Repurposing FDA-approved drugs: a high-throughput screen for ExoU phospholipase inhibitors

Recombinant ExoU bearing a C-terminal 6 × His tag was expressed in E. coli and purified using immobilised metal affinity chromatography (IMAC) followed by size-exclusion chromatography, as described previously [25] (S1 Fig). ExoU in vitro phospholipase activity was analysed using a real-time, high-throughput phospholipase assay based on a modified version of the Cayman Chemical PLA₂ kit, adapted for 96- and 384-well formats [25] (S1 Fig). The assay was conducted in the presence of PIP₂ and ubiquitin, cofactors essential for ExoU activation and catalysis.

In the initial screen of 3,042 FDA-approved compounds from the Selleckchem library (Fig 1A), those that reduced ExoU activity to 40% or less, relative to DMSO-treated controls, were advanced to dose-response analysis. These were performed to eliminate false positives and determine IC₅₀ values of confirmed inhibitors (Fig 1B), with compound structures illustrated in Fig 1C.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. High-throughput screening identifies FDA-approved ExoU inhibitors with activity in vitro and in cell-based infection models.

(A) A high-throughput in vitro phospholipase assay with 175 nM ExoU (specific activity 5.85 ± 0.26 μmol/min/μg) and absorbance read at 414 nm was used to screen 3,042 compounds (50 µM) from the Selleckchem FDA-approved drug library to identify novel inhibitors of ExoU. The plot shows percentage inhibition of ExoU phospholipase activity after a 1 h enzymatic reaction, with reference to 0.1% (v/v) DMSO controls (100% activity). (B) Dose-response curves of selected compounds were generated to determine IC₅₀ values for ExoU inhibition in vitro. (C) Chemical structures of lead compounds. (D) High-content imaging screen of lead ExoU inhibitors using a high-throughput scratch and infection assay. Human corneal epithelial (HCE-T) cells were infected with P. aeruginosa PA103 (MOI 10). After 4 h, cell lysis was quantified by ethidium homodimer staining. The bar graph shows the percentage of lysed cells, reflecting ExoU-mediated cytotoxicity, following treatment with cyclofenil (Cyclo), mivacurium chloride (MivC), chlorhexidine (Chlor), bismuth subcitrate (Bis), zinc pyrithione (ZP), and polymyxin B (Pol). White bars indicate compounds alone, while black bars indicate compounds in the presence of 2.5 µM moxifloxacin (mox). No infection indicates vehicle controls of 0.1% (v/v) DMSO control without the addition of PA103.

https://doi.org/10.1371/journal.ppat.1014474.g001

Of the 35 initial screening hits, we first excluded any compounds that either absorbed at ~414 nm or otherwise interfered with the DTNB secondary reporter reaction or were unsuitable for therapeutic development (e.g., chemotherapy agents). Six showed clear inhibitory activity in the low micromolar range. Zp emerged as the most potent inhibitor identified to date, with an IC₅₀ of 1.4 ± 0.4 μM (Fig 1B) and a Kᵢ of 2.4 μM (S2 Fig). Several additional compounds showed strong to moderate inhibition, including mivacurium chloride (MivC; IC₅₀ = 6.0 ± 1.8 μM), bismuth subcitrate (Bis; 12.8 ± 1.0 μM), polymyxin B (Pol; 13.1 ± 1.7 μM), chlorhexidine (Chlor18.5 ± 3.3 μM), and cyclofenil (20.2 ± 2.8 μM).

With the exception of Cyclofenil, a selective estrogen receptor modulator with reported adverse effects [26] that could make it unsuitable for ocular application, the other hits represent attractive candidates for repurposing as treatments for P. aeruginosa keratitis. These included polymyxin B, a cationic lipopeptide antimicrobial, with dose-limiting neurotoxicity when administered systemically, and chlorhexidine, a broad-spectrum antiseptic used in ophthalmic and periocular disinfection. Notably, several compounds, including polymyxin B and chlorhexidine are already used topically in ophthalmology [1,27], enhancing their translational potential. Zp, although not currently used for eye infections has been used as an active ingredient in anti-dandruff shampoos. Its applications span across medical treatments (anti-fungal), household products, and industrial materials [28]. Zp exhibited promising potency against ExoU, prompting us to investigate it further. The full data from the initial screen comprising initial ‘hits’, compounds that interfered with assay readouts and dose response validated compounds are in S1 Table.

Determining efficacy of lead ExoU inhibitors in a cellular wound infection model

ExoU is synthesised in P. aeruginosa as an inactive complex with the SpcU chaperone [29]. Dissociation from SpcU and significant unfolding are required for its translocation through the T3SS into host cells, where is it refolded and then exerts its activity. Thus, it was important to evaluate the efficacy of ExoU inhibitors in a cellular context. A high-throughput epithelial infection model was developed based on our previously reported scratch wound assay in HCE-T corneal epithelial cells [25]. The assay was adapted to a 96-well plate format compatible with high-content live-cell imaging using the IncuCyte S3 system. Uniform scratch wounds were generated using the IncuCyte WoundMaker, and HCE-T cells were infected with ExoU-producing P. aeruginosa PA103 (MOI 10). Cell viability and lysis were continuously monitored for 15 h at 30 min intervals using calcein-AM (viable cell marker) and ethidium homodimer-1 (membrane damage/lysis marker) (Fig 1D).

Because P. aeruginosa replicates rapidly in vitro, as before [25], we incorporated sub-MIC moxifloxacin (2.5 μM; 50% MIC) to slow bacterial growth, extend the infection window and enhance detection of ExoU-dependent cytotoxicity. Moxifloxacin does not inhibit ExoU nor the T3SS [30], ensuring that observed effects are not due to interference with toxin delivery or activity.

DMSO-treated, uninfected monolayers exhibited minimal background cell death (8.0 ± 1.8% ethidium-positive cells at 10 h). Sub-MIC Moxifloxacin was confirmed to be non-toxic to HCE-T cells, as ethidium homodimer uptake remained unchanged compared to untreated controls (<1% difference at 10 h). In contrast, infection with PA103 in DMSO-treated wells induced robust ExoU-dependent lysis, with ~40% of cells staining positive for ethidium homodimer at 10 h (Fig 1D). The inclusion of sub-MIC moxifloxacin did not appreciably reduce cytotoxicity (38.8% lysis), confirming, in line with our previous studies [30], that ExoU-mediated epithelial injury occurs even under conditions of partially restricted bacterial growth (Figs 1D and S3 Fig).

We next evaluated the six lead compounds in the infection model (Fig 1D). Cyclofenil, MivC, and chlorhexidine did not reduce ExoU-mediated lysis, whereas Bis, Zp, and Pol conferred clear protection of HCE-T cells.

To determine whether this protection reflected antibacterial activity, we assessed bacterial growth across a range of concentrations (S3 Fig). While Zp, chlorhexidine, and Pol exhibited dose-dependent bactericidal activity at higher concentrations, the concentrations used in the infection assays were below these thresholds and did not measurably affect bacterial viability. This was confirmed by both growth curves and endpoint CFU enumeration following IncuCyte imaging (S3 Fig), demonstrating that compounds combined with 2.5 µM moxifloxacin did not increase bactericidal activity (S3 Fig).

Polymyxin B exhibited an IC₅₀ of 0.75 µM in LB and 2.1 µM in DMEM/F-12 supplemented with 10% (v/v) FBS, likely reflecting serum interactions. The concentration used in the cell-based infection assays (1 µM) did not significantly reduce bacterial viability under these conditions (S3 Fig).

Collectively, these findings identified Pol, Bis, and Zp as ExoU targeting inhibitors capable of protecting corneal epithelial cells from P. aeruginosa cytotoxicity independent of bacterial killing. This also establishes the infection model as a robust platform for future screening of antivirulence therapeutics.

Selectivity of the ExoU inhibitors

To establish the selectivity of the ExoU inhibitors, they were also screened against PLA2G7 and PLA2G4C which represent two distinct and major classes of human calcium-independent PLA₂ [31,32], to evaluate the selectivity of our hits for bacterial ExoU over host phospholipases. Using 2-thio-PAF as the substrate for human PLA2G7 we found that neither Bis, Pol, nor Zp inhibited enzymatic activity, whereas darapladib, a clinical PLA2G7 inhibitor, effectively suppressed it (S4 Fig). In contrast, darapladib did not inhibit ExoU activity when arachidonoyl thio-PC was used as the substrate (S4 Fig). Similarly, using arachidonoyl thio-phosphatidylcholine (thio-BGPC) as a substrate for PLA2G4C we observed no inhibition by Bis, Pol or Zp (S4 Fig). The use of different thio-ester substrates also underscored this selectivity: while darapladib effectively blocked PLA2G7 with its preferred substrate, it had no effect on ExoU. The absence of inhibition by Bis, Pol and Zp of these enzymes suggests that they selectively targeted ExoU without inhibiting human PLA₂ activity, reducing the likelihood of host toxicity.

The cytotoxicity of the compounds was measured directly in primary human corneal epithelial cells (S5 Fig). The LDH assays showed that Bis had no measurable toxicity up to 160 µM, while Zp and Pol had LD₅₀ of 32.6 ± 12.6 µM and 42.5 ± 10.6 µM, respectively. Importantly, these LD₅₀ values were an order of magnitude higher than the concentrations applied in the infection assays.

