Reactive oxygen species drive evolution of pro-biofilm variants in pathogens by modulating cyclic-di-GMP levels

The host immune system offers a hostile environment with antimicrobials and reactive oxygen species (ROS) that are detrimental to bacterial pathogens, forcing them to adapt and evolve for survival. However, the contribution of oxidative stress to pathogen evolution remains elusive. Using an experimental evolution strategy, we show that exposure of the opportunistic pathogen Pseudomonas aeruginosa to sub-lethal hydrogen peroxide (H2O2) levels over 120 generations led to the emergence of pro-biofilm rough small colony variants (RSCVs), which could be abrogated by l-glutathione antioxidants. Comparative genomic analysis of the RSCVs revealed that mutations in the wspF gene, which encodes for a repressor of WspR diguanylate cyclase (DGC), were responsible for increased intracellular cyclic-di-GMP content and production of Psl exopolysaccharide. Psl provides the first line of defence against ROS and macrophages, ensuring the survival fitness of RSCVs over wild-type P. aeruginosa. Our study demonstrated that ROS is an essential driving force for the selection of pro-biofilm forming pathogenic variants. Understanding the fundamental mechanism of these genotypic and phenotypic adaptations will improve treatment strategies for combating chronic infections.


Background
Bacterial pathogens can colonize human hosts for years and cause persistent/ chronic infections [1,2]. These infections predominantly result from biofilm formation, whereby secreted extracellular-polymeric substances (EPS) such as adhesive proteins, biosurfactants, extracellular DNA and exopolysaccharides act as physical barriers to protect bacterial cells from the host immune clearance and antimicrobial treatments [3][4][5].
Pseudomonas aeruginosa is the leading cause of chronic lung infection and morbidity of patients with cystic fibrosis (CF) [19,20]. It is able to survive and form biofilms despite the presence of a functional host immune system, antibiotic treatments and competing pathogens, such as Staphylococcus aureus in the CF lung environment [21][22][23][24]. Investigations of adaptive evolution of P. aeruginosa in CF lung infections have provided valuable information for our current understanding of chronic infections. P. aeruginosa can colonize the CF lungs for decades and usually gives rise to pro-biofilm subpopulations after adaptive evolution [25,26]. The occurrence of rough small colony variants (RSCVs) and mucoid strains is often reported from chronic CF infections, implying these variants have better fitness than their ancestors [27,28].
Although the characteristics of RSCVs have been studied [29,30], the stimuli and mechanisms leading to the evolution of such adapted sub-populations are unclear. Given the CF environment consists of oxidative stress, high antibiotic concentrations, high pro-inflammatory cytokine levels and poor nutrient conditions [31][32][33][34][35], it is highly likely that each of the CF-derived environmental factors can drive the adaptive evolution of pathogens differently.
Experimental evolution assays have been used in previous studies to investigate bacterial adaptation to various conditions such as antibiotic treatments and carbon sources [18,36]. Here, we employed the adaptive experimental evolution assay to evolve P. aeruginosa against an important host-derived antimicrobial, reactive oxygen species (ROS), resulting in the occurrence of RSCVs with a strong capability for biofilm formation and ROS stress resistance.
The RSCVs isolated from H 2 O 2 treated cultures showed increased intracellular c-di-GMP content. Next-generation sequencing (NGS) analysis revealed that wspF mutation was associated with these isolated RSCVs. The wspF is part of the chemosensory-like system Wsp (wrinkly spreader phenotype) [8], whose product acts as a repressor against WspR (DGC), thus its mutagenesis leads to the de-repression of WspR. The increased production of exopolysaccharides (especially Psl) as the result of wspF mutation conferred resistance of RSCVs to H 2 O 2 treatment. Hence, this work strongly suggests that exposure to ROS imposes a strong selective pressure on P. aeruginosa during chronic colonization and accounts for the occurrence of RSCVs in clinical isolates obtained from CF patients.

