|Home | About | Journals | Submit | Contact Us | Français|
Patients recovering from traumatic injuries or surgery often require weeks to months of hospitalization, increasing the risk for wound and surgical site infections caused by ESKAPE pathogens, which include A. baumannii (the ESKAPE pathogens are Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species). As new therapies are being developed to counter A. baumannii infections, animal models are also needed to evaluate potential treatments. Here, we present an excisional, murine wound model in which a diminutive inoculum of a clinically relevant, multidrug-resistant A. baumannii isolate can proliferate, form biofilms, and be effectively treated with antibiotics. The model requires a temporary, cyclophosphamide-induced neutropenia to establish an infection that can persist. A 6-mm-diameter, full-thickness wound was created in the skin overlying the thoracic spine, and after the wound bed was inoculated, it was covered with a dressing for 7 days. Uninoculated control wounds healed within 13 days, whereas infected, placebo-treated wounds remained unclosed beyond 21 days. Treated and untreated wounds were assessed with multiple quantitative and qualitative techniques that included gross pathology, weight loss and recovery, wound closure, bacterial burden, 16S rRNA community profiling, histopathology, peptide nucleic acid-fluorescence in situ hybridization, and scanning electron microscopy assessment of biofilms. The range of differences that we are able to identify with these measures in antibiotic- versus placebo-treated animals provides a clear window within which novel antimicrobial therapies can be assessed. The model can be used to evaluate antimicrobials for their ability to reduce specific pathogen loads in wounded tissues and clear biofilms. Ultimately, the mouse model approach allows for highly powered studies and serves as an initial multifaceted in vivo assessment prior to testing in larger animals.
Acinetobacter baumannii has become a problematic, nosocomial species of bacteria responsible for various types of infections ranging in severity from urinary tract infection to skin and soft tissue infections, as well as ventilator-associated pneumonia (VAP); all of these infections predominantly arise in immunocompromised patients (1, 2). Complications in treatment have grown in recent years due primarily to the bacterium becoming increasingly antibiotic resistant; strains are often multidrug resistant (MDR) or extensively drug resistant (XDR) (3). As new treatments are developed for A. baumannii and other MDR bacteria, animal models to evaluate these treatments and various indications are also required in order to generate the required preclinical data before moving on to human trials.
Perhaps less appreciated, due to a lower mortality rate than VAP or cases of septicemia (4, 5), are the wound infections caused by A. baumannii, which represent a large fraction of the clinical indications observed (1, 2, 4, 6) and that need to be considered when testing novel antimicrobials against this bacterial pathogen. Despite significant advancements in wound and burn care, wound infections caused by MDR bacteria, including A. baumannii, still remain a major problem with regard to morbidity and mortality in both civilians and wounded military service members (1, 2, 6, 7, 8). Wound infections occur at higher rates among military service members, possibly due to issues associated with the type of traumatic injury, time until treatment, and the fact that the patients will pass through multiple medical facilities before arriving at a U.S. treatment facility (7). In the civilian sector, A. baumannii has been found associated with pressure ulcers and wounds in people with diabetes (9, 10), and it was even isolated from a case of necrotizing fasciitis (11). Implicating A. baumannii as the cause of these wound infections may be the result of improved diagnostic bacteriology and the increase in the number of MDR infections on the whole; regardless, there is an urgent need to develop effective strategies to prevent wound infection and promote wound healing after injury.
Other research laboratories have developed animal models to evaluate the virulence of A. baumannii and treatment regimens against this pathogen. These efforts include both murine and rat pulmonary models of infection. The murine models have relied upon mucin (12), cyclophosphamide (13), or morphine (14) to establish the infection, and yet another model used diabetic mice (15). Besides the rat pulmonary model developed by Russo et al. (16), that group also developed an abscess model of infection in the same study (16). Other A. baumannii small rodent models have utilized intravenous inoculums to induce sepsis and evaluate treatments (17, 18, 19). Lastly, while other groups have developed wound models of infection, those models relied on burns and/or were of short duration (20, 21, 22). Therefore, because of the lack of a cutaneous wound model, we attempted to develop a novel murine excision model in which antimicrobials could be evaluated from multiple quantitative/qualitative and microbiological/wound healing endpoints throughout a longer-duration, aggressive A. baumannii infection.
