Glutathione-Disrupted Biofilms of Clinical Pseudomonas aeruginosa Strains Exhibit an Enhanced Antibiotic Effect and a Novel Biofilm Transcriptome

Abstract

Pseudomonas aeruginosa infections result in high morbidity and mortality rates for individuals with cystic fibrosis (CF), with premature death often occurring. These infections are complicated by the formation of biofilms in the sputum. Antibiotic therapy is stymied by antibiotic resistance of the biofilm matrix, making novel antibiofilm strategies highly desirable. Within P. aeruginosa biofilms, the redox factor pyocyanin enhances biofilm integrity by intercalating with extracellular DNA. The antioxidant glutathione (GSH) reacts with pyocyanin, disrupting intercalation. This study investigated GSH disruption by assaying the physiological effects of GSH and DNase I on biofilms of clinical CF isolates grown in CF artificial sputum medium (ASMDM+). Confocal scanning laser microscopy showed that 2 mM GSH, alone or combined with DNase I, significantly disrupted immature (24-h) biofilms of Australian epidemic strain (AES) isogens AES-1R and AES-1M. GSH alone greatly disrupted mature (72-h) AES-1R biofilms, resulting in significant differential expression of 587 genes, as indicated by RNA-sequencing (RNA-seq) analysis. Upregulated systems included cyclic diguanylate and pyoverdine biosynthesis, the type VI secretion system, nitrate metabolism, and translational machinery. Biofilm disruption with GSH revealed a cellular physiology distinct from those of mature and dispersed biofilms. RNA-seq results were validated by biochemical and quantitative PCR assays. Biofilms of a range of CF isolates disrupted with GSH and DNase I were significantly more susceptible to ciprofloxacin, and increased antibiotic effectiveness was achieved by increasing the GSH concentration. This study demonstrated that GSH, alone or with DNase I, represents an effective antibiofilm treatment when combined with appropriate antibiotics, pending in vivo studies.

INTRODUCTION

Infections with Pseudomonas aeruginosa are associated with significant rates of morbidity and death for patients, regardless of infection site (1, 2). They are particularly problematic for people with cystic fibrosis (CF), in whom infections can persist for substantial periods of time. The formation of a biofilm is a seminal event in the establishment of persistent populations of P. aeruginosa. The extracellular polymeric substance (EPS) matrix of P. aeruginosa biofilms has been widely characterized as being a variable combination of secreted polysaccharides that enhance surface attachment (3), extracellular DNA (eDNA) (4), and extracellular structural proteins (5). The resultant matrix is intrinsically resistant to external environmental parameters such as antibiotics and host immune factors (6, 7). Biofilms that form in CF sputum are unique in their structure, existing as intercellular aggregates near the air-surface interface of sputum (8), rather than typical “mushroom-shaped” microcolonies (9). Populations within these persistent biofilms undergo substantial within-host evolution over time, with seeding dispersion from biofilms resulting in phenotypically divergent strains throughout the CF lung and the emergence of antibiotic-resistant strains (10, 11).

Traditional P. aeruginosa eradication therapies typically have involved untargeted antibiotic usage. Such therapies are not effective, since charged antibiotics, such as aminoglycosides, are repelled by the structure of the biofilm and other antibiotics, such as ciprofloxacin, are ineffective due to low metabolic activity in the anoxic areas biofilms often occupy (7, 12–14). Furthermore, the increasing prevalence of multidrug-resistant P. aeruginosa acquired nosocomially limits treatment options (15, 16). Therefore, methods to disrupt biofilms and to enhance antibiotic effectiveness have been developed, including the use of dispersal agents such as metal chelators and high-affinity anti-EPS peptides (17, 18). However, dispersal agents may actually result in poorer outcomes for the hosts, as dispersed cells have been shown to become more virulent (19). Thus, there is an urgent need for novel antibiofilm agents that are efficacious for in vivo use.

The redox virulence factor pyocyanin and eDNA within the biofilm matrix of P. aeruginosa are both essential for biofilm formation, cellular aggregation, and ongoing biofilm structural integrity (4, 20). Pyocyanin intercalates directly with eDNA (21), thereby conferring structural integrity to the biofilm through increased cell surface hydrophobicity, which enhances intercellular aggregation (22). We showed previously that biofilm formation by P. aeruginosa UCBPP-PA14 (PA14) could be inhibited by DNase I, which removes the eDNA component of biofilms (21). Glutathione (GSH), a host antioxidant, has been shown to interfere directly with the ability of pyocyanin to intercalate with eDNA (21), by reacting with pyocyanin to form a cell-impermeable structure (23).

In this study, we investigated the activity of GSH and DNase I with clinical CF isolates of P. aeruginosa grown in a medium mimicking CF sputum (i.e., artificial sputum medium [ASMDM]) (24). To account for the phenotypic heterogeneity of in vivo strains, we assayed two Australian epidemic strain isogens of P. aeruginosa isolated from the same patient during acute (AES-1R) and chronic (AES-1M) infection and not eradicated in the interim (25). Changes in biofilm structure were assessed by confocal scanning laser microscopy (CSLM) after treatment, and transcriptional changes were assessed by RNA-sequencing (RNA-seq) analysis. Antibiotic effectiveness in biofilms treated with GSH and DNase I was assessed in these isogens and several other CF epidemic strains isolated throughout the world. This work provides a proof of concept for a novel biofilm eradication treatment for P. aeruginosa infections.

MATERIALS AND METHODS

P. aeruginosa strains, culture conditions, and chemotherapy reagents.

The strains used in this study can be seen in Table S1 in the supplemental material. All strains were maintained on Luria-Bertani (LB) agar and grown overnight in LB broth prior to inoculation into ASMDM+. ASMDM was prepared as described previously (24) with the modification of added ferritin (26) and was designated ASMDM+. For all experiments, overnight cultures of P. aeruginosa were grown in LB broth, with shaking at 200 rpm, and diluted to a MacFarland standard of 0.5. Fifty microliters of diluted culture was inoculated into ASMDM+ by pushing the pipette tip 3 to 5 mm below the surface and expelling the bacterial mixture. ASMDM+ was then incubated at 37°C, with loosened caps. Bottles were incubated statically as there was abundant oxygen for biofilm growth in the airspace above the surface. For all plate-based experiments, incubation was accompanied by moderate shaking at 90 rpm, to oxygenate the small air volume in each well. Chemicals used for treatment of all cultures included reduced glutathione (GSH) (Sigma) resuspended in distilled water, 40 U amplification-grade DNase I (Sigma), and ciprofloxacin (Sigma) solubilized in distilled water with HCl.