ExoU is inhibited by zinc and bismuth ions

To determine the ExoU inhibitory active pharmacophore in Zp, we assessed copper pyrithione, pyrithione alone and Zn²⁺ in vitro phospholipase assays (Fig 2A). Neither copper pyrithione, nor pyrithione inhibited ExoU (Fig 2A), suggesting that ExoU activity was sensitive to zinc. To further probe this, a panel of metal salts was tested and only zinc sulphate inhibited ExoU phospholipase activity (Fig 2B). We therefore, then determined whether Zn²⁺ alone could inhibit ExoU-mediated cytotoxicity (S6 Fig). LDH release assays in PA103-infected HCE-T cells showed that ZnSO₄ (2.5 µM) did not reduce cell lysis compared to the DMSO control (mean lysis: 89% vs. 85%), whereas Zp (2.5 µM), comprising a pyrithione zinc ionophore, significantly reduced cytotoxicity (mean lysis: 55%). These findings indicate that zinc ions required a carrier to enter host cells and inhibit ExoU activity and that the Zn²⁺chelate Zp is itself active.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Chemical features associated with ExoU inhibition: metal ions and fatty acid peptides.

(A) Real-time phospholipase assay showing ExoU inhibition in the presence of pyrithione and pyrithione-coordinated zinc or copper complexes. (B) ExoU phospholipase activity detected in the presence of a panel of selected metal salts. (C) Dose-response phospholipase assays used to determine IC₅₀ values for bismuth-based ExoU inhibitors and subcitrate alone. (D) Chemical structures of native polymyxin B, the cyclic peptide fragment of polymyxin B (nanopeptide), and custom-synthesized polymyxin-derived fatty acid peptides: Poltide 1 (truncated polymyxin mimic), Poltide 2 (negatively charged sequence), and Poltide 3 (linear polymyxin-like analog). (E) Real-time phospholipase activity assay demonstrating ExoU inhibitory profiles of each polymyxin compound.

https://doi.org/10.1371/journal.ppat.1014474.g002

The ExoU inhibitory active pharmacophore in Bis was determined using other bismuth containing compounds, bismuth subsalicylate, bismuth chloride and subcitrate in phospholipase assays. All bismuth containing compounds inhibited ExoU activity, whereas subcitrate alone did not (Fig 2C). Thus, the Bi³⁺ ion was the pharmacophore in the chelate and the chelator subcitrate likely a carrier.

Inhibition by Polymyxin B depends on its peptide and acyl tail combined

Pol comprises a cyclic cationic decapeptide linked to a hydrophobic fatty acyl tail (Fig 2D) and the inhibitory activity of these structural elements was assayed individually. The isolated hydrophobic tail ((S)-6-methyloctanoic acid) failed to inhibit ExoU (Fig 2E). The cyclic peptide portion of polymyxin B alone (nanopeptide) also did not inhibit ExoU. This indicated that Pol’s inhibitory activity likely arose from synergy between its cationic peptide and lipid tail.

The role of the fatty acyl-peptide linkage in ExoU inhibition, was examined by using three custom “poltide” analogues (Fig 2D), each bearing an N-terminal lauric acid (C₁₂) lipid attached to a simplified polymyxin-inspired peptide. Poltide 1 contained an N-terminal lauric acid followed by –Dab–Thr–Dab. In Poltide 2, the charge of the first 2,4-diaminobutyric acid residue was reversed by replacing it with an aspartic acid (poltide 2: lauric acid–Asp–Thr–Dab), while Poltide 3 was extended by consisting of lauric acid followed by Dab–Thr–Dab–Dab–Dab–Phe. Poltide 1 inhibited ExoU, albeit more weakly than Pol (IC₅₀ 40 µM), whereas Poltide 2, bearing an opposite charge in the first amino acid residue, had no effect on ExoU activity (Fig 2E). Poltide 3 was more potent than Poltide 1 (IC₅₀ 20 µM) but did not recapitulate fully the full inhibitory activity of Pol (IC₅₀ 13 µM). These data highlight that both peptide length and charge, and potentially its cyclic structure are critical for ExoU inhibition.

In vitro biochemical effects of ExoU inhibition by Bis, Pol and Zp

The effects of Bis, Pol and Zp on the thermal stability of ExoU, a proxy for its structure/conformation, and oligomeric assembly were measured using nanoDSF and BS³ crosslinking.

Thermal stability of ExoU

ExoU had a melting temperature (Tₘ) of 40.8 °C (S7 Fig). Addition of 5 µM PIP₂ increased the Tₘ to 43.1 °C, indicating that PIP₂ binding alters the apparent thermal stability of ExoU under assay conditions. Zp decreased the Tₘ of ExoU by 1.8 °C in the absence of PIP₂ and by 1.3 °C in its presence, suggesting a modest destabilising effect that is largely independent of lipid binding. Pol produced a more pronounced decrease in thermal stability, lowering the Tₘ by 3.8 °C without PIP₂ and by 6.2 °C in its presence. Bis had no measurable effect on Tₘ in the absence of lipid, but in the presence of PIP₂ reduced the Tₘ by ~2 °C. Collectively, these data indicate that all three compounds interact with ExoU and modulate its thermal stability in a PIP₂-dependent (Bis) or PIP₂-independent (Zp and Pol) manner.

After reacting ExoU with the homobifunctional cross-linker BS³, the protein migrated on SDS-PAGE as a smear around 74 kDa (Fig 3B). The smearing may have been the result of the reaction products, whereby BS³ may react with a single amino group (generally lysine side chains) or form a loop between two, so generating a range of products of different sizes. While Bis did not alter this profile, in the presence of Zp the ExoU migrated as a smear centred on a band >250 kDa, with only a small amount of monomer remaining. In contrast in the presence of Pol, ExoU migrated as a diffuse smear centred around ~100 kDa (Fig 3B). This likely represents a dimer which migrated anomalously due to BS³ cross linking.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. ExoU destabilisation, oligomerisation and inhibition in the presence of inhibitors and PIP₂.

(A) Nano differential scanning fluorimetry (nanoDSF) analysis of ExoU thermal stability in the presence of compounds (10 µM) without (left) and with (right) PIP₂ (5 µM) present. Chart summarises ΔTm values with standard deviation. (B) BS³ cross-linking assay of recombinant ExoU followed by SDS-PAGE and anti-6 × His immunoblotting, showing compound-dependent changes in oligomerisation in the presence or absence of PIP₂. ExoU was incubated with 10 µM inhibitor (Bis, Zp, or Pol) or DMSO control ± PIP₂ for 30 min at 25 °C prior to cross-linking. Molecular weight markers (kDa) are indicated (maximum 250 kDa). Lanes 1-4: DMSO, Bis, Zp, Pol (−PIP₂). Lanes 5-12: DMSO, Bis, Zp, Pol (+PIP₂; conditions shown in duplicate). Monomeric ExoU migrates at ~74 kDa, with higher molecular weight species corresponding to dimeric (~150 kDa), tetrameric, and higher-order oligomeric complexes. (C) Combinatorial compound treatment enhances ExoU inhibition in vitro. Phospholipase activity of ExoU was measured in real-time assays in the presence of individual inhibitors (Bis, Zp, or Pol) or equimolar combinations of thereof. Dose-response curves were generated by fitting log[inhibitor] versus activity data using a three-parameter nonlinear regression model in GraphPad Prism. IC₅₀ were derived from the fitted curves to quantify the extent of ExoU inhibition.

https://doi.org/10.1371/journal.ppat.1014474.g003

In the presence of PIP₂ (Fig 3B, lanes 5–12) BS³ crosslinking resulted in a dominant ExoU species migrating at ~300 kDa, with bands at ~400 kDa, ~ 150 kDa and ~74 kDa also present (Fig 3B, lanes 5–6).

In the presence of Bis and PIP₂ (lanes 7–8) crosslinking of ExoU with BS³ resulted in a single discrete high‐molecular‐weight band matching the Zp (-PIP₂) complex and a smear around ~74 kDa. The absence of lower‐order dimers or a smear around 300/400 kDa suggested that Bis constrained the PIP₂‐induced assembly of ExoU into one oligomeric form. By contrast, in the presence of Zp and PIP₂ (lanes 9–10) crosslinked ExoU migrated mainly as two uniform bands at ~300 kDa and >400 kDa, the latter being the more intense species and representing the bulk of crosslinked protein. Finally, in the presence of Pol and PIP₂ (lanes 11–12) the smear observed in the absence of PIP₂ migrated slower (~80–150 kDa), indicative of the presence of monomers and dimers.

Compound combination enhances in vitro ExoU inhibition

Given that Bis, Pol, and Zp inhibited ExoU through distinct mechanisms, we hypothesized that combining these compounds could enhance inhibitory potency. To test this, we performed dose-response phospholipase assays in the presence of Bis, Zp, Pol, and equimolar compound combinations (Fig 3C). Individually, Bis, Zp, and Pol inhibited ExoU activity with IC₅₀ values of 13.3 ± 2.9 µM, 1.4 ± 0.2 µM, and 11.6 ± 0.9 µM. All pairwise combinations exhibited increased potency relative to single agents, with IC₅₀ values in the sub- to low-micromolar range. Notably, the triple combination displayed markedly enhanced inhibition, yielding an IC₅₀ of 0.48 ± 0.1 µM. These findings suggest that the compounds act through complementary mechanisms and provide a rationale for evaluating combination treatments in subsequent efficacy studies.

