Direct Evaluation of Pseudomonas aeruginosa Biofilm Mediators in a Chronic Infection Model

P. aeruginosa strains were routinely grown in Luria-Bertani (LB) broth (10 g/liter tryptone, 5 g/liter yeast extract) without NaCl and on 1.5% LB agar plates without NaCl (LANS). For chinchilla experiments, bulla and brain homogenates were plated on 1.5% Pseudomonas isolation agar (Difco). Strains used in this study are shown in Table S1 in the supplemental material and have been described previously. In each experiment comparing PAO1 and mutant P. aeruginosa strains, the parental PAO1 strain from which the mutant(s) was derived was used as the wild-type strain.

Strains were evaluated in vitro by a 30-min polyvinyl chloride (PVC) microtiter dish attachment assay, as described previously (9), and by a procedure modified from the MBEC P&G assay (Innovotech). In the modified MBEC assay, 100 μl of mid-log-phase (optical density at 600 nm, 0.5) cultures was added to wells of a microtiter plate; the lid, containing 96 polystyrene pegs that fit into the wells of the microtiter plate, was placed on the plate; and the plate was incubated at 37°C for 3 h. Pegs were rinsed twice in phosphate-buffered saline (PBS) with agitation to remove loosely attached cells, the lid was placed on another 96-well microtiter plate containing 100 μl of 0.1% crystal violet per well, and the plate was incubated at room temperature for 15 min. Pegs were rinsed three times in PBS with agitation and allowed to air dry, and then the lid was placed on a final microtiter plate containing 200 μl of 95% ethanol per well for 30 min at room temperature. One hundred microliters of solubilized crystal violet from each well was transferred to a flat-bottom microtiter plate, and the absorbance was read at 540 nm.

A chinchilla otitis media model was used to evaluate persistence of P. aeruginosa in vivo. Healthy adult chinchillas (Chinchilla lanigera; body weight, 400 to 500 g) from Rauscher's Chinchilla Ranch (Larue, OH) were allowed to acclimate to the vivarium for 7 to 10 days prior to infection. Animals were assessed and found to be healthy before any studies were undertaken. Chinchillas were anesthetized by isoflurane inhalation, and both left and right bullae were inoculated via transbullar injection of approximately 100 CFU of P. aeruginosa into the middle ear, in accordance with protocols approved by the Animal Care and Use Committee of Wake Forest University Health Sciences and as described previously (3, 24). All inocula were confirmed by plating a single dose (100 CFU in 0.1 ml sterile PBS) on LANS. For survival experiments, at least 4 to 6 animals per strain were used; 3 animals per strain per time point were used in those experiments with defined harvest times. Animals were euthanized either at defined time points (1 day, 2 days, 3 days, or 7 to 8 days) or, for survival experiments, whenever the animals began to show signs of distress requiring euthanasia. In these experiments, distress requiring euthanasia includes severe head tilt, trouble righting, and/or severe lethargy coupled with loss of appetite. Otoscopies were performed regularly to monitor redness, drainage, and tympanic membrane bulging. Otoscopies showing abundant drainage correlated well with distress, typically requiring euthanasia within 12 to 24 h. Left bullae were aseptically removed and homogenized. In some experiments, right bullae were homogenized, while in others, the right bullae were fixed intact or had biofilm material removed and stained. In some experiments brain tissue was aseptically removed prior to opening of the bullae and homogenized or fixed in 2% paraformaldehyde for histology. Bullae and brain were homogenized in 10 ml sterile PBS for CFU determination by plate count (power setting 4 for bullae, power setting 3 for brain; Power Gen 1000; Fisher Scientific). Effusion was removed from the bullae and combined with a 1-ml PBS wash for determination of the numbers of CFU of bacteria not associated with biofilm material.

For scanning electron microscopy (SEM), bullae were fixed whole in 2.5% glutaraldehyde-PBS, trimmed of excess bone, and prepared as described previously (42). Samples were visualized with a Philips SEM-515 scanning electron microscope. For histopathology, bullae were fixed intact in 4% formalin for at least 24 h and then decalcified for 6 h using a Decalcifier II reagent (Surgipath Medical Industries). The tissues were sagitally sectioned, embedded in paraffin, processed routinely for histology, cut at 6 μm, and stained with hematoxylin-eosin (H&E) according to standard techniques (3). Slides were examined by routine light microscopy using a conventional Nikon Eclipse 50i microscope in a blinded fashion by a board-certified veterinary pathologist.

