Pseudomonas aeruginosa strain PAO1 (Holloway1C Stanier131) was obtained from the National Collection of Industrial, Food and Marine Bacteria (NCIMB), United Kingdom. PAO1 was grown in Luria-Bertani (LB, Oxoid) broth and on nutrient agar (Oxoid) at 37°C.

Corneal epithelium was scraped from corneo-scleral rims of healthy donors and the stromal layer was dissected and digested with collagenase type-1 (1 mg/mL; Life Technologies) for 3 h at 37°C. Digested stroma was cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Sigma–Aldrich) supplemented with 5% (v/v) decomplemented Fetal Calf Serum (dFCS, Life Technologies), 100 Units/mL of penicillin and 100 Units/mL of streptomycin (Life Technologies) and 0.5 μg/mL of amphotericin B (Lonza) and fibroblasts were characterized as described previously (Wong et al., 2011). The number of cell passages used in this study was kept between 3 and 6 to minimize fibroblast differentiation to myofibroblast, as seen in other studies (Masur et al., 1995; Bargagna-Mohan et al., 2015). Fibroblasts from each donor were kept separate (not pooled) and used for independent experiments. Cells were incubated in a humidified atmosphere containing 5% (v/v) CO₂ at 37°C.

In experiments to examine the efficacy of different antibiotics against extracellular and intracellular PAO1, hCFs were cultured in antibiotic-free DMEM supplemented with 5% (v/v) dFCS for at least one passage prior to bacterial challenge, in order to ensure the absence of possible traces of antibiotics.

Patients provided written informed consent in accordance with the Declaration of Helsinki to use surplus corneal tissue specimens for research via the NHS Blood Transplant Eye Retrieval Service. Protocols were approved by the NRES Committee South Central - Berkshire 06/Q1602/56.

The antibiotics chosen for this in vitro study were based on those commonly used in United Kingdom clinical practice, namely the bactericidal antibiotics ciprofloxacin (CIP, Sigma–Aldrich), cefuroxime 0.5% (w/v) (CXM, Hampshire Hospitals NHS), polymyxin B (PMB, Sigma–Aldrich), gentamicin (GEN, Sigma–Aldrich), ofloxacin (0.3% w/v) (OFX, Allergan), levofloxacin (0.5% w/v) (LVX, Santen) and the bacteriostatic antibiotic chloramphenicol 0.5% (w/v) (CHL, Bausch and Lomb). A broth dilution assay was done to determine PAO1 susceptibility to each antibiotic. Various concentrations of antibiotic (0.01, 0.1, 1, 10, 50, 100, and 200 μg/mL) were added to equal volumes of LB broth medium with various concentrations of bacteria (∼10⁵, ∼10⁶ and ∼10⁷ Colony Forming Units (CFU)/mL). Absorbance (Optical Density (OD) λ₆₀₀ nm) as a measure of bacterial growth was assessed hourly up to 9 h and then at 24 h of incubation at 37°C with shaking (200 rpm). Comparison of the 24 h OD readings for bacterial growth with and without antibiotics was used to calculate the lowest concentration of antibiotic that completely reduced OD growth by 100% and the concentration of antibiotic that reduced OD growth by 50%.

Extracellular and intracellular PAO1 bacteria were quantified in the presence of each antibiotic used at concentrations of 50 and 200 μg/mL. Confluent hCF monolayers were challenged in triplicate in 24-well cell culture plates (Greiner Bio-one, ∼10⁵ hCF/well) with different Multiplicities Of Infection (MOI = 1, ∼10⁵ CFU/ml; MOI = 10, ∼10⁶ CFU/mL and MOI = 100, ∼10⁷ CFU/mL) of PAO1. At given time-points, cell monolayers were washed gently four times with phosphate buffered saline (PBS), pH7.4 and associated PAO1 bacteria were quantified by viable counting on nutrient agar after hCF lysis with a buffer of PBS containing 0.1% (w/v) saponin and 0.1% (v/v) dFCS, as described previously (Hung et al., 2013).

