By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
The Escherichia coli NsrR protein is a nitric oxide-sensitive repressor of transcription. The NsrR-binding site is predicted to comprise two copies of an 11 bp motif arranged as an inverted repeat with 1 bp spacing. By mutagenesis we confirmed that both 11 bp motifs are required for maximal NsrR repression of the ytfE promoter. We used chromatin immunoprecipitation and microarray analysis (ChIP-chip) to show that NsrR binds to 62 sites close to the 5′ ends of genes. Analysis of the ChIP-chip data suggested that a single 11 bp motif (with the consensus sequence AANATGCATTT) can function as an NsrR-binding site in vivo. NsrR binds to sites in the promoter regions of the fliAZY, fliLMNOPQR and mqsR-ygiT transcription units, which encode proteins involved in motility and biofilm development. Reporter fusion assays confirmed that NsrR negatively regulates the fliA and fliL promoters. A mutation in the predicted 11 bp NsrR-binding site in the fliA promoter impaired repression by NsrR and prevented detectable binding in vivo. Assays on soft-agar confirmed that NsrR is a negative regulator of motility in E. coli K12 and in a uropathogenic strain; surface attachment assays revealed decreased levels of attached growth in the absence of NsrR.
Nitric oxide (NO) is synthesized by the inducible nitric oxide synthase in phagocytic cells and is an important component of the innate immune response to infection (Fang, 2004). NO is also made by some bacteria, either as a by-product of nitrite reduction to ammonia, or as an intermediate of denitrification (Watmough et al., 1999). Thus, pathogenic bacteria can potentially be exposed both to endogenously generated NO and to the NO produced by host cells. In Escherichia coli, the transcriptional regulators NorR and NsrR mediate adaptive responses to NO by controlling the expression of genes encoding enzymes that reduce or oxidize NO to less toxic species (Mukhopadhyay et al., 2004; Bodenmiller and Spiro, 2006; Spiro, 2007). The key NO detoxifying enzymes are the flavohaemoglobin (encoded by the hmp gene) and the flavorubredoxin (encoded by norVW), which are regulated by NsrR and NorR respectively (Hutchings et al., 2002; Gardner et al., 2003; Bodenmiller and Spiro, 2006), and the respiratory nitrite reductase, Nrf, which reduces both nitrite and NO to ammonia (Poock et al., 2002). The σ54-dependent transcriptional activator NorR is stimulated by the formation of a nitrosyl species at a mono-nuclear iron site in its signalling domain (D′Autréaux et al., 2005). In the case of NsrR, the binding site for NO is likely to be an [Fe-S] cluster (Isabella et al., 2008; Tucker et al., 2008; Yukl et al., 2008).
We initially identified NsrR as a repressor of the ytfE, hmp and ygbA genes (Bodenmiller and Spiro, 2006). The product of the ytfE gene is a di-iron protein, which has been implicated in the repair of damaged [Fe-S] clusters (Justino et al., 2007). A YtfE homologue (NorA) from Ralstonia eutropha has been shown to bind NO, and has been suggested to function to lower the cytoplasmic NO concentration (Strube et al., 2007). The ygbA gene is of unknown function, while hmp encodes a well-characterized NO detoxification system (Poole and Hughes, 2000). Subsequently described targets for NsrR regulation include the hcp-hcr and yeaR-yoaG genes, and the nrf operon that encodes Nrf (Filenko et al., 2007; Lin et al., 2007). Known and predicted targets for NsrR regulation have in their promoter regions an 11 bp inverted repeat with a spacing of 1 bp (Rodionov et al., 2005; Bodenmiller and Spiro, 2006; Lin et al., 2007). In this paper we confirm that this sequence is required for NsrR-mediated repression of the ytfE promoter, but also present evidence to suggest that a single copy of the 11 bp motif may be sufficient for NsrR binding.
The full extent of the NsrR regulon of E. coli has been assessed computationally (Rodionov et al., 2005), and by analysis of the transcriptome of a strain in which NsrR was depleted by the presence of multiple copies of a cloned NsrR-binding site (Filenko et al., 2007). As a complementary approach to identifying genes that might be regulated by NsrR, we describe in this paper the use of chromatin immunoprecipitation and microarray analysis (ChIP-chip) to locate NsrR-binding sites in the E. coli genome. Computational analysis of newly identified targets revealed additional insights into the requirements for a functional NsrR target site. Unexpectedly, we found NsrR-binding sites associated with the promoter regions of three transcription units containing genes with well-established or suspected roles in motility and/or biofilm development. We confirmed that two of the three promoters are subject to regulation by NsrR and NO, and showed that NsrR is a negative regulator of motility in E. coli.