To do so, we evaluated known effective antimicrobials against a previously characterized clinical isolate of A. baumannii, strain AB5075. This strain causes severe clinical disease in a murine pulmonary model and demonstrates a high level of resistance to most clinically used antibiotics (submitted for publication). With the introduction of AB5075 in a neutropenic murine excision model, we can successfully simulate a traumatic wound infection and test novel antimicrobial compounds that could eventually become a part of future clinical treatments.
A. baumannii clinical isolate AB5075 was used in all experiments. AB5075 was isolated from a patient from Walter Reed Army Medical Center during 2008 to 2009 (23). AB5075 is MDR but susceptible to both rifampin (MIC, 4.0 μg/ml) and doxycycline (MIC, 0.125 μg/ml). AB5075 was propagated on Lennox Luria-Bertani (LB) medium (Becton, Dickinson and Co., Sparks, MD) for all experiments. To prepare inocula for animal infection, 100 μl of AB5075 overnight culture was subcultured into 10 ml of LB and then grown at 37°C and shaking at 250 rpm in a 250-ml Erlenmeyer flask. Cells in the mid-exponential growth phase were harvested at an optical density at 600 nm of 0.7. Cells were washed twice with sterile phosphate-buffered saline (PBS) and then resuspended in PBS so that 25 μl of the suspension contained 5.0 × 104 cells. The cell concentration of the suspension was verified via use of a Petroff-Hauser counting chamber prior to inoculation of mice and confirmed by serial dilution and plating on LB agar by using a spiral plating system (Autoplate; Advanced Instruments, Inc., Norwood, MA).
The neutropenic agent cyclophosphamide was purchased in powdered form from Baxter (Deerfield, IL) and dissolved in sterile 0.9% sodium chloride injection solution (Hospira Inc., Lake Forest, IL) to a final concentration of 10 mg/ml. The antibiotics rifampin and doxycycline were purchased in powdered form from Sigma-Aldrich (St. Louis, MO). Rifampin was dissolved in 100% dimethyl sulfoxide at 50 mg/ml and then diluted in 0.9% sterile sodium chloride injection solution to a final concentration of 20 mg/ml. Doxycycline was diluted in 0.9% sterile sodium chloride injection solution to a final concentration of 20 mg/ml. All treatment solutions were kept on ice or refrigerated until use.
Female BALB/c mice were purchased from the National Cancer Institute, Animal Production Program (Frederick, MD). The mice used in these experiments were 6 to 10 weeks of age and weighed 14 to 20 g. All mice received sterile food and water ad libitum, and dry rodent chow was supplemented with DietGel Recovery (ClearH2O, Portland, ME) during the 48 h following wounding. All mice were housed in groups of three, in sanitized cages on sterile paper bedding, and were provided with environmental enrichment, including in-cage plastic housing.
All procedures were performed in accordance with protocol IB02-10, which was approved by the Walter Reed Army Institute of Research (WRAIR)/Navy Medical Research Center (NMRC) Institutional Animal Care and Use Committee (Silver Spring, MD). All of the research presented here was conducted in compliance with the Animal Welfare Act and other federal statutes and regulations relating to animals and experiments involving animals, and it adhered to principles stated in the Guide for the Care and Use of Laboratory Animals (46). Mice received 150 mg/kg of body weight and 100 mg/kg cyclophosphamide via intraperitoneal (i.p.) injections, before wounding and infection, on days −4 and −1, respectively (24). On day 0, the day of wounding and inoculation, mice were anesthetized with injections of ketamine at 130 mg/kg (Ketaset; Fort Dodge Animal Health, Fort Dodge, IA) and xylazine at 10 mg/kg (AnaSed; Lloyd Inc., Shenandoah, IA), and buprenorphine at 0.05 mg/kg (Hospira Inc., Lake Forest, IL) was given via injection for pain management. Hair was clipped from the cervical to mid-lumbar dorsum, and the skin was scrubbed with iodine solution followed by an ethanol rinse. A 6.0-mm disposable skin biopsy punch (VisiPunch; Huot Instruments, LLC, Menomonee Falls, WI) was used to create a full-thickness skin defect overlying the thoracic spinal column and the adjacent musculature. Aliquots of 25 μl containing 5.0 × 104 AB5075 cells in a PBS suspension were pipetted into the wound and allowed to absorb for 3 min. A circular cutout (30 mm in diameter) of transparent dressing (TegadermRoll; 3M Health Care, St. Paul, MN) was placed over the wound and secured with tissue adhesive (Vetbond; 3M Animal Care, St. Paul, MN).
