Strains used are listed in Table 1. For all pyoverdine production and biofilm formation assays, bacteria were seeded in M9 media (M9 salts (1% w/v) and casamino acids (1% w/v), supplemented with 1 mM MgSO₄ and 1 mM CaCl₂) in static six-well plates (Greiner, North Carolina) at 30°C. SK media was composed of 0.35% (w/v) Bacto-Peptone and 0.3% (w/v) NaCl, supplemented with 1 mM MgSO₄ and 1 mM CaCl₂ (Conery et al., 2014).

PA14 transposon mutants were inoculated into 96-well plates with LB media containing 15 μg/mL gentamicin. Inoculated plates were incubated overnight at 35°C. 10 μL of LB culture from each well were transferred into 96-well, clear, flat-bottom plates (Greiner, North Carolina) containing 90 μL of M9 media per well. Bacterial plates were grown at room temperature for 24 h. Pyoverdine production (Ex 405 nm, Em 460 nm) and bacterial growth (O.D.₆₀₀) were measured every hour in these plates using a Cytation5 (BioTek, Vermont) multimode plate reader.

Bacterial strains were grown in LB media with appropriate antibiotics overnight with shaking. Two milliliters M9 media were dispensed into each well in six-well plates (Greiner, North Carolina) and inoculated with 100 μL from overnight LB cultures. The plate was incubated at 30°C inside a plate reader for 24 h with pyoverdine fluorescence measurements and bacterial growth absorbance measurements made every 30 min. Each experiment consisted of at least three biological replicates.

This procedure was adapted from (Merritt et al., 2005). In brief, bacterial strains were grown under conditions identical to those described above. After incubation at 30°C for 24 h, bacterial cultures were aspirated and the biofilm matrix on the bottom of the plate was stained with 2 mL of 0.1% (w/v) crystal violet in 20% (v/v) ethanol/water for 30 min. The stain was removed and excess stain was washed with two consecutive rinses of PBS (Gibco, Maryland). Plates were dried and then photographs were taken. For biofilm quantification, the remaining crystal violet was solubilized in 30% acetic acid and absorbance was measured at 550 nm. Each experiment consisted of at least three biological replicates. Statistical significance was determined using Student's t-test.

After 8 h growth in six-well plates, planktonic cells were collected from 1.5 mL of supernatant. RNA was extracted and purified using Trizol reagent (Invitrogen, California) according to manufacturer's protocols with minor adjustments. To ensure cell lysis, cells resuspended in Trizol reagent were heated at 95°C for 15 min prior to phase separation. Purified RNA was treated with DNase I (Thermo Scientific, Massachusetts). Reverse transcription was performed using random decamers and Retroscript kit (Ambion). qRT-PCR was conducted using SYBR green PerfeCTa SYBR Green Fastmix (Quantabio, Massachusetts) in a CFX-96 real-time thermocycler (Bio-Rad, California). Fold-changes were calculated using a ΔΔCt method, and compared to expression from wild-type P. aeruginosa. Primer sequences are available upon request.

Two milliliters of M9 media were dispensed into three wells of a six-well plate (Greiner, North Carolina) and inoculated with 100 μL of PA14 dsRed/PpvdA-gfp LB overnight culture. After incubation at 30°C for 16 h, supernatant was carefully collected from each well. Planktonic cells were collected from 1 mL of media. Cells in biofilm were collected by scraping all three wells into 1 mL of PBS buffer using a cell scraper. The two cells samples were extensively washed and resuspended in 1 mL of PBS. GFP and dsRed fluorescence from 100 μL of each sample was measured in a black 96-well plate (Greiner, North Carolina). Each experiment consisted of at least three biological replicates. Statistical significance was determined using Student's t-test.

Two milliliters of M9 media were dispensed into each of three wells in a six-well plate (Greiner, North Carolina) and inoculated with 100 μL of P. aeruginosa grown overnight in LB. After 8 or 16 h incubation at 30°C, supernatant was carefully collected from all wells. Planktonic cells were collected from 2 mL of media. Planktonic cells and biofilm-associated cells attached to the bottom of the plate were washed twice with PBS. Planktonic cells were resuspended in 0.5 mL PBS buffer and dispensed into a six-well plate. Biofilms and concentrated planktonic cells were imaged using a custom filter (445/45 excitation, 510/42 emission, 482 nm dichroic) using a Cytation5 multimode reader (Biotek, Vermont). All images were taken under identical conditions, and each experiment consisted of at least three biological replicates.