Confocal scanning laser microscopy of biofilm structures.

AES-1R and AES-1M were cultured for 24 h at 37°C in 3 ml ASMDM+ in polystyrene micropetri dishes, with shaking at 90 rpm (n = 4). After 24 h of incubation, 1× phosphate-buffered saline (PBS) (control), 40 U DNase I, 2 mM GSH, or 2 mM GSH plus 40 U DNase I was added, and the plates were incubated for an additional 8 h. ASMDM+ was then washed off (three times) with ice-cold 1× PBS. The remaining surface-attached biofilm biomass was stained using the Live/Dead BacLight bacterial viability kit for microscopy (Life Technologies, USA), according to the manufacturer's instruction. To demonstrate the effect of pyocyanin in enhancing the biofilm of AES-1M, ASMDM+ was supplemented with 100 μM pyocyanin (Sigma). Cultures were allowed to grow for 24 h, washed, and stained as outlined above. Staining, CSLM (Olympus FV1200; Olympus, Australia), and image quantification of thickness (in micrometers), live and dead biovolumes (square micrometers per cubic micrometers), and substratum surface coverage (in square micrometers), using ImageJ (v1.49), were performed as described previously (21). Images were taken in technical triplicate for each individual replicate and were quantified for features as outlined, to give an average for each individual replicate. Results were analyzed with one-way analysis of variance (ANOVA), with Tukey's post hoc correction for significance after multiple comparisons.

Crystal violet quantification of biofilm biomass.

AES-1R and AES-1M were cultured for 72 h in 2 ml ASMDM+ in 24-well flat-bottom clear polystyrene plates (Corning) at 37°C, with shaking at 90 rpm (n = 3). After 72 h of incubation, 1× PBS (control), 40 U DNase I, 2 mM GSH, or 2 mM GSH plus 40 U DNase I was added, and the plates were incubated for an additional 8 h. A crystal violet biofilm assay was then performed as described previously (27). Results were analyzed using one-way ANOVA, with Tukey's post hoc correction for significance after multiple comparisons.

RNA isolation and sequencing.

AES-1R was cultured for 72 h at 37°C in 10 ml ASMDM+ in McCartney bottles (n = 4), to generate a mature biofilm of respiring cells visible with the addition of 2,3,5-triphenyltetrazolium chloride (TTC); 1× PBS (control) or 2 mM GSH was added after 72 h, and cultures were incubated for an additional 8 h. For cell extraction, cultures were pooled within group in technical triplicate in Falcon tubes, to reduce biological variance, prior to the addition of ice-cold 1× PBS, vortex-mixing for 10 s, and centrifugation at 12,000 × g for 20 min at 4°C. The supernatant, containing medium and loose cells, was decanted, and the washing steps were repeated several times, until only a firm mass of red-stained cells (respiring cells stained with TTC) remained on the tube walls. The decanted supernatants were stored at 4°C for further assay. The cell mass was removed, and RNA was extracted in duplicate extracts using the miRNeasy kit (Qiagen), according to the manufacturer's protocols. The integrity of the RNA was assessed through bioanalysis using the Agilent Bioanalyzer 2100 platform (Agilent Technologies). An RNA integrity number (RIN) score of 7 was considered acceptable for downstream use. Total RNA was then depleted of rRNA using the Ribo-Zero Gold (Bacteria) kit (EpiCentre), according to the manufacturer's specifications. cDNA library preparation for RNA-seq analysis was performed using the Illumina TruSeq Stranded kit (catalog no. RS-122-2101) according to the manufacturer's specifications, with the following modifications: preparation began at fragmentation, and the number of PCR cycles was reduced from 15 to 12. RNA-seq analysis was performed on the NextSeq 500 platform using NextSeq500 v2 reagents (catalog no. FC-404-2001) with 75-bp paired-end (PE) reads, for a total of 30 million reads per sample, at the Ramaciotti Centre for Genomics, University of New South Wales (Kensington, Australia).

RNA-seq analysis.

Quality control of raw RNA-seq reads was performed with Illumina BaseSpace using the FastQ Toolkit (v2.0). Raw reads were trimmed of Illumina TruSeq sequencing adapters and filtered for an average Phred Q score minimum of >30 and a minimum read length of 32 nucleotides after trimming. Reads were further processed by trimming the 3′ ends of bases with Q scores of <30. Processed output files were aligned with the AES-1R genome (GenBank accession no. CP013680) using Bowtie2 (v2.2.5). Output sam files were converted to the bam format using Samtools (v0.1.19) and were name-sorted prior to input into HTSeq (v0.6.1). HTSeq counting was performed in union mode against a Gene Transfer Format (GTF) annotation of AES-1R. Reads were counted against coding DNA sequences (CDS), and noncoding RNA sequences were annotated based on homology to PAO1 (37). Differential gene expression analysis was performed using both the edgeR (29) and DESeq2 (74) R packages. Low-expression reads were filtered from analysis, and a minimum false discovery rate (FDR) of <0.05 was accepted as indicating differentially expressed (DE) genes after Benjamini-Hochberg post hoc correction. DE genes commonly identified by both programs are reported in Table S2 in the supplemental material. A principal-component analysis (PCA) plot was generated and used to assess reproducibility. Genes were binned according to known functionality, and totals were generated for upregulated and downregulated genes.

Quantitative PCR validation of transcriptomic results.

First-strand cDNA synthesis and quantitative PCR (qPCR) analysis were conducted with (i) total RNA samples used for RNA-seq analysis (n = 2) and (ii) total RNA from a separate set of biological replicates (n = 3) grown identically to those for RNA-seq analysis. First-strand cDNA synthesis was performed with SuperScript II reverse transcriptase (Life Technologies), using standard laboratory protocols. qPCR was performed using Kapa SYBR Fast qPCR master mix (Life Technologies), on a Rotor-Gene 6000 thermal cycler (Qiagen). Primers used for qPCR are presented in Table S5 in the supplemental material. A linear regression analysis was performed to assess the correlation of qPCR fold changes for target genes with RNA-seq fold changes for target genes, as reported by both edgeR and DESeq2.

