3D printed bioceramics for dual antibiotic delivery to treat implant-associated bone infection

Abstract

Surgical implant-associated bone infections (osteomyelitis) have severe clinical and socioeconomic consequences. Treatment of chronic bone infections often involves antibiotics given systemically and locally to the affected site via poly(methyl methacrylate) (PMMA) bone cement. Given the high antibiotic concentrations required to affect bacteria in biofilm, local delivery is important to achieve high doses at the infection site. PMMA is not suitable to locally deliver some biofilm-specific antibiotics, including rifampin, due to interference with PMMA polymerization. To examine the efficacy of localized, combinational antibiotic delivery compared to PMMA standards, we fabricated rifampin- and vancomycin-laden calcium phosphate scaffolds (CPS) by three-dimensional (3D) printing to treat an implant-associated Staphylococcus aureus bone infection in a murine model.

All vancomycin- and rifampin-laden CPS treatments significantly reduced the bacterial burden compared with vancomycin-laden PMMA. The bones were bacteria culture negative in 50% of the mice that received sustained release vancomycin- and rifampin-laden CPS. In contrast, 100% of the bones treated with vancomycin monotherapy via PMMA or CPS were culture positive. Yet, the monotherapy CPS significantly reduced the bacterial metabolic load following revision compared to PMMA. Biofilm persisted on the fixation hardware, but the infection-induced bone destruction was significantly reduced by local rifampin delivery.

These data demonstrate that, despite the challenging implant-retaining infection model, co-delivery of rifampin and vancomycin from 3D printed CPS, which is not possible with PMMA, significantly improved the outcomes of implant-associated osteomyelitis. However, biofilm persistence on the fixation hardware reaffirms the importance of implant exchange or other biofilm eradication strategies to complement local antibiotics.

Introduction

The biofilm-forming species Staphylococcus aureus (S. aureus) is the most common pathogen in implant-associated infections. These infections are extremely difficult to treat due to the ability of the pathogens in biofilm to resist the immune response and antibiotic treatments, leading to the highest re-infection rates (Darouiche, 2004; Keren et al., 2004; Rani et al., 2007; Yun et al., 2008; Zimmerli et al., 2004). Most often, these infections are associated with bone fracture fixation devices, such as plates and screws, which account for over two-thirds of all surgical implant infections with an estimated 100,000 cases per year in the U.S. alone, and a cost of $15,000-$50,000 per incident (Darouiche, 2004). When revision surgery is required for infection management, surgeons sometimes elect to retain the orthopedic implant for an acute (< 3 weeks) infection case if it is biomechanically stable within the bone, the soft tissue is in good condition, and the organism is identified to be susceptible to biofilm-relevant antibiotics (Darouiche, 2004; Konig et al., 2001; Trampuz and Widmer, 2006; Trebse et al., 2005; Widmer et al., 1992; Zimmerli et al., 1998). In a previous prospective study in which 87 patients with orthopaedic implant-related infections were identified, 28% of the patients met the inclusion criteria for treatment with retention of the implant (including hip, knee, and ankle prostheses as well as internal fracture fixation devices), which resulted in a probability of survival without treatment failure of 86% at three years post revision (Trebse et al., 2005). However, retaining a device that may be contaminated with biofilm can pose an increased challenge for treatment. The clinical algorithms typically recommend multiple-staged surgical interventions, beginning with meticulous debridement of the infected bone and soft tissue, coupled with localized and systemic antibiotics to treat the infection before bone reconstructive procedures can be pursued. Local antibiotic administration has the advantage of augmenting the systemic therapy by improving the antibiotic bioavailability directly at the infection site (Darouiche, 2004; Mihailescu et al., 2014; Tang et al., 2013; Valour et al., 2014; Zimmerli et al., 2004). Antibiotics are typically delivered locally via poly(methyl methacrylate) (PMMA) bone cement spacers (Darouiche, 2004; Jaeblon, 2010). However, PMMA spacers are compatible with only a limited number of antibiotics and have poor release profiles that affect the drug’s bioavailability.

Rifampin is a critical antibiotic for management of Staphylococci in biofilm (Mihailescu et al., 2014; Tang et al., 2013), especially in cases where the implant is retained (Darouiche, 2004; Zimmerli et al., 2004). Unfortunately, PMMA spacers are not suitable for localized delivery of rifampin because this antibiotic impairs the polymerization process by acting as a free radical scavenger (Beeching et al., 1986). Alternative biomaterials for local rifampin delivery, which include polymers (O'hara and Hickey, 2000; Waeiss et al., 2014), ceramics (Molina-Manso et al., 2012) and composites (Ahola et al., 2012; Gu et al., 2012), have been investigated in vitro, but the potential therapeutic efficacy of these rifampin-laden biomaterials has not yet been demonstrated in vivo. Furthermore, the in vivo efficacy of local rifampin delivery has previously been evaluated only as a prophylactic implant coating (Darouiche et al., 2007), but not as a treatment for an established bone infection with biofilm involvement.

Calcium phosphates (CaP) are a class of ceramics that have been investigated as regenerative scaffolds for hard tissues and as drug delivery materials in orthopedic and dental applications (Bose and Tarafder, 2011; Ginebra et al., 2006). CaP are amenable to additive fabrication technologies such as three-dimensional (3D) printing based on the concept of selective granular fusion or binding, which can enable the creation of medical imaging-guided, patient-specific bone substitutes (Klammert et al., 2010), and potentially the incorporation of drugs and bioactive factors. However, sintering of 3D printed CaP materials at ultra high temperatures, which might be required to increase their strength, can obliterate the bioactivity of incorporated drugs. In contrast, low temperature 3D printing of CaP allows for fine spatial distribution of “bioink” (drug) within the printed geometry (Vorndran et al., 2010). Vorndran and colleagues demonstrated direct addition of vancomycin, an important antibiotic against methicillin-resistant S. aureus (MRSA), into CaP during the low temperature 3D printing process, which utilized a highly acidic binder (20 wt% phosphoric acid) (Vorndran et al., 2010). However, while the Vorndran study demonstrated a proof of concept, we have recently shown that the highly acidic binding solution results in residual acidity that can be highly cytotoxic (Inzana et al., 2014), and can significantly degrade the bactericidal activity of the incorporated drug.

