Aztreonam-evolved P. aeruginosa replicates more during macrophage infection

To characterize the virulence of the P. aeruginosa AzEvC10 mutant, we assessed if there was a difference in clearance between the mutant and WT PAO1 by infecting bone marrow-derived macrophages (BMDM), BM-derived neutrophils, or airway epithelial cells (AEC) with 2.5 x 10⁷ CFU/mL of either strain. At 6 h post-infection (hpi) of BMDM, the AzEvC10 mutant had increased in number by about 1 log₁₀ whereas the WT parental strain showed no change from the initial inoculum (p< 0.01) (Fig 1A). In addition, the AzEvC10 mutant induced higher cytotoxicity of BMDM (44% ± 8.2) compared to the WT strain (20% ± 5.2) (p<0.05) (Fig 1B). These differences were specific to BMDM, as there was no difference in the replication of (Fig 1A) or cytotoxicity (S1A Fig) by the two strains in BM-derived neutrophils and AEC. During in vivo lung infections, both alveolar macrophages and infiltrated monocytes kill bacteria. Although alveolar macrophages are the first line of defense against pulmonary pathogens and may clear low concentrations of bacteria, recruitment of circulating monocytes/macrophages was shown to be essential to clear higher concentrations of bacteria in the lung [5,28]. For this reason, we chose to perform our infections using BMDM. Their phenotypes were confirmed for alveolar macrophage markers (S1B Fig). We also confirmed the AzEvC10 hypervirulent phenotype during infection of alveolar macrophages (S1C Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. An aztreonam-evolved P. aeruginosa mutant replicates more during macrophage infection.

A. BMDM, BM-derived neutrophil, or AEC cells were infected with WT PAO1 or the AzEvC10 mutant for 6 h. Dotted lines represent the initial infection inoculum, 2.5 x 10⁷ CFU/mL. Bacterial burden at 6 hpi was determined by viable CFU plate counts. Analysis was matched for experimental repeat. B. BMDM cytotoxicity assessed by LDH assay at 6 hpi by WT or AzEVC10. Analysis was matched for experimental repeat. C. Intracellular survival of bacteria in BMDM at 6 hpi, following amikacin treatment at 1 hpi for a total of 5 h to kill extracellular bacteria. Viable intracellular bacteria were determined by CFU plate counts. D. Escaping bacteria at 6 hpi: Amikacin was added in medium at 1 hpi and left for 1 h. Then cells were incubated in fresh medium (no amikacin) for another 4 h. Escaping bacteria were determined by CFU plate counts. E. Images of BMDM infections at 6 hpi by immunofluorescence showing macrophages in green, P. aeruginosa in red, and DAPI in blue. n = 3–6 independent replicates for each experiment. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See S5 Table for statistical tests used and exact p-values.

https://doi.org/10.1371/journal.ppat.1010942.g001

We tested if increased proliferation of the AzEvC10 mutant was due to its capacity to use growth medium components as a source of carbon and energy. We found that both strains had net lethality in host cell-free medium but that there was a significant survival difference between the strains. At 6 hpi in cell-free medium, the number of viable WT PAO1 was 3.5 log₁₀ less than the initial inoculum, whereas the number of viable AzEvC10 mutants was reduced only 0.8 log₁₀ (S1D Fig). These findings demonstrate that the AzEvC10 mutant has acquired a significant survival advantage–nearly 1,000-fold better than WT–in a nutrient restrictive environment.

We also assessed if the survival and subsequent escape of phagocytosed bacteria contributed to the increased bacterial burden of AzEvC10 mutant in macrophage cultures. To test this, we infected BMDM with the mutant and WT and 1 h later added 200 μg/mL amikacin, an antibiotic that does not enter host cells, to kill extracellular bacteria. This concentration of amikacin effectively killed both WT PAO1 and AzEvC10 mutant bacteria (S1E Fig). At 6 hpi (i.e., 5 h post-amikacin), the number of surviving intracellular AzEvC10 mutants was significantly greater (about 1 log₁₀) than the number of intracellular viable WT PAO1 (p<0.05) (Fig 1C). This could be due to either higher phagocytosis of mutant bacteria by macrophages, increased mutant bacterial proliferation inside host cells, or increased intracellular survival. Regardless, to determine if these intracellular surviving bacteria could escape from BMDM and proliferate extracellularly, we treated BMDM medium with amikacin at 1 hpi for 1 h, then washed the cells and added fresh medium without antibiotic. Compared to WT infected cells, we found that extracellular bacterial counts were about 1 log₁₀ increased in AzEvC10 mutant infected cultures at 6 hpi (Fig 1D and 1E), proportional to the difference in viable intracellular bacteria (Fig 1C). These data suggest that both WT PAO1 and AzEvC10 mutant can escape BMDMs and proliferate extracellularly with similar efficiency. Although the specific cause of the increased intracellular bacterial recovery needs to be defined, overall, our studies indicate that AzEvC10 adapts better than WT PAO1 to the host environment and that this growth advantage was enhanced by the presence of macrophages. Below we explore the bacterial and host pathways involved in the mutant’s growth advantage.

AzEvC10 mutant hypervirulence is caused by a compensatory mutation in RNase E

The aztreonam-evolved AzEvC10 mutant first acquired a loss-of-function point mutation in the nalD gene (S2A Fig) in response to the aztreonam selective pressure, which caused overexpression of the MexAB-OprM efflux pump and a multidrug resistance phenotype [26]. Recently, we found that mutation of another mexAB-oprM regulator was sufficient to increase P. aeruginosa virulence [27]. Thus, we hypothesized that the AzEvC10 mutant’s virulence was being driven by its nalD mutation. To test this, we complemented a functional wild-type nalD gene in the AzEvC10 mutant and infected BMDM to assess if this would reverse the hypervirulence. Complementation of the nalD gene was confirmed by mRNA expression quantification and aztreonam MIC (S2B and S2C Fig). Unexpectedly, nalD complementation in AzEvC10 did not reverse bacterial growth and cytotoxicity during BMDM infection (S2D and S2E Fig). This was further confirmed when we infected BMDM with mutants in which mexAB genes were deleted (S2F–S2I Fig).

