animal models for post-surgical IASI provide an in vivo approach to study the pathophysiology and to evaluate preventative measures, diagnostic tools and treatment strategies. Current animal models in mice, rats and rabbits use a variety of surgical procedures, which may not fully recapitulate the complex surgical techniques and orthopedic hardware in humans (Guiboux et al., 1998; Laratta et al., 2017a,b; Ofluoglu et al., 2007; Poelstra et al., 2000; Stavrakis et al., 2015). In addition, most of these prior animal models of IASI were evaluated in an acute time period following surgery and infection (i.e. 5-14 days) and there was not an opportunity to evaluate the persistent infection and inflammation that occur in human IASI. Despite being an eminent part of IASI, only one study directly evaluated biofilm formation on the spinal hardware (Hazer et al., 2016). Prior studies have also utilized a mouse model of IASI in conjunction with a bioluminescent S. aureus strain and in vivo bioluminescence imaging (BLI) (Dworsky et al., 2017). This useful model was employed to evaluate local antibiotic treatment (Hu et al., 2017), implant antimicrobial coating (Hegde et al., 2018) and multimodal imaging for surgical management (Zoller et al., 2019).

In order to better simulate the surgical procedures and implants involved in human IASI, we developed a post-surgical IASI model in rabbits using a surgical approach and orthopedic-grade implants. Using this model, in vivo bacterial burden, infection-induced inflammation and bone remodeling were monitored using non-invasive imaging [bioluminescence, PET, computed tomography (CT)] and ex vivo analysis to study both acute and chronic disease.

The rabbit IASI was performed as described in Fig. 1 with three different inocula [1×10⁴, 1×10⁵ or 1×10⁶ colony-forming units (CFUs)] of the bioluminescent MRSA strain SAP231. Non-invasive in vivo whole-animal BLI (on post-operative days 0, 3, 7 and once weekly thereafter for a total of 56 days) demonstrated a peak bacterial burden by day 14, when wound dehiscence also occurred (Fig. 2A-C). Note that for 1×10⁴ CFUs, BLI signal appeared only after wound dehiscence occurred and within the open wound (Fig. 2A,B, top row). For higher inocula, the in vivo BLI signals were apparent before wound dehiscence. The BLI signals decreased gradually over the 56-day experiment with all inocula (Fig. 2B,C). By post-operative day 56 there were no in vivo BLI signals detected in any of the rabbits and all of the wounds overlying the post-operative sites spontaneously healed (Fig. 2A-C). On post-operative day 14, ex vivo CFU enumeration confirmed high burden of infection for soft tissue and bone as well as the hardware (Fig. 3A). In contrast, on post-operative day 56, there were no ex vivo CFUs recovered from the tissue or implants. To identify the presence or absence of any remaining bacteria in the tissue specimens or adherent to the implants, soft tissue and bone homogenates as well as implant sonicates were cultured in shaking broth for an additional 48 h followed by detection of CFUs after overnight culture on plates. On post-operative day 14, 100% of the tissue and implant specimens had bacterial CFUs present (Fig. 3B). However, on day 56, only 50% of the tissue specimens and 100% of the implant specimens had bacteria present, and all recovered bacterial colonies were bioluminescent as expected. On post-operative days 28 and 56, structures suggesting biofilm formation with viscous fibers were identified on all of the ex vivo screws and plates (Fig. 4), although these structures were less extensive on day 56 compared with day 28. Bacterial biofilms were deemed present if characteristic features such as viscous fibers were present.

On post-operative days 7, 21 and 56, ¹⁸F-FDG-PET/CT imaging was performed (Fig. 5). The ¹⁸F-FDG-PET imaging demonstrated distinct and localized uptake at the site of infection (Fig. 5A). In agreement with in vivo BLI signals (Fig. 2B,C), the standard uptake value (SUV) ratios of the ¹⁸F-FDG-PET imaging signals between the infected site surrounding the hardware and a non-infected reference increased from 3.99±0.23 on post-operative day 7 to 5.37±0.41 on post-operative day 21 (Fig. 5B, left). ¹⁸F-FDG uptake decreased by day 56 (SUV ratio=3.88±0.04), but the ¹⁸F-FDG uptake signals were still visible and localized to the site of infection, likely indicating persistent inflammation. During the 56-day course of infection, the ¹⁸F-FDG-PET signal from bone did not statistically change, with bone-to-non-infected reference SUV ratios of 2.55±0.14, 2.30±0.07 and 2.21±0.13 on post-operative days 7, 21 and 56, respectively. However, bone-to-soft tissue ratios had statistically significant changes as they were 0.86±0.12 and 1.22±0.07 on post-operative days 21 and 56, respectively (P<0.05) (Fig. 5B, right), suggesting an increase in bone inflammation at the later time point.

