The reference strain P. aeruginosa PAO1 (DSM22644) is from the German Collection of Microorganisms and Cell Cultures. The bacterial strain was grown at 37°C in lysogeny broth (LB, also commonly called Luria Bertani medium), LB agar or LB soft agar overlays containing 0.6% (w/v) of agar. The three phages used in this work include phiIBB-PAA2 (Pires et al., 2014), vB_PaeM_CEB_DP1 (Pires et al., 2015a), and LUZ19 (Lavigne et al., 2013) (LUZ19 was kindly provided by Professor Rob Lavigne, KU Leuven, Belgium).

Phages were propagated using the double agar overlay technique (Azeredo et al., 2014). Briefly, 10 μL of isolated phage lysate was added to LB agar plates containing the host bacterial lawn and spread using a paper strip. The Petri dishes were then incubated at 37°C for 16–18 h. After incubation, 3–5 mL of saline magnesium buffer (SM buffer) (5.8 g/L NaCl, 2 g/L MgSO₄ 7H₂O, 50 mL/L 1 M Tris-HCl pH 7.5) were added to the lysate from a clear Petri dish and the plates were placed under agitation at 4°C for 18 h. Subsequently, the SM buffer with the eluted phages was collected, centrifuged (9,000 × g, 4°C, 10 min) and filtered (0.22 μm). Phage concentration was performed using established protocols described elsewhere (Azeredo et al., 2014). Briefly, 58.4 g/L of NaCl were added to the phage lysates and incubated at 4°C for 1 h under agitation (90 rpm). After incubation, this solution was centrifuged (9,000 × g, 4°C, 10 min) and the supernatant collected. Then, 100 g/L of PEG 8000 were mixed with the supernatant and incubated at 4°C for 16 h with agitation (90 rpm). The suspension was centrifuged again (9,000 × g, 4°C, 10 min), the supernatant discarded and the pellet resuspended in 5 mL of SM buffer per 50 mL of centrifuged sample. Chloroform was added in 1:4 (v/v) proportion and the suspension was centrifuged (3,500 × g, 4°C, 5 min). The aqueous phase (upper phase), which contained the phages was recovered, filtered (0.22 μm) and stored at 4°C until further use.

Plaque forming unit (PFU) counting was performed using double agar overlay technique (Kropinski et al., 2009). Serial dilutions of phage stock solutions in SM buffer were performed. Then, 100 μL of the diluted phage solution was mixed with 100 μL of the bacterial host and 3 mL of LB soft agar into a Petri dish already containing a layer of LB agar. The plates were incubated overnight at 37°C and the PFUs were counted.

Biofilms were formed on 96-well polystyrene microtiter plates (Orange Scientific) using established protocols (O’Toole and Kolter, 1998b; Friedman and Kolter, 2004) with minor modifications. P. aeruginosa PAO1 cultures were first grown for 16 h at 37°C and 120 rpm and then diluted 1:100 in LB. Each well was inoculated with 200 μL of this bacterial suspension and the microtiter plates were incubated in an orbital incubator for 24 h at 37°C and 120 rpm.

Biofilm infection experiments with phages phiIBB-PAA2, vB_PaeM_CEB_DP1 and LUZ19, individually or in three-phages cocktail, were performed using a multiplicity of infection (MOI) of 0.1 of each phage, according to a previously described protocol (Pires et al., 2013), with minor modifications. After biofilm formation for 24 h, the medium and planktonic bacteria were removed, and the wells were washed with fresh LB medium. Afterward, 200 μL of LB already containing the phage, the phage cocktail or the SM buffer (control) were added to each well. The plates were incubated at 37°C with agitation (120 rpm) and the colony forming units (CFU) were evaluated after 2, 4, 6, 8, 10, 12, 24, and 48 h of biofilm infection, as described below. For each time point three wells were surveyed.

