PA3049 homologs are widespread in γ-proteobacteria

To uncover the role of PA3049 (BatR) in P. aeruginosa, we first analysed the distribution of batR-like genes in bacterial genomes, and compared its predicted structure to E. coli RMF, which shares ~49.09% sequence similarity with PA3049. Alignment of batR homologs from different Pseudomonas species with the RMF sequence from E. coli shows a 15-residue C-terminal extension encoded by batR that is absent from RMF (Fig 1a). AlphaFold3 [35] three-dimensional protein structure prediction of P. aeruginosa BatR highlighted a marked difference in the C-terminal fold compared to its E. coli homolog. Specifically, BatR contains a predicted 12 residue α-helix at the C terminus (red box), in addition to the two α helices connected by a 13-amino acid linker region that characterise E. coli RMF (Fig 1b).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. BatR is phylogenetically distinct from RMF.

a. Sequence alignment of BatR and RMF. ClustalW alignment of RMF from E. coli str. K-12 substr. MG1655 and BatR from 20 different Pseudomonas spp. Different colours mark conserved residues in all 21 proteins. The 15-residue C-terminal extension on BatR sequences that is absent from RMF is shown in the red box. b. AlphaFold 3 model of PAO1 BatR (magenta), overlaid onto the structure of E. coli RMF (cyan). The additional alpha-helix at the C-terminus of the predicted BatR structure is indicated in red. c. The phylogenetic relationship between BatR/RMF homologs. The tree is based on 765 γ-proteobacteria homologous and was constructed using the maximum likelihood method, with 100 bootstrap replicates.

https://doi.org/10.1371/journal.ppat.1013832.g001

A phylogenetic analysis of batR homologs revealed that these genes are confined to the γ-proteobacteria class (Fig 1c), with distinct evolutionary paths across diverse bacterial orders.

Notably, batR in Pseudomonadales forms a distinct cluster that diverges significantly from homologs in Enterobacterales, including E. coli and Yersinia pestis; and Vibrionales, including Vibrio cholerae, where RMF plays a role in ribosome hibernation [36]. Unlike other characterised RMF genes, batR deletion mutants do not exhibit impaired ribosomal integrity during starvation [32]. Thus, while rmf and batR are likely to share an ancestral root, their divergent structure and phylogeny (Fig 1) are consistent with an alternative functional role for BatR [32].

batR protects P. aeruginosa biofilms from sub-inhibitory concentrations (SIC) of antibiotics in vitro

Given the heightened abundance of batR transcripts in dormant P. aeruginosa cells within biofilms [11], we assessed the importance of batR in biofilm-linked antimicrobial sensitivity in vitro. To do this, we generated a non-polar deletion mutant (ΔbatR) in P. aeruginosa PAO1 (hereafter, PAO1) and examined its response to antimicrobial agents when grown under biofilm conditions. Assays were conducted using antibiotics targeting distinct metabolic processes, i.e., β-lactams (piperacillin, PIP) and quinolones (ciprofloxacin, CIP).

The minimal inhibitory concentration (MIC) [37] of the tested antibiotics for cells grown in liquid culture was unaffected in ∆batR, (64 µg/mL for piperacillin and 0.5 µg/mL for ciprofloxacin). Interestingly however, ΔbatR showed a small but significant increase in sensitivity to both PIP and CIP (12–17%), when grown on a solid surface (Fig 2a). We further confirmed the previously observed effect of the aminoglycoside tobramycin (TOB) ([11], referred to as rmf), with ΔbatR again showing a significant increase in sensitivity on agar (Fig 2a). Next, we used a glass bead biofilm model [38], to assess the survival of cells within established biofilms challenged with PIP. This model revealed a substantial reduction in the survival of established ΔbatR biofilms compared to WT PAO1, for samples exposed to a sub-inhibitory concentration (SIC) of PIP (Fig 2b). Furthermore, the ΔbatR phenotype could be fully rescued by the heterologous expression of batR (Fig 2b). Growth curves confirmed that the presence of the plasmids pME-empty and pME-batR did not affect growth of the strains WT and ΔbatR in shaking cultures (S1a Fig.).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. batR influences biofilm-associated antimicrobial sensitivity.

a. Antibiotic disk diffusion assay. Results are shown as the normalized width of the antimicrobial halo (NW halo), calculated as described in [39]. Results were analysed by a one-way ANOVA and inhibition differed significantly between WT and ∆batR strains (F_5,24 = 35.99, p < 0.0001). Tukey's multiple comparisons between WT and ∆batR strain under PIP, CIP and TOB treatments are indicated (p < 0.001 **, p < 0.0001 ***). b. Glass Beads Biofilm survival. Bacterial recovery (CFU/bead) from established biofilms grown on glass beads for 24h following treatment with SIC of PIP for 90 min. Strains used were WT PAO1 carrying the empty vector pME6032 (WT-pME-Empty), ∆batR carrying pME6032 (∆batR-pME-Empty), and ∆batR overexpressing batR (∆batR-pME-batR). Normalised values represent the estimated proportion of cells surviving PIP exposure, as described previously [38]. Results were analysed by a one-way ANOVA showing significant differences (F2,12 = 6.414; p < 0.05 *). Tukey's multiple comparisons between WT-pME6032 (PIP) and ∆batR-pME6032 (PIP), and between ∆batR-pME6032 (PIP) vs ∆batR-pME-batR (PIP) are indicated (p < 0.05 *).

https://doi.org/10.1371/journal.ppat.1013832.g002

Since the antibiotics tested in this study have different modes of action, our results suggest that BatR affects P. aeruginosa antimicrobial sensitivity through a general, rather than drug-specific mechanism. A simple explanation was a nonspecific change in membrane permeability. However, we ruled this out by measuring the intracellular concentration of resazurin, a fluorescent dye used to assess permeability and efflux (S2 Fig), which showed little difference between WT and ∆batR strains.