Beginning at 4 h postinoculation, mice were treated with either rifampin at 25 mg/kg via i.p. injection once daily, doxycycline at 25 mg/kg via i.p. injection twice daily, or an equivalent vehicle control over the course of a 6-day treatment period, similar to previous studies (25, 26). On day 7, the transparent dressing was removed, treatment was discontinued, and the wound was monitored for closure through day 25.
To measure viable bacterial cells within the wound, mice were euthanized via ketamine (250 mg/kg) and xylazine (25 mg/kg) overdose according to the protocol, and a 4-mm disposable skin biopsy punch (Acuderm Inc., Fort Lauderdale, FL) was used to remove a disc of wound bed material. The disc was placed in 1.0 ml of sterile PBS in a stomacher bag and manually disrupted. Serial 10-fold dilutions of homogenate were plated via spiral plater (Autoplate; Advanced Instruments, Inc., Norwood, MA) onto eosin methylene blue (EMB) agar (Becton, Dickinson and Co., Sparks, MD). Plates were incubated overnight at 37°C, and then CFU were enumerated.
Three mice from each treatment group (placebo, rifampin, and doxycycline) and from representative time points (4 hours postwounding and on days 1, 3, and 7) were sacrificed for wound bed bacterial community profiling. Wound bed tissue was excised, placed in 250 μl PBS, and frozen at −80°C. Samples were evaluated for bacterial species diversity by using pyrosequencing-based analysis of 16S rRNA genes (Research and Testing Laboratories, LLC, Lubbock, TX).
The 16S rRNA sequences in FASTA format were analyzed using mothur software (27). For quality filtering, sequences that had a quality score (Q) of less than 20, lacked an accurate primer sequence, contained ambiguous characters, or contained more than 8 homopolymers were removed. Sequences were then aligned against the SILVA (28) alignment template. Sequences that did not map to overlapping regions in the alignment were also removed. Potential chimeric sequences were detected by using Uchime (29), embedded in the mothur software (chimera.uchime), and removed. The high-quality sequences were then preclustered (30) to further reduce potential influences of sequencing errors. Operational taxonomic units (OTUs) were determined by using average neighbor clustering of sequences with 97% sequence identity. To assess the taxonomic distributions across each sample, a weighted representative sequence from each OTU was selected and subsequently classified by using a locally running RDP classifier program (50% bootstrap threshold) (31) and the mothur formatted version of the greengenes reference taxonomy (32).
Wound area measurements were taken on the day of wounding and at subsequent time points by using a Silhouette wound measurement device (Aranz Medical Ltd., Christchurch, New Zealand). Time course wound photographs to assess gross pathology were taken with a 5-megapixel iSight camera (Apple Inc., Cupertino, CA).
Dressings and wound bed tissue were evaluated by scanning electron microscopy (SEM). A representative mouse from the placebo, rifampin, and doxycycline treatment groups was sacrificed at 4 h postwounding and on days 1, 3, and 7. The transparent dressings and a 4-mm tissue disc for each animal were fixed in 4% formaldehyde, 1% glutaraldehyde, 0.1 M PBS. The samples were washed three times with 0.1 M PBS and then postfixed in 1% osmium tetroxide in 0.1 M PBS for 1 h. The samples were dehydrated in a graded series of ethanol solutions and then dried (model 28000 critical point dryer; Ladd Research Industries, Burlington, VT). The samples were mounted to specimen stubs by using double-sided carbon tape and ion coated with gold-palladium (30:70; Hummer X sputter coater; Anatech Ltd., Alexandria, VA). The samples were visualized by using an Amray 3600 FE scanning electron microscope (Bedford, MA) operated at a voltage of 3 kV. Samples were analyzed by scanning 10 or more 1,000×-magnified fields within the wounded tissue and on the portion of the dressing overlying the wounded area. Photomicrographs representative of the observed biofilm density were taken at 2,500× magnification.
