Diabetic murine models for Acinetobacter baumannii infection

Abstract

Objectives

Extremely drug-resistant (XDR; i.e. resistant to all antibiotics except colistin or tigecycline) Acinetobacter baumannii has emerged as one of the most common and highly antibiotic-resistant causes of infection. Diabetes is a risk factor for acquisition of and worse outcomes from A. baumannii infection. We sought to develop diabetic mouse models of A. baumannii bacteraemia and pneumonia and validate these models by comparing the efficacy of antibiotic treatment in these models with the established neutropenic mouse models.

Methods

Diabetic or neutropenic mice were infected via intravenous inoculation or inhalation in an aerosol chamber with an XDR A. baumannii. Treatment with colistin started 24 h after infection and continued daily for 7 days. Survival served as the primary endpoint while tissue bacterial burden and histopathological examination served as secondary endpoints.

Results

Lethal infection was achieved for the neutropenic and diabetic mice when infected intravenously or via inhalation. Neutropenic mice were more susceptible to infection than diabetic mice in the pneumonia model and equally susceptible in the bacteraemia model. Both models of bacteraemia were sensitive enough to detect virulence differences among different clinical strains of A. baumannii. In the pneumonia model, colistin treatment was effective in improving survival, reducing lung bacterial burden and histologically resolving the infection compared with placebo only in diabetic mice.

Conclusions

We developed novel models of A. baumannii bacteraemia and pneumonia in diabetic mice. These models can be used to study mechanisms of infection, develop immunotherapeutic strategies and evaluate drug efficacies against highly lethal A. baumannii infections.

Introduction

In the last decade, Acinetobacter baumannii has become one of the most common and highly antibiotic-resistant pathogens in the USA and throughout the world.^1,2 Extremely drug-resistant (XDR; i.e. resistant to carbapenems and all other antibiotics except colistin or tigecycline) A. baumannii causes a wide range of nosocomial infections, including ventilator-associated pneumonia, skin and soft-tissue infections, wound infections, urinary tract infections, secondary meningitis and bloodstream infections.^1,3,4 These infections are associated with prolonged hospitalization, tremendous healthcare costs and high rates of death despite treatment.^1,3,4 For instance, bacteraemia caused by XDR A. baumannii causes >50%–60% mortality rates even with antibiotic therapy.^5,6

Animal models are very useful for studying mechanisms of infection, developing novel strategies to prevent or treat diseases and evaluating the efficacy of antibiotic use. Recent clinical data indicate that diabetes is a risk factor for acquisition of and worse outcomes from A. baumannii infection.^7–9 Although robust models of A. baumannii infection have been previously described (including those utilizing immunocompetent and immunocompromised mice),^10–15 none of them utilized diabetic animals. We have extensively published the streptozotocin-induced diabetic mouse model for studying mucormycosis.^16–19 Additionally, we published extensively on using the cyclophosphamide-induced neutropenia mouse model for a variety of opportunistic infections.^17,20 Therefore we sought to compare diabetic mice with immunocompetent or neutropenic mice in their susceptibility to A. baumannii infections. Furthermore, we validated the model by comparing A. baumannii pneumonia in diabetic or neutropenic mice in their susceptibility to colistin treatment.

Methods

Bacterial strains and growth medium

Six clinical isolates of A. baumannii were used. Five of the strains (HUMC1, HUMC4, HUMC5, HUMC6 and HUMC12) were isolated from patients seen at Harbor-UCLA Medical Center.²¹ HUMC1 was isolated from the blood and sputum of the patient. HUMC4, 5 and 6 were isolated from patients with ventilation-associated pneumonia, whereas HUMC12 was isolated from a diabetic stump wound. A. baumannii ATCC 17978 was isolated in 1951 from a 4-month-old infant with fatal meningitis.²² All HUMC isolates were found to be resistant to all antibiotics except for colistin (XDR).²¹ Briefly, these strains were resistant to the following antibiotics with MICs (mg/L) as follows: amikacin, >128; gentamicin, >128; aztreonam, ≥32; ampicillin, ≥16; sulbactam, ≥8; piperacillin, <128; tazobactam, 4; cefepime, 16; meropenem, ≥4; imipenem, ≥2; ertapenem, ≥16; doripenem, ≥4; ciprofloxacin, ≥64; tigecycline, 4; and colistin, 2. Bacteria were grown at 37°C in tryptic soy broth (Difco Laboratories, Sparks, NV, USA). Cells were passaged for 3 h in the same medium (logarithmic-phase cells) prior to use in the animal models.