To explore if the AzEvC10 mutant’s hypervirulence could be explained by a second-site compensatory mutation, we used whole genome sequencing and found that the strain evolved a 50-bp deletion (bases 3121-3170/3174) in the rne gene (Fig 2A and 2B) encoding RNase E, an endonuclease involved in RNA turnover. Because compensatory mutations can restore fitness and/or virulence, and due to the nature of RNase E activity, we hypothesized that the rne variation was involved in the virulence and growth of the AzEvC10 mutant during macrophage infection. To begin testing the virulence consequences of the rne mutation, we created single nalD_T158P and rne_Δ50bp mutants in a wild-type P. aeruginosa background and used an rne mutant containing a transposon inserted at the position 2945 of the gene (rne::Tn) [29]. We also reversed the rne mutation directly in the AzEvC10 chromosome at its native position (AzEvC10 rne_WT) as a rescue. As expected, the aztreonam resistance seen in AzEvC10 mutant was caused by the mutation in nalD gene (Fig 2C). However, we found that bacterial burden during BMDM infection and BMDM cytotoxicity were increased in AzEvC10, rne_Δ50bp, and rne::Tn mutants compared to WT, nalD_T158P, and AzEvC10 rne_WT strains (Fig 2D–2F). These results confirmed the role of the RNase E variant in the hypervirulent phenotype and growth of the AzEvC10 mutant during macrophage infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. AzEvC10 mutant hypervirulence is caused by an RNase E mutation.

A. Genome diagram showing coverage of DNA sequencing reads aligning to the rne gene. The deletion of bases 3121-3170/3174 nt of the rne gene in the AzEvC10 mutant are enlarged in the bottom panel. Genome coverage plots generated from sequencing read alignments to the PAO1 reference genome are indicated in grey. B. Schematic representation of the rne gene and corresponding RNase E protein domain description. WT PAO1, nalD_T158P, and AzEvC10 rne_WT mutants carry a wild-type rne gene. AzEvC10 and rne_Δ50bp mutants have a deletion of the bases 3121-3170/3174 nt in the rne gene. The rne::Tn mutant carries an ISphoA/hah transposon inserted at the position 2945 nt of the gene. C. Aztreonam MIC was assessed by Etest for the indicated bacteria. D-E. BMDM cells were infected with WT PAO1 or different nalD and rne mutants at MOI:100 for 3 h (D) and 6 h (E). Bacterial burden was determined by viable CFU plate counts. Dotted lines represent the initial infection inoculum, 2.5 x 10⁷ CFU/mL. F. BMDM cytotoxicity was assessed by LDH assay at 6 hpi with indicated bacteria. n = 5–8 independent replicates for each experiment *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See S5 Table for statistical tests used and exact p-values.

https://doi.org/10.1371/journal.ppat.1010942.g002

Iron acquisition genes are upregulated in the AzEvC10 mutant during BMDM infection

While the rne variation was responsible for the increased virulence of the AzEvC10 mutant, the mechanisms driving the rne variant’s increased virulence were unclear. To understand the mechanisms underlying the AzEvC10 mutant phenotype, we infected BMDM and performed RNA sequencing at 3 hpi and 6 hpi. We first compared the AzEvC10 mutant gene expression to WT PAO1 during the BMDM infections at the two timepoints. We found a marked transcriptional upregulation of the two P. aeruginosa siderophore biogenesis pathways at both 3 hpi and 6 hpi in the AzEvC10 mutant (Fig 3A and 3B). These pathways include genes involved in pyoverdine biosynthesis (pvdA, pvdD, pvdE, pvdF, pvdG, pvdH, pvdJ, pvdL, pvdN,pvdO, pvdP, pvdQ), secretion (pvdR, pvdT, opmQ), and ferric-pyoverdine import (fpvA). mRNA from genes encoding the pyochelin biosynthesis enzymes (pchABCD, pchEFG), ferric-pyochelin import (fptA), and their transcriptional regulator (pchR) were also upregulated at 3 hpi and even more so at 6 hpi. Transcriptional upregulation of heme acquisition genes (hasAp, hasD, hasE and hasR) was observed but only at 6 hpi. Interestingly, none of these pathways were upregulated at the mRNA level when we grew the strains in LB broth (Fig 3B), suggesting that this mutant’s response was specific to the macrophage infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Iron acquisition genes are upregulated in the AzEvC10 mutant during macrophage infection.

A-B. Genes involved in iron acquisition were upregulated in the AzEvC10 mutant compared to WT PAO1 during BMDM infection at 3 hpi (n = 2) and 6 hpi (n = 4–5), but not when grown in LB broth (n = 2). A. Volcano plot highlighting differentially expressed iron acquisition genes in red during BMDM infection. B. Heat map indicates expression of indicated genes in individual biological replicates expressed in normalized number of reads (TPM) during WT and AzEvC10 BMDM infections at indicated time points. C-F. Pyochelin and pyoverdine production by given strains measured by fluorescence (Ex350/Em430 for pyochelin and Ex400/Em460 for pyoverdine) at 3 hpi (C-D) and 6 hpi (E-F) and normalized to log₁₀(CFU). n = 5–8 independent replicates for each experiment. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See S5 Table for statistical tests used and exact p-values.

https://doi.org/10.1371/journal.ppat.1010942.g003

To confirm the effect of the RNase E variation on increased siderophore biogenesis, we measured pyochelin and pyoverdine secreted in the growth medium during BMDM infection. To do this, we measured the fluorescence at Ex350/Em430 for pyochelin and Ex400/Em460 for pyoverdine. The two siderophores had enough fluorescence emission in the macrophage growth medium to be detected (S3A–S3D Fig). A limitation of this technique is that the fluorescence of ferric-pyochelin and ferric-pyoverdine is quenched thus, undetectable (S3–S3D Fig) [30]. Thus, for both siderophores, the fluorescence measurements reflect only a relative quantification of the metal-free forms that was normalized on the log₁₀(CFU). At 3 hpi and 6 hpi, both siderophores were barely detectable in WT, nalD_T158P, and AzEvC10 rne_WT strains (Figs 3C–3F and S3E). In contrast, the AzEvC10 and rne::Tn mutants had produced detectable pyochelin and pyoverdine by 3 hpi, and at 6 hpi, the concentrations increased further (Figs 3C–3F and S3E). Surprisingly, the rne_Δ50bp mutant did not secrete siderophores at 3 hpi, as seen in AzEvC10 and rne::Tn mutants (Figs 3C–3D and S3E). However, the mutant did secrete pyochelin and pyoverdine at 6 hpi (Figs 3E–3F and S3E). While the reason for this difference is not clear, it is possible that the dynamics of increased siderophore production vary slightly based on genetic background and the nature of the rne variation.

The AzEvC10 mutant carries specific gene expression signatures during BMDM infection

It was unclear if the increased iron acquisition capacity in the AzEvC10 mutant preceded or followed its increased proliferation during BMDM infection. This distinction is important to resolve because either: 1) the increased capacity to capture iron could improve bacterial fitness under stressful conditions such as being in contact with host immune cells, or 2) increased mutant proliferation could create a greater need for iron and trigger the biosynthesis of pyochelin and pyoverdine siderophores. To resolve these two possibilities, we first performed Gene Ontology Enrichment Analyses to identify processes that were differentially regulated by the AzEvC10 mutant at 3 hpi and 6 hpi relative to WT infections. Our analyses revealed two distinct gene expression patterns between the two timepoints. At 3 hpi, the AzEvC10 mutant showed increased gene expression of siderophore biosynthesis and secretion processes, including siderophore and pyoverdine metabolic processes, T3SS, cell projection and assembly, protein membrane transport and secretion (Fig 4A). In contrast, at 6 hpi, genes in metabolic pathways such as isoprenoid, lipid, terpene carboxylic acid, and amino acid catabolism were differentially expressed (Fig 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The AzEvC10 mutant upregulates virulence factor expression before metabolic gene expression during macrophage infection.