The number of viable cells present in the biofilms were determined as previously described (Pires et al., 2013). Briefly, after the removal of growth medium, the wells were washed with saline [0.9% (w/v) NaCl]. Then, 200 μL of fresh saline was added to each well. After that, biofilms were scraped off and samples were collected and serially diluted in saline containing 5 mM of ferrous ammonium sulfate (FAS), to prevent further infection of bacteria by phage (McNerney et al., 2004; Swift et al., 2014; Alves et al., 2015). One drop (10 μL) was placed on a Petri dish and allowed to run down the plate, to obtain single colonies. CFU counts were carried out after overnight incubation at 37°C.

A P. aeruginosa PAO1 culture grown for 16 h was diluted 1:100 in a final volume of 25 mL and incubated for 24 h at 37°C and 120 rpm. After that time, cells were harvested by centrifugation (6,000 × g, 4°C, 5 min) and resuspended in fresh LB medium with phages or phage cocktail, at an MOI of 0.1. Control experiments were done using SM buffer instead of the phage. The suspensions were incubated at 37°C with agitation (120 rpm) and CFU counts were determined after 2, 4, 6, 8, 10, 12, 24, and 48 h, as for the biofilm infection experiments.

Nine single colonies from each time point were recovered in each independent experiment of biofilm and planktonic cell infection, and their susceptibility to the three phages was tested as described by Moons et al. (Supplementary Figure S1; Moons et al., 2013). The goal of this assay was to screen for bacterial resistance to phage infection, and select phage-resistant variants of P. aeruginosa, for posterior analysis. Phage susceptibility assays were also performed on the complemented strains that have restored galU function.

Swimming, swarming and twitching motilities were analyzed for ten P. aeruginosa PAO1 variants (isolated from 48-h biofilm infection with phage cocktail) using different media and protocols described elsewhere (Ha et al., 2014a,b; Turnbull and Whitchurch, 2014).

Five P. aeruginosa PAO1 variants that displayed resistance to the three phages used in this study were chosen for genomic DNA isolation and subsequent sequencing. The P. aeruginosa cultures were grown overnight at 37°C and 1.5 mL of these bacterial suspensions were centrifuged (9,000 × g, 4°C, 5 min). The pellets were resuspended in 500 μL of extraction buffer (80 mM Tris pH 8.5, 200 mM NaCl, 0.5% SDS, 5 mM EDTA, 1 mg/mL Proteinase K) and incubated overnight at 65°C. After incubation, 500 μL of phenol:chloroform:isoamyl alcohol (25:24:1, v/v) were added and the tubes were placed under gently agitation for 10 min at room temperature. The solution was then centrifuged (13,000 × g, 4°C, 10 min) and the supernatant was transferred into new tubes. One volume of isopropyl alcohol 100% (v/v) was mixed with the recovered supernatant and the solution was centrifuged again (13,000 × g, 4°C, 15 min). The supernatant was discarded and the pellet was washed with ethanol 75% (v/v). After drying, the pellets were resuspended in sterile water and stored at -20°C.

DNA sequencing libraries were produced from 1 μg of genomic DNA, following the recommendations of the TruSeq DNA protocol (Illumina). Genomic DNA was sheared by sonication using a Covaris S2 instrument. Sizes and concentrations of DNA sequencing libraries were determined on a Bioanalyzer 2100 (DNA1000 chips, Agilent). Paired-end sequencing (2 × 50 bp) was performed on one lane on a Hiseq1500 (Illumina) platform using TruSeq PE Cluster KIT v3 – cBot – HS and TruSeq SBS KIT v3 – HS. Cluster detection and base calling were performed using RTAv1.13 and quality of reads assessed with CASAVA v1.8.1 (Illumina). The sequencing resulted in at least 25 million pairs of 50-nt long reads for each sample, with a mean Phred quality score > 35. The raw genomic sequence data of the individual samples is available from the European Nucleotide Archive (ENA¹) under the accession number PRJEB15059.