In this study, we test the hypothesis that utilizing a low acidity binder solution would enable low-temperature 3D printing of vancomycin- and rifampin-laden CPS while maintaining the bactericidal activity; and that these dual-antibiotic CPS would significantly reduce the bacterial burden and osteolytic bone loss of an established implant-associated bone infection compared with vancomycin-laden PMMA, which is considered a clinical standard of care. This hypothesis was studied using a previously developed mouse model of established implant-associated bone infection (Inzana et al., 2015). Prior to treatment, this murine model of established bone infection mimics the salient aspects of the challenging clinical conditions of implant-related osteomyelitis, including biofilm and bone loss. This mouse model was designed to simulate the clinical algorithms for implant retention, as described above, considering that a revision surgery takes place 7 days post-infection, the fixation plate remains mechanically stable, and an extensive debridement of both the bone and soft tissue is performed, along with a local and systemic antimicrobial treatment, as previously described (Inzana et al., 2015).

Materials and Methods

Preparation of 3D printed antibiotic-laden CPS and PMMA materials

A modified ZPrinter® 450 (3D Systems; Andover, MA) was used to 3D print the CPS materials as previously described (Inzana et al., 2014). In this method, a phosphoric acid-based solution is selectively sprayed via inkjets that raster over a layered bed of calcium phosphate powder to initiate a dissolution-precipitation reaction, which forms 3D CPS after repeated layer-by-layer binding (Gbureck et al., 2007b). The calcium phosphate powder was 99% phase pure α-tricalcium phosphate (α-TCP; InnoTERE GmbH; Radebeul, Germany). The powder was bound during the 3D printing process by delivering a solution of 8.75 wt% phosphoric acid + 0.25 wt% Tween 80 from the inkjets, which was previously demonstrated to produce cytocompatible and osteoconductive scaffolds (Inzana et al., 2014). The powder layer thickness was set to 89 μm and the binder liquid-to-powder ratio was set as 0.46 in the ZPrint™ software. The CPS materials were supplemented with antibiotics by either mixing vancomycin (Alfa Aesar; Ward Hill, MA) or rifampin (Sigma-Aldrich; St. Louis, MO) powder into the α-TCP or by infusing an antibiotic solution into the CPS via a second set of ‘color’ inkjets. Vancomycin was mixed into the powder to achieve a concentration of 10 wt%, which yielded CPS with 5.2 wt% vancomycin (~365 μg/CPS; Vanco_5%-CPS). Rifampin was mixed into the powder substrate at a concentration of 1 wt% to yield CPS with 0.5 wt% rifampin (~35 μg/CPS; Vanco_5%+Rif_0.5%-CPS or Rif_0.5%-CPS). Alternatively, a solution of 10 mg/mL rifampin dissolved in 50% DMSO/H₂O vol/vol was delivered via the inkjets to achieve a concentration of 0.05 wt% rifampin (~3.5 μg/CPS; Vanco_5%+Rif_0.05%-CPS or Rif_0.05%-CPS). To create a sustained antibiotic release formulation, Vanco_5%+Rif_0.5%-CPS were coated with poly(D,L-lactide-co-glycolide) (PLGA; Vanco_5%+Rif_0.5%-CPS/PLGA). The PLGA (50:50 L:G, MW 24k-38k; Sigma-Aldrich; St. Louis, MO) was dissolved into dichloromethane at a concentration of 5, 12.5, or 20 wt% in combination with 20 mg/mL rifampin. The CPS were soaked in this solution for 1 hour to adsorb the rifampin-laden PLGA and were then dried at 40°C in a vacuum chamber overnight to evaporate the volatile solvent. The CPS/PLGA composites that were implanted into mice used 12.5 wt% PLGA. PMMA bone cement (SmartSet® MV Bone Cement; DePuy Synthes, Warsaw, IN) samples were prepared by hand mixing according to the manufacturer’s instructions as a placebo (PBO-PMMA) or mixed with 5 wt% vancomycin (0.5 mg per murine spacer; Vanco_5%-PMMA) and formed in a custom mold.

Characterization of antibiotic release and functionality from CPS and PMMA materials

The total antibiotic loaded into the 3D printed CPS was measured by dissolving scaffolds in 3M HCl and measuring the concentration of vancomycin and rifampin (n = 6 samples/group) by optical absorbance at 280 nm using a spectrophotometer (Synergy Mx Microplate Reader, BioTek Instruments Inc.; Winooski, VT). The limit of quantification at 280 nm was 7 μg/mL for vancomycin and 3 μg/mL for rifampin.

Antibiotic functionality was measured through a Kirby-Bauer disc diffusion assay (n = 6 samples/group measured in duplicate). Vancomycin and rifampin were eluted into PBS from CPS or PMMA samples and the concentrations were measured by optical absorbance at 280 nm after passing the eluate through a 0.2 μm filter. Fresh antibiotic solutions that spanned the concentrations of interest were used to generate a linear standard curve. Filter paper discs (7 mm diameter) were placed into 1 mL of each solution overnight at 4°C to adsorb the antibiotics. The discs were dabbed dry and then placed onto Luria agar plates, which were inoculated with a bioluminescent strain of methicillin-sensitive S. aureus (Xen36; PerkinElmer, Inc.; Waltham, MA), and the plates were incubated overnight at 37°C. The diameters of the bacterial inhibition zones surrounding each disc were measured and the active antibiotic concentration was determined by comparing to the linear standard curve as previously described (Bonev et al., 2008). Statistically significant differences in the active antibiotic concentration were assessed by comparing to the non-printed control antibiotic solution through one-way ANOVA with Tukey’s test for multiple comparisons.

To measure the antibiotic release kinetics, 1.8 mm diameter × 3 mm murine femoral scaffolds were 3D printed with either vancomycin or rifampin independently (n = 6 samples/group). The cylinders were placed into sealed vials with 0.5 mL of PBS in static conditions at 37°C. The cylinders were transferred to new vials with fresh PBS after 1, 2, 4, 8, and 24 hours and then daily for up to 2 weeks. The vancomycin or rifampin concentrations were measured by optical absorbance at 280 nm.

Biomechanical properties of 3D printed antibiotic-laden CPS materials

The flexural properties of the 3D printed CPS were assessed by 3 point bending to determine if supplementing these materials with antibiotics affects the mechanical strength (n = 6-10/group). Samples were 3D printed as 4 × 1 × 14 mm bars and tested on a 10 mm support span using an Instron 8841 DynaMight™ Axial Testing System (Instron Corp.; Canton, MA) with a 50 N load cell. Flexural tests were performed at a displacement rate of 1 mm/min until failure.

