In vitro Lung Epithelial Cell Model Reveals Novel Roles for Pseudomonas aeruginosa Siderophores

Abstract

Multidrug-resistant Pseudomonas aeruginosa is a common nosocomial respiratory pathogen that continues to threaten the lives of patients with mechanical ventilation in intensive care units and those with underlying comorbidities such as cystic fibrosis or chronic obstructive pulmonary disease. For over 20 years, studies have repeatedly demonstrated that the major siderophore pyoverdine is an important virulence factor for P. aeruginosa in invertebrate and mammalian hosts in vivo. Despite its physiological significance, an in vitro, mammalian cell culture model to characterize the impact and molecular mechanism of pyoverdine-mediated virulence has only been developed very recently. In this study, we adapt a previously-established, murine macrophage-based model for human bronchial epithelial cells (16HBE). We demonstrate that conditioned medium from P. aeruginosa induced rapid 16HBE cell death through the pyoverdine-dependent secretion of cytotoxic rhamnolipids. Genetic or chemical disruption of pyoverdine biosynthesis decreased rhamnolipid production and mitigated cell death. Consistent with these observations, chemical depletion of lipid factors or genetic disruption of rhamnolipid biosynthesis was sufficient to abrogate conditioned medium toxicity. Furthermore, we also examine the effects of purified pyoverdine exposure on 16HBE cells. While pyoverdine accumulated within cells, the siderophore was largely sequestered within early endosomes, showing minimal cytotoxicity. More membrane-permeable iron chelators, such as the siderophore pyochelin, decreased epithelial cell viability and upregulated several proinflammatory genes. However, pyoverdine potentiated these iron chelators in activating proinflammatory pathways. Altogether, these findings suggest that the siderophores pyoverdine and pyochelin play distinct roles in virulence during acute P. aeruginosa lung infection.

Introduction:

Multidrug-resistant Pseudomonas aeruginosa is one of the most common Gram-negative, respiratory pathogens, and infects mechanically-ventilated patients in intensive care units or those with cystic fibrosis (CF) or chronic obstructive pulmonary disease (COPD) (1–5). This pathogen’s intrinsic resistance to several classes of antibiotics and exceptional ability to form biofilms on medical devices and airway tissue pose a serious challenge for medical intervention (6, 7). In addition to colonizing the respiratory tract, P. aeruginosa actively deploys numerous virulence factors and toxins that damage host tissue, affecting pulmonary function (8). Two of the major virulence factors produced by this pathogen are the siderophores pyoverdine and pyochelin.

Several have proposed possible mechanisms of siderophore-dependent virulence during P. aeruginosa lung infection (9–12). As siderophores, both pyoverdine and pyochelin scavenge ferric iron and provide the pathogen with this essential micronutrient during infection. However, pyoverdine exhibits orders of magnitude higher affinity for ferric iron and is distinctly able to chelate the metal from host ferroproteins such as transferrin and lactoferrin (13, 14). Generally, iron acquisition serves an important function during infection by promoting bacterial growth and biofilm formation (15, 16), and P. aeruginosa mutants lacking various iron uptake systems exhibit attenuation of virulence during murine lung infection (17). It is important to note however that these iron uptake systems do not contribute equally, and of the two siderophores, pyoverdine plays a greater role in P. aeruginosa virulence (17).

Pyoverdine-mediated iron uptake further promotes P. aeruginosa virulence by derepressing the alternative sigma factor PvdS, which activates the transcription of several virulence genes such as those encoding the translational inhibitor exotoxin A, the exoprotease PrpL (protease IV), and pyoverdine biosynthetic enzymes (12, 18). Furthermore, we have recently used a Caenorhabditis elegans nematode model to demonstrate that pyoverdine likely directly chelates host iron, disrupting mitochondrial homeostasis (19–21). Pyoverdine’s well documented role in acute lung infection is likely mediated by a combination of these various pathogenic functions (17, 22–25).

Recently, we established the first-reported in vitro cell culture model for pyoverdine-dependent virulence, where murine macrophages were treated with conditioned medium from P. aeruginosa grown in serum-free cell culture medium (26). Under these conditions, P. aeruginosa exhibited robust pyoverdine production, yet the siderophore was not required for bacterial growth (Fig. 1A, B; Fig. S1A, B), allowing for the study of pyoverdine’s role in virulence. This pyoverdine-rich conditioned medium from wild-type P. aeruginosa PAO1 was cytotoxic towards murine macrophages, including murine alveolar macrophages (Fig. S1C); in clinical isolates; pyoverdine content in the conditioned medium positively correlated with cytotoxicity (26).

Fig. 1. Pyoverdine-rich conditioned medium kills 16HBE cells and damages the epithelial monolayer.

(A, B) Bacterial growth (A) or pyoverdine production (B) by WT P. aeruginosa PAO1 or pyoverdine biosynthetic mutant (PAO1ΔpvdF) in serum-free EMEM. (C) Fluorescent micrographs of 16HBE human bronchial epithelial cells after 30 min exposure to conditioned medium from WT PAO1 or PAO1ΔpvdF grown in EMEM. Cells were prelabeled with CellMask Orange plasma membrane stain. (D) Quantification of percentage micrograph area covered by fluorescent cells. (E) Fluorescent micrographs of 16HBE cells after 15 min exposure to conditioned medium from WT PAO1 or PAO1ΔpvdF in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (F) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342. (G) Fluorescent micrographs of 16HBE cells after 15 min exposure to conditioned medium from WT PAO1 in the presence of Sytox Orange nucleic acid stain (red). Conditioned medium was pretreated with 200 μM Ga(NO₃)₃, 200 μM FeCl₃, or 100 μg/mL proteinase K for 24 h, had macromolecules depleted via a 10 kDa centrifugal filter, or had lipids depleted by chloroform (CHCl₃) extraction. Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (H) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342. (I) Quantification of lipids by FM 1–43 fluorescent labeling in conditioned medium from WT PAO1 (control) or macromolecule- (<10 kDa) or lipid- (CHCl₃) depleted material. (J) Quantification of lipids in conditioned medium from WT PAO1 or PAO1ΔpvdF by FM 1–43 labeling. All error bars represent SEM from at least three biological replicates. * corresponds to p < 0.01 and NS corresponds to p > 0.05 based on Student’s t-test (D, F, J) or one-way ANOVA with Dunnett’s multiple comparisons test (H, I).