Quantification of pyocyanin generated in 72-h biofilms.

Supernatants removed after the centrifugation of cultures were assayed for total pyocyanin concentrations, from cultures used for RNA-seq analysis (n = 2) and cultures from the repeat experiment performed for qPCR analysis (n = 3). Supernatants were filtered twice with 0.22-μm filters, to obtain cell-free supernatants. A standard curve for pyocyanin was generated by the addition of purified pyocyanin (Sigma) to 1× PBS, in a clear 96-well flat-bottom plate (Corning), and absorbance was assayed at 691 nm using a Tecan M1000 plate reader (Tecan), as described previously (21). This curve was used to quantify pyocyanin present in supernatants, on the basis of absorbance at 691 nm.

Quantification of catalase production and activity.

Seventy-two-hour cultures identical to those for RNA-seq analysis were generated (n = 3). After 8 h of incubation with or without GSH, cells were centrifuged at 12,000 × g for 20 min at 4°C. The supernatant was decanted and filtered twice with 0.22-μm filters and was used to assay total supernatant catalase and intracellular catalase in the cell pellets. Quantification of catalase activity (units per milligram of protein) was performed as described previously (31). A two-tailed Student's t test was performed to assess significant differences between control and GSH-treated groups.

Quantification of pyoverdine activity.

Seventy-two-hour cultures identical to those for RNA-seq analysis were generated (n = 6). After 8 h of incubation with or without GSH, cells were centrifuged at 12,000 × g for 20 min at 4°C. The supernatant was decanted and filtered twice with 0.22-μm filters, to obtain cell-free supernatant. Quantification of apo-pyoverdine and ferric pyoverdine was performed as described previously (32). A two-tailed Student's t test was performed to assess significant differences between control and GSH-treated groups.

Quantitative assay of nitrate/nitrite contents.

Seventy-two-hour cultures identical to those for RNA-seq analysis were generated (n = 5). After 8 h of incubation with or without GSH, 80-μl samples of ASMDM+ supernatant were removed separately from the air-surface interface and the anoxic base of biofilms of cultures. These samples were transferred to a 96-well flat-bottom clear plate in technical triplicate for assay of nitrate and nitrite. A colorimetric assay was used according to the manufacturer's specifications (Cayman Chemical Co., Ann Arbor, MI). Absorbance was measured at 540 nm using a Tecan M1000 plate reader (Tecan), blank corrected appropriately, and quantitated using a standard curve of nitrate alone and nitrite alone. Results are presented as averages of nitrate/nitrite contents at the air-surface interface and anoxic base of biofilms. A two-tailed Student's t test was performed to assess significant differences between control and GSH-treated groups.

Antibiotic susceptibility after treatment with GSH and DNase I.

All strains were assayed for antibiotic susceptibility using a standard disc diffusion assay (see Table S1 in the supplemental material). The ASMDM+ was modified so as to not include antibiotics (24). Strains were incubated for 72 h in 2 ml ASMDM+ in 24-well flat-bottom clear polystyrene plates (Corning) at 37°C, with shaking at 90 rpm (n = 4). At the 72-h time point, 1× PBS (control), 40 U DNase I, 2 mM GSH, or 2 mM GSH plus 40 U DNase I was added, and the cultures were incubated for 8 h. Ciprofloxacin (5 μg/ml) was added to a subset of control, GSH-treated, and GSH/DNase I-treated cultures, and the cultures were incubated for an additional 12 h. Biofilms contained in each well were thoroughly homogenized by pipetting and then were serially diluted in 1× PBS at the 92-h time point, before dilutions were plated on LB agar. After overnight growth, colonies were counted and expressed as CFU per milliliter. A one-way ANOVA was performed on log₁₀-transformed counts for treatment groups within each strain, with Sidak's post hoc correction for significance after multiple comparisons.

Contribution of GSH alone to enhanced antibiotic activity in P. aeruginosa biofilms.

A replicate of the antibiotic susceptibility assay described above was performed in LB broth instead of ASMDM+ to use an absorbance-based assay. Single colonies of each strain were resuspended in 500 μl of full-strength LB broth, and wells were inoculated with 50 μl to an optical density at 600 nm (OD₆₀₀) of 0.045 ± 0.005. After 72 h of incubation with shaking at 90 rpm, 1× PBS, 2 mM GSH, 5 mM GSH, or 10 mM GSH was added, and the plates were incubated for an additional 8 h. Cultures were then treated with either 1× PBS or ciprofloxacin (5 μg/ml) and incubated for an additional 12 h. TTC was added to the wells to a final concentration of 0.1% (wt/vol), and the cells were incubated at 37°C for an additional 4 h. Cells were then assayed for absorbance at 540 nm using a Tecan M1000 plate reader (Tecan) and were blank corrected appropriately. Percent survival was calculated as the OD₅₄₀ of treated biofilms divided by the OD₅₄₀ of untreated biofilms.

Data set accession number.

All data (for control and GSH-treated AES-1R) are accessible through NCBI Gene Expression Omnibus (GEO) Series accession number GSE75334.

RESULTS AND DISCUSSION

GSH and DNase I disrupt P. aeruginosa biofilms in artificial sputum medium.

We first investigated how treatment of biofilms with GSH and DNase I was reflected at the microscopic level, utilizing CSLM with 24-h biofilm cultures of AES-1R and AES-1M. All treatments (DNase I, GSH, and GSH plus DNase I) resulted in significant qualitative effects on the biofilm architecture for both AES-1R and AES-1M (Fig. 1A to D and H to K). Both GSH alone and DNase I alone significantly reduced the live and dead biovolumes, biofilm surface coverage, and biofilm thickness for AES-1R (Fig. 1E and F). With respect to AES-1M, GSH alone significantly reduced live cell biovolume, biofilm thickness, and biofilm surface coverage, while a significant decrease in biofilm surface coverage occurred with DNase I treatment; however, there was no significant effect on either live or dead biovolume (Fig. 1L to N). Considering both strains, these findings demonstrate that GSH alone is able to disrupt biofilms. GSH was shown previously to interfere with the intercalation of pyocyanin with eDNA by forming a glutathionyl-pyocyanin conjugate (21, 23), whereas pyocyanin would normally function by intercalating with eDNA to afford enhanced intercellular aggregation (22). Our findings here suggest that GSH alone is able to disrupt biofilms of these clinical strains through this mechanism, with disruption being enhanced by the presence of DNase I. These findings are consistent with previous studies showing the disrupting effects of GSH and DNase I on biofilms of the laboratory strain PA14 (4).