Three mice from each treatment group (placebo, rifampin, and doxycycline) and from representative time points (days 1, 3, 7, 15, and 23) were sacrificed in order to characterize wound model histopathology. The dorsal wounds were removed en bloc by severing the cervical and lumbar spinal column and trimming the tissue to >2 cm beyond the wound edge. The tissue was immediately fixed in phosphate-buffered formalin (10%) for >72 h. The wound tissues, consisting of the spinal column and surrounding soft tissues, were then demineralized for 24 h by using Decal Stat (Decal Chemical Corp., Tallman, NY), rinsed with water for 3 to 5 min, trimmed in a dorsal-ventral plane bisecting the spinal column, and placed back into 10% phosphate-buffered formalin. The wound tissue specimens were embedded in paraffin, cut in a dorsal-ventral plane bisecting the spinal column, mounted on positively charged glass slides (Colormark Plus; Thermo Scientific, Portsmouth, NH), and stained with hematoxylin (Astral Diagnostics, Inc., West Deptford, NJ) and eosin (Astral Diagnostics, Inc., West Deptford, NJ) for light microscopic examination.
Wounds were histologically assessed for the presence and of dissemination of bacteria, host immune response, indications of wound healing, i.e., extent of epithelial migration, coverage, maturation, and amount of granulation tissue present within the wound, and whether the wound and associated inflammation extended into underlying vertebrae, the spinal canal, and/or spinal cord.
One mouse from each treatment group (placebo, rifampin, and doxycycline) and from representative time points (days 1, 3, and 7) was sacrificed in order to characterize the dissemination of AB5075 within wound tissue via PNA-FISH (33). Tissue was trimmed at 3 μm. One drop of the PNA probe Acinetobacter PNA CP0050 (AdvanDx, Inc., Woburn, MA) was added to each slide, and coverslips were applied. Slides were put on a heating block for 90 min at 55°C in the dark. Slides were then immersed in preheated, 55°C deionized water for less than 1 min, while the coverslips (AdvanDx, Inc., Woburn, MA) were removed. Slides were then immersed in preheated, 55°C 60× wash solution (AdvanDx, Inc., Woburn, MA) for 30 min. Slides were then overlaid with coverslips by using mounting medium (AdvanDx, Inc., Woburn, MA). Separately, cultures of A. baumannii and Escherichia coli were kept as positive and negative controls, respectively. One drop of GN fixation solution (AdvanDx, Inc., Woburn, MA) was added to a slide. One small drop of culture was added to the fixation solution. The slides were then placed on a heating block for 20 min at 55°C to fix the smear. One drop of Acinetobacter PNA CP0050 (AdvanDx, Inc., Woburn, MA) was added to each slide, and slides wre then overlaid with coverslips. Control slides were also prepared, as described above.
All statistical analyses were carried out using GraphPad Prism software. Wound sizes, weight changes, and CFU burdens were compared via the Kruskal-Wallis test followed by Dunn's multiple comparison test. All results were considered significant if P was <0.05.
In a pilot study, using cyclophosphamide-treated mice, we inoculated wounds with AB5075 or left wounds uninoculated and then measured the wound area over time. In the A. baumannii-infected mice, the wound became infected, and the time to wound closure was delayed at least 5 days compared to mice that received no A. baumannii inoculum (see Fig. S1 in the supplemental material). These results and other pilot studies that tested other strains of mice and A. baumannii as well as different dressings (data not shown) allowed us to continue experiments with the conditions set forth in Materials and Methods and validate the model with antibiotic treatment.
The first assessment of the model included a gross pathology comparison of wounds in animals that were infected with A. baumannii and received antibiotic treatment versus those in animals that received no treatment. Photographs of the dorsal, full-thickness wounds were taken on days 3, 8, 15, and 21. On day 3 (Fig. 1A to toC),C), open wound beds without contraction or reepithelization were observed, and on day 8 devitalized epidermal tissue peripheral to wound edges was lightly colored and correlated with post-Tegaderm removal necrosis and wound expansion. On day 15 (Fig. 1D to toF),F), a variation in the wound and the serocellular crust covering the wound was dependent upon treatment, where the treated wounds appeared smaller and less inflamed. On day 21 (Fig. 1G to toI),I), the untreated control wounds maintained serocellular crust with minimal contractile healing, whereas, in contrast, the treated wounds displayed a contraction of the wound and the wounds were either fully closed or had less than 5 mm2 of overlying serocellular crust (Fig. 2).