In this report, we adapted this in vitro pyoverdine virulence model for human bronchial epithelial cells to examine the consequences of pyoverdine production (i.e. exposure to pyoverdine and/or pyoverdine-regulated virulence factors) during P. aeruginosa lung infection. Conditioned medium from P. aeruginosa caused acute cell death and severe damage to the epithelial monolayer in a pyoverdine-, but not pyochelin-, dependent manner. Interestingly, this damage did not require host iron chelation nor production of the two known pyoverdine-regulated toxins, exotoxin A or PrpL. Instead, pyoverdine production promoted the secretion of cytotoxic rhamnolipids that have previously been shown to permeabilize host membranes (27). Consistent with this observation, chemical depletion of lipids or genetic disruption of rhamnolipid production was sufficient to abrogate toxicity from conditioned medium on 16HBE cells. Importantly, the pyoverdine biosynthetic inhibitor 5-fluorocytosine effectively inhibited rhamnolipid production and mitigated P. aeruginosa virulence in two highly virulent clinical isolates. We also examined the effects of exposing 16HBE cells to purified pyoverdine. While pyoverdine accumulated within cells, the siderophore was largely sequestered within early endosomes, showing minimal cytotoxicity. More membrane-permeable iron chelators such as the siderophore pyochelin, decreased epithelial cell viability and upregulated several proinflammatory pathways. However, pyoverdine potentiated these iron chelators in activating proinflammatory pathways. Altogether, these findings suggest that the siderophores pyoverdine and pyochelin play distinct roles in virulence during acute P. aeruginosa lung infections.

Results:

Pyoverdine-rich, conditioned medium induces rapid cell death and damages the epithelial monolayer

To investigate the role of pyoverdine production during P. aeruginosa lung infection, we treated human bronchial epithelial cells (16HBE) with bacteria-free, pyoverdine-rich conditioned medium from P. aeruginosa PAO1 grown in serum-free cell growth medium (Eagle’s Minimum Essential Medium or EMEM). To visualize the integrity of the epithelial monolayer, we prelabeled cells with a plasma membrane stain. Within 30 min, the conditioned medium severely damaged the monolayer, causing detachment of more than half of the cells (Fig. 1C, D). This disruption was significantly attenuated in 16HBE cells treated with identically prepared material from an isogenic pyoverdine biosynthetic mutant (PAO1ΔpvdF) but not a pyochelin mutant (PAO1ΔpchBA) (Fig. S1D, E). Preventing the biosynthesis of both pyoverdine and pyochelin (PAO1ΔpvdFΔpchBA) conferred no further protection to 16HBE cells than the disruption of pyoverdine alone (Fig. S1D, E). To determine whether cell detachment was caused by cell death (rather than from degradation of the extracellular matrix via bacterial proteases and other factors), we labeled cells with cell-permeant (Hoechst 33342; labels all cells) and cell-impermeant (Sytox Orange; labels only dead cells) nucleic acid stains. Exposure to pyoverdine-rich, conditioned medium from wild-type cells caused rapid (within 15 min) membrane permeabilization and internalization of the cell-impermeant nucleic acid stain, suggesting cell death (Fig. 1E, F). In contrast, conditioned medium from the pyoverdine mutant exhibited substantially less cytotoxicity (Fig. 1E, F).

One key difference between pyoverdine and pyochelin is their affinity for ferric iron. Due to an exceptionally high affinity for the metal, pyoverdine is uniquely able to remove iron from host ferroproteins (13, 14) and induce a lethal hypoxic response in a C. elegans nematode model (19). We thus examined whether host iron chelation was important for the cell death we observed in the epithelial monolayer. To hinder pyoverdine’s ability to bind iron, we pretreated the conditioned media with gallium (Ga³⁺) or ferric iron (Fe³⁺), either of which would prevent pyoverdine from scavenging iron from the epithelial cells. Surprisingly, even with the addition of excess metal, there was no significant rescue (Fig. 1G, H; Fig. S2). Interestingly, we observed that removing material with a molecular mass greater than 10 kDa (via centrifugal filtration) prevented damage to the epithelial monolayer (Fig. 1G, H, Fig. S2), suggesting that macromolecules or molecular complexes were responsible for conditioned medium cytotoxicity. To investigate the role of proteinaceous toxins, we pre-treated the conditioned medium with proteinase K. Degradation of secreted proteins did not significantly attenuate conditioned medium cytotoxicity (Fig. 1G, H). We could not determine whether proteinase K treatment mitigated cell detachment; treatment alone caused considerable damage to the extracellular matrix, making it very difficult to unambiguously assess contributions (Fig. S2). Consistent with this observation, genetic disruption of the two pyoverdine-regulated toxins, exotoxin A and PrpL, or the type II secretion system through which these toxins are secreted, did not affect conditioned medium toxicity (Fig. S3). We did however observe that removing lipids from the conditioned medium by chloroform extraction abrogated cytotoxicity and prevented damage to the epithelial monolayer (Fig. 1G, H; Fig. S2). Based on these results, we directly measured lipid content in the conditioned medium using the lipophilic dye FM 1–43. This probe is nonfluorescent in aqueous solution but is highly-fluorescent upon binding to lipid membranes (28). Chloroform extraction and removal of macromolecules (>10 kDa) significantly depleted lipid content (Fig. 1I). Importantly, conditioned medium from the pyoverdine biosynthetic mutant also contained significantly less lipid material than that from wild-type PAO1 (Fig. 1J).

Pyoverdine regulates the production of cytotoxic rhamnolipids

Based on previous studies (27, 29), we posited that the relevant secreted lipid factors were rhamnolipids. We recently demonstrated that P. aeruginosa secretes rhamnolipids that rapidly induce membrane rupture and permeabilization in a wide range of host cells, including murine macrophages, human bronchial epithelial cells, and erythrocytes (27). Chloroform extraction and centrifugal filtration, two treatments that would likely remove these rhamnolipids from conditioned medium, decreased cytotoxicity (Fig. 1G, H). To test this hypothesis, we used the rhamnolipid biosynthetic mutant MPAO1rhlA to measure the lipid content of conditioned medium. This mutant lacked lipid content in its conditioned medium but showed no apparent defects in bacterial growth or pyoverdine production compared to the control strain MPAO1cat (transposon inserted in an extraneous gene encoding a chloramphenicol acetyltransferase, Fig. 2A–C), suggesting that much of the lipid material being detected by FM 1–43 was due to rhamnolipid production. Importantly, the conditioned medium from MPAO1rhlA neither damaged the 16HBE epithelial monolayer nor induced cell death (Fig. 2D–G). We observed similar results (i.e., reduced lipid content and lower toxicity) for mutants with transposons inserted into genes encoding the RhlRI quorum sensing system (MPAO1rhlR, MPAO1rhlI) (Fig. 2A–G), indicating that rhamnolipid production is primarily regulated via this signaling pathway in EMEM (30).

Fig. 2. Pyoverdine regulates the production of rhamnolipids.

