Magnesium Limitation Is an Environmental Trigger of the Pseudomonas aeruginosa Biofilm Lifestyle

Biofilm formation is a conserved strategy for long-term bacterial survival in nature and during infections. Biofilms are multicellular aggregates of cells enmeshed in an extracellular matrix. The RetS, GacS and LadS sensors control the switch from a planktonic to a biofilm mode of growth in Pseudomonas aeruginosa. Here we detail our approach to identify environmental triggers of biofilm formation by investigating environmental conditions that repress expression of the biofilm repressor RetS. Mg2+ limitation repressed the expression of retS leading to increased aggregation, exopolysaccharide (EPS) production and biofilm formation. Repression of retS expression under Mg2+ limitation corresponded with induced expression of the GacA-controlled small regulatory RNAs rsmZ and rsmY and the EPS biosynthesis operons pel and psl. We recently demonstrated that extracellular DNA sequesters Mg2+ cations and activates the cation-sensing PhoPQ two-component system, which leads to increased antimicrobial peptide resistance in biofilms. Here we show that exogenous DNA and EDTA, through their ability to chelate Mg2+, promoted biofilm formation. The repression of retS in low Mg2+ was directly controlled by PhoPQ. PhoP also directly controlled expression of rsmZ but not rsmY suggesting that PhoPQ controls the equilibrium of the small regulatory RNAs and thus fine-tunes the expression of genes in the RetS pathway. In summary, Mg2+ limitation is a biologically relevant environmental condition and the first bonafide environmental signal identified that results in transcriptional repression of retS and promotes P. aeruginosa biofilm formation.


Introduction
In the natural environment and during infection of susceptible hosts bacteria predominantly grow as biofilms. Biofilms are surface-associated, microbial communities, which are embedded in an extracellular matrix composed primarily of bacterial-derived exopolysaccharides (EPS) and DNA [1][2][3]. Biofilms have been intensively studied in recent years due to their significance in industrial, natural and medical settings. A multicellular biofilm lifestyle for bacteria has survival advantages compared to bacteria living as planktonic individual cells. During infection, growth as a biofilm confers resistance to multiple environmental stresses, antibiotics and the immune system [4][5][6].
The environmental bacterium and opportunistic human pathogen Pseudomonas aeruginosa PAO1 is one of the most widely used model organisms for studying bacterial biofilm formation. P. aeruginosa is a versatile Gram-negative bacterium that grows in soil and marine environments, as well as on plant and animal tissues [7]. P. aeruginosa has also emerged as a major opportunistic human pathogen during the past century [8] and individuals with the genetic disease Cystic Fibrosis (CF) are particularly susceptible [9]. A large body of in vitro and in vivo data in the literature indicates that P. aeruginosa exists as a biofilm in the lungs of CF patients [10][11][12][13][14][15].
Biofilm formation protects bacteria from harsh and stressful conditions [16] and the ability of bacteria to adapt to changing environmental conditions is essential for survival. P. aeruginosa is capable of sensing environmental conditions and adapting to changing conditions through modification of gene expression. A common mechanism of adaptation, which is both rapid and reversible, utilizes two-component systems (TCS) [17]. In many bacterial systems, TCS are involved in the regulation of biofilm formation. Typical TCS are comprised of a membrane-anchored histidine kinase sensor and a cytoplasmic response regulator (RR). After the sensor detects specific environmental signals, a signal transduction cascade is initiated that results in phosphorylation of a RR, which activates or represses necessary target genes. In P. aeruginosa multiple sensor proteins have been identified, mostly through genetic screens, which are important for P. aeruginosa biofilm formation. These include the sensor proteins GacS, RetS, LadS, SadS and PhoQ [18][19][20][21][22][23] as well as the BfiS, BfmS and MifS sensors, which control progression through distinct stages of biofilm maturation [24]. However, the exact signals detected by most of these environmental sensors are unknown.
RetS is required for expression of the virulence-associated type III secretion system (T3SS) and for repression of the pel and psl biofilm matrix EPS biosynthesis genes through both transcriptional and post-transcriptional regulation in P. aeruginosa [19,20,25,26].
RetS directly interacts and forms heterodimers with the GacS sensor protein, preventing activation of the GacAS pathway and repressing biofilm formation [27,28]. Reduced levels of RetS favors the formation of GacS homodimers, autophosphorylation of GacS and activation of the GacA-controlled the small regulatory RNAs (sRNA), rsmZ and rsmY [29,30]. RsmZ and rsmY bind and sequester the post-transcriptional regulatory protein RsmA, which normally functions to bind other target mRNAs. RsmA-mediated regulation can be direct, through mRNA binding and preventing initiation of translation such as that observed for psl [26] or indirect, by interfering with the translation of specific regulatory factors [25]. However in addition post-transcriptional regulation of psl genes, genome-wide transcriptional profiling identified retS as a transcriptional regulator of the pel and psl EPS operons [19].
It has been proposed that the RetS sensor responds to environmental conditions encountered during acute infections maintaining P. aeruginosa in a planktonic growth state capable of Type III secretion [20]. In contrast to this model, we report here that Mg 2+ limitation causes transcriptional repression of retS and promotes a switch from the planktonic to a biofilm lifestyle. Repression of retS occurred through direct repression by the cation sensing PhoPQ two-component system. Recent work from our laboratory has identified extracellular DNA as a chelator of divalent cations that activates the PhoPQ two-component system, resulting in the expression of antibiotic resistance genes [31]. To our knowledge this is the first demonstration of a specific environmental signal that promotes a switch to a biofilm mode of growth by acting through the RetS/LadS/GacS pathway. As DNA is abundant both in the natural and host environment [32][33][34][35][36][37], Mg 2+ limitation is a relevant biological signal and is encountered ubiquitously by P. aeruginosa.