The read pairs resulting from sequencing were mapped to the reference genome of P. aeruginosa PAO1 (RefSeq Accession No. NC_002516) using the Bowtie 2 aligner (Langmead and Salzberg, 2012), resulting in an overall alignment rate of at least 95% for each of the samples. Small sequence variations (SNPs and short insertions/deletions) were detected using SAMtools (Li, 2011) with a minimum read depth of five reads covering a putative variation site. Candidate sequence variations were inspected manually using Integrative Genomics Viewer (Thorvaldsdóttir et al., 2013) and compared with previously reported intra-strain variations in the P. aeruginosa PAO1 genome sequence (Dötsch et al., 2009; Klockgether et al., 2011) to identify and validate variations that are specific to the phage-resistant strains. Large deletions were detected by de novo genome assembly of the reads using IDBA-UD (Peng et al., 2012) in hybrid mode with the reference sequence of PAO1 as a template, and subsequently aligning the resulting contigs with the reference genome using Mauve (Darling et al., 2004).

The samples of bacterial LPS were prepared according to the protocol described by Hitchcock and Brown (H&B) (Hitchcock and Brown, 1983).

Acrylamide running gels at 12% were prepared according to a modified Laemmli procedure with resolving gels devoid of SDS (Laemmli, 1970). Three microliter samples of H&B LPS, except for the O6 B-band blot in which the sample used was only 1 μL, were loaded on the acrylamide gel and run at 120 V for 100 min. Blotting was performed for 60 min at 200 mA. Skimmed milk (5%) in PBS was used to block the nitrocellulose (NC) blots (Rocchetta and Lam, 1997). Primary monoclonal antibodies (mAb) specific against P. aeruginosa LPS were from cell lines of mouse hybridoma, and the supernatants of the culture of these were used undiluted as described previously (Emara et al., 1995) to incubate with the NC blots overnight at room temperature. The mAb used included MF15-4 (serotype O5 B-band specific), MF83-1 (serotype O6 OSA specific), N1F10 (CPA specific), 5c-101 (outer-core specific), and 5c-7-4 (inner-core specific). Secondary antibodies used were goat anti-mouse F(ab)₂-alkaline phosphatase conjugated and incubated at room temperature for 60 min. The blots were developed with the BCIP/NBT as per the manufacturer’s protocols (Sigma). The LPS banding patterns were visualized using the ultrafast silver nitrate-staining method that was described previously (Fomsgaard et al., 1990).

A single colony from a Tryptic Soy Agar plate was used to inoculate 5 mL of LB, and grown statically at 37°C for ∼68 h. Using a sterile inoculum loop a small amount (∼5 μL) of the pellicle that formed at the air-liquid interface was removed and resuspended in 50 μL of 0.1 M HEPES buffer pH 7.4. To disperse cellular clumps, the cell suspension was agitated/mixed by re-pipetting repetitively. Five microliter of the clarified suspensions were removed and applied to 200-mesh carbon-coated copper grids. The liquid was wicked away using filter paper (#1 Whatman) and stained with 5 μL of 2% (w/v) uranyl acetate (UA) for 30 s. The UA was wicked away and the grid was imaged using a single-tilt holder on an FEI Tecnai G2 F20 transmission electron microscope operating at an accelerating voltage of 200 kV and equipped with a bottom-mount Gatan 4k charge-coupled-device (CCD) camera.

The galU gene was amplified by PCR from P. aeruginosa PAO1 genomic DNA using the primers 5′ CATGTCGAAGAATTCATGATCAAGAAATGTCTTTTC and 5′ CTTCATCGGGAACGGAAGCTTTCAGTGAGCCTTGCC. The resulting PCR product was digested with the restriction enzymes EcoRI/HindIII and ligated into the vector pHERD26T. The resulting construct pHERD26T::galU was verified by sequencing.