Wounds were measured via Aranz on days 0, 3, 8, 15, and 21 postinfection. Wound sizes between groups were not statistically different for day 0 when compared via the Kruskal-Wallis test (Fig. 2). When wounds were measured on day 3 postinfection, the placebo group (n = 12) had a median wound size of 39.0 mm2, while rifampin-treated (n = 8) and doxycycline-treated (n = 9) mice had median wound sizes of 35.0 mm2 and 31.0 mm2, respectively (P < 0.05). By day 8, 1 day after the occlusive dressing was removed, the median wound size for placebo-treated mice had increased to 65.5 mm2, while the rifampin-treated wounds stabilized at a median of 34.5 mm2 (P < 0.01) and doxycycline-treated mice exhibited a median decrease to 29.0 mm2 (P < 0.001). On days 15 and 21, the placebo group had median wound sizes of 97.0 mm2 and 22.0 mm2, while the rifampin-treated mice had medians of 5.5 mm2 (P < 0.001) and 0.0 mm2 (P < 0.001). Similarly, the doxycycline-treated mice had medians of 2.0 mm2 (P < 0.001) and 0.0 mm2 on days 15 and 21, respectively (P < 0.001); these results were indicative of the significant differences seen between antibiotic-treated and untreated wounds.
Aside from gross pathology observations and wound measurements performed on the mice, we were also cognizant of clinical signs of infection. One of the signs that showed a significant difference among groups was weight measurement over time. Mice treated with the placebo (n = 12) lost a median of 11.6% of their infection day body weight at 1 day postinfection and lost a maximum median of 20.3% by 2 days postinfection (Fig. 3). Rifampin-treated (n = 8) and doxycycline-treated (n = 9) mice only lost 6.7% (P < 0.01) and 5.3% (P < 0.01) of their body weights, respectively, on day 1 postinfection, and maximally they lost 6.8% (P < 0.001) and 11.8% (P < 0.01) (Fig. 3). When compared via the Kruskal-Wallis test followed by a Dunn's posttest, rifampin-treated mice maintained the lesser weight loss for 5 days postinfection, while doxycycline-treated mice remained significantly different for only 4 days compared to the untreated mice.
In addition to the significant differences in gross pathology among treatment groups, suggesting that an A. baumannii infection could be established and treated in the model, we also wanted to investigate what was happening at the cellular level. Therefore, both standard histopathology and some nonstandard methods to discover the localization of bacteria and the impact of antibiotic treatment on the bacteria and host healing were utilized. Photomicrographs of hematoxylin and eosin (H&E)-stained dorsal wound longitudinal sections were prepared and evaluated. In Fig. 4A to toC,C, the day 7 slides at a 12.5× low magnification demonstrated a varied defect in the widths of the wound bed under the Tegaderm, and all the wounds were devoid of epidermis, with significant inflammatory cell infiltrate at the wound edge. At 40× magnification, the day 15 untreated, control wound lacked reepithelization and was covered by a serocellular crust (Fig. 4D). In contrast, on day 15 in the rifampin and doxycycline treatment groups, a reepithelization over the wound was observed, with granulation tissue surrounding the necrotic cellular debris (Fig. 4E and andF).F). On day 23, the untreated control wound was fully reepithelized but lacked evidence of contractile healing and maintained evidence of an expansive overlying serocellular crust (Fig. 4G). In contrast, the rifampin and doxycycline treatment groups on day 23 showed signs of contractile healing and reepithelized wound beds with less overlying crust and granulation tissue surrounding necrotic cellular debris (Fig. 4H and andII).
We also evaluated the wound and dressing by using PNA-FISH in order to understand where the bacteria were localized in animal tissue sections. The untreated control (Fig. 5) was evaluated on day 1 at 12.5× magnification with H&E staining and at 40× in Acinetobacter-specific PNA probe-stained photomicrographs. The designations A to E in the larger view of the H&E-stained photomicrographs correlate to individual PNA probe pictures. The green fluorescent probes demonstrated A. baumannii cells at all levels of the wound bed from the superficial serocellular crust at the surface to the deep paraspinal musculature. This was in striking contrast to the effects of rifampin treatment at day 1 (see Fig. S2 in the supplemental material). The absence of green fluorescence demonstrated a clearance of A. baumannii cells below the level of detection at all levels of the wound bed, from the superficial serocellular crust to the deep paraspinal musculature (see Fig. S2).