Bis and Pol destabilise EGFP-tagged ExoU (S142A) expressed in HEK293T cells

To examine the effects of Bis, Pol and Zp on ExoU in host cells, HEK293T cells were transfected with a plasmid encoding EGFP-tagged catalytically inactive ExoU (S142A), as WT is toxic to transfected cells. After 16 h transfection, detection of GFP-S142A ExoU by Western blotting and confocal fluorescence microscopy were used to establish inhibitor effects on ExoU abundance and localisation (Fig 4).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Bismuth subcitrate and polymyxin B promote lysosome localisation and degradation of GFP-S142A ExoU in HEK293T cells.

(A) Left: HEK293T cells were transfected with pEGFP-C3, encoding S142A ExoU with an N-terminal EGFP tag, in the presence of bismuth subcitrate (Bis), polymyxin B (Pol) or zinc pyrithione (Zp) for 16 hours. Western blotting of whole cell lysates (two independent samples per group; three separate experiments) was used to detect abundance of GFP-S142A ExoU in transfected HEK293T cells, using an α-GFP antibody. Right: Confocal fluorescence microscopy was used to demonstrate plasma membrane distribution of GFP-S142A ExoU in the presence of DMSO and Zp but reduced in the presence of Bis and Pol (full z-stacks are available in supplementary). (B) HEK293T cells were co-transfected with GFP-S142A ExoU and the RFP-tagged lysosomal marker LAMP1, followed by 16 h compound treatment. Fluorescence microscopy with split-channel imaging (GFP, RFP, brightfield, and merged), together with Costes’ mask images (colocalised pixels shown in white), was used to visualise S142A ExoU-lysosome colocalization. (C) Quantification of ExoU-lysosome colocalization in the presence of DMSO, Bis Zp or Pol was performed from three independent experiments using the Coloc 2 plugin in Fiji (ImageJ) to calculate Pearson’s correlation coefficients. Data were plotted in GraphPad Prism, and statistical significance was assessed by one-way ANOVA with multiple comparisons to DMSO-treated controls. Original uncropped and unadjusted images underlying all blot and gel results are presented in S1 Raw Gel.

https://doi.org/10.1371/journal.ppat.1014474.g004

In HEK293T cells transfected with the plasmid encoding EGFP-tagged ExoU (S142A) produced consistent quantities of EGFP-tagged ExoU (S142A) (Fig 4A). The larger form was possibly ubiquitinated, consistent with previous studies, that found that ExoU undergoes modification within the cell by addition of two ubiquitin molecules at lysine-178 [33]. In transfected cells incubated with Zp, the amount of EGFP-ExoU (S142A) was similar. When the transfected cells were treated with 5 µM Bis or Pol, however, the level of EGFP-tagged ExoU (S142A) was substantially reduced (Fig 4A, left panel). Using confocal microscopy, EGFP-ExoU (S142A) fluorescence in control and Zp-treated cells was predominantly localised to the plasma membrane (Fig 4A, right panel). In contrast, treatment with Bis or Pol resulted in a marked reduction in EGFP–ExoU (S142A) fluorescence (S1 Video, S2 Video, S3 Video and S4 Video), consistent with the decreased protein abundance observed by Western blot (Fig 4A, left panel).

As we previously showed that ExoU turnover is mediated by PMSF-sensitive proteases rather than the proteasome [30], the reduction in ExoU levels suggested involvement of an alternative degradation pathway. Notably, prior studies have reported that ExoU (S142A) can traffic through endosomal and lysosomal compartments, where it accumulates in punctate structures [33, 34]. Based on these observations, we hypothesised that Bis and Pol promote trafficking of ExoU to lysosomes, leading to its degradation.

To test this, HEK293T cells were co-transfected with EGFP-ExoU (S142A) and RFP-tagged LAMP1, a lysosomal marker. In DMSO-treated cells, EGFP-ExoU (S142A) remained predominantly associated with the plasma membrane at 16 h post-transfection (Fig 4B). In contrast, Bis- or Pol-treated cells showed reduced membrane-associated fluorescence, with the remaining EGFP-ExoU (S142A) redistributed into punctate structures that co-localised with RFP-LAMP1 (Fig 4B), consistent with lysosomal trafficking.

Analysis of the images showed that Bis and Pol treatments (Pearson’s correlation coefficient r = 0.49 and r = 0.64, respectively) substantially increased colocalisation of ExoU with lysosomes compared with the DMSO control (Pearson’s correlation coefficient r = 0.05) (Fig 4C). In contrast, Zp treatment did not enhance colocalisation (Pearson’s correlation coefficient r = 0.09). Taken alongside the differences observed in the effects of each compound on the Tₘ of ExoU (Fig 3A), on the complexes observed upon BS³ crosslinking (Fig 4B) and the enhanced inhibition seen when the compounds were used together (Fig 3C), these data support the argument that the compounds inhibit ExoU through complementary mechanisms.

Efficacy of ExoU inhibitors in a high-content microscopy screen of clinical P. aeruginosa keratitis isolates

Whereas Bis, Pol, and Zp reduce ExoU-mediated cytotoxicity in cell infection models using the laboratory PA103 strain, it was important to determine if they did likewise with clinically relevant strains. To this end we employed a fully sequenced, phenotyped, and genotyped library of 52 P. aeruginosa clinical keratitis isolates from the UK Microbiology Ophthalmic group [35]. High-throughput phenotypic screening used a CellDiscoverer 7 (CD7) platform for live-cell imaging.

HCE-T cells were seeded in 96-well plates, stained with Calcein-AM and ethidium homodimer, and infected at ~80% confluence with each clinical isolate (MOI 10) in the presence of individual inhibitors (Bis, Pol or Zp) and combinations thereof (Bis + Zp, Bis + Pol, Zp + Pol and Bis + Zp + Pol) (Fig 5). Importantly, none of the compounds, alone or in combination, affected bacterial viability at the assayed concentrations (S8 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. High-content imaging screen evaluating ExoU inhibitor efficacy across clinical P. aeruginosa keratitis isolates.

A high-content microscopy screen using the Cell Discoverer 7 (CD7) platform was developed to assess the efficacy of ExoU inhibitors against PA103 and a panel of 29 exoU⁺ and 24 exoS⁺ clinical P. aeruginosa keratitis isolates. HCE-T cells were seeded into 96-well plates and cultured to ~80% confluence prior to infection at MOI 10. Cells were treated with 10 µM Bis, 2.5 µM zinc pyrithione (Zp), 1 µM polymyxin B (Pol), or the indicated combinations (Bis + Zp, Bis + Pol, Zp + Pol, and Bis + Zp + Pol). Cells were stained with calcein-AM and ethidium homodimer to quantify cell viability and cytotoxicity, enabling assessment of cell lysis (ExoU activity) or cell rounding (ExoS activity) over 4.5 h. (A-C) Mean live-cell area at 0.5, 2.5, and 4.5 h post-infection, expressed as a mean percentage across all isolates (± standard error of mean) relative to non-infected controls (live-cell area, µm²). (D) Live-cell area percentage across all conditions following infection with exoS⁺ clinical isolates, which do not induce rapid cell lysis and therefore serve as controls for ExoU-dependent cytotoxicity. (E) Quantification of cell rounding following infection with exoS⁺ clinical isolates at 4.5 h post-infection, expressed as rounded-cell area (convexity, µm²). For individual strain data, see S9 Fig. Statistical analysis was performed using one-way ANOVA with multiple-comparisons testing in GraphPad Prism, comparing each condition to infected vehicle (DMSO) controls.

https://doi.org/10.1371/journal.ppat.1014474.g005

Time-lapse imaging was performed every 0.5 h over 4.5 h. Among the 52 isolates, 28 were exoU⁺ and 24 exoS ⁺ . For nearly all exoU⁺ strains, ExoU-mediated cytolysis was evident as progressive loss of Calcein-AM fluorescence and membrane rupture (S5 Video). The sole exception was isolate 68228, which displayed fewer lysed cells than uninfected controls (S9 Fig), indicating an absence of ExoU-dependent cytotoxicity. In contrast, exoS⁺ isolates induced cell rounding while maintaining Calcein-AM signal, consistent with cytoskeletal disruption but preserved membrane integrity (S1 Video 6).

Compound efficacy in ExoU producing clinical P. aeruginosa clinical isolates

To quantify inhibitor efficacy across the exoU⁺ cohort, live-cell area was measured over time and normalised to non-infected controls to account for inter-experimental variability (Fig 5A-C).

At early timepoints (0.5 h post-infection), minimal cytotoxicity was observed, with no significant differences between treatment conditions (mean live-cell area ~90% of non-infected controls; Fig 5A). By 2.5 h, cytotoxic effects became apparent (DMSO ~69%), with significant protection observed only for the combination treatments Zp + Pol and the triple combination (Bis + Zp + Pol), both of which showed significant improvements in cell viability (Fig 5B).

At 4.5 h post-infection, pronounced ExoU-dependent cytotoxicity was evident in DMSO-treated cells (~43% live-cell area). Of the individual compounds, only Bis conferred modest protection (55.8%, P < 0.05), whereas combination treatments resulted in progressively enhanced efficacy. Zp + Pol significantly improved cell viability (67.2%, P < 0.0001), while the triple combination (Bis + Zp + Pol) provided the greatest protection (87.3%, P < 0.0001), approaching levels observed in non-infected controls (Fig 5C). The other combinations (Bis + Zp and Bis + Pol) also showed significant, but less pronounced, effects, consistent with additive or complementary activity.