BatR induces specific changes in the PAO1 proteome

To investigate the physiological changes associated with BatR in P. aeruginosa biofilms, we conducted a comparative proteomic analysis between the PAO1 WT and ΔbatR strains. Strains grown on agar plates have been shown to function as organised and coordinated bacterial communities that share multiple characteristics with biofilms [10]. Therefore, whole-cell lysates were analysed by TMT proteomics following growth as a lawn on solid medium. On average, 4,581 individual proteins were detected per sample (S1 es), representing ~80% of the predicted total P. aeruginosa PAO1 proteome [40]. Surprisingly, the overall differences between WT and ΔbatR proteomes were limited (S1 Data), suggesting that the impact of batR deletion is quite specific under the conditions tested.

Of the 21 proteins decreased in the ΔbatR strain (Table 1), seven (PA0617-PA0633) comprise the R2/F2 pyocin gene locus [41]. In addition, the loss of batR led to reduced abundance of proteins involved in transport & metabolism, two heat shock proteins, and four proteins of unknown function. Conversely, the loss of batR increased the abundance of 14 proteins, including Rubredoxin-1, Type VI secretion system components and the Phenazine-1-carboxylate N-methyltransferase PhzM (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. COG pathway analysis of proteins decreased in the ΔbatR strain (TMT proteomics).

https://doi.org/10.1371/journal.ppat.1013832.t001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. COG pathway analysis of proteins increased in the ΔbatR strain (TMT proteomics).

https://doi.org/10.1371/journal.ppat.1013832.t002

In P. aeruginosa PAO1, the R2/F2 pyocin cluster encodes the structural components for R2 and F2 pyocins, a shared regulatory region and a lytic cassette made up of four proteins (S3a Fig). Among these, the endolysin Lys (PA0629) and the holin (PA0614) are essential for explosive cell lysis, a process in which a subset of cells transition from rod-shaped to round then lyse, releasing intracellular contents including genomic DNA, critical for early biofilm development [23,42]. TMT proteomics revealed increased levels of several R2-type pyocin proteins (PA0617, PA0618, PA0619, PA0620, PA0623 and PA0626) and an F2-type pyocin (PA0633) in the WT strain, suggesting BatR-dependent regulation of the pyocin cluster (S3a Fig).

To validate these assays, the promoter activity of PA0614, the first gene of the R2/F2 pyocin gene cluster, was measured (S3b Fig). We use the pM0614-G plasmid, in which the PA0614 promoter is fused to an egfp reporter, and a promoterless pMEXGFP plasmid as a negative control [43]. In the ΔbatR mutant, PA0614 promoter activity was significantly decreased compared to the WT, supporting our proteomic results that BatR acts as a positive regulator of pyocin production (whether directly or indirectly). Then, to directly analyse the effect of BatR on explosive lysis, we performed phase-contrast time-lapse microscopy using a microfluidic device [44]. We quantified lysis events based on the transition of rod-shaped cells into round morphotypes followed by lysis over a 3h period. Under these conditions, no significant differences in lysis frequencies were observed between WT and ΔbatR strains (S3b, c Fig). Nonetheless, we cannot rule out a BatR-dependent effect on explosive lysis under different growth conditions. BatR appears to function primarily in biofilms, and may not be present or active under the low-density, flow-based conditions of the microfluidic setup.

BatR interacts with a putative stress response kinase

To understand the molecular basis of BatR function, we next investigated its interactions with other PAO1 proteins by performing a co-immunoprecipitation (Co-IP) analysis (S2 Data). BatR-3xFlag co-immunoprecipitated alongside several cytoplasmic proteins, suggesting potential direct regulatory influences (Table 3). Independent validation of strongly co-precipitating proteins was conducted using Bacterial Two-Hybrid (B2H) analysis. This confirmed specific interaction between BatR and the predicted kinase PA0486, annotated as Stress response kinase A (SrkA) [45](Fig 3a).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. BatR interacting proteins (Co-IP assay).

https://doi.org/10.1371/journal.ppat.1013832.t003

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. BatR interacts with a predicted kinase protein.

a. Representative image of qualitative β-galactosidase assays on agar plates. pKT25 and pUT18C fusions are shown in rows and columns, with the indicated protein/empty vector present in each case. Positive control (+): pKT25-zip and pUT18C-zip encoding the two adenylate cyclase fragments, T25 and T18 [46]. b. Predicted interface between the Alphafold3 models of BatR (magenta) and SrkA (green). The conserved residues Ser33 (S33), predicted ATP-binding site, and Asp216 (D216), predicted Mg2 + -binding site, in SrkA are shown.

https://doi.org/10.1371/journal.ppat.1013832.g003

To gain additional insight into the potential interaction between BatR and SrkA, we used AlphaFold3 [35] to model the potential interactions of BatR with the target protein (Fig 3b). Whilst the Alphafold3 interaction prediction for BatR had low confidence (0.270), the prediction did indicate a potential interaction with the active site of SrkA, including the residue D216, predicted to be the Mg²⁺ binding site.

SrkA influences biofilm formation, cell death, and pyocyanin production in P. aeruginosa

SrkA (PA0486) is annotated as a eukaryotic-like serine-threonine protein kinase. Its homolog in E. coli is implicated in protecting cells from stress-induced programmed cell death [45,47]. Since SrkA remains uncharacterized in P. aeruginosa, we overexpressed srkA in both the WT and ΔbatR PAO1 strains to assess its functional role. To test whether the observed phenotypes are dependent on SrkA’s kinase activity, we also overexpressed a catalytically inactive (“null”) version of srkA, in which the conserved Ser33 and Asp216 residues were substituted with alanine. Alphafold3 analysis of the predicted interaction between BatR and SrkA suggests that BatR binding may interfere with Ser33, which is predicted to be the ATP-binding residue (Fig 3b).