(A, B) Bacterial growth (A) or pyoverdine production (B) by MPAO1 transposon mutants in serum-free EMEM. (C) Quantification of lipids in conditioned medium from MPAO1 transposon mutants by FM 1–43 fluorescent labeling. (D) Fluorescent micrographs of 16HBE cells after 30 min exposure to conditioned medium from MPAO1 transposon mutants. Cells were prelabeled with CellMask Orange plasma membrane stain. (E) Fluorescent micrographs of 16HBE cells after 15 min exposure to conditioned medium from MPAO1 transposon mutants in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (F) Quantification of percentage micrograph area covered by fluorescent cells in (D). (G) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in (E). (H) Bioluminescence produced by E. coli JM109 pSB536 (N-butanoyl-L-homoserine lactone reporter strain) grown in media supplemented with conditioned medium from MPAO1 transposon mutants. (I) Quantification of lipids in EMEM with purified rhamnolipids (RHL) by FM 1–43 labeling. (J) Fluorescent micrographs of 16HBE cells after 15 min exposure to purified rhamnolipids from P. aeruginosa in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (K) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in (J). All error bars represent SEM from three biological replicates. * corresponds to p < 0.01, # corresponds to p < 0.05, and NS corresponds to p > 0.05 based on one-way ANOVA with Dunnett’s multiple comparisons test (C, F, G) or Tukey’s multiple comparisons test (H) – see Fig. S4 or Student’s t-test (K).

To test whether quorum-sensing is impaired in pyoverdine biosynthetic mutants, we took advantage of an Escherichia coli-based bioluminescent reporter that responds to extracellular quorum-sensing molecules (31), specifically N-butanoyl-L-homoserine lactone (C4-HSL) that is produced by the RhlRI system (32). Using this reporter, we quantified C4-HSL concentrations in the conditioned medium of pyoverdine mutants but did not observe significant difference in C4-HSL production between the mutants and their pyoverdine-producing counterparts (Fig. 2H; Fig. S4). For the rhlI C4-HSL biosynthetic mutant, the bioluminescent output in the conditioned medium was comparable to that of media control (Fig. S4), indicating that the reporter was selectively responding to extracellular C4-HSL. These results imply that pyoverdine regulates rhamnolipid production through an alternative pathway.

Furthermore, we examined the effects of exposing 16HBE cells to purified rhamnolipids (27). At concentrations comparable to those seen in conditioned medium from wild-type PAO1 or MPAO1cat (Fig. 2I; Fig. S5A), purified rhamnolipids were sufficient to kill 16HBE cells (Fig. 2J, K; Fig. S5C, E). However, cells treated with purified rhamnolipids displayed less detachment (Fig. S5B, D), suggesting that other secreted factors (e.g., proteases) in the conditioned medium were required for full damage to the epithelial monolayer. Consistent with these findings, heat denaturation of the conditioned medium did not affect rhamnolipid content (Fig. S5F) or cytotoxicity of the material (Fig. S5H, J), but it did attenuate cell detachment from the monolayer (Fig. S5G, I). Since the type II secretion system is responsible for the secretion of several P. aeruginosa protein toxins and the majority of secreted proteases (Fig. S3E) (33), we hypothesized that this system would contribute to cell detachment. We treated 16HBE cells with the conditioned medium from a mutant (MPAO1xcpQ) lacking the outer membrane transporter for these secreted proteins but saw no significant attenuation in cell detachment (Fig. S3A, C). It is important to note that while protease production in the type II secretion mutant was substantially impaired, it was not abolished (Fig. 3E). If extracellular proteases were involved in epithelial damage, they could be secreted through other means.

Fig. 3. mexEF overexpression mutant exhibits decreased rhamnolipid production.

(A) Fluorescent micrographs of 16HBE cells after 30 min exposure to conditioned medium from plcH transposon mutants. Cells were prelabeled with CellMask Orange plasma membrane stain. (B) Quantification of percentage micrograph area covered by fluorescent cells. (C) Fluorescent micrographs of 16HBE cells after 15 min exposure to conditioned medium from plcH transposon mutants in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (D) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in. (E) Alignment of MexT protein sequences from MPAO1plcH (PW2536), MPAO1plcH (PW2537), MPAO1pvdF, PAO1, or PAO1ΔpvdF. MexT sequences were acquired from paired-end Illumina sequencing (PW2536, PW2537, PAO1, PAO1ΔpvdF) or PCR-Sanger sequencing (MPAO1pvdF). Amino acid substitution sites are highlighted yellow. Amino acid substitutions in PW2536 are labeled red. (F) mexE and mexF mRNA levels compared to that of WT PAO1. mRNA levels were measured by qRT-PCR. (G) Rhamnolipid content in conditioned medium from 14 h EMEM cultures measured by FM 1–43 labeling. All error bars represent SEM from three biological replicates. * corresponds to p < 0.01 based on Student’s t-test (B, D) or one-way ANOVA with Tukey’s multiple comparisons test (G).

Pyoverdine mutants do not exhibit changes in mexEF expression

While investigating possible secreted toxins that may facilitate rhamnolipid-mediated epithelial damage, we serendipitously identified two presumably isogenic MPAO1 mutants with transposon insertions in plcH, a gene that encodes a hemolytic phospholipase C: PW2536 and PW2537. Sanger sequencing verified the presence of the transposon in the plcH gene in each of the strains. (PW2536 – 1711/2193nt; PW2537 – 1049/2193nt;). Despite this, incongruous results were obtained when conditioned medium was tested for cytotoxicity against 16HBE cells; conditioned medium from PW2536 showed reduced toxicity, while identically-prepared medium from PW2537 did not (Fig. 3A–D). To address this discrepancy, each strain was subjected to whole-genome sequencing, which revealed two monogenic single nucleotide polymorphisms (SNPs) between the two strains. This caused two amino acid substitutions in the transcriptional regulator MexT in PW2536. MexT is known to regulate the mexEF-oprN multidrug efflux pump operon (Fig. 3E) (34). The mutations in PW2536 led to the overexpression of mexEF (Fig. 3F), a phenotype previously associated with decreased rhamnolipid production (Fig. 3G) (35, 36).

Based on these observations, we tested whether pyoverdine biosynthetic mutants with poor rhamnolipid production (Fig. 3G) had background mutations in MexT or increased mexEF transcription. Neither the MPAO1 transposon mutant (MPAO1pvdF) nor the PAO1 in-frame deletion mutant (PAO1ΔpvdF) had altered MexT protein sequence or changes in mexEF expression (Fig. 3E, F), suggesting that pyoverdine regulates rhamnolipid production through an unknown mexEF- and C4-HSL- (Fig. S4) independent pathway.

5-Fluorocytosine inhibits pyoverdine and rhamnolipid production in highly virulent P. aeruginosa clinical isolates

Since genetic disruption of pyoverdine biosynthesis decreased rhamnolipid production, we hypothesized the same could be accomplished using a chemical inhibitor. To that end, we tested whether the pyoverdine biosynthetic inhibitor and FDA-approved antimycotic drug 5-fluorocytosine (5-FC) (23, 37, 38) could inhibit rhamnolipid production in several P. aeruginosa strains, including PAO1 and two clinical strains isolated from pediatric CF patients, PA2–72 and PA2–61. These two strains were selected from a large collection of CF isolates for their high in vitro pyoverdine production and virulence against the nematode host C. elegans (Fig. 4A, B) (22). These isolates also exhibited substantial in vivo pyoverdine production during acute murine lung infection where they caused host mortality (22). When these strains were grown in EMEM, 5-FC significantly impaired pyoverdine and rhamnolipid production without overtly affecting bacterial growth (Fig. 4C–F) with the exception of PA2–61, where the drug induced planktonic cell aggregation (Fig. 4D), confounding bacterial growth measurement by optical density.