A-B. Gene Ontology Enrichment Analyses reveal distinct enriched biological processes in the AzEvC10 mutant at 3 hpi (A) and 6 hpi (B). Percentage indicates relative number of genes in each pathway that were differentially regulated. C. Venn Diagram of AzEvC10 mutant DEGs compared to WT PAO1 at 3 hpi and 6 hpi of BMDM. The 50 most upregulated genes at each timepoint were selected and the expression is presented in mean fold-change compared to WT PAO1 (n = 2–5 independent replicates for each experiment).

https://doi.org/10.1371/journal.ppat.1010942.g004

To confirm these findings, we compared AzEvC10 differentially expressed gene (DEG) sets at 3 hpi and 6 hpi and selected the 50 most upregulated genes at each timepoint (Fig 4C and S1 Table). At 3 hpi, 23 (46%) of the most upregulated genes were involved in pyoverdine biogenesis and ferric-siderophore import. Virulence factor secretion genes were also present in this gene set, with 13 (26%) upregulated genes including 10 genes involved in T3SS. At 6 hpi, 17 (34%) of the most upregulated genes were involved in denitrification processes. The other 24 (48%) genes were involved in carbohydrate, fatty acid and amino acid degradation, metabolism, and transport. These different signatures at 3 hpi and 6 hpi demonstrated how the phenotype of the AzEvC10 mutant changed during the BMDM infection. In the first hours of infection, AzEvC10 secreted pyochelin and pyoverdine siderophores as well as T3SS virulence factors and regulators. At 6 hpi, the gene expression signature shifted to a highly proliferative state and adaptation to decreased oxygen availability. Interestingly, operons involved in pyochelin biosynthesis pchABCD and pchEFG, as well as the transporter fptA were equally overexpressed at both 3 hpi and 6 hpi (S1 Table). These results support the hypothesis that the AzEvC10 mutant increased proliferation during BMDM infection due to a higher capacity to capture iron and by inducing more damage to the host cells by virulence factor secretion.

Increased siderophore secretion results in ferroptosis in macrophages

Reactive oxygen species (ROS) production is a defensive mechanism used by host cells to kill intracellular pathogens. However, too much ROS can cause host damage and lead to cell death. Therefore, we quantified general ROS production in BMDM during bacterial infection (Fig 5A). As expected, both WT and AzEvC10 infected BMDM increased their production of ROS at 6 hpi (p<0.01 and p<0.0001 compared to 3 hpi, respectively). However, BMDM infected with the AzEvC10 mutant had significantly greater ROS production at 6 hpi compared to macrophages infected with WT PAO1 (p<0.01).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The AzEvC10 mutant induces iron dysregulation, oxidative stress, and ferroptosis in macrophages.

BMDM cells were infected with WT PAO1 or AzEvC10 mutant at MOI:100 for 3 h and 6 h. A. Quantification of ROS in infected BMDM by flow cytometry. Quantification is presented in mean fluorescence intensity (MFI). Analysis was matched for experimental repeat. B. Quantification of lipid peroxidation in BMDM by flow cytometry presented as the reciprocal of the ratio of red (Ex561/Em582)/green (Ex488/Em525) fluorescence intensities normalized to WT PAO1 3 h. C-D. BMDM cell death was measured by flow cytometry and characterized by BMDM cells double positive for annexin V and PI. E-G. Uninfected BMDM were treated with 10 μM of either ferric iron (Fe(III)), pyoverdine (Pvd), ferric-pyoverdine (Pvd_Fe(III)), pyochelin (Pch), or ferric-pyochelin (Pch_Fe(III)) for 6h. Quantification of ROS presented in mean fluorescence intensity (MFI). (E-F). Quantification of lipid peroxidation in BMDM by flow cytometry presented as the reciprocal of the ratio of red (Ex561/Em582)/green (Ex488/Em525) fluorescence intensities (G). n = 3–6 independent replicates for each experiment. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See S5 Table for statistical tests used and exact p-values.

https://doi.org/10.1371/journal.ppat.1010942.g005

We then assessed if this degree of ROS stimulated by infection with the AzEvC10 mutant was sufficient to induce cell damage. One of the consequences of ROS toxicity in cells is lipid peroxidation, in which free radicals attack double bonds of fatty acids leading to oxidative damage of polyunsaturated fatty acids and cell structure damage [31]. We used a ratiometric sensor to measure lipid peroxidation in BMDM and found that lipid peroxidation was significantly greater in AzEvC10 infected BMDM compared to those infected with WT PAO1 (p<0.05) (Fig 5B). Consistent with this, we found that BMDM infected with AzEvC10 exhibited high levels of annexin V and propidium iodide staining (Figs 5C–5D and S4A). The increase of these programmed cell death markers matched the increase in AzEvC10-mediated macrophage death shown in Fig 1B.

High ROS production and lipid peroxidation are hallmarks of ferroptosis, a form of iron-mediated programmed cell death. During bacterial infection, macrophages use different mechanisms to sequester iron and restrict its use by bacteria thereby limiting their pathogenicity [32]. However, iron overload can have deleterious effects on macrophages and impair their bactericidal functions. To determine if there was iron dysregulation in macrophages infected by the AzEvC10 mutant, we quantified the labile iron content in BMDM using a colorimetric method described by Abbasi et al. [33]. Since the labile iron content was similar in BMDM infected with the two strains(S4B Fig), a dysregulation in labile iron content was ruled out as a cause of the increased lipid peroxidation observed in AzEvC10-infected BMDM.

We then hypothesized that bacterial siderophores may be directly involved in the ROS production and/or lipid peroxidation in BMDM. In that sense, it has been shown that ferric-pyochelin could be an efficient catalyst for hydroxyl radical formation and induced cell injury [34,35]. Pyoverdine was also recently shown to translocate into host cells and disrupt iron mitochondrial homeostasis [36]. To test this hypothesis, we added 10 μM or 100 μM ferric iron (Fe(III)), pyochelin, ferric-pyochelin, pyoverdine or ferric-pyoverdine to healthy uninfected BMDM for 6 h. We then measured ROS production and lipid peroxidation in these cells. Interestingly, 10 μM pyochelin was sufficient to induce ROS production and lipid peroxidation at the same level as the Fe(III), and ROS production was even greater when BMDM were treated with 10 μM ferric-pyochelin (Fig 5E–5G). At 100 μM, Fe(III) and pyochelin did not induce more ROS and lipid peroxidation than the lower concentration, while 100 μM of ferric-pyochelin resulted in ROS production and lipid peroxidation exceeding levels of detection (S4C–S4E Fig). On the contrary, neither pyoverdine nor ferric-pyoverdine induced ROS or lipid peroxidation at both concentrations (Figs 5E–5G and S4C–S4E).