Animals and surgical procedures

All animal studies were performed in accordance with protocols approved by the University of Rochester’s Committee on Animal Resources. Female BALB/cJ mice were purchased from Jackson Research Labs (Bar Harbor, ME) at 13-15 weeks of age and acclimated for one week prior to surgery. The surgeries were performed according to a previously described model (Inzana et al., 2015). Briefly, the right femur was exposed by a direct lateral approach and a 6-hole radiolucent polyether ether ketone (PEEK) plate with a 40 nm titanium coating was installed across the anterolateral surface using 4 titanium screws (RISystem; Davos, Switzerland). To prevent septic loosening from osteolysis around the screws, the fixation was reinforced in a cerclage-fashion by tying a 5-0 nylon monofilament suture around the bone and plate in between each pair of proximal and distal screws. A 0.7 mm transverse osteotomy was cut through the femoral mid-diaphysis using a 0.67 mm wire gigli saw and a cutting guide (RISystem; Davos, Switzerland). A 1 mm radius semi-circle of fibrillar collagen sheet (Kensey Nash; Exton, PA) that was soaked for at least 2 hours in an overnight culture of bioluminescent methicillin-sensitive S. aureus (Xen36; PerkinElmer, Inc.; Waltham, MA) was placed into the defect. Xen36 was modified from the parental strain ATCC 49525, which is a clinical isolate from a bacteremia patient, to contain the luxABCDE operon. Xen36 was previously found to be a good candidate for longitudinal imaging in murine models of chronic implant-associated osteomyelitis (Pribaz et al., 2012) and was confirmed to form robust biofilms in vivo on the plates and screws of this murine model (Inzana et al., 2015). The viable bacterial load on each collagen sheet was previously determined to be at least 8.0 ± 2.9 × 10⁴ colony forming units (CFU) (Inzana et al., 2015). The muscle and skin were closed with 5-0 nylon monofilament sutures and the infection was allowed to establish for 7 days until the revision surgery (Figure 3a). At the time of revision, the infected tissue was debrided and the bone defect was widened to 3 mm to place a 1.8 mm diameter × 3 mm antibiotic CPS or PMMA spacer into the defect (n = 8-15/group; Figure 3b). One mouse from the Vanco_5%-PMMA group was over-infected during the inoculation procedure and was therefore excluded, resulting in n = 7 for this group. The muscle and skin were closed and the mice were allowed to heal for an additional 21 days. Select groups also received systemic vancomycin beginning at the revision surgery (110 mg/kg twice daily subcutaneously (SC Vanco)) until sacrifice. Upon tissue harvest, the bone, soft tissue, fixation hardware, and antibiotic implants were placed into separate vials with 1 mL of sterile PBS and frozen at −80°C for CFU analysis.

Figure 3.

Low dose local antibiotic delivery via 3D printed CPS reduced the bacterial burden after revision. (a) A full-thickness femoral osteotomy (to simulate a fracture) is stabilized with a fixation plate and screws and then infected via a collagen sheet (white triangle) inoculated with a bioluminescent strain of methicillin-sensitive S. aureus (Xen36). (b) Seven days after the initial osteotomy and the establishment of infection, an implant-retaining revision surgery is performed, in which the infected tissue is debrided, infected screws 3 and 4 are replaced with sterile screws 2 and 5, and an antibiotic-laden spacer (yellow arrow) is placed within the critical bone void. (c) Local antibiotic treatment with Vanco_5%+Rif_0.05%-CPS significantly reduced the bacterial load compared with Vanco_5%-PMMA, independent of subcutaneous vancomycin (SC Vanco) administration as observed through longitudinal bioluminescent imaging (BLI). The area under the radiance curve (AUC) following revision surgery further demonstrates the reductions in bacterial metabolic load with each treatment (d). *denotes multiplicity-adjusted p<0.05 for the effect of Vanco_5%+Rif_0.05%-CPS vs. Vanco_5%-PMMA by 2-way ANOVA. ^#denotes multiplicity-adjusted p<0.05 vs. Vanco_5%-PMMA after ANOVA with Dunnett’s test for post-hoc comparisons. n = 7-15/group.

Bioluminescent imaging of the active bacterial infection

The Xen36 bacterial infection was evaluated longitudinally through bioluminescent imaging (BLI) using an IVIS Spectrum imaging system (PerkinElmer, Inc.; Waltham, MA) with a five-minute dorsal exposure. The average luminescent radiance was quantified through the Living Image software (PerkinElmer, Inc.) by using a fixed-size region of interest (ROI) around the right thigh of each mouse. The area under the radiance curve post-revision surgery (Post-Rev Radiance AUC) was computed to quantify the pathogenic burden of the infection.

Radiographic imaging and analysis of osteolysis and reactive bone formation

To monitor changes to the bone, X-rays (LX-60 X-ray Cabinet, Faxitron Bioptics LLC; Tucson, AZ) were taken at 26 kV with an exposure time of 5 seconds. One day after the revision surgery (day 8) and on the final day of the study (day 28), the right femur of each mouse was scanned in vivo by micro-computed tomography (micro-CT; VivaCT 40, Scanco Medical; Bassersdorf, Switzerland) with a 17.5 μm isotropic voxel size using an integration time of 300 ms, energy of 55 kV, and intensity of 145 μA. 3D reconstruction and quantitative analysis of the micro-CT images was performed using Amira software (Amira 5.4.5, FEI Visualization Sciences Group; Burlington, MA). For each mouse, micro-CT images that were collected on day 8 were co-registered with images collected on day 28 to measure the volumetric changes to the bone. The proximal ROI and distal ROI encompassed the bone from the edge of the 3 mm osteotomy to the nearest screw on the proximal and distal sides of the defect, respectively (Figure 6a). The images were binarized using a global bone mineral density threshold of 435 mg HA/cm³ to measure the volumetric bone changes (reactive bone formation as well as osteolytic resorption), which were normalized to the original bone volume at day 8 in each ROI.

Figure 6.