Fig. 4. 5-Fluorocytosine (5-FC) inhibits pyoverdine and rhamnolipid production in highly virulent cystic fibrosis isolates.

(A, B) Pyoverdine production (A) or C. elegans death (B) by multidrug-resistant P. aeruginosa strains isolated from pediatric cystic fibrosis patients. Black bar represents the median pyoverdine production or C. elegans death. Red bar represents PA2–72, blue bar represents PA2–61. Survey data was adapted from (22). (C) Bacterial growth of PAO1, PA2–72, or PA2–61 in EMEM with or without 100 μM 5-FC. (D) Photograph of EMEM culture after 18 h incubation. (E, F) Pyoverdine (E) or rhamnolipid (F) production of PAO1, PA2–72, or PA2–61 in EMEM with or without 100 μM 5-FC. (G) Fluorescent micrographs of 16HBE cells after 40 min (PAO1, PA2–72) or 20 min (PA2–61) exposure to EMEM conditioned medium. Cells were prelabeled with CellMask Orange plasma membrane stain. (H) Fluorescent micrographs of 16HBE cells after 20 min (PAO1, PA2–72) or 10 min (PA2–61) exposure to EMEM conditioned medium in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (I) Quantification of percentage micrograph area covered by fluorescent cells in (G). (J) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in (H). All error bars represent SEM from four biological replicates. * corresponds to p < 0.01 and NS corresponds to p > 0.05 based on one-way ANOVA with Sidak’s multiple comparisons test.

5-FC significantly attenuated 16HBE cell detachment and death after exposure to conditioned medium from each of the three strains (Fig. 4G–J). These findings are consistent with previous work where the inhibition of pyoverdine biosynthesis by 5-FC was sufficient to rescue invertebrate and mammalian hosts from P. aeruginosa virulence (22, 23, 39).

Pyoverdine translocates into 16HBE cells but is sequestered within early endosomes

Next, we wanted to investigate the consequences of exposing 16HBE cells to pyoverdine in the absence of other virulence factors. In brief, pyoverdine-rich bacterial filtrate was subjected to two purification steps to separate small molecules by polarity: absorption and elution from a nonpolar polymeric resin (Amberlite XAD-4) and high-performance liquid chromatography (HPLC) via a C-18 reverse-phase column (Fig. 5A–C). We tested whether this purified material was toxic to 16HBE cells using a resazurin-based cell viability assay and compared its toxicity to other known iron-chelating molecules, namely the ferric iron chelator ciclopirox olamine, the ferrous iron chelator 1,10-phenanthroline, or the siderophores pyochelin (from P. aeruginosa) or deferoxamine (from Streptomyces spp.). While all other iron chelators exhibited time- and dose-dependent cytotoxicity towards 16HBE cells, pyoverdine remained largely nontoxic (Fig. 5D) even after 72 h treatment at 200 μM (Fig. S6A, B).

Fig. 5. Pyoverdine translocates into 16HBE cells and localizes to early endosomes.

(A) Summary of the pyoverdine purification pipeline. (B) Representative chromatogram from the HPLC purification step of the pipeline. Red box depicts the predominant pyoverdine-containing fractions that were collected. (C) Analysis of the final purified product via HPLC. (D) 16HBE cell viability after 48 h treatment with 100 μM ciclopirox olamine, 1, 10-phenanthroline, or pyochelin, or 72 h treatment of 100 μM pyoverdine or deferoxamine in serum-free EMEM. (E) Confocal micrographs of 16HBE cells exposed to 100 μM purified pyoverdine, pyoverdine with excess Ga(NO₃)_3, pyoverdine with excess FeCl₃, or media control for 24 h. Cells were trypsinated prior to imaging. (F) Quantification of pyoverdine fluorescence within 30 individual cells. (G) Confocal micrographs of 16HBE cells treated with 100 μM pyoverdine-gallium and dextran-Texas Red (10,000 MW). Error bars in (D) represent SEM from three biological replicates. Error bars in (C) represent standard deviation. * corresponds to p < 0.01 based on one-way ANOVA with Dunnett’s multiple comparisons test.

Because pyoverdine is considerably larger than these iron chelators, with a molecular weight of ~1,365 g/mol, it may be unable to translocate across cellular membranes. We took advantage of pyoverdine’s intrinsic spectral properties to examine whether pyoverdine could translocate into 16HBE cells. After 24 h, 16HBE cells treated with purified pyoverdine showed considerable internalization of pyoverdine (Fig. 5E, F). Consistent with previous studies, this intracellular fluorescence was enhanced when pyoverdine was presaturated with gallium and was quenched when pyoverdine was presaturated with iron (Fig. 5E, F) (21, 26). Importantly, pyoverdine fluorescence did not colocalize with that of the plasma membrane stain, indicating that pyoverdine was within the cell (Fig. S6C). However, we also observed that pyoverdine fluorescence formed distinct punctae within the cell. Based on previous observations in murine macrophages (26), we hypothesized that pyoverdine was sequestered within early endosomes. Supporting this hypothesis, pyoverdine colocalized with fluorophore-conjugated 10 kDa dextran, a well-established endosomal marker (Fig. 5G, Fig. S6D) (40). These results are consistent with our observations that pyoverdine, unlike other iron chelating molecules, exhibited low cytotoxicity towards 16HBE cells.