FIG 1.

Disruption by GSH and GSH plus DNase I of the architecture of 24-h AES-1R and AES-1M biofilms grown in ASMDM+. Green indicates live cells, and red indicates dead cells. (A to D) CSLM images. (A) Control 24-h AES-1R biofilm. (B) Twenty-four-hour AES-1R biofilm treated with 40 U DNase I. (C) Twenty-four-hour AES-1R biofilm treated with 2 mM GSH. (D) Twenty-four-hour AES-1R biofilm treated with 2 mM GSH and 40 U DNase I. (E) Surface areas of 24-h AES-1R biofilms. All treatments significantly reduced the surface coverage, relative to control (P < 0.0001), with GSH plus DNase being the most effective treatment (P < 0.001). (F) Total biovolumes of 24-h AES-1R biofilms. All treatments significantly reduced the total live biovolume of biofilms (P < 0.0001), with GSH plus DNase I being the most effective treatment (P < 0.0001). (G) Thicknesses of 24-h AES-1R biofilms. All treatments significantly reduced the biofilm thickness, with the combination of GSH and DNase I resulting in the largest reduction, compared to control (P < 0.0001). (H to K) CSLM images. (H) Control 24-h AES-1M biofilm. (I) Twenty-four-hour AES-1M biofilm treated with 40 U DNase I. (J) Twenty-four-hour AES-1M biofilm treated with 2 mM GSH. (K) Twenty-four-hour AES-1M biofilm treated with 2 mM GSH plus 40 U DNase I. (L) Surface areas of 24-h AES-1M biofilms. All treatments significantly reduced the surface coverage, relative to control (P < 0.0001), with GSH plus DNase I being more effective than DNase I alone (P < 0.01). (M) Total biovolumes of 24-h AES-1M biofilms. GSH and GSH plus DNase I significantly reduced the total live biovolume (P < 0.0001), with GSH plus DNase I being more effective than DNase I alone (P < 0.001). Treatment with GSH alone resulted in a significant increase in dead cells stained, relative to control (P < 0.0001), DNase I (P < 0.001), and GSH plus DNase I (P < 0.01). (N) Thicknesses of 24-h AES-1M biofilms. All treatments significantly reduced biofilm thickness, with the combination of GSH and DNase I resulting in the largest reduction, compared to control (P < 0.0001). All biofilms were cultured in ASMDM+. CSLM images were taken 8 h posttreatment and are representative of all biological replicates (n = 4). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Results are shown as mean ± standard deviation (SD).

The finding that DNase I treatment did not affect AES-1M biovolume was unexpected, given that there was a small decrease in biofilm thickness after DNase I activity; however, there was a visible difference in biofilm architecture between AES-1R and AES-1M under the control conditions (Fig. 1A and H). This was reflected in differences in overall biovolume and biofilm thickness between the strains (Fig. 1F, G, M, and N). AES-1M forms more distinct and developed microcolonies than AES-1R, which may influence DNase I access to the biofilm. Previously published transcriptomic comparisons between the two isogens suggested that AES-1M biofilms differ in structural constituents, with significant upregulation of adhesive polysaccharide, alginate, and other biofilm development genes (25), which may be influencing this result. Overall, GSH plus DNase I was the most effective disruptor of both AES-1R and AES-1M biofilms, resulting in pronounced significant decreases in live cell biovolume, biofilm thickness, and biofilm surface coverage (Fig. 1E to G and L to N).

Notably, AES-1M did not produce any noticeable pyocyanin when cultured in LB broth or ASMDM+ (see Fig. 5 and 6 for photographs). Thus, it is plausible that the effects of GSH on AES-1M are due to autooxidation of GSH to oxidized glutathione (GSSG) to produce H₂O₂. H₂O₂ generated in this manner may act as a dispersant of P. aeruginosa biofilms, as has been observed for biofilms of several other Gram-negative species (28). In this sense, it is likely that GSH can act as a biofilm-disrupting agent independent of the presence of pyocyanin.

FIG 5.

Antibiotic susceptibility of biofilms of clinical CF strains of P. aeruginosa disrupted with GSH alone or combined with DNase I. The contributions of GSH alone and GSH plus DNase I in enhancing the effect of ciprofloxacin were determined by serially diluting 90-h biofilms of clinical CF P. aeruginosa strains grown in ASMDM+, to obtain CFU values (n = 4 each strain) (upper). Isolates were phenotypically heterogeneous regarding both the amounts and types of phenazine produced (lower). Eight hours of exposure to either GSH alone or GSH plus DNase I prior to 12 h of incubation with ciprofloxacin (5 μg/ml) resulted in highly significant reductions in CFU values, relative to control (P < 0.0001 for all except AES-1M [P < 0.001]). Ciprofloxacin reductions did not differ significantly between pretreatment with GSH alone and pretreatment with GSH plus DNase I for any of the strains assayed. Statistical analyses were performed for the strains individually, using a one-way ANOVA with Sidak's post hoc correction for significance. ***, P < 0.001; ****, P < 0.0001. Results are shown as mean ± SD.

FIG 6.

Antibiotic susceptibility of clinical and laboratory strains of P. aeruginosa treated with a molar excess of GSH. The effects of GSH concentrations on biofilm disruption were assayed by 8-h pretreatment of 72-h biofilms in LB medium with 2 mM, 5 mM, or 10 mM GSH, followed by 12 h of incubation with 5 μg/ml ciprofloxacin (n = 3 for each strain). To quantify respiration, cells were then incubated for 4 h at 37°C with 0.1% (wt/vol) 2,3,5-triphenyltetrazolium chloride, followed by OD₅₄₀ measurements. Percent survival was calculated as the OD₅₄₀ of treated biofilms divided by the OD₅₄₀ of untreated biofilms (upper). Treatment with 10 mM GSH resulted in the most significant decreases, irrespective of strain. The absence of red formazan color in wells denoted a lack of respiration (lower).

