Article focus
Key messages
Strengths and limitations
Introduction
Rapid and accurate diagnosis of bacterial bone infections is essential for early intervention, surgical planning, and rational use of antimicrobial treatments. Commonly used diagnostics, such as pathology, microbiological cultures, and molecular testing, are dependent on invasive sampling of suspected sites of infection. Sampling is prone to error and contamination, may be harmful, and delays the diagnosis.1 This is particularly consequential when diagnosing bone infection, because invasive diagnostic procedures are often incorporated into the curative surgical interventions (e.g., removal of infected bone and/or implant) and thus interventions precede the final diagnosis of bone infection.
Patients with orthopaedic implants are at risk of bone infection. For example, 0.8% to 1.9% of prosthetic joint surgeries develop infection accounting for about 10,000 infections per year in the USA alone.2,3 This number is expected to raise as the number of prosthetic joint surgeries is anticipated to quadruple by 2030.4,5 About 10% to 15% of periprosthetic joint infections (PJIs) involve Enterobacterales.6 Moreover, a particular paediatric population with neuromuscular scoliosis is at high risk (~10%) of developing post-surgical spine infection, and 30% of these involve Enterobacterales.7 Therefore, there is an urgent unmet clinical need to develop noninvasive tools for rapid diagnosis and monitoring of bone infections.1,8,9
While CT and MRI can provide rapid diagnosis of bone pathology, they are non-specific and cannot differentiate infections from non-infectious processes such as inflammatory changes, reactive bone remodelling, and neoplasm. Recently, positron emission tomography (PET) imaging of 2-deoxy-2-[18F]fluoro-D-glucose (abbreviated [18F]-FDG), has been used for detecting infections,10,11 and may be particularly valuable in diagnosing implant-associated infections. Although highly sensitive, [18F]-FDG lacks specificity.
For example, increased uptake is detected even a year after surgery with orthopaedic implants without infection,12,13 likely due to residual inflammation. Finally, all current imaging tools for infections are dependent on host responses that could be significantly altered in immunocompromised states, as well as diabetes.14
Ordonez et al15,16 have developed an approach for detection of bacterial infection by utilizing small molecules to target bacteria-specific metabolic pathways. They showed that sorbitol-based PET imaging is a useful, specific, and readily available method to diagnose and monitor Enterobacterales spp. infection.17 Sorbitol, a sugar alcohol used as a ‘sugar free’ sweetener, is metabolized by bacteria-specific pathways on selective media. It is also approved by the Food and Drug Administration (FDA) as a laxative. [18F]fluoro-d-sorbitol, abbreviated [18F]-FDS, is a radio-labelled form of sorbitol that shows excellent performance and safety, and represents the first radiotracer which is both specific for a bacterial class and readily translatable to humans.15-18
Here, we developed a novel mouse model of implant-associated spinal infection with Enterobacterales spp. and show [18F]-FDS PET-CT provides a useful, rapid, and specific approach for diagnosis and monitoring of infection.
Methods
Bacterial strains
E. coli ATCC 25922 or USA300 bioluminescent Staphylococcus aureus strain SAP231 were used for all experiments. SAP231 was previously derived from the NRS384 community-acquired methicillin-resistant S. aureus isolated from an outbreak in the Mississippi prison system.19 SAP231 has a stably integrated modified luxABCDE operon from the bacterial insect pathogen Photorhabdus luminescens and was used previously to study orthopaedic implant infections.20 Bacteria were streaked onto tryptic soy agar (TSA) plates (tryptic soy broth (TSB; TSA plus 1.5% Bacto Agar)) (BD Biosciences) and grown overnight at 37°C in a bacterial incubator. Two to three colonies were picked up and cultured in TSB at 37°C in a shaking incubator (MaxQ HP 420, Thermo Fisher) (240 rpm) overnight (16 hours), followed by a 1:50 subculture at 37°C for two hours to obtain midlogarithmic phase bacteria.
The bacteria were pelleted, washed three times, and resuspended in sterile phosphate-buffered saline (PBS). The absorbance (optical density 600) was measured to estimate the number of colony-forming units (CFUs), which was verified after overnight culture on TSA plates.
Mice and surgical procedure
All procedures were approved by the Animal Care and Use Committee (ACUC). We have adhered to the ARRIVE guidelines and have included the ARRIVE checklist in the Supplementary Material. We used 48 C57BL/6 male mice aged between ten and 12 weeks in all experiments.