Overexpression of srkA led to an increase in biofilm formation in both solid (Congo red binding) and liquid (Crystal Violet staining) media (Fig 4a, b), as well as an increase in pyocyanin production (Fig 4d). This increase in biofilm was independent of the SrkA kinase activity, with the overexpression of the SrkA null mutant showing identical phenotypes to the WT version of SrkA. Overexpression of the kinase-inactive variant also produced a distinctive, highly aggregative colony phenotype on solid media and the emergence of multiple escape mutants, manifesting as regions of reduced dye binding and WT PAO1 morphology (Fig 4a).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Phenotypic effects of srkA overexpression in PAO1 WT and ∆batR strains.

a. Congo-red binding. Representative images of WT-pME-Empty, WT-pME-srkA, WT-pME-srkA null, ∆batR-pME-Empty; ∆batR-pME-srkA and ∆batR-pME-srkA null biofilms grown using the colony biofilm morphology assay. Only plates supplemented with IPTG are shown (7 days). Escape mutants are indicated with red arrows. b. Biofilm formation. Strains were grown statically in LB medium for 24h at 37°C. Biofilm biomass was quantified by Crystal Violet staining and measured spectrophotometrically at 590 nm (A_590nm). Values represent the mean of five biological replicates with two technical replicates each; error bars indicate SD. c. Cell viability. Cells were scraped from LB agar plates, resuspended in PBS, and enumerated via serial dilution and plating. d. Pyocyanin production. Top view images of cell lawns showing pyocyanin production (blue-green colouration). Results were analysed by a one-way ANOVA showing significance between strains in b (F_5,54 = 45.17, p < 0.0001) and c (F_7,32 = 28.01, p < 0.0001). Tukey's multiple comparisons between strains are indicated (p < 0.05 *, p < 0.001 **, p < 0.0001 ***).

https://doi.org/10.1371/journal.ppat.1013832.g004

Curiously, the overexpression of srkA led to a kinase-independent significant increase in cell death, with even a small induction of SrkA leading to a 10³-fold reduction in cell viability. Growth curves confirmed this srkA-killing effect in shaking cultures, with impaired growth in strains overexpressing srkA and more pronounced effect in strains overexpressing the null version of srkA (S1 b Fig.).

Given the link between BatR and the R2/F2 pyocin cluster suggested by the proteomic data, we hypothesised that the srkA-mediated killing effect occurs through a similar mechanism. To test this, we overexpressed srkA in a mutant deficient in production of the endolysin Lys (PA0629) [23]. While the impact of srkA expression was markedly reduced from that seen in WT PAO1, the ∆lys pME-srkA mutant still exhibited an approximately 10²-fold reduction in viable cell counts compared to the empty vector control (Fig 4c). This suggests that srkA induced cell death is partially endolysin-dependent but also implicates an additional lethal pathway that operates independently of the R2/F2 pyocin lysis system.

We next generated a non-polar deletion mutant (ΔsrkA) and the double mutant (ΔbatR ΔsrkA) in P. aeruginosa PAO1 and examined their phenotypes. Interestingly, despite SrkA’s dramatic effect when overexpressed, ΔsrkA showed no phenotypic differences from WT at baseline under the conditions tested. However, ΔbatR ΔsrkA displayed several strong phenotypes: reduced biofilm formation, increased cell death and increased pyocyanin production (S4 Fig). Growth curves confirmed that the deletion of batR and srkA did not result in growth impairment in shaking cultures (S4 c Fig.). The fact that these traits emerge only in the absence of both genes is consistent with an additional, unidentified regulator whose activity is suppressed by BatR and SrkA.

SrkA activates multiple cell death mechanisms in P. aeruginosa

To visualise the SrkA-mediated killing effect, we repeated the phase-contrast time-lapse microscopy using a microfluidic device. Lysis events increased from 2-3 per 1,000 cells in the control strains (WT and ∆batR) (S3 c Fig) to 30–40 per 1,000 cells upon srkA overexpression (Fig 5a). Overexpressing cells appeared highly compromised, with frequent round cells and evidence of explosive lysis (Fig 5b). Additionally, srkA overexpression induced distinct morphological changes, including the emergence of highly elongated filamentous cells (Fig 5b).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. a. srkA overexpression increases explosive cell lysis in P. aeruginosa.

Cells were classified as rod-shaped or round over an 11-hour period. The frequency of round cells was calculated as the number of round cells divided by the total number of cells counted at the final time point. No significant differences were observed between strain backgrounds under the conditions tested. b. srkA overexpression alters cell morphology in P. aeruginosa. Phase contrast images from an 11h microfluidics assay of WT pME-Empty, pME-srkA ∆batR, pME-Empty and pME-srkA strains. Cells were incubated for 1h in LB, followed by 9h in LB + 0.1mM IPTG, and a final 1 h in LB. Each experiment was performed independently at least three times. Two primary phenotypes were observed upon srkA overexpression: increased frequency of explosive cell lysis (red arrows) and pronounced cell elongation (blue arrows). Time point: 11 h; scale bar: 2 μm. c. Bacteriophage activity using cell-free supernatants from WT PAO1 and ∆batR strains. Top: 20 µL drops from the filtered supernatants of WT, ∆batR, and ∆batRc (∆batR::batR) PAO1 strains were spotted onto the indicator strain, WT PAO1. Zones of clearing on the indicator lawn indicate cell lysis. Bottom: a control LB agar plate was spotted with 20 µL of each supernatant to ensure they were free of bacterial growth. Plates were incubated at 37°C overnight. d. Extracellular DNA quantification. Double stranded extracellular DNA (eDNA) concentration in the filtered supernatants was quantified on a Qubit Fluorometer using the high sensitivity dsDNA assay kit. The bars denote means of five biological replicates and a T-test shows significant differences between WT and ∆batR (t = 3.735 df = 6, p = 0.0097).