Bacterial strains, plasmids, media and growth conditions
Escherichia coli DH5a strain was used for standard DNA manipulations. Luria-Bertani (LB) medium was used to cultivate E. coli strains. Batch cultivation of P. aeruginosa strains was carried out at 378C in ABTG (ABT minimal medium supplemented with 5 g l 21 glucose) or ABTGC (ABT minimal medium supplemented with 2 g l 21 glucose and 2 g l 21 casamino acids The populations were then cryopreserved with 50% glycerol (cryoprotectant) for revival at a later time. To observe the emergence of unique phenotypes arising from treatment with 2 mM H 2 O 2 , the populations were grown on LB agar plates at 378C overnight. Ten RSCV isolates were identified from the different replicates and frozen with 50% glycerol.

H 2 O 2 resistance assay
The P. aeruginosa cultures were grown overnight and prepared by adjusting the optical density to OD 600 ¼ 0.3 in ABTGC with and without 4 mM H 2 O 2 . The cultures were grown at 378C, 200 rpm for 4 h. The cell culture was serially diluted and plated on LB agar plates. The culture plates were incubated in 378C for 16 h. The numbers of colonies were counted for tabulation of log 10 CFU ml 21 . The log 10 CFU ml 21 is calculated by log 10 (average number of colonies Â dilution factor Â volume used to spread on LB agar plate). Three independent experiments were performed in triplicate, one-way ANOVA and Student's t-tests were used to determine statistical significance and the results were shown as the mean + s.d.

Competitive mixed-treatment assay of rough small colony variant isolates and PAO1
The evolved RSCVs and PAO1 cultures were prepared by adjusting the optical density to OD 600 ¼ 0.5 in ABTGC. The genomic DNA was then sequenced on an Illumina (San Diego, CA, USA) MiSeq V3 platform, generating 300 bp long paired-end reads using the method described in Chua et al. [38]. The average insert sizes were 490-544 bp, and average genomic coverage depths were 40-186 fold. Nucleotide differences were generated from the CLC GENOMICS WORKBENCH 8.0 (CLC bio, Aarhus, Denmark). Briefly, adapters and low quality reads were trimmed. Paired-end reads in FASTQ format for RSCV and control genomes were mapped against the P. aeruginosa PAO1 genome (NC_002516). Both the mappings of RSCV and control strains were compared with their ancestor PAO1, and variants were detected using the quality based variant detection method with the required frequency of 35%.

Measurement of mutation rates
PAO1 was grown in ABTGC with or without 2 mM H 2 O 2 until stationary phase, and plated on LB agar plates with and without 0.78 mg ml 21 ciprofloxacin, using the method described in Mandsberg et al. [39]. The mutated colonies cultured on LB agar with ciprofloxacin were enumerated to account for point mutations on the gyr gene that can easily develop in P. aeruginosa exposed to ciprofloxacin [40]. The mutation rate was then calculated by dividing the number of mutations by the final CFU of cells grown on LB agar plates without ciprofloxacin. Three independent experiments were performed in triplicate, the Student's t-test was used to determine statistical significance and the results were shown as the mean + s.d.

The p cdrA -gfp reporter assay
The cultures of PAO1 with p cdrA -gfp reporter fusion were grown in ABTGC with or without H 2 O 2 . In total, 200 ml of cell culture was then transferred into each well of a 96-well plate (triplicates). The OD 600 and GFP fluorescence (excitation 485 nm/emission 535 nm) were measured using the Tecan Infinite 200 microplate reader (Tecan, Austria). The relative fluorescence intensity count was calculated by dividing GFP values by OD 600 values. Three independent experiments were performed in triplicate, one-way ANOVA and Student's t-tests were used to determine statistical significance and the results were shown as the mean + s.d.

Quantification of pyoverdine
The cultures were grown in ABTGC with or without H 2 O 2 . In total, 200 ml of cell culture was then transferred into each well of a 96-well plate (triplicates). As previously described [13], the OD 600 and pyoverdine fluorescence (excitation 400 nm/ emission 460 nm) were measured using the Tecan microplate reader, and the relative fluorescence intensity count calculated by dividing pyoverdine fluorescence values by OD 600 values. Three independent experiments were performed in triplicate, one-way ANOVA and Student's t-tests were used to determine statistical significance and the results were shown as the mean + s.d.