Local rifampin therapy and systemic vancomycin significantly reduced the infection-induced osteolysis. (a) Micro-CT images collected on day 8 were co-registered with images collected on day 28 to measure the volumetric changes to the bone in two ROIs encompassing the osteotomy to the nearest screw on the proximal and distal sides of the defect, respectively, to measure the volumetric bone changes (osteolytic resorption (red) and reactive bone formation (blue)). Treatment with Vanco_5%+Rif_0.05%-CPS and SC Vanco each resulted in significantly reduced osteolytic bone resorption (b) and more positive net bone changes (d) after 3 weeks of treatment (day 28 of the study). The levels of reactive bone formation were not significantly different (c). *denotes p<0.05 for the effect of Vanco_5%+Rif_0.05%-CPS vs. Vanco_5%-PMMA by 2-way ANOVA. ^#denotes p<0.05 for the effect of SC Vanco by Sidak’s test after 2-way ANOVA with a significant interactive effect. Vanco_5%-CPS was not significantly different from Vanco_5%-PMMA by Dunn’s test after nonparametric ANOVA (Kruskal-Wallis) in each metric. n = 7-15/group.

Scanning electron microscopy (SEM) imaging of the implants

The plate and screws from each mouse were fixed in 2.5% glutaraldehyde/4% paraformaldehyde in 0.1M Cacodylate buffer, post-fixed in 1% osmium tetroxide, dehydrated, and gold sputter coated. The samples were imaged by SEM (LEO 982 FE-SEM, Carl Zeiss SMT; Thornwood, NY) for qualitative assessment of bacterial colonization and biofilm formation on the implant surfaces.

Measurement of bacterial colonization

To measure the viable bacterial load remaining in the bone and soft tissue, each sample was thawed and homogenized in 3 mL of sterile PBS. Residual bacteria were dislodged from the fixation hardware and antibiotic implants by sonicating the samples in 1 mL sterile PBS for 3 minutes at 35 kHz (VWR Symphony Ultrasonic Cleaner; VWR Intl.; Radnor, PA) and then vortexed. The tissue homogenates and sonicated suspensions were spread on Luria agar plates in 10-fold dilutions and incubated overnight at 37°C for CFU counts, which were normalized to the mass of each sample. CFU data were transformed using log₁₀(1+X) to achieve normal distributions that retain zero values.

Measurement of functional antibiotics from retrieved spacers

Following sonication of the antibiotic implants that were retrieved after 21 days of implantation in the mice, the CPS and PMMA spacers were vortexed in 1 mL PBS to elute any residual antibiotics. Two filter paper discs (7 mm diameter) were placed into each vial of eluate overnight at 4°C. The discs were placed onto Xen36 inoculated Luria agar plates and incubated for 12 hours at 37°C. The plates were imaged by BLI to measure the bacterial viability by placing a 7 mm diameter region of interest around each disc.

Data analysis

Two in vivo studies were performed. First, Vanco_5%-PMMA was compared against Vanco_5%+Rif_0.05%-CPS, with or without SC Vanco, in a two-way ANOVA with Sidak’s test for multiple comparisons when the interactive term was significant. Second, Vanco_5%-PMMA was compared with the lower rifampin dose (Vanco_5%+Rif_0.05%-CPS), the higher rifampin dose (Vanco_5%+Rif_0.5%-CPS), and the sustained release formulation (Vanco_5%+Rif_0.5%-CPS/PLGA), all with concomitant SC Vanco, using ANOVA with Dunnett’s post-hoc test for multiple comparisons against Vanco_5%-PMMA if the ANOVA was significant. Data were confirmed to satisfy assumptions of a Gaussian distribution based on the D'Agostino-Pearson normality test or a nonparametric alternative was used where appropriate. Considering the lack of a nonparametric alternative for two-way ANOVA and the bimodal distribution of CFU data (due to culture negative samples), Fischer’s exact test was also performed to complement the two-way ANOVA tests of the CFU data. The 2×2 contingency table included mice that received Vanco_5%-PMMA and those that received Vanco_5%+Rif_0.05%-CPS local treatment, regardless of whether SC Vanco was also administered. The lowest CFU measurement in the Vanco_5%-PMMA group was set as the threshold for binarizing the outcome data. For assays where a culture negative sample was observed in the Vanco_5%-PMMA group, the data was analyzed based on the number of culture negative samples. Bonferonni’s correction was applied to control the family-wise error rate for the multiple ANOVAs conducted at each time point over the BLI time courses. Differences were considered significant for multiplicity-adjusted p < 0.05. All p values presented are multiplicity-adjusted.

Results

3D printing and physical characterization of antibiotic-laden CPS

Antibiotic-laden CPS were 3D printed using the previously optimized lower acidity binding solution (8.75% phosphoric acid + 0.25% Tween 80) (Inzana et al., 2014). The 3D printed CPS had a coarser surface morphology and increased porosity compared to PMMA spacers that are typically used in the management of osteomyelitis (Figure 1). The bioactivity of the antibiotics within the low acidity 3D printed CPS was retained, whether vancomycin in solution was infused via the ‘color’ inkjets (82% ± 2% bioactive; Figure 2a) or vancomycin powder was mixed into the α-TCP powder (70% ± 11% bioactive; Figure 2b). In contrast, a highly acidic (20 wt% phosphoric acid) binding solution significantly degraded the bactericidal activity of the printed vancomycin (30% ± 9% active; p < 0.0001; Figure 2a). Since PMMA tends to sequester the majority of impregnated antibiotic, the 3D printed CPS delivered over 6-fold more functional vancomycin than equivalently loaded PMMA (Figure 2c). Further, the addition of antibiotics did not compromise the mechanical strength of the CPS materials (Table 1). Rifampin also retained 35 to 50% of its bactericidal activity when mixed into the α-TCP powder (0.5 wt%) or infused via the inkjets (0.05 wt%) and 3D printed with the lower-acidity binder solution (Figure 2d). In all subsequent in vivo experiments, rifampin was always delivered in combination with vancomycin, because the bacteria can develop resistance to rifampin very rapidly when it is used as a monotherapy (Vergidis et al., 2011; Zimmerli et al., 2004).

Figure 1.

Differences in the surface m01phology and porosity between 3D printed CPS and PMMA spacers. Macroscopic and microscopic (SEM) images of 3D printed CaP scaffolds loaded with vancomycin and rifampin compared to a manually molded PMMA spacer mixed with vancomycin.

Figure 2.

