Bacteria are often found associated with surfaces as sessile bacterial communities called biofilms, and the formation of a biofilm can be split up into different stages each requiring the expression of specific genes. The production of extracellular polysaccharides (EPS) is important for the maturation of biofilms and is controlled by the Rcs two-component pathway in Escherichia coli (and other Gram-negative bacteria). In this study, we show, for the first time, that the RcsC sensor kinase is required for normal biofilm development in E. coli. Moreover, using a combination of DNA macroarray technology and transcriptional fusion analysis, we show that the expression of > 150 genes is controlled by RcsC in E. coli. In silico analyses of the RcsC regulon predicts that 50% of the genes encode proteins that are either localized to the envelope of E. coli or have activities that affect the structure/properties of the bacterial surface, e.g. the production of colanic acid. Moreover, we also show that RcsC is activated during growth on a solid surface. Therefore, we suggest that the RcsC sensor kinase may play an important role in the remodelling of the bacterial surface during growth on a solid surface and biofilm formation.
It has been estimated that at least 90% of all bacteria in the environment reside attached to a solid surface, and many of these bacteria form sessile communities called biofilms. The medical and industrial importance of biofilms has now been firmly established, and much work has been done on the genetic analysis of biofilm formation in several bacterial species, e.g. Escherichia coli, Pseudomonas aeruginosa and Vibrio cholerae (O’Toole and Kolter, 1998; Pratt and Kolter, 1998; O’Toole et al., 1999). Biofilm formation can be divided up into different stages, from the initial attachment of the bacteria to the surface to the maturation of the biofilm into a stable three-dimensional structure, and each stage requires the expression of different genes (O’Toole et al., 2000; Danese et al., 2001). Indeed, there is evidence that the expression of as many as 38% of the genes present in the genome of a bacterium change during the transition from free-living, planktonic growth to growth on a solid surface (Prigent-Combaret et al., 1999). It is therefore likely that many different signalling pathways will be involved in controlling these changes in gene expression.
Many bacteria produce extracellular polysaccharides (EPS) that play an important role in the maturation of biofilms. E. coli mutants that are unable to produce the EPS colanic acid can still attach to surfaces; however, these biofilms do not develop into stable three-dimensional structures (Danese et al., 2000a). In E. coli, the genes encoding the proteins required for the production and secretion of colanic acid are grouped together in the cps operon (recently renamed the wca operon; Whitfield and Roberts, 1999). Expression of the cps operon in E. coli is controlled by the Rcs two-component pathway, which is composed of the membrane-associated proteins RcsC and YojN and the cytoplasmic response regulator, RcsB (Stout and Gottesman, 1990; Stout, 1994; Takeda et al., 2001). RcsC is a hybrid sensor kinase protein that contains a transmitter domain and a receiver domain. Transmitter domains are associated with histidine kinase activity, and recent genetic evidence suggests that, on receiving a signal, RcsC autophosphorylates on residue His-463 in the transmitter domain (Clarke et al., 2002). The phosphoryl group is then transferred to Asp-859 in the RcsC receiver domain and then to His-842 in the HPt domain of YojN. The last step in the Rcs phosphorelay is the phosphorylation of the RcsB response regulator by YojN (Takeda et al., 2001). The binding of phosphorylated RcsB to DNA is modulated by the availability of an accessory protein called RcsA, the level of which is post-translationally controlled by the Lon protease (Torres and Gottesman, 1987; Stout et al., 1991).
The Rcs two-component pathway has been implicated in the regulation of bacterial responses to surfaces. Mutations in both rcsC and yojN affect the temporal regulation of swarming motility in E. coli, Proteus mirabilis and Salmonella typhimurium (Belas et al., 1998; Toguchi et al., 2000; Takeda et al., 2001). Although the signal for RcsC remains uncharacterized, several studies have shown that this sensor kinase responds to complex signals such as dessication, changes in osmolarity and alterations in the levels of certain proteins with functions that are associated with the cell surface (Parker et al., 1992; Ophir and Gutnick, 1994; Sledjeski and Gottesman, 1996). Moreover, it has been shown recently that the rcsC gene in S. typhimurium also regulates the expression of ugd, a gene that is required for the production of l-aminoarabinose, an amino sugar that is added to the lipid A moiety of lipopolysaccharide (LPS) to induce resistance to antimicrobial cationic peptides (Gunn et al., 1998; Mouslim and Groisman, 2003). Taken together, these data suggest that the Rcs two-component pathway contributes to the modification of the bacterial surface in response to changes in the environment.
In this study, we show that RcsC is required for normal biofilm development in E. coli K-12. In addition, we show that RcsC regulates the production of colanic acid, and other genes, during growth on a solid surface. Several other reports have shown that colanic acid is expressed in response to a solid surface (Prigent-Combaret et al., 1999; 2000), but this is the first report that directly implicates RcsC in this regulation. Moreover, we use nylon-based DNA arrays and transcriptional fusions to identify genes that are regulated by RcsC (i.e. the RcsC regulon). These data suggest that RcsC has an important role in controlling the remodelling of the surface of E. coli in response to growth on a solid surface and biofilm formation.
