Iron and Pseudomonas Aeruginosa Biofilm Formation

Iron and Pseudomonas aeruginosa biofilm formation

Ehud Banin*, Michael L. Vasil†, and E. Peter Greenberg*‡

*Department of Microbiology, School of Medicine, University of Washington, Seattle, WA 98195-7242; and †Department of Microbiology, University of Colorado Health Sciences Center, Aurora, CO 80045

Contributed by E. Peter Greenberg, May 25, 2005 Iron serves as a signal in Pseudomonas aeruginosa biofilm devel- which results in transmission of a signal to a membrane-spanning opment. We examined the influence of mutations in known and anti-␴ factor that governs the activity of PvdS, thereby control- putative iron acquisition-signaling genes on biofilm morphology. ling expression of pyoverdine synthesis. The signal also regulates In iron-sufficient medium, mutants that cannot obtain iron through expression of other genes, including extracellular virulence the high-affinity pyoverdine iron acquisition system form thin factors such as exotoxin A (9, 11, 12). There are at least 10 other biofilms similar to those formed by the parent under low iron gene clusters in the P. aeruginosa genome that may encode conditions. If an iron source for a different iron acquisition system iron-responsive ECF ␴ factor-regulated systems (12). For exam- is provided to a pyoverdine mutant, normal biofilm development ple, there is one gene cluster encoding proteins with extensive occurs. This enabled us to identify iron uptake gene clusters that sequence similarity to the FecIR, ferric citrate system in Esch- likely serve in transport of ferric citrate and ferrioxamine. We erichia coli (12, 13). suggest that the functional iron signal for P. aeruginosa biofilm As in many other bacteria, P. aeruginosa has a Fur protein that development is active transport of chelated iron or the level of functions as a global regulator of iron-responsive genes (14, 15). internal iron. If the signal is internal iron levels, then a factor likely Fur has both negative and positive regulatory effects on gene to be involved in iron signaling is the cytoplasmic ferric uptake expression. It represses gene expression by direct binding to the regulator protein, Fur, which controls expression of iron-respon- operators of iron starvation-inducible genes. It activates gene sive genes. In support of a Fur involvement, we found that with expression indirectly through control of a pair of small regula- low iron a Fur mutant was able to organize into more mature tory RNAs (sRNAs), PrrF1 and PrrF2 (16). In P. aeruginosa, fur biofilms than was the parent. The two known Fur-controlled appears to be an essential gene, and the only fur mutants small regulatory RNAs (PrrF1 and F2) do not appear to mediate available either produce reduced levels of wild-type Fur or have iron control of biofilm development. This information establishes missense mutations that exhibit high reversion rates (17, 18). a mechanistic basis for iron control of P. aeruginosa biofilm We sought to better understand the iron-signaling cascade formation. critical for normal biofilm development. We asked whether any of the known or putative iron-uptake and iron-starvation ECF pyoverdine ␴ factors are involved in regulating biofilm development. Our data indicate that any iron-uptake system that can provide iofilms consist of groups of bacteria attached to surfaces and sufficient levels of internal iron to P. aeruginosa can function in Bencased in a hydrated polymeric matrix. Pseudomonas the iron-signaling pathway. This finding suggests that a critical aeruginosa biofilms cause persistent infections in individuals with level of intracellular iron serves as the signal for biofilm devel- underlying health problems. For example, most people with the opment. Our data also indicate that intracellular iron signaling genetic disease cystic fibrosis are plagued by chronic P. aerugi- is mediated by Fur but not by means of its regulation of PrrF1 nosa biofilm infections of their lungs (reviewed in refs. 1–3). Iron and PrrF2. starvation can prevent bacterial growth. Recent work shows that with sufficient iron for growth the levels of this metal serve as Materials and Methods a signal for biofilm development (4, 5). By sequestering iron, Bacterial Strains, Plasmids, and Culture Conditions. We used P. subgrowth inhibitory concentrations of the mammalian iron aeruginosa PAO1 as the wild-type strain. Strain PAO1 and most chelator lactoferrin block the ability of P. aeruginosa biofilms to of the P. aeruginosa transposon insertion mutants used were mature from thin layers of cells attached to a surface into large from the Comprehensive P. aeruginosa transposon mutant li- multicellular biofilm structures (4). The influence of lactoferrin brary at the University of Washington Genome Center (19). We on biofilm development is thought to be related to the fact that used the following mutants with transposons in iron starvation at low iron concentrations P. aeruginosa exhibits incessant ECF ␴-factor homologs: PTL4103 (PA0149), PTL52421 twitching motility on surfaces, and cells do not form sessile (PA0472), PTL9556 (PA0675), PTL16978 (PA1300), PTL37473 structures (4, 5). The mechanism for iron control of twitching (PA1363), PTL16034 (PA1912), PTL18118 (PA2050), motility is unknown, as is the mechanism of iron signaling in PTL51829 (PA3410), PTL7851 (PA3899), and PTL49853 biofilm development. (PA4896). We also used mutants with sequence similarity to Iron is essential for, yet toxic to, bacteria. For most pathogens, various iron receptors: PTL40529 and PTL14626 (both including P. aeruginosa, there is intense competition for iron with PA0151); PTL20210 (PA0470) with similarity to a hydroxamate the host (6–8). P. aeruginosa has multiple systems to sense and type receptor; PTL45373 (PA0674), a ferripyoverdine receptor sequester iron in its environment, and it is able to regulate homolog; PTL18189 (hxuC), a heme receptor homolog; cellular iron acquisition and storage through an assortment of PTL5195 (PA1365), a ferribactin receptor homolog; PTL4009 positive and negative regulatory factors (reviewed in refs. 6, 9, (PA1910); PTL11509 (PA2466), a ferrioxamine receptor ho- and 10). The two best-studied P. aeruginosa iron acquisition molog; PTL9628 (hasR), a heme uptake homolog; and systems are the high-affinity pyoverdine system and the lower- PTL14263 (fecA). The pyochelin mutant (pchA) was PTL5032. affinity pyochelin system. Pyoverdine and pyochelin bind extra- All of the mutants contained either a ISlacZ͞hah or an ISphoA͞ cellular iron (Fe3ϩ), which is then transported into the cell together with these siderophores. Pyoverdine synthesis and secretion are regulated by means of the extracytoplasmic func- Abbreviations: ECF, extracytoplasmic function; Tc, tetracycline; Fur, ferric uptake regulator; tion (ECF) ␴ factor PvdS. Expression of PvdS is regulated by iron TSB, Bacto Tryptic Soy Broth; sRNA, small regulatory RNA. and the ferric uptake regulator (Fur). In this acquisition system, ‡To whom correspondence should be addressed. E-mail: [email protected]. pyoverdine binds to an outer membrane-associated receptor, © 2005 by The National Academy of Sciences of the USA

