Do Biofilm Formation and Interactions with Human Cells Explain the Clinical Success of Acinetobacter baumannii?

Background The dramatic increase in antibiotic resistance and the recent manifestation in war trauma patients underscore the threat of Acinetobacter baumannii as a nosocomial pathogen. Despite numerous reports documenting its epidemicity, little is known about the pathogenicity of A. baumannii. The aim of this study was to obtain insight into the factors that might explain the clinical success of A. baumannii. Methodology/Principal Findings We compared biofilm formation, adherence to and inflammatory cytokine induction by human cells for a large panel of well-described strains of A. baumannii and compared these features to that of other, clinically less relevant Acinetobacter species. Results revealed that biofilm formation and adherence to airway epithelial cells varied widely within the various species, but did not differ among the species. However, airway epithelial cells and cultured human macrophages produced significantly less inflammatory cytokines upon exposure to A. baumannii strains than to strains of A. junii, a species infrequently causing infection. Conclusion/Significance The induction of a weak inflammatory response may provide a clue to the persistence of A. baumannii in patients.


Introduction
With the recent description of Acinetobacter bereziniae (genomic species (gen. sp.) 10), A. guillouiae (gen. sp. 11) [1], A. venetianus [2] and A. soli [3], the genus Acinetobacter comprises 23 named species and 11 species with provisional designations. Of these, A. baumannii and the closely related A. gen. sp. 3 and 13TU are clinically the most relevant. Strains of these species have the ability to colonize and spread among critically ill hospitalized patients. Outbreaks of multidrug resistant A. baumannii strains have been observed worldwide [4,5]. A striking manifestation is the dramatic increase in the frequency of imipenem resistant Acinetobacter isolates in US hospitals [6] and the recent occurrence of infection in severely injured soldiers during the conflicts in Iraq and Afghanistan [7]. Three major lineages of genetically highly related A. baumannii strains, designated European clone I, II and III, have been found to be frequently implicated in outbreaks [8,9]. Other Acinetobacter species, such as A. junii, A. johnsonii and A. lwoffii that can frequently be found on the human skin are only incidentally involved in infection, which usually has a mild course [4]. This suggests differences in the pathogenic potential among Acinetobacter species.
The high prevalence of A. baumannii strains in the hospital in epidemic and endemic situations might be explained by several factors, including their resistance to antibiotics [10] and desiccation [11], their ability to form biofilms on medical devices [12], and to colonize skin and mucosal surfaces of vulnerable hosts [13,14]. Adherence of bacteria to host cells is generally considered to be an essential initial step in the colonization process [15]. Once the primary colonizing bacteria have attached to a surface, microcolonies are formed after which bacteria may secrete exopolysaccharides resulting in a highly structured sessile microbial community, the biofilm [16]. Several studies have documented the ability of A. baumannii to adhere to epithelial cells and to form biofilms on glass and plastic surfaces [12,17,18]. Adherent bacteria can interact with cells of the host defense systems resulting in the release of cellular mediators and effector molecules, such as interleukin (IL)-6 and IL-8 and antimicrobial peptides, which help to eradicate the pathogen [19].
Little is known about the pathogenicity of A. baumannii. Moreover, the scarce reports on the virulence of A. baumannii are focused on one or a few strains only. The purpose of the present study was to obtain insight into the factors that might explain the clinical success of A. baumannii. To this aim, biofilm formation was investigated for a large set of well-described A. baumannii strains that differed in epidemicity and clonality. Next, biofilm formation by A. baumannii was compared to that of other Acinetobacter species, including A. gen. sp. 3 and 13TU, A. calcoaceticus and A. junii. For a subset of A. baumannii and A. junii strains, adherence to airway epithelial cells and induction of inflammatory cytokine production by these cells and cultured human macrophages was quantitated. Furthermore, the presence of pilus-like structures that may play a role in adherence and biofilm formation was assessed with scanning electron microscopy (SEM).