RcsC is required for normal biofilm development in E. coli K-12
The Rcs two-component pathway regulates the expression of the cps (wca) operon in E. coli. The cps operon contains 19 genes that are responsible for the production and secretion of colanic acid from the cell (Stevenson et al., 1996). It has been shown previously that the production of colanic acid in E. coli K-12 is important for the maturation of biofilms (Danese et al., 2000a). Therefore, we hypothesized that RcsC would be important for the regulation of genes involved in the maturation of biofilms. To test this, we used phage P1cml to transduce the E. coli K-12 strain ZK2686 with the rcsC52::Tn10 allele, resulting in strain BMM520. ZK2686 and BMM520 were then tested for their ability to form biofilms on the wells of a PVC microtitre plate (see Experimental procedures). The cells were incubated at 30°C, without shaking, for 48 h, and crystal violet (CV) was used to determine the extent of biofilm formation. No difference was observed between the growth rates of ZK2686 and BMM250 when the cells were grown in liquid broth (data not shown). As expected, the CV staining for ZK2686 was strong, indicating that this strain forms a good biofilm on PVC (see Fig. 1). BMM520 also formed a biofilm, although quantitative analysis revealed that the CV-stained band was not as intense as the wild type (see Fig. 1B). Moreover, the defect in BMM520 biofilm formation was complemented by a plasmid containing the rcsC gene, pPSG980. On the other hand, BMM520 cells carrying plasmids encoding mutated RcsC proteins, RcsCH463Q and RcsCD858Q, were unable to form normal biofilms (see Fig. 1). The residues H463 and D859 are implicated in the RcsC phosphorelay, and this result indicates that RcsC signalling is required for normal biofilm formation in E. coli (Clarke et al., 2002).
DNA macroarray experiments
Given the potential role for RcsC in controlling normal biofilm development, we decided to use array technology to identify the RcsC regulon in E. coli K-12. In order to compare the expression profiles of E. coli in the RcsC ON state and the RcsC OFF state, we needed a mechanism by which we could conditionally activate RcsC. We, and others, have shown previously that the moderate overproduction of DjlA, a membrane-linked DnaJ-like protein, induces the expression of cpsB–lacZ in a RcsC-dependent manner (Zuber et al., 1995; Clarke et al., 1997; Kelley and Georgopoulos, 1997). Importantly, the overproduction of DjlA results in a strong induction of cpsB–lacZ expression that is not transient and is clearly dependent on the presence of rcsC (Kelley and Georgopoulos, 1997; Clarke et al., 2002). Therefore, our strategy was to use DjlA as a signal for activating RcsC and compare the expression profiles of isogenic rcsC+ and rcsC– strains of E. coli in the presence of DjlA overproduction. Isogenic strains PSG1031 and PSG1038 were transformed with pPSG961-31, a pBAD33-derived plasmid containing the djlA gene under the control of the paraBAD promoter (Guzman et al., 1995; Clarke et al., 1996; 1997). The transformed cells were grown overnight, inoculated into fresh LB broth, and the cultures were incubated at 30°C with aeration. At an OD600 = 0.1, the cultures were supplemented with 0.2% (w/v) l-arabinose, and samples were removed at regular intervals for OD600 readings and β-galactosidase assays. We observed that, as reported earlier, there is a strong, non-transient and rcsC-dependent increase in the level of cpsB–lacZ expression (see Fig. 2A). We also confirmed previous observations that there was a significant delay (≈ 100 min, see Fig. 2A) between the addition of arabinose and the induction of cpsB–lacZ expression (Clarke et al., 1997). The reason for this delay is not fully understood, but it does result in the bacterial cells entering stationary phase before cpsB–lacZ expression, and presumably RcsC activity, reached maximum levels. To maximize the reproducibility of our experiments, we wanted to isolate RNA from exponentially growing cells and, therefore, cells induced with arabinose for 2 h (see arrow in Fig. 2A) were diluted to an OD600 = 0.1 into fresh LB broth supplemented with arabinose, and samples were taken for OD600 reading and β-galactosidase assays. Under these conditions, it can be seen that cpsB–lacZ expression reaches high levels during the exponential phase of growth (see Fig. 2B). Importantly, there was no cpsB–lacZ expression in PSG1038/pPSG961-31 cells cultured under identical conditions, confirming that DjlA overproduction induces cpsB–lacZ expression in a rcsC-dependent manner. Both PSG1031/pPSG961-31 and PSG1038/pPSG961-31 cells were harvested during exponential phase at an OD600 of 0.5–0.6 for RNA isolation (see arrow in Fig. 2B).
The Rcs regulon
For the array analysis, we used the PanoramaTME. coli Gene Array nylon membranes supplied by Sigma-Genosys. RNA from each sample to be studied was isolated, converted to cDNA and hybridized to the membrane. The RNA from each sample was isolated twice, hybridized to each membrane twice and, as every gene has two spots on the array, data analysis was carried out on eight experimental values for each gene, a method adapted from that described previously (Conway et al., 2002). The data were normalized, the eight values obtained for each gene were averaged, and the induction ratio (IR) was calculated as the ratio of the average expression level of each gene in the rcsC+ and rcsC– backgrounds. The ratio was calculated such that genes that are more highly expressed in the rcsC+ background (upregulated by RcsC) are given a positive value and genes that are more highly expressed in the rcsC– background (downregulated by RcsC) are given a negative value. Moreover, we calculated the mean and standard deviation of the log10-transformed induction ratio between the rcsC+ and rcsC– backgrounds. Genes that have an expression ratio > 3 standard deviations from the mean and display a P-value < 0.05 were considered as RcsC-regulated genes (see Experimental procedures).