PFGE

To determine whether the six A. baumannii strains used in this study are distinct from each other, PFGE was carried out using a modification of a previously described method.²³ In brief, genomic DNA was extracted from A. baumannii strains grown in brain heart infusion medium. The DNA was digested with the restriction endonuclease ApaI (New England Biolabs Ipswich, MA, USA). DNA fragments were subjected to PFGE in 1.0% agarose (w/v) in 1.0× TBE (0.5 M Tris/0.1 M boric acid/0.2 mM EDTA, pH 8.0) buffer on a CHEF-DR III system (Bio-Rad Laboratories, Richmond, CA, USA). The running conditions were 6 V/cm at 14°C for 20 h with an initial switch time of 5.0 s and a final switch time of 42 s, and included an angle of 120°. Salmonella braenderup H-9812 digested with XbaI restriction enzyme was used as a molecular size marker.

Animal models

BALB/c, ICR or A/J mice were all purchased from Taconic (Germantown, NY, USA). Mice were made diabetic as previously described by intraperitoneally (ip) injecting a single dose of streptozotocin (200 mg/kg) 10 days prior to infecting with A. baumannii.^16–19 Although streptozotocin is selectively toxic to the β-cells of the pancreatic islets, it has been shown to have no effect on phagocyte counts when administered to rodents.^24,25 We also published extensively on using the cyclophosphamide-induced neutropenic mouse model for a variety of opportunistic infections.^17,20 We adapted this model to develop A. baumannii infections because it is relevant to highly lethal A. baumannii infections in patients with cancer/neutropenia.^26,27 Mice were immunosuppressed with cyclophosphamide (200 mg/kg administered ip) and cortisone acetate (250 mg/kg administered subcutaneously) on day −2 and +3 relative to infection. This treatment regimen results in 9 days of leucopenia with total white blood cell count dropping from 3000 to <500/mm.^3,26

For the bacteraemia model, mice were infected intravenously (iv) via the tail vein. For the pneumonia model, mice were infected by aerosolizing bacterial cells in an inhalational chamber through a nebulizer, as we previously described.²⁸ Briefly, mice were introduced via a hinged doorway to a Plexiglas exposure chamber (South Bay Plastics, Torrance, CA, USA). The bacteria cells (1.0 × 10¹¹ cells/mL) were introduced by aerosolizing a 12 mL suspension with a small-particle nebulizer (Hudson Micro Mist, Hudson RCI, Temecula, CA, USA) driven by compressed air at 100 lb/in². The nebulizer was connected to a channel that ran along the top of the chamber and vented the aerosol from the middle of the chamber ceiling. A standard exposure time of 1 h was used for all experiments to allow time for complete aerosolization and uniform exposure of the mice. The entire apparatus was contained within a laminar flow hood in a negative pressure room. To determine the inhaled inoculum, three mice from each experiment were sacrificed immediately after the procedure and their lungs collected and quantitatively cultured on tryptic soy agar (‘TSA’) plates.

Drug treatment

Colistin treatment was evaluated in the diabetic or neutropenic pneumonia models. Colistin (2.5 mg/kg) treatment was given twice daily via ip injection for 7 days starting on day +1 relative to infection. Placebo mice received diluent (PBS). Survival time was used as the primary endpoint, while tissue bacterial burden and histopathological examination served as secondary endpoints. All procedures involving mice were approved by the institutional animal use and care committee, according to the NIH guidelines for animal housing and care.

Statistical analysis

The Mann–Whitney U-test was utilized to compare tissue bacterial burden. The non-parametric log-rank test was utilized to determine differences in survival time. P < 0.05 was considered significant.

Results

Diabetic or neutropenic mice are susceptible to A. baumannii bacteraemia

Consistent with previously published results, in which A. baumannii pneumonia and bacteraemia spontaneously cleared in immunocompetent mice,²⁹ pilot studies confirmed that immunocompetent mice were resistant to fatal A. baumannii bacteraemia and pneumonia regardless of the mouse strain used (i.e. BALB/c, ICR or A/J mice) (data not shown). Thus we utilized the diabetic or the neutropenic mouse models of bacteraemia. In both neutropenic (Figure 1a) and diabetic (Figure 1b) mouse models, lethal infection was established in an inoculum-dependent manner. For example, in neutropenic mice inocula of A. baumannii HUMC1 at 8.9 × 10⁷, 1.2 × 10⁷ and 7.3 × 10⁶ caused 100%, 75% and 25% mortality, respectively, by day 21 post-infection (Figure 1a). Similarly, in the diabetic mouse model, inocula of ≥4.1 × 10⁷ caused 100% mortality within 48 h of infection, whereas smaller inocula resulted in no deaths (Figure 1b).