Biofilm formation
Biofilm formation on plastic at 28uC and 37uC was first investigated for a comprehensive set of A. baumannii strains. The results revealed a large variation in biofilm formation among A. baumannii isolates; the results at 28uC and 37uC did not differ (Fig. 1). There was no difference in the median biofilm size between strains from outbreaks (0.9; 0-1.8 a.u.) and those not assumed to be from outbreaks (0.8; 0.1-2.8 a.u.). Strains of European clone II (1.1; 0.6-1.8 a.u.) formed larger (p#0.05) biofilms than strains of clone I (0.8; 0-1.0 a.u.), but not larger than strains of clone III (1.0; 0.6-1.3 a.u.). Multidrug resistant strains (0.8; 0-1.8 a.u.) did not form larger biofilms than susceptible strains (0.8; 0.1-2.8 a.u.). Furthermore, no association between biofilm formation and body site of isolation was found (data not shown).
Next, biofilm formation by A. baumannii strains was compared to that by strains of other Acinetobacter species. No significant differences in the median size of the biofilms at 28uC were found between clinically relevant species, i.e., A. baumannii (0.8; 0-2.8 a.u.), A. gen. sp. 3 (1.4; 1.3-1.5 a.u.) and A. gen. sp. 13TU (0.4; 0.2-0.7 a.u.), and the other Acinetobacter species, i.e., A. calcoaceticus (1.1; 1.0-1.1 a.u.) and A. junii (1.6; 0.5-3.0 a.u.). Similar results were obtained at 37uC except for strains of A. gen. sp. 13TU that formed smaller (p#0.05) biofilms than all other species. Of note, the number of bacteria in the supernatants at the end of each experiment did not differ much between the different strains (data not shown).

Adherence to human airway epithelial cells
For further comparison of clinically relevant to less-relevant Acinetobacter strains and species, experiments exploring the ability of strains to adhere to and induce cytokines by human cells were conducted with a subset of strains (further referred to as subset), including six strains of A. baumannii and six strains of A. junii, a species infrequently causing infections ( Table 1). All selected strains were epidemiologically unrelated and were genotypically diverse as assessed by AFLP genomic fingerprinting.
Adherence to human airway epithelial H 292 cells varied widely among the Acinetobacter strains (Table 1). No significant difference in the percentage of H 292 cells associated with bacteria was observed between A. baumannii (30; 7-55%) and A. junii (23; 11-70%). In addition, no difference in the number of A. baumannii and A. junii per positive H 292 cell was seen; the median number of bacteria per H 292 cell was 3 (range 1-5). Of note, differences in the inoculum (within the set range of 7610 6 -4610 7 ) between the various experiments did not influence the outcome of the adherence assay. Furthermore, cell monolayers remained intact and the morphology of the cells was not affected by the bacteria (data not shown).

Cytokine production by human airway epithelial cells in response to Acinetobacter
Pilot experiments demonstrated that 1610 8 CFU bacteria induced higher (p#0.05) levels of the major cytokines IL-6 and IL-8 in H 292 cells than 1610 6 and 1610 7 CFU did. Furthermore, stimulation with live bacteria resulted in higher (p#0.05) IL-6 and IL-8 production by H 292 cells than heat-inactivated bacteria did. Time-course experiments demonstrated that the levels of IL-6 and IL-8 were higher (p#0.05) after 24 h than after 6 h stimulation. Therefore, further stimulation experiments were performed for 24 h with 1610 8 CFU live bacteria.
Results revealed that H 292 cells produced less (p#0.05) IL-8 in response to A. baumannii strains (1.8; 1.0-3.5 ng/ml) than to A. junii Since live bacteria were used in this assay, there was a possibility of outgrowth that may have caused the difference in cytokine induction. However, the number of bacteria in the supernatants at the end of each experiment did not differ much between A. baumannii and A. junii (range 8610 6 -7610 7 CFU/ml). Of note, cell monolayers remained intact and the morphology of the cells was not affected by the bacteria (data not shown).
Finally, preliminary experiments demonstrated that exposure of primary human bronchial epithelial cells of a single donor to six A. baumannii strains (in triplicate) resulted in lower (p#0.05) levels of IL-8 (0.9; 0.5-3.6 ng/ml) than to six A. junii strains (4.5; 0.9-30.7 ng/ml). Cytomix induced 8.6 (7.4-11.5) ng/ml IL-8 in these cells. Of note, no IL-6 was produced by these primary cells upon exposure to these two Acinetobacter species or to cytomix.
Cytokine production by cultured human macrophages in response to Acinetobacter In tissues, macrophages trigger an adequate innate immune response upon encountering pathogens. In this regulatory process, macrophages serve a dual purpose. Initially, they contribute to the elimination of pathogens and the elicitation of an inflammatory reaction. When the infection recedes due to removal of the pathogens, their function may shift toward resolution of inflammation and tissue repair. In line with this notion, we investigated cytokine production by two clearly distinct types of macrophages, i.e., pro-inflammatory macrophages (further referred to as type 1 macrophages), and macrophages with an anti-inflammatory/proangiogenic phenotype (type 2 macrophages) [20], upon exposure to A. baumannii and A. junii strains. Results revealed that macrophage type 1 produced significantly (p#0.05) less tumor necrosis factor (TNF)a, IL-12p40, IL-10 and IL-8 in response to strains of A. baumannii [195 (108-244) ng/ml TNFa, 7 (2-11) ng/ ml IL-12p40, 22 (0-149) pg/ml IL-10 and 49 (27-66) ng/ml IL-8] than to strains of A. junii [650 (458-812) ng/ml TNFa, 130 (111-155) ng/ml IL-12p40, 764 (126-1587) pg/ml IL-10 and 111 (48-208) ng/ml IL-8; Fig. 3]. A. baumannii strains also induced less (p#0.05) inflammatory cytokines in macrophage type 2 than A. junii strains (Fig. 3). Of note, cell monolayers remained intact and the morphology of the cells was not affected by the bacteria (data not shown).

