Mouse BMMϕ were generated by harvesting bone marrow cells from murine tibias and femurs as described earlier (Godek et al., 2006; Weischenfeldt and Porse, 2008). Briefly, bone marrow cells were flushed from mouse bones (C57 BL6), and then differentiated into a macrophage phenotype by incubating in complete DMEM (cDMEM, 10% heat inactivated fetal bovine serum (FBS, Hyclone^®, Logan, UT, USA), 30% L-929 fibroblast conditioned medium, 1% penicillin–streptomycin (Gibco, Carlsbad, CA, USA), 0.01 M Hepes buffer, 1 mM sodium pyruvate, and 1% of a × 100 MEM non-essential amino acids solution (all from Sigma, St. Louis, MO, USA) in Dulbecco’s modified Eagle’s medium (DMEM, Mediatech, Herndon, VA, USA). Cells were differentiated for 7 days with media changes every 2 days. This study was carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (NIH). The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Miami.

Human MDMs were generated by separating monocytes from peripheral blood obtained from healthy blood donors, as previously described (Welin et al., 2008; Eklund et al., 2010; Meng et al., 2015; Shinzaki et al., 2016). Donors were between age groups of 25–50 years, of either sex and were free from blood-borne diseases. In brief, monocytes were obtained by layering blood onto a density gradient, followed by centrifugation and isolation of the monocyte/lymphocyte fraction. Cells were seeded in culture flasks and allowed to adhere for 2 h before non-adherent lymphocytes were washed away. The monocytes were allowed to differentiate into human MDMs for 5–8 days in DMEM (Gibco, Carlsbad, CA, USA) containing 25 mM Hepes, 100 U/ml penicillin, 100 μg/ml streptomycin and 10% active human serum (all Sigma, St. Louis, MO, USA). All cell cultures were incubated under 37°C, 5% CO₂, 98% humidity. Human whole blood was obtained from commercial vendors including Interstate blood bank and followed NIH guidelines for the protection of human subjects including informed signed consent from all subjects by these vendors. All the information regarding human subjects was deidentified and investigators were having no access to any human subject information. The institutional review board (IRB) of the University of Miami approved the study protocol.

The clinical strains of P. aeruginosa (10 strains) isolated from CSOM patients attending University of Miami Hospital or Hospital for Rehabilitation of Craniofacial Anomalies of the University of São Paulo, Bauru, Brazil were used in this study. The patients exhibited persistent chronic purulent otorrhoea through perforated tympanic membrane for more than 6 weeks confirmed by otolaryngologic diagnosis. The isolation and identification of P. aeruginosa was performed using standard methods (MacFaddin, 1976; Forbes et al., 1998). The oprF mutant and complemented strain of P. aeruginosa was generated as described in previous studies (Woodruff and Hancock, 1989; Horton et al., 1990; Schweizer and Hoang, 1995; Rietsch et al., 2005; Balasubramanian et al., 2012; Yakhnina et al., 2015) (Supplementary Materials). Bacteria were grown overnight at 37°C in Luria broth (LB) (Teknova, Hollister, CA, USA) in a rotary shaker.