Isolation of mutations in the NsrR-binding site
We have previously shown that NsrR regulates the ytfE, hmp and ygbA promoters, and have predicted the sequences of the NsrR-binding sites in these promoters (Bodenmiller and Spiro, 2006). There is also a predicted NsrR-binding site in the promoter region of the hcp-hcr genes (Rodionov et al., 2005), which encode the hybrid cluster (prismane) protein and an associated redox enzyme, and NsrR is a repressor of hcp-hcr transcription (Filenko et al., 2007). We analysed the 5′ non-coding regions of these four transcription units for the occurrence of candidate cis-acting regulatory sequences, using the MEME algorithm (Bailey et al., 2006). MEME detected the previously predicted NsrR-binding sites, and further suggested the presence of a second NsrR-binding site in the ygbA, hcp and hmp promoters. In each case, the primary (previously described) NsrR sites overlap the −10 and/or transcription start site (Fig. 1A), while the secondary sites are further upstream. The seven predicted sites (Fig. 1A) give rise to the sequence logo depicted in Fig. 1B, which is very similar to the logo previously generated for NsrR-binding sites in a group of Enteric bacteria (Rodionov et al., 2005), with the addition of two AT base pairs which are present at the 5′ ends of all seven E. coli sites (Fig. 1). A similar sequence can also be found in the yeaR promoter, which is regulated directly by NsrR (Lin et al., 2007).
The presence in the cell of multiple copies of the putative NsrR-binding site from the ytfE promoter causes de-repression of a ytfE–lacZ reporter fusion, by a repressor titration effect (Bodenmiller and Spiro, 2006). Deletion of a single AT base pair at the centre of the NsrR-binding site eliminated repressor titration (Bodenmiller and Spiro, 2006). We selected the NsrR-binding site in ytfE as a model for further study, and sought to isolate additional mutations in this sequence that impair NsrR binding. We subjected the 205 bp ytfE promoter fragment to random mutagenesis, and screened on lactose indicator media for clones with lower activities in the repressor titration assay. By picking Lac- colonies, we repeatedly isolated the same 1 bp deletion that we had previously made by site directed mutagenesis. We assume that the run of four AT base pairs in the NsrR-binding site in ytfE is prone to deletion by slipped-strand mispairing during the mutagenesis reaction. This mutation eliminates activity in the repressor titration assay (Table 1), and ytfE promoter activity, presumably because the deletion is in the −10 sequence (we were unable to isolate Lac+ fusion phages to assay the activity of this mutant ytfE promoter). By screening for a partial phenotype in the repressor titration assay (pale blue colonies), we isolated two substitutions at positions 2 and 6 in the NsrR-binding site (Fig. 1B). Both caused a defect in the repressor titration assay, and de-repression of the ytfE promoter in both the absence and presence of nitrite (Table 1). The two mutations isolated by random mutagenesis are at positions that are almost completely conserved in known and predicted NsrR-binding sites, and introduce nucleotides that never occur in known and predicted sites (Fig. 1; Table 2), with the single exception of a C at position 6 in the MEME-predicted second site in hmp (Fig. 1A).
Table 1. Characterization of mutations in the NsrR-binding site in the ytfE promoter.
Wild type and mutant ytfE promoter sequences cloned in pSTBlue-1 were transformed into a strain with a ytfE-lacZ reporter fusion. Activities reflect de-repression of the ytfE promoter by repressor titration.
Wild type and mutant promoters were fused to lacZ and integrated in the chromosome. Cultures were grown anaerobically in a minimal medium. For treatment with nitrite, cultures were grown to mid-exponential phase, supplemented with 5 mM nitrite, then assayed 60–90 min later.
Wild type ytfE promoter assayed in a ΔnsrR background.
The deletion mutation eliminates ytfE promoter activity, presumably because it is located in the −10 sequence.
The co-ordinate is shown for the 5′ end of the array feature that gave rise to the largest fluorescence ratio, and refers to a non-current version of the E. coli genome (http://genolist.pasteur.fr/Colibri/).
The enrichment ratio is expressed as the peak mean fluorescence ratio from the triplicate datasets.
Genes flanking the signal and their transcriptional orientations are shown.
The distance from the predicted NsrR-binding site to the start codon is shown. Where the binding site is between divergently transcribed genes, the distance to the closer start codon (indicated in parentheses) is shown.
Additional mutations were introduced into the NsrR-binding site in ytfE by site-directed mutagenesis. Mutations corresponding to those previously isolated at positions 2 and 6 were made in the right half of the inverted repeat, at positions 22 and 18 respectively (Fig. 1B). We also made mutations at the symmetry-related positions 5 and 19, and made three double mutants in which symmetry-related single mutations were combined (Fig. 1B, Table 1). Repressor titration assays showed that all single mutations significantly reduced the ability of the sites to titrate NsrR, and that single mutations at symmetry-related positions had similar effects (mutations 2 and 22 being the most severe). In all three cases, symmetry-related double mutations had larger effects than either of the corresponding single mutations (Table 1).