Based on this macroarray analysis, we determined that, in PSG1031 (a derivative of MC4100), 149 genes were regulated by RcsC, i.e. displaying an induction ratio (IR) < −2.3 and > +2.0; 83 genes were upregulated and 66 genes were downregulated (see Tables 1 and 2). PSG1031 contains a chromosomal cpsB–lacZ fusion and, in our experimental system, RNA was extracted from these cells when β-galactosidase activity was high. Therefore, the level of both lacZ and cps expression serves as an important internal control in these experiments. As expected, the expression of all the cps genes (up to and including cpsB) is clearly upregulated in the presence of RcsC (see Table 1 and Fig. 3) as is the expression of lacZ (see Table 1). Surprisingly, the IR is different for each of the cps genes despite recent work that has shown that a single promoter drives the expression of the entire cps operon (Stout, 1996; Van Dyk et al., 2001). We feel that this variation in the IR values probably results from processing of the large cps transcript. Finally, as we were overproducing DjlA from the plasmid pPSG961-31, it was a formal possibility that some differences might be accounted for by slight differences in the expression of djlA itself, i.e. djlA-dependent genes. However, this is unlikely as the level of djlA expression in both strain backgrounds was identical (IR = 1.0).
. Function of each gene product as described in the database of the National Centre for Biotechnology (http://www.ncbi.nlm.nih.gov/). Genes highlighted in bold are predicted to be involved in the production of surface appendages, and these genes are discussed in more detail in the text.
Interestingly, rcsB (IR = +2.2) and rcsA (IR = +7.6) were both induced by RcsC in our analysis, suggesting that expression of these genes is autoregulated. Autoregulation is often found to occur when a two-component pathway regulates the expression of a large number of genes (Bijlsma and Groisman, 2003). Other genes previously shown to be regulated by the Rcs two-component pathway were also identified, e.g. osmC (IR = +2.7) and tolQ (IR = −2.3) (Clavel et al., 1996; Davalos-Garcia et al., 2001).
The Rcs two-component pathway regulates the surface composition of E. coli
Apart from the genes with unknown function (which account for 35% of the genes identified), the largest functional category described contains genes that are involved in the biosynthesis of capsule and envelope components (17% of the genes identified). This suggests that the Rcs two-component pathway may have an important role in regulating the surface composition of E. coli, a function that would be compatible with biofilm formation. To address this possibility, we carried out further in silico analyses to predict the cellular localization of each protein encoded by an RcsC-regulated gene. Of all the genes identified by the array analysis, 91 (61%) were predicted to encode proteins with a cytoplasmic localization (see Table 3). Therefore, 58 genes (39%) were predicted to encode proteins that are localized to the inner membrane (31 genes), the periplasm (9 genes), are lipoproteins (8 genes) or have an unknown extracytoplasmic location, i.e. are located in the periplasm, secreted outside the cell and/or associated with the outer membrane (10 genes). However, 17 (11%) of the genes localized in the cytoplasm are known to encode proteins that are involved in capsule production and envelope biogenesis. Therefore, in total, at least 50% of the RcsC-regulated genes identified in this study can be classified as envelope-related genes. This confirms that the Rcs pathway plays an important role in controlling the surface composition of E. coli.
Table 3. Summary of the predicted cellular localization of proteins encoded by RcsC-regulated genes.
No. of genes
% of identified genes
. Genes that encode envelope proteins and genes that encode cytoplasmic proteins that are predicted to be involved in envelope biogenesis, e.g. capsule synthesis.
Identification of RcsC-regulated genes using transcriptional fusions
Expression profiling using arrays is unlikely to identify all the genes that are potentially regulated by a signalling pathway. Therefore, we decided to use a promoterless lacZ as a transcriptional reporter in an alternative approach to identify rcsC-regulated genes. In this way, we would (i) provide independent confirmation of the RcsC-dependent regulation of genes already identified by the macroarray analysis and (ii) determine the degree of coverage provided by the array study by comparing the genes identified uniquely by the array or the lacZ studies or both.