Material removed from bullae was fixed in 2% paraformaldehyde overnight for histopathology or immunohistochemistry staining or was immediately live/dead stained. Viability of unfixed material was evaluated using BacLight LIVE/DEAD staining (Molecular Probes). Material was stained with SYTO 9 and propidium iodide (PI) in PBS for 15 min and then washed in PBS (42). Material embedded in OCT (Sakura Finetek) at −20°C was cut into 5-μm sections and stained with rabbit antipseudomonal or rabbit anti-Psl antiserum (each 1:5,000) (9) with a fluorescein isothiocyanate (FITC)-conjugated donkey anti-rabbit IgG secondary antibody (1:50). Some sections were stained with H&E and evaluated by light microscopy as described above. In sections stained with anti-Psl, PI was also used to stain DNA. Fluorescence was visualized using a Zeiss LSM510 confocal microscope.

Animals were monitored for redness and/or effusion in the external ear using a Digital VetScope system with VetDock (version 2.1) software (MedRx). Each ear was evaluated over time and scored positively for having redness or drainage at any time during the experiment.

Statistical analyses were performed using GraphPad Prism (version 5) software. Survival curves were compared by the log-rank (Mantel-Cox) test, and video otoscopy scores were evaluated by Fisher's exact test. Images were processed using Adobe Systems Photoshop CS (version 8.0) and assembled using Photoshop, Prism, and/or ImageJ (version 1.42q) (National Institutes of Health) software.

Initial experiments evaluated the ability of the common laboratory strain PAO1 to cause otitis media in chinchillas and established the optimal inoculum for persistence. To evaluate gross infection, the thin layer of bone comprising the superior bullous surface was removed and the bulla was imaged from above. Thick, opaque exudate is visible (arrows) in the bullae of chinchillas infected by transbullar inoculation (3, 24) with 20 or 200 CFU PAO1 (Fig. 1A to F) at both 3 days postinfection (d.p.i.) (Fig. 1A to D) and 8 d.p.i. (Fig. 1E and F). This material is similar to what is observed in the bullae of chinchillas infected with nontypeable Haemophilus influenzae (NTHi) and Streptococcus pneumoniae, the two most common pathogens causing otitis media in humans (24, 41, 42). Clear effusions (Fig. 1A and B) are visible in a majority of infected animals, and in fewer cases, erythema is visible (Fig. 1C and D). In contrast, middle ears infected with 0.1 ml sterile PBS remained clear of obstructing material (Fig. 1G and H). Infection with 20 CFU increased persistence slightly over that seen with 200 CFU (i.e., the onset of morbidity in animals that did not clear the infection was delayed compared to the time to onset of morbidity in animals receiving the higher dose), but at 8 d.p.i. with 2/4 animals surviving, 1 animal did not have detectable bacteria. Further experiments used an inoculum of 100 CFU, as this reliably results in establishment of infection while reducing the morbidity seen with higher inocula. In this and subsequent experiments, the bacterial load commonly exceeded 10⁹ CFU per bulla at as early as 2 d.p.i., confirming that P. aeruginosa readily exploits the middle ear environment and demonstrating the utility of this model in evaluating the establishment of infection.

In this model, less than 20% of PAO1-infected animals routinely survive to 10 days. Survival is defined as the time to morbidity requiring euthanasia. To investigate the cause of morbidity, three chinchillas were infected with PAO1, and plate counts of blood, bullae, and brain were performed at 3 to 4 d.p.i. (see Table S2 in the supplemental material). No animals had detectable bacteria in the blood in this or subsequent experiments in which blood samples were evaluated. In the two animals with detectable bacteria in the bullae, there was approximately 10-fold less bacteria in the left hemisphere of the brain, which was aseptically removed prior to opening of the bullae. In the one animal that did not have bacteria in the left bulla, there were likewise no bacteria detected in the brain. This suggested that although none of the animals were moribund, P. aeruginosa in the brain could account for the eventual onset of otitis media signs and rapid progression of morbidity often observed in these experiments.