Mutant ytfE sequences made by site-directed mutagenesis were also fused to lacZ for assays of promoter activity. A similar pattern was found to that seen with the repressor titration assays, in that single mutations caused some de-repression of the promoter, with mutations at symmetry-related positions having similar effects (Table 1). Double mutations caused greater de-repression than the corresponding single mutations, showing that both halves of the site are required for optimal repression of the ytfE promoter. Mutations at positions 2 and 22 again caused the most severe phenotypes. Induction by nitrite increased the activities of promoters with mutations at positions 2 and 22 between 2.5- and 3.6-fold, but between 8.8- and 14.7-fold for all of the other promoters (Table 1). The reduced induction ratio is due to particularly high activities for the 2, 22 and 2 + 22 mutants in cultures grown anaerobically in the absence of nitrite (Table 1). The reason why mutations at these positions have a disproportionate effect on promoter activity under non-inducing conditions is not known. All of the promoters remain somewhat inducible by nitrite, demonstrating that none of the mutations completely eliminates NsrR binding.
We tested binding of NsrR to a selection of the mutant ytfE promoters in vivo by chromatin immunoprecipitation (ChIP). For these experiments, the NsrR protein was modified by addition of a C-terminal 3XFlag tag; the modified protein was expressed from a single-copy gene at the nsrR locus on the chromosome (Efromovich et al., 2008). Cultures were transformed with pSTBlue-1 derivatives containing wild type and mutant ytfE promoters (the same plasmids used for the repressor titration assays). After ChIP, the immunoprecipitated DNAs were used as templates for PCRs using vector-specific primers designed to amplify ytfE sequences. Amplification conditions were optimized to allow detection of NsrR binding to the wild-type site in ytfE (Fig. 1D, compare lanes 1 and 2). Under these conditions, binding to the deletion mutant was barely detectable, and all of the single and double point mutations tested showed significantly reduced binding (Fig. 1D). Quantitative conclusions cannot be drawn from this experiment, but it nevertheless shows that all of the mutations reduce binding in vivo sufficiently to severely impair detection by ChIP, under conditions that allow detection of the wild type interaction.
Our results with the ytfE promoter are consistent with previous suggestions that the NsrR consensus-binding site in E. coli consists of two copies of the sequence 5′-AAGATGCYTTT-3′ arranged as an inverted repeat separated by 1 bp (Rodionov et al., 2005; Bodenmiller and Spiro, 2006; Lin et al., 2007). Results presented later in this paper suggest that a single 11 bp motif can also function as an NsrR-binding site.
Genome-wide search for NsrR-binding sites
We next used ChIP-chip to identify NsrR-binding sites in the genome of E. coli K12 strain MG1655. Cultures expressing the 3XFlag-tagged NsrR were grown anaerobically in the presence and absence of nitrate. Under anaerobic growth conditions, nitrate provides a source of endogenously generated NO and causes de-repression of NsrR targets (Bodenmiller and Spiro, 2006). After ChIP, the precipitated DNAs were labelled with Cy5 and Cy3 and hybridized together to a high-density microarray (from Oxford Gene Technology). Peaks in the fluorescence ratio therefore identify regions of the chromosome that are bound by NsrR, with the degree of occupancy of sites being greater in the culture grown in the absence of nitrate. Full technical details of this experiment and statistical procedures used for data analysis have been published (Efromovich et al., 2008). The ChIP-chip data for the three originally identified NsrR targets (ytfE, hmp and ygbA) are shown in Fig. 2.
The ChIP-chip data show only a weak signal for NsrR binding at the ygbA promoter (Fig. 2), which is known to be regulated by NsrR in vivo (Bodenmiller and Spiro, 2006). Thus we required rigorous methods to identify other weak signals in the data that may represent bona fide cis-acting regulatory sites. We adopted three approaches to this problem. First, we scrutinized the three datasets individually, and looked for peaks in which two or more consecutive probes showed a greater than twofold enrichment. If a peak met these criteria in at least two of the three samples, then it was recorded as positive. Twenty-nine peaks were identified in this way, of these nine were not considered further on the grounds that they were between convergently transcribed genes or deep within coding regions (here defined as >300 bp from the start codon). This method did not identify the ygbA peak, so may be too conservative. Next, we analysed the three datasets independently with ChIPOTle (Buck et al., 2005), and scored a peak as positive if it was significant (P < 0.0001) in at least two of the three datasets. This analysis identified an additional 41 peaks (including ygbA) of which 16 were discarded as internal sites or sites between convergent genes. Finally, we used a novel method for peak detection (Efromovich et al., 2008) which, disregarding sites deep within coding regions, between convergently transcribed genes or with very small mean enrichment ratios (< 1.5), identified an additional 17 NsrR-binding sites. The final output of 62 NsrR-binding sites in or close to 5′ non-coding regions is shown in Table 2.