The procedure was adapted from (Kirienko et al., 2014). In brief, glp-4(bn2ts) worms were grown at 25°C on E. coli OP50 until they reached young adulthood. Worms were washed off plates, rinsed with S Basal, and resuspended in M9 media (500 worms/mL). 1,000 worms were dispensed into each well of a six-well plate and inoculated with 100 μL of overnight P. aeruginosa PA14 culture grown in LB. Plates were covered with air-permeable membranes and incubated at 25°C for 30 h. After exposure to P. aeruginosa, worms were collected from the wells and extensively washed in S Basal buffer. Worms were pelleted by gravity between washes to effectively remove most P. aeruginosa bacteria from the media. After washes, 30–40 worms were dispensed into each well of black half-area 96-well clear bottom plates (Greiner, North Carolina). Worms were stained with Sytox Orange nucleic acid stain (Thermo Fisher) for 12 h at room temperature. Excess stain was removed by washing twice in S Basal, and plates were imaged under bright field and RFP channels using a Cytation5 multimode plate reader (BioTek, Vermont). Relative host death was quantified by normalizing RFP/bright field intensity in worms. Killing of C. elegans by biofilm mutants was normalized to that of wild-type P. aeruginosa. Each experiment consisted of at least three biological replicates. Each biological replicate contained 12 wells (~400 worms). Statistical significance was determined using Student's t-test.

glp-4(bn2ts) worms were grown at 25°C on E. coli OP50 until they reached young adulthood. Worms were washed off plates, rinsed with S Basal, and resuspended in M9 media (3,000 worms/mL). 6,000 worms were dispensed into each well of a six-well plate and inoculated with 100 μL of P. aeruginosa PA14 grown overnight in LB. Three wells were used for each condition for each biological replicate. Plates were covered with air-permeable membranes and incubated at 25°C for 36 h. Worms were collected and extensively washed in S Basal buffer. To facilitate lysis, worms were frozen at –80°C, thawed, and sonicated. To quantify iron (III) concentration, 50 μL of pyoverdine-rich bacterial filtrate of known fluorescence was added to 200 μL of lysate. After incubation, pyoverdine fluorescence was measured with a Cytation5 (Biotek, Vermont), and was compared to the same filtrate diluted with media. Since pyoverdine binds to ferric iron in a 1:1 stoichiometric ratio, the reduction of pyoverdine fluorescence reflects the ferric iron content in the worm lysate. Iron concentrations in worms exposed to biofilm mutants and to wild-type P. aeruginosa PA14 were compared. Each experiment consisted of at least three biological replicates. Statistical significance was determined using Student's t-test.

Twenty-five microliters of P. aeruginiosa PA14 grown overnight in LB was inoculated into 475 μL of M9 media in 17 mm diameter glass or plastic culture tubes. Tubes were grown in 30°C under static growth conditions for 16 h. Bacterial supernatant was collected from the tubes, and pyoverdine fluorescence was determined. Biofilms in tubes were stained and crystal violet staining was measured as described above. Each experiment consisted of at least three biological replicates. Statistical significance was determined using Student's t-test.

Using a non-redundant P. aeruginosa PA14 transposon mutant library (Liberati et al., 2006), we conducted a high-throughput, kinetic screen of 5,810 mutants to measure pyoverdine production. The high-throughput screen was performed under static growth conditions in 96-well plates, which resulted in robust biofilm formation in most wells, supporting the potential for a relationship between the generation of biofilm and production of pyoverdine. Fluorescence spectrophotometry (Ex 405 nm, Em 460 nm) was used to monitor pyoverdine biosynthesis in 96-well plates over a 24 h span. Of the mutants screened, 485 showed severe impairments in pyoverdine fluorescence (as defined as production below 30% of wild-type levels). Eliminating targets with growth defects (i.e., mutants where reduced pyoverdine biosynthesis might be caused by poor growth) left 338 hits. These were narrowed further to include only 296 hits that had predicted gene function (Figures 1A,B). 55 of these (18.6%) were associated with various stages of biofilm formation (Figure 1C), including flagellin biosynthesis, chemotaxis, type IV pili assembly, Cup fimbriae biogenesis, and exopolysaccharide synthesis (Table S1).

To verify our findings, we selected genes representative for each stage of biofilm formation (initial surface attachment: flgK, a flagellum biosynthesis gene, and motAB, required for chemotaxis; adhesion factor production: pilY1, a type IV pili mutant; and bioflm maturation: pelA, an exopolysacharide mutant) (Vogeleer et al., 2014). As expected, these strains exhibited impairments in both biofilm formation (Figure 2A) and pyoverdine production (Figure 2B). We also observed a strong linear correlation between the amount of biofilm and pyoverdine produced by these mutants (Figure 2C), indicating a role for biofilm formation in pyoverdine production.