The doxycycline treatment group at day 1 displayed an intermediate phenotype (see Fig. S3 in the supplemental material). As before, the designations A to F in the H&E-stained photomicrograph correlate to the increased magnification of individual PNA probe pictures (see Fig. S3). Green fluorescent probes demonstrated the presence of A. baumannii cells on the surface of the wound and deep to the subdermal fat (see Fig. S3A to D); however, fluorescent cells were absent in the deep paraspinal musculature (see Fig. S3E and F), suggesting systemically delivered (via i.p. injection) doxycycline has less wound bed penetration or less efficacy against bacteria in the wound bed relative to effects in rifampin-injected animals.
The bacterial community composition was determined by 16S rRNA pyrosequencing (Research and Testing Laboratories, LLC, Lubbock, TX). Bacterial communities were sampled from wounds in all groups 24 h postinfection (n = 6, all groups) and showed >95% Acinetobacter spp. community composition in all mice (Fig. 6, top row). By day 3 postinfection, five of seven placebo-treated mice (n = 7) had a wound profile composed of >95% Acinetobacter spp., whereas the other two mice had wounds with Enterobacter spp. present (>10% community composition) (Fig. 6, middle row). Rifampin-treated mice sampled at day 3 (n = 7) all had wounds composed of >95% Acinetobacter spp., while 6/7 wounds from doxycycline-treated mice (n = 7) were composed of >95% Acinetobacter spp., with 1 wound having a significant presence (>10%) of Enterobacter spp. On day 7, the placebo-treated mice (n = 6) all had wounds composed of >95% Acinetobacter spp., while the rifampin-treated (n = 6) and doxycycline-treated (n = 6) groups both had 5/6 wounds comprised of >95% Acinetobacter spp., while one wound from each treatment group was > 60% Staphylococcus (Fig. 6, bottom row).
We conclude from these results that there was not significant contamination from other bacterial species when we utilized the methods of this study. Therefore, the necrotizing tissue damage and inflammation observed (Fig. 1 and and4)4) can be attributed to the A. baumannii that was inoculated into the wound bed. As the bacteria proliferated (see below) over time in the untreated mice, the increase in wound damage and the time to wound closure can also be attributed to the increase in A. baumannii numbers.
An untreated group of mice (n = 6) used to assess the log CFU burden at the time of treatment (4 h postinfection) showed 5.2 × 105 CFU per 4-mm tissue biopsy punch (95% confidence interval [CI], 4.4 × 105 to 5.8 × 105). After 24 h of infection, placebo-treated mice (n = 6) had a median log CFU burden of 6.7 × 106, while rifampin-treated mice (n = 6) had 1-log fewer organisms, with a median log CFU burden of 5.7 × 105 (P < 0.01), and doxycycline-treated mice had a median log CFU of 5.9 × 105 (Fig. 7). Samples taken at 3 days postinfection revealed that placebo-treated mice (n = 7) had a median log CFU burden of 7.0 × 106 per punch, while both log CFU burdens in the rifampin-treated (n = 8) and doxycycline-treated (n = 8) groups decreased to 4.7 × 104 (P < 0.001) and 5.0 × 104 (P < 0.05), respectively. On day 7 postinfection, placebo-treated mice (n = 9) maintained a high median log CFU burden of 7.3 × 106. While still significantly lower than the placebo-treated group, mice in both the rifampin-treated (n = 9) and doxycycline-treated (n = 9) groups showed increases in median log CFU burdens above their day 3 values to 6.1 × 105 (P < 0.01) and 6.2 × 105 (P < 0.05), respectively.
Within 24 h after inoculation, masses of bacteria were visible in punches taken from the wound beds of mice in all treatment groups (Fig. 8A, ,B,B, and andC),C), while initial attachment was visible on the occlusive dressing only in the placebo-treated groups (Fig. 9A) and doxycycline-treated groups (Fig. 9C). There were no substantial numbers of bacteria observed on the dressings removed 1 day postinfection from the rifampin-treated mice (Fig. 9B), and no biofilm formation was observed on dressings on days 3 or 7 postinfection in this group (Fig. 9E and andH).H). By day 3, host-associated matrix proteins, red blood cells, and immune effectors were observed in the wound beds of placebo-treated mice and obfuscated the bacterial cells (Fig. 8D), while a complex biofilm architecture and a continuous substratum of bacteria formed on the occlusive dressing (Fig. 9D). In both rifampin- and doxycycline-treated mice at day 3 postinfection, the wound beds were rife with host-associated matrix proteins, blood cells, and immune effectors, similar to observation for the placebo-treated group (Fig. 8E and andF).F). Doxycycline-treated mice appeared to have pockets of bacteria trapped underneath host-associated matrix proteins on the occlusive dressings on day 3 (Fig. 9F) but lacked the continuous substratum or higher architecture of the biofilms seen in the placebo-treated group.
