Identification of FleQ from Pseudomonas aeruginosa as a c-di-GMP-responsive transcription factor

Authors


*E-mail csh5@u.washington.edu; Tel. (+1) 206 221 2848; Fax (+1) 206 543 8297.

Summary

High levels of the intracellular signalling molecule cyclic diguanylate (c-di-GMP) supress motility and activate exopolysaccharide (EPS) production in a variety of bacterial species. In many bacteria part of the effect of c-di-GMP is on gene expression, but the mechanism involved is not known for any species. We have identified the protein FleQ as a c-di-GMP-responsive transcriptional regulator in Pseudomonas aeruginosa. FleQ is known to activate expression of flagella biosynthesis genes. Here we show that it also represses transcription of genes including the pel operon involved in EPS biosynthesis, and that this repression is relieved by c-di-GMP. Our in vivo data indicate that FleQ represses pel transcription and that pel transcription is not repressed when intracellular c-di-GMP levels are high. FleN, a known antiactivator of FleQ also participates in control of pel expression. In in vitro experiments we found that FleQ binds to pel promoter DNA and that this binding is inhibited by c-di-GMP. FleQ binds radiolabelled c-di-GMP in vitro. FleQ does not have amino acid motifs that resemble previously defined c-di-GMP binding domains. Our results show that FleQ is a new type of c-di-GMP binding protein that controls the transcriptional regulation of EPS biosynthesis genes in P. aeruginosa.

Introduction

Biofilms are surface-associated communities of bacteria that are encased in exopolysaccharides (EPS; Costerton, 1999; Kolter and Greenberg 2006). Cells of the opportunistic pathogen Pseudomonas aeruginosa form biofilms in a variety of situations including on indwelling medical devices as well as in the lungs of cystic fibrosis patients (Costerton et al., 1999; 2003; Singh et al., 2000; Hoiby, 2002; Yahr and Greenberg, 2004). Biofilm infections are difficult to treat because bacteria in biofilms tend to be highly resistant to antimicrobial treatment and are able to more easily evade immune responses than planktonic cells (Stewart and Costerton, 2001; Hoiby, 2002). Because of this there is significant interest in understanding factors important for biofilm formation and maintenance. One such factor is the intracellular signalling molecule cyclic diguanylate (c-di-GMP). Originally discovered as an allosteric effector of cellulose synthesis in Gluconacetobacter xylinus (Ross et al., 1987), c-di-GMP is now recognized as playing a central role in modulating the transition between planktonic and biofilm lifestyles in a large and growing number of bacterial species, including P. aeruginosa, P. fluorescens, Salmonella typhimurium, Escherichia coli and Vibrio cholerae (Ross et al., 1991; Simm et al., 2004; Tischler and Camilli, 2004; Hickman et al., 2005; Romling et al., 2005; Kulasakara et al., 2006; Romling and Amikam, 2006; Goymer et al., 2006; Thormann et al., 2006; Cotter and Stibitz, 2007; Tamayo et al., 2007; Wolfe and Visick, 2008).

In general, increased intracellular c-di-GMP stimulates EPS production and inhibits flagella- and pilus-mediated motility, whereas low intracellular c-di-GMP promotes motility and inhibits EPS synthesis (Boles and McCarter, 2002; Aldridge et al., 2003; Garcia et al., 2004; Simm et al., 2004; Tischler and Camilli, 2004; Hickman et al., 2005; Kader et al., 2006; Lim et al., 2007). Accumulating evidence indicates that c-di-GMP has effects both on enzyme activities and protein assembly as well as on the expression of genes for EPS synthesis, flagella biogenesis and virulence in a variety of bacteria (Hickman et al., 2005; Beyhan et al., 2006; Mendez-Ortiz et al., 2006; Lee et al., 2007; Ferreira et al., 2008). Recently, a number of c-di-GMP binding proteins have been identified that affect the activities of protein complexes including alginate synthase, cellulose synthase, flagella and pili (Ryjenkov et al., 2006; Christen et al., 2007; Lee et al., 2007; Merighi et al., 2007; Pratt et al., 2007). To date no c-di-GMP binding proteins that influence gene expression have been reported.

We set out to identify the mechanism by which c-di-GMP affects gene expression in P. aeruginosa. The observation that flagella biosynthesis genes are expressed at lower levels in cells with high intracellular c-di-GMP prompted us to investigate if c-di-GMP might affect the activity of one of the transcription factors known to regulate flagella gene expression. One candidate is FleQ (PA1097), the master regulator of flagella gene expression in P. aeruginosa (Arora et al., 1997). FleQ sits at the top of a four-tiered transcription hierarchy that controls flagellar gene expression (Dasgupta et al., 2003). It is an enhancer binding protein that contains an N-terminal FleQ domain, an AAA σ54 interaction domain, and a helix–turn–helix DNA binding domain. FleQ activates expression of the two component regulatory genes fleSR as well as genes that are necessary for the assembly of the flagella export apparatus and initiation of flagella basal body-hook formation in conjunction with the alternative RNA polymerase sigma factor, σ54 (Jyot et al., 2002). A fleQ mutant is non-motile (Dasgupta et al., 2000; Dasgupta and Ramphal, 2001). A transcriptome analysis of a fleQ mutant of P. aeruginosa strain PAK showed that not only was FleQ involved in activating the transcription of flagella genes, but FleQ also negatively regulated the expression of a number of genes that are now known to specify EPS synthesis (Dasgupta et al., 2003). The expression levels of these genes are also now known to be increased in cells with high intracellular c-di-GMP (Hickman et al., 2005).

Here we show that c-di-GMP acts directly on FleQ to cause it to derepress the expression of the Pel EPS synthesis operon and other genes. We demonstrate that full-length FleQ, as well as a truncated version of FleQ that lacks the N-terminal FleQ domain, bind c-di-GMP and that the ability of FleQ to bind to the pel promoter is inhibited by c-di-GMP. Our results indicate that c-di-GMP functions to relieve transcriptional repression by FleQ of EPS genes necessary for biofilm formation in P. aeruginosa.