The biofilms of AES-1M were smaller in terms of both thickness and biovolume than were those of AES-1R and resembled those of a phenazine biosynthesis knockout in PA14 (PA14 ΔphzA-G) (20). AES-1M is phenotypically distinct from AES-1R after 10.5 years of within-host evolution and does not produce pyocyanin. However, levels in sputum of chronically infected individuals with CF can reach as high as 130 μM (33). It is thus likely that exogenous pyocyanin is a factor contributing to the structure of AES-1M biofilms in vivo.

To demonstrate the contribution of exogenous pyocyanin to the biofilm integrity of a host-adapted clinical isolate from a chronic infection, we cultured AES-1M in the presence and absence of 100 μM pyocyanin (Fig. 2A and B). Figure 2B demonstrates the typical intercellular aggregates that formed in the presence of pyocyanin. Quantitatively, the presence of pyocyanin significantly increased the total biovolume, biofilm thickness, and biofilm surface coverage of AES-1M biofilms, relative to the control (Fig. 2C to E). These results indicate that the robust biofilms of AES-1M grown in the presence of pyocyanin are more reflective of chronic infection biofilms likely occurring in vivo. Therefore, GSH treatment would most likely disrupt the biofilm of a chronic strain formed in the presence of pyocyanin in a manner similar to that observed for an acute infection strain such as AES-1R. These findings are consistent with previous studies demonstrating the ability of pyocyanin to rescue the biofilms of phenazine-deficient mutants (20, 21), and this is the first such study demonstrating the biofilm-enhancing properties of pyocyanin in a chronic infection CF isolate ex vivo.

FIG 2.

Pyocyanin enhancement of the structure of AES-1M biofilms grown in ASMDM+. Green indicates live cells, and red indicates dead cells. (A and B) CSLM images. (A) Control 24-h AES-1M biofilm. (B) Twenty-four-hour AES-1M biofilm grown in the presence of 100 μM pyocyanin. (C) Surface areas of 24-h AES-1M biofilms. The addition of pyocyanin facilitated a greater amount of substratum surface coverage, relative to control (P < 0.01). (D) Total biovolumes of 24-h AES-1M biofilms. The addition of pyocyanin to ASMDM+ resulted in a significantly greater biovolume, relative to control (P < 0.05). (E) Thicknesses of 24-h AES-1M biofilms. The addition of pyocyanin to ASMDM+ resulted in a significantly greater biofilm thickness, relative to control (P < 0.05). Images are representative of all biological replicates (n = 4). Statistical significance analyses were performed using a two-tailed t test for significance. *, P < 0.05; **, P < 0.01. Results are shown as mean ± SD.

The results demonstrated by CSLM were limited, due to difficulties involved in removing the viscous ASMDM+ at time points after 24 h. To demonstrate these crucial findings in the more structurally developed and resistant mature biofilms that are representative of those found in vivo (34, 35), we assessed whether the addition of GSH and DNase I to established biofilms caused disruption. AES-1R and AES-1M were grown in 24-well plates for 72 h and stained with crystal violet. Both GSH alone and GSH combined with DNase I were sufficient to reduce AES-1R biofilms significantly (Fig. 3A); however, only the combination of GSH with DNase I significantly reduced AES-1M biofilms (Fig. 3B). As mentioned previously, AES-1M is known to differ phenotypically in biofilm-related gene expression (25), which might have enabled it to remain surface attached after treatment with GSH alone.

FIG 3.

Disruption by GSH and GSH plus DNase I of surface-associated 72-h AES-1R and AES-1M biofilms grown in ASMDM+. (A) Biomasses of 72-h AES-1R biofilms grown in 24-well plates (n = 3) and stained with crystal violet. The addition of either GSH (P < 0.05) or GSH plus DNase I (P < 0.01) resulted in significant decreases in surface-attached biofilms. (B) Biomasses of 72-h AES-1M biofilms grown in 24-well plates (n = 3) and stained with crystal violet. The addition of GSH plus DNase I (P < 0.01) resulted in significant decreases in surface-attached biofilms. All biofilms were cultured in ASMDM+. *, P < 0.05; **, P < 0.01. Results are shown as mean ± SD.

GSH-disrupted biofilms exhibit a novel transcriptome.

GSH alone was found to be sufficient to disrupt both immature and mature biofilms of AES-1R. Therefore, we sought to assess the transcriptomic effects of GSH on mature biofilms of this strain. AES-1R was cultured in ASMDM+ for 72 h, to generate a mature biofilm, prior to the addition of GSH to biofilms for 8 h. To control for the possibility of a molar deficit of GSH to pyocyanin (23), the pyocyanin concentration in these cultures was measured and found to average 83 ± 5.2 μM (Fig. 3A). Thus, the 2 mM GSH added was deemed sufficient to enable an effect over the 8 h of treatment (23). This concentration of pyocyanin (83 μM) also reflects the levels typically found in chronically infected CF patients (33).

RNA-seq analysis revealed that a considerable biological change had occurred as a result of the addition of GSH to AES-1R biofilms. A total of 587 genes (∼10% of the AES-1R genome) were differentially expressed (DE) after GSH treatment, compared to the control, at a false discovery rate (FDR) of <0.05, as commonly identified as DE by both edgeR and DESeq2 (see Fig. S1A in the supplemental material). A full list of DE genes, fold changes, and corresponding P values can be found in Table S2 in the supplemental material. The DE genes with known functions were categorized and are shown in Fig. S1D. A brief rationale and analysis of the bioinformatics approach used in the present study can be found in the legend to Fig. S1.

We further sought to validate the transcriptomic results by (i) performing qPCR on a selection of genes with the RNA used for RNA-seq (Fig. 4A) and (ii) extracting biofilm bacteria and generating RNA from another set of ASMDM+ biological replicates and repeating the qPCR on the same gene set, to ensure that the results were biologically reproducible (Fig. 4B). The genes were selected on the basis of relevance to mature biofilm maintenance (glgA and katE), a virulence factor highly expressed in stressed P. aeruginosa (i.e., pyoverdine [pvdM and fpvA]), and antibiotic resistance (oprH and oprJ), as these may indicate changes in resistance in the disrupted biofilm. The correlation coefficients derived indicated that the RNA-seq changes were real and, importantly, were biologically reproducible (Fig. 4B). Fold changes found in qPCR target genes were more strongly correlated technically and biologically with those obtained by DESeq2 than with those obtained by edgeR (see Fig. S1 in the supplemental material).

FIG 4.

