Evaluation of a polymeric composite bone filler scaffold for local antibiotic delivery to prevent Staphylococcus aureus infection in a contaminated bone defect

We previously reported the development of an osteogenic bone filler scaffold consisting of degradable polyurethane (dPU), nano-sized hydroxyapatite (nHA), and decellularized bovine bone particles (DBP). In this report we describe the results of studies aimed at evaluating the use of this scaffold as a means of local antibiotic delivery for the prevention of infection in a segmental bone defect contaminated with Staphylococcus aureus. We evaluated two different scaffold formulations that contained the same components in the same ratios but differed from each other with respect to overall porosity and therefore surface area. Studies done with vancomycin, daptomycin, and gentamicin confirmed that antibiotic uptake was concentration dependent and that increased porosity was correlated with increased uptake and prolonged release of all three antibiotics. Vancomycin could be passively loaded into either scaffold formulation in an amount sufficient to prevent infection, as evidenced by the complete eradication of viable bacteria from the surgical site of most animals in a rabbit model of a contaminated mid-radial segmental bone defect. Even in those few cases in which complete eradication was not achieved, the number of viable bacteria present in the bone was significantly reduced comparison to untreated controls. There was also no radiographic evidence of osteomyelitis in any rabbit treated with vancomycin-loaded scaffold. Microcomputed tomography (μCT) of bone defects up to 84 days of exposure to scaffolds with and without vancomycin also demonstrated that the addition of vancomycin even in the highest concentration did not significantly diminish the osteogenic properties of either scaffold formulation. Together, these results demonstrate the potential utility of our bone regeneration scaffold for local antibiotic delivery.


Introduction 53
Open fractures and penetrating wounds that damage the bone are extremely complex injuries.

54
The resulting bone defects are particularly problematic because they involve a breach of the skin 55 and occur in non-sterile environments, thus making them highly susceptible to infection. The 56 treatment of these injuries requires an intensive interdisciplinary clinical approach that includes 57 long-term systemic antibiotic therapy and surgical debridement to remove damaged and 58 contaminated bone and soft tissues [1][2][3][4][5]. Debridement can exacerbate the bone injury even 59 further, thus increasing the likelihood that reconstruction will be required. This often necessitates release bioactive agents including antibiotics [15][16][17][18]. However, to date, its use as an antibiotic-antibiotic in each sample; the results of these bioassays are reported based on the lower end of 127 this two-fold range. Our vancomycin bioassays were verified using a spectrophotometric assay 128 based on absorbance at 280 nm as previously described [25].

129
In vivo rabbit osteomyelitis model 130 Our initial in vivo studies were done with the K20 scaffold formulation saturated in a solution 131 containing 100 mg/mL of vancomycin based on the results of in vitro elution studies demonstrating 132 maximum antibiotic uptake and elution at this concentration (see below). To test in vivo relevance,

133
we used a rabbit model of post-surgical OM directly assess the ability of vancomycin-saturated 134 scaffolds to prevent infection in a contaminated bone defect [7,8,24]. Specifically, a 1 cm 135 segment of bone was surgically excised from the middle of the right radius leaving the ulna intact.

136
The segment was discarded and 1 X 10 6 colony-forming units (CFU) of the S. aureus strain 137 UAMS-1 [21] in 10 µL of PBS was injected into the medullary canal on each end of the exposed 138 defect, thus resulting in a total inoculum of 2 X 10 6 CFU. The surgical void created in the bone 139 was then filled with scaffold saturated with PBS or PBS containing vancomycin. Control groups 140 included uninfected rabbits exposed to scaffolds saturated with PBS or PBS saturated with 100 141 mg/mL of vancomycin. All groups in each experiment included 3 rabbits per group. Radiographic 142 images were obtained on the surgical day (day 0) and at weekly or biweekly intervals thereafter.