Results

A fleQ mutant forms wrinkly colonies

A fleQ::Tn5 mutant of P. aeruginosa strain PAO1 was non-motile, as expected, and the motility defect was complemented in trans with a wild-type copy of fleQ. In addition to having a motility defect, we found that the fleQ mutant formed colonies that were wrinkled in appearance compared with those of the wild-type on tryptone agar at room temperature (Fig. 1). The fleQ mutant colonies were slightly less wrinkly than wspF mutant colonies (Fig. 1). wspF mutations cause the diguanylate cyclase WspR to be activated, resulting in the accumulation of high intracellular concentrations of c-di-GMP relative to wild-type cells (Hickman et al., 2005; Goymer et al., 2006). The colony morphology phenotype of a fleQ mutant on tryptone agar was reverted by complementation with the fleQ gene in trans (Fig. 1). Several studies have shown that increased production of Pel or Psl EPS by P. aeruginosa results in a wrinkled colony morphology (Friedman and Kolter, 2004a,b; Hickman et al., 2005; Sakuragi and Kolter, 2007). When we generated deletions in the pelA and pslBCD genes in a fleQ mutant background the wrinkled colony morphology typical of the fleQ mutant changed to smooth (Fig. 1), indicating that Pel and Psl EPS are responsible for the wrinkled colony phenotype.

Figure 1.

A fleQ mutant has a wrinkly colony morphology. P. aeruginosa wild-type (strain PAO1), fleQ mutant, wspF mutant, fleQ mutant containing fleQ on a plasmid (pJNFleQ) or control vector (pJN105) and a fleQ pel psl mutant were streaked on tryptone agar containing Congo red, and colonies were imaged after 5 days at 25°C.

A fleQ mutation and high intracellular c-di-GMP have nonadditive effects in causing elevated EPS gene expression

We have previously shown that high intracellular c-di-GMP results in elevated expression of a number of genes in P. aeruginosa PAO1, including genes for EPS biosynthesis (Hickman et al., 2005). An earlier study by Dasgupta et al. (2003) showed that a fleQ mutant of P. aeruginosa strain PAK had increased expression of many of the same genes that we found to be elevated in high c-di-GMP cells. To confirm and extend these findings in P. aeruginosa strain PAO1, we used reverse transcriptase PCR to compare levels of the EPS biosynthesis genes pelA (PA3064) and pslA (PA2231), as well as levels of hypothetical genes PA2441 and PA4625 in a fleQ mutant to those of the wild-type and we also quantified transcript levels in a wspF mutant. In our previous work we found that expression of pelA, pslA, PA2441 and PA4625 were elevated in the wspF mutant (Hickman et al., 2005). We found that the fleQ mutant had a 20-fold increased level of pelA transcript and a 2.8-fold increased level pslA transcript compared with wild-type (Fig. 2A and B). The levels of PA2441 and PA4625 transcription were also higher in a fleQ mutant compared with wild-type (Fig. 2C and D). A wspF strain with high intracellular c-di-GMP had levels of these transcripts that were comparable to those observed with the fleQ mutant. We also generated elevated levels of intracellular c-di-GMP in wild-type and fleQ mutant cells by overexpressing the diguanylate cyclase encoded by PA1120 (Kulasakara et al., 2006). This also resulted in elevated transcript levels of pelA and pslA (Fig. 2A and B). Thus FleQ appears to negatively regulate transcription of pel, psl, PA2441 and PA4625 whereas high c-di-GMP appears to somehow stimulate transcription of these genes.

Figure 2.

FleQ is a negative regulator of genes that are also expressed at high c-di-GMP levels. Relative transcript levels for pelA (A), pslA (B), PA2441 (C), PA4625 (D) and fleR (E) as assayed by quantitative RT-PCR are shown for wild-type (PAO1), wspF mutant (high c-di-GMP), a fleQ mutant, a fleQ wspF mutant (high c-di-GMP), wild-type expressing PA1120 from a plasmid (pJN1120), a fleQ mutant expressing PA1120, fleN mutant and a fleN wspF mutant (high c-di-GMP). Strains with an asterisk have high c-di-GMP levels and display phenotypes consistent with elevated c-di-GMP. Data points are the average of three independent biological samples. Error bars represent the standard deviations between replicates.

If FleQ and c-di-GMP influence gene transcription at the same point, then the effects of a fleQ mutation and high c-di-GMP on transcription should not be additive. Consistent with this, when we quantified transcript levels of pelA, pslA, PA2441 and PA4625 in a wspF fleQ double mutant we saw no increase in expression compared with either single wpsF or fleQ mutants (Fig. 2).

FleN, a known antagonist of FleQ activation of flagella gene expression, also modulates pel and psl transcription

The activity of FleQ is known to be modulated by FleN (PA1454), which has been reported to bind to FleQ and dampen its ability to activate gene transcription (Dasgupta et al., 2000; Dasgupta and Ramphal, 2001). Strains lacking FleN have increased expression of flagella biosynthesis genes and fleN mutants are multiflagellated (Dasgupta et al., 2000). If FleN also dampens the ability of FleQ to repress transcription of EPS biosynthesis and other genes, then fleN mutants would be predicted to show decreased expression of genes that are negatively regulated by FleQ. In keeping with this, the transcript levels of pelA, pslA, PA2441 and PA4625 were 1.5-, 2.8-, 3.0- and 1.6-fold lower in a fleN mutant compared with wild-type (Fig. 2). The transcript levels of our target genes were lower in a wspF fleN double mutant than in a wspF mutant (a strain with elevated c-di-GMP) (Fig. 2). However, the levels of the target gene transcripts were higher in the wspF fleN double mutant than in the fleN single mutant, indicating that the double mutant strain responded to c-di-GMP to modulate transcription. These results are consistent with a model where FleQ acts as a repressor of pel and psl gene transcription and FleN interacts with FleQ to antagonize its repressor activity.