The gentamicin exclusion assay was used to eliminate extracellular bacteria in order to subsequently quantify intracellular bacteria. The assay was done as described previously (Hung et al., 2013), whereby infected cell monolayers are washed with PBS, pH 7.4 and then incubated with a solution of gentamicin (200 μg/mL) for 90 min to kill extracellular bacteria. The monolayers were washed to remove gentamicin and then lysed with a solution of PBS containing saponin (0.1% w/v) and dFCS (0.1% v/v) and the surviving intracellular bacteria enumerated by viable counting on NA.

To quantify the effects of the different antibiotics on survival of intracellular PAO1 bacteria, 1 mL of DMEM containing 50 or 200 μg/mL of the antibiotics listed above was added to hCFs after 3 h of PAO1 infection at the different MOIs, and intracellular PAO1 enumerated after 1.5, 4.5, 7.5, and 24 h of each antibiotic treatment. At these given time points, monolayers were washed four times with PBS, lysed and intracellular bacteria quantified as described above. Extracellular growth of PAO1 during cell monolayer infection with different antibiotic treatments was quantified by viable counting at 0, 1, 3, 6, 9, and 24 h on nutrient agar.

hCF monolayers were infected for 3 h with PAO1 (MOI = 1, 10, and 100), then treated with antibiotics and IL-1β cytokine was quantified from 9 h antibiotic-treated-hCF supernatants using the Meso Scale Discovery (MSD) electro-chemiluminescence assay (MesoScaleDiagnostics), following the manufacturer’s instructions.

The cytotoxicity of CIP, CXM, PMB, GEN, OFX, LVX and CHL (50 and 200 μg/mL) alone to hCFs and of PAO1 infection (MOI = 1, 10, and 100) with and without antibiotics, was examined after 0, 1, 3, 6, 9, and 24 h of incubation. Release of lactate dehydrogenase (LDH) was measured using the Pierce LDH Cytotoxicity Assay Kit (Thermo-Scientific) and the percentage cytotoxicity was calculated following the manufacturers’ instructions.

Data were analyzed by one-way ANOVA with Dunnett’s multiple comparison test or by unpaired t-test. P-values < 0.05 denoted significance.

In the current study, we used an in vitro model of primary hCF cells derived from the stroma, to evaluate current antibiotic treatments used in the United Kingdom to treat eye infections. Initially, we used a broth dilution method to calculate the lowest concentration of GEN, CXM, OFX, PMB, CIP, CHL and LVX antibiotics that completely reduced OD growth of P. aeruginosa strain PAO1 (tested at various concentrations of ∼10⁵ – ∼10⁷ CFU/mL) by values of 50 and 100%. In rank order, the fluoroquinolones CIP and LVX and polymyxin PMB showed the highest antimicrobial activities with 100% reduction in OD with a concentration ≤10 μg/mL and 50% reduction in OD with a concentration ≤1 μg/mL (Table 1). This was followed by GEN and OFX (100% reduction in OD with concentrations ≤50 μg/mL and 50% reduction in OD with concentrations ≤10 μg/mL) and lastly by CXM and CHL treatments, which showed the lowest activities (≥200 μg/mL) and failed to sterilize the PAO1 broth cultures (Supplementary Figure S1 and Table S1). LB broth could support the planktonic growth of P. aeruginosa PAO1 over time in the absence of antibiotics (Supplementary Figure S2).