Results
Identification of Mg 2+ limitation as an environmental signal that represses retS expression RetS is required for repression of EPS biosynthesis genes and preventing biofilm formation in P. aeruginosa [19]. Our strategy was to identify environmental signals that repressed the expression of retS and thus likely promoted biofilm formation in P. aeruginosa. To identify environmental signals that repressed the levels of RetS, we monitored expression of the retS promoter fused to the lux (bioluminescence) reporter in diverse growth conditions. The growth conditions tested were intended to mimic the conditions faced by P. aeruginosa during chronic lung infections (Table S1).
RetS expression was induced between 10-and 50-fold, or repressed between 7-and 125-fold in the conditions tested. Over the course of 20 h-growth, maximal repression (up to 125-fold) of retS was observed in limiting Mg 2+ conditions ( Figure 1A). We were particularly interested in retS gene expression under limiting Mg 2+ growth conditions because we recently showed that extracellular DNA is an efficient chelator of divalent cations including Mg 2+ [31]. Consistent with DNA acting as a chelator of Mg 2+ , retS expression was repressed up to 25-fold in media supplemented with exogenous DNA ( Figure 1A). The expression of retS was repressed in a Mg 2+ concentration-dependent manner ( Figure 1B), with maximal repression observed at 0.02 mM Mg 2+ . Gene expression was normalized to growth as growth rates of cells grown under 2 mM or 0.02 mM Mg 2+ were similar (6.2 h and 6.6 h, respectively).
Mg 2+ limitation results in a gene expression signature consistent with the switch to a biofilm lifestyle Several studies examining the regulation of virulence factor production in P. aeruginosa identified a potential regulatory switch involving the RetS/LadS/GacS two component sensors, that controls the transition between acute and chronic infection related phenotypes [19,20]. Identification of limiting Mg 2+ as an environmental signal that represses expression of the known biofilm repressor gene retS prompted us to investigate the expression of a panel of genes controlled by the RetS/LadS/ GacS pathway that are known to be important for either acute or chronic P. aeruginosa infections.
The data in Figure 2A represents the time course of fold induction (green) or repression (red) of genes in BM2 growth medium with limiting Mg 2+ relative to high Mg 2+ over 18 h of growth. In the cluster analysis depicted in Figure 2A, gene expression profiles cluster into two distinct groups: genes that are upregulated or downregulated in limiting Mg 2+ conditions. The cluster of induced genes included genes previously identified as regulated by Mg 2+ limitation by the PhoPQ or PmrAB TCS. These included PA3553 (arnC) from the LPS modification operon PA3552-3559, the outer membrane protein oprH, and a putative polyamine synthesis gene PA4774 [38][39][40]. Additionally the small regulatory RNAs encoded by rsmZ and rsmY, as well as genes from two independent EPS biosynthesis operons, pelD and pslA, were also induced by Mg 2+ limitation ( Figure 2). In contrast, genes encoding the retS biofilm regulator, type II secretion system genes (xcpR and aprA) and a type III secretion system effector, exoT, clustered together as genes repressed by Mg 2+ limitation. To demonstrate the magnitude of induction or repression, Figure 2B illustrates the maximal change in gene expression over 18 h.
According to the current model of reciprocal regulation of acute and chronic infection related traits by the RetS/LadS/GacS pathway, biofilms and chronic infections are promoted under conditions that induce EPS expression while simultaneously repressing the T2SS and T3SS [19,20]. The expression profiles observed for the genes shown in Figure 2 are consistent with limiting Mg 2+ acting as a signal that promotes a switch from the planktonic to an aggregative, biofilm mode of growth. We observed that maximal repression of the T2SS and T3SS genes occurred in the log phase of growth, prior to maximal repression of retS. This indicates that additional factors, the identification of which are beyond the scope of this study, may be involved in regulating these secretion system genes under Mg 2+ limitation.