Elevated c-di-GMP levels have small effects on transcription of flagella biosynthesis genes

Consistent with previously published results showing that FleQ is an important positive activator of flagella gene expression (Arora et al., 1997), we observed 20-fold lower levels of fleR transcript in the fleQ mutant compared with wild-type (Fig. 2E). The fleR transcript levels were 1.6-fold lower in a wspF mutant compared with wild-type, indicating that elevated c-di-GMP has a much smaller effect on expression of flagella genes than on EPS biosynthesis genes. A wspF fleQ double mutant had the same low levels of fleR transcript as did a fleQ mutant alone. Loss of fleN resulted in an eightfold increase in fleR transcript level (Fig. 2E), consistent with previous reports (Dasgupta et al., 2000). A wspF fleN double mutant showed a 1.6-fold decrease in fleR transcript levels compared with the fleN mutant alone, indicating that the small effect of c-di-GMP on fleR transcript levels is observed in the presence or absence of the fleN gene (Fig. 2E). Similar effects of elevated c-di-GMP on transcript levels of the flhA gene were also observed (data not shown). As the effect of elevated c-di-GMP on the expression of flagella genes was small we decided to focus on the effects of c-di-GMP and FleQ on the pel operon.

fleQ or fleN mutations have minimal effects on c-di-GMP concentrations

One explanation for the effects of FleQ and FleN on expression of the pel and psl genes is that loss of FleQ or FleN affects c-di-GMP levels, which then influence gene transcription. In order to rule out this possibility we quantified c-di-GMP levels in strains lacking FleQ or FleN and compared them with the levels in wild-type cells. We measured c-di-GMP by liquid chromatography-mass spectrometry using a modification of a previously reported method (Thormann et al., 2006). We estimate from our measurements that the wild-type strain PAO1 has an average intracellular concentration of 0.7 μM c-di-GMP and the wspF mutant has an intracellular concentration of c-di-GMP of about 3.0 μM. The fleQ and fleN mutants had intracellular c-di-GMP levels that were slightly lower than those of the wild-type strain (Fig. 3). The intracellular concentration of c-di-GMP in the fleQ wspF double mutant was essentially the same as those measured in a strain with only a wspF mutation. Levels of c-di-GMP in the fleN wspF double mutant were slightly lower than in the wspF mutant alone. These small changes in c-di-GMP levels likely do not cause the large changes in gene expression that we observed in the fleQ and fleN mutants. This indicates that the effects of the fleQ mutation in causing elevated pel, psl, PA2441 and PA4625 transcription are not due to a secondary effect of this mutation causing increases in intracellular c-di-GMP.

Figure 3.

Loss of FleQ or FleN does not cause elevated c-di-GMP levels. C-di-GMP was extracted from whole cells, and levels were measured as described in the Experimental procedures. Strains are wild-type (PAO1), a fleQ mutant, a wspF mutant, a fleQ wspF mutant, a fleN mutant and a fleN wspF mutant. Data represent averages of three independent cultures. Error bars are the standard deviations between independent cultures.

FleQ binds to pel promoter DNA

Although there are relatively few examples of bacterial enhancer binding proteins like FleQ directly repressing transcription, it is not unprecedented for members this class of protein to act as transcriptional repressors (Weiss et al., 1991; North et al., 1996; Chang et al., 2007). To test the possibility that FleQ directly represses transcription, we purified FleQ and assayed its ability to bind to pelA promoter DNA in an electrophoretic mobility shift assay (EMSA). We incubated purified FleQ with a fragment of DNA spanning −284 to +30 base pairs (bp) relative to translational start of pelA. The addition of FleQ to reaction mixtures caused a shift in the mobility of the pel promoter DNA fragment (Fig. 4A). FleQ did not affect the mobility of a 135 bp fragment from the plasmid pUC19 that we included as a negative control (Fig. 4A). Although FleQ was able to bind specifically to the pel promoter region, relatively large amounts of protein (500 nM) were required to see even a slight shift in DNA mobility, suggesting that FleQ is a relatively poor DNA binding protein. This is in agreement with previous reports showing that purified FleQ bound poorly to the promoters of several genes for flagella biosynthesis (Dasgupta and Ramphal, 2001; Jyot et al., 2002).

Figure 4.

FleQ binds to the pel promoter and binding is inhibited by c-di-GMP.
A. Binding of FleQ or FleN to the pelA promoter in the absence and presence of ATP (10 μM). A fragment from the pUC19 vector is included as a negative control. The migration of unbound pelA promoter or pUC19 DNA though the gel is indicated by arrows. The concentrations of FleQ and FleN are indicated.
B. The effect of FleN and ATP on FleQ–DNA complex formation. The FleQ and FleN proteins were provided in equimolar amounts at the concentrations indicated. ATP was added at a concentration of 10 μM where indicated.
C. The effect of different nucleotides on FleQ–FleN–DNA complex formation. The nucleotide added and its concentration is indicated above each lane.
D. Binding of FleQ and FleN to the pel promoter in the absence and presence of c-di-GMP. FleQ and FleN were provided in equimolar amounts at the concentrations indicated. C-di-GMP was added where indicated. All reactions contained 10 μM ATP.

FleN binds with FleQ to pel promoter DNA in an ATP-dependent manner

Previous work showed that FleN and FleQ physically interact with each other and that addition of FleN, together with FleQ, to EMSA reactions retarded the migration of flhA promoter DNA through acrylamide gels more than did FleQ alone (Dasgupta and Ramphal, 2001). This was interpreted to mean that FleN binds to FleQ at the flhA promoter to form a higher molecular weight complex. To test possible effects of FleN on pel promoter DNA migration though gels, we purified FleN and added it to our EMSA reactions either alone, or in equimolar amounts with full-length FleQ. FleN alone had a barely detectable effect on pelA promoter migration in the EMSA assays (Fig. 4A). FleN has a predicted nucleotide-binding site and is a predicted ATPase. When we added 10 μM ATP to FleN alone or to FleQ alone in gel shift reactions, we saw no effect of ATP on the ability of either of these proteins to bind the pelA promoter (Fig. 4A). In contrast, the inclusion of FleN in reaction mixtures together with FleQ resulted in the formation of a complex with a slower mobility than that observed with FleQ alone when 10 μM ATP was present (Fig. 4B). ADP also enhanced the ability of FleQ–FleN to shift pelA promoter DNA (Fig. 4C). However, the inclusion of AMP, GTP or GMP in reactions had no effect on the ability of FleQ–FleN to shift the pelA promoter (Fig. 4C).