C-di-GMP quantification by liquid chromatography -mass spectrometry
Fifteen millilitres of PAO1 cell culture treated with H 2 O 2 , RSCVs, Burkholderia cenocepacia strains and Klebsiella pneumoniae strains were harvested and washed twice with 1 mM ammonium acetate. An aliquot of cells was used for protein quantification. The remaining cells were lysed in 1 ml acetonitrile/methanol/ddH 2 O (v/v ratio 40 : 40 : 20) using a probe tip ultrasonicator (amplitude 30%; 5 s on, 5 s off) for 1 min on an ice slurry. The cell debris was removed by centrifuging at 13 000g, 4 8C for 3 min, rinsing twice. The supernatant containing the nucleotides was lyophilized with the vacuum concentrator. The lyophilized nucleotides were resuspended in 100 ml 1 mM ammonium acetate. The c-di-GMP standard was also used as a reference to identify the c-di-GMP peak and the concentration of c-di-GMP in the samples. For the detection and quantification of c-di-GMP, a Thermo Accela 1250 series LC system fitted with EQuanMax autosampler and a Thermo Velos Pro Orbitrap mass spectrometer (Thermo Fisher Scientific) were used. Chromatographic separation was achieved using a Nucleodur C18 Pyramid (2 mm Â 50 mm, 3 mm) column (Macherey-Nagel GmbH, Dü ren, Germany) at 408C, with a solvent flow rate of 0.3 ml min 21 and an injection volume of 10 ml. Buffer A was 10 mM ammonium acetate buffer, containing 0.1% acetic acid, and buffer B was acetonitrile, containing 0.1% acetic acid. Solvent gradient conditions were as follows: 0% B from 0 to 3 min; 10% B at 3 min; 90% from 4th to 5th min; 0% B at 5.5th min and equilibrated for 4.5 min. Total run time was 10 min.
Detection was carried out in positive ion electrospray ionization (ESIþ) mode. The heater and capillary temperatures were 3008C. Sheath, auxiliary and sweeper gas flows were 40, 15 and 1 arb. units, respectively. Source voltage was 3.5 kV. For quantitation, scan type in selected ion monitoring mode was used at high-resolution (60 000), with an AGC target of 1 Â 10 6 . Quantification was achieved via an MS/ MS experiment using collision induced dissociation (CID) rsob.royalsocietypublishing.org Open Biol. 6: 160162 with normalized collision energy 20% (of maximum), with isolation width of 1 Da and activation time of 30 ms.
To quantify protein concentration, a small aliquot of cell culture was lysed in 100 ml 5 M sodium hydroxide at 958C for 5 min. The protein concentration was then measured using a Qubitw 2.0 fluorometer (Invitrogen, Thermo Fisher Scientific, CA). The final concentration of c-di-GMP was then normalized with protein quantity. Three independent experiments were performed in triplicate, one-way ANOVA and Student's t-tests were used as to determine statistical significance and results were shown as the mean + s.d.

Psl staining by fluorescent concanavalin-A
Planktonic MiniTn7-gfp-tagged cells were grown in ABTGC þ 0, 0.5, 1 and 2 mM H 2 O 2 at 378C, 200 rpm for 4 h. Additionally, PAO1, mutants and evolved RSCVs with tagged MiniTn7-gfp were grown until stationary phase in ABTGC overnight at 378C, 200 rpm. Ten microlitres of cell culture were then transferred to the glass slide and stained by 5 mM concanavalin-A Alexa Fluor w 647 conjugate (Thermo Fisher Scientific, Cat. No. C21421) [41].
To monitor fluorescence of GFP and Psl stain, the cells were imaged using an LSM780 confocal laser scanning microscope (CLSM; Carl Zeiss, Germany) with 40Â objective or 63Â oil objective (for planktonic cells) and the images were processed using IMARIS software (Bitplane AG, Zurich, Switzerland). Three independent experiments were performed in triplicate and representative images were shown.