11076–11081 ͉ PNAS ͉ August 2, 2005 ͉ vol. 102 ͉ no. 31 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0504266102 Downloaded by guest on September 26, 2021 hah transposon insertion. Further information on these mutants Pseudomonas isolation agar. A TcS strain was isolated, and the is at www.genome.washington.edu͞UWGC͞Pseudomonas͞ 63-bp insertion was confirmed by PCR. This strain was trans- index.cfm. The locations of all of the mutations were confirmed formed with PTL11509 genomic DNA, a transformant was by PCR with primers flanking the insertion sites. We also used selected, and all of the mutations were confirmed by PCR. PAO1⌬pvdS (20); PAO1⌬pvdA (14); PAO1⌬fpvA (11), a strain To construct the pvdS-gfp fusion strain, a 250-bp EcoRI- with a fur point mutation; PAO1furC6Tc (17, 18), a HindIII fragment containing the pvdS promoter was amplified PAO1⌬prrF1-F2 mutant (16); PAO1⌬pvdD ⌬pchEF (provided (from bp 2721925 to bp 2722175) and cloned into an EcoRI- by Urs Ochsner, Replidyne Inc., Louisville, CO); and several HindIII-digested promoterless gfp pPROBE AT (26). The plas- strains with mutations generated as described below. mid was transferred to PAO1⌬pvdD ⌬pchEF by conjugation. For growth curve and flow cell biofilm experiments we used The resulting strain, EB125, contains the pvdS-gfp fusion as a 1% Bacto Tryptic Soy Broth (TSB) (Becton Dickinson). Incu- single chromosomal copy. bation was at room temperature (23–25°C). Where indicated, 20 For complementation of the pvdA mutation, a 2-kb fragment ␮g͞ml human lactoferrin (Sigma) was added to the medium. containing pvdA was amplified (from bp 2638500 to bp 2640501) Also as indicated, 50 or 100 ␮M ferric chloride, 1 ␮M ferric and cloned in EcoRI-digested pJN105 (27). The resulting plas- dicitrate, 1.5 ␮M desferrioxamine mesylate (Sigma), or pyover- mid pEB4 was used to transform PAO1⌬pvdA. For complemen- dine-conditioned medium was added. Pyoverdine-conditioned tation of the pvdS mutation we used pPVD27 (28). medium was prepared by growing P. aeruginosa PAO1 in Kings B medium (21) for 12 h at 37°C. The cells were removed by Biofilm Experiments. We used flow cells to follow biofilm growth centrifugation and filtration through a 0.2-␮m filter. Pyoverdine- on a glass surface (24). The flow cells were inoculated with a 1:50 conditioned medium was added at 1% as indicated. As a control dilution (in 1% TSB) of a P. aeruginosa stationary phase culture, for the pyoverdine experiments, we used medium conditioned by and flow was initiated after 1 h. To image biofilms, we used the PvdA mutant PAO1⌬pvdA. Escherichia coli was used for confocal scanning laser microscopy (CSLM). The CSLM was a recombinant manipulations and was grown in Luria-Bertani Radiance 2100 system (Bio-Rad) with a Nikon Eclipse E600 (LB) broth. We used Pseudomonas Isolation Agar (Becton microscope. Generally, we imaged GFP in P. aeruginosa con- Dickinson) for selection of transconjugants. Antibiotics were taining pMRP9-1, which has a constitutively expressed gfp (29). used at the following concentrations: 100 ␮g͞ml gentamicin Where indicated, we counterstained the biofilm with propidium (Gm) for E. coli and P. aeruginosa,10␮g͞ml tetracycline (Tc) for iodide (4 ␮M) and imaged the propidium iodide fluorescence by