To more accurately assess the breadth of inhibitor efficacy across the exoU⁺ clinical isolate panel, we analysed live-cell area at the 4.5 h endpoint for each strain under all treatment conditions (S9 Fig). Substantial heterogeneity in ExoU-mediated cytotoxicity was observed across isolates, with variable responses to individual compounds. Notably, Zp retained unique activity against isolate 68130, which was unresponsive to other single-agent treatments. For the reference strain PA103, however, Zp and Pol alone did not confer significant protection (S9 Fig), in contrast to the scratch assay (Fig 1D). This likely reflects differences in assay format, with the high-content assay using sub-confluent cells and live-cell area as a readout, versus the scratch assay measuring acute lysis in fully confluent monolayers.

Combination therapy improved both the magnitude and breadth of protection. Dual combinations (Bis + Zp and Bis + Pol) increased the proportion of responsive isolates to 60%, while Zp + Pol provided further enhancement (70%), consistent with its strong efficacy in the primary screen.

Inclusion of Bis in the triple combination (Bis + Zp + Pol) yielded the most robust and consistent protective effect, increasing responder rates to 85% of isolates. Across many strains, live-cell area approached levels observed in non-infected controls, indicating near-complete suppression of ExoU-mediated cytotoxicity. In addition, the triple combination reduced inter-isolate variability and converted several previously non-responsive strains (76230, 86028, 68130, and 68168) into responders, demonstrating expanded coverage across the clinical panel.

Despite this, a small subset of isolates remained only weakly responsive, suggesting intrinsic differences in ExoU activity, expression, or host-pathogen interactions. Nevertheless, these data demonstrate that combination therapy, particularly the triple regimen, substantially enhances both the efficacy and consistency of ExoU inhibition across genetically diverse clinical isolates.

Bis, Pol and Zp do not effectively reduce ExoS mediated cytotoxicity

Whereas the three compounds inhibited ExoU activity, it was possible that they also interfered with type III secretion. Therefore, we examined the 24 exoS⁺ isolates in the panel of clinical isolates and quantified host cell viability (Fig 5D), but more importantly, host cell convexity (Fig 5E) as a readout of ExoS-mediated cytotoxicity. ExoS promotes cell rounding through its GAP and ADP-ribosyltransferase domains, which inactivate Rho family GTPases, disrupt actin polymerisation and focal adhesion dynamics, and thereby increase cell convexity [36,37] (S2 Video).

No significant differences in live-cell area were observed across treatment conditions at 4.5 h (Fig 5D), indicating that the compounds do not broadly affect host cell survival under these conditions. Furthermore, quantification of cell rounding (convexity) demonstrated that compound treatment did not inhibit ExoS-mediated morphological changes (Fig 5E), supporting a mechanism specific to ExoU inhibition.

In addition, we evaluated whether Bis, Zp or Pol affected T3SS gene expression in P. aeruginosa PA103 and PAO1 [30]. Quantitative RT-PCR analysis (S10 Fig) revealed no significant changes in the expression of key T3SS-associated genes exsA, pcrV, and exoU in PA103, or exsA, pcrV, and exoS in PAO1, following treatment with the compounds at concentrations used in the cell-based assays. These results indicated that the compounds did not suppress the transcription of genes associated with T3SS or of the two exotoxins, further supporting that their protective effects observed in cell assays were due to direct inhibition of ExoU activity.

Evaluation of ExoU inhibitors in a porcine ex vivo corneal infection model

Our previously developed ex vivo porcine corneal infection model [38] provided a means to measure the efficacy of Bis, Zp and Pol to mitigate P. aeruginosa PA103-induced damage in a living cornea. The ring-based setup allowed sustained exposure of dissected porcine corneas to both compound and pathogen at the infection site. Infection severity was quantified through measurements of corneal opacity, epithelial ulceration, bacterial burden (CFU per cornea), and further assessed by histological examination.

Dissected porcine corneas were topically infected with 50 µL solutions of P. aeruginosa strain PA103 (1 × 10⁵ CFU) mixed and treated topically with 50 µL solutions of Bis (5 µM), Zp (2.5 µM), or Pol (1 µM), either alone or in combination. After 48 hours, monotherapies had only modest effects on reducing corneal opacity, with Bis performing slightly better than Zp (Fig 6A). Combinatorial treatments yielded progressively improved outcomes: Bis + Zp showed minor improvement over single-agent treatments, while Zp + Pol and Bis + Pol produced greater reductions in opacity. Notably, the triple combination (Bis + Zp + Pol) resulted in corneas that closely resembled uninfected controls, suggesting near-complete protection (Fig 6A).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. ExoU inhibitor combinations reduce P. aeruginosa pathogenicity in ex vivo porcine corneas.

(A) Dissected porcine corneas were infected with 1 × 10⁵ CFU of PA103, followed by topical application of ExoU inhibitors individually (Bis: 5 µM, ZP: 2.5 µM, Pol: 1 µM) or in combination. A 20 µM moxifloxacin treatment served as a positive control. Corneas were imaged after 48 hours of infection using a BioRad ChemiDoc imaging system. (B) Corneal opacity caused by ulceration 48 hours post-infection was quantified by measuring optical density at 400 nm using a spectrophotometer. (C) The ulcerated area (mm²) of infected corneas was measured at 48 hours post-infection using ImageJ. (D) At 48 hours post-infection, corneas were washed with PBS, homogenized, and serially plated onto agar plates to determine the bacterial load (CFU) within the corneal tissue. (E) Infected corneas were fixed in paraformaldehyde (PFA), paraffin-embedded, sectioned, and stained with haematoxylin and eosin (H&E). Sections were imaged using a Ventana DP 200 slide scanner. Black arrows indicate areas of epithelial erosion/ulceration; white arrows highlight regions of stromal oedema.

https://doi.org/10.1371/journal.ppat.1014474.g006

These qualitative observations were supported by quantitative measurement of corneal opacity (Fig 6B). While individual compounds modestly reduced opacity compared to DMSO-treated infected controls, combinations, particularly Bis + Pol and the Bis + Zp + Pol treatment, produced significantly greater improvement. The triple combination was comparable to 20 µM moxifloxacin, a clinically relevant positive control. Assessment of epithelial ulceration (Fig 6C), defined here as the combined loss of epithelial integrity and underlying stromal degradation following infection, mirrored these trends, mirrored these trends. Corneal damage was substantially reduced in combination treatments, with Bis + Pol and Bis + Zp + Pol showing the most dramatic reductions in ulcer size, approaching the resolution achieved with high-dose moxifloxacin. Bacterial loads within the corneas, measured by CFU (Fig 6D), were not significantly altered by compound treatment, indicating that ExoU inhibitors were not directly bactericidal under these conditions. As expected, 20 µM moxifloxacin markedly reduced bacterial burden but did not fully eradicate infection, consistent with previous reports [38].

Histological analysis of haematoxylin and eosin-stained sections (Fig 6E) corroborated these findings. Although the epithelium was initially partially disrupted during model establishment, DMSO-treated infected corneas showed complete epithelial loss, severe stromal erosion, and oedema. In contrast, Bis and Pol treated corneas showed partial epithelial preservation and less stromal damage. The triple combination conferred near-complete epithelial integrity and minimal stromal swelling, closely resembling uninfected tissue.

Together, these data demonstrated that the three ExoU inhibitors, particularly in combination, could significantly reduce the severity of P. aeruginosa PA103-induced keratitis in an ex vivo porcine model, even without directly affecting bacterial viability.

Bis, Zp and Pol enhance survival of PA103 infected Galleria mellonella larvae

To assess the therapeutic potential of ExoU inhibitors in vivo, Galleria mellonella larvae were infected with 1 × 10⁵ CFU of PA103 and treated with individual compounds or combinations thereof. Larval health was evaluated using a validated scoring system (0–9) [39], as cocoon formation was not observed within the 40 h experimental window. Mean health scores at 40 h post-infection were analysed using linear mixed-effects models with repeated measures and post hoc Tukey testing (Table 1; Fig 7A). In addition, larval health was monitored at 8 h intervals (Fig 7B), and survival was assessed by Kaplan–Meier analysis (Fig 7C).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Galleria mean ± SD clinical health scores. The health scores of Galleria larvae (5 per group) were analysed using a validated (0-9) scoring system, across 40h of infection. A linear mixed model was used with repeated measures and post hoc Tukey tests. Statistical differences between all groups are shown (p < 0.05*).

https://doi.org/10.1371/journal.ppat.1014474.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. ExoU inhibitors enhance survival of P. aeruginosa PA103-infected Galleria mellonella.

https://doi.org/10.1371/journal.ppat.1014474.g007

Galleria mellonella larvae were infected with 1 × 10⁵ CFU of P. aeruginosa PA103 and treated with ExoU inhibitors or vehicle control (PBS + 0.1% DMSO). Treatments included bismuth subcitrate (Bis; 0.1 mg/g), zinc pyrithione (Zp; 0.01 mg/g), polymyxin B (Pol; 0.004 mg/g), or combinations thereof. Data represent five independent experiments (mean of 2 larvae per condition per experiment; total n = 10 per condition). Larval health was assessed using a modified 0–9 health index scoring system (maximum score reduced from 10 due to absence of cocoon formation within the 40 h assay period). (A) Mean ± SD health scores at 40 h post-infection. (B) Longitudinal health scores (mean ± SD) recorded at 8 h intervals. Top panel: single-agent treatments; bottom panel: combination treatments. (C) Kaplan–Meier survival curves showing the proportion of surviving larvae over time for vehicle control, polymyxin B alone, and the triple combination (Bis + Zp + Pol). Statistical analysis was performed using linear mixed-effects models with repeated measures and Tukey’s post hoc test (A,B), and log-rank (Mantel–Cox) test for survival (C). Significance thresholds: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****).