Gentamicin protection assays were used to quantify the extent of human MDM and mouse BMMϕ invasion and survival by otopathogenic P. aeruginosa. Briefly, cells were infected with bacteria at 1:1, 5:1, 10:1, 25:1, 50:1, 100:1 multiplicity of infection (MOI) (bacteria:cell) for 1, 2, 4, and 6 h. After incubation, the cells were washed five times with warm RPMI medium (Corning, NY, USA) followed by addition of medium containing gentamicin (200 μg/ml) (Life Technologies, Carlsbad, CA, USA) and further incubated for 1 h at 37°C. All P. aeruginosa isolates we used in this study were sensitive to gentamicin at this concentration. The cells were washed three times with RPMI and then lysed with 1% saponin (Sigma, St. Louis, MO, USA) to release intracellular bacteria. Serial dilutions were then plated on blood agar plates and bacterial colonies were counted the next day. In some experiments, bacteria were pretreated with 20% human or mouse pooled serum or heat inactivated pooled serum and then used to infect human MDMs and mouse BMMϕ, respectively. In some experiments, bacteria were pretreated with monoclonal anti-OprF antibody (kindly provided by Dr. Hancock) and then used in the invasion assay. The monoclonal antibody (mAb) was specific to surface epitopes of OprF and was generated as described previously (Finnen et al., 1992; Rawling et al., 1995). To determine the effect of cytoskeletal inhibitors, macrophages were pretreated with different concentrations of cytochalasin D, vinblastine, nocodazole, or colchicine (all from Sigma, St. Louis, MO, USA) for 30 min before infecting with bacteria, and maintained in the medium for the entire infection period.

Human MDMs and mouse BMMϕ were cultured on glass cover slips and were infected with bacteria for 30 min to 8 h. After incubation, the cells were washed five times with warm phosphate buffer saline (PBS) buffer (Sigma, St. Louis, MO, USA) to remove unbound bacteria and were then processed for scanning electron microscopy (SEM). Samples were fixed in 2% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in PBS buffer followed by three changes of PBS buffer for 10 min each. The samples were then post-fixed in 1% osmium tetroxide (Electron Microscopy Sciences, Hatfield, PA, USA) in PBS buffer for 45 min and rinsed in three changes of PBS buffer for 10 min each. The samples were dehydrated in a graded series of ethanol, dried in hexamethyldisilazane (HMDS) (Electron Microscopy Sciences, Hatfield, PA, USA) and mounted on carbon adhesive tabs fixed to metal stubs. The samples were coated with palladium in a plasma sputter coater and viewed in a SEM (FEI, ESEM-FEG XL-30).

Human MDMs and mouse BMMϕ were infected with bacteria for 30 min to 8 h. After incubation, the cells were washed with PBS and fixed using 2% glutaraldehyde (Electron Microscopy Sciences, Hatfield, PA, USA). The samples were rinsed in three washes of PBS buffer then post-fixed in 1% osmium tetroxide (Electron Microscopy Sciences, Hatfield, PA, USA) in 0.1 M phosphate buffer for 1 h. After buffer rinses, specimens were dehydrated through a series of graded ethanol, placed in two rinses of propylene oxide (Electron Microscopy Sciences, Hatfield, PA, USA) for 5 min each and then put in a 1:1 mixture of propylene oxide: EMbed/Araldite resin (Electron Microscopy Sciences, Hatfield, PA, USA) for overnight incubation at room temperature. Next day, the pellets were placed in fresh EMbed/Araldite and put in a vacuum desiccator for 2–4 h. The samples were changed to fresh EMbed/Araldite and polymerized overnight. Silver/gold sections were then cut on a Leica Ultracut E (Leica, Buffalo Grove, IL, USA), stained in uranyl acetate (Electron Microscopy Sciences, Hatfield, PA, USA) and lead citrate (Electron Microscopy Sciences, Hatfield, PA, USA), and viewed in a JEOL 1400 electron microscope (JEOL, Peabody, MA, USA) with Gatan Orius SC1000 camera (Gatan, Pleasanton, CA, USA).

For staining of bacteria and actin, human MDMs and mouse BMMϕ were cultured in 8-well chamber slides and infected with P. aeruginosa for 30 min to 8 h. After incubation, cells were washed three times with PBS buffer and then fixed and permeabilized with BD cytofix and cytoperm reagent (BD Biosciences, San Jose, CA, USA) for 30 min. After washing, the cells were blocked with 3% normal goat serum (NGS) (Life Technologies, Carlsbad, CA, USA) for 20 min and then incubated with anti-P. aeruginosa antibody (1:200) (Abcam, Cambridge, MA, USA) for 45 min followed by Alexa Fluor 488 antibody (1:1000) (Life Technologies, Carlsbad, CA, USA). After washing, cells were counterstained for actin with rhodamine phalloidin (Life Technologies, Carlsbad, CA, USA) for 45 min, washed and mounted in an antifade Vectashield solution containing 4, 6-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA, USA). The cells were viewed with a Zeiss LSM 710 microscope (Carl Zeiss, Germany) and images were assembled using Adobe photoshop 7.0.