E. coli, and 100 ␮g͞ml Tc for P. aeruginosa. CSLM (30). The image acquisition software was LASERSHARP MICROBIOLOGY 2000 (Bio-Rad). Images were processed with CONFOCAL ASSIS- Construction of Plasmids and P. aeruginosa Mutants. Although most TANT or VOLOCITY (Improvision, Lexington, MA) software. of the P. aeruginosa mutants were from a preexisting library, some were constructed as described here. For construction of the Twitching Motility Assays. For twitching motility assays we used PA2387 mutant (EB101), a 1.4-kb HindIII-NotI DNA fragment Petri plates with 1% TSB plus 1% Noble agar (Becton Dickin- upstream of PA2387 (from bp 2638897 to bp 2640299 on the P. son) with or without 2 mM desferrioxamine to chelate iron as aeruginosa chromosome) and a 1.5-kb NotI-XbaI DNA fragment indicated. Plates were dried overnight at room temperature, and downstream of PA2387 (from bp 2640893 to bp 2674792 on the cells from a colony grown overnight on a LB-agar plate were P. aeruginosa chromosome) were amplified and cloned into a point inoculated at the bottom of the agar plate. After 3 days, the HindIII-XbaI-digested pEX18Tc gene replacement vector (22). twitching motility distance along the plastic–agar interface (at The resulting plasmid was digested with NotI and ligated with a the bottom of the agar plate) was measured. NotI-digested aacC1 cassette to create pEB2. This plasmid was mobilized from E. coli SM10 (23) into P. aeruginosa PAO1 by Results conjugation. Selection for GmR colonies followed by screening Biofilm Development in Pyoverdine and Pyochelin Synthesis Mutants. for TcS colonies yielded a PA2387 mutant, P. aeruginosa EB101. We first followed biofilm development in P. aeruginosa mutants The mutation was confirmed by PCR analysis. The PA2467– incapable of synthesizing the two well known siderophores 2468 mutant (EB102) was constructed in a similar fashion. A pyoverdine and pyochelin, PAO1⌬pvdA and PAO1pchA::TcR, 1.4-kb HindIII-NotI DNA fragment corresponding to a region respectively. Planktonic growth in 1% TSB and twitching mo- upstream of PA2467 (from bp 2783773 to bp 2785221 on the tility (data not shown) of these mutants were indistinguishable P. aeruginosa chromosome) and a NotI-XbaI 1.5-kb fragment from the parent. Biofilm formation of the pyochelin mutant downstream of PA2468 (from bp 2786704 to bp 2788240 on the under a flow of 1% TSB is similar to that of the parent. Both P. aeruginosa chromosome) were amplified and cloned into strains form mushroom-like structures on the glass surface. The HindIII-XbaI-digested pEX18Tc. The resulting plasmid was pyoverdine mutant, however, forms a thin uniform layer of then digested with NotI and ligated with a NotI-digested aacC1 growth on the glass surface that is similar to the biofilms formed cassette to create pEB3. A transconjugant was isolated, and the by the parent and pyochelin mutant in the presence of lactoferrin mutation was confirmed as described above. (Fig. 1). PvdA is required for pyoverdine synthesis, and PvdS We constructed the following double mutants using a method codes for the ␴ factor required to activate pyoverdine synthesis described in ref. 24: EB104 (PAO1 ⌬pvdA,ISlacZ͞hah PA3901), and to regulate expression of other genes. A pvdS mutant, EB107 (PAO1 ⌬pvdA,ISlacZ͞hah pchA), EB110 (PAO1 ⌬pvdA, PAO⌬pvdS, shows a biofilm phenotype similar to that of the ISlacZ͞hah PA0470), and EB111 (PAO1 ⌬pvdA,ISlacZ͞hah pvdA mutant (Fig. 1). Complementation of the pvdA and pvdS PA2466). In all cases we transformed PAO1⌬pvdA with genomic mutations restores the normal biofilm architecture (Fig. 1). DNA (40–80 ␮g) from PTL14263 to construct EB104, PTL5032 Furthermore, addition of pyoverdine-conditioned medium (see to construct strain EB107, PTL20210 to construct strain EB110, Materials and Methods) allows biofilms of the PvdA mutant to or PTL11509 to construct strain EB111. After transformation, form normally (Fig. 1). colonies were selected on LB agar plates containing Tc (100 ␮g͞ml). The mutation in each strain was confirmed by PCR. Biofilm Formation in Flow Cells Requires Active Iron Transport. Com- The triple mutant EB112 (PAO1⌬pvdA, 63-bp insertion in pared to pyoverdine, pyochelin has a low affinity for iron (9). To PA0470, ISlacZ͞hah PA2466) was constructed by using the ask whether the pyoverdine mutant can acquire iron for signaling cre-recombinase system (25). Strain EB110 was mated with biofilm development through pyochelin in the presence of E. coli carrying pCre1. Transconjugants were selected on sufficiently high environmental iron concentrations, we supple-