We constructed a promoter library of MC4100 chromosomal DNA in a reporter plasmid, pFZY1. This plasmid is based on the F factor and has an average copy number of 1–2 in the cell (Koop et al., 1987). This library consists of small fragments of chromosomal DNA fused to a promoterless lacZ gene. Therefore, lacZ will be expressed if a promoter is present on the fragment of E. coli DNA. This library was electroporated into a reporter strain, PSG1066 Δlac cps::Tn10, that was carrying the plasmid pPSG961-31. Genes regulated by the Rcs pathway were identified by replica plating transformed colonies on agar plates, with or without arabinose, and comparing the expression of the lacZ gene. In total, 1000 independent clones were screened, and 16 clones in which the expression of lacZ was affected by DjlA overproduction were identified. The pFZY1-derived plasmid was isolated from these bacteria and retransformed into an rcsC– strain to confirm that the regulation was dependent on RcsC. In this way, we identified four clones (pUP14, pUP16, pUP28 and pUP30) that were regulated by DjlA overproduction in a manner that was RcsC dependent (see Table 4). The insert in each of the four pUP clones was sequenced and submitted to a homology search against the annotated E. coli genome sequences at NCBI using the blastn algorithm (http://www.ncbi.nlm.nih.bov/blast). In our analysis, the first predicted promoter to occur upstream from the lacZ gene was assumed to be the promoter responsible for driving lacZ expression. The plasmid pUP14 contains an insert of < 1 kb that is predicted to contain a single promoter just upstream of b0379 (yaiY ), a gene encoding a putative inner membrane protein of unknown function. Plasmid pUP30 contains an insert of < 1 kb, and sequence analysis suggests that a promoter upstream from b2466 (ypfG) is driving lacZ expression. The ypfG gene is predicted to encode a periplasmic protein and, although ypfG was not identified on the macroarray analysis, the gene immediately upstream (yffH, b2467) had an IR = +3.7. Both yffH and ypfG are transcribed in the same direction, and our data suggest that this locus could be an operon that is regulated by RcsC. The plasmid pUP28 contains an insert of ≈ 2 kb, and sequence analysis predicted that a promoter upstream from b3552 (yiaD) is driving lacZ expression, and this gene is predicted to encode an outer membrane lipoprotein. Finally, plasmid pUP16 contains an insert of ≈ 2 kb, and sequence analysis predicted that the promoter driving lacZ expression was between b4090 and b4092. A small open reading frame (ORF), encoded by b4091 (phnQ), was predicted in this region of the E. coli K-12 genome, and phnQ was identified on the macroarray with an induction ratio of +2.9. However, it is predicted that phnQ is transcribed in the opposite orientation to the lacZ gene in pUP16 (Blattner et al., 1997). Interestingly the genomic sequences of the equivalent region of both E. coli O157 and E. coli CFT073 predict an ORF in the opposite orientation to phnQ (encoding proteins NP_290725 and NP_756948 respectively) (Perna et al., 2001; Welch et al., 2002). This ORF is predicted to encode a 109-amino-acid peptide that is localized to the inner membrane via an N-terminal transmembrane domain. Alignment of the DNA sequences from E. coli K-12, O157, CFT073 and pUP16 revealed that the K-12 genome sequence has a −1 frameshift in this region (data not shown). This mutation effectively interrupts the ORF and results in the inappropriate annotation of this gene in K-12. Moreover, this frameshift appears to be the result of a sequencing error, as it is not found in the insert of pUP16, which originates from MC4100, a K-12 strain. Importantly, all the promoters identified here using pFZY1 are found upstream of genes encoding proteins that are predicted to be associated with the cell surface.
Table 4. RcsC-regulated genes identified by transcriptional fusions.
. Value shown is the mean of three replicates ± standard deviation.
617 ± 65
52 ± 4
474 ± 76
61 ± 6
172 ± 18
38 ± 8
784 ± 23
149 ± 8
545 ± 17
139 ± 19
254 ± 72
11 ± 6
297 ± 48
0.2 ± 1
531 ± 42
3 ± 3
435 ± 72
2 ± 1
684 ± 74
10 ± 7
746 ± 77
3 ± 0.4
155 ± 5
56 ± 8
λp lacMu53 library
The phage λplacMu53 is a derivative of bacteriophage λ that uses the transposition machinery of phage Mu to insert into the chromosome (Bremer et al., 1985). When inserted in the proper manner, λplacMu53 results in a stable transcriptional fusion between the site of insertion and lacZ, which is linked to a gene encoding resistance to kanamycin. Using λplacMu53, we screened 1200 individual KmR colonies of PSG1032/pPSG961-31 for transcriptional fusions in which the expression of lacZ was induced by DjlA overproduction. Potential positive fusions were then moved into clean rcsC+ and rcsC– backgrounds by P1cml transduction and, in this way, we identified nine fusions that were positively regulated by DjlA overproduction in a manner that was dependent on RcsC. The chromosomal location of each fusion was identified by arbitrary-primed polymerase chain reaction (PCR) and is shown in Table 4. Interestingly, four of these fusions were in genes that had previously been identified in our array analysis (wcaG, wzxC, gmd, ykfE ). Of the remaining fusions, two were independent fusions in ycfT, one was in ybcI, one was in the leader sequence of osmY, and the final insertion was in yjbH. This gene is predicted to be the final ORF in an operon (yjbEFGH ) and, although yjbH was not identified by the macroarray studies, the three ORFs upstream were all clearly identified (yjbE, IR = +9.5; yjbF, IR = +4; and yjbG, IR = +3.6), strongly suggesting that this locus is regulated by RcsC. It is interesting to note that the yjbEFGH operon is predicted to encode proteins involved in polysaccharide biosynthesis, and both ycfT and ybcI are both predicted to encode membrane proteins.