On post-operative day 56, rabbits were euthanized and infection-induced changes in the bone were evaluated by high-resolution ex vivo CT imaging. Visualization of lumbar vertebrae showed that the infected vertebra had less dense bone than uninfected vertebra (Fig. 6A). The infected vertebra had decreased amount of dense bone (31.9±1.5%), compared with the uninfected vertebra (36.5±1.0%; P<0.05) (Fig. 6B). It should be noted that even small changes in bone density (as low as 2-3%) are clinically significant (Bonnick, 2008).

Post-surgical IASI is a devastating complication of spinal surgical procedures with potentially long-term sequelae. Accurate diagnosis is crucial, and precision medicine tools are being developed for prognostication and risk stratification of patients. In this study, a rabbit model of a post-surgical MRSA IASI was created to more closely simulate the surgical technique and orthopedic-grade hardware used in spinal implant surgery in humans. This model more accurately recapitulates initial infection that often involves wound dehiscence at the post-surgical site. Even though the soft tissue infection overlying the surgical site resolved in all of the rabbits (irrespective of the bacterial inoculum of 1×10⁴, 1×10⁵ or 1×10⁶ CFUs) over a few weeks, bacteria could still be recovered from tissue/bone and implant specimens with scanning electron microscopy (SEM), suggesting biofilm formation on the implants. At the same time, bacteria could not be imaged by bioluminescence. Although BLI is a powerful tool to monitor infection and can image as low as 500 CFUs (Bernthal et al., 2010), in this late, low-level infection, bacterial load was minimal and below the limit of detection for BLI. This indicates that the surgical site had ongoing bacterial persistence, which is representative of the clinical scenario whereby the initial IASI might seem to resolve clinically but can later relapse as a chronic infection, owing to the presence of bacteria adherent to the implanted materials (Cho et al., 2018). Taken together, these results establish a biphasic rabbit model of IASI: (1) an early purulent post-surgical soft-tissue infection with wound dehiscence, and (2) a chronic and persistent infection with bone remodeling. Small changes in bone density (as low as 2-3%) are clinically significant and may modify the risk for complication, such as pathological fractures (Bonnick, 2008). In the current study, there was a 4.6% reduction in bone density (measured by CT) in infected versus uninfected vertebrae, further supporting the translational relevance of this model.

In this model, longitudinal imaging with ¹⁸F-FDG was able to monitor the infection-induced inflammation as well as localize the infection (i.e. soft tissue versus bone involvement). In recent years, ¹⁸F-FDG-PET/CT is being increasingly utilized for imaging of infectious diseases (Censullo and Vijayan, 2017). In general, for detection of a vertebral infection, ¹⁸F-FDG-PET/CT imaging has a relatively high sensitivity (Gemmel et al., 2010; Kouijzer et al., 2018). However, since ¹⁸F-FDG is a marker for glucose uptake, any metabolically active tissue is detected and this tracer is therefore non-specific and cannot accurately distinguish between inflammation due to infection versus that caused by non-infectious processes. In the future, novel PET-based imaging tracers specific to the pathogen or components of the inflammatory response may serve as precision medicine tools to facilitate acquisition of temporal and spatial information on infection progression, unachievable by current modalities (Censullo and Vijayan, 2017; Gordon et al., 2019; Hammoud, 2016). The extent and location of infection and inflammatory response could facilitate risk stratification of patients with post-surgical IASI. This is crucial for medical decision making involving possible invasive procedures (surgical removal of infected implanted materials, debridement, etc.), and to direct prolonged parenteral or oral antibiotic therapy (Davis et al., 2009; Lin et al., 2016). As PET-based imaging approaches can be easily translated to clinical use (Ordonez et al., 2019), future applications of this technology may contribute valuable information, including delineating bone involvement and detecting persistence of infection.