To determine the composition of the middle ear material, whole bullae were excised, fixed, and trimmed to expose the inner surface and then prepared for SEM (9). At 4 d.p.i. with PAO1, the middle ear epithelium clearly shows the presence of both P. aeruginosa and host cells, the latter having morphology consistent with polymorphonuclear cells (PMNs) (Fig. 2A, top left). At higher magnification, individual P. aeruginosa cells can be seen enmeshed in a matrix (Fig. 2A, top right and bottom), recalling in vitro biofilms, in which, at 48 h under static conditions, PAO1 cells are partially obscured by a matrix of bacterial origin (see Fig. S1A in the supplemental material). In PBS-infected animals, no bacterial or immune cells are visible on an intact epithelial surface (see Fig. S1B and C in the supplemental material). Other bullae were sectioned and stained to observe the material in situ. At 3 d.p.i. with PAO1, there was evidence of PMN infiltration into the middle ear, with a dense network of fibrin surrounding the immune cells (Fig. 2B). The presence of PMNs associated with a fibrous mesh is similar to that seen in NTHi and S. pneumoniae otitis media (3, 42). The fibrin forms layers within the bulla, reflecting waves of infiltration likely attempting to isolate the infection. Material removed from the bulla showed aggregates of bacteria, some associated with a weblike structure that stained positively for nucleic acid (Fig. 2C). Immediately following removal from the bulla, some material was stained using the live/dead reagents. Confocal images showed living and dead cells, identified both as host and as bacterial cells by size and morphology (Fig. 2D). Staining of cryosections with antipseudomonal antiserum revealed both single cells and dense clusters of bacteria distributed throughout an immunoreactive matrix (Fig. 2E) (9). A Psl-specific antiserum, which stains both cell- and matrix-associated biofilm polysaccharide (9), showed Psl colocalization with cells stained with the nonviable indicator PI, indicating that Psl remains associated with dead cells or cells with compromised membranes (Fig. 2F and G). Diffuse Psl and DNA staining between cells agrees with the findings of in vitro analyses of P. aeruginosa biofilms, although the DNA could be of either bacterial or host origin (Fig. 2F) (1, 33, 52). At high magnification, Psl was detected surrounding individual cells, supporting a model of surface-associated Psl mediating cell-cell and cell-surface interactions during biofilm formation (Fig. 2G) (33). From this evidence, we concluded that the bulla material contains living and dead host and bacterial cells in a matrix of host-derived fibrin, P. aeruginosa-specific material, including Psl, and extracellular DNA of undetermined origin.

The second messenger c-di-GMP positively regulates biofilm formation in P. aeruginosa (6, 22, 27, 30). The rugose small colony variant (RSCV) MJK8 is derived from PAO1 but contains a mutation in the wspA gene resulting in high levels of intracellular c-di-GMP (51). Due to overproduction of c-di-GMP, MJK8 forms biofilms with significantly greater biomass than PAO1 biofilms (Table 1). We thus hypothesized that MJK8 would persist longer than its parental PAO1 strain in the chinchilla due to its ability to form biofilms that would potentially limit dissemination and immune system involvement. Chinchillas were inoculated with PAO1 or MJK8 and monitored for signs of otitis media for 9 days. Animals were euthanized when signs became severe, and bullae were removed, homogenized, and plated to determine bacterial load. Chinchillas infected with MJK8 survived significantly longer than those infected with PAO1 (Fig. 3; P = 0.013). Six out of 11 animals infected with MJK8 survived until the end of the experiment without becoming moribund, and of these, 2 had detectable P. aeruginosa, while 4 did not harbor detectable bacteria. In contrast, one-half of the chinchillas infected with PAO1 had to be euthanized by 4 d.p.i., and only 2 out of 10, with neither having detectable bacteria, survived through 9 days. All MJK8-infected animals displaying signs of otitis media at 7 to 9 d.p.i. (5/11) had detectable P. aeruginosa. The presence of MJK8 at later time points suggests that this strain is not simply cleared early in infection but, rather, persists with delayed morbidity compared to PAO1. The bacterial load at the time of killing was not different between the groups (approximately 5 × 10⁸ average CFU per bulla for both groups) and cannot account for the difference in survival (data not shown). Bulla material from MJK8-infected animals at 8 d.p.i. does not appear to be altered or reduced in size compared to material from PAO1-infected animals (see Fig. S2A and B in the supplemental material). SEM analysis of an MJK8-infected bulla at 8 d.p.i. shows both bacterial and host components at the epithelial surface, similar to those seen in PAO1 in Fig. 2A (see Fig. S2C and D in the supplemental material). These data support the argument that differences in MJK8 as a result of c-di-GMP overproduction are sufficient to explain the survival data, the mechanism of which is currently under investigation.