Figure 1.

Cyclophosphamide + cortisone acetate treatment (neutropenic) or diabetes renders BALB/c mice susceptible to A. baumannii HUMC1 bacteraemia. Survival of neutropenic (a) or diabetic (b) mice (n ≥ 10 per arm) following iv challenge with different inocula of A. baumannii. Uninfected neutropenic mice were included in the study to monitor for possible mouse death due to contaminants.

We wanted to evaluate both bacteraemia models in their ability to differentiate virulence among several clinical isolates of A. baumannii. The six strains used in this study belonged to four different groups based on the PFGE analysis, with strains HUMC4, 5 and 6 showing the same DNA splicing pattern (Figure 2a). HUMC4, 5 and 6 also demonstrated similar susceptibility to antibiotics,²¹ thereby confirming their clonal similarity. Neutropenic or diabetic mice were infected with the six strains of A. baumannii at an intended inoculum of 5 × 10⁷. The neutropenic and diabetic models were sensitive enough to detect virulence differences among strains. We found HUMC1 to be the most virulent strain, with 100% mortality within 24 h and 6 days in the neutropenic model and diabetic model, respectively (Figure 2b and c). In contrast, HUMC12 was moderately virulent in the neutropenic model and avirulent in the diabetic model, while strain ATCC 17978 had no to minimal virulence in both models. Finally, HUMC4, 5 and 6 (which are clonally similar) had similar moderate virulence in the neutropenic model, whereas in the diabetic model HUMC6 was more virulent than HUMC4 and 5, but the actual inoculum concentration used to infect mice with HUMC6 was greater than that used for HUMC4 and 5 (1.2 × 10⁸ cells for HUMC6 versus 5 × 10⁷ for HUMC5 and 3.0 × 10⁷ for HUMC4) (Figure 2b and c). Hence these two models can separate isolates into high, intermediate and low virulence groups.

Figure 2.

Neutropenic and diabetic bacteraemia models differentiate the virulence of A. baumannii clinical isolates. PFGE demonstrating the differences in genotypes among the six A. baumannii clinical isolates used in this study (a). Survival of neutropenic mice (b) or diabetic mice (c) infected via tail vein injection with different clinical isolates of A. baumannii at a 5 × 10⁷ inoculum. n = 8 for all strains except for ATCC 17978, HUMC1 and HUMC4 in the neutropenic mice in which n = 7 and HUMC12 in diabetic mice in which n = 6. *P < 0.005 versus all other treatments, **P < 0.05 versus HUMC1 and HUMC12, and ‡P < 0.007 versus HUMC1 and HUMC6 by log-rank test.

Diabetic or neutropenic mice are susceptible to A. baumannii pneumonia

Aerosolization of 1.0 × 10¹¹/mL bacterial cells consistently delivered an inoculum ranging from 1.0–3.3 × 10⁶ cells to the mice lungs. Although this inoculum caused almost 100% mortality in both models, neutropenic mice were more susceptible to infection than diabetic mice (median survival time of 5 days versus 11 days, P = 0.002 by Wilcoxon rank sum test) (Figure 3a). Histopathological examination of lungs harvested from diabetic mice expiring from the infection showed acute inflammation in the alveolar walls and perivascular spaces (Figure 3c) characterized by neutrophil infiltration (Figure 3d), as previously reported.¹⁰ Similarly, lung histopathology from neutropenic mice infected with A. baumannii via inhalation had diffused damage resulting in oedema and polymerized fibrin, which resulted from leaked plasma due to alveolar and capillary damage (Figure 3f). Further, inflamed lungs demonstrated numerous swollen type II pneumocytes (Figure 3g).

Figure 3.

