Establishment of a Novel Rat Model of Gram-Negative Periprosthetic Joint Infection Using Cementless Hip Hemiarthroplasty

Background: Gram-negative periprosthetic joint infections (GN-PJIs) present unique challenges. Our aim was to establish a clinically representative GN-PJI model that recapitulates biofilm formation in vivo. We also hypothesized that biofilm formation on the implant surface would affect its ability to osseointegrate.

Methods: Three-dimensionally-printed medical-grade titanium hip implants were used to replace the femoral heads of male Sprague-Dawley rats. GN-PJI was induced using 2 bioluminescent Pseudomonas aeruginosa strains: a reference strain (PA14-lux) and a mutant biofilm-defective strain (ΔflgK-lux). Infection was monitored in real time using an in vivo imaging system (IVIS) and magnetic resonance imaging (MRI). Bacterial loads were quantified utilizing the viable colony count. Biofilm formation at the bone-implant interface was visualized using field-emission scanning electron microscopy (FE-SEM). Implant stability, as an outcome, was directly assessed by quantifying osseointegration using microcomputed tomography, and indirectly assessed by identifying gait-pattern changes.

Results: Bioluminescence detected by the IVIS was focused on the hip region and demonstrated localized infection, with greater ability of PA14-lux to persist in the model compared with the ΔflgK-lux strain, which is defective in biofilm formation. This was corroborated by MRI, as PA14-lux induced relatively larger implant-related abscesses. Biofilm formation at the bone-implant interface induced by PA14-lux was visualized using FE-SEM versus defective-biofilm formation by ΔflgK-lux. Quantitatively, the average viable colony count of the sonicated implants, in colony-forming units/mL, was 3.77 × 108 for PA14-lux versus 3.65 × 103 for ΔflgK-lux, with a 95% confidence interval around the difference of 1.45 × 108 to 6.08 × 108 (p = 0.0025). This difference in the ability to persist in the model was reflected significantly on implant osseointegration, with a mean intersection surface of 4.1 × 106 ± 1.99 × 106 μm2 for PA14-lux versus 6.44 × 106 ± 2.53 × 106 μm2 for ΔflgK-lux and 7.08 × 106 ± 1.55 × 106 μm2 for the noninfected control (p = 0.048).

Conclusions: To our knowledge, this proposed, novel in vivo biofilm-based model is the most clinically representative for GN-PJI to date, since animals can bear weight on the implant, poor osseointegration was associated with biofilm formation, and localized PJI was assessed by various modalities.

Clinical relevance: This model will allow for more reliable testing of novel biofilm-targeting therapeutics.