Electron microscopy analysis of bacterial surface structures
Pili have been described to be involved in biofilm formation, adherence and the induction of an immune response [21]. Therefore, we performed SEM to assess the presence of such surface structures on four A. baumannii strains that differed in their ability to form biofilm and adhere to human cells. SEM of bacteria cultured for 16 h at 37uC on blood agar plates revealed two types of cell appendages: short pilus-like structures and long extensions (Fig. 4). The latter varied in length, were irregularly distributed over the cell surface, and sometimes connected bacteria. The pilus-like structures were detected in A. baumannii strain LUH1398, a large biofilm former and highly adherent, in LUH7312, a small biofilm former and highly adherent, and RUH3023 T , a large biofilm former but poorly adherent (Fig. 4, white arrows). Long cell extensions were seen in A. baumannii strain RUH3023 T and RUH3239, a small biofilm former and poorly adherent (Fig. 4, black arrows). In addition to these structures, there was a marked surface heterogeneity, from smooth (LUH7312) to pockmarked (LUH1398). In contrast to A. baumannii strain RUH3023 T and RUH3239, strain LUH1398 and LUH7312 formed only a few cell clusters with no more than four cells grouped together on the glass slides (data not shown). No difference in structural features was observed between strains cultured at 37uC and 30uC.

Discussion
This study was undertaken to obtain some insight into the mechanisms underlying the clinical success of A. baumannii. The main conclusion from the present study is that strains of A. baumannii induced a poor inflammatory response in human cells, despite the finding that they adhered well to these cells. This conclusion is based on the following findings.
First, airway epithelial H 292 cells in vitro produced less IL-6 and IL-8 in response to A. baumannii strains than to A. junii strains. Furthermore, cultured human macrophages produced less TNFa, IL-12p40, IL-8 and IL-10 in response to A. baumannii strains than to A. junii strains. In agreement with our in vitro data, Qiu et al documented that the high susceptibility of A/J mice to A. baumannii ATCC17961 infection was associated with a reduced local proinflammatory response and reduced elimination of bacteria from the lungs [22]. Knapp et al showed in an in vivo model that A. baumannii strain RUH2037 induced the release of pro-inflammatory cytokines and chemokines resulting in clearance of bacteria from the lungs of experimentally infected mice [23]. Although study-design and outcome were distinct from our study, the findings emphasize the importance of inflammatory cytokines for clearing of A. baumannii. It is furthermore of note that our finding that clinically relevant strains induced a weak immune response in vitro has also been reported for Haemophilus influenzae [24].
There was a wide variation in biofilm formation among a large set of well-described A. baumannii strains. Although epidemic strains did not form larger biofilms than sporadic strains, it appeared that strains of European clone II formed larger biofilms than strains of clone I. It is of note that clone II, which was less frequently involved in outbreaks during the 1990s than clone I, is now emerging in several European countries with many strains being carbapenem resistant [25,26]. Further to A. baumannii, the intra-strain variation in biofilm formation of A. gen. sp. 3 and 13TU and of the clinically less-relevant species A. calcoaceticus and A. junii was also considerable. Interestingly, there was no difference between these species, except for A. gen. sp. 13TU that formed small biofilms.
Many of the strains used for the current biofilm experiments had previously been used to investigate adherence to H 292 airway epithelial cells [18]. In that study, it was shown that adherence to airway epithelial cells varied considerably among strains of A. baumannii, while strains of clone II had higher adherence values than those of clone I [18]. Likewise, the present study showed a considerable intra-species variation in adherence to human airway epithelial cells but no difference between A. baumannii and A. junii strains. Biofilm formation on plastic was not correlated to adherence to human cells, indicating that different mechanisms  are involved in these colonization processes. Furthermore, the ability to form biofilm on plastic and the capacity to adhere to human cells was not always accompanied by the presence of piluslike cell surface structures and long cell extensions. Taken together, it seems that biofilm formation and adherence to human cells is strain-and not species-specific. Thus, these features do not solely explain the success of A. baumannii in the susceptible host.
In summary, biofilm formation and adherence to airway epithelial cells did not differ between clinically relevant and lessrelevant Acinetobacter strains and species. However, there was a difference in the production of inflammatory cytokines by airway epithelial cells and macrophages between A. baumannii and A. junii. This may be a first clue to explain the difference in clinical behavior between A. baumannii and A. junii. We hypothesize that A. baumannii may survive and persist in the airways of patients and cause disease at least in part by inducing a weak inflammatory response.