Neutropenic or diabetic mice are susceptible to A. baumannii pneumonia. (a) Survival of mice (n = 9 for neutropenic and 10 for diabetic) infected by aerosolization of A. baumannii (inhaled inoculum of 2.1 × 10⁶). Control mice were either neutropenic or diabetic, and were not infected. Histopathological examination of lung sections from diabetic mice (b–d; stained with H&E) or neutropenic mice (e–g; stained with Gram's stain) demonstrate acute perivascular and alveolar inflammation accompanied by diffuse damage resulting in oedema. Arrows in (d) demonstrate neutrophil infiltration and the arrow in (g) illustrates swollen type II pneumocytes. The circle in (f) shows oedema with polymerized fibrin. Magnification is 100× for (b), (c) and (e), 400× for (d) and 1000× for (f) and (g). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Colistin was effective in treating A. baumannii pneumonia in the diabetic but not in the neutropenic mouse models

To evaluate drug treatment efficacy in treating A. baumannii infections and investigate if outbred mice are equally susceptible to infection, the pneumonia model was chosen over the bacteraemia model because the former model allows ample time for drug treatment [median survival time for the pneumonia model ranges from 5 to 11 days versus 1 to 3 days for the bacteraemia models (compare Figure 1 with Figure 3a)]. A. baumannii HUMC1 was chosen because this strain consistently resulted in >90% lethality to infected mice and it was only susceptible to colistin (MIC = 2 mg/L).²¹ Diabetic or neutropenic ICR mice were infected with A. baumannii HUMC1 via inhalation, as above, then treated with colistin after infection. Diabetic or neutropenic ICR mice were equally susceptible to A. baumannii pneumonia as BALB/c mice [50% mortality on day 10 for ICR mice versus day 11 for BALB/c mice in the diabetic model (Figure 3a and Figure 4a) and 50% mortality on day 6 for ICR mice versus day 5 for BALB/c mice in the neutropenic mice (Figure 3a and Figure 5)]. Treatment with colistin improved 40 day survival of infected diabetic mice compared with placebo (Figure 4a) (80% survival for colistin versus 0% survival for placebo, P < 0.007 by log-rank test). However, colistin had no effect on survival time when used in the neutropenic mice (Figure 5). Additionally, colistin therapy resulted in an ∼2.0 log decrease in lung bacterial burden compared with placebo (Figure 4b). Finally, lungs harvested 7 days after colistin treatment had normal histology compared with placebo-treated mice, which had acute inflammatory neutrophil infiltration and increased type II pneumocytes (Figure 4c).

Figure 4.

Colistin treatment is effective in treating A. baumannii pneumonia in ICR diabetic mice. (a) Survival of diabetic mice (n = 18 per arm from two independent experiments with similar results) with pneumonia (average inhaled inoculum = 3 × 10⁶ bacterial cells) treated with colistin or placebo. *P < 0.007 for colistin versus placebo by log-rank test. bid, twice daily. (b) Lung bacterial burden in mice (n = 16 for placebo and 13 for colistin from two independent experiments with similar results) infected via inhalation. **P = 0.002 versus placebo by the non-parametric Mann–Whitney test. The y-axis reflects the lower limit of detection of the assay. Bars = medians. (c) Histopathological examination of lung sections stained with H&E showing resolution of inflammation in the mice treated with colistin. Arrows in ‘Placebo’ denote neutrophils and type II pneumocytes. Magnification is 400×. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Figure 5.

Colistin treatment is not effective in treating A. baumannii pneumonia in ICR neutropenic mice. Mice (n = 8 for placebo and 9 for colistin) were made neutropenic with cyclophosphamide + cortisone acetate then infected with an inhaled inoculum of A. baumannii HUMC1 (3.3 × 10⁶ bacterial cells). Drug treatment started 24 h post-infection and continued for 7 days.