Previously identified and novel NsrR-binding sites
The presence of NsrR-binding sites in or near to 5′ non-coding regions identifies genes that potentially belong to the NsrR regulon. Of the promoters bound by NsrR in vivo (Table 2), eight not previously known to be regulated by NsrR (hypA/hycA, acs, aceE, ydcX, putA, ndh and sodB) show responses to nitrite in transcriptomics experiments that are consistent with positive or negative regulation by NsrR (Constantinidou et al., 2006). Most known NsrR regulon members (ytfE, yeaR-yoaG, hmp, hcp-hcr and ygbA) show differential regulation in an asymptomatic strain of E. coli growing in the urinary tract (Roos and Klemm, 2006). Potential NsrR targets ydcX, dsdX, ndh and tehAB (Table 2) are also upregulated in the urinary tract (Roos and Klemm, 2006) and so share an expression pattern with genes known to be regulated by NsrR. Several of the other targets listed in Table 2 have been reported to respond to sources of NO or to S-nitrosoglutathione in other transcriptomics experiments (Justino et al., 2005; Hyduke et al., 2007; Pullan et al., 2007; Bourret et al., 2008; Jarboe et al., 2008). In total, of the 62 targets implicated by the ChIP-chip data, 33 have been previously shown to be influenced by NsrR and/or sources of NO or nitrosative stress (Table 2).
Nine transcription units were previously suggested by transcriptomics to be repressed by NsrR (Filenko et al., 2007); the promoter regions of seven of these (ytfE, hmp, hcp, nrfA, yccM, ygbA and napF) are bound by NsrR in vivo according to the ChIP-chip data (Table 2), and two (uspF and yeaR) are not. Visual inspection of the raw ChIP-chip data confirmed the absence of signals for uspF and yeaR. Direct regulation of yeaR by NsrR has been demonstrated (Lin et al., 2007), hence this is a true false-negative in the ChIP-chip data. One possible explanation is that the 3X-Flag tag on NsrR is occluded by other proteins bound to the yeaR promoter. Overexpression of NsrR causes reduced expression of the small RNA RybB and of the rpoE gene encoding σE (Thompson et al., 2007). Neither gene is bound by NsrR in its promoter region according to the ChIP-chip data. We assume that these are also false negatives, or that NsrR regulation of rybB and rpoE is indirect. The transcriptomics data provided good evidence of one gene positively regulated by NsrR, ydbC (Filenko et al., 2007). We found no NsrR-binding site in the ydbC promoter, though there is a site in the downstream gene, ydbD (Table 2), which is upregulated by NO (Justino et al., 2005).
More than 50 promoters implicated as NsrR targets by ChIP-chip were not identified in a transcriptomics experiment in which an nsrR mutation was phenocopied by repressor titration (Filenko et al., 2007). In some cases this may simply be because NsrR binding to DNA has no regulatory consequence. More likely explanations are that the repressor titration approach used was not very sensitive (see below), and/or that some NsrR targets are subject to additional layers of regulation, such that they would not be identified in a straightforward analysis of the transcriptome under a limited range of growth conditions. The latter consideration applies, for example, to the tynA and feaB genes, which were identified as potential NsrR targets by ChIP-chip (Table 2), but which are subject to regulation by NsrR only in cultures grown on unusual carbon or nitrogen sources (Rankin et al., 2008).
In several cases, NsrR-binding sites are close to the 5′ ends of genes that are (or are probably) internal to single transcription units, and therefore are not associated with promoters; examples are feoB, hcr and nrdB (Table 2). The regulatory significance, if any, of these sites is not known; hcr is particularly interesting because the promoter-proximal gene of the operon (hcp) also has an NsrR-binding site and is regulated by NsrR (Filenko et al., 2007).
The new potential targets for NsrR regulation (Table 2) include genes and operons involved in carbon and energy metabolism (hycA/hypA, feaB, aceE, mhpT, tynA, caiA and ndh), NO metabolism (norR/norV), proteolysis (clpB, ftsH and ptrA), transport processes (mhpT, yhfC, dsdX and yhfC), stress responses (sodB and sufA) and motility (mqsR, fliL and fliA). In an initial follow-up study, we have confirmed NsrR regulation of tynA and feaB (Rankin et al., 2008).
Computational analysis of NsrR-binding sites
Non-coding regions that contain NsrR-binding sites as revealed by ChIP-chip were initially scrutinized for common potential regulatory sequences using WEEDER (Pavesi et al., 2004). This search suggested that most novel potential NsrR targets do not contain a sequence resembling the long inverted repeat that is present in the ytfE, hmp and ygbA promoters (Fig. 1). However, a sequence resembling half of the inverted repeat could be detected in many cases. To extend this search, we constructed a position-specific score matrix (PSSM) from the six easily detectable half-sites in the ytfE, hmp and ygbA promoters and used the PSSM to search the E. coli genome with Virtual Footprint (Münch et al., 2005). As new half sites in promoters known to be bound by NsrR (Table 2) were detected, they were added to the PSSM and the search was repeated iteratively. In this way, we identified 49 potential NsrR-binding sites in 37 of the intergenic regions to which NsrR binds in vivo (Table 2). The sequence logo for these 49 sites is shown in Fig. 1C, which reveals that the consensus sequence for the suggested NsrR-binding site contains a 12 bp interrupted partial palindrome, 5′-AANATGCATTTN-3′, corresponding to one half of the previously described inverted repeat sequence (Fig. 1A; see above). In a number of promoters, the computational search failed to identify significant matches to the PSSM (Table 2). Similarly, in several other ChIP-chip studies, binding sites have been identified in chromosomal regions that do not contain a good match to the consensus sequence for the regulatory protein concerned (reviewed in Wade et al., 2007).