Banin et al. PNAS ͉ August 2, 2005 ͉ vol. 102 ͉ no. 31 ͉ 11077 Downloaded by guest on September 26, 2021 (Fig. 2). We obtained similar results with a pvdD,pchEF double mutant (data not shown). Addition of desferrioxamine or ferric dicitrate, both of which P. aeruginosa can use as an iron source (31, 32), restores biofilm formation by the double mutant (Fig. 2). These experiments suggest that an iron uptake system is required for normal biofilm development on glass and that passive diffusion of iron does not provide sufficient intracellular iron for biofilm development. To test the hypothesis that P. aeruginosa biofilms cannot obtain sufficient iron for signaling by passive diffusion even at high external iron concentrations, we examined the expression of an iron-repressed, Fur-regulated promoter (pvdS)-driven gfp in our pyoverdine, pyochelin double mutant (EB125). With or without added iron (100 ␮M FeCl3), biofilms are flat (Ͻ15 ␮m), and the cells express GFP (Fig. 3). In the presence of added iron, the cells on the biofilm surface express less GFP than those at the base of the biofilm (Fig. 3). This finding suggests that there is sufficient iron for Fur to serve as a repressor in cells near the biofilm surface, but without an iron uptake system passive diffusion does not provide sufficient iron for Fur repression throughout the biofilm. When the double mutant is grown in the presence of desferrioxamine, which can be taken up by the ferrioxamine transport system, mushroom- like structures form and the cells express low levels of GFP (Fig. 3). These data are consistent with the hypothesis that an active iron transport system is needed for normal biofilm development.

Do Mutations in Other Iron-Responsive ECF ␴-Factor Systems Influ- ence Biofilm Formation? Visca et al. (12) identified 19 ORFs encoding putative ECF ␴-factors in the P. aeruginosa genome. Thirteen of these show substantial sequence similarity with iron starvation ␴ factors, and they contain putative Fur binding sites in their promoter region. Several are in clusters that include genes predicted to encode outer-membrane receptors involved in iron transport. We wanted to determine whether any of these putative iron acquisition signaling systems is involved in biofilm development; therefore, we studied mutants with defects in the putative iron-responsive ECF ␴-factor genes and the putative Fig. 1. Bioﬁlm formation of P. aeruginosa pyoverdine and pyochelin mu- FecA iron receptor homologs. Mutants with insertions in the tants. The parent without (A) and with (B) lactoferrin (20 ␮g͞ml), the pyoch- following ECF ␴-factor genes were examined: PA0149, PA0472, elin mutant PAO1pchA::TcR without (C) and with (D) lactoferrin (20 ␮g͞ml), PA0675, PA1300, PA1363, PA1912, PA2050, PA2387, PA2468, the pyoverdine synthesis mutant PAO1⌬pvdA (E), the pyoverdine ECF ␴-factor ⌬ PA3410, PA3899, and PA4896. Iron receptor mutants with PvdS mutant PAO pvdS without lactoferrin (F), the pyoverdine mutant car- insertions in PA0151, PA0470, PA0674, PA1302, PA1910, rying the pvdA expression vector pEB4 without lactoferrin (G), and the pyoverdine mutant grown in pyoverdine-conditioned medium without lactoferrin PA1365, PA2398, PA2466, PA3408, and PA3901 were also (H). The P. aeruginosa cells contained the gfp plasmid pMRP9-1. Images are examined. None of the mutations has an appreciable affect on from 6-day bioﬁlms (the squares are 61 ␮m on a side). planktonic growth or twitching motility (data not shown). With a single exception, biofilm formation of the mutants is indistinguishable from the parent in the presence (thin, flat biofilms) or mented the medium with 50 ␮M FeCl3. In this iron-rich medium absence (large, mushroom-like structures) of lactoferrin. The the pyoverdine mutant produces biofilms that are similar to the single exception is the PA2398 mutant, PAO1fpvA::GmR. This parent. A pvdA,pchA double mutant, EB107, forms thin, flat fpvA mutant forms thin, flat biofilms in the absence of lactoferrin biofilms in high-iron medium, even at 100 ␮M ferric chloride that are similar in appearance to those formed by pvd mutants

Fig. 2. Active iron transport is required for normal biofilm development in flow cells. (A) Biofilm of the pvdA mutant PAO1⌬pvdA grown with added ferric chloride (50 ␮M). Biofilm of a pyoverdine, pyochelin double mutant (EB107) grown with 100 ␮M ferric chloride (B)or1.5␮M desferrioxamine (C) (similar results were obtained when 1 ␮M ferric dicitrate was added; data not shown). All strains contained the gfp plasmid pMRP9-1. Images represent 6-day biofilms (the squares are 30 ␮m on a side).