Iron chelation activates a proinflammatory response in 16HBE cells

While pyochelin exhibits lower affinity towards ferric iron than pyoverdine, it is also substantially smaller, with a molecular weight of ~325 g/mol. We hypothesized that pyochelin may be able to enter 16HBE cells and chelate intracellular iron. While we were not able to visualize pyochelin within cells (due to its lack of distinct spectral properties), one likely consequence of iron deprivation in epithelial cells would be a proinflammatory transcriptional response. Several studies have demonstrated that iron chelation by various siderophores such as deferoxamine or enterobactin promotes the production of proinflammatory cytokines, most notably interleukin (IL)-8 in lung epithelial cells, intestinal epithelial cells, or oral keratinocytes (41–43). To reaffirm these findings, we treated 16HBE cells with various iron chelators and measured the mRNA levels of genes involved in neutrophilic inflammation. We first observed that total RNA yield (from phenol-chloroform extraction) in these cells corresponded with the resazurin-based cell viability assay (Fig. 5D). Cells treated with small molecule (< 1,000 MW), cytotoxic iron chelators (ciclopirox olamine, phenanthroline, pyochelin, or deferoxamine) yielded lower quantities of RNA, while cells treated with pyoverdine had RNA quantities comparable to that of media control (Fig. 6A). Using qRT-PCR, we measured the expression of genes encoding components of the inflammasome, namely NLRP3 and NLRP1, and those encoding the major proinflammatory cytokines produced by lung epithelial cells, specifically IL-1β (IL1B), IL-8 (IL8), and tumor necrosis factor alpha (TNF). All of these genes have been associated with inflammation during lung infection (44). With the exception of pyoverdine, all iron chelators induced the expression of these proinflammatory genes (Fig. 6B). For IL8, we validated the qRT-PCR results by ELISA to determine whether transcriptional activation led to increased cytokine production. Cells treated with cytotoxic iron chelators exhibited time-dependent increase in IL-8 secretion that correlated to IL8 mRNA levels (Fig. 6C, Fig. S7). Cells treated with pyoverdine exhibited IL-8 secretion comparable to that of media control (Fig. 6C). To ensure that the observed proinflammatory response was due to iron chelation, we presaturated pyochelin and deferoxamine with excess gallium (1:2 stoichiometric ratio) prior to exposure. Cells treated with gallium-bound pyochelin or deferoxamine yielded RNA quantities comparable to that of media control, suggesting that gallium inhibited the cytotoxic effects of the siderophores (Fig. 6D). Furthermore, pretreating the siderophores with gallium resulted in a significant decrease in proinflammatory gene expression (Fig. 6E), demonstrating that the siderophore-induced inflammatory response was due to iron chelation rather than other nonspecific reactions or contaminants in the commercially-sourced material.

Fig. 6. Small molecule iron chelators promote the expression of proinflammatory genes in 16HBE cells.

(A) Total RNA yield in 16HBE cells treated with ciclopirox olamine (CPX), 1,10phenanthroline (PHE), or pyochelin (PCH) for 30 h or cells treated with pyoverdine (PVD) or deferoxamine (DFO) for 60 h normalized to that of media control. All treatments were at 100 μM in serum-free EMEM. (B) Proinflammatory gene expression (NLRP3, IL1B, NLRP1, IL8, TNF) in cells treated with iron chelators compared to that of media control. mRNA levels were measured by qRT-PCR. (C) IL-8 protein concentration in the supernatants of 16HBE cells treated with iron chelators. IL-8 was quantified by ELISA. (D, E) Total RNA yield (D) or proinflammatory gene expression (E) in 16HBE cells treated with iron chelators with or without excess Ga(NO₃)₃ supplementation. (F) Proinflammatory gene expression after 60 h treatment with pyoverdine, deferoxamine, or both molecules. All error bars represent SEM from three biological replicates. * corresponds to p < 0.01, # corresponds to p < 0.05, and NS corresponds to p > 0.05 based on one-way ANOVA with Dunnett’s (A, C), Sidak’s (D, E), or Tukey’s (F) multiple comparisons test.

Finally, we investigated whether pyoverdine can indirectly promote lung inflammation by potentiating other iron chelating molecules. Due to its exceptionally high affinity for iron, pyoverdine is likely to remove iron from other, more cell-permeable siderophores or outcompete them for trace iron in the extracellular milieu, increasing the pool of apo-siderophores that can promote inflammation. To test this hypothesis, we treated 16HBE cells with deferoxamine, pyoverdine, or both. Cells treated with both siderophores exhibited higher expression of proinflammatory genes compared to those treated with deferoxamine alone (Fig. 6F). Considering that pyoverdine alone did not affect the transcription of proinflammatory genes, these results suggest that pyoverdine enhanced deferoxamine-mediated damage by removing iron from deferoxamine and effectively increasing the pool of apo-deferoxamine.

Discussion:

Arguably one of the greatest challenges to combating P. aeruginosa infections is the sheer multitude of virulence factors produced by the bacterium that contribute to pathogenesis. These include small molecule virulence factors (e.g., siderophores, quorum-sensing molecules), factors involved in biofilm formation and motility (e.g, exopolysaccharides, type IV pili, flagella), and more than twenty toxins that either directly kill host cells (e.g., exotoxin A, exoenzyme S, exotoxin T, exotoxin U) or damage host tissue (e.g., elastase LasA, elastase LasB, PrpL, alkaline protease) (32, 45–48). This complexity casts a dark shadow over the prospects of epidemiological or therapeutic intervention. Ideally, we would be able to reliably predict a pathogen’s ability to cause disease through our evolving molecular surveillance tools such as whole-genome sequencing and mass spectrometry and therapeutically impair pathogenesis through antivirulence drugs that inhibit the production or function of key virulence factors and toxins. In P. aeruginosa, the only feasible way to approach these strategies would be to unravel the regulation of virulence factors in the bacterium and to target virulence networks rather than individual factors.

The results we report in this study suggest that the alternative sigma factor PvdS may be a promising target for therapeutic intervention during P. aeruginosa lung infections. PvdS has already been shown to regulate the production of several secreted toxins, such as the translational inhibitor exotoxin A and the secreted protease PrpL. Exotoxin A, arguably one of the most extensively studied toxins in P. aeruginosa, inhibits protein synthesis (46, 49), inducing airway epithelial cell death (50) and inhibiting cell junction repair in the presence of P. aeruginosa elastase (51). Exotoxin A also contributes to P. aeruginosa virulence in various murine infection models (52–54). PrpL degrades host defense factors, like surfactant proteins and IL-22, that contribute to lung innate immunity (55–57). PrpL has also been shown to directly contribute to P. aeruginosa virulence during ocular infections (58, 59).

PvdS is best known for its role in pyoverdine biosynthesis and is indispensable for the production of pyoverdine biosynthetic enzymes. In addition to scavenging trace iron in the environment or directly from host ferroproteins, pyoverdine is involved in a positive feedback loop where the uptake of iron-bound pyoverdine by its outer membrane receptor, FpvR, derepresses PvdS, increasing the production of pyoverdine, exotoxin A, and PrpL (12, 18). Here, we have demonstrated that pyoverdine promotes the production of an additional secreted material, rhamnolipids, that rapidly induce cell death. Secreted rhamnolipids have been shown to assemble into micellar structures (60) that directly interact with host membranes, causing rapid plasma membrane rupture and cell death (29, 61, 62) and damaging the lung epithelium (63, 64). It remains unclear however how pyoverdine regulates rhamnolipid production and whether this mechanism is linked to quorum-sensing, the primary mode of rhamnolipid regulation in the bacterium. While certain studies suggest that P. aeruginosa quorum-sensing affects pyoverdine production (65–69), an inverse relationship has yet to be explored.