C-di-GMP abrogates binding of the FleQ–FleN complex to pel promoter DNA

If c-di-GMP and FleQ regulate the transcription of the pel operon through a common system, then one would predict that the addition of c-di-GMP to EMSA reaction mixtures might influence the binding of FleQ and FleN to pel promoter DNA. We found that the addition of c-di-GMP at concentrations as low as 10 μM resulted in a decrease in the ability of FleQ–FleN to bind pel promoter DNA as assayed by retarded migration of the protein–DNA complex though gels (Fig. 4D). These results indicate that FleQ–FleN constitute a c-di-GMP-responsive transcription regulatory complex. C-di-GMP addition did not influence the small shift in pelA mobility observed with 500 nM FleQ alone (data not shown).

We also tested if c-di-GMP addition affected the ability of FleQ–FleN to shift the promoter of the flagella genes fleSR. While FleQ–FleN shifted fleSR promoter DNA in the presence of ATP, we saw little to no effect of c-di-GMP on the ability of FleQ–FleN to shift fleSR promoter DNA (Fig. S1A and B). This result is consistent with our observation that elevated c-di-GMP resulted in a very small decrease of fleR transcript levels in vivo (Fig. 2E).

N-terminally truncated FleQ has enhanced DNA-binding properties

Enhancer binding proteins like FleQ often have an N-terminal domain that controls the DNA binding properties of the protein in response to a variety of stimuli including small molecules or other proteins (Keener and Kustu, 1988; Little and Dixon, 2003; Martinez-Argudo et al., 2004; Chen et al., 2005). We constructed and purified a truncated version of FleQ that lacked its 126-amino-acid N-terminal domain and tested it in EMSA reactions. We found that truncated FleQ bound to the pel promoter region with much higher affinity than the full-length protein (Fig. 5A). A concentration of 10-fold less truncated FleQ (50 nM) gave a greater mobility shift than that observed with full-length FleQ. As with full-length FleQ, addition of ATP to truncated FleQ alone did not affect the ability of this protein to bind the pelA promoter (Fig. 5A). The addition of FleN and ATP along with truncated FleQ in reaction mixtures resulted in a small increase in the shift of pelA promoter (Fig. 5A), but the effect was much less than that seen with full-length FleQ. In contrast to the situation with pelA, truncated FleQ did not bind with increased affinity to the fleSR promoter (Fig. S1C).

Figure 5.

FleQ lacking its N-terminal domain has increased DNA binding and its binding is inhibited by c-di-GMP.
A. Binding of truncated FleQ (FleQ-trunc) or FleN to the pelA promoter in the absence and presence of ATP (10 μM). The amount of protein added is indicated above the lane.
B. Binding of truncated FleQ to the pelA promoter in the presence of c-di-GMP, GTP and GMP. The nucleotide and amount of each added is indicated above each lane. All reactions contained 50 nM truncated FleQ and 1 mM ATP.
C. Binding of truncated FleQ to the PA4625 promoter in the presence of c-di-GMP, GTP and GMP. The nucleotide and amount added is indicated above each lane. All reactions contained 50 nM truncated FleQ and 1 mM ATP.

C-di-GMP inhibits DNA binding by N-terminally truncated FleQ

We found that the addition of c-di-GMP to truncated FleQ abrogated the ability of this form of the protein to retard the migration of pelA promoter DNA in EMSA reactions. C-di-GMP concentrations as low as 10 μM affected the shift, with an almost complete inhibition of DNA binding at 100 μM in both the presence and absence of 1 mM ATP (Fig. 5B and data not shown). GTP or GMP at concentrations of 10 μM had no effect on the gel mobility shift of pelA promoter DNA by truncated FleQ, and 100 μM of either of these nucleotides had a very small effect on DNA binding by truncated FleQ (Fig. 5B). We also tested the ability of truncated FleQ to bind to the promoter of PA4625, another gene whose expression is coregulated by FleQ and c-di-GMP. Truncated FleQ also specifically retarded the migration of this promoter fragment through gels and addition of c-di-GMP inhibited the binding of FleQ to PA4625 promoter DNA (Fig. 5C). These results demonstrate that c-di-GMP acts directly on FleQ to control DNA binding at multiple promoters, FleN is not required for this effect, and that the N-terminal domain is dispensable for c-di-GMP-mediated inhibition of DNA binding by FleQ.

FleQ binds c-di-GMP in vitro

We carried out direct binding assays to confirm and expand our results indicating that FleQ binds c-di-GMP. FleQ (2 μM) incubated with 20 μM [32P]-c-di-GMP in filter binding assays bound approximately 20 times more c-di-GMP than a no-protein control or reactions containing the P. aeruginosa transcriptional regulator QscR (Fig. 6A). QscR is not known to bind c-di-GMP or any other nucleotide compounds, and did not bind more c-di-GMP than the no-protein control (Chugani et al., 2001; Ledgham et al., 2003; Lee et al., 2006; Lequette et al., 2006). The diguanylate cyclase WspR bound approximately 1.9-fold more c-di-GMP than FleQ under the same conditions (Fig. 6A). WspR has a c-di-GMP binding site, the I-site, which is involved in product inhibition of enzymatic activity by c-di-GMP (De et al., 2008). Truncated FleQ bound as much c-di-GMP as full-length FleQ under these conditions, consistent with the model that the N-terminal domain has no role in c-di-GMP binding by FleQ (Fig. 6A). Both FleQ and truncated FleQ bound c-di-GMP in a concentration-dependent manner over a range of 0.5–100 μM c-di-GMP (Fig. 6B). Both FleQ and truncated FleQ displayed binding characteristics consistent with first order kinetics. Half-maximal binding of c-di-GMP by FleQ occurred at a concentration of approximately 15–25 μM (Fig. 6B).