Therefore, from our transcriptional fusion studies, we have identified 12 genes that were regulated by RcsC. Interestingly, seven of these genes (58.3%) were in loci previously identified in our macroarray analysis. The remaining five genes (41.7%) represent loci that were not identified by the macroarray analysis, suggesting that other members of the Rcs regulon remain to be identified.
The insertion of the λplacMu53 element into a gene generally results in the formation of a knock-out mutation of that gene. Therefore, the different λplacMu53 fusions identified in this study were transduced into strain ZK2686 and tested for their ability to form biofilms on PVC. Using this test, all the mutations formed biofilms that were indistinguishable from the wild-type strain (data not shown). Interestingly, one of the mutants tested, BMM526, has the λplacMu53 insertion in gmd, a gene in the cps operon. It has been reported previously that the cps operon is required for normal biofilm development in E. coli (Danese et al., 2000a). However, we have not carried out any microscopic analyses of the biofilm formed by these mutants and, therefore, we cannot rule out the possibility that these mutants are affected in biofilm formation. Nevertheless, these data indicate that the defect in biofilm formation on PVC observed with the rcsC mutant (see Fig. 1) is probably a result of the combined action of many gene products or that the RcsC-regulated gene responsible for the defective biofilm phenotype has not yet been identified.
The RcsC sensor kinase responds to growth on a solid surface
We wanted to show that RcsC was regulating gene expression in response to a physiological signal that might be relevant to biofilm formation. Previous work had indicated that RcsC was more sensitive to DjlA activation when cells were cultured on agar plates, suggesting that RcsC may be responding to growth on a solid surface (D. Clarke, unpublished observation). Therefore, in order to determine whether growth on a solid surface was activating the RcsC sensor kinase, we cultured PSG1031 cpsB–lacZ and PSG1038 rcsC52::Tn10 cpsB–lacZ until mid-exponential phase in LB at 30°C before the cells were transferred to a nitrocellulose membrane by filtration. The cells were washed in situ with PBS, and the membrane was transferred to LB agar plates and incubated at 30°C. At the appropriate time, the membrane was removed from the agar plate, the cells growing on the membrane were harvested by immersing the membrane in LB, the cell suspension was stored on ice, and an aliquot of the resuspended cells was then used for a β-galactosidase assay. The results shown in Fig. 4A clearly show that the expression of the cpsB–lacZ fusion in PSG1031 increased over 200-fold after growth on a solid surface, starting immediately after transfer and peaking ≈ 150 min after transfer, before slowly decreasing. Moreover, this increase in cpsB–lacZ expression was dependent on the presence of the rcsC gene, as no significant increase in cpsB–lacZ expression was observed in PSG1038 (see Fig. 4A). We did not observe any difference in the growth rate of PSG1031 and PSG1038 during these studies (data not shown).
Previous studies have shown that the expression of cpsB–lacZ does not increase during growth in LB broth (Clarke et al., 1997). To confirm that the response observed during the filtration assay is not a consequence of either bacterial growth or filtration per se, we cultured PSG1031 to mid-exponential phase in LB broth and, at t0, the cells were filtered onto two membranes as described in Experimental procedures. One membrane was placed on a prewarmed LB agar plate, and the other was placed in prewarmed LB broth, and both the agar plate and the liquid culture were incubated at 30°C for 150 min. Samples were then taken for β-galactosidase assays, and the results are shown in Fig. 4B. After filtration and growth in liquid, the level of cpsB–lacZ expression had increased 4.5-fold relative to the level of expression at t0, suggesting that filtration has a relatively small effect on cpsB–lacZ expression. On the other hand, when the membrane was incubated on LB agar, the increase in cpsB–lacZ expression is 90-fold compared with t0, suggesting that the full increase in cpsB–lacZ expression requires growth on a surface, i.e. LB agar. Therefore, the RcsC sensor protein is regulating the expression of cpsB–lacZ in response to growth on a solid surface.
RcsC is not sensing osmotic shock, but may be sensing membrane perturbations, during growth on a solid surface
It has been shown previously that the RcsC sensor kinase can respond to both osmotic shock and membrane perturbations (Parker et al., 1992; Sledjeski and Gottesman, 1996; Ebel et al., 1997). We were interested in finding out whether either of these signals contributed to the RcsC-dependent increase in cpsB–lacZ expression that we observed in our filter assay. EnvZ is a well-characterized sensor kinase that, via the response regulator OmpR, regulates the expression of ompC in response to changes in osmolarity (Pratt et al., 1996). CpxA is a sensor kinase that, via the response regulator CpxR, regulates the expression of cpxR in response to extracellular stresses and membrane perturbations (Raivio and Silhavy, 2001). Therefore, we reasoned that, if the filter assay is affecting either osmolarity or membrane integrity, we would expect to observe changes in ompC expression or cpxR expression during growth on a surface. PSG1041 ompC–lacZ and TR235 cpxR–lacZ were grown to mid-exponential phase and filtered onto membranes that were placed either on an LB agar plate or in LB broth, incubated at 30°C for 150 min before the cells were harvested for β-galactosidase assays. From these results (see Fig. 5), it is clear that the relative level of ompC–lacZ expression increases to the same extent during growth in liquid broth or on LB agar, suggesting that the small increase in ompC expression observed is not dependent on a solid surface. Moreover, if this observed increase in expression resulted from a change in osmolarity, we would expect that the relative level of expression would decrease to 1 in the absence of the envZ gene. However, the increase in ompC–lacZ expression during growth on LB agar is in fact greater when the envZ gene is interrupted, strongly suggesting that the observed change in ompC–lacZ expression is not a response to either growth on a surface or a change in osmolarity but, rather, this change is simply growth dependent. Therefore, osmotic shock does not appear to be a signal that is being generated by the filter assay. On the other hand, the relative level of cpxR–lacZ expression increases twofold during growth on LB agar compared with growth in liquid broth. Moreover, this increase in expression is dependent on cpxA, suggesting that the CpxA sensor kinase is being activated during the filter assay. Therefore, the CpxA-dependent increase in cpxR–lacZ expression suggests that the membrane of E. coli is being perturbed during the filter assay, and this perturbation may be an important contributing factor to the activation of RcsC.