In addition to regulating secreted toxins, pyoverdine may also indirectly contribute to inflammation by removing iron from other more cell permeable siderophores such as pyochelin, deferoxamine, or enterobactin, the latter in the context of polymicrobial infections with Enterobacteriaceae such as the respiratory pathogen K. pneumoniae (70). Importantly, while P. aeruginosa may lose the ability to produce pyoverdine during lung infection with the emergence of social cheaters or due to a transition in iron acquisition strategy (71–73), several surveys of patient sputum samples and clinical isolates have revealed that a large fraction of strains still exhibit substantial pyoverdine production (22, 74–76), demonstrating that pyoverdine may be an important target for therapeutic intervention.

We took advantage of the FDA-approved antimycotic drug, 5-fluorocytosine (5-FC), that inhibits pvdS expression in P. aeruginosa and attenuates virulence during murine lung infection. Imperi and colleagues first identified 5-FC in a screen for small molecules that inhibit pyoverdine production (23). We have independently identified a chemical analogue of 5-FC, 5-fluorouracil – another pvdS inhibitor, in a small molecule screen for compounds that rescue C. elegans from P. aeruginosa in a pyoverdine-dependent pathogenesis model (37, 77). We also recently reported that 5-FC synergizes with another FDA-approved drug, gallium nitrate, to inhibit P. aeruginosa growth and virulence against C. elegans (38). Our findings in this study suggest that in addition to its bactericidal and biofilm-inhibitory activities (78, 79), gallium could also function as an anti-inflammatory agent during lung infection by inhibiting intracellular iron chelation by pyochelin and mitigating not only epithelial cell death, but also activation of proinflammatory pathways such as the NLRP3 inflammasome and IL-8 production. These newly discovered roles for pyochelin further demonstrate that pyoverdine and pyochelin play distinct roles in P. aeruginosa virulence. Previous studies have shown that pyochelin triggers reactive oxygen species (ROS) production in an iron-dependent manner in host cells during infection (80) or in other microbes during interbacterial competition (81–83). While pyochelin may not regulate additional virulence pathways, its ability to permeate membranes makes this siderophore a distinctly effective tool for direct host cell damage.

The benefits of suppressing pyochelin-mediated neutrophilic inflammation during lung infection, particularly chronic lung infection, has been well documented. While the mechanisms neutrophils employ to kill and remove pathogens such as the production of neutrophil elastases are important for host defense, they can also cause tissue damage by degrading extracellular matrix proteins (84, 85). During chronic infections (such as those in CF patients), these host defense factors continue to cause airway damage while the pathogen persists, exacerbating the decline in pulmonary function (86). While lung inflammation is mediated by many factors (cytokines and chemokines), for CF patients, a strategy to specifically inhibit the NLRP3 inflammasome by therapeutics such as MCC950 is currently being investigated with promising results in murine infection studies (87, 88). It is important to note that NLRP3 inflammasome priming (i.e., transcriptional upregulation of NLRP3, encoding the major component of the inflammasome, and IL1B, encoding pro-IL-1β) (89) was a pathway that was activated by intracellular iron chelation in lung epithelial cells (Fig. 6B). We observed this transcriptional response in not only wild-type 16HBE cells but also those carrying mutations in the cystic fibrosis transmembrane conductance regulator (CFTR G551D, CFTR ΔF508), two of the most frequently identified mutations in CF patients (Fig. S8) (90). While gallium has been broadly associated with anti-inflammatory properties (91–93), studies have yet to specifically explore gallium’s role in inhibiting pathogen-associated inflammation. Considering recent findings that bacterial siderophores promote inflammation (41–43, 70), this therapeutic avenue may merit consideration.

Materials and Methods:

Bacterial Strains and Growth Conditions

See list of bacterial strains in Table 1. All MPAO1 transposon insertion sites were verified by Sanger sequencing (94). Insertions were determined by adapting a previously established method for the MAR2xT7 transposon library in PA14 (95) and primers in Table S2. To produce pyoverdine-rich conditioned medium, an LB overnight culture of P. aeruginosa was diluted 20-fold into 2 mL of serum-free Eagle’s Minimum Essential Medium (EMEM) (Millipore Sigma, St. Louis, MO) in a 6-well plate. The plate was sealed with a Breathe-Easy sealing membrane (Diversified Biotech, Dedham, MA) and grown statically at 37 °C for 18 h. Pyoverdine production (Ex. 405 nm; Em. 460 nm) and bacterial growth (Abs. 600 nm) were measured spectrophotometrically on a Cytation5 Multimode Reader (Biotek, Winnoski, VT). Bacteria was then removed by centrifugation and the supernatant was treated with an antibiotic cocktail to kill residual bacteria (100 μg/mL amikacin, 100 μg/mL gentamicin, 100 μg/mL tobramycin).

Table 1.

List of bacterial strains and cell lines used in this study.

Strains	Relevant Genotype	Source or Reference
Pseudomonas aeruginosa Strains
PAO1	WT	D. Haas
PAO1ΔpvdF	ΔpvdF	D. Haas
PAO1ΔpchBA	ΔpchBA	D. Haas
PAO1ΔpvdFpchBA	ΔpvdFpchBA	D. Haas
MPAO1 PW2280	cat (ISlacZ/hah)	Jacobs et al., 2003
MPAO1 PW5033	pvdF (ISlacZ/hah)	Jacobs et al., 2003
MPAO1 PW3078	toxA (ISphoA/hah)	Jacobs et al., 2003
MPAO1 PW8077	prpL (ISphoA/hah)	Jacobs et al., 2003
MPAO1 PW6887	rhlA (ISphoA/hah)	Jacobs et al., 2003
MPAO1 PW6882	rhlR (ISlacZ/hah)	Jacobs et al., 2003
MPAO1 PW6880	rhlI (ISphoA/hah)	Jacobs et al., 2003
MPAO1 PW5085	pvdS (ISlacZ/hah)	Jacobs et al., 2003
MPAO1 PW6223	xcpQ (ISphoA/hah)	Jacobs et al., 2003
MPAO1 PW2537	plcH (ISlacZ/hah)	Jacobs et al., 2003
MPAO1 PW2536	plcH (ISlacZ/hah)	Jacobs et al., 2003
PA2–72	WT (CF Isolate)	Kang et al., 2019
PA2–61	WT (CF Isolate)	Kang et al., 2019
Other Bacterial Strains
Escherichia coli JM109 pSB536	PahyI ::luxABCDE regulated by constitutive AhyR	Winson et al., 1998
Human Bronchial Epithelial Cells
16HBE	WT CFTR	G. Bao
16HBE G551D	CFTR G551D	G. Bao
16HBE ΔF508	CFTR AF508	G. Bao

Cell Culture

Wild-type and mutant 16HBE cells (Table 1) were passaged in Eagle’s Minimum Essential Medium (EMEM) supplemented with 10% fetal bovine serum (Corning, Corning, NY), penicillin/streptomycin (Millipore Sigma), and MEM non-essential amino acids (Millipore Sigma).