11078 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0504266102 Banin et al. Downloaded by guest on September 26, 2021 Fig. 4. Ferric citrate and ferrioxamine mediate development of mushroom- like structures in the pyoverdine synthesis mutant PAO1⌬pvdA. The pvdA mutant biofilm in medium supplemented with 1 ␮M ferric dicitrate (A)or1.5 ␮M desferrioxamine (B). (C) A biofilm of the pvdA, PA3901 double mutant EB104 in medium containing 1 ␮M ferric dicitrate. (D) A biofilm of EB112 (the pyoverdine, PA2466, and PA0470 triple mutant) in medium with 1.5 ␮M desferrioxamine. All strains contained pMRP9-1. Images represent 6-day biofilms (the squares are 62 ␮m on a side). MICROBIOLOGY even in the presence of ferric dicitrate (Fig. 4). Similar to the results obtained with ferric dicitrate as an alternative iron scavenger, addition of desferrioxamine (1.5 ␮M) also restores mushroom formation to the pvdA mutant. In the P. aeruginosa genome there are two gene clusters (PA0470–0472 and Fig. 3. Expression of pvdS-gfp in biofilms of a pyoverdine, pyochelin double PA2466–2468) that show extensive sequence identity with fer- mutant (EB125) in the presence of FeCl3 or desferrioxamine. (A) Biofilms grown without FeCl3.(B) Biofilms grown with 100 ␮M FeCl3.(C) Biofilms grown rioxamine uptake systems from other bacteria (15). We con- with 1.5 ␮M desferrioxamine. Images represent 6-day biofilms counterstained structed two double mutants that contain a deletion in pvdA and with propidium iodide. (A–C Upper) A reconstructed side view of the com- an insertion in PA0470 or PA2466 (EB110 and EB111, respec- bined red and green channels. Red shows the biofilm, and green shows tively). Both mutants form mushroom-like structures in the regions where cells have expressed GFP. (A–C Lower)A side view of GFP presence of desferrioxamine (data not shown). A triple mutant ␮ expression alone. (Scale bar in Lower,15 m; all of the side images are the (EB112) that contains deletions in pvdA and PA2466 and an same magnification.) insertion in PA0470 forms a sparse biofilm (Fig. 4). This triple mutant exhibits normal twitching motility, but its growth in broth in the absence of lactoferrin (data not shown). FpvA is known to containing desferrioxamine is poor (data not shown). These data function in the pyoverdine signaling and uptake cascade (33). are consistent with the conclusion that with sufficient levels of cytoplasmic iron P. aeruginosa biofilms form mushroom-like Alternate Sources of Iron Can Signal Biofilm Development in a structures. These experiments also provide evidence suggesting Pyoverdine Synthesis Mutant. Without sufficiently high levels of that PA3899–3901 are involved in ferric dicitrate uptake and that environmental iron, pyoverdine mutants form thin, flat biofilms the PA0470 and PA2466 gene clusters are involved in ferriox- (Fig. 1), and we believe that cytoplasmic iron levels or transport amine uptake. Further work is required to critically establish of chelated iron cues biofilm development. However, we know these gene clusters as ferric dicitrate and desferrioxamine uptake that P. aeruginosa can use a variety of chelated forms of systems. environmental iron and that it can acquire iron chelated by other organisms (9). For example, it can use iron chelated with citrate Intracellular Iron Signaling Is Mediated by Fur. The experiments or with desferrioxamine (31, 32, 34). If iron signaling in biofilm described above are consistent with the hypothesis that biofilm formation is mediated by a cytoplasmic receptor or active development is regulated by intracellular iron. The major intra- transport of chelated iron, then provision of ferric dicitrate or cellular iron regulator in Proteobacteria including P. aeruginosa desferrioxamine should allow the pvdA mutant to form biofilms is Fur (6, 15). The P. aeruginosa Fur appears to be an essential with the wild-type mushroom-like structures. In fact, biofilms of gene; however, Fur missense mutants have been isolated (17, the pvdA mutant grown in the presence of 1 ␮M ferric dicitrate 18). Fur is an iron-dependent transcription repressor. Thus, one show the mushroom-like appearance of the parent strain (Fig. 4). might predict that a P. aeruginosa Fur mutant could build The P. aeruginosa PA3899–3901 gene cluster codes for structured biofilms even under lactoferrin-induced iron limita- polypeptides showing extensive sequence identity with the E. coli tion. Accordingly, we followed biofilm development of a Fur ferric dicitrate uptake polypeptides, the Fec polypeptides (35). mutant in the presence and absence of lactoferrin. Without Thus, we constructed P. aeruginosa EB104, which contains an lactoferrin the mutant forms mushroom structures that resem- insertion in PA3901 and a deletion in pvdA. This mutant shows bled parent structures without lactoferrin (Fig. 5). In the pres- no appreciable differences in growth and twitching motility ence of lactoferrin, the mutant forms structured biofilms compared with the parent; however, it forms thin, flat biofilms whereas the parent forms the expected flat, thin biofilms (Fig. 5).