The expression of RcsC-regulated genes responds to a solid surface
We have shown that the expression of the cps operon is activated by growth of the cells on a solid surface in an RcsC-dependent manner. We wanted to know whether this surface-induced expression was common to other RcsC-regulated genes. To address this, we took advantage of the transcriptional fusions that we had identified as being regulated by RcsC. The appropriate reporter (either the pUP plasmid or the λplac Mu53 fusion) was moved into isogenic rcsC+ and rcsC – backgrounds, grown to mid-exponential phase and filtered onto nitrocellulose membranes as described in Experimental procedures. The membranes were placed on LB agar plates and incubated at 30°C for 150 min before the cells were harvested and assayed for β-galactosidase activity. All the pUP plasmids showed an increase in the level of lacZ expression 150 min after transfer to the membrane, although for pUP16 (phnQ) and pUP28 (yiaD), this increase is small (1.5- to 2.5-fold) (see Fig. 6A). On the other hand, all the λplacMu53 fusions show large increases in lacZ expression ranging from fivefold to the 180-fold induction observed with the gmd–lacZ fusion (see Fig. 6B). The gmd gene is part of the cps operon, and the increase in expression observed with this fusion agrees with the 200-fold induction observed previously with a cpsB–lacZ fusion (see Fig. 4A). We tested the expression of these transcriptional fusions during growth in LB broth and, as expected, the expression did not change during exponential growth in LB broth (data not shown). Therefore, the expression of all nine genes tested was induced when the bacteria were grown on a solid surface, and this induction was dependent on RcsC.
In this study, we have shown, for the first time, that the RcsC sensor kinase is required for normal biofilm formation in E. coli. Moreover, we have combined DNA array studies and transcriptional fusion analysis to identify over 150 genes that are regulated by RcsC in response to the overproduction of the membrane protein, DjlA, in E. coli K-12. Analysis of the RcsC regulon reveals that at least 50% of the genes regulated by RcsC are predicted to encode proteins with functions related to the bacterial surface. All the genes identified in this study that were tested further are also regulated by RcsC during growth on a surface, i.e. agar. Therefore, we propose that RcsC plays an important role in the remodelling of the bacterial surface in response to certain environmental signals during growth on a surface and biofilm formation.
The Rcs two-component pathway has recently been shown to activate the expression of a small RNA, rprA, that is involved in the post-translational regulation of σS levels in E. coli (Majdalani et al., 2002). Therefore, it is possible that some of the effects on gene expression that we have observed here may result from alterations in the cellular level of σS. Moreover, there is some evidence that some RcsC-regulated genes might be regulated post-transcriptionally as > 40% of the genes identified by transcriptional fusion analysis were not originally identified by the array analysis. However, by analysing the list of all the genes identified in this study, we have only found three genes that have been shown previously to be regulated by σS, i.e. katE, osmB and osmC (Schellhorn and Hassan, 1988; Hengge-Aronis et al., 1991; Gordia and Gutierrez, 1996). Therefore, we do not think that the effect of RcsC activity on rprA transcription, and thus σS levels, plays a significant role in our studies.
As expected, the expression of the cps operon, encoding the proteins required for colanic acid biosynthesis and secretion, is upregulated by RcsC. The importance of colanic acid in the maturation of E. coli biofilms is well established, leading to the hypothesis that the Rcs two-component pathway will play a role in biofilm formation (Danese et al., 2000a; Prigent-Combaret et al., 2000). Given the important role that RcsC has in regulating the expression of the cps operon, it is possible that other genes positively regulated by RcsC will encode proteins that are required for the late stages of biofilm formation. One interesting locus identified during this study is the yjbEFGH locus. The yjbEFG genes were all identified as being regulated by RcsC during the array studies, whereas yjbH was identified as a λplacMu53 fusion. Although the role of this locus is unknown, homology suggests that the proteins encoded by the locus are involved in the production of polysaccharides. The gene yjbE is predicted to encode a lipoprotein, and the genes yjbFGH have homology with the otnEFG genes from Vibrio cholerae O139, a virulent serotype associated with a large outbreak of cholera in India and Bangladesh in 1992 (Bik et al., 1996; Mooi and Bik, 1997). It has been shown that the O139 serotype originated from the more common O1 serotype by the acquisition of a region of DNA, called the otn (one-three-nine) DNA, which replaced the rfb genes encoding the proteins required for the biosynthesis of O-antigen. Two genes from the otn region (otnA and otnB) have been shown to encode proteins involved in capsule biosynthesis but not lipopolysaccharide biosynthesis. However, the presence of a JUMPstart sequence in the otn DNA suggests that the other genes in this region are also involved in polysaccharide biosynthesis (Bik et al., 1996). Although preliminary studies suggest that the yjbEFGH operon is not required for E. coli biofilm formation in the wells of PVC microtitre plates, we are currently undertaking further studies to determine the role, if any, of this operon in E. coli biofilm formation.