Quantitative real-time PCR (qRT-PCR) using RNA harvested from WT P. aeruginosa PA14, ΔpilY1, and ΔflgK planktonic cells after 8 h growth corroborated this finding. We used qRT-PCR to analyze expression of the alternative sigma factor pvdS and pvdS-dependent pyoverdine biosynthesis genes [including pvdA and pvdF (which are responsible for generating the non-standard amino acid N-formyl-N-hydroxyornithine), pvdE (predicted to transport nascent pyoverdine into the periplasm), pvdP and pvdQ (involved in periplasmic maturation)] (Visca et al., 2007; Drake and Gulick, 2011). All genes were significantly downregulated in ΔflgK mutant (Figure 2D). As may be hypothesized by the weaker pyoverdine biosynthesis phenotype (Figure 2B), the ΔpilY1 mutant showed significantly less impact, although several genes were still disrupted, including pvdF, pvdP, and pvdQ. The PvdRT-OpmQ system is thought to be responsible for recycling pyoverdine back out of the periplasmic space (Imperi et al., 2009). Unsurprisingly (pvdT is also transcriptionally regulated by pvdS), pvdT expression was diminished in the ΔflgK mutant, although regulation of opmQ, which is in the same operon as pvdR and pvdT) was unaffected. Our qRT-PCR data also suggest a possible role for ferripyoverdine uptake in biofilm-mediated regulation of pyoverdine production. fpvA, the ferripyoverdine receptor (Shen et al., 2002), and tonB1, which provides the energy for ferripyoverdine translocation into the cell (Shirley and Lamont, 2009), were significantly downregulated in the ΔflgK mutant (Figure 2D).

To determine whether the regulatory relationship between pyoverdine and biofilm is bidirectional (i.e., whether pyoverdine production was necessary for biofilm formation), we assayed biofilm development in two mutants (PA14ΔpvdA and PA14pvdE). Pyoverdine production was abolished in PA14ΔpvdA (Figure 2B), while PA14pvdE produced a small amount of pyoverdine (Figure 2B). Both mutants showed nearly normal biofilm formation (Figure 2A). This is consistent with previous results from Banin and colleagues, who showed that active iron acquisition was necessary for biofilm formation, but that pyoverdine itself is not (Banin et al., 2005).

P. aeruginosa strains exhibit tremendous variation in the regulation of virulence genes (Lee et al., 2006). Therefore, we tested whether biofilm formation was necessary for pyoverdine production in PAO1, the reference strain for P. aeruginosa. PAO1 mutants with transposons inserted into analogous genes involved in biofilm formation (fliF and motA to disrupt attachment, pilG to disrupt adhesion, and pelB to disrupt biofilm maturation) generally exhibited compromised pyoverdine production (Figures S1A,B). However, unlike PA14ΔpelA mutant, the PAO1pelB strain was capable of forming dense biofilms comparable to wild-type bacteria (Figure S1A), resulting in wild-type levels of pyoverdine production (Figure S1B). This is likely due to the fact that PAO1 produces two different types of exopolysaccharides, Pel and Psl, and the disruption in Pel alone is insufficient to effectively decrease biofilm formation and pyoverdine biosynthesis. This is consistent with our observations that PAO1 appears to create more biofilm than PA14 (Figure S2A). Interestingly, PAO1 also secretes more pyoverdine and at a faster rate than PA14 (Figure S2B). However, in genes that most disrupt biofilm formation (flgK for PA14 and motAB for PAO1), pyoverdine production was attenuated to similar levels in the two strains (Figure 2B, Figure S1B).

Based on these observations, we hypothesized that sessile and planktonic cells would exhibit differences in pyoverdine biosynthetic machinery. We used a GFP-based reporter to assay expression of the pvdA gene, which was normalized to constitutively-expressed dsRed. To our surprise, pvdA transcription was indistinguishable between planktonic and biofilm-associated cell populations (Figure 2E). However, we observed significant differences in pyoverdine production between sessile and planktonic cells in earlier stages of bacterial growth when pyoverdine biosynthesis is initiated. Using fluorescence microscopy, we visualized intracellular pyoverdine levels in sessile cells aggregated in the biofilm matrix and planktonic cells collected from the growth media after 8 h (when pyoverdine production is initiated), and 16 h of growth (Figure 2F). At 8 h, pyoverdine fluorescence is detected only in cells in the biofilm matrix, but at 16 h, pyoverdine fluorescence is present in both biofilm matrix and concentrated planktonic cells (Figure 2F). In a ΔpvdA mutant, neither biofilm matrix cells nor planktonic cells exhibit fluorescence, verifying that the fluorescence detected is from pyoverdine (Figure S3A). These results suggest one possible model for biofilm-dependent regulation of pyoverdine. First, pyoverdine production is initiated in the microcolonies that will nucleate biofilm formation. Secreted pyoverdine binds to ferric iron, returning to the pathogen via the ferripyoverdine receptor protein FpvA (Shen et al., 2002). This increases the activity of PvdS, upregulating pyoverdine biosynthesis genes in both sessile and planktonic cells (Lamont et al., 2002; Llamas et al., 2014). Attenuated pyoverdine fluorescence in the extracellular matrices of biofilm mutants (Figure S3B) and the downregulation of ferripyoverdine uptake genes in ΔpilY1 and ΔflgK planktonic cells (Figure 2D) are consistent with this model.