To evaluate the cytotoxic effects of P. aeruginosa on human MDMs and mouse BMMϕ during the course of infection, cells were seeded in 96 well plates and infected with bacteria for 2–24 h at an MOI of 10. Lactate dehydrogenase (LDH) levels were then determined in cell culture supernatants using commercially available kit, as per the manufacturer’s instructions (Cayman Chemical, Ann Arbor, MI, USA). Uninfected cells were included as negative control. Maximum LDH release induced by treatment of macrophages with 1% Triton X-100 (Sigma, St. Louis, MO, USA) was used as a positive control. Results were expressed as percentage LDH release compared to the positive control.

The viability of P. aeruginosa infected human MDMs and mouse BMMϕ was also determined by ethidium homodimer-1 staining (Eidet et al., 2015; McCanna et al., 2015; Yoeruek et al., 2016). Ethidium homodimer-1 enters inside dead cells through the disrupted plasma membrane and binds to nucleic acids that increase its fluorescence intensity leading to the production of red fluorescence. Live cells exclude ethidium homodimer-I staining due to the intact plasma membrane. Human MDMs and mouse BMMϕ were infected with bacteria for 2–24 h time periods at an MOI of 10 and then stained with ethidium homodimer-1. Fluorescence was then determined using a microplate reader with appropriate filters (excitation at 535 nm and emission at 620 nm). The percentage dead cells were then calculated using the following formula:

where F(620 nm) is fluorescence intensity at a wavelength of 620 nm, F(620 nm)_sample is fluorescence intensity of human MDMs and mouse BMMϕ infected with P. aeruginosa, F(620 nm)_minimum is fluorescence intensity of uninfected human MDMs and mouse BMMϕ, and F(620 nm)_maximum is fluorescence intensity of human MDMs and mouse BMMϕ treated with 1% Triton X-100 (Sigma, St. Louis, MO, USA).

Statistical significance was determined by a paired, two-tailed Student’s t-test and ANOVA using SPSS software. Values of P < 0.05 were considered to be statistically significant.

Since P. aeruginosa is the major etiological agent of CSOM, we hypothesize that this pathogen hijacks host innate immunity that forms the first line of defense. To test this hypothesis, we examined the ability of a clinical isolate of P. aeruginosa from CSOM patient to enter and survive inside primary human and mouse macrophages by gentamicin protection assay at different MOIs for 2 h. We observed that P. aeruginosa is able to enter and survive inside macrophages. At an MOI of 1, log 1.45 colony forming unit (cfu) bacteria were recoverable from human MDMs whereas log 1.12 cfu bacteria were demonstrable inside mouse BMMϕ (Figure 1a). The number of bacteria inside human MDMs increased from 2.23 to 4.11 log cfu with a corresponding increase in MOI from 5 to 10 (Figure 1a). Similar increase in number of bacteria was observable inside mouse BMMϕ. The number of bacteria increased from log 2.11 log cfu at an MOI of 5 to log 3.48 cfu at an MOI of 10 (Supplementary Figure S1a). Further increase in MOI to 25, 50, and 100 caused only marginal increase in number of bacteria inside human MDMs and mouse BMMϕ (P > 0.05). Therefore, we selected a MOI of 10 and post-infection time period of 2 h for further experiments.