For experiments with P. aeruginosa conditioned medium, 4×10⁶ cells were seeded into each well of a collagen (type I from calf skin - Millipore Sigma)-coated 12-well plate and grown at 37 °C for 24 h in a CO₂-jacketed incubator by which they reached 100% confluence. To visualize the epithelial monolayer, cells were stained with 2.5 μg/mL CellMask Orange plasma membrane stain (Invitrogen, Carlsbad, CA) for 1 h prior to conditioned medium exposure. Following treatment, the medium was aspirated and the monolayer was imaged on a Cytation5 Multimode Reader using a RFP filter cube. Percentage image area covered by fluorescent cells was quantified using ImageJ. To visualize cell death, cells were prelabeled with 20 μM Hoechst 33342 (ThermoFisher Scientific, Waltham, MA) for 30 min then exposed to conditioned medium in the presence of 2.5 μM Sytox Orange (Invitrogen). Following treatment, the medium was aspirated, and the monolayer was imaged on a Cytation5 Multimode Reader using DAPI (for Hoechst 33342) and RFP (for Sytox Orange) filter cubes. Images were exported and quantified for mean blue or red fluorescence intensity on ZEN Blue image analysis software (Zeiss, Oberkochen, Germany).

For 16HBE cell viability measurements following iron chelator treatment, 440 μM resazurin (ThermoFisher Scientific) in phosphate buffered saline was diluted 10-fold into the treatment medium, and cells were incubated for 1.5 h. The medium was collected and briefly centrifuged to remove cells. 150 μL of the supernatant was transferred to a 96-well plate, and resorufin (reduced resazurin) fluorescence (Ex. 560 nm; Em. 590 nm) was measured on a Cytation5 Multimode Reader.

N-butanoyl-L-homoserine lactone (C4-HSL) Measurement

To measure C4-HSL concentration, one part spent culture medium from P. aeruginosa grown in EMEM was mixed with three parts E. coli JM109 pSB536 in LB medium. This LB medium was inoculated with the E. coli reporter strain by diluting an overnight culture 5-fold. 150 μL of the P. aeruginosa conditioned medium-reporter strain mixture was transferred to each well in a 96-well plate and incubated at 37 °C for 2 h in a Cytation5 Multimode Reader. Bioluminescence measurements were taken every 15 min.

P. aeruginosa whole genome sequence analysis

Bacterial genomic DNA was purified from overnight culture using DNeasy UltraClean Microbial Kit (Qiagen, Hilden, Germany). Paired end Illumina sequencing was performed by the Microbial Genome Sequencing Center (MiGS, Pittsburgh, PA) for at least 40X genome coverage. To compare the two plcH transposon mutants, raw sequencing reads from PW2537 were first assembled via SPAdes (96) and annotated via Prokka using PAO1 (NC_002516.2) as the reference genome (97). Genomic polymorphisms in PW2536 were identified using breseq (98).

Pyoverdine Purification

A LB overnight culture of P. aeruginosa PAO1 was diluted 100-fold into 300 mL of M9 medium (1% w/v 5X M9 Salts (BD Difco, Franklin Lakes, NJ), 1.5% w/v Bacto Casamino Acids with low iron and salt content (BD Difco), 1 mM MgSO₄, 1 mM CaCl₂) in a 2 L flask and grown aerobically for 24 h at 37 °C. Bacteria were then removed by centrifugation and filtration through a 0.22 μm membrane. The filtrate was incubated with 10% w/v amberlite XAD-4 resin (MilliporeSigma) at room temperature for 4 h with constant agitation. After rinsing the resin with copious amounts of water, pyoverdine was eluted in 50% methanol. This eluent was diluted in water to 15% methanol and loaded onto a Luna Omega 5 μm Polar C18 LC prep column (Phenomenex, Torrance, CA) for high-performance liquid chromatography on a 1220 Infinity LC system (Agilent Technologies, Santa Clara, CA). Pyoverdine was eluted from the column by a 0–100% methanol gradient across 4 h at a flowrate of 5 mL/min. Fractions were collected every other minute for pyoverdine content analysis (Fig. 5B). The fractions with the highest pyoverdine content were pooled. Methanol was evaporated using a SpeedVac vacuum concentrator. The final purified product was analyzed by HPLC on an analytical column to verify sample purity (Fig. 5C).

Confocal Laser Scanning Microscopy

8×10⁶ 16HBE cells were seeded into each well of a collagen-coated 6-well plate and grown at 37 °C for 24 h in a CO₂-jacketed incubator by which they reached 100% confluence. After treatment, cells were washed in serum-free EMEM and detached from the microtiter plate by trypsin-EDTA solution (MilliporeSigma). After inactivating the trypsin with media containing 10% fetal bovine serum, the cells were concentrated via centrifugation and transferred onto a glass side with a 3% noble agar pad. These slides were visualized under a LSM800 AiryScan confocal laser scanning microscope (Zeiss). Pyoverdine fluorescence was visualized via a 405 nm laser line using the channel conditions for Pacific Blue. Dextran-Texas Red (Invitrogen) fluorescence was visualized via a 561 nm laser line using channel conditions for Texas Red. CellMask Deep Red plasma membrane stain (Invitrogen) fluorescence was visualized via a 640 nm laser line using channel conditions for Alex Fluor 660.

qRT-PCR

For P. aeruginosa, bacterial cells were collected from 12 mL of EMEM culture by centrifugation. The pellet was resuspended in 2 mL of TRI reagent (Molecular Research Center, Cincinnati, OH) for phenol/chloroform/guanidinium thiocyanate RNA extraction according to manufacturer’s protocols (bromochloropropane phase separation followed by isopropanol RNA precipitation). Prior to phase separation, bacterial cells were homogenized with 0.1 mm zirconia beads by vigorous vortexing. To remove DNA contaminants in bacterial RNA extracts, samples were treated with DNase I (ThermoFisher Scientific) at 37 °C for 30 min, followed by 75 °C enzyme heat denaturation for 10 min. For 16HBE cells grown and treated in 6-well plates, the treatment medium was aspirated, and cells were incubated in TRI reagent at room temperature for 15 min to lyse cells prior to RNA extraction.

For both bacterial and human cell RNA, cDNA synthesis was performed on a Bio-Rad T100 Thermo Cycler (Bio-Rad, Hercules, CA) using a reverse transcription kit (Applied Biosystems, Waltham, MA). qRT-PCR was performed on a Bio-Rad CFX Connect Real-Time System (Bio-Rad) using a universal qPCR master mix (New England Biolabs, Ipswich, MA). All qPCR primer sequences are in Table S2. For P. aeruginosa genes, cDNA amplification (Ct value) was normalized to that of housekeeping gene gyrB. For 16HBE genes, cDNA amplification was normalized to that of ACTB.