Chemically competent P. aeruginosa PAO1 (wild-type and phage-resistant variants) were prepared using the rubidium chloride method (Green and Rogers, 2013) with mid-log (OD₆₀₀ 0.5-0.6) grown cells. Competent cells were stored at -80°C prior to use. Transformation of pHERD26T::galU into P. aeruginosa PAO1 occurred via heat shock at 42°C for 90 s followed by incubation in SOC media at 37°C with shaking at 200 rpm, for 1 h. Transformed cells were isolated on tryptic soy agar containing tetracycline at 100 μg/mL.

The statistical analysis of biofilm and planktonic cells infection experiments was done using two-way ANOVA with Bonferroni’s multiple comparisons test using GraphPad Prism 6. All tests were performed with a confidence level of 99%.

Phage infection of P. aeruginosa PAO1 biofilms and planktonic cultures was performed over the course of 48 h with phages phiIBB-PAA2, vB_PaeM_CEB_DP1 and LUZ19, either individually or in a cocktail composed by the three phages. Samples for determining the number of CFU were taken every 2 h during the first 12 h of phage infection. After that, two more samples were taken at 24 and 48 h post biofilm treatment. This was performed to help us better understand the dynamics of phage infection and bacterial resistance over the course of experiment. We hypothesized that at the later time points the population is mostly comprised of P. aeruginosa phage-resistant variants.

Overall, in the biofilm infection experiments (Figure 1A), phage LUZ19 was most effective among the phages tested in its ability to eliminate biofilm cells, followed by phages phiIBB-PAA2 and vB_PaeM_CEB_DP1. The latter two had similar levels of activity in depletion of P. aeruginosa biofilms. When the cocktail of the three phages was used, the results achieved during the first 8 h of biofilm infection were very similar with the ones obtained with phage LUZ19 alone. However, beyond 8 h, the phage cocktail was significantly more efficient (p < 0.01) than the other treatments with a particular phage in eliminating biofilm cells.

Although phage LUZ19 was able to significantly reduce the number of cells present in P. aeruginosa PAO1 biofilms, achieving a maximum reduction of ∼1.5 orders-of-magnitude after 6 h of treatment compared to the control (p < 0.01), a gradual increase in the number of biofilm cells was observed in subsequent time points (Figure 1A), which might indicate the fast proliferation of phage-resistant variants toward this phage. The use of the phage cocktail for biofilm treatment apparently resulted in two reduction phases: there is a reduction of ∼1.8 orders of magnitude within 4 h followed by a short period of regrowth, and then a maximum reduction of ∼2.1 orders-of-magnitude in the number of biofilm cells is achieved between 10 and 12 h of biofilm treatment versus the control (p < 0.01). This biphasic behavior might be a consequence of the early development of variants that get resistant to phage LUZ19, which has a shorter latent period and a higher burst size than the other two phages and consequently replicates faster in the first period of infection (Ceyssens et al., 2011; Pires et al., 2015a).

Although an increase in the number of biofilm cells was observed in the last time points, the number of viable cells present in biofilms 24 and 48 h after treatment with phage cocktail remained significantly low compared to the control (∼1.9 and ∼0.8 orders-of-magnitude reduction, respectively; p < 0.01) (Figure 1A).

The individual use of phages phiIBB-PAA2 and vB_PaeM_CEB_DP1 revealed to be the less effective treatments. The maximum reductions of biofilm cells achieved with those phages were ∼1.2 and ∼1.4 orders-of-magnitude after 12 and 24 h of biofilm infection, respectively (Figure 1A). Forty-eight hours post-infection of the biofilm, no statistic differences in the number of biofilm cells could be discerned when compared to the control.

In planktonic cells infection experiments, greater reductions in P. aeruginosa PAO1 cell counts were observed for all treatments when compared to that of the biofilm infection experiments (Figure 1B). Regarding planktonic cultures infection with phage LUZ19, a similar behavior as in biofilm infection experiments was observed: the maximum reduction of planktonic cells was achieved 6 h after phage treatment, showing a reduction of ∼3.2 order-of-magnitude vs. control (p < 0.01). Then, a gradual increase in the number of planktonic cells was observed in subsequent time points and no statistic differences were observed between the treatment with phage LUZ19 and the control after 24 and 48 h of infection (Figure 1B).