Regulation of motility genes by NsrR
We were particularly interested to observe NsrR-binding sites associated with the promoter regions of transcription units containing genes that are known (fliAZY and fliLMNOPQR) or suspected (mqsR-ygiT) to have roles in motility and sessile growth (Fig. 2). The fliA gene encodes the alternative sigma factor, σ28, which is required for the transcription of Class III motility and chemotaxis genes (Chilcott and Hughes, 2000). The fliZ gene, which is co-transcribed with fliA, encodes a protein which acts as a positive regulator of motility, and as an inhibitor of the expression of curli fimbriae, which are required for surface-attached growth (Pesavento et al., 2008; Saini et al., 2008). The fliLMNOPQR operon encodes structural components of the flagellum and the flagellin export apparatus (Chilcott and Hughes, 2000). The mqsR gene (which is very likely co-transcribed with ygiT, a predicted regulatory gene) has been described as a regulator of motility and biofilm formation (Gonzalez Barrios et al., 2006).
The predicted NsrR-binding site in the fliA promoter overlaps the start site for transcription by RNA polymerase containing σ28, the product of the fliA gene (Fig. 3). The site is therefore well situated to mediate negative regulation of fliAZY expression. To test the functionality of this site, we substituted the highly conserved G at position 6 (Fig. 1C) with a C. We chose to mutate position 6 because the equivalent mutation in the ytfE promoter causes a severe phenotype (Table 1) and this nucleotide is located such that the mutation is unlikely to affect either the σ70 or the σ28 promoter of fliA (a substitution at position 2 would change the σ28 transcription start site). The G to C mutation is on the bottom strand of the fliA promoter (Fig. 3A); the mutant promoter is designated fliAc. NsrR binding to the fliA promoter in vivo was examined by ChIP. When PCR amplification of immunoprecipitated DNA was optimized to allow detection of NsrR binding to the wild-type fliA promoter, binding to the fliAc promoter was undetectable above background levels (Fig. 3B). Thus, these experiments confirm the presence of an NsrR-binding site in the fliA promoter as was suggested by ChIP-chip (Fig. 2) and bioinformatic analysis, and provide experimental support for the revised consensus sequence for NsrR-binding sites.
To quantify the regulation of motility genes by NsrR, we constructed lacZ reporter fusions to the fliA, fliL and mqsR promoters and measured their activities in an nsrR mutant and a strain containing multiple copies of the nsrR gene, in the presence and absence of a source of NO. We found no evidence for regulation of the mqsR promoter by NsrR (data not shown), possibly because we have yet to identify suitable growth conditions that reveal regulation by NsrR. The fliA and fliL promoters were 1.9- and 1.7-fold upregulated in an nsrR mutant respectively, and had moderately increased activities in the presence of a source of NO (Table 3). In the presence of multiple copies of nsrR, the activities of both promoters were 6–7 fold reduced, an effect that was partially reversed by the addition of a source of NO to growth media (Table 3). Taken together, these data indicate that NsrR is a negative regulator of both the fliA and the fliL promoters.
Table 3. Activities of reporter fusions to the fliA and fliL promoters.
Cultures were grown aerobically in minimal medium with glucose as the carbon and energy source. For treatment with NO, cultures were grown to mid-exponential phase, supplemented with 50 μM spermine-NONOate, then sampled 90 min later.
57 ± 8
78 ± 6
84 ± 8
82 ± 8
66 ± 4
90 ± 9
108 ± 10
109 ± 15
103 ± 8
102 ± 6
114 ± 16
106 ± 8
41 ± 2
53 ± 5
48 ± 4
52 ± 5
52 ± 4
66 ± 7
7 ± 1
29 ± 4
13 ± 2
37 ± 5
11 ± 2
41 ± 5
Assay of a fliAc–lacZ fusion revealed that the single base pair mutation in the fliAc promoter mimicked the effect of NO in an otherwise wild-type strain (Table 3). The partial de-repression caused by the fliAc mutation could be overcome by the presence of multiple copies of nsrR (Table 3). These results are consistent with the fliAc mutation lowering the affinity of the NsrR-binding site in the fliA promoter, as was suggested by ChIP.