Validation of RNA-seq results by qPCR and biochemical assays. (A) qPCR analysis of AES-1R target genes using the same RNA as extracted for RNA-seq (n = 2). The R² values for correlations indicated very good reproducibility between techniques, with DESeq2 (R² = 0.8742) being a more accurate indicator of fold changes than edgeR (R² = 0.8486). (B) qPCR analysis of AES-1R target genes using RNA generated from three biological replicates in an identical experiment. The R² values for correlations of target genes demonstrated a biologically reproducible event after the addition of GSH to mature biofilms. qPCR results demonstrated better correlation with DESeq2 results (R² = 0.8691) than edgeR results (R² = 0.8283). (C) Catalase activity in AES-1R cultures. The catalase activity in both cytosolic fractions (P < 0.05) and culture supernatants (P < 0.01) was significantly decreased after the addition of GSH (n = 3). (D) Pyoverdine concentrations in AES-1R cultures. The concentrations of pyoverdine, in both the apo-pyoverdine (apo-PVD) (P < 0.05) and ferric pyoverdine (Fe-PVD) (P < 0.05) forms, in AES-1R culture supernatants were significantly increased after the addition of GSH (n = 6). (E) Nitrite and nitrate content in AES-1R cultures. The average nitrite concentration was significantly decreased in cultures (P < 0.05) after the addition of GSH (n = 5). All experiments were performed with 80-h AES-1R biofilms in ASMDM+. Linear regression analysis was performed for qPCR experiments. Statistical significance for the catalase, pyoverdine, and nitrate/nitrite assays was analyzed with a two-tailed Student's t test. *, P < 0.05; **, P < 0.01. Results are shown as mean ± SD.

The findings for two catalase genes that were shown to be significantly DE by RNA-seq analysis were validated by a biochemical assay. Catalase activity (both secreted and intracellular) was measured spectrophotometrically and was significantly decreased in both the culture supernatant and the cytosol after treatment with GSH, compared to control (Fig. 4C). Pyoverdine biosynthesis genes that were found to be significantly DE by RNA-seq were validated by spectrophotometric measurement of both apo-pyoverdine and ferric pyoverdine in GSH-treated cultures, which contained significantly higher concentrations, compared to control (Fig. 4D).

All DE genes were then compared to DE genes from other studies that investigated the transcriptomes of biofilms versus planktonic cells and dispersed cells (19, 36–41) (see Table S2 in the supplemental material). From this comparison, it was evident that, after GSH treatment, biofilms displayed a transcriptomic profile that was distinctly different from those of both mature biofilms and dispersed cells.

Among the metabolic processes demonstrating this distinct difference is cyclic diguanylate (c-di-GMP) cyclase expression. Dispersion is a process that occurs in mature biofilms, during which planktonic cells are released to seed distal locations and to form new biofilms (42). Dispersal agents induce dispersion by acting directly on receptors for environmental cues such as increases in nitric oxide (NO) (43). NO-responsive phosphodiesterase (PDE) family proteins then enzymatically degrade c-di-GMP (43, 44). c-di-GMP gates the transition between motile and sessile lifestyles, and increases in c-di-GMP levels result in upregulation of exopolysaccharide production and the type VI secretion system (T6SS) and generally enhanced biofilm attachment and formation (45, 46).

Contrary to this, we found that GSH-induced biofilm disruption did not lead to differential expression of any genes coding for PDE family proteins, while genes involved in c-di-GMP synthesis (siaA and siaD) were significantly upregulated (1.56- and 1.75-fold, respectively) (47). This upregulation runs counter to previous profiling of dispersed cells, which showed downregulation of siaA and siaD, relative to biofilms (19). However, siaA and siaD are implicated in enhancement of cell aggregation after detergent-induced disruption of biofilms with sodium dodecyl sulfate (48), which suggests that a similar nondispersal form of disruption is occurring after GSH treatment.

In concordance with the known upregulation of T6SS by c-di-GMP, several T6SS genes (tssBCEFG1, hcp1, and clpv1) from the Hcp secretion island I (HSI-I) cluster that encode structural proteins were significantly upregulated (1.5- to 2.2-fold) in GSH-treated biofilms, as were HSI-II (hsiC2) and HSI-III (hsiB3) genes (1.6- and 2.4-fold, respectively) (46, 49, 50). Interestingly, a pair of cognate effector-immunity phospholipase genes (tle4 and tli4) that are regulated through the HSI-II pathway (51) were also found to be upregulated in GSH-treated cells (1.8- and 1.4-fold, respectively). T6SS phospholipases usually function to facilitate interspecies competition and target phosphatidylcholine in the membranes of bacterial competitors (51). These genes are immediately upstream of a gene coding for an uncharacterized Vgr-containing protein, AES1R_01014 (homologous to PA14_44900). In the PAO1 genome, tle4 and tli4 are upstream of vgrG2a. The coexpression of these genes is in line with previous studies showing that secretion of effector proteins is dependent on the presence of VgrG proteins (52). Our findings are particularly interesting as the T6SS upregulation observed may be a defensive mechanism facilitating survival of disrupted biofilms in environments containing bacterial competitors, such as the polymicrobial CF lung.

c-di-GMP is also known to repress flagellar biosynthesis (53), and the flagellar genes flgF and fliK were downregulated in the GSH-containing biofilms (1.5- and 1.6-fold, respectively). This is in contrast to dispersed cells, which typically become motile when they are removed from the biofilm matrix (19, 36). Therefore, rather than dispersing, the GSH-disrupted biofilm biomass appears to be seeking to enhance biofilm formation by increasing the c-di-GMP biosynthetic capability, thereby driving T6SS upregulation concomitant with downregulation of flagellar biosynthesis.

Genes encoding components of the alternate Tad pilins, i.e., tadABCDZ, rcpAC, flp, and pprA (54), were all downregulated (1.5- to 1.9-fold) in GSH-treated biofilms. In other Gram-negative organisms, these Tad pilins contribute to virulence through their capacity to form tightly packed biofilms (55). This finding is particularly interesting, given that these genes have consistently been shown to be upregulated in biofilms and subsequently downregulated in dispersed cells (19, 37, 40). Therefore, Tad pilins may play a role in P. aeruginosa biofilms through maintenance of tightly aggregative structures. The upregulation of type IVa pilin subunit pilA in GSH-treated biofilms was also in line with biofilm disruption rather than dispersal, since PilA is known to bind directly to eDNA with high affinity, to facilitate biofilm formation (9, 56). Thus, it is tempting to surmise that the cells of the disrupted biofilm are attempting to reattach to eDNA present in the remaining biofilm EPS.