Although it is possible that other programmed or non-programmed cell death pathways (pyroptosis, necroptosis, necrosis, etc.) are also involved in BMDM death, together our results suggest that the increased pyochelin produced by the AzEvC10 mutant during infection induce excessive ROS and lipid peroxidation leading to ferroptosis of macrophages.

Gallium treatment efficiently kills multidrug-resistant and hypervirulent P. aeruginosa

Gallium is an ion with similar properties to iron and has been proposed as an antimicrobial agent against P. aeruginosa [37]. Because virulence of the AzEvC10 mutant included a higher capacity to acquire iron and induce toxicity with siderophores, we sought to test if gallium was efficient against this multidrug-resistant mutant during BMDM infection. We first confirmed inhibition of bacterial growth with 150 μM gallium in LB broth (Fig 6A). We then tested gallium treatment during BMDM infection (Fig 6B–6E). As shown above (Fig 1B), the AzEvC10 mutant was markedly more toxic to macrophages than WT PAO1. Although gallium significantly decreased BMDM cytotoxicity caused by either strain in a dose-dependent manner (Fig 6B), we saw a significant difference in the protective effectiveness of this iron-like ion. Whereas cytotoxicity caused by WT PAO1 was completely abolished with the lowest dose of gallium (100 μM), BMDM death caused by AzEvC10 remained significantly elevated at all doses. We performed our next experiments using 750 μM gallium since this concentration was effective against the AzEvC10 mutant (Fig 6B) without significant toxicity in the BMDM (S5A Fig). Interestingly, at 750 μM gallium pyochelin and pyoverdine levels were significantly decreased in the AzEvC10 mutant without impacting bacterial growth at 3 hpi (Figs 6C and S5B). Siderophore production was still inhibited at 6 hpi but was accompanied by significant inhibition (~2.5–3.5-fold) of bacterial growth (Figs 6D and S5B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Gallium efficiently inhibits AzEvC10 siderophore production and proliferation during macrophage infection.

A. WT PAO1 and AzEvC10 mutant growth curves in LB broth in the presence or absence of 150 μM gallium. B. BMDM were infected with either WT PAO1 or the AzEvC10 mutant (MOI:100) and treated with different concentrations of gallium for 6 h. BMDM cytotoxicity assessed by LDH assay: #represents the comparisons between WT with gallium to the WT no gallium baseline; *represents the comparisons between AzEvC10 with gallium to AzEvC10 no gallium baseline; $represents the comparisons between AzEvC10 and WT at the same gallium concentration. C-D. BMDM were infected with either WT PAO1 or the AzEvC10 mutant (MOI:100) and treated with 750 μM gallium for 3 h (C) and 6 h (D). Bacterial burden was determined by viable CFU plate counts. Dotted lines represent the initial infection inoculum, 2.5 x 10⁷ CFU/mL. Pyochelin and pyoverdine production was measured by fluorescence (Ex350/Em430 for pyochelin and Ex400/Em460 for pyoverdine) normalized to log₁₀(CFU). E. BMDM were infected with either WT PAO1 or AzEvC10 mutant (MOI:1) and treated with 750 μM gallium for 24 h. Bacterial burden was determined by viable CFU plate counts. Dotted lines represent the initial infection inoculum, 2.5 x 10⁵ CFU/mL. Pyochelin and pyoverdine production was measured by fluorescence (Ex350/Em430 for pyochelin and Ex400/Em460 for pyoverdine) normalized to log₁₀(CFU). n = 6 independent replicates for each experiment. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See S5 Table for statistical tests used and exact p-values.

https://doi.org/10.1371/journal.ppat.1010942.g006

We then infected BMDM for 24 h using a lower MOI (MOI:1) with or without gallium. Gallium was efficient at killing both WT (~1000-fold decrease, p<0.001) and the AzEvC10 mutant (~100-fold decrease, p<0.0001) P. aeruginosa (Fig 6E). As shown at the higher MOI, production of pyochelin and pyoverdine was greater in the AzEvC10 mutant compared to WT PAO1 (p<0.0001) and both were abolished with gallium at 24 hpi (Figs 6E and S5B). It is noteworthy that pyochelin and pyoverdine display a greater fluorescence when bound to gallium than the metal-free forms (S3A–S3D Fig) [30,38]. Thus, the fluorescence detected in the gallium treatment conditions may overestimate the real concentrations of the two siderophores.

Since gallium decreased siderophore production and bacterial load, we wanted to further clarify whether it had a direct impact on macrophage ferroptosis. In that sense, gallium could only act by blocking iron intake by the bacteria, leading to their death, or it could have additional direct protective effects on BMDM. To test this, we measured ROS production and lipid peroxidation in infected BMDM with or without gallium (S5C–S5H Fig). Interestingly, gallium slightly increased lipid peroxidation in BMDM at 3 hpi without any increase in ROS production (S5C and S5D Fig), which suggests modest toxicity in BMDM. Although this increase was significant, it was relatively stable over time (S5E–S5H Fig). Moreover, infected BMDM treated with gallium had lower ROS production and lipid peroxidation compared to untreated infected BMDM at 6 hpi and 24 hpi (S5E–S5H). This supports a protective role for gallium during bacterial infection that could outweigh its toxicity. Finally, we found that gallium-saturated pyochelin caused significantly less damage to BMDM compared to equivalent ferric-pyochelin concentrations (S5I and S5J Fig).

Altogether, our results suggest that gallium 1) has bactericidal potential against multidrug-resistant and hypervirulent P. aeruginosa strains, including strains like the AzEvC10 mutant, and 2) protects BMDM against ferroptosis during infection by counteracting ferric-pyochelin toxicity.

Virulence and metabolic adaptation are common features of pulmonary infections

P. aeruginosa is a versatile bacterium that adapts to different environments. While the in vitro BMDM infections recapitulate interactions between P. aeruginosa and macrophages during in vivo infections, this model lacks several key environmental characteristics that mirror real-world human infections. These missing features include other host cells like neutrophils and epithelial cells, as well as the physical and chemical properties of sputum in CF airways. Thus, gene expression measured during in vitro experiments may not always accurately predict bacterial functioning during clinical infections.