Mg 2+ limitation promotes biofilm formation and aggregation
Pel and Psl EPS are essential biofilm matrix components that are important for adhesion and biofilm formation [41,42]. As the expression of pel and psl EPS biosynthesis genes were strongly induced by Mg 2+ limitation, we predicted that biofilm formation would be increased under these conditions. Mg 2+ limitation strongly promoted biofilm formation, as measured by crystal violet (CV) staining of the total biomass adhered to the polystyrene pegs ( Figure 3A) or to glass surfaces ( Figure 3B). Additionally, large

The cation chelators DNA and EDTA induce biofilm formation
We have recently shown that extracellular DNA can function as a cation chelator that activates the cation-sensing PhoPQ TCS, leading to increased antimicrobial peptide resistance [31]. Since limiting Mg 2+ is an environmental condition that also promoted biofilm formation, we proposed that cation chelation by extracellular DNA may ultimately impose a cation limitation on cells and promote biofilm formation.
To test this hypothesis, we cultivated ring biofilms in media with increasing concentrations of exogenous DNA. The addition of salmon sperm DNA promoted biofilm formation in a concentration-dependent manner ( Figure 4A). To confirm that the chelating activity of DNA promoted biofilm formation, as opposed to the adhesive capacity of DNA, exogenous Mg 2+ was added to cultures. The biofilm promoting effect of exogenous DNA was significantly reduced by addition of 10 mM excess Mg 2+ , to a final concentration of 12 mM Mg 2+ ( Figure 4A). The cation chelator EDTA also caused a concentration-dependent increase in biofilm formation that was also completely neutralized by 10 mM excess Mg 2+ ( Figure 4B).