In order to determine the effect of increasing infection time on the cell invasion, macrophages and P. aeruginosa were incubated for several time-points. By 1 h post-infection time period, 1.04 log cfu bacteria were demonstrable inside human MDMs that increased to 2.96 log cfu bacteria by 8 h post-infection at an MOI of 1 (Figure 1b). At an MOI of 5 and 10, there were 1.56 and 2.12 log cfu bacteria at 1 h post-infection that increased to 3.39 and 6.98 at 8 h post-infection time period respectively. At a high MOI of 100, 7.55 log cfu bacteria were recoverable from human MDMs at 8 h post-infection time period. Mouse BMMϕ also demonstrated increase in bacterial load with the corresponding increase in post-infection time period from 1 to 8 h at all MOIs (Supplementary Figure S1b). Similar results of human MDM and mouse BMMϕ invasion was observed with nine additional P. aeruginosa CSOM clinical isolates at an MOI of 10 (Supplementary Figures S2 and S3). There was increase in bacterial load in both human MDMs and mouse BMMϕ with increase in post-infection time period from 2 to 6 h. Taken together, these results suggest that otopathogenic P. aeruginosa invades human MDMs and mouse BMMϕ in a time and dose dependent manner.

Complement proteins are constitutively present in the serum and can opsonize bacteria non- specifically promoting pathogen killing and clearance of infection (Daha, 2010; Merle et al., 2015; Varela and Tomlinson, 2015). Therefore, we evaluated whether complement affects the phagocytosis of P. aeruginosa by macrophages. To determine the effect of opsonins on the survival of P. aeruginosa, two CSOM bacterial isolates were opsonized with fresh normal or heat-inactivated human or mouse pooled serum or were treated with medium alone prior to infection of human MDMs and mouse BMMϕ. Interestingly, there was no statistical difference in the number of non-opsonized and opsonized bacteria in human MDMs and mouse BMMϕ (P > 0.05; Figures 2a,b). The heat-inactivated serum also had no significant effect on the ability of otopathogenic P. aeruginosa to invade macrophages (P > 0.05). These results suggest that complement did not play a significant role in invasion of human MDMs and mouse BMMϕ by P. aeruginosa.

To examine the interaction of P. aeruginosa with macrophages in detail, human MDMs and mouse BMMϕ were subjected to SEM. During the first 15 min of post-infection, P. aeruginosa was observed to adhere to human MDMs through pseudopod like structures (Figure 3a). There was an increase in the formation of these pseudopod like structures at 30 min post-infection (Figure 3b). Bacteria were seen to be adhered to the human MDMs mostly in clusters. However, individually attached bacteria were also observed. By 1 h post-infection, few bacteria were demonstrable on the surface of human MDMs accompanied with membrane ruffling (Figure 3c). The number of bacteria on the surface of human MDMs increased with increase in post-infection time period (Figures 3d,e). A large number of bacteria were observed on the surface of macrophages by 8 h post-infection (Figure 3f). Similarly few bacteria were demonstrable on the surface of mouse BMMϕ at 60 min post-infection that increased in number with increase in post-infection time period to 8 h. As observed with human MDMs, we also observed pseudopod formation and membrane ruffling in P. aeruginosa infected mouse BMMϕ (Supplementary Figure S4).

Transmission electron microscopy (TEM) demonstrated that human MDMs form plasma membrane protrusions in response to P. aeruginosa infection that can act as bacterial adhesion sites (Figure 4a). The bacteria were found in the enclosed membrane protrusions by 30 min post-infection in human MDMs (Figure 4b). At 60 min post-infection, P. aeruginosa was internalized into human MDMs inside membrane bound vacuoles (Figure 4c). In some vacuoles, the membrane was closely apposed to the bacteria whereas in others there was a space between bacteria and the membrane. In addition, actively dividing bacteria inside human MDMs were also observed (Figure 4d). There was increase in bacterial number inside human MDMs at 2 h post-infection (Figure 4e). At 4 h post-infection, the vacuoles containing bacteria increased in size (Figure 4f). At 8 h post-infection, there was clustering of bacteria inside human MDMs. In some vacuoles, multiple bacteria were demonstrable inside them (Figure 4g). At this post-infection time-period, P. aeruginosa disrupted the membrane and many free bacteria were observed in the cytoplasm of human MDMs (Figure 4h). Mouse BMMϕ displayed similar phenotypic changes in response to otopathogenic P. aeruginosa infection (Supplementary Figure S5).