Bacterial strains, culture conditions and antimicrobial susceptibility testing
Forty-five A. baumannii, 3 A. gen. sp. 3, 3 A. gen. sp. 13TU, 3 A. calcoaceticus, and 7 A. junii isolates, were selected from the Leiden University Hospital Acinetobacter collection for this study. Of the A. baumannii strains, 18 were from outbreaks, 16 presumably not from outbreaks on basis of time-space-origin, and 11 of which the association with an outbreak was unknown. Eight of the A. baumannii strains belonged to European clone I, 11 to clone II, and 3 to clone III. All isolates had previously been identified to species by one or more validated genotypic identification methods [27,28,29]. Bacteria were preserved for prolonged periods in nutrient broth supplemented with 20% (vol/vol) glycerol at -80uC. Prior to each experiment, inocula from frozen cultures were grown overnight at 30uC [30] on sheep blood agar plates (BioMerieux, Boxtel, the Netherlands). For experiments, fresh subcultures were made either under these conditions or in Luria-Bertani (LB) medium. Susceptibility to antimicrobial agents was determined by disc diffusion according to CLSI recommendations [31]. Strains resistant to more than two of the following drug classes were defined as multidrug resistant: cephalosporins, carbapenems, ampicillin-sulbactam, quinolones, aminoglycosides.

Biofilm formation
Biofilm formation in 96-wells polyvinylchloride microtiter plates (Falcon, BD, Breda, the Netherlands) was assayed as described [32]. Briefly, bacteria from an overnight culture in LB medium were suspended to 1610 6 CFU/ml as calculated from the absorbance of a suspension at 600 nm. Five ml of this suspension was inoculated in 100 ml of M63 medium consisting of KH 2 PO 4 (12 g/l), K 2 HPO 4 (7 g/l), (NH 4 ) 2 SO 4 (2 g/l), glucose (0.2% w/v), MgSO 4 (1 mM) and casaminoacids (0.5% w/v). After 24 h incubation at 28uC and 37uC, wells were washed and biofilms attached to the wells were stained with crystal violet (1% w/v). The optical density at 590 nm, expressed in arbitrary units (a.u.), was taken as a quantitative measure of biofilm mass. To determine the bacterial concentration after 24 h, serial dilutions of the supernatants were made in PBS and plated onto blood agar.

Adherence to human airway epithelial cells
Adherence of bacteria to human bronchial epithelial H 292 cells (ATCC CRL-1848, Manassas, VA, USA) was determined as described [18]. Briefly, H 292 cells were incubated for 1 h at 37uC with 1610 7 (range 7610 6 -4610 7 ) CFU of an overnight culture on blood agar. Bacterial adherence to H 292 cells was quantified by light microscopy using two parameters: (i) percentage of epithelial cells associated with at least one bacterium; and (ii) average number of bacteria per epithelial cell [18].