Arabinose-inducible expression of Psl by PAO1/ p BAD -psl
A starting culture of PAO1/p BAD -psl cells containing an L-arabinose-inducible promoter for psl operon expression was grown in ABTGC with increasing concentrations of L-arabinose at 378C, 200 rpm for 3 h until OD 600 ¼ 0.3 was established, as described in Irie et al. [42].

Macrophage phagocytosis assay
As previously described [12], 5 Â 10 5 macrophages were grown in each well of a 24-well culture plate (Nunc, Denmark). Macrophages were infected with bacterial suspension at a multiplicity of infection (MOI) of 100 : 1. The co-cultures were incubated at 378C and 5% CO 2 for 2 h. The macrophages were washed three times with phosphatebuffered saline (PBS) to remove extracellular bacteria, and then lysed with ddH 2 O containing 0.5% Triton X-100. The cell lysates were then serially diluted, and 100 ml of each dilution was plated on triplicate LB agar plates and incubated overnight at 378C. The number of colonies was enumerated, and CFU ml 21 was tabulated. Experiments were performed in triplicate, one-way ANOVA and Student's t-tests were used to determine statistical significance and the results were shown as the mean + s.d.

Macrophage cytotoxicity assay
RAW264.7 macrophages (5 Â 10 5 ml 21 in each well) were grown in 24-well culture plates as previously described [12]. To determine if oxidative stress provokes the adaptive evolution of P. aeruginosa, P. aeruginosa was cultivated in the presence of a sub-lethal concentration of H 2 O 2 (2 mM, 0.5 Â minimal inhibitory concentration (MIC)) for 120 generations. This simulated the presence of ROS produced by polymorphonuclear leucocytes (PMNs) in the CF lung, which generate oxidative stress as a selective pressure on P. aeruginosa cells [43]. Although a millimolar range of H 2 O 2 is required to kill most bacteria and bacteria internalized in the phagosome are usually exposed to micromolar concentrations, the rates of free radical generation within the phagosome are estimated to be 10 000 times faster, leading to higher concentrations of ROS than have been experimentally tested [44]. Hence, the sufficiently high levels of ROS present can be potentially damaging to bacterial cells. After 120 generations, colonies with RSCV phenotype were observed on the agar plates inoculated with 2 mM H 2 O 2 treatment (electronic supplementary material, figure S1), while no distinctive phenotypes were observed in the control populations without H 2 O 2 treatment. The percentage of RSCVs in H 2 O 2 exposed P. aeruginosa populations reached around 40% (figure 1a), implying that ROS could significantly contribute to the occurrence of RSCVs. We also showed that P. aeruginosa was able to evolve within 120 generations, when compared with a shorter 47.5 generation period in the evolution of RSCVs with antibiotics [18]. As P. aeruginosa can persist in CF lungs for decades, spanning 200 000 bacterial generations [33,45], there is more than sufficient time for the metabolically active [46] P. aeruginosa to adapt in CF lungs.
The resistance of randomly selected RSCV clones to H 2 O 2 was then measured. The RSCVs showed up to 16   We determined if the selection of RSCV by oxidative stress could be neutralized by the addition of L-glutathione, a commonly used antioxidant [47]. As expected, L-glutathione reduced the proportion of RSCVs formed in a dose-dependent manner, compared with that in the absence of antioxidants (figure 1d ).

Increased c-di-GMP content confers resistance of rough small colony variants to H 2 O 2
As c-di-GMP signalling is implicated in RSCV formation [27], we examined the intracellular c-di-GMP content in the selected RSCV isolates. Liquid chromatography-mass spectrometry (LC-MS) quantification of the whole cell extract showed that the RSCV isolates contained higher intracellular c-di-GMP than wild-type P. aeruginosa PAO1 (figure 2a). This was in accordance with their higher expression levels of the c-di-GMP reporter fusion p cdrA -gfp [48,49] (figure 2b) and higher pyoverdine production [50,51] (figure 2c), compared with that of ancestral PAO1. To confirm that the increased H 2 O 2 resistance of RSCVs was attributed to enhanced intracellular c-di-GMP content, we transformed a p lac -yhjH plasmid [52] into the RSCVs to deplete their intracellular c-di-GMP.