Pyoverdine is mainly involved in iron transport, and its exceptionally high affinity for ferric iron grants the pathogen a significant advantage over the host in the competition for iron acquisition. We recently reported that exposure to pyoverdine significantly disrupts host iron homeostasis, inducing a lethal hypoxic response in C. elegans (Kirienko et al., 2013). Based on these findings, we predicted that the attenuation of pyoverdine production in P. aeruginosa biofilm mutants would limit their virulence and their ability to remove iron from the host. We used a fluorometric assay to determine relative ferric iron concentration in C. elegans (see Materials and Methods for details). As expected, C. elegans exposed to pelA, flgK, or pilY1 mutants exhibited significantly greater ferric iron retention than pyoverdine-competent controls, similar to worms exposed to PA14pvdE (Figure 3A).

To determine whether this was sufficient to alleviate pyoverdine-mediated pathology, we exposed worms to wild-type or biofilm mutants of P. aeruginosa for 30 h, washed them extensively to remove the pathogen, and then stained them with Sytox Orange, a cell-impermeant dye that stains only dead worms. C. elegans exposed to biofilm mutants displayed significant reductions in mortality compared to wild-type P. aeruginosa (Figures 3B,C). Interestingly, biofilm mutations compromised virulence to a greater extent even than the pvdE mutant (Figures 3B,C). This suggests that biofilm formation may contribute to virulence through multiple means.

To identify the effects of biofilm formation in a pathogenesis model that is not dependent upon pyoverdine, we tested whether biofilm mutants showed attenuation in an agar-based, intestinal colonization assay known as Slow Killing (Kirienko et al., 2013). With the exception of PA14 ΔpilY1, the biofilm mutants showed wild-type levels of virulence (Figure 3D). These data were also corroborated by screening results from Feinbaum and colleagues (Feinbaum et al., 2012). A majority of biofilm mutants (again, with the exception of mutants that compromised formation of the type IV pili) had no effect in P. aeruginosa's ability to kill C. elegans.

Cyclic diguanylate monophosphate (c-di-GMP) is a crucial secondary messenger in P. aeruginosa that transcriptionally regulates a wide variety of virulence factors, including type III and type VI secretion (Moscoso et al., 2011). Previous research has also linked intracellular c-di-GMP concentration to biofilm formation. For example, diguanylate cyclases (i.e., SadC) are involved in c-di-GMP synthesis and support increased formation of biofilms (Kuchma et al., 2007; Merritt et al., 2007), while phosphodiesterases (i.e., BifA) hydrolyze c-di-GMP and limit biofilm formation (Kuchma et al., 2007; Merritt et al., 2007). Under our screening conditions, PA14ΔsadC (low c-di-GMP) and PA14ΔbifA (high c-di-GMP) mutants exhibited less and more biofilm respectively (Figures 4A,B), which is consistent with previously published data (Kuchma et al., 2007; Merritt et al., 2007). As predicted, pyoverdine production was attenuated in ΔsadC and enhanced in ΔbifA mutants (Figure 4C).

To test the regulatory relationship between c-di-GMP concentration and biofilm formation in pyoverdine production, we repeated the kinetics experiment with a PA14ΔbifAΔpelA double mutant that exhibits high intracellular c-di-GMP concentrations (Kuchma et al., 2007), but poor biofilm formation (Figures 4A,B). Since pyoverdine production was not completely abolished in the PA14ΔpelA mutant background, the ramifications of bifA deletion on pyoverdine production (whether augmented or decreased) can be observed. Pyoverdine production in the PA14ΔbifAΔpelA double mutant was similar, if not lower, to ΔpelA single mutant (Figure 4C), indicating that regulation of pyoverdine via intracellular c-di-GMP concentrations is hypostatic to biofilm formation. This confirms that c-di-GMP's effect on pyoverdine production is mediated indirectly, via biofilm formation.