To test if the pathoadaptive phenotype observed in the AzEvC10 mutant during macrophage infection was reflective of P. aeruginosa gene expression profiles during real-world infections, we analyzed RNA-sequencing data from CF sputum and subsequently isolated P. aeruginosa clinical isolates grown in vitro [39] (Fig 7A). In accordance with published findings [39], our analyses showed that pyochelin biosynthesis enzymes (pchABCD, pchEFG) and its transporter (fptA) were significantly overexpressed during pulmonary infections of CF patients but not when grown in vitro (Fig 7B and S2 Table). We then compared the genes that were differentially expressed between the sputum samples and in vitro grown P. aeruginosa with the DEGs of our WT PAO1 or AzEvC10 strains between macrophage infection and LB growth. As expected, common upregulated genes between AzEvC10 mutant 3 hpi and sputum included pyoverdine and pyochelin biosynthesis enzymes and transporters (Fig 7C and S2 Table). Other common genes were involved in the response to oxidative stress (ahpC), T1SS (aprA, aprFED), and regulation of the T3SS (rsmZ and pcrG, respectively). Interestingly, expression of the alkaline protease (aprA) and its secretory apparatus (aprFED) is regulated by iron levels and they facilitate iron acquisition by the proteolytic cleavage of transferrin [40,41]. This highlights the importance of iron acquisition during infection and a role for rne mutants in iron-mediated bacterial virulence. A subset of commonly upregulated genes shared by the AzEvC10 mutant 6 hpi and sputum P. aeruginosa are known to be expressed under anaerobic conditions, including the nir and nor operons, cytochrome C oxidase subunits (ccoP2 and ccoQ2), arginine deaminase (arcA), and malate/L-lactate dehydrogenase (dpkA) (Fig 7D and S2 Table). This switch to anaerobic metabolism contributes to increased fitness of P. aeruginosa under hypoxic conditions. The open reading frame PA3271 was also overexpressed in both AzEvC10 at 6 hpi and sputum and encodes a probable two-component sensor. Because more than 50% of two-component systems have been implicated in virulence and in vivo fitness and colonization ability [15], this suggests that the sensor may be important for P. aeruginosa responses to macrophages. Hence, our results show that the pathoadaptive changes seen in AzEvC10 during macrophage infection share similarities with P. aeruginosa behavior during pulmonary infections.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Virulence and metabolic adaptation are common features of P. aeruginosa pulmonary infections.

A. Schematic representation of the experimental design for RNA and whole genome sequencing of P. aeruginosa clinical isolates. B. Individual expression of iron acquisition genes in CF sputum (n = 12) and in vitro grown clinical isolates (n = 8) expressed in normalized number of reads (TPM) C-D. Expression of selected iron and metabolic genes commonly upregulated by clinical isolates in sputum and by the AzEvC10 mutant during BMDM infection at 3 hpi (C) and 6 hpi (D) (n = 2–5). E. List of non-synonymous mutations detected in the rne gene in P. aeruginosa clinical isolates. F. Representation of the genomic location of rne mutations in clinical isolates. Fisher’s exact test was used to compare the prevalence of mutations in amino-terminal vs carboxy-terminal halves (p<0.0001) (S5 Table).

https://doi.org/10.1371/journal.ppat.1010942.g007

Clinical isolates evolve multiple mutations in the rne gene

The rne gene encodes RNase E endonuclease, a 1,057-residue protein involved in RNA processing and decay [42]. Because the virulent phenotype of AzEvC10 mutant is caused by a 50-bp deletion in the of 3’ end of the rne gene encoding the enzyme’s carboxy-terminus, we assessed if mutations in this region were present in human clinical isolates. We performed whole genome sequencing on P. aeruginosa isolates from both CF (n = 16) and non-CF (n = 29) respiratory infections (Fig 7A). We found that all isolates (n = 45) had multiple non-synonymous mutations in the rne gene (Fig 7E). All non-synonymous mutations except for two were found in the second half of the gene, which encodes the carboxy-terminal half of the RNase E protein (p<0.0001) (Fig 7F). Based on our observations in the AzEvC10 mutant during macrophage infection, these results suggest that rne variations evolved in P. aeruginosa clinical isolates may have contributed to their colonization and/or fitness during pulmonary infections. However, because clinical isolates likely have additional mutations in other genes, it is difficult to directly demonstrate the role of rne mutations on the gene expression and fitness of these clinical isolates. Further investigation will be needed to determine which variants affect RNase E functions and bacterial phenotypes during infection.

Bacterial strains and growth conditions

Bacterial strains and plasmids are listed in S3 Table. P. aeruginosa PAO1 AzEvC10 and AzEvC10 pMQ72::nalD mutants were obtained from Pradeep K. Singh [26]. The rne Tn mutant (rne::Tn) was obtained from Colin Manoil’s laboratory at the University of Washington and corresponds to the PW5993 mutant containing the genotype PA2976H05:: ISphoA/hah [29]. This mutant was grown from freezer stock on Lysogeny broth (LB) agar (BD, cat# 244520) with 10 mg/mL tetracycline. All strains were maintained at 37°C in LB (BD, cat# 244620) unless otherwise specified. Where necessary, gentamicin (Gm) was added to growth medium at the following concentrations: 10 μg/mL for E. coli and 30 μg/mL for P. aeruginosa.

P. aeruginosa clinical isolates

P. aeruginosa clinical isolates from CF (n = 11) or non-CF (n = 29) ICU patients with pulmonary infection were a generous gift from the Pulmonary Translational Research Core at University of Pittsburgh.

Mutant constructions

Gene deletion mutants listed in S3 Table were generated with suicide plasmids as described previously [88]. Suicide plasmids carrying mexAB deletion, nalD_T158P point mutation, 50bp deletion in rne gene, or WT rne gene were constructed in a similar way. First, two PCR fragments were generated from chromosomal DNA using corresponding primer pairs to the up- and down-stream regions of each gene (see S4 Table for primer nucleotide sequences). The two fragments were then assembled into the suicide vector pEX18Gm between using NEBuilder HiFi DNA Assembly Cloning Kit (cat# E5520). Suicide plasmids were then transformed into E. coli DH5α according to the NEB protocol and verified by Sanger sequencing. P. aeruginosa strains were transformed by electroporation as published previously [89] and incubated on LB + Gm 30 μg/mL at 37°C overnight. For counter selection, isolated clones were streaked on low-salt LB containing 15% sucrose as published [88] and incubated at room temperature for 48 h. Gene deletions were confirmed by PCR using primers listed in S4 Table and whole genome sequencing.