Antibiotics retain their bactericidal activity after 3D printing into calcium phosphate scaffolds (CPS). Bactericidal activity was detennined by the size of the bacterial inhibition zone in Kirby-Bauer disc diffusion assays of antibiotic eluate delivered via filter paper discs (inset in a), compared against linear standards . (a) Vancomycin retains up to 80% bactericidal activity when infused into 3D printed CPS via the inkjets in parallel with an 8.75 wt-0/o phosphoric acid binding solution. * denotes p<0.05 vs. Not-printed by Tukey 's test after ANOVA (n=6/group). (b) Bactericidal activity of vancomycin is also retained when the antibiotic is mixed into the calcium phosphate powder phase (Vancos%-CPS) and 3D printed with 8.75 wt% phosphoric acid + 0.25% Tween 80 as the binder solution. (c) Vancos%-CPS released over 6-fold more functional vancomycin than loaded PMMA (Vancos%-PMMA) . (d) Rifampin bactericidal activity is also retained with this binder solution when the CPS are supplemented with rifampin via the inkjets (Rifoos%-CPS) or by mixing into the powder phase (Rifo.s%-CPS). Data presented as mean ± standard enor of the mean; n = 6/group).

Table 1.

Flexural Properties of Antibiotic-laden CPS

	Maximum Stress (kPa)	Elastic Modulus (MPa)	Toughness (mJ/cm3)
CPS	63.0 ± 4.8	18.9 ± 3.7	0.12 ± 0.03
Vanco5%-CPS	66.7 ± 8.9	17.4 ± 3.4	0.18 ± 0.05
Vanco5%+Rif0.05%-CPS	62.3 ± 11.1	16.6 ± 4.0	0.14 ± 0.04
Vanco5%+Rif0.5%-CPS	69.1 ± 9.2	19.1 ± 4.0	0.16 ± 0.05
Vanco5%+Rif0.5%-CPS/PLGA	3134 ± 586	721 ± 90	9.9 ± 3.6

Flexural strength was measured by 3-point bending with 1 × 4 × 14 mm bars on a 10 mm support span at a displacement rate of 1 mm/min.

Values are represented as means and standard deviation. n ≥ 6 per group.

In vivo efficacy of low dose rifampin- and vancomycin-laden CPS scaffolds

The pathogenic burden, which was measured longitudinally using BLI, was significantly reduced by the Vanco_5%+Rif_0.05%-CPS treatment compared with Vanco_5%-PMMA on days 8-12 (1700 ± 240 vs. 8770 ± 4720 photons/s/cm²/sr, respectively at day 10; p = 0.001 by 2-way ANOVA; Figure 3c) and in the overall area under the post-op BLI curve (41000 ± 8000 vs. 64000 ± 36000 photons/s/cm²/sr×day; p = 0.0001 by 2-way ANOVA; Figure 3d). Further, only mice treated with Vanco_5%+Rif_0.05%-CPS local therapy had a significantly reduced BLI signal from Day 7 post-op through Day 10 (mean change of −2957 ± 2747 photons/s/cm²/sr; p = 0.0003 by one sample t-test vs. zero; Figure 3c). For results regarding the effect of the local treatment strategy (Vanco_5%+Rif_0.05%-CPS vs. Vanco_5%-PMMA) from 2-way ANOVA, the mean and SD values and percent of culture negative samples reported in this section are pooled from the groups of each local treatment, irrespective of SC Vanco treatment. The Vanco_5%-CPS treatment also significantly reduced the overall area under the post-op BLI curve relative to Vanco_5%-PMMA without SC Vanco (53000 ± 7000 vs. 94000 ± 48000 photons/s/cm²/sr×day; p = 0.0003 by ANOVA; Figure 3d). Quantitative bacterial colonization determined by CFU assays confirmed the findings from the longitudinal BLI. The bones (Figure 4a) and soft tissues (Figure 4b) from animals that received Vanco_5%+Rif_0.05%-CPS local treatment (regardless of SC Vanco) were bacteria culture negative in 19% (3/16) and 63% (10/16) of the mice, respectively, compared with 0% (0/22) and 5% (1/22) of the mice that received Vanco_5%-PMMA local treatment (regardless of SC Vanco) or 0% (0/8) and 0% (0/8) that received Vanco_5%-CPS. The importance of co-treatment with rifampin was demonstrated by the lesser efficacy of the vancomycin monotherapy and the lack of significant differences in CFU counts between Vanco_5%-CPS and Vanco_5%-PMMA. Vanco_5%+Rif_0.05%-CPS treatment significantly reduced the bacterial load compared with Vanco_5%-PMMA on the antibiotic implants (1.22 ± 1.90 vs. 2.78 ± 1.94 log(1+CFU/g), respectively; p = 0.013 by 2-way ANOVA; p = 0.020 by Fischer’s exact test; Figure 4c). The fixation plate and screws were colonized in 100% of the mice and the bacteria colonizing them may not have been significantly affected by the Vanco_5%+Rif_0.05%-CPS local treatment compared with Vanco_5%-PMMA (5.53 ± 1.09 vs. 6.08 ± 0.62 log(1+CFU/g); p < 0.001 by 2-way ANOVA; p = 0.072 by Fischer’s exact test; Figure 4d). Biofilm on the plates and screws was confirmed through scanning electron micrographs (Figure 5).

Figure 4.

Bacterial colonization, measmed by end-point colony forming unit (CFU) assays, was reduced by low dose local antibiotic delivery via 3D printed CPS. (a) CFU quantification from the homogenized femoral bone, (b) soft tissue of the thigh, (c) the sonicated antibiotic implants, and (d) the fixation plate and screws. *denotes p<0.05 for the effect ofVancos%+RifooswCPS vs. Vancos%-PMMA by 2-way AN OVA. $denotes p<0.05 for the effect ofVancos%+Rifoos%-CPS vs. Vancos%-PMMA by Fischer's exact test. Vancos%-CPS was not significantly different from Vanco₅%-PMMA by Dunn's test after nonparametric ANOVA (Kruskal-Wallis) in each tissue and implant. n = 7-15/group. Bars indicate the median of each group.

Figure 5.

Bacteria were observed embedded within biofilm on the plates and screws from all groups at day 28 of the study through scanning electron micrographs (SEM).