We hypothesized that chemical inhibition of biofilm formation, like genetic disruption, would reduce pyoverdine production. Several compounds are known to compromise P. aeruginosa biofilm formation, including the aromatic heterocycle 2-amino-5,6-dimethylbenzimidazole (2-ABI) (Frei et al., 2012). Addition of 25 μM 2-ABI to P. aeruginosa culture reduced biofilm formation by 40% compared to the solvent control (Figure 5A), with a concomitant decrease (60%) in pyoverdine production (Figure 5B). The decrease in these exoproducts was not a consequence of bacterial growth inhibition; bacterial titer was unaffected by the presence of the compound (Figure 5C). As expected, disruption of biofilm formation and pyoverdine production by the addition of 2-ABI were sufficient to rescue C. elegans (Figures 5D,E), offering an exciting possibility of utilizing this molecule as an anti-virulent.

Another way to significantly alter biofilm formation is by changing the surface characteristics of the vessel for P. aeruginosa static cultures. In this case many of the off target effects that may be induced by chemical inhibitors or genetic manipulations will be eliminated. On a plastic surface, P. aeruginosa PA14 formed approximately one-half of the biofilm as a glass surface (Figure S4A). This led to ~25% decrease in pyoverdine production (Figure S4B).

Arguably, the most important cellular function for pyoverdine is iron acquisition. Since biofilm formation is necessary for pyoverdine production, we hypothesized that biofilm mutants would be compromised under conditions where pyoverdine is required for growth. To test this hypothesis, we assessed their growth alongside PA14pvdE mutant in the presence of 1,10-phenanthroline, a ferrous iron chelator (Figure 6A). When grown in media supplemented to a final concentration of 250 μM 1,10-phenanthroline, growth was compromised in all strains, but it was most severely curtailed in pvdE mutant. To our surprise, neither growth nor pyoverdine production in biofilm mutants in low iron conditions were significantly different from wild-type bacteria (Figures 6A,B). This was not due to improvement of biofilm formation; mutants still showed significantly reduced biofilm under these conditions (Figures 6C,D). This observation is consistent with findings by Hunter et al., who reported that inhibiting ferrous iron acquisition in P. aeruginosa reduces biofilm formation (Hunter et al., 2013). The recovery of pyoverdine production in biofilm mutants treated with 1,10-phenanthroline demonstrates that biofilm formation and pyoverdine biosynthesis can be unlinked. As predicted, 1,10-phenanthroline also abolished the protection provided by biofilm gene mutation in the Liquid Killing assay (Figures 6E,F). In contrast, virulence in the pvdE mutant remained attenuated, even in the presence of 1,10-phenanthroline. This demonstrated that adding 1,10-phenanthroline does not abolish the importance of pyoverdine in virulence.

We observed similar results for biofilm mutants grown in the presence of the ferric iron chelator ethylenediamine-N,N′-bis(2-hydroxyphenylacetic acid (EDDHA). Addition of 1 mg/mL EDDHA significantly reduced bacterial growth and biofilm formation (Figures S5A,C). As a result, neither growth nor pyoverdine production was attenuated in biofilm mutants, but it was severely curtailed in the pvdE mutant (Figures S5A,B). Observations from biofilm mutants treated with 1,10-phenanthroline or EDDHA suggest that under normal conditions, biofilm production is necessary for robust pyoverdine biosynthesis, but when iron concentrations become strongly limited, pyoverdine production becomes independent of biofilm genes.

In addition to iron starvation, nutrient availability also contributes to pyoverdine regulation. Previously, we reported that biofilm mutants showed no attenuation of virulence in C. elegans Liquid Killing (Kirienko et al., 2013). The media used for those studies is comprised of low concentrations of sodium chloride and peptone, diluted into buffered inorganic salts to match the osmolarity of the host. This contrasts with the M9 media used in this study, which is better defined and more nutritionally rich. Kinetic measurements of biofilm mutants grown in SK media showed no significant decrease in pyoverdine fluorescence (Figure 7A). In contrast to the M9 media used in this study, SK media does not support dense bacterial growth nor formation of dense biofilms (Figures 7B,C). Much like low-iron conditions, this alleviates pyoverdine biosynthesis from biofilm regulation.

Likewise, pyoverdine production in c-di-GMP biosynthesis mutants (PA14ΔsadC, PA14ΔbifA) in the presence of 1,10-phenanthroline (Figure 6G) or SK media (Figure 7D) also did not significantly differ from that of wild-type bacteria. This further supports our model where c-di-GMP affects pyoverdine production indirectly via biofilm formation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.