The fliA–lacZ reporter fusion was ∼2-fold upregulated in a strain transformed with a high copy number plasmid containing the cloned ytfE promoter (data not shown). This effect of the ytfE promoter was abolished by the deletion at position 12 of the NsrR-binding site (Fig. 1B), suggesting that multiple copies of the NsrR-binding site in ytfE de-repress the fliA promoter by repressor titration (hence providing additional confirmation of the presence of an NsrR-binding site in fliA). The small magnitude of the repressor titration effect on fliA likely explains why fliA was not identified as an NsrR target in the transcriptomics analysis (Filenko et al., 2007). In the reciprocal experiment, multiple copies of the fliA promoter failed to cause de-repression of the ytfE–lacZ reporter fusion (and also failed to de-repress the fliA promoter, data not shown). One possible explanation is that the inverted repeat sequence in ytfE provides a higher affinity NsrR-binding site than the single half site in fliA. The same consideration may also explain why deletion of the central base pair in the inverted repeat in ytfE abolishes repressor titration despite preserving two intact half sites.
Regulation of motility by NsrR
Flagella-based motility was assayed on soft agar plates. An nsrR mutant showed a small though reproducible and significant increase in motility (1.3-fold; P < 0.0001) as compared with the wild-type strain (measured as the diameter of the motility ring; data not shown). Addition of an NO source caused a similar increase in motility in a wild-type strain but not in an nsrR mutant. These observations are consistent with the negative regulation of motility genes by NsrR that was measured in reporter fusion assays. In a strain containing multiple copies of nsrR, the motility ring was ∼2-fold smaller (P < 0.0001) than in a control strain with a single chromosomal copy of nsrR (Fig. 4). This effect of NsrR on motility was reversed in plates supplemented with a slow-releasing source of NO (Fig. 4), suggesting that inactivation of NsrR alleviates negative control of motility. The NsrR protein contains three conserved cysteine residues thought to be involved in the co-ordination of an [Fe-S] cluster, which is the likely site of NO sensing (Isabella et al., 2008; Tucker et al., 2008; Yukl et al., 2008). We have substituted cysteine 96 with serine, and found that the NsrR-C96S variant is unable to repress fully the NsrR targets ytfE, hmp, hcp and ygbA (J. Partridge and S. Spiro, unpubl. data). The C96S protein also has no negative effect on motility (data not shown), confirming that the effect of NsrR on motility requires the protein to be in a form that is competent to control transcription. This excludes the possibility that inhibition of motility by multiple copies of nsrR is a non-specific consequence of protein over-production. Motility is a variable and strain-specific phenotype in E. coli. In similar assays to those described above, we showed that multiple nsrR copies inhibit motility to a similar extent (∼2-fold; P < 0.0001) in an E. coli K12 strain (RP437) that is frequently used for assays of motility and chemotaxis.
It has recently been shown that hmp mutants of E. coli are non-motile (Stevanin et al., 2007), though on the succinate medium used the phenotype would also be consistent with a defect in aerotaxis. As the hmp gene is negatively regulated by NsrR (Bodenmiller and Spiro, 2006), one possible interpretation of our results is that the motility defect associated with an increased nsrR copy number results from downregulation of hmp. We assayed the motility of hmp mutants of MG1655 and RP437, using both tryptone and succinate soft agar (Stevanin et al., 2007). With these strains, we found no detectable effect of hmp on E. coli motility. Thus, the motility phenotype that we observe when nsrR is deleted or overexpressed is not a consequence of hmp up or downregulation respectively.
NsrR regulates motility and surface attachment in a uropathogenic strain of E. coli
We were interested to determine whether the effects of NsrR (and NO) on motility that we observed in K12 strains are generalizable to pathogenic strains of E. coli. We focused on a uropathogenic strain of E. coli (UPEC) that is associated with urinary tract infections. The amino acid sequence of NsrR, and the nucleotide sequence of the fliA promoter shown in Fig. 3 are identical in the UPEC strain CFT073 and MG1655. In CFT073, flagella-based motility is important for the organism's ability to ascend the urinary tract and disseminate further in the host (Lane et al., 2007). Furthermore, some genes that are regulated by NsrR in K12 strains are upregulated during urinary tract infection, notably the hmp gene encoding the NO detoxifying haemoglobin (Snyder et al., 2004). Thus, there is evidence that both motility and NsrR might have important roles in vivo, and transcriptomics data suggest that CFT073 is exposed to NO (Snyder et al., 2004).
We found that nsrR overexpression exerted a greater negative effect on motility in CFT073 (> 3-fold; P < 0.0001) than was the case in K12 strains, an effect that was reversed by addition of NO to the medium (Fig. 4). NO caused a small though significant (P < 0.0001) stimulation of CFT073 motility (Fig. 4), as did deletion of the nsrR gene (Fig. 4). Thus, NsrR is a negative regulator of motility in CFT073, and NO influences motility via NsrR. As for K12 strains, we found no motility phenotype associated with an hmp mutation in CFT073 (data not shown).
As motility and attached growth are typically subject to reciprocal regulation, we measured the ability of CFT073 (and derivatives deleted for nsrR or containing multiple copies of the nsrR gene) to adhere to the surface of glass tubes. Deletion of nsrR or addition of a source of NO significantly reduced attached growth (Fig. 5). The presence of multiple copies of nsrR stimulated attached growth, an effect that was partially reversed by the addition of NO (Fig. 5). These results indicate that NsrR regulates attached growth in CFT073 (most likely indirectly) and that NO influences attached growth via NsrR.