Host cytoskeleton can play a crucial role in bacterial cell invasion. Therefore, we determined whether invasion of macrophages by P. aeruginosa relies on microfilament and microtubule dependent pathways. Human MDMs and mouse BMMϕ were infected with P. aeruginosa in the presence of increasing concentrations of cytochalasin D, an inhibitor of actin polymerization. There was a significant dose-dependent decrease in the invasion of human MDMs and mouse BMMϕ in the presence of cytochalsin D compared to DMSO treated or untreated cells (P < 0.01) (Figure 5a). A 30% decrease in invasion in the presence of 2 μM cytochalasin D was observed whereas 90% decrease in invasion was observable at a concentration of 20 μM in human MDMs (Figure 5a). Microtubule disrupting compounds, vinblastine, colchicine, and nocodazole also caused a dose-dependent decrease in the invasion of human MDMs by P. aeruginosa (P < 0.01). Human MDMs pretreated with 10 μM vinblastine showed 40% decrease in invasion whereas cells pretreated with 50 μM demonstrated more than 90% decrease in invasion compared to DMSO treated or untreated macrophages (Figure 5b). Colchicine caused an 80% decreased in invasion at a concentration of 25 μM, whereas a 95% decrease in invasion was observed in the presence of 30 μM nocodazole (Figures 5c,d). Similar significant decrease in invasion of mouse BMMϕ was observed in the presence of cytochalsin D, vinblastine, colchicine, and nocodazole (P < 0.01) (Figures 5a–d). We observed that there were no toxic effects of these reagents on bacteria or on cells at the tested concentrations (data not shown). These results with inhibitory compounds suggest that P. aeruginosa invades human MDMs and mouse BMMϕ through both microfilament and microtubule dependent uptake mechanisms.

Actin cytoskeleton rearrangement is a common strategy employed by pathogens to invade host cells (Yang et al., 2012; Navarro-Garcia et al., 2013; de Souza Santos and Orth, 2015; Zheng et al., 2015). To determine whether P. aeruginosa induces alterations to the cytoskeleton during invasion of human MDMs and mouse BMMϕ, the distribution of F-actin in macrophages was examined. Human MDMs and mouse BMMϕ were infected with P. aeruginosa and stained with rhodamine phalloidin to detect F-actin. Uninfected human MDMs showed spatial distribution of F-actin throughout the cell (Figures 6a–d). However, infected human MDMs showed actin condensation in response to otopathogenic P. aeruginosa infection (Figure 6). At 30-min post-infection, a lot of actin accumulation underneath the bacterial binding sites was observed (Figures 6e–h). There was a lot of interaction of bacteria with the actin filaments. At 60-min, there was further increase in actin accumulation that colocalized with the bacteria as indicated by the yellow color (Figures 6i–l). Similar pattern of actin accumulation was observed in mouse BMMϕ infected with otopathogenic P. aeruginosa (data not shown). Taken together, these findings suggest that P. aeruginosa induces actin cytoskeleton rearrangements that facilitate its entry inside macrophages.