143
At the completion of each experiment, rabbits were humanely euthanized and the surgical 144 site exposed. To preserve the surgical site for µCT analysis, bacteriological sampling in these 145 early experiments was limited to thoroughly swabbing the exposed surgical site and plating on  formalin were then subjected to bacteriological analysis. All µCT components were cleaned with 157 70% ethanol before and after use.

159
The reconstructed cross-sectional slices were processed with the Skyscan CT-analyzer software 160 using the following steps: first, bone tissue was isolated from the soft tissue and background using 161 a global thresholding (low = 85; high = 255). We then delineated regions of interest (ROI) by 162 manually contouring the areas that contained scaffold. Next, we obtained the bone volume 163 present in the drawn ROIs using the 3D analysis tool. In addition, we performed µCT on the S1 164 and K20 scaffolds prior to implantation (Fig 1) and calculated the bone volume, which allowed us 165 to calculate the percent of scaffold remaining after implantation by comparison to the volume 166 observed before implantation.

Assessment of bacterial burdens in bone 168
In short-term 14 day experiments in which µCT was not done and in the later longer term 169 study in which µCT was done on frozen samples rather than formalin-fixed samples, dissected

Results and discussion
189 The dry weight of individual S1 and K20 scaffolds ranged from 84.78-90.61 and 84.62-90.27 190 mg, respectively. However, µCT imaging of each formulation before antibiotic loading provided 191 visual evidence that the S1 scaffold was more dense than the K20 scaffold (Fig 1), and based on 192 the amount of PBS remaining after saturation of each scaffold formulation the K20 scaffold was 193 found to take up an average of approximately 2.30-fold more fluid than the S1 scaffold (range = 194 195-363 and 105-138 µL, respectively), thus confirming that the modifications made to the K20 195 scaffold increase its surface area and porosity relative to the S1 scaffold.

196
With both scaffold formulations, the amount of fluid uptake was consistent irrespective of

223
To determine whether the amount of vancomycin observed with the K20/100 formulation could 224 be defined as saturating, we also exposed the K20 scaffold to a solution containing 250 mg/mL 225 of vancomycin, resulting in a greater initial release as defined after a 30 minute elution period (~32 mg/mL). However, within 24 hours the amount of antibiotic in the eluate fell to a concentration 227 comparable to that observed with the K20 scaffold saturated with 100 mg/mL antibiotic (~8.0 228 mg/mL). Thus, as defined by the overall elution profile, we considered the K20 scaffold to be 229 saturated when it was exposed for 24 hours to a solution containing 100 mg/mL of vancomycin.

230
Although our in vivo studies were limited to vancomycin-saturated scaffold (see below), using this 231 assay we also demonstrated that the K20 scaffold could be saturated with daptomycin ( Fig S1) 232 and gentamicin ( Fig S2) in a concentration-dependent manner, although saturation required

246
Given an experimental group size of 3 rabbits per group, these experiments collectively 247 included 9 rabbits per experimental group. In the experimental group in which S. aureus was 248 introduced and no antibiotic was added to the scaffold, the soft tissues surrounding the surgical 249 site were confirmed to be heavily infected in 8 of 9 rabbits (Fig 3, Group 1). The exception was a 250 single rabbit evaluated at 84 days post-infection, thus suggesting that the infection in this rabbit 251 spontaneously resolved over time. These results confirm that the inoculum was sufficient to establish infection, particularly given that all isolates obtained at the end of each experiment were 253 confirmed by PCR to be the same S. aureus strain used to initiate the infection (data not shown).

254
In contrast, none of the 9 rabbits in the experimental group inoculated with S. aureus and treated 255 with vancomycin-loaded scaffold were found to be infected at any time point (Fig 3, Group 2).

256
Although bacteriological analysis in these studies was limited to swabs taken from the surgical 257 site and did not include the bone or scaffold itself, these results nevertheless demonstrate that 272 the K20/100 scaffold formulation. A 1 cm defect (red arrow) was created in the right radius of 273 rabbits as previously described [24]. The excised bone was discarded, and the distal and proximal 274 ends of the remaining bone were either infected with S. aureus or inoculated with sterile PBS.