As described previously by Dworsky et al,21 mice were surgically instrumented with ‘L-shaped’ surgical-grade 0.1 mm orthopaedic grade titanium implant (Modern Graining, Custom Wire Technologies, USA) at the level of L4. A 2 cm midline incision was made, and carried down through fascia and muscle to the osseous spine. The dissection was directed laterally along the right side of the spinous processes, visualizing the lateral aspect of the process, and developing a pocket for the implant.
A 25-gauge needle was inserted through the L4 spinous process. The implant was inserted into the spinous process defect. With the short arm of the implant in the spine, the long arm was placed longitudinally along the spine heading cranially.
An inoculation of bacteria (E. coli ATCC 25922 or S. aureus SAP231) was pipetted onto the 90° bend of the implant (1 × 106 CFUs). When indicated, antibiotic treatment with intraperitoneal (IP) ciprofloxacin 20 mg/kg /dose twice daily was started on day of surgery and infection, and continued for five days.
CT imaging
CT imaging was obtained following surgery to validate the exact location of the implants before continuing further analysis. Bone remodelling was evaluated by measuring bone density in Hounsfield units (HU). Spherical volumes of interest (VOIs) were drawn around the implant and surrounding bone.
A threshold of over 700 HU was set to measure bone density, and the mean HU of bone was then calculated. CT imaging was performed multiple times for each animal together with PET imaging and used to measure bone density.
Bioluminescence imaging
In vivo bioluminescence imaging was performed using the IVIS Lumina III Imaging System (PerkinElmer, USA) in live mice at several timepoints to noninvasively monitor bacterial burden.20 Bioluminescent signals were localized on a grayscale image of the mice and quantified within an oval 0.5 × 0.75 cm region of interest using maximum flux (photons per second per square centimetre per steradian) (level of detection = 2.5 × 103 photons s−1 cm−2 sr−1).
Quantification of ex vivo bacterial burden
Mice were euthanized on day 14 or following five days of antibiotic treatment (IP ciprofloxacin 20 mg/kg/dose twice a day) for ex vivo quantification of bacterial burden. Whole tissues (bone and surrounding soft tissues) were homogenized in 1 ml of PBS (Pro200 Series homogenizer; Pro Scientific, USA). Bacteria adherent to the implant were isolated by sonication in 1 ml of 0.3% Tween solution (Sigma-Aldrich, USA) for ten minutes followed by vortexing for two minutes.
The CFUs were counted after overnight culture with serial dilutions on plates of 100 μl of homogenate or sonicate. CFUs were isolated and enumerated after overnight cultures of homogenized bone specimens and solutions from sonicated implants.
PET‐CT imaging with [18F]-FDG and [18F]-FDS in mice
PET‐CT imaging with [18F]-FDG and [18F]-FDS was obtained on the indicated days post-surgery. [18F]-FDG and [18F]-FDS imaging was done on separate days. Mice were placed under anaesthesia and approximately 9.25 MBq of either [18F]-FDG or [18F]-FDS were injected via the tail vein as previously described.16 Static images were acquired at 45 minutes following tracer injection for [18F]-FDG and at 120 minutes following tracer injection for [18F]-FDS, as previously determined.15 PET images were obtained using a single static 15-minute acquisition followed by CT imaging for attenuation correction and anatomical co‐registration. Both PET and CT imaging were performed using the nanoScan PET/CT (Mediso) small-animal imager.
Image analysis
[18F]-FDG and [18F]-FDS PET/CT were visualized using VivoQuant 2020 (Invicro, USA). 3D VOIs were manually drawn using CT as a guide and applied to the PET data.15,16 PET signals in blood (left ventricle), liver, and infected bone (visualized on CT) were measured. The implants (identified on the basis of radiodensity > 5,000 HUs) were excluded from the image analysis in all studies.
Standardized uptake value (SUV) ratios were calculated by dividing the mean SUV of the infected bone by the mean SUV of the heart or the liver, for [18F]-FDS and [18F]-FDG, respectively. To differentiate between inflammation due to the surgical procedure and titanium implant and bacterial infection, results for E. coli and S. aureus were further corrected by dividing them by the results of the non-infected implanted controls (PBS).
Statistical analysis
GraphPad Prism v. 11 (GraphPad Software, USA) was used to compare data among groups. Data are presented as mean (SD).
CFU counts before and after antibiotics were compared using the two-sided Mann-Whitney U test. Bone density and tracer uptake were compared using the Kruskal-Wallis test, adjusted for multiple comparisons to preserve the false-discovery rate. Tracer uptake before and after antibiotics was compared using the one-sided Mann-Whitney U test.