Figure 6.

FleQ binds c-di-GMP in vitro.
A. Binding of radiolabelled c-di-GMP by FleQ, truncated FleQ, WspR (positive control), QscR (negative control) and no-protein control. All reactions contained 20 μM [32P]-c-di-GMP. Data are presented as cpm retained on the nitrocellulose filter.
B. Concentration-dependent binding of [32P]-c-di-GMP by FleQ (squares) and truncated FleQ (triangles). All data are the average of at least three independent binding reactions. Error bars represent the standard deviations between replicates.

The addition of unlabelled c-di-GMP at 10-fold excess to reaction mixtures inhibited the binding of [32P]-c-di-GMP to FleQ or truncated FleQ, indicating that FleQ interacts specifically with this dinucleotide (Table 1). By contrast, addition of ATP, GMP or GTP at 10-fold excess concentration had little effect on the binding of [32P]-c-di-GMP to FleQ or truncated FleQ (Table 1). ATP or GTP partially competed with c-di-GMP for binding to full-length FleQ when provided at 100-fold excess relative to radiolabelled c-di-GMP, however, these two nucleotides had a much smaller effect on c-di-GMP binding by truncated FleQ (Table 1). GMP at 100-fold excess had no effect on c-di-GMP binding by full-length FleQ, and a relatively small effect on binding by truncated FleQ (Table 1).

Table 1.  Relative c-di-GMP binding by FleQ in the presence of competing nucleotides.
CompetitoraFleQ full-lengthbFleQ truncatedc
  • a. 

    Nonradioactive nucleotides added to binding reactions at 10-fold (10×) and 100-fold (100×) excess concentrations over [32P]-c-di-GMP (20 μM). Results are the average of at least three independent binding reactions.

  • b. 

    Relative binding of [32P]-c-di-GMP by full-length FleQ in the absence or presence of competing nucleotides. Numbers indicate per cent binding relative to the no-competitor control, which is set to 100%. Standard deviations between replicates are indicated in parentheses.

  • c. 

    Relative binding of [32P]-c-di-GMP by truncated FleQ in the absence or presence of competing nucleotides. Numbers indicate per cent binding relative to the no-competitor control, which is set to 100%. Standard deviations between replicates are indicated in parentheses.

None100 (35)100 (26)
10× c-di-GMP23 (9)20 (4)
10× ATP97 (12)83 (9)
10× GMP109 (14)112 (33)
10× GTP87 (17)111 (27)
100× ATP47 (17)77 (20)
100× GMP104 (3)84 (15)
100× GTP42 (10)76 (23)

Discussion

Here we have presented three lines of evidence that c-di-GMP binds to the transcriptional regulator FleQ to cause it to derepress the expression of pel and other EPS genes necessary for biofilm formation. First, a fleQ mutation and high c-di-GMP each caused an increase in the relative levels of pelA, pslA, PA2441 and PA4625 transcripts, and the effects of these two conditions were not additive (Fig. 2). This suggests that FleQ and c-di-GMP act through a common mechanism. Second, electromobility shifts of pelA promoter DNA in the presence of FleQ and FleN, or FleQ lacking its N-terminal domain, were abrogated by addition of c-di-GMP. This suggests that c-di-GMP prevents binding of FleQ to pelA promoter DNA (Figs 4 and 5). The effect of c-di-GMP on the binding of truncated FleQ to the promoter region of PA4625 was similar to that seen at the pel promoter. Finally, radiolabelled c-di-GMP bound specifically to purified FleQ protein (Fig. 6 and Table 1). Also, truncated FleQ bound c-di-GMP as well as full-length protein, indicating that the N-terminal domain is dispensable for c-di-GMP binding (Fig. 6 and Table 1).

Our data on FleQ-mediated regulation of pelA expression are consistent with the model shown in Fig. 7. We propose that in the absence of FleN or ATP, FleQ maximally represses pelA expression (Fig. 7A). When FleN and ATP (or ADP) are present they partially inhibit the ability of FleQ to repress transcription of pelA (Fig. 7B). FleN is a predicted ATPase with 35% sequence identity to the cell division inhibitor MinD and 25% identify to nitrogenase reductase NifH. By analogy with the known modes of action of these proteins (Schindelin et al., 1997; Lutkenhaus, 2007), we hypothesize that ATP/ADP promotes multimerization of FleN and binding of FleN to FleQ, leading to inhibition of FleQ activity. C-di-GMP stimulates the derepression of pel transcription. One possible mechanism for this is that binding of c-di-GMP to FleQ stimulates a change in protein complex conformation that results in the dissociation of FleQ from promoter DNA (Fig. 7C).

Figure 7.

Model for the regulation of gene expression by FleQ, FleN and c-di-GMP.
A. FleQ in the absence of FleN or c-di-GMP maximally represses pel transcription.
B. The situation in wild-type cells. FleQ binding at the pelA promoter is reduced by FleN and ATP/ADP, resulting in less pel repression than the situation in (A).
C. C-di-GMP binds to FleQ to cause it to dissociate from DNA, thereby causing derepression of transcription from the pel promoter.

Our data indicate that FleQ has different effects at promoters it represses (like pel) as compared with promoters it activates (like fleSR). While removal of the N-terminal domain of FleQ enhanced DNA binding at pel and PA4625, it had little effect on binding to the fleSR promoter. Also, elevated c-di-GMP had a much greater effect on promoters that FleQ represses (pel) compared with those that it activates (fleSR). The mechanism of transcriptional regulation by FleQ at activated and repressed promoters is likely to differ in significant ways because transcriptional activation by FleQ is dependent on the alternative RNA polymerase sigma factor σ54, whereas published array data indicate that transcriptional repression of pel, psl, PA4625 and other genes by FleQ does not depend on RNA polymerase σ54 (Dasgupta et al., 2003). These differences in sigma factor participation may account for why c-di-GMP has such a small effect on activation of σ54-dependent genes including flagella genes and much larger effects, as much as 20-fold effects, on derepressing expression of genes that are transcriptionally repressed by FleQ in a σ54-independent manner. In addition, it seems plausible that the number and position of FleQ binding sites may be different between repressed and activated promoters. Only a few FleQ binding sites have been identified to date, all of which are upstream of flagella genes that FleQ activates (Jyot et al., 2002). No consensus DNA sequence for FleQ binding was discernable from these previous studies. Therefore, it is difficult at this point to use bioinformatics to predict possible FleQ binding sites in the promoter regions of repressed genes.