Comparative genomic analysis of rough small colony variants reveal parallel evolution traits
To gain insights into the underlying genetic changes contributing to c-di-GMP-linked RSCV phenotype development with prolonged oxidative stress via c-di-GMP signalling, we sequenced the genomes of 10 randomly selected RSCV isolates and 2 control PAO1 colonies using the Illumina MiSeq platform. Two random control PAO1 colonies were picked from cultures grown in a medium without H 2 O 2 for 120 generations. A series of point mutations were evident in the sequences of 10 randomly selected RSCVs and control PAO1 colonies, compared with the ancestor PAO1 sequence (electronic supplementary material, tables S2 and S3). The distribution of the identified mutations across the PAO1 genome is illustrated in figure 3a. The functional annotation of these RSCV-associated mutations is listed in the electronic supplementary material, tables S2 and S3. Several parallel evolution traits were observed from both RSCVs and control PAO1 colonies, including mutations in PA0727 and PA0720 (encoding phage-related genes), PA1458 (encoding a two component sensor), exsC (encoding an anti-anti-activator of the type III secretion system), PA2877 (encoding a probable transcriptional regulator), intergenic region 4 699 910 bp and intergenic region 5 242 141 bp. As P. aeruginosa commonly accumulate mutations even in planktonic cultures and the mutations are common between control and ROS-evolved strains, the above-mentioned mutations are not deemed to be generated by ROS pressure. We also found that the mutations were not attributed to highly mutagenic growth conditions, as there was no significant increase in mutation rates with ROS treatment (electronic supplementary material, figure S2) compared with control medium, with the mutation rates similar to previous reports [38]. Moreover, the cell populations did not develop rsob.royalsocietypublishing.org Open Biol. 6: 160162 mutations in the mutS gene, which is important in mismatch repair and its mutation can lead to the emergence of multiple phenotypic variants [53].
Interestingly, there was one parallel evolution trait, wspF gene mutation, which could only be detected in the genomes from the RSCVs and not in the control PAO1 colonies. All RSCV genomes had loss of function mutations in the wspF gene at different sites (figure 3b; electronic supplementary material, table S1). The DwspF mutation caused the de-repression of the WspR DGC, leading to increased c-di-GMP synthesis and high intracellular c-di-GMP content, resulting in bacterial cell clumping (electronic supplementary material, figure S1) [54]. This indicated that enhanced intracellular c-di-GMP content might play a positive role in P. aeruginosa's adaptation to oxidative stress.

Regulatory roles of wspF on ROS-mediated resistance of Pseudomonas aeruginosa
To investigate whether intracellular c-di-GMP content is increased under conditions of ROS stress, we measured the intracellular c-di-GMP content and expression of c-di-GMP reporter fusion p cdrA -gfp with H 2 O 2 treatment. Both intracellular c-di-GMP content and p cdrA -gfp expression levels were increased in PAO1 strain after exposure to 4 mM H 2 O 2 for 4 h ( figure 4a,b). Given the increase in intracellular c-di-GMP content in adaptation to H 2 O 2 stress and the presence of wspF mutations in our experimental evolution assay, we hypothesized that the genetic basis of adaptation to ROS was directly linked to the mutation in wspF, which resulted in the de-repression of wspR DGC [8] and a subsequent increased production of c-di-GMP. Hence, we used an isogenic DwspF mutant as previously described [51] to ascertain ROS resistance. We also used 'locked' high and low c-di-GMP content strains of PAO1 by using the p lac -yedQ and p lac -yhjH plasmids, respectively. The yedQ gene encodes an E. coli DGC that synthesizes c-di-GMP [50,55,56], thus the exogenous addition of the p lac -yedQ would enable the constitutive production of c-di-GMP within the species, resulting in the highly aggregative nature of the colonies (electronic supplementary material, figure S1). The yhjH gene encodes an E. coli PDE that degrades c-di-GMP [56,57], so the presence of p lac -yhjH plasmid would cause the reduction of intracellular c-di-GMP levels.
The DwspF and PAO1/p lac -yedQ mutants were more resistant to 4 mM H 2 O 2 (1Â MIC) than PAO1, while PAO1/p lac -yhjH was more sensitive to 4 mM H 2 O 2 than PAO1 (figures 4c and 5a). Epistatic mutations of wspR gene restored p cdrA -gfp expression (figure 4d) to that of the wild-type of a few selected RSCV isolates. This also caused the RSCV isolates to lose their resistance to H 2 O 2 (figure 4e), confirming that PAO1 could adapt to ROS stress simply via the mutagenesis of wspF.
As clinical RSCV isolates with wspF mutations were isolated from CF patients [27], we also tested the ROS resistance of clinical RSCV isolates with known wspF mutations (CF173-2005 isolate of lineage H and a related isolate CF273-2002) [45]. Such RSCV isolates from the CF lungs possess higher c-di-GMP levels [8]. We showed that they were resistant to 4 mM H 2 O 2 and the insertion of the p lac -yhjH plasmid reduced their resistance to ROS (figure 4f). This confirmed that the clinical RSCV isolates were adapted to the presence of ROS in the human body and cemented the involvement of c-di-GMP regulation in biofilm ROS resistance.