Bacterial OMPs play a crucial role in interaction of pathogens with host cells (Toma et al., 2014; Alzahrani et al., 2015; Bulir et al., 2015). OprF is the major OMP/porin of P. aeruginosa that has been demonstrated to facilitate biofilm formation under anaerobic conditions and adhesion to host cells (Azghani et al., 2002; Yoon et al., 2002). Therefore, we examined whether OprF play a role in the invasion of macrophages by P. aeruginosa. Both human and mouse macrophages were infected with wild-type (WT), ΔoprF and trans-complemented (pOprF) strains of P. aeruginosa. At 2 h post-infection, there was no significant difference (P > 0.05) in the phagocytosis of ΔoprF mutant strain of P. aeruginosa compared to WT or pOprF strains (Figures 7a,b). However, the number of ΔoprF bacterial mutant decreased significantly inside both human MDMs and mouse BMMϕ at 4 h post-infection (P < 0.01). At 6 h post-infection, no viable ΔoprF bacteria were recoverable from human MDMs and very low number of bacteria from mouse BMMϕ. In contrast, high numbers of bacteria were recoverable from WT or pOprF infected human MDMs and mouse BMMϕ.

To further confirm the role of OprF in invasion of macrophages by P. aeruginosa, WT bacteria were pretreated with anti-OprF monoclonal antibody or isotype control or left untreated and then used to infect human MDMs and mouse BMMϕ. Pretreatment of bacteria with anti-OprF monoclonal antibody significantly decreased the number of bacteria recoverable from human MDMs at 4 and 6 h post-infection compared to isotype control treated or untreated bacteria (P < 0.01) (Figure 7c). Similar decrease in number of intracellular P. aeruginosa was observed at 4 and 6 h post-infection when pretreated bacteria were used to infect mouse BMMϕ (Figure 7d). These results suggest that OprF plays a critical role in the survival of P. aeruginosa inside macrophages.

To determine whether infection of macrophages with P. aeruginosa causes cell death. Human MDMs and mouse BMMϕ were infected with P. aeruginosa at an MOI of 10 and the levels of LDH released were determined in the cell culture supernatants. LDH release is the most acceptable and reliable marker to determine cell viability^34-36. There was not much cell damage up to 8 h post-infection as demonstrated by minimal LDH release in cell culture supernatants of both human MDMs and mouse BMMϕ infected with WT otopathogenic P. aeruginosa (Figures 8a,b). However, a further increase in time resulted in significant cell death (P < 0.01). LDH levels increased from 12.5% at 10 h post-infection to 46.7% by 16 h post-infection in culture supernatants of infected human MDMs. Similarly, LDH levels increased from 15.7% at 8 h post-infection to 39.8% at 16 h post-infection in mouse BMMϕ. At 24 h post-infection, there was even a higher level of cell death, as indicated by high LDH levels in human MDMs and mouse BMMϕ. Interestingly, deletion of oprF abrogated the ability of otopathogenic to induce significant cell death in macrophages (P < 0.01). However, complementation with the pOprF plasmid restored the ability of mutant strain to exert cytopathic effects on human MDMs and mouse BMMϕ.

To further confirm that macrophages undergo apoptosis, infected human MDMs and mouse BMMϕ were stained with ethidium homodimer-1, a high affinity, membrane-impermeant dye that exclusively stains the DNA of dead cells (Eidet et al., 2015; McCanna et al., 2015; Yoeruek et al., 2016). In agreement with the LDH assay results, at 8 h post-infection very little ethidium homodimer-1 staining of infected human MDMs was observed (P > 0.05) (Supplementary Figure S6). However, 25% of human MDMs were stained with membrane-impermeant dye at 10 h post-infection. At 16 h post-infection, 50% of infected MDMs failed to exclude the dye. By 24 h post-infection, almost 75% of human MDMs had been stained with ethidium homodimer-1 suggesting considerable cell death. Similar results were obtained with mouse BMMϕ demonstrating extensive cell death at 24 h post-infection (Supplementary Figure S6). On par with our LDH data, human MDMs and mouse BMMϕ infected with ΔoprF mutant strain did not demonstrate significant ethidium homodimer-1 staining even at 24 h post-infection compared to WT infected cells (P < 0.01). This phenotype was restored upon complementation with pOprF. These results suggest that otopathogenic P. aeruginosa exerts cytopathic effects on macrophages for which bacterial OprF expression is necessary.