FleQ homologues transcriptionally regulate flagella gene expression in other polarly flagellated γ-proteobacteria including other Pseudomonas species, Vibrio and Legionella species (McCarter, 2006), and it is possible these homolgous proteins also regulate EPS gene transcription in response to c-di-GMP. Recently, P. fluorescens FleQ was reported to negatively regulate expression of the wss operon necessary for cellulose production. Interestingly, expression of these genes is also activated by high c-di-GMP levels (Giddens et al., 2007).

FleQ represents the first member of a new class of c-di-GMP binding proteins that are transcriptional regulators. The fact that FleQ lacking its N-terminal domain binds c-di-GMP as well as the full-length protein indicates that c-di-GMP binds somewhere outside the N-terminal domain. The in vitro system that we have established for examining binding of FleQ to pelA promoter DNA provides an opportunity to examine in detail the mechanism by which c-di-GMP controls the biochemical activity of a transcriptional regulator. The FleQ protein does not contain any regions that have a predicted secondary structure that resembles those of known c-di-GMP binding regions, including the PelD protein from P. aeruginosa, PilZ domains or the I-sites of diguanylate cyclases (Chan et al., 2004; Amikam and Galperin, 2006; Wassmann et al., 2007; De et al., 2008). A detailed characterization of FleQ is expected to reveal alternative determinants of c-di-GMP binding.

Experimental procedures

Strains and growth conditions

Strains and plasmids used in this study are listed in Table 2. Primer sequences are available upon request. P. aeruginosa and E. coli strains were routinely cultivated on LB medium at 37°C unless otherwise indicated. Colony morphology was visualized on tryptone-Congo red agar plates (10 g tryptone, 40 mg Congo red, 10 mg comassie Brilliant blue R-250 per litre) after 5 days growth at 25°C (Friedman and Kolter, 2004b). Antibiotics for E. coli and P. aeruginosa were added where appropriate. These were 100 μg ml−1 ampicillin for E. coli and 50 μg ml−1 gentamycin for P. aeruginosa.

Table 2.  Strains and plasmids.
Strain or plasmidRelevant phenotype or genotypeSource or reference
P. aeruginosa
 PAO1Wild-type strain; twitching motility+Jacobs et al. (2003)
 PAO1100PAO1 derivative, in-frame deletion of wspFHickman et al. (2005)
 fleQ::Tn5PAO1 derivative, fleQ::Tn5; TcR.Jacobs et al. (2003)
 fleN::Tn5PAO1 derivative, fleN::Tn5; TcR.Jacobs et al. (2003)
 PAO1110fleQ::Tn5 with in-frame deletion of wspFThis study
 PAO1111fleN::Tn5 with in-frame deletion of wspFThis study
 PAO1114fleQ::Tn5 with deletion of pelA and pslBCDThis study
E. coli
 DH5αsupE44ΔlacU169 (φ80 lacZΔM15) hsdR178 recA1 endA1 gyrA96 thi-1 relA1Gibco-BRL
 S17-1C600::RP-4 2-(Tc::Mu) (Kn::Tn7) thi pro hsdR hsdM+recASimon et al. (1983)
 ER2566F- lamda-fhuA2[lon]ompT lacZ:: T7 gene1 gal sulA11Δ(mcrC-mrr) 114::IS10 R(mcr-73::miniTn 10 – TetS)2 R(zgb-210::Tn10) (TetS) endA1 [dcm]NEB
Plasmids
 pTYB12Apr; vector used for N-terminal Intein–chitin-binding domain fusionsNEB
 pFleQ-Int1Apr; NdeI–EcoRI fragment containing fleQ cloned into NdeI–EcoRI digested pTYB12This study
 pFleN-Int1Apr; NdeI–EcoRI fragment containing fleN cloned into NdeI–EcoRI digested pTYB12This study
 pFleQ-trunc-Int1Apr; NdeI–EcoRI fragment containing a truncated fleQ cloned into NdeI–EcoRI digested pTYB12This study
 pJN105Gmr; araC-PBAD cassette cloned in pBBR1MCS-5Newman and Fuqua (1999)
 pJNFleQGmr; fleQ cloned as a 1.7 kb EcoRI–SmaI fragment into EcoRI–SmaI digested pJN105This study
 pJN1120Gmr; PA1120 cloned as a 1.3 kb NheI–XbaI fragment into NheI–XbaI digested pJN105This study
 pMPSL-KO1Apr, Gmr; Pseudomonas suicide vector containing deletion construct for pslBCD genesM. Starkey, unpublished
 pMPELAApr, Gmr; Pseudomonas suicide vector containing deletion construct for pelA geneM. Starkey, unpublished

Plasmid and strain construction

All primer sequences are available upon request. Plasmid pJNFleQ was constructed by ligating an EcoRI–SmaI fragment from plasmid pAT6 containing the FleQ gene into EcoRI–SmaI digested pJN105 (Tart et al., 2005). Plasmid pJN1120 was constructed by amplification of PA1120 with primers homologous to the 5′ and 3′ ends of PA1120 containing restriction sites for NheI and XbaI respectively. The PCR product was cloned into NheI–XbaI digested pJN105, and the resulting plasmid, pJN1120, was verified by sequencing. Plasmids were electroporated into wild-type PAO1 and the fleQ mutant using established protocols (Choi et al., 2006). Deletions of pelA and pslBCD in a fleQ mutant were constructed using standard procedures for allelic exchange in P. aeruginosa. Two plasmids, pMPSL-KO1 and pMPELA, containing the deletion constructs (M. Starkey and M. Parsek, unpublished) were mated into a fleQ mutant and strains were selected on Pseudomonas isolation agar containing 50 μg ml−1 gentamycin. Double recombinant mutants were selected on LB plates containing 5% sucrose. Mutants were confirmed by PCR.