C-di-GMP-dependent exopolysaccharides are required for Pseudomonas aeruginosa ROS resistance
To understand how the wsp operon mediates ROS resistance, we tested whether exopolysaccharides could interfere with the action of ROS on P. aeruginosa, as the wsp operon regulates the expression of Pel and Psl exopolysaccharide synthetic     rsob.royalsocietypublishing.org Open Biol. 6: 160162 genes [8]. We showed that both DwspFDpelADpslBCD and DpelADpslBCD/p lac -yedQ mutants, which could not produce Pel and Psl exopolysaccharides, were highly sensitive to H 2 O 2 , even though they contain high intracellular c-di-GMP levels (figure 5a). To identify which of the two exopolysaccharides were more important in conferring ROS resistance, we further showed that the DpslBCD/p lac -yedQ was more sensitive to ROS than DpelA/p lac -yedQ, thus providing the first indication that Psl is more important than Pel in conferring ROS resistance (figure 5b). We then stained the Psl using a fluorescent lectin and observed the formation of small aggregates and enhanced production of Psl in PAO1 with 0.5, 1 and 2 mM H 2 O 2 treatment (figure 5c). Moreover, the evolved RSCVs, DwspF and PAO1/p lac -yedQ mutants produced higher amounts of Psl than PAO1, as indicated by the fluorescence intensity of the Psl stain (figure 5d). We further validated the requirement of Psl for ROS resistance by using a PAO1/p BAD -psl mutant by replacing its native psl operon promoter with an L-arabinose-inducible promoter [58]. Upon addition of L-arabinose, there was an increase in  rsob.royalsocietypublishing.org Open Biol. 6: 160162 biofilm formation via Psl production (figure 5e). Resistance of PAO1/p BAD -psl to ROS improved with increasing L-arabinose concentrations (figure 5f), indicating that Psl production is positively correlated with ROS resistance. Treatment with cellulase, which degrades Psl [58], rendered the PAO1/p BAD -psl sensitive to ROS treatment (figure 5g). Hence, Psl plays a major role in P. aeruginosa ROS resistance. Moreover, Psl acts as a signal for biofilm formation [43], thus could possibly provide positive feedback for biofilm formation and the emergence of RSCVs upon ROS exposure.
3.6. C-di-GMP signalling confers resistance to macrophage phagocytosis ROS are mostly produced by host leucocytes, such as PMNs and macrophages, to kill the invading pathogens, which in turn induce c-di-GMP content to confer ROS resistance. Because the evolved RSCVs and the DwspF mutant showed increased resistance to H 2 O 2 in vitro, we examined whether c-di-GMP signalling could protect P. aeruginosa from phagocytosis by RAW264.7 macrophages. Here, the DwspF mutant and evolved RSCVs were more efficient in evading phagocytosis by macrophages than PAO1 ( figure 6a,b). Although the PAO1/p lac -yhjH strain was subjected to lower phagocytosis levels by the macrophages than DwspF mutant (figure 6b), we found that it was highly cytotoxic to macrophages compared with PAO1 and the DwspF mutant ( figure 6c). Hence, the lower macrophage phagocytosis level could be attributed to increased macrophage killing by PAO1/p lac -yhjH. Moreover, the PAO1/p lac -yhjH was exposed to higher ROS production from macrophages than PAO1 and the DwspF mutant (figure 6d). Further, Psl played a role in evading phagocytosis by macrophages, as DpslBCD/p lac -yedQ was internalized at a higher rate than PAO1/p lac -yedQ and DpelA/p lac -yedQ (figure 6e). However, Psl was not involved in the cytotoxicity against macrophages, as Psl was not a virulence factor and there were no significant differences between PAO1/p lac -yedQ and DpslBCD/p lac -yedQ in killing the macrophage cells ( figure 6f). Nonetheless, increases in c-di-GMP levels by p lac -yedQ plasmid insertion reduced the cytotoxicity to macrophages, which corroborated the DwspF mutant results (figure 6f).