Local antibiotic treatment with Vanco_5%+Rif_0.05%-CPS significantly reduced the osteolytic bone resorption compared with Vanco_5%-PMMA and Vanco_5%-CPS (23 ± 7% vs. 34 ± 11% resorption, respectively; p < 0.0001; Figure 6b) and produced a significantly more positive net bone change (-8.9% ± 12% vs. −21% ± 16%, respectively; p = 0.0012; Figure 6d). No significant differences were observed in the volume of bone formation between all of the treatment groups (Figure 6c). Radiographic changes to the bone were not significantly different between Vanco_5%-CPS and Vanco_5%-PMMA in all of the measurements.

In vivo efficacy of antibiotic-laden CPS with a higher rifampin dose and sustained release

Next, we explored whether the efficacy of rifampin- and vancomycin-laden CPS could be enhanced by increasing the rifampin dose 10-fold to 0.5 wt% (Vanco_5%+Rif_0.5%-CPS) or by coating the higher dose rifampin-loaded CPS with PLGA to extend the time course of antibiotic release (Vanco_5%+Rif_0.5%-CPS/PLGA). The PLGA coated designs displayed an initial burst release of rifampin (51 μg in the first 24 hr, 56% of total released for 12.5 wt% PLGA), followed by a sustained release (1.8 ± 0.8 μg/day after day 1, excluding the secondary burst at days 8-9) for at least 2 weeks in vitro (Figure 7a,b). The design of the PLGA coating for in vivo analysis was selected to be 12.5 wt% based on a similar burst release of rifampin during the first 24 hours between the uncoated Vanco_5%+Rif_0.5%-CPS and the Vanco_5%+Rif_0.5%-CPS/PLGA_12.5% groups (Figure 7a). Over the two weeks of elution in vitro, all of the PLGA groups maintained daily rifampin concentrations that were well above the planktonic S. aureus MIC (0.08 μg/mL (Mihailescu et al., 2014)) (Figure 7b). In addition, the PLGA coating significantly enhanced the flexural strength, stiffness, and toughness of the 3D printed CaP scaffolds (Table 1).

Figure 7.

Release kinetics of rifampin as a :function of PLGA coating. (a) Rifo swCPS that were coated by dipping in a 12.5% PLGA solution, supplemented with 20 mg/mL rifampin, released a similar amount of rifampin as uncoated Rifo swCPS during the first 24 hours, followed by a sustained release of rifampin for at least 2 weeks (n=5). The Vancos%+Rifos%-CPS that were uncoated or coated with the 12.5% PLGA solution were studied in vivo. (b) The daily rifampin concentrations were well above the minimum inhibitory concentration (MIC; 0.08 µg/mL) of rifampin-sensitive S. aureus for the :full 2 weeks.

The in vivo efficacy of the higher rifampin concentration and sustained release formulations was examined alongside the original low dose design and compared to Vanco_5%-PMMA, with all of the groups receiving concomitant systemic vancomycin. None of the mice possessed gross pathological signs of infection at the end of the 3-week treatment time course. All of the rifampin-laden CPS treatments significantly improved the early infection management at 1 and 3 days after revision surgery compared with the Vanco_5%-PMMA (2080 ± 560 vs. 3720 ± 1400 photons/s/cm²/sr, respectively at 3 days post-revision; p < 0.001; Figure 8).

Figure 8.

The bactericidal efficacy of rifampin- and vancomycin-laden CPS is enhanced by higher rifampin dose and slow-release antibiotic delivery at 1 and 3 days after revision surgery compared with the Vanco_5%-PMMA. (a) In vivo, the bacterial burden was significantly reduced at early time points (Days 8 and 10) after revision surgery by each of the rifampin-laden CPS treatments compared to Vanco_5%-PMMA, as determined by longitudinal bioluminescence. (b) The Vanco_5%+Rif_0.05%-CPS and Vanco_5%+Rif_0.5%-CPS/PLGA treatments each reduced the post-revision radiance AUC compared to Vanco_5%-PMMA spacers. All treated groups concomitantly received systemic vancomycin via subcutaneous injection. *denotes multiplicity-adjusted p<0.05 vs. Vanco_5%-PMMA by Dunn’s test after nonparametric ANOVA (Kruskal-Wallis). Bonferonni’s correction was applied for analysis of the multiple time points across the bioluminescence time course. n = 8-15/group. Bars and data points are medians and error bars are interquartile range.

The bacterial colonization of the bone, as measured by endpoint CFU assays, was significantly reduced by the sustained release CPS/PLGA treatment compared with the vancomycin-laden PMMA (2.1 ± 2.3 vs. 5.2 ± 0.8 log(1+CFU/g), respectively; p = 0.002; Figure 9a). The bones were bacteria culture negative in 50% (4/8) of the mice that received Vanco_5%+Rif_0.5%-CPS/PLGA, 25% (2/8) that received Vanco_5%+Rif_0.5%-CPS, 12.5% (1/8) that received Vanco_5%+Rif_0.05%-CPS, and 0% (0/15) that received Vanco_5%-PMMA (Figure 9a). All of the rifampin-laden CPS treatments significantly reduced the bacterial burden in the soft tissue relative to Vanco_5%-PMMA as well (1.0 ± 1.3 vs. 3.8 ± 1.4 log(1+CFU/g); p < 0.003), which were 63% culture negative within each CPS group (Figure 9b). The PMMA spacers were susceptible to colonization by the bacteria, with 60% (9/15) of the implants culture positive at the end of the study (Figure 9c). In contrast, the Vanco_5%+Rif_0.5%-CPS and Vanco_5%+Rif_0.5%-CPS/PLGA formulations were 100% culture negative (Figure 9c). In fact, the retrieved Vanco_5%+Rif_0.5%-CPS/PLGA still released active antibiotics at the end of the study (Figure 10), which suggests that these CPS would deliver antibiotics for more than 3 weeks in vivo. Bacteria were not eradicated from the fixation plates and screws in any of the mice, but the bacterial load was significantly reduced by local rifampin therapy compared to Vanco_5%-PMMA (4.7 ± 0.9 vs. 5.9 ± 0.7 log(1+CFU/g); p = 0.03; Figure 9d).

Figure 9.

Bacterial colonization, measured by end-point colony-forming unit (CFU) assays after 3 weeks of treatment (day 28 of study), in the homogenized femoral bone (a) and soft tissue of the thigh (b) as well as on the sonicated antibiotic implants (c) and on the fixation plate and screws (d) was significantly reduced by local rifampin treatments compared to Vanco_5%-PMMA. All of the groups concomitantly received systemic vancomycin via subcutaneous injection. *denotes multiplicity-adjusted p<0.05 vs. Vanco_5%-PMMA by Dunn’s test after nonparametric ANOVA (Kruskal-Wallis). n = 8-15/group. Bars indicate the median of each group.