PBS-injected controls maintained relatively high health scores throughout the experiment (mean 7.4 ± 2.0), with a modest decline over time likely reflecting injection-associated stress rather than infection. In contrast, PA103 infection caused rapid deterioration in larval health, resulting in complete mortality by 32 h (Fig 7A-C). Consistent with this, infection significantly reduced health scores in mock-treated larvae (PBS + 0.1% DMSO; p < 0.01), as well as in all single-agent (Bis, Zp, Pol) and dual-treatment groups (Bis + Zp, Zp + Pol, Bis + Pol; p < 0.01) when compared to uninfected controls.

Single-agent treatments provided only modest and transient benefit. Bis (0.1 mg/g) and Zp (0.01 mg/g) delayed disease progression by approximately 8 h in a subset of larvae, while Pol (0.004 mg/g) showed slightly greater efficacy, with some larvae surviving beyond 24 h; however, none of the individual treatments prevented mortality by 40 h (Fig 7B).

Dual combinations improved outcomes relative to single agents but did not fully prevent disease progression. Bis + Zp delayed larval decline, though mortality still occurred by 32 h. Zp + Pol extended survival further, with some larvae retaining partial activity up to 40 h, while Bis + Pol provided the most consistent protection among dual treatments, with a small number of larvae maintaining measurable health scores at 40 h (Fig 7B).

The triple combination (Bis + Zp + Pol) conferred the greatest benefit. Larvae maintained high health scores through 16 h, followed by a progressive decline, although this was delayed relative to all other treatment groups. Several larvae retained moderate health scores at 24 h, and one larva maintained a score of 4.5 at 40 h (Fig 7B). At the 40 h endpoint, mean health scores for the triple combination (5.8 ± 3.4) did not reach statistical significance compared to uninfected controls (p = 0.27); however, variability between larvae remained high and outcomes did not fully match baseline health. Kaplan–Meier analysis supported an overall improvement in survival relative to infection alone (Fig 7C), but not complete restoration to uninfected levels.

Together, these data indicate that combination therapy, particularly the triple regimen, improves larval health and survival in this model, although protection remains partial under these conditions.

ExoU inhibitor combinations improve clinical outcome in an in vivo mouse model of keratitis

To evaluate the therapeutic potential of ExoU inhibitors during P. aeruginosa corneal infection, we used a murine scratch and infection model [40,41]. Mice received topical treatment with individual inhibitors or inhibitor combinations at 30 min, 4 h, and 24 h post-infection. Disease severity was evaluated at 24 and 48 h using a standard 16-point clinical scoring system (Fig 8A), with representative corneal images shown at 24 and 48 h (Fig 8B). Bacterial load was quantified by CFU enumeration (Fig 8C). In parallel, flow cytometric analysis of digested corneas was performed to assess host responses, including corneal cell viability, total leukocyte infiltration, and neutrophil recruitment (Fig 8D-G).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 8. ExoU inhibitor combinations reduce disease severity and bacterial burden in a murine P. aeruginosa corneal infection model.

Corneal infections were established in female C57BL/6J mice (8–10 weeks old) by creating three parallel scratches with a 25-gauge needle, followed by topical inoculation with 1 × 10⁴ CFU of P. aeruginosa strain PA48386. After 30 min, mice received topical treatment with ExoU inhibitors Bis (50 µM), Zp (20 µM), Pol (1 µM) or PBS vehicle control (5 µL per cornea). Treatments were repeated at 4 h and 24 h post-infection. Eyes were imaged at 24 h and 48 h, and corneas collected at 48 h for analysis. (A) Clinical severity of corneal infection at 48 h, scored on a standardized 16-point scale by two independent, masked assessors. Bar graphs show severity scores (mean and ±SD). (B) representative corneal images are shown below for each treatment at 24 h and 48 h. (C) Bacterial burden (CFU) in whole corneal homogenates at 48 h. (D-G) Flow cytometry of corneal single-cell suspensions. (D) Quantification of non-immune cells (CD45⁻). (E) Proportion of dead non-immune cells (Zombie Violet⁺). (F) Quantification of neutrophils (CD45 ⁺ Ly6B⁺F4/80⁻). (G) Proportion of dead neutrophils across treatment groups. Statistical significance for D-G was assessed by one-way ANOVA with post hoc correction for multiple comparisons against mock-treated controls. *p < 0.05, **p < 0.01, ***p < 0.001.

https://doi.org/10.1371/journal.ppat.1014474.g008

Combination therapy, but not single agents, significantly reduced disease severity

Non-infected corneas exhibited minimal pathology, whereas mock-treated infected corneas showed severe disease (mean clinical score 10.53 at 24 h; Fig 8A; Table 2). Treatment with individual inhibitors (Bis, Zp, or Pol) did not significantly improve clinical scores at either 24 h or 48 h (Tables 2 and 3). In contrast, combination therapies significantly reduced disease severity at both timepoints, with Bis + Pol and Zp + Pol showing the greatest benefit. The triple combination (Bis + Zp + Pol) also improved outcomes compared to mock-treated controls, although it did not confer a significant advantage over the most effective dual combinations (Table 3).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Clinical severity scores at 24 hours post-infection in a murine P. aeruginosa corneal infection model. Female C57BL/6J mice (8–10 weeks old) were scratched and topically inoculated with P. aeruginosa strain PA48386 (1 × 10⁴ CFU per cornea). Mice received topical treatment with ExoU inhibitors Bis (50 µM), Zp (20 µM), Pol (1 µM), or combinations thereof at 30 min, 4 h, and 24 h post-infection. Control animals received PBS with 0.1% (v/v) DMSO. Corneal images were captured 24 h post-infection and graded by two masked observers using a standardized 16-point scoring system. Values represent mean ± SD (median and interquartile range (IQR) from two independent, masked assessments. Statistical analysis was performed using a General Linear Model (GLM) followed by Tukey’s post hoc test (p < 0.05). The probability values for comparison of each condition are provided. Combination treatments reduced clinical severity compared to mock treatment, with the lowest scores observed in the Bis + Pol and Zp + Pol groups.

https://doi.org/10.1371/journal.ppat.1014474.t002

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Clinical severity scores at 48 hours post-infection in a murine P. aeruginosa corneal infection model. Clinical scores were assessed 48 h post-infection in the same experimental cohort described in Table 2. Each image was independently graded twice by two masked observers (Cohen’s weighted κ = 0.74–0.84, indicating strong inter- and intra-rater agreement). Data represent mean ± SD (median and interquartile range (IQR). Statistical analysis was performed using a General Linear Model (GLM) with Tukey’s post hoc test (p < 0.05). The probability values for comparison of each condition are provided. Compared to mock-treated infected controls, combination treatments (Bis + Pol, Zp + Pol, Bis + Zp + Pol) significantly improved clinical outcome, achieving scores comparable to uninfected controls.

https://doi.org/10.1371/journal.ppat.1014474.t003

Therapeutic benefit occurred independently of bacterial clearance

Bacterial burden, quantified by CFU at 48 h, was not significantly different between treatment groups (Fig 8C), indicating that the observed improvements in clinical outcome were not due to direct antibacterial effects.

Combination therapy preserves corneal cell viability and modulates neutrophil responses

Flow cytometric analysis revealed no significant differences in total CD45⁻ cell numbers across groups (Fig 8D), indicating comparable tissue recovery. However, combination treatments (Bis + Zp, Zp + Pol, and Bis + Zp + Pol) significantly reduced the proportion of dead CD45⁻ cells relative to mock-treated controls (Fig 8E), consistent with protection of corneal epithelial cells.

Analysis of immune infiltration showed that neutrophil numbers were significantly reduced in the Bis + Pol and Zp + Pol groups (Fig 8F). In addition, the proportion of dead neutrophils was decreased in the Bis + Pol and triple combination groups (Fig 8G), suggesting improved neutrophil viability.

Ethics statement

Primary human corneal epithelial cells were derived from donor corneas obtained through the Liverpool Research Eye Bank with approval from the Central University Research Ethics Committee under the study “The Eye and its Repair” (Dr Carl Sheridan; IRAS project ID 239185). Written informed consent for research use was obtained from donors.

Porcine eyes were obtained from CS. Morphet & Sons abattoir (Widnes, UK) as by-products of routine commercial slaughter; no pigs were sacrificed specifically for this study. Work with Galleria mellonella does not currently require regulatory approval in the UK.

All mouse experiments were approved by the Cleveland Clinic Institutional Biosafety Committee (IBC protocol 1419) and the Cleveland Clinic Institutional Animal Care and Use Committee (IACUC protocol 0000–2324). Animal studies were performed in accordance with Public Health Service (PHS) Policy on Humane Care and Use of Laboratory Animals, National Institutes of Health Office of Laboratory Animal Welfare guidelines, and the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. The Cleveland Clinic animal care and use program is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC International).