Cytokine induction in human airway epithelial cells
Cytokine production by airway epithelial cells in response to bacteria was determined as described previously [33] with minor modifications. In short, bacteria were cultured overnight at 30uC on blood agar and suspended in RPMI-1640 to a concentration of 5610 8 CFU/ml as assessed spectrophotometrically. Approximately 2610 5 H 292 cells were cultured in 0.5 ml of RPMI-1640 supplemented with 2 mM L-glutamine, 50 mg/ml streptomycin, 1000 U sodium penicillin G, and 10% (vol/vol) heat-inactivated foetal calf serum (FCS i ), further referred to as culture medium, in 24-wells plates. At 85-90% confluency, H 292 cells were washed with prewarmed sterile phosphate buffered saline (PBS; pH 7.4) and incubated at 37uC and 5% CO 2 in culture medium without FCSi for 24 h. Subsequently, cells were washed with prewarmed PBS and incubated for 1 h at 37uC and 5% CO 2 with 1610 6 , 1610 7 , or 1610 8 CFU live or heat-inactivated (by 1 h incubation at 100uC) bacteria. H 292 cells were washed five times with prewarmed PBS to remove non-adherent bacteria and fresh RPMI-1640 with 2 mM L-glutamine was added. After additional 5 or 23 h incubation at 37uC, supernatants were collected and stored at 220uC. After 24 h, bacterial CFU count in the supernatants was determined. In each experiment, RPMI-1640 with 2 mM L-glutamine alone was added to the cells to determine background values. A mixture of cytokines (100 ng/ ml TNFa, 20 ng/ml IL-1b and 10 ng/ml IFNc; all from Biosource, Nivelles, Belgium) with 10 ng/ml rough type lipopolysaccharide (Escherichia coli J5, Sigma-Aldrich, Zwijndrecht, the Netherlands), further referred to as cytomix, was added to the cells as a positive control.

Cytokine induction in cultured human macrophages
Buffy coats from healthy human donors were purchased from Sanquin bloodbank, Amsterdam, the Netherlands, upon written consent with regard to scientific use. The current study did not require approval from an ethics committee according to the Dutch Medical Research Involving Human Subjects Act. Monocytes were isolated from buffycoats by Ficoll amidotrizoate density centrifugation and magnetic sorting using anti-CD14-coated beads (Miltenyi Biotec, Auburn, CA, USA) according to the manufacturer's instructions. Monocytes were resuspended in culture medium and cultured at a concentration of 2610 5 cells/ml in wells of a 24-wells plate at 37uC, 5% CO 2 in the presence of either 5 ng/ml recombinant human granulocyte-macrophage colonystimulating factor (Biosource International, Camarillo, CA, USA) to induce differentiation of monocytes into macrophage type 1 or in the presence of 50 ng/ml recombinant human macrophage colony-stimulating factor (R&D Systems, Minneapolis, MN, USA) for type 2 macrophages [20]. After three days, 0.5 ml of fresh culture medium was added. On the sixth day following isolation, cells were washed once in PBS and 0.5 ml RPMI-1640 with 2 mM L-glutamine was added. Bacteria were cultured overnight at 30uC on blood agar and suspended in PBS to a concentration of 5610 7 CFU/ml as assessed spectrophotometrically. Cells were stimulated with 1610 7 CFU of live bacteria for 24 h, after which supernatants were collected and stored at 220uC. In each experiment, PBS alone was added to the cells to determine background. Lipopolysaccharide (100 ng/ml, Escherichia coli J5, Sigma-Aldrich) was added to the cells as a positive control.

Determination of levels of inflammatory mediators
The levels of IL-6, IL-8, TNFa, IL-12p40 and IL-10 in culture supernatants were determined by ELISA (Biosource) according to the manufacturer's instructions. The lower limit of detection was 15 pg/ml for IL-6, 7 pg/ml for IL-8, and 25 pg/ml for TNFa, IL-12p40 and IL-10.

Electron microscopy analysis of bacterial surface structures
For SEM, bacteria from an overnight culture on blood agar at 30uC were suspended in PBS and fixed for 1 h at room temperature with 1.5% (w/v) glutaraldehyde in 0.1 M sodium cacodylate buffer (1:1). Fixed bacteria were transferred to poly-Llysine-coated glass slides. After 1 h incubation at room temperature, bacteria were fixed to the slides with 1.5% glutaraldehyde for 30 min at room temperature. Thereafter, slides were washed twice in PBS, dehydrated through a series of graded ethanol, critical-point-dried, and coated with a layer of palladium-gold.
Bacteria were examined using a JEOL JSM-6700F field emission scanning electron microscope.

Statistical analysis
Results are expressed as medians and ranges. The Kruskal-Wallis one-way analysis of variance and the Wilcoxon rank sum test were used to evaluate differences in distribution. Spearman rank correlation coefficients were calculated to evaluate possible associations between epidemicity and adherence to epithelial cells and biofilm formation on plastic. P values of #0.05 were considered significant.