Figure 10.

After 3 weeks in vivo, the retrieved scaffolds were vortexed in 1 mL PBS and filter paper discs were allowed to adsorb the eluate overnight before placing the discs onto Xen36 inoculated Luria agar plates (a). Only eluate from the Vanco₅%+Rifos%-CPS/PLGA samples produced a bacterial inhibition zone around the respective disc. Note the lack of positive bioluminescent radiance sunounding the Vanco₅%+Rifos%-CPS/PLGA sample. (b) Only the Vancos%+Rifos%-CPS/PLGA group had significantly reduced radiance values around the disc compared to the placebo PMMA sample (PBO-PMMA). *denotes p<0.05 vs. PBO-PMMA by Dunnett's test after ANOVA. n = 4/group in duplicates.

The osteolytic bone resorption was significantly reduced by the higher rifampin dose (20.0 ± 7.7% resorption; p = 0.045) and the sustained release CPS formulations (18.1 ± 7.1% resorption; p = 0.012) compared to Vanco_5%-PMMA (28.8 ± 9.3% resorption; Figure 11b). Similar levels of reactive bone formation were observed across all of the treatment groups (Figure 11c), and the higher rifampin dose achieved a significantly more positive net bone change compared to Vanco_5%-PMMA (+1.0 ± 12.8% vs. −14.0 ± 12.0% net change; p = 0.017; Figure 11d).

Figure 11.

Infection-induced bone loss is significantly reduced in mice that were treated with the higher rifampin dose or extended release formulations. Volumetric bone changes were measured by overlaying micro-CT scans that were collected one day after the revision surgery (day 8) and on the last day of the study (day 28). (a) The higher rifampin dose and sustained release groups significantly reduced the volumetric bone resorption compared with Vancos%-PMMA. (b) No significant differences in bone formation were observed. (c) The net bone change was significantly more positive in the Vancos%+RifoswCPS group. *denotes multiplicity-adjusted p<0.05 vs. Vanco₅wPMMA by Dunnett's test after AN OVA.

Discussion

Treatment of implant-associated bone infections remains a significant clinical challenge despite the aggressive surgical and antibiotic treatments that are currently used. The general advantages of local antibiotic delivery are well established for osteomyelitis (Gogia et al., 2009), but the potential efficacy of local rifampin therapy against an implant-associated bone infection with biofilm involvement had not previously been studied. In vitro studies (Mihailescu et al., 2014; Parra-Ruiz et al., 2010; Saginur et al., 2006; Tang et al., 2013) have demonstrated evidence for rifampin’s efficacy against S. aureus in biofilms. Studies in experimental animal models (Garrigos et al., 2010; John et al., 2009; Niska et al., 2013; Saleh-Mghir et al., 2011; Vergidis et al., 2011) and clinical studies (Drancourt et al., 1997; Widmer et al., 1992; Zimmerli et al., 1998) have further corroborated the benefits of rifampin, administered systemically, in cases of implant-associated staphylococcal infections with likely biofilm involvement. Therefore, this study examined the concomitant local delivery of vancomycin and rifampin from 3D printed CPS for treatment of a Staphylococcal bone infection in the presence of an orthopaedic implant contaminated with biofilm.

Considering the potential toxicity of rifampin to local osteogenic cells (Duewelhenke et al., 2007; Rathbone et al., 2011) and the very low MIC for planktonic S. aureus (0.08 μg/mL (Mihailescu et al., 2014)), we first evaluated a low dose of rifampin (0.05 wt% or 3.5 μg per murine scaffold). Even with this low rifampin dose, significant reductions in the bacterial load and osteolysis were observed compared with vancomycin-laden PMMA. Yet, the higher rifampin dose and sustained release formulations tended to further enhance the efficacy of the rifampin-laden CPS. Although the differences in mean CFU were not statistically significant, the Vanco_5%+Rif_0.5%-CPS and Vanco_5%+Rif_0.5%-CPS/PLGA resulted in higher rates of culture negative bones compared with the lower dose Vanco_5%+Rif_0.05%-CPS. In addition, the retrieved Vanco_5%+Rif_0.5%-CPS/PLGA scaffolds were still bactericidal at the end of the 3-week treatment study. In contrast, the PMMA spacers often became inert after the initial elution of antibiotics from the surface and were susceptible to colonization by the remaining bacteria. This problem was observed in 60% (9/15) of the mice that received PMMA in this study and has also been documented clinically (Mariconda et al., 2013; Schmolders et al., 2014).

Previous in vitro studies have found that rifampin is more cytotoxic to primary human osteoblasts than many other antibiotics (Duewelhenke et al., 2007; Rathbone et al., 2011). In this septic bone defect study, reactive bone formation was not significantly reduced by any of the rifampin treatments compared to vancomycin-laden PMMA. These data suggest that the local rifampin delivery, regardless of dose or elution time in this study, did not significantly impair the activity of local osteogenic cells.

The significant reduction in BLI signal from day 7 post-op to day 10 in the groups that received local rifampin, and the lack thereof in every vancomycin monotherapy group (including Vanco_5%-CPS), suggests that rifampin may be important for attenuating the bacterial metabolic load shortly after the initiation of treatment. The use of BLI in this infection model is limited to quantification of metabolically active bacteria, which may not include the more quiescent viable bacteria within the biofilm. Therefore, these BLI data can only evidence that rifampin better controls the planktonic bacteria following revision. A limitation to this comparison is that systemic antibiotic therapy would have been initiated prior to the revision surgery in a clinical situation, which may have resulted in a lower BLI signal in the vancomycin monotherapy cases since new planktonic bacteria that could be released during debridement likely would have been controlled by pre-existing systemic antibiotic. Nonetheless, the observed benefits of local rifampin and vancomycin co-therapy are consistent with many in vitro (Mihailescu et al., 2014; Parra-Ruiz et al., 2010; Saginur et al., 2006; Tang et al., 2013) and in vivo (Garrigos et al., 2010; John et al., 2009; Niska et al., 2013; Saleh-Mghir et al., 2011; Vergidis et al., 2011) studies, which have found that combinational therapies that include rifampin exceed the efficacy of each monotherapy.