Next, we examined the bactericidal ability of the antibiotic panel to control PAO1 extracellular infection of cultured hCF cells. The antibiotic concentrations to be tested were chosen as 50 and 200 μg/mL based on their ability to reduce planktonic growth of PAO1 in broth (Table 1). For some of the antibiotics, e.g., CIP, LVX and PMB, these concentrations were significantly higher than the concentrations that could completely inhibit bacterial growth as judged by the broth dilution assay, but could be seen as potentially representing accumulated antibiotic concentrations in clinical use with repeated dosing of infected eye tissue. hCF cultures were infected with PAO1 (MOI = 1, = 10, = 100), antibiotics added at time = 0 h and extracellular growth of the organism quantified (CFU) over time (Figure 1). Concentrations of 50 and 200 μg/mL of CIP, LVX, PMB, GEN and OFX killed all PAO1 (>99.99%; P < 0.0001) in the extracellular environment by 1 h (Figure 1A–E). Treatment with CXM at 50 μg/mL resulted in a >99% reduction of extracellular bacterial CFU by 6 h (P < 0.05), but this was not maintained and bacterial CFU increased thereafter to levels similar to bacterial growth observed in PAO1-infected monolayers not treated with antibiotics (≥10⁸–10⁹ CFU/mL; P > 0.05, using an unpaired t-test; Supplementary Figure S3). CXM treatment only killed extracellular PAO1 when used at a concentration of 200 μg/mL (P < 0.05 up to 6 h and P < 0.001 at 24 h), with sterility occurring between 6–9 h (Figure 1F). By contrast, CHL was ineffective at killing PAO1 at both concentrations (Figure 1G) (P > 0.05), with extracellular CFU of PAO1 increasing over time to levels (≥10⁸ CFU/mL; P > 0.05, using an unpaired t-test) similarly observed in PAO1-infected monolayers not treated with antibiotics (Supplementary Figure S3).

PAO1 infection is cytotoxic to hCFs in vitro, with MOI-dependent cell death occurring between 9–24 h, as judged by loss of monolayer integrity and increased release of LDH due to membrane damage (Supplementary Figure S4A). We next examined whether the antibiotic treatments protected hCF monolayers from PAO1-induced cell death. In general, antibiotics alone were not cytotoxic to hCFs in vitro, except for treatment with PMB, which increased LDH release by 30–40% by 24 h, and with OFX to a lesser extent (∼10% increased LDH release) (Supplementary Figure S4B). By visual inspection, PAO1-infected hCF monolayers treated and maintained with the antibiotics CIP, LVX, GEN and PMB were intact after 24 h. OFX-treated hCF cells showed signs of apoptosis by morphology, i.e., cell rounding and shrinkage, although the monolayers were still intact. By contrast, CXM or CHL treatments failed to protect hCF monolayers from PAO1 infection, with cell death and monolayer detachment occurring between 9 and 24 h.

The protective effect of antibiotics was confirmed by measuring LDH release (Figure 2). Monolayers were infected with PAO1 (MOI = 1, 10, 100), antibiotics were added at 3 h and LDH release measured at 9 and 24 h. At a MOI = 1, there was no significant difference at 9 h in the cytotoxicity levels between CIP and GEN concentrations and no antibiotic (P > 0.05), whereas cytotoxicity levels were increased marginally following LVX, OFX, CXM, CHL (P < 0.05) and significantly with PMB (P < 0.0001) treatments. By 24 h, significant reductions in cytotoxicity were observed with CIP, LVX, GEN and OFX (P < 0.0001), PMB and CXM (P < 0.05) and CHL (P < 0.01).

At a MOI = 10, there was a significant reduction by 9 h in percentage cytotoxicity with 50 and 200 μg/mL concentrations of CIP, LVX and GEN (P < 0.001) and OFX (P < 0.05) and with 200 μg/mL concentration only of CXM (P < 0.05). With PMB treatment, there was no significant difference in cytotoxicity levels with the 50 μg/mL concentration (P > 0.05), but the 200 μg/mL concentration increased cytotoxicity significantly (P < 0.001). By 24 h, highly significant reductions in cytotoxicity were observed with 50 and 200 μg/mL of CIP, LVX, PMB, GEN, OFX and CXM (P < 0.0001). CHL used at a concentration of 50 μg/mL had no significant effect on LDH release at either 9 or 24 h (P > 0.05), but percentage cytotoxicity was reduced with the higher concentration tested (P < 0.05) (Figure 2).

At a MOI = 100, there was a significant reduction by 9 h in percentage cytotoxicity with concentrations of 50 and 200 μg/mL of CIP and LVX (P < 0.0001) and GEN and OFX (P < 0.01) and with CHL (P < 0.01, 200 μg/mL concentration only). By contrast, PMB used at 200 μg/mL significantly increased cytotoxicity (P < 0.05). By 24 h, a 50 μg/mL concentration of CIP, LVX, OFX, PMB, GEN and OFX significantly reduced cytotoxicity (P < 0.0001), but both CXM and CHL had no significant effect (P > 0.05). With the higher concentration of 200 μg/mL, significant reductions were observed for all antibiotic treatments (CIP, LVX, GEN P < 0.0001; PMB, OFX, CXM and CHL P < 0.05).