This study has some limitations. The study was a proof of concept and group sizes were not determined by power calculations. Evaluation of biofilm formation was performed by SEM. This method is limited because it cannot definitely identify bacteria or components of the extracellular matrix and may, in fact, reflect adherent bacteria. Finally, bone remodeling and changes in bone density can be a result of surgery and/or implants rather than infection alone, but this was not addressed in the current study.

Taken together, this study establishes a rabbit model of IASI in which the surgical approaches and orthopedic-grade implants are used to more closely simulate human IASI. By using advanced in vivo BLI and PET/CT imaging, the in vivo bacterial burden, infection-induced inflammation and bone remodeling could also be measured at acute and chronic time points. This rabbit model could serve a valuable preclinical model of IASI to study the pathogenesis and novel diagnostic and therapeutic modalities prior to studies in larger animals and humans.

The bioluminescent USA300 community-acquired MRSA strain SAP231 was previously derived from the clinical isolate NRS384 (Plaut et al., 2013). SAP231 was previously used in preclinical mouse models, including a prosthetic joint infection model (Bernthal et al., 2010; Pribaz et al., 2012). SAP231 possesses a stable bioluminescent construct integrated into the bacterial chromosome that is maintained in all progeny without selection. Only live and actively metabolizing SAP231 bacteria emit light. SAP231 were streaked onto plates containing tryptic soy broth and Bacto agar (1.5%) (BD Biosciences). Colonies of SAP231 were grown overnight at 37°C in a shaking incubator (240 rpm) in tryptic soy broth. Mid-logarithmic-phase bacteria were obtained after a 2-h subculture of a 1:50 dilution of the overnight culture.

All procedures were approved by the Johns Hopkins Animal Care and Use Committee. A total of fourteen 10- to 16-week-old male Dutch Belted Rabbits (Robinson Services) were used. All rabbits were housed under specific pathogen-free conditions at an American Association for the Accreditation of Laboratory Animal Care (AAALAC)-accredited animal facility at Johns Hopkins University and were used according to procedures described in the Guide for the Care and Use of Laboratory Animals [National Research Council (US) Committee for the Update of the Guide for the Care and Use of Laboratory Animals, 2011].

Of the 14 rabbits used in this study, two were inoculated with 1×10⁴ CFUs and followed by BLI up to day 56 (Fig. 2). Nine rabbits were inoculated with 1×10⁵ CFUs. Of these nine rabbits, three were sacrificed at day 14 for ex vivo CFU enumeration (Fig. 3). Three were sacrificed at day 28 for SEM analysis (Fig. 4) and three were sacrificed at day 56 for CFU enumeration followed by SEM (Figs 3 and 4). Three more rabbits were inoculated with 1×10⁶ CFUs and sacrificed at day 56 for ex vivo CT imaging (Fig. 6). All animals were followed with BLI and two to four rabbits from each inocula group were subjected to ¹⁸F-FDG-PET/CT on post-operative days 7, 21 and 56.

Rabbits were initially anesthetized with intramuscular ketamine (20-30 mg/kg) and xylazine (1-2 mg/kg) and general anesthesia was subsequently maintained with inhalation isoflurane (∼1.5%). Sterile ophthalmic ointment (Rugby, Livonia, MI, USA) was applied to the eyes. For analgesia, sustained-release buprenorphine (0.2 mg/kg) and meloxicam (ZooPharm^®; 0.6 mg) were administered subcutaneously prior to surgery. The surgical procedure is described in Fig. 1. The overlying back skin was shaved and prepped with isopropyl alcohol (70%) followed by povidone-iodine (10%), each applied thrice. A longitudinal incision was made with a 15-blade scalpel in the midline, ∼3 cm in length, over the lower back at the level of the fifth and sixth lumbar vertebrae (L5-L6). An incision of the fascia was then performed in order to expose the L6 vertebra. Using a small rongeur, the entire spinous process, with surrounding musculature and ligaments, was excised from the base, creating a hollow self-contained defect, mimicking a partial laminectomy defect. Orthopedic-grade self-drilling pedicle screws [1.5 mm (width)×4 mm (length)] were used to fix an orthopedic-grade compatible plate (0.6 mm width) between the two transverse processes of the L6 vertebra (both Zimmer Biomet). The surfaces of the screws and plate were then inoculated with either 1×10⁴, 1×10⁵ or 1×10⁶ CFUs of SAP231 in 100 μl PBS. The surgical site and overlying skin was closed with interrupted 3-0 Vicryl sutures (Ethicon). CT imaging was obtained following surgery to validate the correct location of the implants before continuing further experiments.