The application of a phage cocktail against planktonic cultures resulted in a fast reduction of cells in the first 6 h of infection followed by a slower reduction until a maximum reduction of ∼4.3 order-of-magnitude vs. control (p < 0.01) within 12 h. This reduction was followed by an increase in the number of planktonic cells and 48 h after treatment, no statistic differences comparatively with the control were detected (Figure 1B).

The individual treatment of planktonic cultures with phages phiIBB-PAA2 and vB_PaeM_CEB_DP1 resulted in a reduction of ca. 1.5 orders-of-magnitude vs. control (p < 0.01), which was obtained 24 h post-treatment. Although a slight increase in the number of planktonic cells was observed in the last time point for both treatments, the number of planktonic cells remained significantly lower comparatively with the control (p < 0.01) (Figure 1B).

To determine the amount of time required post-treatment for the number of phage-resistant variants to begin to increase, ten colonies were isolated from each independent experiment, at each time point of biofilm and planktonic cells infection experiments and the isolates were characterized, particularly regarding their susceptibility to the phages.

The percentages of phage-resistant variants isolated at each time point of the experiment were depicted in Figure 2. According to these results, the first phage-resistant variants were isolated from biofilm and planktonic cultures infected with phage LUZ19. P. aeruginosa PAO1 variants resistant to phage LUZ19 were isolated as early as 6 h after biofilm treatment with LUZ19 both individually (Figure 2A) and in cocktail (Figure 2B). In planktonic cultures, phage-resistant variants emerged within 4 h post-infection with LUZ19 (Figure 2C). However, when planktonic cultures were infected with phage cocktail, LUZ19-resistant variants were isolated 10 h post-treatment (Figure 2D). The emergence of resistant variants toward phages phiIBB-PAA2 and vB_PaeM_CEB_DP1 was significantly later compared to infection with phage LUZ19, i.e., between 12 and 24 h post-treatment of the biofilm (Figures 2A,B) and between 24 and 48 h post-infection of the planktonic cultures (Figures 2C,D). It noteworthy to point out that in both planktonic and biofilm infection experiments, the use of the cocktail of the three phages resulted in a lower percentage of phage-resistant variants than infection of the PAO1 cultures with individual phage.

Ten of the P. aeruginosa PAO1 variants, isolated after 48 h of biofilm treatment with the phage cocktail were selected for further characterization.

The susceptibility profiles of the ten P. aeruginosa variants to each phage are indicated in Table 1. P. aeruginosa PAO1 variants 1, 2, 8, and 9 showed resistance to phage LUZ19 but remained susceptible to phages phiIBB-PAA2 and vB-PaeM_CEB_DP1, whereas variants 3, 4, 5, 6, 7, and 10 were resistant to the three phages. These variants were streaked several times and their susceptibility profiles were maintained throughout the generations.

It was further observed that the colonies of variants 6, 7, and 10, all resistant to the three phages, had a brown color (data not shown), which is indicative of the production of a brown pigment, called pyomelanin, in the surrounding agar (Hosseinidoust et al., 2013b).

Overall, all ten of the selected P. aeruginosa PAO1 variants exhibited reduced twitching (p < 0.01) compared to the wild-type strain, which might indicate mutations in type IV pili of these variants. The swimming motility of all variant strains was also significantly inferior (p < 0.01) than the wild-type strain (Table 2). Concerning swarming motility, although all variant strains were statistically different from the wild-type strain, it increased in some cases and decreased in others. Therefore it was not possible to infer about the impact of phage resistance in swarming motility.

The genomes of five P. aeruginosa PAO1 variant strains, which displayed resistance to the three phages used in this study and defects in motility were sequenced and compared to the wild-type genome in order to identify genetic variations specific to the resistance phenotype (Figure 3 and Supplementary Table S1).