Mg 2+ limitation induces production of EPS in the biofilm matrix
Biofilm formation is induced under limiting Mg 2+ conditions ( Figure 3). To confirm increased EPS production under limiting Mg 2+ conditions, we compared EPS production in planktonic cultures grown in 0.02 and 2 mM Mg 2+ . Using the congo red assay for measuring EPS production [43], P. aeruginosa produced significantly more total EPS under Mg 2+ limiting conditions relative to high Mg 2+ conditions ( Figure 5A). Single mutants in either the pel or psl EPS biosynthesis operons showed significantly decreased congo red binding, 40% of wild-type levels ( Figure 5B). Furthermore, the double pel/psl mutant exhibited 60% reduced congo red binding relative to PAO1, confirming the requirement of Pel and Psl for EPS production ( Figure 5B). Both rsmA and retS mutants (PAZH13, retS::lux) exhibited elevated EPS levels and were included as positive controls for EPS production.
Cellulase has been used previously to degrade Psl EPS and effectively reduces P. aeruginosa biofilm formation [43]. The addition of cellulase during biofilm cultivation had no effect on  biofilms cultivated in high Mg 2+ conditions but caused a significant reduction in biofilm formation under Mg 2+ limitation ( Figure 5C). This data indicates the importance of EPS in biofilm formation under limiting Mg 2+ conditions as degradation of the EPS matrix with cellulase results in significantly lower biofilm formation during growth in limiting Mg 2+ .
To correlate EPS production with surface attachment, we tested biofilm formation phenotypes of EPS synthesis mutants. Individual pel or psl and double pel/psl mutants exhibited significantly reduced biofilm formation (up to 80%) relative to PAO1 ( Figure S1). Additionally, we found that biofilm formation under Mg 2+ limitation was not significantly reduced in mutants lacking type IV pili or flagella production ( Figure S1). This suggests that Mg 2+limited biofilm formation does not require the presence of pili or flagella, which are generally important adhesins for biofilm formation [44].
Calcofluor binds sugars with b-1,4 linkages and has previously been shown to positively correlate with EPS production in a number of bacterial species including P. aeruginosa [41,43,45], Salmonella enterica serovar Typhimurium, and Escherichia coli [46][47][48][49]. Microscopic analysis indicated calcofluor bound to P. aeruginosa aggregates cultured under limiting Mg 2+ and the absence of both aggregation and calcofluor binding in high Mg 2+ conditions ( Figure 5D). Aggregates formed under limiting Mg 2+conditions were EPS-dependent as pelD, pslH and pel/psl mutants failed to aggregate or stain positively for EPS under these conditions ( Figure S2). Scanning electron microscopy indicated that biofilms grown under high Mg 2+ conditions resulted in diffuse cell clusters with no observable interconnecting matrix or conditioning layer ( Figure 5E). However, biofilms grown in limiting Mg 2+ conditions clearly produced an extracellular matrix that coated the plastic surface and connected cells in web-like patterns ( Figure 5E). Taken together a combination of phenotypic assays and microscopic analysis confirmed EPS overproduction in limiting Mg 2+ conditions and that increased expression of EPS promoted biofilm formation in limiting Mg 2+ conditions.
Aggregation is not the signal that leads to retS repression under Mg 2+ limiting conditions Our data was consistent with the hypothesis that Mg 2+ limitation led to increased EPS production and biofilm formation, as a result of repression of the biofilm repressor RetS. An alternative interpretation to this data could be that aggregation itself may serve as the environmental cue for repressing retS transcription. To test this possibility, we examined the effects of adding cellulase, which degrades EPS, on retS expression. Cellulase was added to limiting Mg 2+ cultures at the beginning of growth, as cellulase is known to degrade the EPS matrix and reduce cell-cell aggregation [43] and biofilm formation ( Figure 5C). Cellulase treatment did not result in an increase in retS expression ( Figure 5F). Similarly, retS expression was still repressed in the non-aggregating pel/psl double mutant strain compared to wildtype levels (data not shown) indicating that Mg 2+ limitation represses retS expression independent of aggregation.

Overexpression of the RetS sensor prevents biofilm formation under Mg 2+ limitation
Previous studies showed mutation of retS results in a hyperbiofilm phenotype due to hyperproduction of Pel and Psl exopolysaccharides [19,20]. The expression of retS is repressed in Mg 2+ limitation, which correlates with increased EPS production and biofilm formation. To definitively show that reduced levels of retS expression under Mg 2+ limitation was essential for promoting biofilm formation, the retS gene was cloned (without its native promoter) under the control of a rhamnose-inducible promoter [50]. Biofilms were cultivated in the presence of increasing amounts of rhamnose to induce retS expression in a concentrationdependent manner. In the absence of rhamnose, expression of RetS from pSCRhaB2RetS caused a reduction in biofilm formation, most likely due to low, basal levels of expression of RetS from the plasmid ( Figure 6). In support of our hypothesis, rhamnose-induced expression of RetS from pSCRhaB2RetS inhibited biofilm formation in limiting Mg 2+ conditions ( Figure 6), a condition which otherwise promotes biofilm formation (Figure 3), as a result of reduced retS expression (Figure 1).