Tracer uptake over 14-day follow-up was compared between groups using the two-way analysis of variance (ANOVA). p-values of 0.05 were considered statistically significant.
Results
A murine model for Enterobacterales implant-associated spinal infection
Following the surgical technique described by Dworsky et al,21 a mouse model for post-surgical spine infection was developed (Figure 1a). First, we re-created a model of Gram-positive infection with S. aureus in C57BL/6 male mice. Weekly, non-invasive in vivo bioluminescence (BLI) demonstrated sustained spine infection over 14 days (Figures 1b and 1c).
On day 14, mice were euthanized, and tissue homogenates and implant-sonicates were cultured to quantify bacterial burden ex vivo, demonstrating significant bacterial burden (Figures 1d and 1e). Next, we adapted this model for Gram-negative infection with E. coli (strain ATCC 25922) using the same surgical technique. Similar to the Gram-positive model, Gram-negative infection resulted in a sustained implant-associated infection as demonstrated by development of skin wounds around the surgical site in the E. coli infected, but not in mice inoculated with PBS (Figure 2a).
Bacterial burden, evaluated ex vivo, showed significant bacterial burden in tissue and implant (Figures 2b and 2c). To evaluate response to therapy in this model, we treated mice with antibiotics (IP ciprofloxacin 20 mg/kg/dose twice a day for five days). The bacterial load was significantly reduced in treated mice (Figures 2b and 2c).
Finally, we evaluated bone remodelling by CT and were able to demonstrate that both S. aureus and E. coli induced substantial bone pathology with reduced bone density compared with mice imaged prior to surgery and infection (Figure 2d). Antibiotic treatment in the Gram-negative model, however, did not mitigate this bone remodelling (Figure 2d).
[18F]-FDS PET/CT imaging of Enterobacterales implant-associated spinal infection
We inoculated mice with either E. coli, S. aureus, or no bacteria (PBS) and performed [18F]-FDS PET/CT (Figure 3a). [18F]-FDS could specifically detect implant-associated spinal infection due to E. coli with mean SUV ratio of 2.2 (SD 1.3), which was substantially lower in S. aureus (0.7 (SD 0.2); p = 0.002, Kruskal-Wallis test) and uninfected mice (1.0 (SD 0.1); Figure 3b).
Comparison of [18F]-FDS and [18F]-FDG PET/CT imaging
To directly compare pathogen specific with non-specific PET imaging, we performed [18F]-FDS and [18F]-FDG PET/CT on the same infected mice. In contrast to [18F]-FDS, imaging with [18F]-FDG could not differentiate between E. coli and S. aureus or the uninfected control (p = 0.497, Kruskal-Wallis test; Figure 4).
[18F]-FDS PET/CT imaging can monitor infection over time and evaluate the response to antibiotic treatment
We induced implant-associated spinal infection with E. coli and treated mice with antibiotics for five days (IP ciprofloxacin 20 mg/kg/dose twice a day). We performed [18F]-FDS PET/CT before and after antibiotic treatment and compared the results (Figure 5a). Following the last antibiotic dose, we euthanized the mice and cultured tissue homogenates and implant-sonicates to quantify the bacterial burden ex vivo.
We compared this to the bacterial inoculum pipetted onto the implant initially, with significant reduction following treatment (p = 0.014, two-way ANOVA; Figure 5b). [18F]-FDS could predict antibiotic efficacy with diminished signal following treatment (p = 0.032, two-way ANOVA; Figure 5b). We next imaged infected mice with [18F]-FDS over a two-week period. [18F]-FDS could accurately monitor infection over this time frame with significantly higher SUV ratios compared to control mice inoculated with S. aureus (p = 0.006, two-way ANOVA; Figure 5c).