In addition to controlling the expression of the cps genes, which are known to be involved in the late stages of biofilm formation, we have also shown that RcsC represses the expression of several genes that are involved in the production of surface appendages such as curli and some pili (see Table 2). These appendages are well known for their important role in bacterial attachment to surfaces and the initiation of biofilm formation. Antigen 43 (Ag43) is an outer membrane protein that has been shown to mediate autoaggregation in E. coli (Henderson et al., 1997a). More recently, Ag43 has been shown to contribute to biofilm formation in E. coli, probably through the mediation of cell–cell and cell–surface interactions during the development of the biofilm (Danese et al., 2000b). Ag43 is encoded by the flu (agn43) gene, and the expression of flu is regulated, in a phase-variable manner, by OxyR and Dam methylation (Henderson et al., 1997b; Haagmans and van der Woude, 2000; Wallecha et al., 2002). In our DNA array studies, we observed a 2.4-fold repression of flu expression, suggesting that RcsC also regulates the expression of this gene. Curli are surface-associated appendages that, under certain conditions, play an important role in biofilm formation in E. coli. The genes required for the biosynthesis of curli are encoded in two unlinked operons, csgBA and csgDEFG. The csgBA operon encodes the structural protein CsgA and the CsgB nucleator protein. The csgDEFG operon encodes CsgD, a transcriptional regulator of csgBA operon, and three proteins involved in the curli assembly apparatus. In our array, we observed a 2.5-fold repression in the expression of csgD when RcsC was activated by DjlA overproduction. Moreover, the Rcs two-component pathway has been shown recently to negatively regulate the expression of flhDC, and therefore the production of flagella, in E. coli (Francez-Charlot et al., 2003). However, we did not observe any effect on the expression of any of the genes known to be involved in flagella biosynthesis as we used a derivative of the non-motile E. coli strain MC4100 in our studies. MC4100 is non-motile on account of an uncharacterized mutation (called flbB5301) in the flhDC operon encoding the master regulator of flagellar biosynthesis. Our array analysis did not reveal any change in the expression of the flhDC operon during our studies, suggesting that the flbB5301 mutation is affecting the stability and/or transcription of the flhDC mRNA. Therefore, given the positive regulation of genes involved in polysaccharide production and the negative regulation of genes encoding pili and flagella, it is interesting to speculate that RcsC might be involved in controlling the transition from attached cells to mature biofilm.
The normal environmental signal(s) for RcsC activation is not yet fully characterized. In this study, we have used the overproduction of DjlA to activate RcsC, although we admit that this is unlikely to be a physiologically important signal (Zuber et al., 1995; Clarke et al., 1997; Kelley and Georgopoulos, 1997). We also show that RcsC responds to the growth of cells on a solid surface, and we confirmed, where possible, that the RcsC-regulated genes identified by DjlA overproduction were also regulated by growth on a solid surface. Therefore, the overproduction of DjlA may mimic the physiological signals encountered by bacteria growing on surfaces. In this study, we have shown that E. coli are subjected to envelope stress during growth on a solid surface. Interestingly, it has been shown previously that RcsC can be activated by alterations in the outer membrane of E. coli (Parker et al., 1992; Ebel et al., 1997). It has also been shown that RcsC transiently activates cpsB–lacZ expression during osmotic shock and dessication, conditions that might be expected to induce membrane perturbations (Ophir and Gutnick, 1994; Sledjeski and Gottesman, 1996). Moreover, we have shown previously that the overproduction of DjlA results in an increase in the sensitivity of E. coli to EDTA and novobiocin, suggesting that DjlA overproduction is affecting the integrity of the bacterial envelope (Bernard et al., 1998). Therefore, it is possible that RcsC is monitoring and responding to membrane integrity during growth on a surface. Interestingly, the Cpx two-component pathway is also associated with the bacterial response to envelope stress and membrane perturbations (Hung et al., 2001; Prigent-Combaret et al., 2001; Raivio and Silhavy, 2001). Moreover, it has been shown recently that the Cpx two-component pathway responds to the presence of a solid surface, and mutants in this pathway formed reduced biofilms through defects in bacterial adhesion to surfaces (Dorel et al., 1999; Otto and Silhavy, 2002). Therefore, it appears that membrane integrity may play a major role controlling the cellular response to growth on solid surfaces. We are currently undertaking molecular and microscopic studies to elucidate further the role of membrane integrity and RcsC in the regulation of biofilm formation.