https://doi.org/10.1371/journal.ppat.1013832.g005

To gain further insight into the SrkA mechanism of action, we re-isolated six escape mutants -three from WT PAO1-pME-srkA null and three from ∆batR PAO1-pME-srkA null (Fig 4a) and subjected them to whole-genome sequencing. Comparative analysis of escape mutant genomes revealed that all three mutants derived from the ∆batR background had deleted the srkA null gene from the overexpression plasmid. In the WT background, one of the three mutants harboured a plasmid-srkA null deletion, while a second contained a single-nucleotide polymorphism (SNP) that introduced a premature stop codon at position 147 of the srkA null open reading frame, resulting in a truncated and likely nonfunctional protein. The third WT-derived escape mutant sequence harboured multiple unique SNPs in the chromosome, but not in the plasmid, suggesting that bypassing srkA null toxicity (without inactivating the gene itself) involves multiple genetic changes. Interestingly, many of these SNPs were in genes encoding oxidoreductases, biofilm-regulators (gacA, bfmS) [48], and DNA repair proteins.

We detected a substantial number of SNPs within the PA0717-PA0727 gene cluster, which encodes components of a Pf1-like filamentous bacteriophage (Pf4) in all six strains. This mutational enrichment coincided with increased read depth across the region, consistent with active replication of the prophage (S5 Fig). This observation is notable, as phage Pf4 has previously been implicated in promoting cell death and eDNA release in mature PAO1 and PA14 biofilms [25,49,50]. Pf4 is unusual among bacteriophages because it lacks a canonical endolysin and is capable of infecting cells within the clonal population from which it originates [49]. Furthermore, the highly elongated filamentous cells observed during srkA overexpression (Fig 5b), resemble phenotypes associated with Pf4 activity during biofilm development [51]. Together, these findings suggest that Pf4 may represent the additional SrkA-triggered lethal pathway that operates independently of the classical R2/F2 pyocin lysis system. To test this hypothesis, we overexpressed srkA in PAO1 ∆PA0728, which lacks the Pf4 integrase [52]. Similar to the ∆lys mutant, the effect of srkA overexpression on cell viability was attenuated in ∆PA0728 strain compared to WT PAO1 (Fig 4c). These results indicate that srkA-induced cell death is mediated through at least two distinct pathways, with contributions from both endolysin-dependent lysis and a Pf4-associated mechanism.

We therefore asked whether BatR also contributes to Pf4 production or activity. To test this, WT and ΔbatR strains were grown under conditions used for the proteomic assays and filtered culture supernatants were spotted onto lawns of PAO1 WT cells. Interestingly, only WT supernatants produced plaques, whereas ΔbatR supernatants did not (Fig 5c). This phenotype was fully rescued by heterologous expression of batR (ΔbatRc), implicating BatR in phage-mediated cell death and eDNA release (Fig 5c). No differences in growth between these strains were observed in shaking cultures (S1c Fig.). Proteomic analysis of these supernatants identified the Pf4 coat protein A among the most enriched proteins in WT samples (Table 4, S3 Data). Consistent with this, quantification of eDNA revealed significantly higher levels in WT supernatants compared to those from the ΔbatR strain (Fig 5d).

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Proteins increased in WT supernatants (Label free proteomics).

https://doi.org/10.1371/journal.ppat.1013832.t004

Next, we used RT-qPCR to examine the influence of batR and srkA on the regulation of the R2/F2 pyocin operon and the Pf4 bacteriophage at a transcriptional level. RT-qPCR analysis of PA0614 (lys) and one of the core genes of Pf4, repG4 (PA0717), transcription showed very low expression levels in the WT and the single mutants but a strong upregulation in the double mutant ∆batR ∆srkA, consistent with phenotypes seen for this mutant (S4 e, f Fig., respectively).

BatR affects the ability of P. aeruginosa biofilms to withstand antibiotic challenge in clinically validated infection models

Given BatR’s role in biofilm-mediated antibiotic sensitivity under laboratory conditions, we next determined whether this effect extends to clinically relevant infection models. First, we used an in vitro synthetic chronic wounds (SCW) model, which mimics the environment of diabetic foot infections [34], to assess biofilm formation following exposure to increasing concentrations of CIP (0–64 µg/mL).

In the absence of antibiotic, no differences in biofilm formation were observed between the WT and ∆batR strains. However, upon CIP treatment, PAO1 ∆batR biofilms showed significantly reduced viavility compared to WT biofilms (Fig 6a), with a nearly 40-fold reduction in CFU recovery in the ∆batR mutant at higher CIP concentrations.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. BatR contributes to antibiotic resistance and pyocyanin production in clinically validated models.

a. Synthetic Chronic Wounds (SCW) Model. Viability of P. aeruginosa PAO1 and the ΔbatR strain living in established biofilms (24 h), treated with different concentrations of CIP in the cell suspension on top of the matrices. Dots represent the mean from three biological replicates and error bars indicate the standard deviation. Two-way ANOVA showed significant differences for WT and ΔbatR under 64 µg/mL CIP treatment (p < 0.001 **). b. Ex Vivo Pig Lung (EVPL) Model. Growth of P. aeruginosa PAO1 strain and ΔbatR, on 15 pieces of EVPL bronchiole (three replicate pieces of tissue from five independent lungs infected per strain) plus Synthetic Cystic Fibrosis Medium (SCFM), with and without 16 µg/mL CIP treatment [16]. Colony forming units (CFU) were retrieved from biofilms after 2 d growth at 37°C. Bars denote mean for each genotype across all five lungs, and asterisks denote a significant difference under that condition. Two-way ANOVA analysis showed significant differences for WT vs ∆batR; and WT + CIP vs ∆batR + CIP (p < 0.0001 ***). c. Biofilm pigmentation. The characteristic blue-green colouration of P. aeruginosa results from a mixture of the exoproducts pyoverdine and pyocyanin. Notably, pigmentation intensity was higher in the WT strain when grown on bronchiolar tissue sections in SCFM and as biofilm aggregates in a collagen gel matrix in the SCW model. d. Quantification of pyocyanin by P. aeruginosa WT and ∆batR in the EVPL model. Differences in pyocyanin production (A₆₉₅) between WT and ΔbatR under CIP treatment were significant (p < 0.0001***).