Although the local rifampin delivery significantly reduced the mean bacterial load and produced higher rates of culture negative tissues, culture positive mice still existed in each treatment group and 100% of the fixation plates or screws were culture positive at the end of the study. As in every infection treatment study, the culture negative sample rate is an important factor considering that residual viable bacteria will pose a risk for re-infection after cessation of treatment. Yet, complete eradication is extremely challenging in implant-related S. aureus infections. Many other pre-clinical models of implant-associated S. aureus infections were also refractory to conventional antibiotic treatments that employed systemic rifampin combined with vancomycin, daptomycin, linezolid, or levofloxacin, or the respective monotherapies (Garrigos et al., 2010; John et al., 2009; Niska et al., 2013; Saleh-Mghir et al., 2011; Vergidis et al., 2011). Only one study observed a 100% clearance, which used daptomycin plus rifampin against a knee peri-prosthetic infection model in rabbits (Saleh-Mghir et al., 2011). Niska et al. systemically administered rifampin in combination with vancomycin to a mouse model with a femoral intramedullary pin, with a Xen36 infection in the knee joint, and found 100% clearance of bacteria from the pins and 67% clearance from the peri-implant tissue (Niska et al., 2013). In contrast to the current study, however, Niska et al. measured approximately 500 CFU per pin and only a 67% culture positive rate on the pins in the untreated group, which is extremely low compared to the 6×10⁶ CFU and 100% colonization in the untreated group of the trauma infection model utilized herein (Inzana et al., 2015). The fundamental differences between these knee joint peri-prosthetic models and the current trauma model, as well as the differences in CFU counts in untreated animals, makes direct comparison with the current results very challenging. Evaluating systemic rifampin plus vancomycin in the current model may have provided some additional insights, but the current study was focused on local delivery strategies. Since the goal is to achieve a high concentration of antibiotics in the infected tissues for a sufficient time, the origin of the antibiotic (systemic injection vs. local reservoir) may be less important than the resulting bioavailability at the infection site. Nonetheless, these current and previous results highlight the extreme challenge of treating biofilms and caution against implant retention in the presence of a biofilm-forming species.

This challenging infection model may have been refractory to the current treatments if the vancomycin-rifampin combination was not sufficiently effective against the bacteria in biofilm or the concentrations achieved by the antibiotics were still sub-lethal. Indeed, the MIC of rifampin against S. aureus in biofilm is orders of magnitude greater than planktonic S. aureus (164 μg/mL vs. 0.08 μg/mL (Mihailescu et al., 2014)). Yet, it is important to note that the antibiotic concentrations measured in vitro (Figure 7) are likely lower than the in vivo concentrations based on the local fluid volumes in each scenario. Vancomycin was chosen as the co-therapy for this study based on its common clinical use against MRSA strains (Deresinski, 2009). Although Xen36 is methicillin-sensitive, the vancomycin MIC for Xen36 is no more than 0.5 μg/mL (Niska et al., 2013), compared with 1-2 μg/mL in a variety of MRSA strains (Crandon et al., 2010). Clinically, however, Trampuz and Zimmerli recommend systemically administered rifampin in combination with (flu)cloxacillin for methicillin-sensitive Staphylococci or rifampin plus vancomycin for MRSA, followed by rifampin plus ciprofloxacin or levofloxacin after the first two weeks (Trampuz and Zimmerli, 2005; Trampuz and Zimmerli, 2006). Other antibiotic combinations with rifampin may have been more effective against the bacteria in biofilm than vancomycin (Garrigos et al., 2010; John et al., 2009; Mihailescu et al., 2014; Parra-Ruiz et al., 2010; Tang et al., 2013), but the most efficacious antibiotic combination is inconclusive when considering all of the previous in vitro and in vivo studies together. The similar results from treatment with Vanco_5%-CPS and Vanco_5%-PMMA suggest that the delivery material was less important than the choice of antibiotic in this study. An important future study would be to evaluate these vancomycin- and rifampin-laden CPS in a similar animal model, but with complete exchange of the fixation hardware during the revision surgery, to determine if recolonization of the new plate and screws can be prevented.

While this study demonstrates the specific efficacy of a rifampin- and vancomycin-laden CPS in a murine infection model, the additive manufacturing techniques used herein are highly versatile and easily amenable to the inclusion of alternative antibiotics or extension to other regenerative medicine and tissue engineering applications. As demonstrated in this study, the 3D printing process readily provides adaptive modularity through alternative bioactive ‘inks’ that can be infused via the inkjets (e.g. biofilm dispersal agents to augment the antibiotic efficacy), through variations in the powder phase composition, and through post-printing modifications to the scaffold. This versatility is extended by the expanse of possible geometries as well as the selective spatial distribution of bioactive molecules within each layer, which may provide additional control over the release kinetics (Klammert et al., 2010) or site-specific scaffold bioactivity (Gbureck et al., 2007a). Following successful infection management, the osteoconductive 3D printed CPS may serve as a scaffold for regeneration of the bone defect without a second revision surgery, as is necessary with the use of PMMA. Finally, 3D printing enables easily customizable solutions for medical image-guided, patient-specific therapy, which will greatly facilitate future clinical translation, especially for craniomaxillofacial applications in which the aesthetic aspects of the reconstruction are important.

Conclusions

In this study, we have demonstrated that the concomitant local delivery of rifampin and vancomycin from 3D printed CPS significantly improves the outcomes of an implant-related bone infection in a mouse model compared to the current clinical practice of using PMMA spacers, which cannot carry rifampin. The higher rifampin dose and post-printing PLGA coating further enhanced the efficacy of these materials through prolonged local antibiotic delivery to successfully manage the infection and potentially eradicate the pathogen from the bone in 50% of the mice. Following successful infection management, the osteoconductive 3D printed CPS may serve as a scaffold for regeneration of the bone defect without a second revision surgery, as is necessary with the use of PMMA. Yet, based on these current results and those of many other studies discussed herein, it seems that the only way to completely cure an infection is to remove all of the bacteria and biofilm (implant exchange and aggressive debridement) or induce the bacteria to return to a planktonic state where they are more susceptible to antimicrobials (the goal of experimental biofilm dispersal agents (Kaplan, 2010)). We are not convinced that a particular antibiotic combination or dosage alone will assuredly kill 100% of the bacteria, although the clinical symptoms would likely be suppressed.