DNA extraction, purification, and PCR

Plasmid DNA was prepared using a Monarch Plasmid Miniprep Kit (NEB, cat# T1010). Genomic DNA was prepared using the DNeasy Blood & Tissue Kit (Qiagen, cat# 69504). When necessary, cDNA was purified using the Monarch PCR & DNA Cleanup Kit (NEB, cat# T1030). PCR was performed using either KAPA HIFI 2X ready mix (KAPA Biosystem, cat# KK2602) or Expand Long Template PCR System (Sigma, cat# 11681842001). Primers used for each PCR reaction are listed in S4 Table.

Aztreonam susceptibility testing

Etest strips were purchased from Biomerieux Vitex Inc (cat# 501758) and antimicrobial susceptibility testing was performed with the following modifications to the manufacturer’s instructions. A sterile swab was soaked in an overnight culture for each strain after growth for 18h in LB broth and excess fluid was removed by pressing it against the inside wall of the test tube. Mueller Hinton agar plates were fully streaked 4 times with the swabs. After allowing the plates to dry, an Etest gradient strip was placed in the middle of each plate. Plates were incubated at 37°C for 16–20 h and MICs were read by zone of clearing around the strip.

BMDM isolation and phenotype verification

Femurs and tibias from C57BL/6 mice were harvested and cleaned up with 70% ethanol. Working in a biological safety cabinet, bones were washed with 1X PBS. Bone marrow was flushed with 10 mL of BMDM growth medium pH 7.6 (RPMI 1640 containing 20% L929 conditioned medium, 10% fetal bovine serum [FBS, Omega Scientific cat# FB-11] and antibiotic-antimycotic). Antibiotic-antimycotic solution was purchased from ThermoFisher (cat# 15-240-062) and used at a final concentration of 1% (100 units/mL penicillin, 100μg/mL streptomycin, and 2.5 μg/mL fungizone). Red blood cells were lysed in 5 mL 1X RBC lysis buffer for 3 m at room temperature. Lysis was stopped by adding 30 mL of 1X PBS. Cells were pelleted, counted and plated on treated tissue culture dishes. After 24 h, non-adherent cells were plated into a petri dish and grown in BMDM growth medium at 37°C and 5% CO₂. Fresh medium was added after 3 d. After 7 d, cells were plated for experiments. BMDM phenotypes were confirmed for their positivity for the alveolar macrophage markers CD64 and Siglec F. Briefly, cells were washed with 1X PBS and then harvested with 1mL 5mM EDTA-1X PBS and centrifuged at 8000rpm for 1 m. Supernatants were removed, and cell pellets were washed with 1 mL FC buffer (3% FBS in 1X PBS). BMDM were resuspended in 50 μL blocking buffer (1:25 FC block [BD cat# 553142] in FC buffer) and incubated on ice. After 20 m, 50 μL of the 2% FBS in 1X PBS containing 0.25 μL of the following antibodies was added and incubated on ice for 30 m: CD45 (Biolegend cat# 103137), CD64 (Biolegend cat# 139307) and Siglec F (BD cat# 562681). Cells were then washed in cold 1X PBS and resuspended in 300 μL 1X PBS for flow cytometry analysis.

Bone marrow-derived neutrophil isolation

Femurs and tibias from C57BL/6 mice were harvested and cleaned up with 70% ethanol. Working in a biological safety cabinet, bones were washed with 1X PBS. Bone marrow was flushed with 10 mL of growth medium (RPMI 1640 containing 10% FBS). The EasySep Mouse Neutrophil Enrichment Kit (StemCell, cat# 19762) was used to isolate neutrophils according to the manufacturer’s instructions. Neutrophils were used the same day of isolation.

Mouse airway epithelial cells

Immortalized mouse airway epithelial cells (MLE-12) were a gift from Dr. Peter Chen (Cedars-Sinai Medical Center, Los Angeles). Cells were grown in Dulbecco’s Modified Eagle Medium containing 10% FBS (Omega Scientific cat# FB-11).

L929 conditioned medium

L929 mouse fibroblasts were grown in Iscove’s Modified Dulbecco’s Medium (ThermoFisher, cat# 12-440-061) with 10% FBS. Cells were plated at a density of 4 x10⁶ cells/150 mm tissue culture treated dish and culture for 10 d at 37°C, 5% CO₂. Medium was then collected, filtered through a 0.22 μm filter and frozen at -80°C in 50 mL aliquots.

Alveolar macrophage isolation and infection

Wild-type C57BL/6 mice were anesthetized using 4% isoflurane. Cardiac perfusion was performed with sterile 1X PBS. Then, the trachea was exposed and punctured, and a 20G intratracheal tube was inserted into the trachea. A 1mL syringe filled with sterile 2mM EDTA + 2% FBS in 1X PBS was attached to the tube and gently injected into the lungs. The volume was slowly aspirated and transferred on 15 mL tube on ice. This was repeated 7 additional times for a total of 8 mL. The BAL was centrifuged and resuspended in macrophage growth medium (RPMI 1640 containing 20% L929 conditioned medium and 10% fetal bovine serum [FBS, Omega Scientific cat# FB-11]). Cells were incubated overnight at 37°C and 5% CO₂. The next day, alveolar macrophages were visually confirmed to be adherent. Alveolar macrophages were then infected with WT or AzEvC10 mutant P. aeruginosa MOI: 100 for 6 h and the bacterial loads were determined by CFU counts as described below.

Cell infection with P. aeruginosa

BMDM, neutrophils, or AEC were plated with growth medium without antibiotic-antimycotic for all experiments. Bacterial strains were streaked on LB agar and incubated at 37°C. The day before the infection, bacterial strains were grown in LB broth at 37°C overnight shaking at 250rpm. The day of infection, strains were diluted and grown to mid-exponential phase OD₆₀₀ 0.4–0.6. Bacteria were then pelleted, washed with 1X PBS and resuspended in 1X PBS. BMDM were infected with a MOI: 100 for 3 h or 6 h, or with a MOI: 1 for 24 h. For intracellular survival experiments, BMDM were infected for 1 h, then amikacin was added to a final concentration of 200 μg/mL and left for an addition 5 h of incubation. For the escape experiments, antibiotics were added after 1 h incubation and left for 1 h. Cells were then washed with 1X PBS and fresh BMDM medium without antibiotic was added and cells were incubated for a total of 3 h or 6 h.

Immunofluorescence

BMDM were infected with a MOI: 100 for 6 h. Cells were then washed with 1X PBS and fixed with 2% PFA for 10 m at room temperature. After 3 washes with 1X PBS, cells were permeabilized in 1X PBS + 5% Goat serum + 0.1% Triton X-100 for 5 m at room temperature. Anti-CD68 (BioRad, cat#MCA1957, 1: 250) and anti-P. aeruginosa (ThermoFisher, cat# MA183430, 1:500) primary antibodies were diluted in 1X PBS + 2% Goat serum. Cells were incubated with primary antibodies for 1h at room temperature, washed 3 times with 1X PBS, then incubated for 30–45 m with goat anti-mouse Alexa Fluor 555 (Invitrogen cat#A32727, 1:1000) or goat anti-rat Alexa Fluor 488 (Invitrogen cat#A11006, 1:1000) secondary antibodies. After 3 washes with 1X PBS, cells were mounted with Fluoromount G mounting medium with DAPI (ThermoFisher, cat# 00-4959-52). Images were taken using a Zeiss Apotome fluorescence microscope and analyzed with Bitplane Imaris software version 9.2.1.