All variant strains were found to carry mutations in one of the pil genes involved in the synthesis of type IV pili: variants 3 and 4 carry a nonsense mutation in pilE gene, which encodes a pilin-like protein called “minor” pilin (Giltner et al., 2010; Kuchma et al., 2012); variant 6 carries a nonsense mutation in pilO gene, a gene required for pilin glycosylation (Castric, 1995); and variants 7 and 10 carry frameshift mutations in pilC and pilM, respectively, which encode a probable transmembrane protein PilC required to pilus biogenesis (Nunn et al., 1990; Li et al., 2013), and the cytoplasmic actin-like protein PilM that is likely anchored to the inner membrane by binding to the cytoplasmatic tail of PilN (Li et al., 2013; Tammam et al., 2013). Since there is an extended genomic diversity among P. aeruginosa PAO1 laboratory strains (Klockgether et al., 2010) and to confirm that these mutations were not a specification of the wild-type strain used in this study, the genes pilE, pilO, pilC, and pilM of wild-type strain were amplified and sequenced. The Sanger sequencing confirmed that all the mutations found in pil genes of variant strains were not already present in the wild-type strain, thus being acquired during the experiment (data not shown).

Pseudomonas aeruginosa variants 3 and 4 further revealed a point mutation in the galU gene, while in the other variants (6, 7, and 10), this gene was absent from the genome due to large deletions of 486, 317, and 214 kbp, respectively. Sanger sequencing of the wild-type galU confirmed that it is devoid of mutation when compared to that of the P. aeruginosa genome database (data not shown). Due to the size of these deletions, a large number of genes were lost from the genotypes of these strains (420, 266, and 179 genes, respectively, Supplementary Table S1).

Type IV pili are frequently described to play an important role in the initial stage of biofilm formation (O’Toole and Kolter, 1998a; Giltner et al., 2006, 2012; Barken et al., 2008), and thus the biofilm formation capability of P. aeruginosa PAO1 variant strains 3, 4, 6, 7, and 10 that were isolated from phage infection was analyzed. Nonetheless, the number of viable cells present in biofilms formed by those variants was similar to the numbers obtained with the parental strain and no statistical differences were observed (Supplementary Figure S2).

Differences in the phenotype of LPS between the P. aeruginosa PAO1 wild-type strain and the phage-resistant variant strains were apparent (Figure 4). Comparing to the SDS–PAGE LPS banding profile of the wild-type P. aeruginosa PAO1 control, a fast-migrating core oligosaccharide band is observed in the five phage-resistant variants, indicating that all of these variant bacteria synthesize a truncated LPS core. Consequently, neither Common Polysaccharide Antigen (CPA) nor O-Specific Antigen (OSA) polysaccharide could be assembled onto the core and all variant strains also demonstrated a lack the long chain of LPS bands. This was verified by SDS–PAGE and Western immunoblotting using mAb N1F10 (specific for CPA) and MF15-4 (specific for OSA).

Electron micrographs obtained by TEM of the wild-type and variant strains, which were grown under static conditions, showed the presence of pili on the bacterial cells of the wild-type strain, but none could be discerned among the P. aeruginosa PAO1-derived phage-resistant variants (Figure 5).

To confirm that the loss of galU function was responsible for the acquisition of phage resistance by the variant strains, complementation of the galU gene in these variants was performed by transforming the P. aeruginosa PAO1 wild-type control and phage-resistant variants with pHERD26T::galU in trans. The susceptibility of the complemented bacterial strains to the three phages was evaluated. As anticipated, the susceptibility of the P. aeruginosa variant strains to phages phiBB-PAA2 and vB_PaeM_CEB_DP1 was restored but they remained resistant to phage LUZ19 (Table 3). Although these variants were complemented with galU, they still carry mutations in one of the pil genes and phage LUZ19 is reported to use type IV pili as primary receptors.

The LPS-banding profiles of the complemented strains were analyzed by SDS–PAGE and silver staining and they were practically identical to that of LPS prepared from the wild-type PAO1 strain (Supplementary Figure S3)