RNA isolation and real-time quantitative PCR

Pseudomonas aeruginosa strains were grown in 5 ml cultures in test tubes overnight then subcultured twice into 125 ml baffled flasks to an OD600 of 0.02. After the second subculture and subsequent growth, cells were harvested at an OD600 of 0.4–0.6 for RNA isolation. RNA was isolated and cDNA synthesized using previously published procedures (Schuster et al., 2003). Primers for RT-PCR were designed using the program PrimerExpress (Applied Biosystems). Reactions were performed using SybrGreen master mix (Applied Biosystems) with 1 ng cDNA as template and 200 nM of each primer in a final reaction volume of 25 μl. Cycling parameters were 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. A final dissociation curve of all amplified products was performed as a quality control check. Standard curves for quantification of transcripts were performed using dilutions of P. aeruginosa chromosomal DNA from 1 × 10−4 ng to 10 ng (Lequette et al., 2006). Transcript levels of all genes tested were normalized to transcript levels of the ampR gene (Lequette et al., 2006). Data presented are averages of three independent cultures grown on different days. Error bars represent the standard deviation between samples.

Quantification of c-di-GMP levels by LC-MS

Cells were grown overnight in VBMM (Vogel-Bonner minimal medium) and subcultured the next day to an OD600 of 0.02 in 125 ml baffled shake flasks containing 15 ml of VBMM. VBMM, a defined medium, was used for c-di-GMP quantification because an unknown component of LB medium coeluted with c-di-GMP and masked the c-di-GMP signal during LC-MS analysis. When cells reached an OD600 of 0.4–0.6, 1 ml was removed, centrifuged for 1 min at 15 000 g, and the supernatant was removed. The cell pellet was re-suspended in 100 μl 0.6 M perchloric acid. Samples were incubated on ice for 30 min, and cell debris was removed by microcentrifugation at 4°C for 5 min at 15 000 g. Supernatants (100 μl) were removed and samples were neutralized by addition of 20 μl 2.5 M KHCO3. The resulting precipitate was removed by centrifugation at 4°C for 5 min at 15 000 g. These neutralized supernatants were stored at −80°C until analysed by LC-MS. For protein determinations, three 1 ml samples of each culture were removed and centrifuged to pellet cells. Cell pellets were re-suspended in 100 μl 1 M NaOH and boiled in a water bath for 10 min. Samples were then stored on ice until protein was assayed. Standard Bradford protein assays were carried on all samples in duplicate using the Bio-Rad Protein assay reagent (Bio-Rad, Hercules, CA). Bovine serum albumin (BSA) containing 1 M NaOH was used as a standard.

Samples were analysed at the University of Washington Mass Spectrometry Center using an HPLC/MS/MS system. Protocols for c-di-GMP separations and analysis by mass spectrometry were based on a previously published method (Thormann et al., 2006). Separation was achieved with an Acuity UPLC (Waters; Milford, MA) using a 2.1 × 50 mm Synergi Hydro RP column (Phenomenex; Torrance, CA).A gradient system was used starting from 98% aqueous (10 mM ammonium formate, pH 4.0) and 2% organic (acetonitrile). The aqueous went to 10% at 6.0 min, 95% at 6.5 min, 10% at 7.0 min and 98% at 7.5 min. Temperature was uncontrolled, flow was 0.3 ml min−1 and cycle time was 10 min. The extra step from high-to-low aqueous and back again between 6.5 and 7.5 min served to eliminate carry-over between injections. The c-di-GMP was detected by MS/MS multiple reaction monitoring using a Premiere XL triple quadrapole mass spectrometer (MicroMass; Milford, MA) in positive electrospray ionization (API-ES+). The m/z 691>152 transition was used for quantification; 691>248 and 691>540 were monitored as confirmatory signals. The collision energies were 30, 24 and 24 eV respectively; the cone voltages were 40 V in all cases. For a standard curve, 50, 100, 250, 500 and 1000 fmol pure c-di-GMP (Axxora, San Diego, CA) were analysed by the above method. C-di-GMP levels are normalized to total protein per ml of culture. Data represent averages of three independent cultures. Error bars are the standard deviation between replicate cultures. Intracellular concentrations of c-di-GMP were estimated from these data assuming that 1 mg of protein corresponds to 6.7 × 109 cells and that each P. aeruginosa cell has a volume of 1 × 10−15 l.

Protein purification

FleQ, FleQ lacking its N-terminal domain (FleQ-trunc) and FleN were purified using the IMPACT system from NEB (Beverly, MA). The fleQ, fleQ-trunc and fleN genes were each cloned into pTYB12 as NdeI–EcoRI fragments generating plasmids pFleQ-Int1, pFleQ-trunc-Int1 and pFleN-Int1. This generated N-terminal fusions of the Intein-Chitin binding domain (Intein–CBD) to FleQ, FleQ-trunc and FleN respectively. This resulted in the addition of Ala, Gly and His residues to the N-terminus after cleavage of the Intein–CBD tag. Plasmids were transformed into E. coli ER2566 for expression of fusion proteins. For overexpression of the fusion proteins, E. coli carrying either pFleQ-Int1, pFleQ-trunc-Int1 or pFleN-Int1 were subcultured from overnights into 1 l of fresh LB in 3 l Fernbach flasks. Cultures were grown to an OD600 of 0.4 and then shifted to 15°C. After 30 min at 15°C, 0.3 mM IPTG was added to induce expression of fusion proteins, and cultures were grown for 16 h. Cells were harvested by centrifugation at 5000 g, pellets re-suspended in column buffer with Triton X-100 [20 mM Tris-Cl pH 8.0, 250 mM KCl, 0.1 mM EDTA, 0.1% (v/v) Triton X-100], and cells were lysed by sonication. Lysates were clarified by centrifugation at 15 000 g for 30 min. The solubility of the expressed proteins was checked by analysing soluble and insoluble fractions by SDS-PAGE. Expression at 15°C prevented the insolubility of overexpressed FleN that had been previously described (Dasgupta and Ramphal, 2001).