https://doi.org/10.1371/journal.ppat.1013832.g006

To expand these findings to a more complex system, we next used an ex vivo pig lung (EVPL) model that recapitulates P. aeruginosa biofilm infections in CF bronchioles [33]. This model comprises bronchiolar lung tissue and Synthetic Cystic Fibrosis Medium (SCFM) to replicate the luminal mucus environment [53]. Biofilms were challenged with CIP at a SIC (16 µg/mL, 1/8^th of the MIC in the EVPL model, S6 Fig) and CFUs were enumerated at 2 days postinfection (dpi).

CFU counts in WT biofilms were unaffected by CIP treatment. In contrast, ∆batR biofilm formation was reduced even in the absence of CIP, and further decreased upon CIP treatment, resulting in 2.5-fold reduction in bacterial load under antibiotic stress. Our results confirm that BatR contributes to CIP resistance and supports biofilm establishment within the EVPL model (Fig 6b).

Alongside these effects on biofilm formation, increased blue pigmentation, indicative of pyocyanin production, was evident in WT biofilms compared to ΔbatR following CIP treatment (Fig 6c, d). A similar effect was also observed in WT biofilms relative to ∆batR in the synthetic chronic wound model under antibiotic stress (Fig 6c).

BatR contributes to biofilm architecture

Next, given BatR’s role in promoting eDNA release under laboratory conditions, we examined the biofilm architecture in infected lung tissue using the EVPL model, with and without SIC of the antibiotic CIP. Infected lung pieces were collected in replica sets and fixed at 2- and 7-dpi, then paraffin-embedded. Sections were stained with H & E (Haematoxylin & Eosin) to visualise total biofilm mass and tissue structure. As expected, distinct biofilm structural differences were observed at both time points (S7 and 7a, b, Figs respectively) between ∆batR and WT strains.

At 2 dpi, the WT strain exhibited significantly higher staining intensity, indicative of denser and more extensive biofilm formation, both in the presence and absence of CIP (S7 Fig). This observation aligns with the higher viability counts observed for the WT strain (Fig 6b), suggesting a larger population of viable cells contributing to the biofilm mass.

By 7 dpi, the EVPL biofilms show a characteristic “sponge-like” appearance consisting of extracellular matrix punctuated by gaps, resembling CF biofilms observed in vivo [54,55,56]. Interestingly, qualitative differences in sponge-like architecture were apparent between the ∆batR and WT strains, as shown in Fig 7a, b. The WT biofilms showed a thicker matrix and more pronounced structural features, both in untreated samples and those exposed to SIC of CIP. These features were not observed in the ∆batR biofilms, suggesting that batR influence on biofilm architecture may explain the reduced structural resilience we observe in ∆batR, particularly under antibiotic stress (Fig 6b).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. BatR contributes to biofilm architecture.

Haematoxylin and eosin (H & E) stained sections of EVPL bronchiolar tissue at 7 dpi. EVPL was infected with P. aeruginosa PAO1 WT and ∆batR, with uninfected tissue as a negative control. The x20 magnification images from the sections are shown here for a. non-treated tissues (SCFM) and b. treated with CIP (SCFM + CIP). The cartilage and tissue surface (red bar) stain pink and the bacterial biofilm stain purple, including the bacterial cells and biofilm matrix. The typical sponge-like structure is shown by the dashed boxes on the WT images, and the purple arrow shows the thick layer of matrix covering the biofilm in WT. Representative images of phenotypes at 7 dpi are shown here, but the same results were observed for all biological replicates analysed.

https://doi.org/10.1371/journal.ppat.1013832.g007

Bioinformatic analysis

A phylogenetic tree of RMF proteins was constructed using 765 publicly available protein sequences from the NCBI database. The dataset was curated based on a criterion of 50% sequence identity and 40% query cover to ensure the representation of diverse homologs while maintaining a reasonable level of similarity. The multiple sequence alignment was performed using Clustal Omega (v1.2.4), generating an alignment matrix with 137 columns and 135 distinct patterns. Phylogenetic signal analysis revealed 82 parsimony-informative sites, 33 singleton sites, and 22 constant sites. Manual curation was conducted to remove repetitive sequences and false hits. The phylogeny estimation was done using IQ-TREE, (multicore v1.6.12), employing the maximum likelihood (ML) criterion. The tree visualization was done using iTOL (v6.8.1), representation chosen was an unrooted tree with branch lengths proportional to the inferred evolutionary distances between sequences. The clade colours were assigned at order level of the taxonomy.

Bacterial strains and growth media

Bacterial strains and plasmids used in this study are listed in Table 5. Unless otherwise stated, P. aeruginosa PAO1 and E. coli DH5α strains were routinely cultured in LB (lysogeny broth) [70] at 37°C solidified with 1.5% w/v agar where appropriate. For growth curves, cells were grown at a starting OD₆₀₀ of 0.01 in a 96-well plate. Measurements were taken every 60 min for up to 48 h on a FLUOstar nano plate reader (BMG) with the plate being incubated at 37 °C under planktonic conditions.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Strains and plasmids used in this study.

https://doi.org/10.1371/journal.ppat.1013832.t005

Molecular biology techniques and genetic manipulation of PAO1

These procedures were performed as previously described [76]. All pTS1 plasmid inserts were synthesised and cloned into pTS1 by Twist Bioscience.