Data presented in this paper suggest that NsrR-binding sites in E. coli fall into two classes: those (such as the site in ytfE) comprising two copies of an 11 bp inverted repeat with 1 bp spacing, and those (exemplified by the site in fliA) which have a single copy of the 11 bp element. The available information suggests that the inverted repeat is a higher-affinity site that allows NsrR repression to operate over a larger range. A logical corollary of this suggestion is that the two types of site are occupied by NsrR in different oligomeric states. As the 11 bp motif is a palindrome (Fig. 1C), it may be a binding site for an NsrR dimer, in which case the 23 bp inverted repeat might be occupied by a dimer of dimers. In sedimentation equilibrium experiments, the Streptomyces coelicolor NsrR formed a sequence-specific complex with DNA with a molecular weight consistent with the protein being dimeric (Tucker et al., 2008). However, these experiments were done with protein containing a [2Fe-2S] cluster; the physiologically relevant form of NsrR may contain a [4Fe-4S] cluster (Yukl et al., 2008). The [4Fe-4S] NsrR from Bacillus subtilis is dimeric in solution, its oligomeric state in the presence of a DNA target has not been examined (Yukl et al., 2008). Interestingly, the NsrR homologue IscR also binds to two types of site (Type 1 and Type 2), though in this case they are unrelated sequences. Binding to Type 2 sites does not require the [Fe-S] cluster of IscR, and two IscR dimers bind cooperatively to a Type 2 site (Nesbit et al., 2009).
We have demonstrated that the NO-sensitive repressor protein NsrR is a negative regulator of motility genes and of flagella-based motility in E. coli K12. We propose that NO exerts effects on motility through NsrR-mediated regulation of the fliA promoter. The fliA gene product (σ28) is required for the expression of all Class III flagella and chemotaxis genes (Chilcott and Hughes, 2000), so by regulating fliA NsrR potentially exerts widespread indirect effects on motility and chemotaxis, both of which are involved in migration through soft agar (Wolf and Berg, 1989). The effects of NsrR on fliA promoter activity and on motility are quite small, and are more pronounced in strains containing multiple copies of nsrR than in an nsrR mutant. Similar contrasts between deletion and overexpression have been observed previously for genes regarded as negative regulators of motility. For example, the pefI-srgD genes of Salmonella enterica serovar Typhimurium were recently described as negative regulators of motility, despite the fact that there is no phenotype associated with deletion of the genes, which inhibit motility when expressed from the araBAD or tetA promoters (Wozniak et al., 2009).
We also showed that NsrR regulates motility and attached growth in a UPEC strain. A search of the UPEC strain CFT073 genome with the same PSSM used to search the E. coli K12 genome revealed predicted NsrR-binding sites in the promoter regions of genes involved in the production of pili (papI and papI_2) and fimbriae (sfaB, C1936, ipbA and ipuA). Thus it is possible that the effect of NsrR on attached growth is multifactorial and indirect. One possible component of the effect is that changes in fliZ expression mediated by NsrR regulation of the fliAZY promoter lead to altered levels of expression of curli fimbriae. In E. coli K12, FliZ acts by indirectly causing downregulation of genes involved in the expression of curli fimbriae, which are required for surface attachment (Pesavento et al., 2008).
Circumstantial evidence has previously implicated NO as a regulator of chemotaxis, motility and biofilm development. Haem-containing NO-binding domains of methyl accepting chemotaxis proteins have been characterized (Karow et al., 2004; Nioche et al., 2004), although the prediction that these proteins mediate taxis towards or away from NO has not been tested. In transcriptomic experiments, the expression of some motility genes has been observed to be perturbed by exposure of cultures to sources of NO or nitrosative stress (imposed by S-nitrosothiols), although both positive and negative responses have been reported, and the regulators involved were not identified (Bourret et al., 2008; Constantinidou et al., 2006; Jarboe et al., 2008). In the non-pathogenic organism Nitrosomonas europaea, NO stimulates biofilm formation (Schmidt et al., 2004). In Azotobacter vinelandii, expression of the flhDC genes (which encode the master regulator of motility) is negatively regulated by the oxygen-sensor CydR, an orthologue of the E. coli FNR protein (León and Espín, 2008). CydR is sensitive to NO (Wu et al., 2000), suggesting that exposure to NO might stimulate motility in A. vinelandii via increased expression of flhDC. In Pseudomonas aeruginosa and Staphylococcus aureus, NO inhibits biofilm formation or stimulates dispersal, and NO stimulates motility in P. aeruginosa (Barraud et al., 2006; Van Alst et al., 2007; Schlag et al., 2007). A molecular mechanism which accounts for the effects of NO on biofilm development or motility in these organisms has not previously been described, though there has been some speculation about the regulatory proteins involved that might act as receptors for NO (Romeo, 2006). The mechanism we propose in this paper may not be applicable to P. aeruginosa and S. aureus, because those species do not have obvious orthologues of NsrR. Nevertheless, we suggest that other NO sensing transcriptional regulators (Spiro, 2007; Rodionov et al., 2005) might play an equivalent role in these cases.