Cytotoxicity assay

Cytotoxicity assays were performed using the Pierce LDH Cytotoxicity Assay Kit (ThermoFisher, cat# 88954) according to manufacturer’s instructions.

Reactive oxygen species assay

Infected cells were washed with 1X PBS and then harvested with 1mL 5mM EDTA-1X PBS and centrifuged at 8000rpm for 1 m. Supernatants were removed, and cell pellets were washed with 1 mL FC buffer (3% FBS in 1X PBS). BMDM were resuspended in 10 μM 2′,7′-dichlorodihydrofluorescein diacetate dilution (Sigma cat# D6883) and incubated for 30 m at 37°C protected from light. Cells were then washed in 1 mL FC buffer and resuspended in FC buffer and immediately analyzed by flow cytometry (Becton Dickinson LSR Fortessa). Mean fluorescence intensity (MFI) was quantified using FlowJo software (version 10.6.1, FlowJo LLC).

Lipid peroxidation assay

Infected cells were washed with 1X PBS and then harvested with 1 mL of 5 mM EDTA-1X PBS and centrifuged at 8000rpm for 1 m. Supernatant was removed, and cell pellets were washed with 1 mL FC buffer (3% FBS in 1X PBS). The Lipid Peroxidation Assay Kit (Abcam cat# ab243377) was used and cells were stained according to the manufacturer’s instructions. Cells were then washed in 1 mL Hank’s Balanced Salt Solution (HBSS) and resuspended in HBSS buffer and immediately analyzed by flow cytometry (Becton Dickinson LSR Fortessa). Mean fluorescence intensity (MFI) was quantified using FlowJo software (version 10.6.1, FlowJo LLC). Lipid peroxidation was quantified by calculating the red (Ex561/Em582)/green (Ex488/Em525) fluorescence ratio using FlowJo software (version 10.6.1, FlowJo LLC). Data are presented as the reciprocal of the ratio (1/ratio) and normalized on WT PAO1 at 3 h where specified.

Cell death assay

Infected cells were washed with 1X PBS and then harvested with 1 mL of 5 mM EDTA-1X PBS and centrifuged at 8000rpm for 1 m. Supernatant was removed, and cell pellets were washed with 1 mL FC buffer (3% FBS in 1X PBS). To quantify cell death, TACS Annexin V-FITC Apoptosis Detection Kit (R&D Systems cat# 483001K) was used and cells were stained according to the manufacturer’s instructions. Cells were then washed with the binding buffer and resuspended in binding buffer and immediately analyzed by flow cytometry (Becton Dickinson LSR Fortessa). The percentage of double positive cells and MFI were quantified using FlowJo software (version 10.6.1, FlowJo LLC).

CFU counts to determine total and intracellular bacterial burdens

BMDM were permeabilized with 0.1% Triton X-100 for 5 m, then cells were collected using a cell scraper. For each condition, 10 μL was used to make a serial dilution and plated in duplicate on LB agar. Plates were incubated at 37°C overnight. For each condition, CFUs were counted and reported as CFUs/mL.

Pyochelin and pyoverdine measurements

For each condition, BMDM were permeabilized with 0.1% Triton X-100 for 5 m and 100 μL of infected cell supernatant was transferred in duplicate into a black clear-bottom 96-well plate. The plate was read using a Varioskan Lux plate reader at Ex350/Em430nm (pyochelin) and Ex400/Em460nm (pyoverdine). For each sample, the medium fluorescence (background) was subtracted from the raw fluorescence values and the subtracted fluorescence values (RFU) were normalized on log₁₀(CFU).

Metal complexes with pyochelin and pyoverdine for BMDM treatments

Pyochelin I & II (Toronto Research Chemicals cat# P840365) and pyoverdine (Sigma cat# P8124) were purchased and used for the BMDM experiments. For pyochelin, 100 μM or 1 mM pyochelin was incubated with either 50 μM or 500 μM iron(III) chloride (Fe(III)) respectively, or 50 μM or 500 μM gallium(III) nitrate hydrate respectively, in 50mM Tris-HCl buffer pH 7.6 for 16h at room temperature. For pyoverdine, 100 μM or 1 mM pyoverdine was incubated with either 100 μM or 1 mM Fe(III) respectively, or 100 μM or 1 mM gallium(III) nitrate hydrate respectively, in 50mM Tris-HCl buffer pH 7.6 for 16h at room temperature. The solutions were then diluted 1:10 in macrophage growth medium, filtered sterilized and added to BMDM for 6 h.

Pyochelin and pyoverdine fluorescence spectrum

For pyochelin spectra, 1 mM pyochelin was incubated with either 500 μM Fe(III) or 500 μM gallium(III) nitrate hydrate in 50mM Tris-HCl buffer pH 7.6 for 16h at room temperature. For pyoverdine spectrum, 1 mM pyoverdine was incubated with either 1 mM Fe(III) or 1 mM gallium(III) nitrate hydrate in 50mM Tris-HCl buffer pH 7.6 for 16h at room temperature. The solutions were then diluted 1:10 in macrophage growth medium and 100 μL of each solution was transferred in duplicate into a black clear-bottom 96-well plate. The plate was read using a Varioskan Lux plate reader at Ex350 for pyochelin and Ex400 for pyoverdine.

Labile iron pool quantification

To quantify the labile iron pool in infected BMDM, we used a colorimetric method described by Abbasi et al. [33] with the following modifications. Briefly, BMDM were washed with 1X PBS and harvested in 5mM EDTA 1X EDTA. The supernatant was removed and cells were lysed in 200 μL Pierce RIPA buffer (ThermoFisher Scientific cat# 89900) with sonication for 20 min. After lysis, samples were centrifuged at 16,000 g for 10 min at 4°C and supernatants were transferred into clean tubes for further analysis. Iron standards were prepared from Fe(III) in 2% nitric acid, ranging from 0 to 500 μM. Samples (100 μL) and iron standards (100 μL) were transferred into different clean tubes. Ammonium acetate buffer (pH 4.5, 2.5 M) (100 μL) and labile iron working solution (5 mM ferene [Sigma cat# 82940] and 10 mM ascorbic acid [Sigma cat# 255564] prepared in ammonium acetate buffer pH 4.5, 2.5 M) (120 μL) were added to all tubes. The mixtures were then vortexed and incubated overnight at room temperature. The next day, the samples and standards were spun at 16,000 g for 5 m, and 200 μL were transferred into a 96 well plate for absorbance measurements. Absorbance was recorded at 595 nm using a Varioskan Lux plate reader. Iron concentrations were interpolated from the standard curve generated from the iron standards. These concentrations were normalized to the amount of protein analyzed (i.e., nmol of iron per mg of protein).