One of the most striking changes observed in GSH-treated biofilms was upregulation of the full suite of pyoverdine biosynthesis genes, as validated by both qPCR and biochemical assays. This contrasts with previous studies showing downregulation of pyoverdine biosynthesis after biofilm dispersal (19, 37). Active expression of pyoverdine is crucial for complex biofilm structure formation (57). Low iron levels in the ASMDM+ after biofilm disruption are likely at this particular time point (8 h) and could be positively regulating the production of pyoverdine, as well as the HSI-II cluster of the T6SS, which is also known to be induced in periods of iron starvation (58, 59); this may explain the upregulation of hsiC2 and tle4/tli4 noted above. Pyoverdine upregulation may also be due to increased permeability of the disrupted biofilm to ASMDM+ components such as serum albumin (60). We showed previously that AES-1R biofilm growth in ASMDM+ coincided with upregulation of the low-affinity siderophore pyochelin, relative to reference strains PAO1 and PA14 (26). However, P. aeruginosa preferentially produces pyoverdine over pyochelin under severe iron limitations, due to the higher affinity of pyoverdine for iron, which may explain the different siderophore production profiles in GSH-disrupted biofilms (26, 61). Thus, pyoverdine upregulation here may constitute an immediate response to ASMDM+ exposure after biofilm disruption, in order to facilitate biofilm reformation.

An interesting putative marker of mature biofilm growth that was found to be downregulated (1.5- to 4.7-fold) in GSH-disrupted biofilms was the locus AES-1R_02209-02271 (homologous to locus PA2134-2190 in PAO1), which is located on P. aeruginosa genomic island 1 (PAGI-1) and is associated with pathogenic isolates (62). Termed the putative biofilm maintenance locus (PBML), it is likely involved in biofilm maintenance (40) and, given the characteristics of some of its genes, it is likely expressed during periods of nutrient starvation. PBML includes several genes involved in glycogen and trehalose metabolism (glgABPX and treSYZ, respectively), as well as two catalases (katE and katN). The downregulation of the PBML in GSH-treated AES-1R biofilms was in contrast to its upregulation in biofilms, relative to planktonic cells, observed elsewhere (37, 40, 41) and similar to its downregulation in cells dispersed from biofilms (19). Disruption of the biofilm with GSH would be expected to allow for greater nutrient permeability to the biomass and thus is the likely cause of the observed downregulation of the PBML genes. The downregulation of catalase genes in GSH-disrupted AES-1R biofilms was in agreement with qPCR and biochemical assay results. This is interesting, as PBML catalases may also be involved in biofilm protection in response to external factors (63), and an increase in catalase expression in response to hydroxyl radicals from GSH autooxidation would have been expected.

Biofilm cells of GSH-disrupted AES-1R biofilms upregulate nitrogen metabolism and translational machinery genes.

Many of the genes encoding the proteins of the 30S (rpsCDEGHJKLMQSU) and 50S (rplBCDFJLMNOPRVWX and rpmCDHJ) ribosomal subunits were upregulated (1.5- to 3.2-fold) after GSH disruption, as were genes involved in translation initiation (infA) and tRNA modification (gatBC and yleA) (1.6- to 1.8-fold). These findings are consistent with previous gene expression studies showing upregulation of these genes in dispersed cells (19, 37, 40), as well as in cells on the external metabolically active zone of intact biofilms, relative to the interior (64). This would suggest that the GSH-disrupted biofilm is metabolically active. Further evidence for a metabolically active biofilm was provided by the significant upregulation of most nitrite metabolism genes (nirFGHJLMNS), as well as a nitric oxide reductase gene (norC) (1.5- to 2.5-fold) and microaerophilic respiration genes (ccoP2Q2O2N2) (2.0- to 2.2-fold). Nitrite contents were subsequently found to be significantly decreased in cultures after the addition of GSH (Fig. 4E), which indicates that this metabolic change takes place during AES-1R biofilm disruption. This suggests that the disrupted biofilm is utilizing available nitrite for metabolic processes, by upregulating the nitrite reductase genes nirFGHJLMNS.

It was shown previously that dispersed cells upregulate nitrogen metabolism and microaerophilic respiration genes, relative to biofilm (19, 41, 44). It is well established that the ability to utilize nitrate or nitrite as a terminal electron acceptor in the absence of oxygen is crucial for biofilm formation (65, 66). A recent analysis of expectorated CF sputum found that oxygen tension decreased exponentially inward from the surface and oxygen tension was inversely correlated with bacterial load (30). This yields a hypoxic interior, rich in nitrate (8), which results in upregulation of the nitrite reductase locus nirFGHLMNS (67). Similar upregulation of these and related genes (ccoP2Q2O2N2 and norC) was found under conditions of reduced oxygen tension (66). Such conditions would be expected to occur in ASMDM+, which closely mimics the consistency and nutrient content of CF sputum. The upregulation of nitrogen metabolism and the translational machinery suggests that P. aeruginosa biofilms disrupted with GSH have adapted to increased nutrient availability as the biofilm has become more permeable, thus shifting environmental parameters. In concordance with this theory, indicators of the cell growth phase in the GSH-disrupted biofilm, i.e., the exponential growth phase sigma factor rpoD and the RNA polymerase subunit rpoA, were both upregulated almost 2-fold (68, 69). rpoA was shown previously to be significantly downregulated in stationary-phase planktonic cultures and in biofilms (40, 69). Additionally, these sigma factors are upregulated following dispersal of metabolically active cells (19).

Disruption of biofilms with GSH results in upregulation of metabolic and ribosomal protein genes, as well as sigma factors, which suggests that the biofilm population is transitioning to a metabolically active state and increasing the capacity for ribosomal synthesis while attempting to reform the disrupted biofilm. We showed previously that a mutant deficient in phenazine production in PA14 (PA14ΔphzA-G) had significantly lower cell surface hydrophobicity, compared to the wild type, as well as unfavorable physiochemical interactions between cells (22). Subsequently, we showed that GSH could disrupt the intercalation of pyocyanin into eDNA (21) and decrease intercellular aggregation (22). In combination, GSH and pyocyanin form a cell-impermeable structure, which abrogates the physiological functions of pyocyanin (23). On the basis of those findings together with the present findings, we propose that GSH activity results in a physiochemically disrupted biofilm with a novel cell physiology. Significantly, the GSH-disrupted biofilms do not display a transcriptomic profile consistent with increased virulence, which has been observed for dispersed biofilms (19); rather, the disrupted biomass appears ready to reform as a biofilm without first reverting to a dispersed cell phenotype.