We sought to develop a wound infection model that utilized A. baumannii as a sole infectious agent and that included multiple measurable outcomes with effects that entailed quantitative endpoints, which permitted small sample sizes for antimicrobial evaluation. In order to develop a new model of A. baumannii wound infection, three important selections were made from the outset of the study. First, female BALB/c mice were chosen, because immune responses in BALB/c mice are skewed more toward a Th2 than a Th1 response (34, 35), and this immune response favors the establishment of infection by the Gram-negative ESKAPE pathogens (36; Mark Shirtliff, personal communication). Second, a cutaneous wound model was selected based on findings of a previous study, where an open wound did not adversely affect animal health during a >15-day protocol (37), which allowed for a large window of time to quantitatively and qualitatively measure drug efficacy in the model. Lastly, it was important that the chosen A. baumannii strain was virulent enough to cause an infection after a small inoculating dose. For this purpose, AB5075 was chosen for this model, as our laboratory has shown this strain is more virulent than the other A. baumannii isolates that were tested (submitted for publication) and, because the antibiotic susceptibilities of AB5075 were known, we could validate the model via i.p. treatment with doxycycline or rifampin. It should be noted, however, that we have evaluated other A. baumannii strains in this model as well (in our pilot studies). In particular, we evaluated AB5711 (23), which was also able to cause an infection that retarded wound healing past 15 days postinoculation (data not shown).
Previous work has shown cyclophosphamide-induced neutropenia allows A. baumannii to establish infections in a pulmonary model when using smaller inoculums than in those for immunocompetent mice (12). We utilized a similar model to identify AB5075 as an isolate that we could use as a model strain for all of our subsequent studies (submitted for publication), and in a similar immunocompromised background we were able to establish infections with AB5075 that resulted in wounds that took an excess of 21 days to close (Fig. 2). In contrast, uninfected wounds closed no later than 13 days postinfection (see Fig. S1 in the supplemental mateiral), and antibiotic-treated wounds closed between day 15 and day 21 (Fig. 2).
The retarded wound closure rate that we observed in our model was accompanied by a weight loss of up to 25% from the infection day body weight at 2 days postinfection (Fig. 3). Therefore, we can use weight loss as an indicator of A. baumannii infection in this model. Gross pathology of infected wounds indicated tissue may begin to devitalize as soon as 3 days postinfection around the wound perimeter and remain visibly swollen until 15 days after inoculation with AB5075. An evaluation of the wound using standard histopathology showed inflammation throughout the large wound perimeter that extended down to the spinal column by day 7 (Fig. 4 and and5).5). Histopathology from infected mice on day 23 revealed wounds had reepithelialized but showed little evidence of wound perimeter contraction (Fig. 4). In order to investigate the dissemination of AB5075 throughout the wound bed, sections of tissue were fixed 24 hours postinfection and interrogated with Acinetobacter-specific PNA probes. The PNA probes revealed that the bacteria had spread rapidly beyond the wound bed through the underlying muscle and down to the spinal column (Fig. 5). A. baumannii appeared to localize in the interstitial spaces between muscle fibers and between epithelial cells (Fig. 5).
The localization of A. baumannii in the interstitial spaces between cells is an interesting observation, given previous reports of A. baumannii invasion of human epithelial cells at low levels in tissue culture experiments (38, 39, 40, 41). While there is agreement between study groups that this invasion of tissue culture cells takes place (38), it is not clear if invasion occurs in vivo or in human patients. We did not observe intracellular bacteria in any of our histopathological evaluations of mouse tissue (Fig. 4 and and55 and unpublished data), but we hypothesize that the proteins identified in these previous studies may also be involved in the movement of A. baumannii into these interstitial spaces. For example, phospholipase D is believed to facilitate bacterial invasion by breaking down the phospholipid membrane; however, this process could also be utilized to allow passage of bacteria through the junctions between eukaryotic cells (39). Further study of the mechanisms involved are required to determine if this is true.