275
The K20 formulation saturated with either PBS or vancomycin (100 mg/mL) was then placed into

282
The µCT analysis confirmed that the scaffold was largely destroyed in infected rabbits that 283 received PBS saturated scaffold (Fig 4, Group 1), while it remained in place and intact in all other 284 experimental groups (Fig 4, Groups 2-4). These data suggest that the utility of the scaffold would interosseous membrane that runs between the medial aspects of the bones (the radioulnar fibrous 298 joint), thus providing structural stability even after creating a 1 cm segment defect in the radius.

299
However, the presence of the intact ulna can also present a disadvantage in that it makes it 300 difficult to accurately assess bone regeneration across the segmental defect itself. Second, the 301 scaffold itself contains bone particles that are readily detectable by µCT (Fig 1). To compensate 302 for these factors, we focused on the scaffold itself rather than the bone, and the results of these 303 studies led us to conclude that approximately 50% of the scaffold remained after 84 days in both of the uninfected experimental groups (Fig 5, Groups 3 and 4). The amount of scaffold remaining 305 in infected rabbits exposed to vancomycin-loaded scaffold was slightly higher (Fig 5, Group 2), 306 which suggests that the transient presence of bacteria may have slowed decomposition of the 307 scaffold. However, the difference did not reach statistical significance by comparison to the 308 uninfected experimental groups. As suggested by visual analysis of µCT images (Fig 4), the 309 analysis also confirmed that the scaffold was essentially destroyed in infected rabbits exposed to 310 PBS-saturated scaffold (P = 0.0274) (Fig 5, Group 1). with UAMS-1 and exposed to PBS-saturated scaffold (Fig 5, Group 1), the same bacterial strain 325 used to initiate the infection was isolated from 8 of 9 rabbits at the end of each experiment, there 326 was clear evidence of OM as assessed by radiographic analysis (Fig 3), and the segmental defect 327 was not filled as assessed by µCT (Fig 4). In contrast, no bacteria were isolated from any of the 328 27 rabbits in the other three experimental groups, there was no radiographic evidence of the 329 development of OM in any of these rabbits (Fig 3), and there was no significant difference in bone regeneration in rabbits in any of these three groups (Fig 4 and Fig 5). Thus, the results of these

346
To address these issues, we carried out two experiments with three significant modifications 347 that allowed us to focus specifically on determining the minimum amount of vancomycin required 348 to prevent infection. First, infected rabbits were treated using the K20 scaffold saturated with PBS

364
The results of these spectrophotometric assays were generally consistent with the results of 365 our bioassays, although when viewed collectively there was evidence to suggest that our 366 bioassays may have underestimated the amount of antibiotic (Fig S4). This finding is not 367 surprising given our use of two-fold dilutions in our bioassays and the fact that we chose to report 368 the lower concentration of the range defined by these dilutions. Analysis also confirmed that 369 antibiotic uptake and release is dependent on the concentration of antibiotic in the solution used 370 to saturate the scaffold (Fig S4). These results demonstrate that antibiotic loading can be 371 controlled by modifications to the scaffold itself (Fig 2) and by altering the concentration of 372 antibiotic used to load the scaffold (Fig S4), thus providing further flexibility in the use of our 373 scaffold as an anti-infective. of the complete area did allow for exhaustive bacteriological analysis. As in our earlier experiments, all six rabbits in the untreated control group were culture positive based on swabs 383 of the exposed surgical site, while no bacteria were isolated from any rabbits in any of the other 384 experimental groups (Fig 6). The bone in all six rabbits in the infected but untreated experimental 385 group was also found to be heavily colonized. Viable bacteria were also isolated from the bone 386 and/or scaffold of 7 of 24 of the infected rabbits treated with scaffold loaded with varying 387 concentrations of vancomycin. Of these 7, four were treated with the K20/25 scaffold, with the 388 K20/50, K20/75, and K20/100 experimental groups containing 0, 1, and 2 culture-positive rabbits, 389 respectively (Fig 6). However, the presence of even a single colony was counted as a positive,