Supplementary Material

Supplement 1

Fig. S1. Disruption of pyochelin biosynthesis does not mitigate epithelial damage. (A, B) Bacterial growth (A) or pyoverdine production (B) of P. aeruginosa PAO1 siderophore biosynthetic mutants (PAO1ΔpvdF – pyoverdine; PAO1ΔpchBA – pyochelin; PAO1ΔpvdFΔpchBA – pyoverdine and pyochelin) after 18 h growth in serum-free EMEM. (C) MH-S murine alveolar macrophage viability after 2 h exposure to conditioned medium from PAO1 siderophore mutants grown in EMEM. Cell viability was measured using a resazurin-based assay. (D) Fluorescent micrographs of 16HBE cells after 30 min exposure to conditioned medium from PAO1 siderophore mutants. Cells were prelabeled with CellMask Orange plasma membrane stain. (E) Quantification of percentage micrograph area covered by fluorescent cells. All error bars represent SEM from four biological replicates. * corresponds to p < 0.01 and NS corresponds to p > 0.05 based on one-way ANOVA with Tukey’s multiple comparisons test.

Fig. S2. Lipid extraction abrogates epithelial damage by pyoverdine-rich conditioned medium. (A) Fluorescent micrographs of 16HBE cells after 30 min exposure to conditioned medium from WT PAO1. Conditioned medium was pretreated with 200 μM Ga(NO₃)₃, 200 μM FeCl₃, or 100 μg/mL proteinase K for 24 h, had macromolecules depleted via a 10 kDa centrifugal filter, or had lipids depleted by chloroform (CHCl₃) extraction. Cells were prelabeled with CellMask Orange plasma membrane stain. (B) Quantification of percentage micrograph area covered by fluorescent cells. Error bars represent SEM from three biological replicates. * corresponds to p < 0.01 and NS corresponds to p > 0.05 based on one-way ANOVA with Dunnett’s multiple comparisons test.

Fig. S3. Type II secretion system toxins do not contribute to lung epithelial damage. (A) Fluorescent micrographs of 16HBE cells after 30 min exposure to conditioned medium from MPAO1 transposon mutants. Cells were prelabeled with CellMask Orange plasma membrane stain. (B) Fluorescent micrographs of 16HBE cells after 15 min exposure to conditioned medium from MPAO1 transposon mutants in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (C) Quantification of percentage micrograph area covered by fluorescent cells in (A). (D) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in (B). (E) Protease activity in conditioned medium from MPAO1 transposon mutants. Proteolytic activity was measured by fluorescence release from cleavage of FITC-conjugated casein. All error bars represent SEM from three biological replicates. * corresponds to p < 0.01 and NS corresponds to p > 0.05 based on one-way ANOVA with Dunnett’s multiple comparisons test.

Fig. S4. Pyoverdine biosynthetic mutants do not exhibit impaired C4-HSL quorum-sensing. (A) Bioluminescence produced by E. coli JM109 pSB536 (N-butanoyl-L-homoserine lactone reporter strain) grown in media supplemented with conditioned medium from wild-type PAO1 or PAO1ΔpvdF. (B) Bioluminescence at the kinetics (A) endpoint. (C) Bioluminescence produced by the E. coli reporter strain grown in media supplemented with conditioned medium from MPAO1 transposon mutants. (D) Bioluminescence at the kinetics (C) endpoint. All error bars represent SEM from at least three biological replicates. * corresponds to p < 0.01 and NS corresponds to p > 0.05 based on one-way ANOVA with Tukey’s multiple comparisons test.

Fig. S5. Purified rhamnolipids are sufficient for 16HBE cell death but not for epithelial monolayer damage. (A) Quantification of rhamnolipids in purified samples (RHL) or P. aeruginosa conditioned medium from WT PAO1, PAO1ΔpvdF, MPAO1cat, or MPAO1pvdF. (B) Fluorescent micrographs of 16HBE cells after 30 min exposure to purified rhamnolipids in EMEM. Cells were prelabeled with CellMask Orange plasma membrane stain. (C) Fluorescent micrographs of 16HBE cells after 15 min exposure to purified rhamnolipids in EMEM in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (D) Quantification of percentage micrograph area covered by fluorescent cells in (B). (E) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in (C). (F) Quantification of rhamnolipids in conditioned medium from WT PAO1 or conditioned medium after heat denaturation (85 °C for 1 h). (G) Fluorescent micrographs of 16HBE cells after 30 min exposure to conditioned medium from WT PAO1 or conditioned medium after heat denaturation. Cells were prelabeled with CellMask Orange plasma membrane stain. (H) Fluorescent micrographs of 16HBE cells after 15 min exposure to conditioned medium from WT PAO1 or conditioned medium after heat denaturation in the presence of Sytox Orange nucleic acid stain (red). Cells were prelabeled with Hoechst 33342 nucleic acid stain (blue). (I) Quantification of percentage micrograph area covered by fluorescent cells in (G). (J) Quantification of Sytox Orange mean fluorescence intensity normalized to that of Hoechst 33342 in (H). All error bars represent SEM from three biological replicates. # corresponds to p < 0.05 and NS corresponds to p > 0.05 based on Student’s t-test.

Fig. S6. Pyoverdine accumulates in early endosomes of lung epithelial cells. (A, B) 16HBE cell viability after 48 (A) or 72 h (B) treatment with ciclopirox olamine, 1, 10-phenanthroline, pyoverdine, or deferoxamine in serum-free EMEM. (C) Confocal micrographs of 16HBE cells exposed to 100 μM purified pyoverdine, pyoverdine with excess Ga(NO₃)_3, pyoverdine with excess FeCl₃, or media control for 24 h. Cells were labeled with CellMask Deep Red plasma membrane stain and trypsinated prior to imaging. (D) Confocal micrographs of 16HBE cells treated with 100 μM pyoverdine-gallium and dextran-Texas Red (10,000 MW). Bottom row shows an enlarged micrograph of one representative cell. All error bars represent SEM from three biological replicates.

Fig. S7. Small molecule iron chelators promote IL-8 production in 16HBE cells. (A) IL8 mRNA levels in 16HBE cells treated with iron chelators (ciclopirox olamine – CPX, 1,10-phenanthroline – PHE, pyochelin – PCH, pyoverdine – PVD) for 30 h compared to that of media control. mRNA levels were measured by qRT-PCR. (B) IL-8 protein concentration in the supernatants of 16HBE cells treated with iron chelators for 48 h. IL-8 was quantified by ELISA. (C) Correlation between IL8 gene expression and IL-8 protein production. All error bars represent SEM from three biological replicates.

Fig. S8. Deferoxamine promotes the expression of proinflammatory genes in 16HBE CFTR mutants. (A) Total RNA yield in wild-type 16HBE cells and 16HBE cells carrying mutations (G551D, ΔF508) in the cystic fibrosis transmembrane conductance regulator (CFTR) after 48 h treatment with 100 μM deferoxamine in serum-free EMEM. (B) Expression of proinflammatory genes (NLRP3, IL1B, NLRP1, IL8, TNF) or iron-regulated genes (NDRG1, TFRC) in 16HBE cells treated with deferoxamine compared to that of media control. mRNA levels were measured by qRT-PCR. All error bars represent SEM from three biological replicates.