The ORF of srkA with S33 and D216 substituted by an Ala in the srkA null mutant was synthesised by Twist Bioscience. The ORFs of batR and srkA / srkA null were amplified by PCR with primers batR_EcoRI_F / batR_XhoI_R and srkA_EcoRI_F / srkA_XhoI_R, respectively (Table 6), and ligated between the EcoRI and XhoI sites of pME6032.

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Primers used in this study.

https://doi.org/10.1371/journal.ppat.1013832.t006

To introduce batR gene into the PAO1 att::Tn7 site in the strain ∆batRc, the batR locus was amplified from the PAO1 genome using primers batR_EcoRI_F and batR_XhoI_R. The resulting PCR product was cloned into the multiple cloning site of pME6032 [74]. Primers ptacFP and ptacRP [76] were then used to amplify batR in addition to the tac promoter and terminator of pME6032. This product was then cloned into the pUC18-mini-Tn7T-Gm vector [75] and used to transform PAO1 via electroporation.

For the flag-tagged batR construction in pME6032, primers batR_EcoRI_F and 3xflag_batR_XhoI (Table 6) were used. Bacterial-2-hybrid fusion proteins were created by fusing the N-terminus of BatR, AcpP, SrkA, PslC, PilG, GlpK and RibA to pKT25 and combined with pUT18C-Empty and pUT18C-batR (Table 6). Genes were amplified using primers that introduce either XbaI-KpnI (stop codon) or XbaI-EcoRI (stop codon) restriction sites.

MIC determination

Minimum inhibitory concentrations of antibiotics were determined by the broth microdilution method [37] following the EUCAST guidelines, using Mueller-Hinton broth. The Sub Inhibitory Concentration (SIC) was defined as being 1/8 of the lowest antibiotic concentration that inhibited visible growth after overnight incubation at 37°C.

Inhibition disc assay

Bacterial cultures were grown in LB medium at 37°C to mid-log phase, A_600nm=0.5-0.7, and 100 µL were spread on each plate. Whatmann filter paper discs containing the antibiotics were gently placed on the agar and plates were incubated inverted overnight. The normalized width of the antimicrobial “halo” (NW_halo) of each disk was determined after [39].

Glass beads biofilms

These assays were performed as described previously [38,79] with the following modifications. Ten independent biological replicates were included: five PIP-exposed biofilm lineages (challenged with SIC of PIP for 90 min) and five unexposed control lineages. Cells recovered from the beads were serially diluted and spotted onto LB plates for CFU counting. Results are shown as previously described by [38], where means in PIP-exposed biofilm were then normalised by the average number of cells across all unexposed conditions for plotting. Thus, the values represent the estimated proportion of cells that would survive each exposure for each strain. For complementation assays, culture media were supplemented with 0.5mM IPTG and Tet. The experiment was repeated three times.

Membrane permeability assays

Differences in membrane permeability to antibiotics were assessed using the resazurin accumulation assay [80]. Strains of interest were grown to exponential phase under shaking conditions, using a 1:100 inoculum from overnight cultures. Cells were washed, resuspended in PBS, and normalized for cell density before being mixed with resazurin in round-bottom microtiter plates to a final volume of 100 µL (10 µg/mL resazurin). Fluorescence was measured with an Omega FLUOstar plate reader at an excitation wavelength of 544 nm and an emission wavelength of 590 nm. Five replicates were included per strain, and resazurin-only wells were used as controls. The assay was repeated at least twice, yielding reproducible results each time.

Quantitative proteomics (TMT) for expression analysis

The experiment was essentially performed as previously described [81] with some modifications, as detailed in Supplementary Information (S4 Data).

PA0614 promoter activity assay

Promoter activity was assessed using the egfp reporter plasmid pMEXGFP [43]. Cells harbouring the recombinant plasmid were cultured on LB agar medium supplemented with Gn at 37 °C for 48 h. Following incubation, cells were scraped from the agar surface and resuspended in 1mL PBS. Fluorescence was measured using a fluorometer with an excitation wavelength of 488 nm and an emission wavelength of 509 nm. The eGFP fluorescence intensity of each sample was normalized against cells containing the promoterless pMEXGFP plasmid to account for background fluorescence.

Time-lapse experiment using the CellASIC microfluidic system and image analysis

Time-lapse imaging of P. aeruginosa strains was performed as previously described [44], with the following modifications. Briefly, P. aeruginosa strains grown overnight in LB medium at 37°C and 250 rpm were sub-cultured to reach an OD₆₀₀ of 0.2. Bacterial cells were loaded into B04A microfluidic plates (ONIX, CellASIC), grown by perfusing LB for 6 h in the case of PAO1 WT and ∆batR strains; or perfusing LB for 1 h, then switched to 0.1 mM IPTG LB for 9 h, and LB for 1h in the case of PAO1 strains overexpressing srkA. The media flow rate and temperature were maintained at 2 psi and 37°C. Time-lapse imaging was started from the beginning of the experiment and images were acquired every 5 min.

P. aeruginosa strains were visualized using a Zeiss Axio Observer Z.1 inverted epifluorescence microscope fitted with a sCMOS camera (Hamamatsu Orca FLASH 4), a Zeiss Colibri 7 LED light source, a Hamamatsu Orca Flash 4.0v3 sCMOS camera, and a temperature-controlled incubation chamber. Images were acquired using a Zeiss Plan Apochromat 100x/NA 1.4 Ph3 objective. Still images and time-lapse images series were collected using Zen Blue (Zeiss) and analysed using Fiji [82].

Co-immunoprecipitation and mass spectrometry analysis

This protocol is detailed in Supplementary information (S5 Data).

Bacterial 2 hybrid assays

These assays were performed as described elsewhere [81] with some modifications. The ORFs of batR, acpP, pilG, pslC, ribA, srkA and glpK were cloned into pKT25, and pUT18C using either conventional restriction enzyme cloning or Gibson assembly, as indicated in Table 6.