Recently, we have shown with adherence and gentamicin exclusion assays that P. aeruginosa can adhere to primary hCFs and subsequently invade (Cendra et al., 2017). In the current study, MOI-dependent levels of PAO1 adherence and subsequent invasion were observed after 3 h of infection (Figure 3). The numbers of invasive PAO1 were quantified as ∼5 × 10² CFU/monolayer with an initial MOI = 1, ∼10³ CFU/monolayer for an initial MOI = 10 and ∼5 × 10⁴ CFU/monolayer for an initial MOI = 100 (Figure 3).

We next tested the hypothesis that the ability of PAO1 to invade hCFs enabled the organism to evade antibiotic chemotherapy, thereby creating a potential reservoir for reactivation of bacterial infection. In order to examine the effects of antibiotics on intracellular bacteria, hCF monolayers were infected with PAO1 (MOI = 1, 10, 100) for 3 h to allow bacterial invasion to occur (Figure 3), after which antibiotics (50 or 200 μg/mL) were added to the cultures. The numbers of intracellular PAO1 were then quantified after 1.5, 4.5, 7.5, and 24 h of antibiotic treatment. A clear rank order for antibiotic efficacy was observed against intracellular PAO1 in vitro. The fluoroquinolones CIP, LVX and OFX demonstrated the highest antimicrobial activity (Figure 4A). Treatment with CIP and LVX (50 and 200 μg/mL) killed >99.99% of all intracellular bacteria by 3 h, regardless of MOI used to infect the monolayers (Figures 4A,B). Treatment with OFX (50 or 200 μg/mL) eradicated >99.99% of intracellular bacteria by 3 h from hCF monolayers infected initially with an MOI = 1 (Figure 4C). With increasing MOI = 10–100, between ∼10² – ∼10³ intracellular bacteria were recovered by 24 h after OFX (50 or 200 μg/mL) treatment, which still represented high antimicrobial activity of >99.9% reduction in CFU (Figure 4C).

By contrast to the fluoroquinolones, treatment with GEN, PMB, CXM or CHL failed to clear intracellular PAO1 over the 24 h time course of the experiments. With cells infected with MOI = 1, 10, or 100, ∼10³ – ∼10⁵ intracellular bacteria were recovered after treatment for 3 h with GEN or PMB (50 and 200 μg/mL), and these intracellular PAO1 numbers essentially remained unchanged by 24 h (Figures 4D,E). CXM treatment at 50 μg/mL could not clear the infection (intracellular or extracellular), and exponential increases in bacterial CFU were observed (Figure 4F), and the numbers of intracellular PAO1 enumerated at 3 h (∼10³ – ∼10⁵ CFU) in the presence of 200 μg/mL of CXM remained essentially unchanged by 24 h (Figure 4F). These experiments confirmed also that PAO1 was resistant to CHL, with exponentially increasing numbers of bacteria recovered over time intracellular and extracellularly (Figures 2, 4F and Supplementary Table S2), with cell death occurring early by 9 h.