Regardless of the role invasion plays in vivo, the sequelae of infection that we observed postinoculation are consistent to what is observed in skin and soft tissue infections of human patients. For example, we found that A. baumannii penetrated the layers of tissue from the initial wound bed inoculum all the way to bone tissue (Fig. 5), and this was consistent with the tissue penetration observed in A. baumannii-infected patients (42), as well as the consideration of A. baumanii as a potential cause of osteomyelitis (43).
Biopsy punches of infected tissue further revealed the CFU burden reached ~1 × 107 CFU/4-mm punch within 24 h of infection (Fig. 7); this burden was maintained or exceeded that level on day 3 and day 7 postinfection. One interesting observation was that the CFU burden increased on day 7 in the antibiotic-treated animals. We believe this increase in CFU is due to the scab formation over the wound bed after the dressing is removed. The bacteria are concentrated in the scab, so that when a punch biopsy is taken through the scab into the wound tissue, an increase in CFU is seen. It should be noted that despite the rise in CFU, the wounds still closed faster (Fig. 1), and these results were likely due to the immune system (neutrophils) and wound healing, which is similar to the scab formation observed in another murine wound infection model (44).
To ensure the infection should be attributed to AB5075 and not to other commensals or contaminant pathogens, 16S rRNA community profiling was conducted in parallel to the CFU enumerations. Nearly all samples taken from infected but untreated wounds showed a >95% dominance of Acinetobacter with only a minor fraction of communities composed of Enterobacter in two mice from this group (Fig. 6). While immunosuppression is required to establish an infection and it is agreed that A. baumannii is a mild pathogen compared to other bacterial species, it is evident from the microbiome evaluation in this study that A. baumannii can establish a wound infection without other bacterial species being major contributors. In pilot studies, we also found that higher inoculums (CFU of ≥106) resulted in sepsis and severe animal morbidity (data not shown), which also suggests that increased A. baumannii numbers can lead to more severe sequelae that do not require other bacterial species.
A final observation from this wound model of infection was the formation of robust biofilms within the wound bed (Fig. 8) and on the occlusive dressing above the wound bed (Fig. 9). While it is not surprising that A. baumannii formed biofilms in this model, given previous work showing the importance of biofilms (38, 40), it was somewhat unexpected that AB5075 achieved the levels of biofilm formation observed, given that this strain does not form robust biofilms in vitro (unpublished data). Therefore, perhaps there are cues in the in vivo environment that trigger biofilm assembly, in particular with this strain, that have not been discovered. Regardless, biofilm formation in the model provides another metric for measuring antimicrobial efficacy. In fact, given that many laboratories are now developing antibiofilm strategies (45), the SEM data provided in the implementation of this model provide a qualitative assessment of an antibiofilm product.
Overall, the wound infection model we have developed uses all of the endpoints described above to provide a robust data set for evaluating infection and antimicrobials used for treatment. We believe the model will be a powerful tool to evaluate new antimicrobials for not only A. baumannii but also for many difficult bacterial infections caused by the MDR ESKAPE pathogens. We are currently using this model as a platform to test novel antimicrobials against Klebsiella pneumoniae and Staphylococcus aureus infections as well (unpublished data). While in this study only a systemic application was evaluated, other studies to determine the efficacy of new topical antimicrobial treatments are also ongoing (unpublished data). Furthermore, we are currently developing bioluminescent strains that can be utilized to monitor bacterial numbers in vivo without sacrificing animals. However, it is unclear how predictive this murine wound model will be with regard to efficacy in humans, until extensive pharmacokinetic and pharmacodynamic experiments are performed with known antibiotic regimens. That said, the antibiotic dosing that was used herein was similar to what has been utilized in other animal models (12, 25, 26) and somewhat similar to what is used clinically (47, 48, 49). Therefore, we hope that the employment of such a model could be an important first, preclinical step in determining the efficacy of antibiotics in development for skin and soft tissue infections caused by bacterial pathogens.
We kindly acknowledge Anna Jacobs for critical reading of the manuscript. We also thank AdvanDx, Inc., for their technical assistance with the A. baumannii PNA-FISH probes. Lastly, we thank Mark Shirtliff for his expertise and advice with animal modeling.
The research presented here was supported and funded via multiple grants from the Military Infectious Diseases Research Program (MIDRP) and the Defense Medical Research and Development Program (DMRDP).
The findings and opinions expressed herein belong to the authors and do not necessarily reflect the official views of the WRAIR, the U.S. Army, or the Department of Defense.
Published ahead of print 16 December 2013
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AAC.01944-13.