ExoU inhibitors

The full FDA-approved compound library was purchased from Selleckchem, Houston, USA (3,034 compounds, Catalog No.L1300). Follow up hits for downstream analysis were also purchased separately from Selleckchem. Custom synthesised “Poltide” fatty acid peptide analogues were purchased from Biomatik, Kitchener, Ontario, Canada.

Recombinant protein production

Full-length exoU was cloned into pUCP20T to generate a construct encoding ExoU with a C-terminal 6 × His tag. Human phospholipases PLA2G7 and PLA2G4C were cloned into pET28a, each encoding an N-terminal 6 × His tag. All plasmid constructs were verified by Sanger sequencing (Eurofins Genomics, Ebersberg, Germany). For protein expression, transformed C43(DE3) E. coli were cultured in 8L Terrific broth (Melford Laboratories Ltd, Ipswich, UK) supplemented with ampicillin (for pUCP20T) (100 μg/mL) or kanamycin (for pET41a) (50 μg/mL) and grown to an optical density (OD₆₀₀) of 0.8 before induction of recombinant protein expression using 0.4 mM isopropyl-β-d-thiogalactopyranoside (IPTG). ExoU was expressed for 3 hours at 30°C [25] and Lipoprotein-associated phospholipase A₂ (PLA2G7) and cytosolic PLA₂γ, (PLA2G4C) were expressed overnight at 18 ˚C. E. coli were isolated by centrifugation (5,000 g, 10 minutes) and lysed by sonication in 20 mM Tris-HCl pH 8.2, 300 mM NaCl, 0.1% (v/v) Triton-X-100, 10 mM imidazole, 10% (v/v) glycerol and a complete protease inhibitor cocktail tablet (Roche, Welwyn Garden City, UK). Proteins in the cleared lysate were initially purified immobilised nickel affinity chromatography, followed by size-exclusion chromatography (SEC) (16/600 Superdex 200, GE Healthcare, Amersham, UK) in 20 mM Tris-HCl pH 8.2, 100 mM NaCl and 10% (v/v) glycerol. Recombinant proteins were frozen in liquid nitrogen and stored at −80°C.

Phospholipase assays

IC₅₀ Determination

Lead compounds were assayed in a 9point serial dilution series in triplicate. Realtime absorbance readings at 414 nm were recorded every minute for 60 min. For each well, the initial linear portion of the A₄₁₄ vs. time curve (5–15 min) was fitted by least squares regression to yield a slope (ΔA₄₁₄/min). Product formation rates (µmol/min) were calculated using point serial dilution series in triplicate. Real-time absorbance readings at 414 nm were recorded every minute for 60 min. For each well, the initial linear portion of the A-squares regression to yield a slope (ΔA

Rates were then normalized to ExoU mass in each well (µg) to give specific activity (µmol/min/µg). Normalized rates were plotted against the logarithm of compound concentration, and IC₅₀ values were determined by fitting a three-parameter nonlinear regression dose-response model in GraphPad Prism.

Human phospholipase assays

Recombinant human PLA₂G7 and PLA₂G4C activities were measured in 96well, -well, UVtransparent microplates (Corning) in a final volume of 50 µL per well. Reactions were carried out in 20 mM -transparent microplates (Corning) in a final volume of 50 µL per well. Reactions were carried out in 20 mM TrisHCl-HCl (pH 8.2), 100 mM NaCl, and 0.01% Tween 20, with a 1% (v/v) final DMSO concentration. For PLA₂G7, substrate cleavage was monitored using 0.6 mM 2-thio-PAF (Cambridge Bioscience Ltd), and for PLA₂G4C, 0.6 mM arachidonoyl thio-bis-glycerophosphocholine (thioBGC-BGC; Cambridge Bioscience Ltd) was employed. In both assays, 1 mM DTNB (5,5′-dithiobis-(2nitrobenzoic acid), -nitrobenzoic acid), SigmaAldrich-Aldrich) served as the chromogenic reporter. Enzyme (500 nM final) was preequilibrated with buffer at 25 °C for 5 min before initiating-equilibrated with buffer at 25 °C for 5 min before initiating the reaction by substrate addition. Absorbance at 414 nm was recorded every minute for 1 h on a Hidex plate reader.

Secondary Incucyte S3 screen to assess compound efficacy in a wound infection assay

Human corneal epithelial (HCE-T) cells (donated by Kaoru Araki-Sasaki, Japan) were seeded into ImageLock 96-well plates (Sartorius, Epsom, UK) and cultured until confluence in Dulbecco’s modified Eagle’s medium (DMEM)/F-12 medium (Gibco, Thermo Fisher Scientific, Loughborough, UK) with 10% (v/v) fetal bovine serum (FBS; Gibco). Uniform scratch wounds were generated using the IncuCyte WoundMaker (Sartorius) and plates were washed twice with warm phosphate buffered saline (PBS) to remove detached cells.

Cells were infected with P. aeruginosa strain PA103 (multiplicity of infection (MOI) 10). In some conditions, moxifloxacin (2.5 µM; 50% minimal inhibitory concentration (MIC)) was added to restrict bacterial overgrowth and extend the assay window [25,30]. Test compounds were added simultaneously at the following final concentrations: Bis 5 µM, Zp 1 µM, Pol 1 µM, cyclofenil (Cyclo) 5 µM, mivacurium chloride (MivC) 5 µM, and chlorhexidine (Chlorh) 5 µM. All compounds were diluted in DMSO, and vehicle controls (0.1% DMSO) were included in each plate.

To monitor cell viability and lysis, calcein-AM (1 µM; Invitrogen, Thermo Fisher Scientific, Loughborough, UK) and ethidium homodimer-1 (2 µM; Invitrogen) were added immediately following infection to stain live and membrane-compromised cells respectively. Plates were imaged every 30 minutes for 15 hours using the IncuCyte S3 live-cell imaging system (Sartorius) equipped with a 10x objective. Image acquisition and analysis were performed using IncuCyte S3 software. Quantitative metrics included total live cell area (calcein+) and dead cell count (ethidium+).

Each experimental condition was assayed in duplicate. Post-assay bacterial CFU enumeration was performed in parallel, by serial dilution and agar plating, to confirm that observed cytoprotective effects were not due to antimicrobial activity of the compounds at the concentrations used.

Nano Differential Scanning Fluorimetry (nanoDSF)

Thermal stability of recombinant His₆-ExoU was assessed using a Prometheus NT.48 instrument (NanoTemper Technologies, Cambridge, UK) without extrinsic dyes, as these were not compatible with the use of PIP₂. ExoU was diluted to 5 µM in 20 mM TrisHCl-HCl (pH 8.2), 100 mM NaCl, and 10% (v/v) glycerol. To evaluate ligand effects, four samples were prepared: ExoU alone, ExoU with 5 µM PIP₂, ExoU with 10 µM inhibitor, and ExoU with both 5 µM PIP₂ and 10 µM indicated inhibitor, each maintaining 1% (v/v) DMSO. Ten microliters of each mixture were loaded into highsensitivity-sensitivity nanoDSF glass capillaries and subjected to a thermal ramp from 25 °C to 95 °C at 1 °C/min. Intrinsic tryptophan fluorescence emission at 330 nm and 350 nm was recorded continuously, and the ratio F₃₅₀/F₃₃₀ was plotted against temperature. Melting temperatures (Tₘ) were determined from the inflection point of the first derivative of the fluorescence ratio using the manufacturer’s analysis software. All conditions were measured in triplicate. Changes in melting temperature (Δ Tₘ) compared to DMSO, or DMSO PIP₂ controls were plotted using GraphPad.

BS³ crosslinking and SDS-PAGE analysis of ExoU oligomerization

To assess the impact of PIP₂ and small molecule inhibitors on ExoU oligomerization, recombinant His₆ ExoU (5 µM) was first buffer‐exchanged into 50 mM HEPES pH 7.4, 100 mM NaCl, 10% (v/v) glycerol using desalting spin columns to remove Tris. Four reaction mixtures (50 µL each) were prepared at room temperature: ExoU alone; ExoU with 5 µM PIP₂; ExoU with 10 µM inhibitor; and ExoU with both 5 µM PIP₂ and 10 µM inhibitor for 30 minutes at 25 ºC. Bis(sulfosuccinimidyl) suberate (BS³) was added fresh to 1 mM and the samples were incubated for 1 h at room temperature. Reactions were quenched by addition of Tris HCl pH 7.4 to 50 mM and a further 15 min incubation at room temperature. Each sample was then mixed 1:1 with SDS PAGE loading buffer and resolved on 8% (w/v) Bis Tris SDS-PAGE gels at 200 V. Proteins were transferred to nitrocellulose membranes in Tris-glycine-methanol transfer buffer at 100 V for 1 h at 4 °C. Membranes were blocked in TBST containing 5% (w/v) non-fat milk for 1 h, then incubated overnight at 4 °C with mouse anti-His monoclonal antibody (Bio Rad; 1:2,000). After washing in TBST, membranes were incubated with HRP-conjugated goat anti-mouse IgG for 1 h at room temperature. Signal was detected by chemiluminescence with X ray film exposure, and band patterns were analysed to determine the distribution of monomeric and oligomeric ExoU species under each condition.