390
and in all 7 cases the bacterial load was dramatically reduced by comparison to the number 391 isolated from all 6 rabbits in the infected, untreated experimental group (Fig 6). This reduction, can be achieved with the K20 scaffold loaded with only 50 mg/mL of vancomycin. In our initial antibiotic loading studies (Fig 2), the S1 scaffold was found to have a loading capacity ~50% of 409 that observed with the K20 scaffold. The prevention of infection by K20 at half the potential loading 410 capacity suggests that the S1 scaffold saturated with vancomycin would also be sufficient to 411 minimize infection in a contaminated bone defect. To test this idea, we carried out a final 412 experiment that differed from previous experiments in two respects. First, we used the S1 scaffold 413 saturated in 100 mg/mL of vancomycin rather than the K20/100 formulation. Second, radiographs 414 were taken every 21 days throughout the 84 day post-infection period, at which point we employed 415 a modified protocol utilizing frozen rather than formalin-fixed bones in an attempt to obtain µCT 416 and comprehensive bacteriological data from the same rabbit.

417
In this study, the only experimental group in which viable bacteria were isolated from soft 418 tissue swabs of the exposed surgical site were in the infected experimental group in which the 419 surgically-removed defect was replaced with PBS-saturated scaffold, with all three rabbits in this 420 group being heavily culture positive (Fig 7, Group 1). Based on bacteriological analysis of the 421 bone/scaffold, we did confirm one culture positive rabbit in the S1/100 experimental group, but 422 the number of viable bacteria obtained from this rabbit was <10 2 , while in the corresponding 423 experimental group treated with the PBS-saturated scaffold the number ranged from 10 4 to 3 X 424 10 7 (data not shown). These results are consistent with radiographic analysis, which revealed 425 clear signs of soft tissue infection and osteonecrosis in all rabbits in this experimental group (Fig   426  7, Group 1) and the absence of these signs in rabbits from all other experimental groups (Fig 7,

427
Groups 2-4). Radiographic analysis also confirmed that the bone defect remained unfilled in the 428 Group 1 rabbits but that this was not the case in rabbits in any other experimental group (Fig 7).

429
As with the K20 scaffold, the amount of the S1 scaffold remaining after 84 days was also 430 significantly lower (P = 0.0375) in all Group 1 rabbits by comparison to all other experimental 431 groups, and comparable among all other experimental groups irrespective of the presence or 432 absence of vancomycin (Fig 8). However, as with the K20/100 scaffold, there was a trend suggesting that the presence of bacteria may have delayed incorporation of the scaffold, although 434 once again this trend did not reach statistical significance (Fig 8).
435 436 Fig 7. Radiographs from representative rabbits in each experimental group treated with 437 the S1/100 scaffold formulation. Representative X-rays are shown from infected and uninfected 438 rabbits exposed to the S1 scaffold saturated with PBS or 100 mg/mL vancomycin as indicated to 439 the left. After 84 days, rabbits were humanely euthanized and the surgical limb harvested and 440 frozen. Upon removal from the freezer, the surgical site was exposed and swabs taken from the 441 surrounding soft tissues. The bone and its associated scaffold were then imaged by µCT before 442 being homogenized and plated on tryptic soy agar to recover viable bacteria. The collective results

443
observed with soft tissue samples and the bone/scaffold are shown to the right.

454
In summary, the studies we report here were carried out as six independent experiments that 455 differed with respect to the scaffold formulation used (K20 vs. S1), the time after surgery at which 456 the results were evaluated (14, 28, 56 and 84 days), the amount of vancomycin used to saturate 457 the scaffold (25-100 mg/ml), and the method employed for bacteriological analysis (soft tissue vs.

458
soft tissue and bone/scaffold). These changes were implemented to address specific issues as