Expression of P. aeruginosa biofilm polysaccharides Psl and Pel (17, 25, 33, 35) is elevated in MJK8 due to increased intracellular c-di-GMP (51). Biofilms produced in vitro by MJK8 psl and pel mutants contain significantly less biomass than MJK8 (Table 1). To test whether loss of individual polysaccharides and, thus, loss of biofilm-forming ability affects persistence of MJK8, chinchillas were infected with MJK8, MJK8 Δpsl, MJK8 Δpel, or PAO1. While the median survival of animals infected with MJK8 Δpsl was 4 d.p.i., compared to 6 d.p.i. for MJK8-infected animals, the survival curves were not significantly different (P = 0.234; data not shown). The progression of infection with MJK8 Δpel was identical to that of MJK8, suggesting that Pel does not contribute to the persistence of MJK8 in this model. All animals infected with PAO1, however, had severe otitis media signs and had to be euthanized by 5 d.p.i., while half of the MJK8-infected animals survived to the end of the experiment (data not shown). Since Psl and Pel are associated with biofilm formation, we sought to determine if the bulla material was influenced by loss of either polysaccharide. Macroscopically visible material was present in at least half of the bullae in each group: PAO1, 2/3; MJK8, 2/4; MJK8 Δpsl, 4/4; MJK8 Δpel, 3/4. The composition of the material was similar to that seen with PAO1 (Fig. 2B) and was not different among the groups by H&E staining (Fig. 4A to F); however, these samples were not evaluated with Psl-specific antiserum. These data suggest that biofilm phenotypes displayed in vitro are not necessarily recapitulated in vivo but may be overshadowed by an immune response that includes host cells, fibrin, and other host-derived factors. As in the previous experiment, the bacterial loads at the time of killing were not significantly different among the groups (data not shown).

Since Psl has previously been shown to be important in PAO1 biofilm architecture (9, 17, 25, 30, 33–35), we sought to determine if the loss of Psl results in differences in bacterial growth rate or spatial distribution within bullae early in infection. To address this, chinchillas were infected with MJK8 or MJK8 Δpsl, and the bacterial load was determined in the surface-attached, or biofilm, material, in middle ear effusion, and in the brain at 2, 3, and 7 d.p.i. (Fig. 4G). A distinction between effusion and the numbers of biofilm CFU was made because of the possible role of Psl in maintaining the association of bacteria with biofilm components present within bullae. There was no significant difference in bacterial load at any time point tested in both biofilm and effusion samples, suggesting that Psl does not affect the growth rate, nor does it influence the distribution of P. aeruginosa within the middle ear space (Fig. 4G). Bacterial load in the brain was variable at each time point and was not affected by expression of Psl (data not shown). The reduced size of the MJK8 Δpsl group at 7 d.p.i. was due to two animals being euthanized at 5 d.p.i., with 4/4 bullae containing copious exudate and effusion. Signaling by c-di-GMP regulates an array of biofilm-associated responses beyond polysaccharide expression (6, 22, 27, 30); therefore, it is likely that future studies addressing these c-di-GMP-regulated components will elucidate the mechanism of persistence of MJK8 and similar clinical isolates (30, 51).

Flagella are required for swimming and swarming motility and reversible surface attachment in vitro (38, 44), as well as proinflammatory signaling via interaction with the innate immune receptor TLR5 (21). WFPA850, a fliC deletion mutant, adheres to human A549 airway epithelial cells at less than 35% of the rate of PAO1 (8) and displays a severe defect in biofilm initiation on PVC surfaces (Table 1). Interestingly, in the 3-h peg assay on a polystyrene surface, the fliC mutant displayed a less severe biofilm defect than on PVC, suggesting that the biofilm phenotype may be confined to biofilm initiation or to certain surfaces (Table 1). Flagellar mutants of P. aeruginosa have attenuated virulence in acute mouse burn and pneumonia models (4, 14, 16, 36). We hypothesized that deletion of fliC would result in increased survival of infected chinchillas with a decreased bacterial burden compared to that of PAO1, due to the reduced motility and induction of inflammation characteristic of this strain outweighing the potential loss of flagellum-dependent biofilm formation. Six chinchillas per group were infected with PAO1 or with the isogenic fliC mutant WFPA850 and monitored for signs of otitis media over 10 days. Surprisingly, there was no difference in survival between chinchillas infected with PAO1 and those that received WFPA850 (Fig. 5A; P = 0.218). At 10 d.p.i., 2/6 PAO1-infected animals had survived and did not have detectable bacteria in the bullae or brain, whereas 0/6 WFPA850-infected animals survived, and all had detectable bacteria at the time of killing both in the bullae and in the brain; however, these counts were not significantly different from those of PAO1 (Fig. 5A and B). Bacteria in the brain are likely not contamination because, although not all brains were examined histologically, encephalitis was diagnosed in one WFPA850-infected animal, making it reasonable to propose that clinical signs occurred as a result of infection spreading from the middle ear to the brain. The presence of bacteria in the brain, especially in animals infected with the motility-impaired WFPA850, indicates that P. aeruginosa can escape the site of infection without flagella, but the mechanism for this is not understood. Video otoscopy of the auditory canal revealed both redness (PAO1, 4/12 ears; WFPA850, 3/12) and drainage (PAO1, 4/12; WFPA850, 6/12) during the course of infection for both groups. These results suggest that, unlike in mouse models of acute infection, in the chinchilla otitis media model, deletion of fliC is not sufficient to attenuate PAO1. The absence of flagella may affect immune recognition by PMNs and other cell types expressing TLR5. It is also possible that loss of flagellum-mediated attachment to host surfaces contributed to dispersal to the brain, in effect compensating for the motility defect that we had predicted would limit dispersal.