Alphafold3 model predictions

Protein structures were predicted from amino acid sequences using the Alphafold3 server (alphafoldserver.com). For PAO1 BatR (AF_AFQ9HZF9F1) and E. coli Rmf (AF-P0AFW2-F1-v4), initially protein structures were retrieved from AlphaFoldDB with interactions modelled independently using the AlphaFold3 server. For interactions, both the bait (BatR WT) and target (SrkA) amino acid sequences were input into the sever and predictions were modelled using default settings. Models were ranked based on their respective PTM/iPTM score and the highest ranked score was taken forward for analysis. Structural analysis and alignments were conducted in Pymol.

RNA purification and RT-qPCR

Four biological replicates of each strain were set up as described above. A single colony from each strain was picked and grown overnight in 10 mL LB medium at 37 °C, 250 rpm. The OD600 of these cultures was measured and 1 mL LB cultures were set up at OD600 = 0.2 and plated on LB agar and incubated at 37 °C for 48h. Cells were scraped from the agar surface and harvested for RNA extraction. Pellets were resuspended in 150 μL 10 mM Tris-HCl (pH 8) and mixed with 700 μL of ice cold RLT + BME (RLT buffer (Qiagen) supplemented with 1% β-mercaptoethanol) and cells were lysed using a Fastprep (MP Bio) using Lysis matrix B beads (MP Bio). Lysis matrix was removed by centrifugation (13,000 × g, 3 min) and the supernatant was added to 450 μL of ethanol. The supernatant was applied to an RNeasy column and RNA extraction was performed as per the manufacturer’s instruction including the on-column DNA digest. After extraction, a Turbo DNase (Promega) digest was performed as per the manufacturer’s instruction and total RNA yield was quantified using a Qubit RNA broad spectrum assay kit as per the manufacturer’s instructions.

Absolute quantification of target genes alongside the normalising housekeeping gene (rpoD) was performed using the provided CFX Maestro software. The relative quantity of the sample of interest was calculated using a calibration curve of serial dilutions of purified genomic DNA (CFX Maestro Software User Guide, Appendix A). The primers used are listed in Table 6. Relative qPCR units were defined as the ratio between the absolute levels of each gene and the housekeeping gene rpoD.

Colony biofilm morphology assay

Agar plates for colony morphology experiments were prepared as previously described [83]. Briefly, a mixture of 1% agar and 1% tryptone was autoclaved and cooled to 60°C before 20 μg/ml Coomassie blue, 40 μg/ml Congo red, Tetracycline, and 0.05mM IPTG were added. For colony spotting, five microliters of cultures at an OD₆₀₀ of 0.5 were spotted on plates and incubated for up to 7 days at 23–25°C.

Cell viability and pyocyanin production phenotypes

Five biological replicates of each strain were grown in LB medium at 37°C to mid-log phase (A_600nm=0.2-0.3). Then, 25 µL of each culture were spread onto wells of 24-well plates. When required, tetracycline and IPTG were added at the appropriate concentrations. Plates were incubated for 24h at 37°C and scanned. For cell viability measurements, cells were scraped from the agar and resuspended in 1mL of 1X PBS. These resuspensions were serially diluted and spotted onto LB agar plates for CFU enumeration.

Crystal Violet (CV) assays

For the CV assay, five biological replicates of each strain were grown overnight in LB medium supplemented with Tet and then diluted into 200 μL of fresh LB to give an OD_600nm of 0.50 in microtiter plates. The expression of recombinant PAO1 srkA, was induced with 0.05mM IPTG. The microtiter plates were incubated at 37°C for 24 h, after which the wells were emptied and rinsed three times with water before staining. For staining, 200 μL of 0.1% CV was added to each well and incubated for 15 min at room temperature. The crystal violet dye was then removed, and the wells were rinsed with water. The dye bound to the cells was then dissolved in 70% ethanol, and the A_590nm was measured using a SPECTROstar nano plate reader (BMG Labtech).

Genomic analysis

Six escape mutants -three from WT PAO1-pME-srkA null and three from ∆batR PAO1-pME-srkA null- were chosen for whole-genome sequencing. These colonies were re isolated from the colony biofilm morphology assay at day 7. As controls, the genomes of WT and ∆batR PAO1 strains were also sequenced. Whole-genome sequencing was performed by Plasmidsaurus using the Oxford Nanopore long-read technology. Variants were called against the ancestral reference genome using the Breseq computational pipeline using the polymorphic settings [84]. All variants were validated visually using the alignment viewer IGB [85].

Bacteriophage experiments

Strains of interest were grown overnight in 10 mL of LB broth at 37°C. Cultures were adjusted to an OD₆₀₀ of 1.0, and 100 µL of the normalized cultures were spread onto LB agar plates. Plates were incubated for 24 hours at 37°C and the resulting bacterial lawns were scraped from the plates and resuspended in 5 mL of 1 X PBS. These homogenates were passed through a 0.2 μm syringe filter to remove bacterial cells.

Indicator strains were grown overnight in 10 mL of LB broth and diluted 1,000-fold in PBS. Then, 50 µL of the diluted indicator cultures was spread evenly onto LB agar plates. Next, 10 µL of the filtered supernatants was spotted onto the indicator strain plates. Plates were incubated overnight at 37°C. As a control, filtered supernatants were spotted onto LB agar plates to confirm that they were cell-free.

Label free proteomics

This protocol is detailed in Supplementary information (S6 Data)

Synthetic chronic wound infection model

These assays were performed as previously described [34]. Briefly, batch cultures were grown at 37°C in Synthetic Wound Fluid (SWF) overnight to early mid log phase (approx. 6 h). The synthetic wounds were prepared as follows: For 10 mL collagen solution (2 mg/mL), 1 mL 0.1% acetic acid was mixed with 2 mL collagen stock solution (10 mg/mL) and kept on ice. Then, 6.0 mL of cold Synthetic Wound Fluid (SWF) (50% foetal bovine serum (Gibco 10270) and 50% Peptone water (Fluka 70179)) was added followed by 1 mL 0.1 M NaOH. After mixing, 200 μL of the collagen solution was added to each well (24-well plates). To achieve a complete polymerization of the collagen, the plates were placed in an incubator at 37°C for 1 h.