Clinically, a high incidence of recurrent BK has been observed in patients after cessation of antibiotic treatment (Kaye et al., 2013). We next tested the hypothesis that culture sterility and monolayer integrity were unaffected after removal of the antibiotics at 24 h. hCFs were infected with PAO1 at a MOI = 1 in order to detect logarithmic changes in PAO1 growth, for 3 h prior to the addition of antibiotics, as described above. At 24 h, antibiotic-containing media were removed, the cell monolayers washed and fresh antibiotic-free medium added. Extracellular and intracellular PAO1 growth was then measured at time intervals up to 48 h after removal of antibiotics. We compared the fluoroquinolones CIP, LEV and OFX along with GEN and PMB, but excluded the CXM and CHL antibiotic treatments due to their inability to limit infection. Intracellular PAO1 growth was eventually detected in all monolayers after removal of antibiotics at 24 h. With CIP and LVX-treated monolayers, intracellular PAO1 growth was detected as early as 3 h after antibiotic removal and increased exponentially over time to 24 h (Figures 5A,B). With OFX, intracellular PAO1 were recovered from infected hCF-monolayers at 24 h (Figure 4) and after antibiotic removal, intracellular CFU numbers showed similar exponential increases up to 24 h (Figure 5C). With these three treatments, extracellular bacteria were recovered by 9 h onward after antibiotic removal and CFU numbers increased exponentially (Figures 5A–C). For both PMB and GEN, which failed to clear intracellular bacteria after 24 h (Figure 4), removal of the antibiotics resulted in increased CFU numbers intracellularly and early recovery of extracellular bacteria (Figures 5D,E). Regardless, after removal at 24 h of all the different antibiotics, by 24–48 h later, all the monolayers were destroyed as a consequence of uncontrolled bacterial growth. All the bacteria recovered at 48 and 96 h now represented growth in host cell-free cultures.

Infection of hCFs with P. aeruginosa PAO1 induces significant extracellular release of the cytokine IL-1β (Wong et al., 2011; Cendra et al., 2017). Since exacerbation of the inflammatory response is a critical event during eye infection, we tested the hypothesis that antibiotic treatments not only reduced bacterial infection but were also anti-inflammatory, by influencing IL-1β cytokine production. Our cytotoxicity data demonstrated that after 9 h, untreated PAO1 infection appeared to compromise hCF monolayer integrity; therefore, IL-1β cytokine was measured at this time-point in supernatants of PAO1-infected (MOI = 1, 10, 100) and un-infected hCFs (MOI = 0) monolayers following addition of the different antibiotics after 3 h of infection (Figure 6). Baseline levels of IL-1β in uninfected monolayers without antibiotic treatments were ≤0.25 pg/mL (Figure 6A). Addition of some of the antibiotics, i.e., PMB, OFX, CXM and CHL, did statistically increase IL-1β release from uninfected monolayers, but the levels never exceeded 6 pg/mL (Figure 6A and Supplementary Table S4).

Increased levels of IL-1β production were observed with increasing PAO1 MOI: with MOI = 1, IL-1β levels with no antibiotic were ∼6 pg/mL and addition of all the antibiotics at both concentrations significantly reduced these low levels to ≤3 pg/mL (Figure 6B and Supplementary Table S4). With MOI = 10, IL-1β levels with no antibiotic were ∼15–20 pg/mL and addition of all the antibiotics at a concentration of 50 μg/mL either reduced (CIP, LVX, GEN, OFX) or had no significant stimulatory effect (PMB, CXM, CHL) on IL-1β secretion (Figure 6C and Supplementary Table S4). Similarly, addition of 200 μg/mL of the antibiotics either reduced (CIP, PMB, GEN) or had no significant stimulatory effect (LVX, OFX, CXM, CHL) on IL-1β levels, except for PMB, which significantly increased (P < 0.0001) IL-1β levels (≥40 pg/mL) in the presence of bacteria (Figure 6C). With MOI = 100, ∼50–60 pg/mL IL-1β was induced (Figure 6D) and none of the antibiotics used at a concentration of 50 μg/mL statistically reduced IL-1β release (Supplementary Table S4). Only CIP and GEN treatment at 200 μg/mL significantly decreased IL-1β release by ≥50% (Supplementary Table S4). PMB, CXM and CHL had no anti-inflammatory effect with similar levels of IL-1β release (P > 0.05) measured, whereas addition of 200 μg/mL of OFX was pro-inflammatory, significantly (P < 0.001) increasing IL-1β release by ∼2-fold (>100 pg/mL) compared to no antibiotic, infected hCF cells (Figure 6D and Supplementary Table S4). Notably, cells infected with MOI = 100 and treated with OFX (200 μg/mL) were morphologically apoptotic, which was not observed in PAO1 infected cells treated with any of the other antibiotics.