Discussion

Murine models of A. baumannii pneumonia have used cyclophosphamide to induce neutropenia or porcine mucin to establish a foreign body that is then superinfected with A. baumannii.^10,30,31 Most of these studies focused on evaluating drug treatment against highly resistant strains to determine efficacies of colistin,^{10,12,32–34} tigecycline¹⁴ or rifampicin¹⁵ monotherapies or combination therapies^13,15,33 against carbapenem-resistant strains. We established a novel diabetic mouse model of A. baumannii bacteraemia and pneumonia because patients with diabetes are at increased risk for acquisition of and worse outcomes from A. baumannii infection.^7–9 Lethal infection developed within 1 week for bacteraemia and ∼1–3 weeks for pneumonia. Further, our data confirmed that diabetic mice are as susceptible as neutropenic mice and far more susceptible to infection than immunocompetent mice since A. baumannii pneumonia and bacteraemia spontaneously clear in immunocompetent mice.³⁰ Although diabetes does not result in low phagocyte counts like the neutropenic mouse model,^25,28 several studies have shown that phagocyte functions including chemotaxis and oxidative burst are impaired by diabetes,^24,35,36 thus explaining the susceptibility of diabetic mice to A. baumannii infection. Moreover, in the absence of a robust immunocompetent mouse model of A. baumannii infection, the diabetic mouse model with its intact leucocyte count²⁵ represents a suitable model to study immunotherapeutic approaches against the disease. In this regards, we recently utilized the diabetic bacteraemia model in studying active and passive vaccination approaches against A. baumannii infection using outer membrane protein A (OmpA) as a vaccine antigen.²¹

Acinetobacter is considered to be a low-grade pathogen, and can remain on or in the human body without causing illness. Patients who are at risk of developing the disease are generally somehow immunocompromised.²⁹ For example, nosocomial infections are usually seen in patients who are subjected to an invasive procedure, especially in the ICU setting.^29,37 Additionally, community-acquired infections are commonly seen as acute pneumonia generally in patients with a history of alcohol abuse, diabetes, cancer or bronchopulmonary disease.²⁹ These clinical observations are corroborated by the fact that immunocompetent animals are generally resistant to infection with A. baumannii, even when the organism is introduced at high inocula.²⁹ Our results showed that neutropenic mice were gradually susceptible to infection with strain HUMC1 (Figure 1a). In contrast, the same strain was virulent only at high inocula, with rapid death occurring within 48 h when the diabetic model was used (Figure 1b). These results can be explained by the decreased pathogenicity of A. baumannii. In the absence of host defence against the disease, as in the neutropenic mouse model, the mortality of mice is likely to be dependent almost entirely on the bacterial inoculum introduced. In contrast, in the presence of a host defence that is somewhat compromised by underlying factors, as in the diabetic model, phagocytes are expected to be efficient in clearing infections caused by lower inocula but not those caused by higher inocula.

We also demonstrate that the diabetic bacteraemia model can be equally effective as the neutropenic model in differentiating the virulence of A. baumannii clinical strains, with A. baumannii HUMC1 being the most virulent in both models. This can be explained by the fact that our PFGE studies clearly show HUMC1 to be different from the other tested isolates, with HUMC1 being typed to ST206, while the other HUMC strains belong to ST208 and the ATCC 17978 strain belongs to ST112.²¹ Another finding is that strain HUMC12 was not very virulent in the diabetic bacteraemia model but was the second most virulent strain in the neutropenic model. Similar findings with varying virulence in neutropenic versus diabetic murine models have been reported for other human pathogens^17,38 and could reflect the varying susceptibility of the studied organism to host defence mechanisms, including phagocyte killing.

Prior studies using immunocompetent and cyclophosphamide-treated mice infected with A. baumannii demonstrated efficacy when colistin was administered three or four times daily.^10,12,15,33 In the current study we were able to show that colistin was effective in prolonging survival, reducing lung bacterial burden and resolving pneumonia symptoms in diabetic mice but not in the neutropenic model when administered twice daily. An explanation for the difference in colistin efficacy between the two models and with prior studies is the possible effect of the drug on enhancing the immune response to infection. Colistin has been reported to have an immunomodulatory effect on neutrophils, which can increase the ability of the phagocyte to kill bacteria by enhancing neutrophil elastase activity.³⁹ This explanation is further supported by our data showing a lack of activity of colistin in the neutropenic model. Thus the activity of colistin in the diabetic model is likely not limited to directly killing the bacterium.

In summary, the diabetic mouse models of A. baumannii bacteraemia and pneumonia add to the robust animal models currently available to study the pathogenesis of A. baumannii infection and the development of therapeutic strategies against this highly lethal disease. The advantage of the diabetic model for A. baumannii infection, aside from its relevance to human infection, is the use of a less extreme immune defect to unmask virulence factors in the organism and enable testing of therapeutic regimens in the presence of leucocytes.

Funding

This work was supported by Public Health Service grants R01 AI063503 and R21 AI082414 to A. S. I.

Transparency declarations

None to declare.