Banin et al. PNAS ͉ August 2, 2005 ͉ vol. 102 ͉ no. 31 ͉ 11079 Downloaded by guest on September 26, 2021 activity as an iron chelator. These experiments indicated that iron serves as an environmental signal for P. aeruginosa biofilm development. To better understand the role of iron in P. aeruginosa biofilm formation, we studied a variety of mutants with defects in known or suspected iron-uptake and regulatory functions. In the absence of a functional iron uptake system, P. aeruginosa forms thin, flat biofilms in the absence of lactoferrin (Fig. 1). We show that P. aeruginosa can acquire iron and support normal biofilm formation by using the endogenous siderophores pyoverdine (at relatively low external iron concentrations) or pyochelin (at higher iron levels). It can also use specific iron-responsive ECF ␴-factor systems to acquire ferric citrate or ferrioxamine for biofilm formation (Figs. 1–4). It cannot acquire sufficient amounts of iron for normal biofilm development by passive diffusion in the flow-cell system (Fig. 3). All of our data on biofilm formation by iron acquisition mutants are consistent with the conclusion that biofilm formation in flow cells requires active iron transport and sufficient intracellular iron serves as a cue for development of mushroom-like structures. Although the pyoverdine uptake system has a dual function in that it also serves to regulate expression of a number of virulence genes directly (11, 36), this does not appear to be its role in iron regulation of biofilm formation. Strains with mutations in the other predicted iron uptake systems formed normal biofilms. This finding is inconsistent with a recent report that a specific virulence mutant had a defect in a putative iron receptor (PA0151) and had a defect in biofilm formation (37). We tested two PA0151 mutants, and both formed normal biofilms in the absence of lactoferrin. This discrepancy could result from the differences in experimental protocols or strain variation. Our results show that, under the conditions we used and with derivatives of strain PAO1, normal biofilms form as long as there is a functional pyoverdine system and there is a sufficient level of environmental iron for the Fig. 5. Biofilm formation by Fur and PrrF mutants. (Top) Biofilms of the pyoverdine system to function. parent PAO1. (Middle) Biofilms of the Fur mutant PAO1furC6Tc. (Bottom) As discussed above, a mutant strain that could not produce the ⌬ Biofilms of the PAO1 prrF1-F2 double mutant. Where indicated, biofilms high-affinity siderophore pyoverdine or the low-affinity sid- were grown in the presence of 20 ␮g͞ml lactoferrin. Strain PAO1 and PAO1⌬prrF1-F2 contained the gfp plasmid pMRP9-1. The Fur mutant was erophore pyochelin is unable to form mushroom-like structures stained with propidium iodide (we could not introduce pMRP9-1 into the even in the presence of high levels of inorganic iron, but it can strain). Images in Left are of 6-day biofilms, and the squares are 60 ␮mona form these structures when given ferric dicitrate or desferriox- side. Images in Right are also images of 6-day biofilms, but the squares are 30 amine (Fig. 2). We know that P. aeruginosa can use ferric citrate ␮m on a side. and other chelated forms of iron (9, 31, 33), but there is no experimental evidence linking specific genes to ferric dicitrate utilization. Among the various mutants with specific defects in These observations suggest that Fur controls genes that are putative iron-responsive ECF ␴-factor systems that we tested, crucial in biofilm development. there is one with a mutation in the PA3899–3901 gene cluster. Wilderman et al. (16) recently identified two Fur-repressed The polypeptides coded by genes in this cluster show Ͼ50% sRNA molecules in P. aeruginosa. These sRNAs (PrrF1 and identity to the E. coli ferric uptake gene products FecA, FecI, PrrF2) are Ͼ95% identical to each other, and expression of both and FecR (35). This E. coli system is responsible for ferric is repressed by iron. However, each can also be independently dicitrate uptake. We thus constructed a mutant incapable of regulated by other factors (e.g., heme). Deletion of both sRNAs producing pyoverdine and this putative ferric citrate system. This is required to affect the iron-dependent regulation of an array of mutant produced thin, flat biofilms even when ferric dicitrate genes, including those involved in resistance to oxidative stress, was provided (Fig. 4). This finding suggests that the PA3899– iron storage, and intermediary metabolism. Unlike the Fur 3901 cluster codes for a ferric dicitrate uptake system. We also mutant, our PrrF1F2 double mutant formed biofilms similar to have similar data suggesting that two gene clusters PA2466–2468 the parent mushroom-like structures in the absence of lactofer- and PA0470–0472 facilitate ferrioxamine uptake (Fig. 4). Fur- rin and thin, flat biofilms in the presence of lactoferrin (Fig. 5). ther characterization of these gene clusters is required to verify Thus, we conclude that these sRNAs are not required for biofilm their role in iron transport. We are currently employing this formation under the conditions used in this study. screening approach to correlate specific iron-responsive ECF ␴-factor systems with the ability to use different chelated forms Discussion of iron for biofilm development. Singh et al. (4) discovered that the mammalian iron chelator If, as we conclude, cytoplasmic iron functions as the signal for lactoferrin restricts the development of P. aeruginosa biofilms. In development of mushroom-like structures, a next question is the presence of lactoferrin, P. aeruginosa forms a thin layer of what is the intracellular iron response regulator? A logical cells on glass surfaces and the individual cells exhibit incessant candidate is Fur, the major iron-responsive transcriptional reg- twitching motility. In the absence of lactoferrin, P. aeruginosa ulator in P. aeruginosa (15). If Fur controls a gene or genes constructs thicker biofilms consisting of cells clustered in mush- required for biofilm development into mushroom-like struc- room-like structures. The lactoferrin effect was correlated to its tures, one might predict that a Fur mutant could form such