Intein–CBD fusions were purified as previously described with the following modifications (Hickman et al., 2005). Proteins were loaded in column buffer plus Triton X-100 onto a 5 ml chitin bead column. Columns were washed with 50 ml column buffer plus Triton, then washed with 50 ml column buffer lacking Triton X-100. Cleavage of the Intein–CBD was induced by addition of column buffer containing 50 mM DTT or β-mercaptoethanol and incubating overnight at room temperature. When necessary, protein was concentrated using Amicon ultrafiltration devices (Millipore, Billerica, MA.) Proteins were dialysed into column buffer containing 50% glycerol and stored at −20°C. Both FleQ and FleN were stable under these storage conditions for up to 2 months.

In vitro DNA binding analysis

The pel promoter region spanning −284 to +30 bp relative to translational start of pelA was PCR amplified. As a negative control a 135 bp fragment of pUC19 vector was also PCR amplified. These DNA fragments were end-labelled with T4 polynucleotide kinase and [γ-32P]-ATP. Equal amounts of each of these labelled DNA fragments (4 fmol, final total probe concentration 100 pM) were added to binding reactions with varying amounts of FleQ or FleQ lacking its N-terminal domain (FleQ-trunc) in binding buffer (10 mM Tris, pH 7.8, 8 mM magnesium acetate, 50 mM KCl, 5% glycerol, 250 ng μl−1 BSA, 20 μl total reaction volume). FleQ was incubated with DNA for 30 min on ice. Reactions carried out with FleN were performed as above with the addition of equimolar amounts of FleN and FleQ. Where indicated, 10 μM ATP or other nucleotides were also added to the reactions and incubated with the proteins for 30 min on ice prior to addition of DNA. Reactions containing c-di-GMP were performed as described above except that c-di-GMP was incubated with proteins for 30 min before addition of DNA. All reactions mixtures were loaded onto a 5% acrylamide gel containing 10 mM Tris-Cl pH 8.0, 400 mM glycine, 5 mM EDTA and electrophoresed at 70 V at room temperature for 1–1.5 h. Gels were dried and exposed to a phosphorimaging screen overnight, visualized on a Storm phosphorimager (GE Healthcare, Pistcataway, NJ), and images analysed using ImageQuant software (GE Healthcare).

C-di-GMP binding assays

[32P]-c-di-GMP was generated using purified WspR protein using reaction conditions previously described (Hickman et al., 2005; Merighi et al., 2007) Reactions (100 μl total volume) contained 10 μM WspR protein, 250 μCi [α-32P]-GTP and 2.5 mM acetyl phosphate in reaction buffer (75 mM Tris-Cl, pH 7.8, 250 mM NaCl, 25 mM KCl, 10 mM MgCl2) and were incubated at 37°C for 2 h. Antarctic phosphatase (NEB, Beverly, MA) was added and reactions were incubated an additional 30 min to remove any residual [α-32P]-GTP. Reactions were then heated to 95°C for 5 min and spun at 15 000 g for 5 min to precipitate protein and supernatants were spun through a Millipore Ultrafree filter device (5000 MWCO) at 5000 g for 30 min. The filtrate was loaded onto a Sep-Pak Light C18 cartridge (Waters, Milford, MA), washed with 0.5× reaction buffer containing 1% MeOH, and [32P]-c-di-GMP was eluted with 50% MeOH. Fractions from the Sep-Pak cartridge were dried using a Speed-Vac and re-suspended in 50 μl water. Samples were run on PEI-cellulose TLC plates using 1.5 M KH2PO4 pH 3.5 to determine that pure c-di-GMP was obtained (Hickman et al., 2005).

The binding of c-di-GMP by proteins was assayed using a filter binding technique. Protein (2 μM FleQ, FleQ truncated, WspR or QscR) was incubated with 20 μM [32P]-c-di-GMP (0.2 Ci mmol−1) in binding buffer (40 mM Tris, pH 7.8, 10 mM magnesium acetate, 50 mM KCl). Reactions were incubated on ice for 25 min then filtered through a 0.2 μm nitrocellulose membrane (Whatman, Dassel, Germany), using a slot blot apparatus (PR 600 SlotBlot, Hoeffer Scientific). Samples were washed with 1 ml cold reaction buffer, the membrane was removed, dried, and then individual wells were cut out and radioactivity was determined using a scintillation counter. Purified QscR from P. aeruginosa was used as a negative control and showed no binding of c-di-GMP above the no-protein control. For competition experiments with ATP, GTP, GMP or c-di-GMP, 2 μM FleQ or FleQ truncated was incubated with 20 μM [32P]-c-di-GMP for 12.5 min, then competing substrate was added and the reactions incubated another 12.5 min and processed as above. For binding curves, 2 μM FleQ or FleQ truncated was incubated with varying amounts of [32P]-c-di-GMP and reactions were processed as above.

Acknowledgements

Public Health Service Grant GM56665 from the National Institute of General Medical Sciences supported this work. J.W.H. was supported by a Cystic Fibrosis Foundation Postdoctoral Fellowship (R565-CR02). We would like to thank Thomas F. Kalhorn and the University of Washington Mass Spectrometry Center for their invaluable assistance with c-di-GMP quantification. Plasmids pMPSL-KO1 and pMPELA were generous gifts from Melissa Starkey and M. Parsek. Purified QscR protein was a generous gift from Ken-Ichi Oinuma and E. P. Greenberg. We would also like to thank Amy L. Schaefer and Breck A. Duerkop for reading the manuscript and for insightful discussions.

Ancillary