The Mg 2+ sensing PhoPQ TCS influences the RetS regulatory pathway
Our results above identified a novel environmental signal that induced biofilm formation through repression of RetS. RetS acts as part of a complex signaling pathway that reciprocally regulates the expression of numerous virulence related genes [19]. It is unlikely that RetS itself responds to Mg 2+ levels, since RetS is predicted to bind carbohydrates [51,52]. Furthermore as PhoPQ is the only known TCS that is capable of sensing and responding to cation limitation in P. aeruginosa [53,54], we hypothesized that PhoPQ repressed retS expression, and thus influenced the RetS/ GacS/LadS pathway. If this model was true, it would be predicted that the expression profile of genes in the RetS pathway would be reversed in a phoP::xylE mutant relative to PAO1, and that high levels of retS expression in phoP::xylE would correspond to decreased levels of the small regulatory RNAs rsmZ and rsmY and the pel and psl EPS biosynthesis genes.
Expression of retS was derepressed in a phoP::xylE mutant compared to PAO1 in BM2 0.02 mM Mg 2+ , indicating that PhoP is required to repress retS ( Figure 7A). No differences in retS expression were observed between PAO1 and phoP::xylE in BM2 2 mM Mg 2+ (data not shown), a condition where the PhoPQ system is inactive. Expression of the psl EPS genes was reduced (60-fold), as predicted, in the phoP::xylE mutant ( Figure 7A). In support of the link between PhoP and EPS production, a phoP::xylE mutant produced significantly less EPS as measured by the congo red binding assay and by direct visualization using transmission electron microscopy ( Figure S3). Both rsmY and rsmZ had increased expression under limiting Mg 2+ conditions ( Figure 2) and it was hypothesized that levels of both rsmY and rsmZ would decrease in a phoP::xylE mutant. This hypothesis was true for rsmY expression, which was repressed in the phoP::xylE mutant. In contrast, rsmZ expression was more highly expressed in a phoP::xylE mutant ( Figure 7A). While rsmY and rsmZ are both induced under limiting Mg 2+ conditions in PAO1 (Figure 2), the relative levels of rsmY are higher (45-fold) ( Figure 7A). In a phoP::xylE mutant this expression pattern is reversed, with higher relative levels of rsmZ observed (47fold) ( Figure 7A), indicating that the PhoPQ system differentially regulates rsmZ and rsmY expression.

PhoP directly represses retS and rsmZ expression
We previously characterized the PhoP regulon and identified PhoP binding sites in promoters of genes directly controlled by PhoPQ [40]. We examined the retS promoter for the presence of a PhoP consensus binding site (GTTCAGNNNNNRTTCAG) and found a candidate PhoP binding site between the start codon and the 210 promoter region ( Figure 7B); a position consistent with PhoP acting as a repressor of gene expression. PhoP binding assays were performed using purified His 6 -PhoP with a 366 bp promoter fragment. PhoP bound this fragment causing a shift ( Figure 7C). Using a truncated promoter fragment (255 bp), which excluded the predicted PhoP box, we observed that PhoP did not bind to the retS promoter lacking the predicted PhoP binding site ( Figure 7C).
To confirm that PhoP repression of retS expression required the putative PhoP binding site, we constructed site-directed mutations in the PhoP box within the retS promoter. We initially constructed a retS promoter fragment that lacked the entire 18 bp PhoP box and this promoter-lux fusion had no promoter activity (data not shown), likely as a result of deleting the transcription start site. To ensure the mutation strategy did not interfere with transcription, we created site-directed changes in the second direct repeat of the PhoP box ( Figure 7D). Substitution of the wildtype sequence cgttcc for cgggac, cgcctc or cgactc resulted in derepression of retS under limiting Mg 2+ conditions in wildtype PAO1, thus confirming the identity of this sequence as the site of PhoP-mediated repression ( Figure 7D).
Our data suggested that PhoP controlled multiple genes in this pathway by repressing both retS and rsmZ ( Figure 7A). Analysis of the rsmZ promoter revealed a candidate PhoP binding site, between the start codon and the 210 promoter region ( Figure 8A), similar to the PhoP-repressing binding site identified in the retS promoter ( Figure 7B). Binding assays indicated that His 6 -PhoP was capable of direct binding to the rsmZ promoter ( Figure 8B). Taken together, cultures of PAO1 and phoP::xylE using wildtype retS promoter sequences and in PAO1 following mutation (underlined) of the second repeat in the PhoP box in the retS promoter. Data shown are representative of 3 independent experiments and error bars represent the standard error of the mean (SEM). **, significant difference (p,0.01, ANOVA); ***, significant difference (p,0.001, ANOVA) in PAO1 compared to that of phoP::xylE. doi:10.1371/journal.pone.0023307.g007 these data indicate that Mg 2+ limitation is sensed by the PhoPQ twocomponent system, which acts as a direct negative regulator of both retS and rsmZ expression, resulting in enhanced biofilm formation.