11080 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0504266102 Banin et al. Downloaded by guest on September 26, 2021 clusters in the presence of lactoferrin where the parent forms hypothetical checkpoints and form mushroom-like structures even thin, flat biofilms. We tested a strain with a point mutation in this in the presence of lactoferrin. essential gene. This mutation results in low Fur activity (17). The Previous studies showed that the formation of mushroom-like influence of lactoferrin on the biofilm development of this structures by P. aeruginosa biofilms depended on the carbon and mutant was not as severe as the influence of lactoferrin on the energy source provided in the growth medium (39, 40). For parent. Whereas the parent formed flat, thin biofilms in the biofilms grown in minimal medium, glucose supported forma- presence of lactoferrin, the mutant formed large clusters of cells tion of mushroom-like structures, but citrate did not. This finding reminiscent of, but not identical to, the parent in the absence of may suggest that the iron-regulatory pathway and the carbon lactoferrin (Fig. 5). Fur appears to mediate iron signaling in source-regulatory pathway might converge at a point down- biofilm development through its regulation of multiple iron stream of Fur involvement. acquisition systems and their regulatory genes (e.g., ECF ␴ It is not surprising that multiple iron uptake systems play a role factors). The fact that Fur mutants produce high amounts of in biofilm formation. It is likely that the different environments pyoverdine (17) may account in part for the biofilm architecture where P. aeruginosa resides dictate which forms of iron might be of our Fur mutant. We do not know what other Fur-regulated available. For example, ferric citrate or ferrioxamine are not factors might be involved in biofilm formation. Nevertheless, we likely to be available iron sources in a mammalian host. Finally, ruled out a contribution of the Fur-repressed sRNAs in biofilm it must be said that no one yet understands the significance of the growth (Fig. 5). formation of mushroom-like structures on glass surfaces to Biofilm development is thought to occur through a series of biofilm formation in natural environments. Nevertheless, under- discrete steps. The first step involves attachment of cells to a standing the pathways leading to formation of these structures surface. The second step is microcolony formation. A third step informs us about regulatory networks involved in biofilm devel- involves maturation of microcolonies into mushroom-like struc- opment. It seems clear that iron plays a critical role in biofilm tures. Finally, there is a detachment step. Previous work suggests formation. It may be that there are several iron-regulated steps that iron sensing might serve as a checkpoint in the first two steps. in biofilm formation. We show that at least one of these depends O’Toole and Kolter (38) reported that the defect in certain attach- on the ability of cells to import sufficient levels of iron for Fur ment mutants can be overcome if sufficient iron is added to the activity. medium. Singh et al. (4) showed that low iron stimulates twitching motility and blocks the formation of microcolonies on glass sur- We thank Michael Jacobs, Colin Manoil, and Maynard Olson for

faces. Our data suggest that there may be another iron checkpoint providing us with the P. aeruginosa mutants from the University of MICROBIOLOGY for development of microcolonies that is not related to twitching. Washington Genome Center P. aeruginosa PAO1 transposon mutant library. We thank Pradeep Singh for his advice and comments. This In support of a third checkpoint, the pyoverdine mutants (PvdA, work was supported by a grant from the W. M. Keck Foundation (to FpvA, and PvdS) did not show a twitching motility phenotype. E.P.G.), National Institute of General Medical Sciences Grant GM59026 Furthermore, time-lapse microscopy of the mutants developing as (to E.P.G.), and National Institute of Allergy and Infectious Diseases biofilms did not show the incessant twitching described by Singh Grant AI15940 (to M.L.V.). E.B. was partially funded by the Fulbright et al. (data not shown). The Fur mutant appears to bypass all three program.