Strains, media and growth conditions
The strains and plasmids used in this work are listed in Table 4. The rich medium was L Broth (tryptone, 10 g l−1 yeast extract, 5 g l−1; NaCl, 5 g l−1). A mineral salts medium (Spencer and Guest, 1973) supplemented with glucose (0.5% and 0.2%, w/v, for anaerobic and aerobic cultures respectively), casamino acids (0.05%, w/v) and thiamine (5 μg ml−1) was used for growth of cultures for β-galactosidase assays. Ampicillin (100 μg ml−1) and kanamycin (25 μg ml−1) were added as required. Cultures were grown aerobically or anaerobically as previously described (Bodenmiller and Spiro, 2006). For β-galactosidase assays (Miller, 1992), aerobic cultures were treated with 50 μM spermine-NONOate, and anaerobic cultures with 5 mM nitrite when in early exponential phase (OD600 = 0.15–0.3), then were assayed 90 min later while still in log phase. Spermine-NONOate liberates two equivalents of NO with a half-life of 39 min at 37°C (Cayman Chemicals), and under our culture conditions caused little or no growth inhibition at this concentration.
The mqsR, fliA and fliL promoter regions were amplified by PCR (all primer sequences are available from the authors on request) and fused to lacZ in pRS415, transferred to λRS45, and integrated into the chromosome as described previously (Simons et al., 1987; Bodenmiller and Spiro, 2006). Genes were disrupted by replacing the coding region with a kanamycin-resistance cassette using the λred recombinase method with pKD4 as the template; mutations were converted to unmarked deletions using pCP20 (Datsenko and Wanner, 2000). The nsrR plasmid pJP07 is a derivative of p2795 (Husseiny and Hensel, 2005) and has been described previously (Rankin et al., 2008). NsrR with a C96S substitution was expressed from the equivalent plasmid pJP09. The fliA promoter (on a 474 bp fragment in pSTBlue) was mutated in the putative NsrR-binding site, and the mutant promoter was fused to lacZ in pRS415 as described above. Site-directed mutants were introduced using the QuickChange Site-Directed Mutagenesis Kit (Stratagene) according to manufacturer's instructions.
For random mutagenesis, pGIT1 (Bodenmiller and Spiro, 2006) was mutagenized with the GeneMorph PCR mutagenesis kit (Stratagene) according to the manufacturer's instructions. After mutagenesis, plasmid DNA was transformed into strain JOEY19 (λytfE-lacZ), and transformants were screened on L agar containing Xgal. Colonies with a white or pale blue phenotype were selected, plasmid DNA was purified and the sequence of the ytfE fragment determined. Clones with multiple mutations were not further analysed. Mutant DNAs of interest generated by random or site-directed mutagenesis were cloned into pRS415, and integrated on to the chromosome at the lambda attachment site, as previously described.
ChIP and ChIP-chip
For ChIP analysis of the ytfE and fliA promoters, chromatin was precipitated from cultures of strains expressing 3XFlag-tagged NsrR (or an untagged control) and transformed with pSTBlue-1 derivatives containing wild type or mutant ytfE or fliA sequences. Precipitated DNAs were purified and equal amounts of templates (1 ng for ytfE, 2 ng for fliA) were amplified by 16 cycles of PCR with primers flanking the cloning site in pSTBlue. ChIP-chip was performed and data analysed as described previously (Efromovich et al., 2008). Chromatin immunoprecipitation and microarray data have been deposited in the GEO database (accession GSE11230).
Motility and attachment assays
Motility was assayed on soft agar swim plates inoculated with 4 μl of an exponential phase (OD600∼0.5) culture. The plates contained 1% tryptone, 0.25% NaCl, 0.3% Difco agar, and antibiotics as required. The plates were incubated in a wet box for 20 h at 30°C. Surface attachment to 16 mm glass tubes was assayed in standing cultures grown for 24 h at 30°C in L broth, using the crystal violet staining method (Pratt and Kolter, 1998). For treatment with NO, the plates or standing cultures were supplemented with 100 μM (for K12 strains) or 250 μM (for CFT073) diethylenetriamine-NONOate, which liberates two equivalents of NO with a half-life of 20–56 h under the conditions of these experiments (Cayman Chemicals).
We are grateful to Sandy Parkinson, Harry Mobley, Michael Hensel, Barry Wanner and Valley Stewart for generously providing strains and plasmids, to Gladys Alexandre, Sam Efromovich, Mike Manson and David Grainger for helpful discussions, and to Ray Dixon for comments on the manuscript. This work was supported in part by Grant MCB-0702858 from the National Science Foundation.