Discussion
Resistance to oxidative stress is essential for pathogens to survive in the host environment. Here, we employed the adaptive experimental evolution assay to investigate P. aeruginosa's ROS resistance via the emergence of RSCVs with enhanced biofilm forming capacity ( figure 7). The production of superoxide in the phagosome was estimated at 1 -4 M [59,60], implying that the use of a millimolar range of H 2 O 2 to induce pathogen evolution in our study was relevant to the host environment. We observed an adaptive evolution trait of mutagenesis in the wspF gene, resulting in c-di-GMP signal induction. Our findings are clinically significant as RSCVs with wspF mutations are frequently isolated from CF patients and these showed similar resistance to oxidative stress as our evolved RSCVs.
The regulatory role of c-di-GMP signalling on oxidative stress response is extendable to other pathogenic species, including Burkholderia cenocepacia and Klebsiella pneumoniae (electronic supplementary material, figure S3).The ROSinduced evolution of RSCVs could potentially explain the emergence of RSCVs by B. cenocepacia and K. pneumoniae during infections [61,62]. The induction of c-di-GMP signalling has been linked to increased production of Psl   rsob.royalsocietypublishing.org Open Biol. 6: 160162 exopolysaccharide in P. aeruginosa [50]. While Psl stiffens the biofilm structure, act as signal for biofilm formation and offers protection against antibiotics [43,63], the role of Psl as a barrier in the biofilm possibly reduces ROS penetration across bacterial membranes and causes cell damage. This confers bacterial cells with a biofilm-based protective mechanism from host immune cell generated ROS and/or antimicrobials. As the oxidative stress resistance mechanism is commonly utilized by various pathogens, we propose the following strategies to modulate c-di-GMP signalling and treat biofilm-associated infections: (i) employing matrix-degrading enzymes, such as cellulase or the PslG hydrolase [64] to degrade Psl-like polysaccharides followed by effective killing of biofilm cells by antibiotics; and (ii) administering antioxidant drugs during early infection stages to prevent pathogen adaptation. To nullify P. aeruginosa's oxidative stress adaptation in our evolution assays, we used the antioxidant L-glutathione, which reduced the emergence of RSCVs. This also indirectly reduced the resistance of P. aeruginosa to host immune clearance strategies such as phagocytosis. Given the lower levels of glutathione and increased oxidative stress from neutrophils in CF patients' lungs, antioxidant therapy could potentially prove effective against inflammation [65,66].
In summary, the interactions between the immune system and pathogens are highly complex, driving the pathogen to adapt accordingly to different host immune system stimuli.
Pseudomonas aeruginosa appears to draw a complex yet fine balance between its numerous DGCs and PDEs, thus leading to phenotypic variation, and ensuring the survival of its species. Hence, chemical manipulation of the c-di-GMP signalling pathway represents a promising strategy to manage chronic bacterial infections.
Data accessibility. The Whole Genome Shotgun bioproject for P. aeruginosa adaptation to ROS has been deposited at in the NCBI Short Read Archive (SRA) database with accession code SRP062804. rsob.royalsocietypublishing.org Open Biol. 6: 160162