EGFP-ExoU (S142A) Transfections

HEK293T cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM; Gibco) supplemented with 10% (v/v) FBS, 4 mM L-glutamine, 100 U/mL penicillin and 100 µg/mL streptomycin. The gene encoding the catalytically inactive ExoU S142A was synthesized by GenScript (Piscataway, NJ, USA) and cloned in-frame into the pEGFP-C3 vector using BamHI and NotI restriction sites. Correct insertion and sequence integrity were confirmed by Sanger sequencing (Eurofins Genomics). For analysis of total EGFP-ExoU (S142A) abundance, 2.2 × 10⁶ cells were seeded into 10 cm dishes 24 h prior to transfection. Cells were transfected with 30 µg of pEGFP-C3 vector encoding an N-terminal EGFP-tagged catalytically inactive ExoU (S142A) using polyethylenimine (PEI) (Sigma-Aldrich, Gillingham, UK) at a 3:1 PEI:DNA ratio. After 24 h, cells were lysed and subjected to SDS-PAGE followed by Western blotting using an anti-GFP antibody (Cell Signaling Technology, London, UK) and HRP-conjugated anti-goat secondary antibody. GAPDH was used as a loading control. Protein detection was performed using enhanced chemiluminescence and imaged with a Bio-Rad ChemiDoc imaging system.

Confocal Microscopy of EGFP-ExoU and RFP-LAMP1 Co-Localization

HEK293T cells were seeded in 4-compartment glass-bottom imaging dishes (Ibidi, Gräfelfing, Germany) and transfected at approximately 70% confluency with constructs encoding EGFP-ExoU (S142A) (pEGFP-C3 backbone) using polyethylenimine (PEI) at a 3:1 PEI:DNA ratio, and RFP-LAMP1 (CellLight Lysosomes-RFP, BacMam 2.0, Thermo Fisher Scientific, Loughborough, UK) by modified baculovirus delivery. Two hours post-transfection, cells were treated with 0.1% (v/v) DMSO or the indicated compounds. Live-cell imaging was performed at 16 h post compound treatment using a Zeiss LSM780 confocal microscope equipped with a 63 × oil immersion objective (NA 1.4), a temperature and CO₂ controlled incubation chamber (37°C, 5% (v/v) CO₂), and laser lines at 488 nm and 561 nm for excitation of EGFP and RFP, respectively. Z-stacks were acquired at 0.5 µm intervals with 1024 × 1024 resolution and a pinhole size of 1 Airy unit.

Colocalisation between EGFP-ExoU(S142A) and lysosomes (Lysosome-RFP, BacMam) was quantified in ImageJ (Fiji) using the Coloc2 plugin. Images were background-subtracted using a 1-pixel rolling ball filter, and Pearson’s correlation coefficients were calculated between the green (ExoU) and red (lysosomal) channels. Data were obtained from three independent experiments.

Inhibitor toxicity analysis in human primary corneal cells

Donor human corneas were obtained from the Liverpool Research Eye Biobank (LREB). Corneal rims were quartered and incubated with 1.2 U/mL Dispase II (Roche) in PBS for 2 h at 37 °C. The limbal region was gently scraped using fine forceps to isolate limbal epithelial cells, which were then resuspended and triturated in corneal epithelial cell medium (CECM; DMEM:F12 supplemented with 10% (v/v) FBS, penicillin-streptomycin (100 µg/mL), epidermal growth factor (10 ng/mL), hydrocortisone (0.4 mg/mL), insulin (5 mg/mL), adenine (0.18 mM), transferrin (5 mg/mL), triiodothyronine (T3, 2 nM), and cholera toxin (0.1 nM; Sigma-Aldrich)). Cells were seeded onto mitotically inactivated 3T3-J2 feeder layers, pretreated with 4 mg/mL mitomycin C for 2 h. Cultures were maintained in CECM, with medium changes three times per week. After 12–14 days of growth, epithelial cells were passaged using 0.5% trypsin-EDTA (Invitrogen) and transferred to 96-well plates for compound cytotoxicity assays.

To assess the cytotoxicity of candidate inhibitors (Bis, Zp, and Pol), lactate dehydrogenase (LDH) release was measured in primary human limbal epithelial cells following 48 hours of compound exposure. Cells were seeded into 96-well plates at a density of 8,000 cells/well in CECM and incubated overnight at 37 °C to allow attachment. After 48 hours of treatment with compounds diluted in CECM, LDH release into the culture supernatant was quantified using the Cytotoxicity Detection Kit (Thermo Fisher) according to the manufacturer’s protocol. Absorbance was measured at 490 nm with a reference wavelength of 620 nm using a Hidex microplate reader. Background signal was subtracted from all wells, and values were normalized to the maximal LDH release control provided in the kit. Each condition was tested in duplicate wells, and the experiment was repeated independently three times (n = 3). IC₅₀ values were calculated by fitting the data to a nonlinear regression model (inhibitor vs. response, variable slope) using GraphPad Prism.

Tertiary CellDiscoverer 7 high-content microscopy screen of clinical P. aeruginosa keratitis isolates

HCE-T cells were seeded at 8,000 cells/well into clearbottom-bottom 96-well plates (Corning). Plates were incubated at 37 °C, 5% (v/v) CO₂ and monitored until cultures reached ~80% confluence (~48 h). Immediately prior to infection, DMEM F12 was replaced with DMEM F12 containing 1 µM calceinAM-AM and 2 µM ethidium homodimer-1 and incubated for 30 min at 37 °C. A panel of 52 clinical P. aeruginosa keratitis isolates (28 exoU⁺; 24 exoS⁺) was used. Bacteria were grown to midlog-log phase in LB, washed, and resuspended in PBS. Cells were infected at MOI 10 in the presence of Bis 10 µM, Zp 2.5 µM, or Pol 1 µM or indicated combinations thereof. All compounds were prepared in DMSO and added to each well at 0.1% v/v DMSO; vehicle controls received 0.1% DMSO.

Plates were transferred to a Zeiss CellDiscoverer 7 (CD7, Carl Zeiss Microscopy GmbH, Germany) equipped with environmental control (37 °C, 5% (v/v) CO₂) and imaged at 5 × magnification (0.25 NA). Three-channel images (transmitted light (oblique), Calcein AM and ethidium homodimer1) were captured automatically at 30-minute intervals over 4.5 h (Table 4). A bespoke image analysis pipeline developed by Dr. Marie Held at the University of Liverpool Centre for Cell Imaging (CCI) was used. This automated workflow, implemented in Zen Blue and Python, is freely available via Jupyter Notebook. All raw image data are deposited in the EMBL-EBI BioImage Archive: DOI: 10.6019/S-BIAD2625) [53].

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Imaging parameters for live/dead cell microscopy using the Zeiss CellDiscoverer 7: Detailed imaging settings used for automated time-lapse acquisition of transmitted light, Calcein AM, and ethidium homodimer-1 signals over 8 hours. The table lists excitation and emission filter specifications, objective and numerical aperture, detectors, and light source parameters for each imaging channel.

https://doi.org/10.1371/journal.ppat.1014474.t004

Images were pre-processed and analysed in Zen Blue (3.7.5, Zeiss, Germany) using an automated macro (Github link to Zen macro). The image was split into the fluorescence channels and the oblique channel. The fluorescence channels were processed via background subtraction (radius = 30) and then merged again with the raw oblique channel. Segmentation was performed in Zen Blue (Github link to analysis settings) using the built-in multichannel segmentation algorithm with background subtraction and thresholding (Otsu method) applied to identify regions of interest (ROIs). Objects were filtered based on size, intensity, and shape metrics to ensure accurate quantification. A minimum object area of 50 µm² and a maximum of 1000 µm² were applied to exclude small artifacts and excessively large structures. Intensity-based filtering was performed on Calcein and EtHD1 fluorescence channels, where objects with mean intensity below 500 (Calcein) or above 2000 (EtHD1) were excluded to remove low-signal noise and overexposed regions. Additional morphological constraints were applied, including convexity, circularity, and compactness thresholds, to refine object selection and remove irregularly shaped debris. Objects overlapping by more than 60% were separated using a watershed-based approach to prevent merging of distinct structures. The final dataset included extracted features such as object count, mean area, relative image area coverage, and integrated density. Image processing and measurements were conducted uniformly across all images the Zen macro which calls the Zen analysis pipeline automatically.

Downstream processing of the measured features, e.g., population average and live/dead ratio calculations were performed in batch in Python, segmented live (calcein⁺) and dead (ethidium⁺) cells and computed total livecell area (µm²) for assessment of ExoU mediated cell lysis. Cell convexity measurements were performed to assess the impact of ExoS mediated cell rounding. Each condition was assayed in duplicate with data plotted as mean ± SD of livecell area (µm²) at 5 h. Figure generation was performed in Python and GraphPad Prism.-cell area (µm²) for assessment of -cell area (µm²) at

Quantitative RT-PCR analysis of T3SS gene expression

P. aeruginosa strains PA103 and PAO1 were grown overnight in LB at 37 °C with shaking in the presence of Bis 10 µM, Zp 2.5 µM, or Pol 1 µM. EGTA (2.5 mM) was included as a positive control to induce T3SS expression.

Total RNA was extracted from cultures using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions, followed by DNase I treatment to remove residual genomic DNA. First-strand cDNA synthesis was performed using a Reverse Transcription kit (Promega). Quantitative PCR was carried out using SYBR Green Master Mix (Bio-Rad) on an Applied Biosystems StepOnePlus Real-Time PCR instrument.