Discussion
Previous studies of P. aeruginosa have identified multiple signaling networks that control biofilm formation. These include the sensor kinases GacS, RetS and LadS that comprise a sophisticated regulatory network involved in reciprocal regulation of biofilmassociated and classical virulence-associated genes [19,20]. However, the central question regarding the role of TCS in biofilm formation remains the identification of the actual environmental signals that are sensed.
The gene expression profiling approach described here, successfully identified Mg 2+ limitation as an environmental condition that simultaneously repressed retS expression and induced expression of the pel/psl EPS biosynthesis genes. Our observation of transcriptional effects on the EPS biosynthesis genes that correlated with decreased levels of retS expression is consistent with a previous microarray study that detected transcriptional regulation of pel and psl in a retS mutant [19]. An additional study has identified post-transcriptional regulation of psl through RsmA [26], which lies downstream of RetS in the regulatory pathway. Thus, RetS is capable of regulating EPS production by both transcriptional and post-transcriptional mechanisms.
The kinetics of gene expression in Figure 2 revealed three gene expression patterns during the early, middle and late stages of growth. The known PmrA-and PmrA/PhoP-regulated genes PA4774 and PA3553 (arnC) were induced early in growth at 1.5 and 5 h, respectively. The exclusively PhoP-controlled gene oprH was maximally induced at the middle stage of growth, 13.5 h. Maximal induction of rsmZ, rsmY, pel and psl occurred at later stages of growth following repression of retS between 7.5 and 14 h (Figure 2A). Other sensor/regulator genes tested but which did not show a significant fold change under high and limiting Mg 2+ conditions included rocR, rocA1, ladS, gacS, fleSR and pilSR. In agreement with the gene expression profile, growth under Mg 2+ limitation caused a switch to the biofilm mode with increased aggregation (Figure 3), EPS production ( Figure 5) and a hyperbiofilm phenotype (Figure 3).
Aggregation effects observed during growth in limiting Mg 2+ conditions were controlled by PhoP-mediated repression of retS and fine-tuning of rsmZ/rsmY levels. Analysis of rsmZ and rsmY expression in phoP::xylE indicated that PhoP acts differently on the expression of these two sRNA-encoding genes, directly controlling the expression of rsmZ and indirectly controlling the levels of rsmY. This is similar to previous studies where rsmY and rsmZ were differentially regulated by other TCS [55,56] and further suggests that rsmZ and rsmY are not functionally redundant but may exert different effects depending on the environmental conditions. In the wildtype PAO1, both rsmZ and rsmY are induced under Mg 2+ limitation ( Figure 2). This is likely a consequence of reduced RetS levels and activation of the GacAS pathway [19]. However the relative level of rsmY in PAO1 is significantly higher compared to that of rsmZ (45-fold) ( Figure 7A), suggesting that expression of rsmZ is limited by PhoP repression (Figure 8). It may be that direct PhoP regulation of rsmZ serves to fine-tune the relative levels of the small regulatory RNAs, in order to maintain higher relative levels of rsmY, which corresponds to increased biofilm formation. This is consistent with a recent report showing that the overexpression of rsmZ, but not rsmY, impairs P. aeruginosa biofilm formation [56]. Extracellular DNA has also been shown to be important for the early stages of biofilm formation [37] and Mg 2+ sequestration by DNA may play a role in maturation of P. aeruginosa biofilms by inducing EPS matrix production. Extracellular DNA, of both bacterial and eukaryotic origin, can be found at high concentrations (up to 20 mg/ml) in the CF lung [33,35,36]. DNA also accumulates in other environmental niches of P. aeruginosa, including soil ( [57] and aquatic environments [34], and therefore likely contributes to P. aeruginosa biofilm formation in the natural environment. Cation chelation by both DNA and EDTA can induce biofilm formation ( Figure 4B). This observation is in contrast to a previous study, which showed that EDTA is a potent biofilm disrupter in P. aeruginosa [58]. This apparent contradiction can be explained by opposite effects of subinhibitory and lethal concentrations of chelators. Since EDTA is lethal at a concentration of 1 mM by disrupting membrane integrity [31], it is not surprising that 50 mM EDTA disrupts the biofilm matrix structure and kills biofilm cells [58].
This study identifies Mg 2+ limitation as an important environmental trigger of P. aeruginosa biofilm development and increases our understanding of the potential role for extracellular DNA and cation chelation in the regulation of antibiotic resistant and aggregative biofilms. Novel approaches to block the PhoPQ and/or RetS signaling pathways, or to neutralize extracellular DNA, may be an effective and novel treatment strategy to prevent or reduce biofilm formation during infections and in the natural environment.