The starter bacterial cultures were diluted at an OD₆₀₀ of approx. 0.05 – 0.1 in SWF. For small wounds, 50 μL of the diluted starter culture was added to each synthetic wound. Plates were incubated at 37°C for 24 h. After this time, 100 μL of antibiotic CIP at different concentrations was added to the wounds and plates were returned to 37°C for a further 24h. For bacterial recovery, 0.5 mg/mL collagenase (made in PBS) was added to each wound and incubated at 37°C for 1 h. Serial dilutions of these homogenates were plated on LB agar for CFU counting.

EVPL infection model

EVPL was prepared as previously described [53,86]. Briefly, porcine lungs were obtained from two local butchers (Quigley and Sons, Cubbington and Taylor’s Butcher, Earlsdon) and dissected on the day of delivery under sterile conditions. The pleura of the ventral surface was heat sterilised using a hot pallet knife. A sterile razor blade was then used to make an incision in the lung, exposing the bronchiole. A section of the bronchiole was extracted, and the exterior alveolar tissue removed using dissection scissors. Bronchiolar sections were washed once in a 1:1 mix of Dulbecco’s modified Eagle medium (DMEM) and RPMI 1640 supplemented with 50 μg/mL ampicillin (Sigma-Aldrich) then cut into approximately 5 mm wide longitudinal strips. The bronchiolar strips were placed in a second 1:1 DMEM, RPMI 1640 supplemented with 50 μg/mL ampicillin wash and cut into squares approximately 5 x 5 mm in size. The tissue squares were washed for a third time in 1:1 DMEM, RPMI 1640 containing 50 μg/mL ampicillin. Bronchiolar pieces were then further washed in SCFM, UV sterilised for 5 min and transferred to individual wells of a 24-well plate containing 400 μL SCFM supplemented with 20 μg/mL ampicillin and solidified with 0.8% (w/v) agarose per well.

A sterile 29G hypodermic needle (Becton Dickinson Medical) was touched to the surface of a P. aeruginosa colony grown on LB agar overnight at 37°C and used to pierce an individual piece of bronchiolar tissue. Uninfected control tissue sections were mock inoculated with a fresh, sterile needle. Following infection of bronchiolar tissue pieces, 500 μL of SCFM ± 16 μg/mL CIP was added to each well. Tissue pieces were incubated at 37°C with a Breathe-Easier membrane (Diversified Biotech) for 2 and 7 d.

EVPL biofilm recovery, assessment of bacterial load and pyocyanin production

EVPL biofilm recovery and assessment of bacterial load and virulence factor production were determined as described [53]. Briefly, bronchiolar tissue pieces were removed from the 24-well plate following incubation, and each briefly washed in 500 μL PBS in a fresh 24-well plate to remove planktonic cells. Tissue pieces were then transferred into sterile homogenisation tubes (Fisherbrand) containing eighteen 2.38 mm metal beads (Fisherbrand) and 1 mL PBS. Tissue was bead beaten in a FastPrep-24 5G (MP Biomedicals) for 40 s at 4 m/s to recover the bacteria and virulence factors from the tissue-associated biofilm. To determine the bacterial load, the homogenate was serially diluted in PBS and plated on LB agar. Plates were incubated overnight at 37°C and colony counts used to calculate colony forming units (CFU) per tissue piece.

To quantify pyocyanin produced by the P. aeruginosa biofilms, the homogenate was diluted 1:4 in PBS to obtain a sufficient volume for further experiments. The diluted homogenate was passed through a 0.2 μm filter to remove bacterial cells and tissue debris. Total pyocyanin was quantified by measuring absorbance of homogenates at 695 nm [33].

Haematoxylin & eosin staining

P. aeruginosa infected EVPL tissue pieces and uninfected control tissue were fixed in 10% (v/v) neutral buffered formalin (VWR Chemicals), as previously described [53]. The infected/uninfected EVPL tissue pieces were fixed and sent to the University of Manchester’s Histology Core Facility for paraffin wax embedding, sectioning, and mounting. Mounted tissue sections were de-paraffinized in xylene for 20 min. To re-hydrate the tissue, slides were transferred to 95% (v/v) ethanol followed by 70% (v/v) ethanol. Any residual ethanol was removed by washing slides in distilled water. Samples were stained in Mayer’s hemalum solution (Merck Millipore) then washed in running tap water (1 L tap water with approximately 10 g sodium carbonate). Samples were counterstained in eosin Y solution (Merck Millipore) for then dehydrated by dipping the slides in 95% (v/v) ethanol. Samples were transferred to fresh 95% (v/v) ethanol then placed in 100% (v/v) ethanol. The samples were then placed in xylene. The samples were mounted using DPX mounting fluid and images were taken using a Zeiss Axio Imager Z2 light microscope with the Zeiss AxioCam 506 and Zeiss Zen Blue v2.3 pro software.

Data presentation and statistical analyses

All graphs and statistical analyses, including one-way and two-way ANOVA followed by post-hoc Tukey's Multiple Comparison Test, when appropriate, were performed using GraphPad Prism version 5.04 for Windows.

Data availability

The TMT proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [87] partner repository with the dataset identifier PXD050997 and 10.6019/PXD050997.

The label-free proteomic data have been deposited to the ProteomeXchange Consortium via the PRIDE [85] partner repository with the dataset identifier PXD062857 and 10.6019/PXD062857.

Co-IP data are available via ProteomeXchange with identifier PXD050995.