GSH, alone and with DNase I, enhances antibiotic activity for mature biofilms.

P. aeruginosa biofilms that have been dispersed through a variety of mechanisms, including iron chelation (17), c-di-GMP depletion (19), and nitric oxide (NO) treatment (70), are more susceptible to antibiotics. To test whether GSH- and GSH/DNase I-disrupted biofilms were also more susceptible to antibiotics, we assayed biofilms of clinical epidemic strains of P. aeruginosa from Australia and the United Kingdom and a globally isolated strain, clone C, for antibiotic effects. All strains were first checked for antibiotic susceptibility and were sensitive or intermediately sensitive to ciprofloxacin at 5 μg/ml (see Table S1 in the supplemental material). Seventy-two-hour biofilms were grown in ASMDM+ and then treated with either 1× PBS (control), GSH alone, or GSH plus DNase I. After 8 h of incubation to allow biofilm disruption, ciprofloxacin (5 μg/ml) was added to a subset of control, GSH-treated, and GSH/DNase I-treated biofilms. After an additional 12 h of incubation, all cultures were evaluated for CFU per milliliter by serial dilution (Fig. 5).

Treatment with GSH alone, GSH plus DNase I, or ciprofloxacin alone did not significantly affect CFU values. However, when GSH-treated or GSH/DNase I-treated cultures were treated with 5 μg/ml ciprofloxacin, significant log reductions in CFU values, relative to control values, were recorded for all strains assayed (P < 0.0001 for all except for AES-1M [P < 0.001]). The reductions were conserved irrespective of strain, despite differing levels of phenazine production (AES-1R, AES-1M, LES431, LESB58, MID8916, and clone C) or production of phenazines other than pyocyanin (AES-2 and MANC8799) (Fig. 5, lower). GSH alone followed by 5 μg/ml ciprofloxacin resulted in an average log reduction across all strains of 2.29 ± 0.79, ranging from 1.63 (AES-1M) to 4.11 (MANC8799). Treatment with GSH plus DNase I followed by 5 μg/ml ciprofloxacin resulted in an average log reduction across all strains of 2.59 ± 0.90, ranging from 2.04 (AES-1R) to 4.79 (MANC8799). When GSH alone was compared with GSH plus DNase I, the addition of DNase I did not significantly enhance ciprofloxacin activity; the reasons for this remain unclear. It is possible that other elements of the EPS matrix, such as exopolysaccharides that colocalize with eDNA (71), may be preventing DNase I activity.

GSH disruption alone is sufficient to enhance antibiotic killing in a dose-dependent manner.

The concentration of GSH used here (2 mM) is only slightly higher than levels found in the epithelial lining fluid of healthy individuals (0.25 to 0.8 mM) (72). Furthermore, it was shown previously that a molar excess of GSH to pyocyanin could increase the rate of reaction between the two compounds (23). In order to determine whether the 2 mM GSH used was a rate-limiting factor in biofilm disruption, we repeated the pretreatment step with 5 mM and 10 mM GSH, followed by ciprofloxacin treatment. TTC was used to quantify respiration in 72-h biofilms grown in LB medium. Increasing the GSH concentration enhanced the antibiotic effects on bacterial survival, with complete inhibition of respiration at 10 mM GSH (Fig. 6). A check of phenazine production showed that the strains were highly heterogeneous regarding the types and quantities of phenazines. Clone C and AES-2 could not be assayed due to high levels of phenazine production (see Fig. S2 in the supplemental material), and the laboratory reference strains PAO1 and PA14 were substituted. These results and those of the previous experiment in ASMDM+ indicate that, given the structural similarity of phenazines, GSH is able to react structurally with more than one type. This is important, given the typical heterogeneity of phenotypes encountered in vivo (10, 11).

The effects of 2 mM GSH on biofilms partly explain the results of previous clinical trials of GSH in CF patients, in which inhaled GSH did not improve the median forced expiratory volume in 1 s (FEV₁) (73). Posttreatment levels of free GSH in the sputum of patients receiving GSH treatment reached an average of 20.4 pM, marginally higher than placebo-treated levels (17.6 pM). Furthermore, 53% of the cohort involved in that study had at least one sputum sample that was culture positive for P. aeruginosa in the previous year. Thus, compared to the 2 mM used in the present study, this small quantity of free GSH would have had a minimal impact on any P. aeruginosa biofilms present.

CONCLUSIONS

This study has demonstrated that GSH alone is sufficient to disrupt immature and mature biofilms of clinical P. aeruginosa. DNase I enhances this effect on both immature and mature biofilms. We also showed, for the first time, that pyocyanin enhances biofilm formation for a host-adapted phenotype of P. aeruginosa from a chronic infection, suggesting that the resultant biofilm is more representative of in vivo biofilms. RNA-seq analysis revealed a novel transcriptome for physiochemically disrupted biofilms, which is distinct from those of previously characterized mature and dispersed biofilms. GSH, alone or combined with DNase I, was shown to enhance the activity of ciprofloxacin in reducing CFU values. GSH alone had a dose-dependent relationship regarding biofilm disruption and, in turn, antibiotic effectiveness. These findings suggest that the activity of GSH results in a biofilm structure that is permeable to the surrounding environment. Furthermore, GSH disrupted the biofilms of a variety of phenotypes, irrespective of the type or quantity of phenazines produced.

This study indicates that GSH, alone or combined with DNase I, is a viable novel treatment for persistent P. aeruginosa biofilms, pending appropriate in vivo trials. The upregulation of pyoverdine biosynthesis in GSH-disrupted biofilms suggests that combining treatment with iron chelation may enhance the efficacy by preventing biofilm reformation (17, 19). GSH treatment combined with antibiotic treatment also would be appropriate for non-CF biofilms, such as those of burns, wounds, or postoperative infections or on abiotic surfaces such as medical prosthetic devices.