As previously described, predesigned gene-specific primers were used to amplify exsA, pcrV, and exoU in PA103, and exsA, pcrV, and exoS in PAO1 [25,30]. RNA polymerase β-subunit gene (rpoB) was used as the internal reference for normalisation. Relative gene expression levels were calculated using the ΔΔCt method, and fold changes were expressed relative to vehicle-treated controls. All reactions were performed in technical triplicates from three independent cultures.

Ex vivo porcine corneal infection studies

Porcine corneas were processed and infected as previously described [38]. Briefly, porcine eyes were obtained from a local abattoir (CS Morphet & Sons, Widnes, UK) within 6 h of slaughter and transported on ice. All procedures were conducted under sterile conditions, as described [38]. Corneas with a 2–3 mm scleral rim were excised, immersed in 10% (v/v) iodinated povidone for 2 min, then rinsed twice in PBS and blotted dry. A single flange of a sterile plumbing ring (B&Q, UK: inner diameter 10 mm) was coated with LIQUIBAND topical skin adhesive and pressed centrally onto the epithelial surface.

To support the stroma and endothelium, 0.5% (w/v) UltraPure agarose in DMEM was prepared by dissolving at 65.5°C, cooling to ~37°C, and pipetting 1 mL onto the endothelial side. Once solidified, each cornea was transferred, epithelial side up, into a 12‑well plate. A defined epithelial wound was generated by applying 10 µL 70% ethanol (v/v in PBS) into the ring (corneal apex) for 10 s, then gently debriding the treated area with a scalpel blade. Corneas were rinsed three times in PBS to remove debris and residual ethanol, and wells were filled with 1 mL DMEM, ensuring the epithelium remained hydrated without submerging the ring.

PA103 was cultured to an OD₆₀₀ of 0.8 in LB, washed twice in PBS, and adjusted to 2 × 10⁶ CFU/mL. A 50 µL bacterial inoculum (1 × 10⁵ CFU) was combined with inhibitors Bis 5 µM, Zp 2.5 µM, Pol 1 µM, alone or in combination, or 20 µM moxifloxacin (positive control), all in 0.1% DMSO. Each treatment (or vehicle) was added into the ring, and corneas were incubated at 37°C and 5% CO₂ for 48 h.

Corneal opacity was measured by scanning whole mounts at 400 nm on a Hidex microplate reader and expressed as optical density (OD₄₀₀). Epithelial ulceration was quantified by photographing corneas on a Bio‑Rad ChemiDoc system and tracing wound areas in ImageJ.

For bacterial burden, rings and adhesive were removed, corneas rinsed in PBS, homogenized in 1 mL PBS using a tissue disruptor, and serial dilutions plated on LB agar. Plates were incubated overnight at 37 °C, and CFU per cornea were determined.

For histology, corneas were fixed in 10% (v/v) neutral‑buffered formalin for 24 h, dehydrated, and embedded in paraffin. Five‑micron sections were stained with hematoxylin and eosin and scanned on a VENTANA DP 200 slide scanner (Roche).

All conditions were tested in triplicate using corneas from at least three different animals. Data are presented as mean ± SD. Statistical analyses were performed by one‑way ANOVA with Tukey’s post hoc test.

Galleria mellonella infection studies

To assess the therapeutic potential of ExoU inhibitors in vivo, Galleria mellonella larvae (Pets at Home, UK) weighing 250–300 mg were used. Larvae were stored at 15 °C in the dark and acclimated to room temperature prior to experimentation. Infections were carried out using PA103, grown to mid-log phase in LB, washed twice in PBS, and adjusted to 1 × 10⁷ CFU/mL. Using a 10 μL Hamilton syringe (Hamilton 702 N, 30 gauge), larvae were injected with 10 μL of bacterial suspension (1 × 10⁵ CFU) into the last left proleg. Control larvae were injected with 10 μL PBS.

Immediately following infection, larvae were treated topically at the injection site with individual or combined ExoU inhibitors: Bis 0.1 mg/g, Zp 0.01 mg/g, or Pol 0.004 mg/g, alone or in combination. Inhibitors were prepared in PBS containing 0.1% (v/v) DMSO, and 10 μL of each treatment was applied using a micropipette directly to the injection site. DMSO-only vehicle controls were included.

Larvae were incubated in the dark and monitored at 8 h intervals over 40 h. Health was assessed every 8 hours using a validated scoring system [39] which evaluates activity, melanisation, responsiveness to touch, and cocoon formation (maximum score = 10). Since the infection period in our assays was limited to 40 h, no cocoon formation was observed in either infected or control larvae. Consequently, cocoon formation (1 point) was excluded from the scoring, and the maximum achievable score in our experiments was 9. In parallel, survival was recorded for Kaplan-Meier survival analysis [54].

Clinical score data were analysed using a linear mixed-effects model with repeated measures and Tukey’s post hoc test for multiple comparisons. Survival differences were evaluated using the Kaplan-Meier method with log-rank testing.

In vivo mouse keratitis studies

A previously established scratch infection model was employed in C57BL/6J mice [40]. Female mice (10–12 weeks old; Jackson Laboratory) were anesthetised via intraperitoneal injection (50 μL/25g body weight) of ketamine (50 mg/kg body weight) and dexmedetomidine (0.375 mg/kg body weight), followed by atipamezole (3.75 mg/kg body weight) for reversal. Ethiqa XR (extended-release buprenorphine) was administered to alleviate potential discomfort associated with infection.

A clinical P. aeruginosa keratitis isolate strain, PA48386, was cultured in LB broth overnight and then subcultured and grown to log phase (OD₆₀₀ 0.8). PA48386 cells were collected by centrifugation (5,000 g for 5 min) and resuspended in PBS to yield 2 × 10⁶ CFU/mL suspension. On anesthetised C57BL/6J mice corneal infections were induced by creating three parallel epithelial scratches in the central cornea of one eye only using a 25-gauge needle, followed by topical application of 5 μL PA48386 in PBS (inoculum containing 1 × 10⁴ CFU). After 30 minutes, mice were treated topically with 5 μL of individual ExoU inhibitors or combinations: Bis 50 μM, Zp 20 μM, Pol 1 μM, or vehicle control (PBS with 0.1% v/v DMSO). Treatments were repeated at 4 h and 24 h post-infection.

Corneal images were acquired at 24 h and 48 h post-infection using a stereomicroscope (Amscope SM-2TZ) equipped with a 10 MP Aptina MT9J003 CMOS digital camera under anaesthesia, after which animals were euthanised for downstream analyses. A total of 564 images were collected across both timepoints.

Corneal disease severity was scored independently by two masked observers on two separate occasions using a previously described extended 16-point grading system [40,41,55]. This composite system evaluates four parameters: (i) area of opacity, (ii) density of central opacity, (iii) density of peripheral opacity, and (iv) epithelial surface integrity. Individual parameter scores were summed to generate a total score ranging from 0 (no infection) to 16 (severe infection).

Inter- and intra-observer reliability were assessed using Cohen’s weighted kappa (SPSS v25), demonstrating high reproducibility (intra-rater κ = 0.77–0.84; inter-rater κ = 0.74–0.78).

Statistical analysis was performed using a general linear model (GLM), with clinical score at 24 h (Table 2) and 48 h (Table 3) as dependent variables and treatment group as the independent factor. Post hoc comparisons were performed using Tukey’s multiple comparison test, with significance defined as p < 0.05.

Flow cytometry

Mouse corneas were processed for flow cytometry analysis and CFU enumeration as previously described [41]. Mouse corneas were dissected and digested in collagenase type I (82 U/cornea in 150 μL PBS; Millipore Sigma) at 37 °C for 2 hours. A portion of the digest (5 μL) was serially diluted and plated on LB agar for CFU enumeration. The remaining suspension was processed for flow cytometry to assess immune cell infiltration and viability. Cells were washed and incubated on ice for 10 min in 100 μl ice cold FACS buffer (PBS with 1% FBS) containing 2 μg Fc blocker (anti-mouse CD16/CD32 antibody clone 93; Biolegend). Cell surface staining was conducted on ice for 1 h with APC/Cy7 anti-mouse CD45 (clone 30-F11; Biolegend), PE anti-mouse neutrophil Ly6B.2 (clone 7/4; Abcam), PE/Cy5 anti-mouse F4/80 (clone BM8; Biolegend) antibodies. All samples were also stained with Zombie Violet fixable dye (Biolegend) for dead cells. Stained cells were washed twice with FACS buffer and resuspended in 1% PFA, then detected by BD LSRFortessa flow cytometer. Cells were gated by forward and side scatter, viability, CD45⁺ population, and Ly6B.2⁺ and F4/80⁺ subpopulations. Unstained samples and fluorescence minus one (FMO) control were used to set boundaries for background and positive populations. Analysis of flow cytometric data was performed using BD FlowJo software (v.10). Neutrophils were defined as CD45⁺Ly6B.2⁺F4/80⁻, and non-immune cells as CD45⁻.

Statistics

Results were obtained from at least three independent experiments unless otherwise stated and are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism and SPSS (version 31). Where appropriate, one-way ANOVA was used to compare multiple groups, followed by Tukey’s post hoc test with correction for multiple comparisons. Significance was defined as * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.