Discussion
While Gram-positive bacteria are the leading cause of bone infections, certain patient populations are at increased risk for Gram-negative bone infection, particularly with Enterobacterales spp., which represent the largest group of bacterial pathogens in humans.2,6,7 Current diagnosis and treatment of bone infection is challenging, and often necessitates surgical removal of the bone or implant, resulting in severe disability. While the sensitivity of MRI to detect bone involvement is high (96%), specificity is suboptimal (84%).22 Current nuclear medicine tools are cumbersome, and non-specific. For example, imaging of 111In-oxine-tagged white blood cells using single-photon emission CT (SPECT).12 Furthermore, labelled leucocytes may have limited penetration into diseased tissues.23,24 [18F]-FDG-PET is not affected by metallic implant artifact and, when compared with other radionuclide approaches, has superior spatial resolution, sensitivity, and specificity.12,25 It can help define the anatomical site of infection and facilitate surgical planning,26 which may be improved in combination with CT imaging (PET/CT).27 Even so, [18F]-FDG is not specific, and may show increased uptake a year after surgery with orthopaedic implants, even with no infection.12,13
In previous studies, in vitro uptake assays indicated that [18F]-FDS accumulated selectively in Enterobacterales and reached ~ 1,000-fold higher concentrations than in Gram-positive bacteria, mammalian cells, and cancer cells.15,16 Moreover, it has been shown that uptake of [18F]-FDS was significantly lower in E. coli mutant strains with an impaired sugar-transporting phosphotransferase system.17 In a murine myositis model, [18F]-FDS PET rapidly differentiated infection from sterile inflammation with a mean limit of detection of 6.2 log10 CFUs (SD 0.2) for E. coli,16 similar to the bacterial burden that can be found in human skin or bone infections.28,29 Furthermore, in a phase I clinical study by Zhu et al,30 [18F]-FDS PET was safe, well tolerated, and had a low background signal in normal tissues in six healthy subjects. In a phase I-II clinical study, whole-body [18F]-FDS PET/CT was performed in 26 prospectively enrolled patients with either microbiologically confirmed Enterobacterales infection or other pathologies (e.g., cancer, inflammatory diseases).17 The study found that [18F]-FDS PET/CT was safe, could rapidly detect and localize Enterobacterales infections due to drug-susceptible or drug-resistant strains, and differentiated them from sterile inflammation or cancerous lesions.17 Finally, this aforementioned clinical study showed that [18F]-FDS PET monitored the efficacy of antimicrobial treatment, demonstrating a signal proportional to the bacterial burden.17
Here, we demonstrate for the first time the feasibility of [18F]-FDS PET/CT imaging for detection and monitoring of bone infections. In a mouse implant-associated spine infection model, 18F-FDS PET/CT could diagnose a bacterial infection and differentiate Gram-negative from Gram-positive bacteria (Figure 3). This was not the case for [18F]-FDG that was non-specific for either infection or pathogen (Figure 4).
Moreover, [18F]-FDS could evaluate the response to antibiotic therapy and monitor infection progression over time (Figure 5).
While this is the first reported validation of [18F]-FDS PET/CT for detection of bone infection, this study has some limitations. First, we evaluated only E. coli (compared with S. aureus). Further studies are needed to validate this novel method for other Enterobacterales species involved in bone infection.
It should be noted, however, that we previously showed [18F]-FDS uptake by many other bacterial species (Klebsiella pneumoniae, Salmonella spp., Enterobacter cloacae, Yersinia enterocolitica, etc.).16 Second, given the small size of vertebrae in our mouse model, we could not determine if [18F]-FDS differentiated between superficial skin and soft-tissue infection alone from involvement of deep-seated structures including bone and implant. However, this should not be an issue in clinical studies. Third, given that all procedures were done within 14 days of surgery and infection, this model represents an acute infection.
Further studies are required to evaluate the relevance to chronic implant-associated infections. Finally, we did not evaluate other types of bone infection, in particular diabetic ulcer-associated bone infection. This is important, as diabetic foot infections account for approximately 25% of hospitalizations in diabetic patients, 20% of whom have underlying osteomyelitis.31,32 This results in over 200,000 amputations every year in the USA alone.33-35Enterobacterales spp. is involved in at least 30% of diabetic foot infections,36 and in up to 50% if bone is infected.37
Improved synthesis of the tracer further highlights feasibility. Li et al38 published a scheme to rapidly (30 mins) synthesize [18F]-FDS from the readily and commercially available [18F]-FDG, by chemical reduction. Later, Mota et al17,18 developed a self-contained, solid-phase cartridge for an even faster (< 10 mins) production of ready-to-use [18F]-FDS from commercially available [18F]-FDG at room temperature.
With these advances, [18F]-FDS may readily be available at any medical centre performing [18F]-FDG PET scans.
Establishing a rapid bacterial-specific PET-based imaging modality to determine bone involvement in implant-associated infection and diabetic ulcers would better inform medical decision-making, reduce the need for surgical interventions, and limit unnecessary antibiotic exposure in a large and growing patient population. Further studies are needed to validate this novel method in human patients and determine its sensitivity, specificity, and clinical usefulness.