Strains, plasmids and media conditions
All strains and plasmids are listed in Tables 1 and 2, respectively. For details on construction of promoter fusions see Methods S1 primer Table S2. P. aeruginosa strains were routinely grown and maintained on Luria-Bertani (LB) plates or LB broth at 37uC and cultured in defined basal medium 2 (BM2) media [59] containing 0.02 mM Mg 2+ (limiting) and 2 mM Mg 2+ (high). Trimethoprim (Sigma-Aldrich) was added at 300 mg/ml for selection and 150 mg/ml for plasmid maintenance. Unless otherwise stated succinate (20 mM) was used as the carbon source. The source of DNA was fish sperm DNA-potassium salt (USB, Cleveland, OH).

Gene expression assays
Gene expression assays were carried out as previously described [31]. Gene expression values (counts per second, CPS) were normalized to cell number (optical density, OD). Gene expression profiles were grouped by hierarchical clustering using complete linkage. Analysis was performed using Cluster 3.0 and visualized using Treeview [60].

Biofilm, aggregation, congo red and calcofluor binding assays
Biofilm formation was quantified by crystal violet (CV) staining (OD 600 ) as previously described [44]. For details see supplementary information. Bright field microscopy was used to assess aggregation in mid-log cultures grown in BM2 media with 2 mM or 0.02 mM Mg 2+ . Congo red binding assays were performed as previously described [43] with minor modifications (see supplementary information). Bacterial aggregates were stained with calcofluor (EPS), 1 mM SYTO9 (live cells) and 10 mM propidium iodide (dead cells and extracellular DNA), mounted on agarose beds. All microscopy was performed using a Leica DMIREB2 inverted microscope equipped with an ORCA-ER digital camera and Openlab software (Improvision) and analysed using Adobe Photoshop.
Scanning electron microscopy P. aeruginosa peg-adhered biofilms were grown as previously described in high or limiting Mg 2+ BM2 for 48 h, fixed as previously described [61], gold coated and visualized using a XL30 environmental scanning electron microscope.

Overexpression of RetS
The coding region of retS was PCR amplified, digested with XbaI and HindIII and ligated into the pSCrhaB2 vector under the control of a rhamnose-inducible promoter (pSCrhaB2RetS) [50]. Biofilms were cultured in BM2 media with 2 mM or 0.02 mM Mg 2+ in the presence of 0-2% rhamnose.

Statistical Analysis
Statistical analysis was performed using GraphPad Prism 5 software. 2-way ANOVA was used to calculate significant differences between PAO1 and mutant strains. Figure S1 Limiting magnesium-induced biofilm formation is dependent on EPS production but independent of pili or flagella production. Attachment of PAO1 and relevant mutants to polystyrene pegs was assessed by crystal violet staining and OD 600 measurement in BM2 2 mM Mg 2+ or 0.02 mM Mg 2+ at 24 h. Bars represent the average values obtained from eight pegs and the error bars represent the standard deviation. Significant differences were observed between strains grown in BM2 2 mM Mg 2+ and BM2 0.02 mM Mg 2+ (a, p,0.05, ANOVA) and between PAO1 and mutant strains grown in BM2 0.02 mM Mg 2+ (b, p,0.05, ANOVA). PAO1w, (Wozniak laboratory strain, Ohio State University) is the parent strain of the pslH and pel/pslH double mutant. (TIF) Figure S2 EPS mutants grown in BM2 0.02 mM Mg 2+ failed to aggregate or stain with calcofluor. Bacteria were grown in BM2 0.02 mM Mg 2+ supplemented with 200 mg/ml calcofluor (blue, EPS stain). At 24 h cells were removed, stained with 1 mM syto9 (green, live cells) and visualized on agarose beds by fluorescence microscopy. Merged blue/green fluorescence images are representative of three independent experiments. (TIF) Figure S3 PhoP regulates EPS production in P. aeruginosa. (A) Quantification of EPS production using congo red binding (corrected A490) in PAO1 and phoP::xylE grown in BM2 0.02 mM Mg 2+ at 24 h. (B) Transmission electron microscopy of PAO1 and phoP::xylE. Bacteria were grown at 37uC overnight on BM2 0.02 mM Mg 2+ 0.5% agar plates. Cells were prepared and stained as described by Hyland et al. 2006 [61] and examined using a Hitachi S-7000 transmission electron microscope. (TIF) Methods S1 Additional information on strains and methods.

Supporting Information
(DOC)