In vivo BLI was performed to non-invasively monitor the bacterial burden longitudinally over time using a Lumina III IVIS system (PerkinElmer, Hopkinton, MA, USA). Rabbits were imaged on post-operative days 0, 3, 7 and once weekly thereafter for a total of 56 days. Bioluminescent signals were overlaid on a grayscale image of the backs of the rabbits and quantified by using total flux (photons/s) within an oval 6×8 cm region of interest (ROI) using Living Image software (PerkinElmer).

Rabbits were euthanized on days 14 and 56 for ex vivo CFU quantification of bacterial burden. The screws and plate were removed, and adherent bacteria were isolated by sonication in 0.3% TWEEN solution (Sigma-Aldrich) for 10 min followed by vortexing for 2 min, as previously described (Miller et al., 2019). The infected vertebra with adjacent upper and lower vertebrae and surrounding soft tissue were also harvested and homogenized in a blender container with 200 ml PBS, as previously described (Miller et al., 2019). The CFUs were counted after overnight culture with serial dilutions on plates. To further identify any bacteria remaining in the bone and soft tissue or the implants, tissue homogenates and implant sonicates were cultured for an additional 48 h in shaking tryptic soy broth (240 rpm at 37°C) and the presence or absence of CFUs were determined after overnight culture on plates.

Rabbits were euthanized on post-operative days 28 and 56 for SEM, as previously described (Pribaz et al., 2012). Briefly, the ex vivo implants were carefully extracted as not to disturb the surface structures and were immediately fixed in buffered formaldehyde (4%) and glutaraldehyde (2.5%) for 48 h and rinsed with buffer thrice (10 min each). Secondary fixation was performed for 1 h with osmium tetroxide (1%) in PBS. Dehydration was carried out by incubating the implants (15 min each) in increasing concentrations of ethanol (30%, 50%, 67%, 80%, 90%, 100% and 100%). Subsequently, the implants were immersed in decreasing mixtures of ethanol and hexamethyldisilazane (HMDS) (ethanol:HMDS=3:1, 1:1 and 1:3) followed by 100% HMDS dehydration twice (15 min each). Samples were air-dried overnight to ensure evaporation of HMDS. All implants were mounted on aluminum stub mounts, sputter-coated with gold-palladium prior to imaging under a field-emission scanning electron microscope (JSM-6700F FE-SEM; JEOL, Tokyo, Japan) at 35×, 70×, 350× and 1500× magnification.

On post-operative days 7, 21 and 56, rabbits were placed under anesthesia, weighed and 14.04±1.7 MBq ¹⁸F-FDG was injected intravenously via the ear vein. Static images were acquired 45 min following tracer injection. PET images were obtained using a single static 15-min acquisition followed by CT imaging for attenuation correction and anatomical co-registration using the nanoScan PET/CT (Mediso Systems, USA), modified from previously described methods (Davis et al., 2009). PET-CT images were reconstructed and co-registered using VivoQuant 3.5 (InviCRO). 3D spherical volumes of interest were drawn to measure ¹¹F-FDG-PET signal surrounding infected hardware and in an uninfected reference point (anteriorly to L4). SUVs were derived. Uptake into bone and soft tissue was determined based on an ROI surrounding the hardware and using CT-derived thresholds for bone [>700 Hounsfield units (HU)]. The implant was excluded from the images prior to analysis. SUVs were derived and ratios between bone and soft tissue calculated.

Rabbits were euthanized on post-operative day 56 and the spine was harvested from L4 to L7, including all four vertebrae and surrounding soft tissue. Ex vivo CT imaging (Mediso Systems, USA) was performed to assess bone remodeling and density. Two ROIs were drawn per vertebra and the percentage of dense bone volume (HU>3000) divided by the total bone volume (HU>700) are presented (Bibb et al., 2015). The implant was removed prior to imaging.

Prism software (GraphPad, La Jolla, CA, USA) was used to analyze the data. Data are presented as mean±s.e.m. Bone-to-soft tissue SUV ratios and percentages of dense bone were compared by two-tailed Mann–Whitney test. P-values <0.05 were considered statistically significant.