Protein quantification

Protein content in cell lysates were measured using the BCA Total protein assay kit (ThermoFisher Scientific cat# 23227), following the manufacturer’s protocol. In brief, working solution was prepared using reagent A and B at a 50:1 ratio, respectively. BSA standards (10 μL) and cell lysates (10 μL) were added into a 96 well plate. Then, 200 μL of Bradford working solution were added to each well. This was kept at 37°C for 30 min, then cooled to room temperature for 5 min. Absorbance was read at 562 nm. Protein concentrations in cell lysates were interpolated using the standard curve generated.

Growth curve in gallium

Overnight cultures were diluted to reach OD₆₀₀~0.005. Ten microliters of these cultures were inoculated in 96-well plates in LB broth containing different concentrations of gallium (III) nitrate hydrate (Sigma cat# 289892). Fifty microliters of mineral oil was added in each well to prevent evaporation. Plates were incubated at 37°C for 20 h, shaking for 5 s every hour and absorbance at 600 nm was recorded using a Varioskan Lux microplate reader (Thermo Fisher Scientific, cat# NC1141718).

BMDM treatment with gallium for BMDM toxicity

Gallium was added to uninfected BMDM to final concentrations of 500, 750, 1000 and 2000 μM for 6 h. Cytotoxicity was performed using a LDH cytotoxicity assay as described above.

BMDM treatment with gallium during infection

BMDM were infected as described above. For cytotoxicity assays, gallium was added immediately after infection to final concentrations of 100, 200, 375, 500, and 750μM for 6 h. For CFU counts, gallium was added immediately after infection to a final concentration of 750 μM for 3 h, 6 h, or 24 h.

RNA-seq library preparation

BMDM were infected with P. aeruginosa strains at MOI: 100 for 3h or 6h. Cells were washed and preserved in RNAlater (Invitrogen cat# AM7021) until RNA extraction. Total BMDM and bacterial RNA was extracted using a RNeasy Mini Kit (Qiagen, cat# 74106) and resuspended in 30 μL of elution buffer. Of this volume, 25 μL was used for bacterial RNA libraries and 5 μL was used for BMDM RNA libraries. For bacterial RNA libraries, RNA from host nucleic acids and bacterial rRNA was depleted using MicrobEnrich Kit (Invitrogen, cat# AM1901) and MicrobExpress Bacterial mRNA Enrichment Kit (Invitrogen, cat# AM1905) respectively, according to manufacturer’s instructions. RNA-seq libraries were prepared for bacteria using the KAPA RNA HyperPrep Kit (KAPA BioSystem, cat# KK8540) and BMDM using the KAPA mRNA HyperPrep Kit (KAPA BioSystem, cat# KK8580), according to manufacturer’s instruction. Libraries were quantified using Quant-iT dsDNA Assay Kit, high sensitivity (Invitrogen, cat# Q33120) and the quality was verified on an Agilent 4150 TapeStation System using D1000 High Sensitivity DNA Tape (Agilent, cat#5067–5584). Libraries were pooled to equimolar amounts and sequenced using a NextSeq 500 instrument to a depth of 400 million reads per sample (75 bp, single-ends) at the Cedars-Sinai Genomics Core Facility.

RNA-seq data analyses

For clinical isolates, publicly available data was analyzed (Accession No. PRJEB24688) [39]. RNA sequencing data from WT and AzEvC10 PAO1 is available in the NCBI Sequence Read Archive (Bioproject Accession # PRJNA934930). RNA-seq data were analyzed using CLC Genomics workbench 20 (version 20.0.1). The P. aeruginosa PAO1 genome (Genbank accession #GCA_000006765.1) was as a reference for RNA sequencing read alignment. Differential Expression for RNA-Seq (version 2.2) analysis was performed to compare whole transcriptomes between the following groups: WT PAO1 vs AzEvC10 mutant at 3 hpi and 6 hpi, each strain during BMDM infection at 3 hpi and 6 hpi vs LB growth, and CF sputum vs in vitro clinical isolates. Reads were normalized in transcripts per million reads (TPM). Gene Set Test (version 1.1) using Gene Ontology (GO) biological processes annotations was performed to reveal enriched biological processes in the AzEvC10 mutant after 3 h and 6 h of BMDM infection. Venn Diagram for RNA-Seq tool (version 0.2) analysis was performed using a minimal absolute fold-change of 1.5 and false discovery rate (FDR) p-value of 0.05. The lists of DEGs are available in S1 and S2 Tables.

Whole genome sequencing of clinical P. aeruginosa isolates

P. aeruginosa clinical isolates from CF (n = 11) and non-CF (n = 29) ICU patients were a generous gift from the Pulmonary Translational Research Core at University of Pittsburgh. These isolates were sent for whole genome sequencing to the Microbial Genome Sequencing Center (MiGs). Genome sequencing data is available in the NCBI Sequence Read Archive (Bioproject Accession # PRJNA934930). These samples were analyzed alongside publicly available whole genome sequencing data from CF sputum (n = 5) (Accession No. PRJEB24688) [39]. Variant calling program Breseq version 0.35.5 was utilized to identify mutations in the isolates’ genomes relative to the PAO1 reference (Genbank accession #GCA_000006765.1) [90]. The identified mutations were filtered for indels and non-synonymous SNPs and then subsequently filtered to identify unique mutations found only in the rne gene.

Statistical analyses

Normality and homogeneity of variance were assessed by Shapiro-Wilk and Brown-Forsythe tests, respectively. Data was log-transformed prior to analysis where necessary to meet assumptions necessary for parametric testing, else non-parametric rank testing was used. Based on data distributions, analysis was performed for matched data with paired Student t-test or Wilcoxon matched pairs rank test; for two groups with a two-sample Student t-test or the nonparametric Mann-Whitney; for three or more groups with ANOVA followed by Tukey’s test or the non-parametric Kruskal-Wallis test with Dunn’s correction. Where data were collected across experimental repeats or repeated measures, data were considered as matched sets, and RM-ANOVA was used. Cytotoxicity dose curves were analyzed with two-way ANOVA. For all parametric testing, residuals were inspected to confirm fit. Fisher’s exact test was used to analyze categorical data. All testing was considered significant at the two-tailed p-value of <0.05. Analysis performed with GraphPad Prism v9. The p-values are listed in S5 Table.