The LasRI/RhlRI quorum-sensing systems in P. aeruginosa regulate expression of numerous virulence factors, including elastase, rhamnolipid, pyocyanin, exotoxin A, and other proteases (7, 50). Production of biofilm matrix components, such as extracellular DNA, Pel, and Psl, is influenced by quorum sensing, yet the requirement for quorum sensing in biofilm formation in vitro is still debated (1, 18, 43, 46). Nutritional factors mitigate the impact of quorum-sensing-regulated swarming motility, which directly affects biofilm structure (46). Biofilms formed in vitro by lasR (PDO-R1) or rhlR (PDO111) mutants or a lasR rhlR double mutant (PAO-JP3) contained a similar biomass as PAO1, although there was a trend toward decreased biomass in double-mutant biofilms relative to that in PAO1 (Table 1). The role of quorum sensing in P. aeruginosa pathogenesis has been well established in both rat and mouse models of acute and chronic infection but has not been evaluated in the chinchilla otitis media model (10, 32, 37, 50). We hypothesized that the loss of quorum-sensing-controlled virulence factors would outweigh any biofilm defects and lead to a more attenuated infection with increased survival. Chinchillas were infected with PAO1, PDO-R1 (ΔlasR), or PDO111 (ΔrhlR) and monitored for signs of otitis media for 10 days. There was no significant difference in survival or bacterial load in bullae and brain for either mutant-infected group compared to the PAO1-infected group (Fig. 6A). One of two brains examined histologically from both the PAO1 and rhlR groups was diagnosed as having meningitis, providing additional evidence for bacteria migrating from the ear to the brain, causing some of the clinical signs observed in this experiment. Video otoscopy revealed that the majority of ears in all three groups had drainage, correlating with the rapid onset of morbidity (Fig. 6C).

We next determined if deletion of both lasR and rhlR would affect survival, since the loss of virulence factors would be greater than that for lasR or rhlR single mutants. Chinchillas were infected with PAO1 or PAO-JP3, an isogenic lasR rhlR double mutant that cannot respond to either the 3OC12-HSL or C4-HSL autoinducer, and monitored for 10 days. Survival was significantly greater for animals infected with PAO-JP3 than PAO1-infected animals (P = 0.001), PAO-R1-infected animals (P = 0.003), and PDO111-infected animals (P = 0.009; Fig. 6A). Animals infected with PAO1 displayed otitis signs earlier than the PAO-JP3 group and had a median survival of 2.5 days, compared to a median survival of 5 days for PAO-JP3-infected animals. Video otoscopy correlated with the survival data in both of these experiments, with what appeared to be higher proportions of PAO1-infected ears having drainage (15/22) and redness (4/22) compared to PAO-JP3-infected ears (4/12 and 5/12, respectively); however, only the incidence of drainage approached significance (P = 0.075). Animals infected with PAO-R1 had a significantly greater incidence of drainage (11/11) than those infected with PAO-JP3 (4/12, P = 0.001), with a concomitantly lesser incidence of redness (P = 0.037), correlating well with the greater survival of the PAO-JP3 group (Fig. 6C). There were no significant differences in the incidences of redness or drainage between the PDO111- and PAO-JP3-infected groups. As in other experiments, deletion of lasR and rhlR did not affect the bacterial load in the bullae at the time of killing (Fig. 6B). The bacterial load in the brain was likewise not different between the two groups, indicating that factors not requiring LasR or RhlR may be responsible for the spread. The one surviving PAO-JP3-infected animal did not have recoverable bacteria in the left bulla or brain but did have a small amount of material present in the right bulla at 10 d.p.i. The presence of material coupled with redness visible by otoscopy at 4 d.p.i. suggests that an infection was established in the right bulla but either had not been established in the left bulla or had been cleared. We concluded from these experiments that the ability to respond to one or both AHLs, which leads to production of regulator-specific and redundantly regulated virulence factors, results in infection similar to that with PAO1 in this model but that loss of both LasR and RhlR results in increased persistence.