1. Costerton, J. W., Stewart, P. S. & Greenberg, E. P. (1999) Science 284, 21. King, E. O., Ward, M. K. & Raney, D. E. (1954) J. Lab. Clin. Med. 44, 301–307. 1318–1322. 22. Hoang, T. T., Karkhoff-Schweizer, R. R., Kutchma, A. J. & Schweizer, H. P. 2. Hall-Stoodley, L., Costerton, J. W. & Stoodley, P. (2004) Nat. Rev. Microbiol. (1998) Gene 212, 77–86. 2, 95–108. 23. de Lorenzo, V. & Timmis, K. N. (1994) Methods Enzymol. 235, 386–405. 3. Lyczak, J. B., Cannon, C. L. & Pier, G. B. (2000) Microbes Infect. 2, 1051–1060. 24. Schuster, M., Hawkins, A. C., Harwood, C. S. & Greenberg, E. P. (2004) Mol. 4. Singh, P. K., Parsek, M. R., Greenberg, E. P. & Welsh, M. J. (2002) Nature 417, Microbiol. 51, 973–985. 552–555. 25. Bailey, J. & Manoil, C. (2002) Nat. Biotechnol. 20, 839–842. 5. Singh, P. K. (2004) Biometals 17, 267–270. 26. Miller, W. G., Leveau, J. H. & Lindow, S. E. (2000) Mol. Plant–Microbe Interact. 6. Ratledge, C. & Dover, L. G. (2000) Annu. Rev. Microbiol. 54, 881–941. 13, 1243–1250. 7. Braun, V. (1997) Biol. Chem. 378, 779–786. 27. Newman, J. R. & Fuqua, C. (1999) Gene 227, 197–203. 8. Griffiths, E. (1999) in Iron and Infection: Molecular, Physiological and Clinical 28. Ochsner, U. A., Johnson, Z., Lamont, I. L., Cunliffe, H. E. & Vasil, M. L. (1996) Aspects, eds. Bullen, J. J. & Griffiths, E. (Wiley, New York), 2nd Ed., pp. 1–26. Mol. Microbiol. 21, 1019–1028. 9. Poole, K. & McKay, G. A. (2003) Front. Biosci. 8, d661–d686. 29. Davies, D. G., Parsek, M. R., Pearson, J. P., Iglewski, B. H., Costerton, J. W. 10. Wandersman, C. & Delepelaire, P. (2004) Annu. Rev. Microbiol. 58, 611–647. & Greenberg, E. P. (1998) Science 280, 295–298. 11. Lamont, I. L., Beare, P. A., Ochsner, U., Vasil, A. I. & Vasil, M. L. (2002) Proc. 30. Yarwood, J. M., Bartels, D. J., Volper, E. M. & Greenberg, E. P. (2004) J. Natl. Acad. Sci. USA 99, 7072–7077. Bacteriol. 186, 1838–1850. 12. Visca, P., Leoni, L., Wilson, M. J. & Lamont, I. L. (2002) Mol. Microbiol. 45, 31. Cox, C. D. (1980) J. Bacteriol. 142, 581–587. 1177–1190. 32. Cornelis, P., Moguilevsky, N., Jacques, J. F. & Masson, P. L. (1987) Antibiot. 13. Braun, V. & Braun, M. (2002) FEBS Lett. 529, 78–85. Chemother. 39, 290–306. 14. Ochsner, U. A., Wilderman, P. J., Vasil, A. I. & Vasil, M. L. (2002) Mol. 33. Poole, K., Neshat, S., Krebes, K. & Heinrichs, D. E. (1993) J. Bacteriol. 175, Microbiol. 45, 1277–1287. 4597–4604. 15. Vasil, M. L. & Ochsner, U. A. (1999) Mol. Microbiol. 34, 399–413. 34. Brock, J. H., Liceaga, J. & Kontoghiorghes, G. J. (1988) FEMS Microbiol. 16. Wilderman, P. J., Sowa, N. A., FitzGerald, D. J., FitzGerald, P. C., Gottesman, Immunol. 1, 55–60. S., Ochsner, U. A. & Vasil, M. L. (2004) Proc. Natl. Acad. Sci. USA 101, 35. Mahren, S., Enz, S. & Braun, V. (2002) J. Bacteriol. 184, 3704–3711. 9792–9797. 36. Beare, P. A., For, R. J., Martin, L. W. & Lamont, I. L. (2003) Mol. Microbiol. 17. Barton, H. A., Johnson, Z., Cox, C. D., Vasil, A. I. & Vasil, M. L. (1996) Mol. 47, 195–207. Microbiol. 21, 1001–1017. 37. Potvin, E., Lehoux, D. E., Kukavica-Ibrulj, I., Richard, K. L., Sanschagrin, F., 18. Ochsner, U. A., Vasil, A. I., Johnson, Z. & Vasil, M. L. (1999) J. Bacteriol. 181, Lau, G. W. & Levesque, R. C. (2003) Environ. Microbiol. 5, 1294–1308. 1099–1109. 38. O’Toole, G. A. & Kolter, R. (1998) Mol. Microbiol. 28, 449–461. 19. Jacobs, M. A., Alwood, A., Thaipisuttikul, I., Spencer, D., Haugen, E., Ernst, 39. O’Toole, G. A., Gibbs, K. A., Hager, P. W., Phibbs, P. V., Jr., & Kolter, R. S., Will, O., Kaul, R., Raymond, C., Levy, R., et al. (2003) Proc. Natl. Acad. Sci. (2000) J. Bacteriol. 182, 425–431. USA 100, 14339–14344. 40. Klausen, M., Aaes-Jorgensen, A., Molin, S. & Tolker-Nielsen, T. (2003) Mol. 20. Ochsner, U. A. & Vasil, M. L. (1996) Proc. Natl. Acad. Sci. USA 93, 4409–4414. Microbiol. 50, 61–68.

Banin et al. PNAS ͉ August 2, 2005 ͉ vol. 102 ͉ no. 31 ͉ 11081 Downloaded by guest on September 26, 2021