Strains, plasmids, phages and growth conditions
The strains used in this study are derivatives of E. coli K-12 MC4100, except where stated, and are listed in Table 5 along with the plasmids and phages. Bacteria were grown at 30°C in LB broth unless otherwise stated. When required, antibiotics were added to the culture medium at the following concentrations: ampicillin (Ap) 100 µg ml−1; chloramphenicol (Cm) 20 µg ml−1; kanamycin (Km) 50 µg ml−1; tetracycline (Tet) 15 µg ml−1 and, where stated, Xgal was added to the LB agar plates to a final concentration of 50 µg ml−1.
The capacity of E. coli strains to form biofilm was assayed as described previously, with a few modifications (Pratt and Kolter, 1998). Briefly, bacteria containing the appropriate plasmids were grown overnight in LB at 30°C with agitation and diluted (1:100) in LB broth containing 0.1 mM IPTG. Aliquots of 125 µl of the diluted culture were added to the wells of a polyvinylchloride (PVC) microtitre plate, and the plate was incubated at 30°C without shaking. After 48 h, the planktonic bacteria were removed by aspiration, and the wells were washed with water. To observe biofilm formation, 200 µl of 1% (w/v) crystal violet (CV) was added to each well, and the wells were incubated at room temperature for 5 min before rinsing in tap water. To quantify biofilm formation, the CV was dissolved into 200 µl of acetone–ethanol (20:80), and CV concentration was determined by measuring the OD570 of a 100 µl sample diluted in 1 ml of H2O.
Overnight cultures of PSG1031/pPSG961-31 and PSG1038/pPSG961-31 were diluted to OD600 = 0.01 in LB broth and incubated at 30°C with aeration. The optical density of the cultures was monitored and, at an OD600 = 0.1, the cultures were supplemented with 0.2% (w/v) l-arabinose. The cultures were incubated at 30°C with aeration and, when the cultures had reached an OD600 of 0.6–0.7, the cultures were diluted in fresh LB broth, supplemented with 0.2% l-arabinose, to an OD600 = 0.1 and incubated at 30°C with aeration. At an OD600 = 0.5–0.6, 10 ml of this culture was removed and transferred into 2 ml of a 5% (v/v) phenol (pH 4.3), 95% (v/v) ethanol solution in order to stop growth and to prevent any enzymatic degradation of unstable RNA molecules during the following steps. The mixture was kept on ice for 30 min, and the cells were subsequently harvested by centrifugation at 4°C and resuspended in prewarmed lysis buffer [0.1 M sucrose, 1.5% (w/v) SDS, 30 mM EDTA, 10 mM sodium acetate, pH 4.2]. After incubation at 65°C for 3 min, one volume of hot acidic phenol (pH 4.3, 65°C) was added, and the sample was mixed vigorously, incubated at 65°C for a further 3 min, chilled on ice for 3 min and centrifuged. The aqueous phase was transferred to a fresh tube, and the procedure was repeated twice more. Two final extractions with one volume of phenol–chloroform–isoamyl alcohol (25:24:1) and one volume of chloroform–isoamyl alcohol (24:1) were then performed at room temperature. Nucleic acids were recovered by ethanol precipitation and digested with RNase-free DNase I to eliminate any contaminating DNA, and successive extractions with hot acidic phenol, phenol–chloroform–isoamyl alcohol (25:24:1) and chloroform–isoamyl alcohol (24:1) were performed, as described above. Finally, RNA was precipitated with ethanol, resuspended into diethylpyrocarbonate (DEPC)-treated water and stored at −80°C until use. The integrity of the RNA preparation was checked by electrophoresis on 1.2% agarose gel. The presence/absence of contaminating DNA was tested by the amplification of the plasmid-borne cat gene by PCR.
Radiolabelled cDNA was generated by mixing 1 µg of total RNA with 4 µl of E coli cDNA labelling primers (Sigma) in a final volume of 16 µl. The mixture was heated at 90°C for 2 min and slowly cooled to 42°C by decreasing the temperature gradually over 10 min. An aliquot of 14 µl of reaction mix (50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl2, 10 mM SDS, 667 µM dATP, 667 µM dTTP, 667 µM dGTP) containing 20 µCi of [α-33P]-dCTP and 200 U of SuperScriptTMII retrotranscriptase (Invitrogen) was added, and the reaction was incubated at 42°C for 2 h before heating to 95°C for 5 min to denature the enzyme. Unincorporated radionucleotides were removed by purification over a Sephadex® G-50 gel filtration NickTM column (Pharmacia Biotech).
Genome-wide analysis was carried out using the PanoramaTME. coli gene arrays nylon membranes (Sigma-Genosys), and radiolabelled cDNA was hybridized to the membrane according to the manufacturer's instructions. Briefly, membranes were rinsed in 2× SSPE buffer (0.18 M NaCl, 1 mM EDTA, 10 mM sodium phosphate, pH 7.7) for 5 min and prehybridized at 65°C in 5 ml of hybridization solution [5× SSPE, 2% (w/v) SDS, 0.02% (w/v) Ficoll (MW 400 000), 0.02% (w/v) PVP (MW 10 000), 0.02% (w/v) BSA, 100 µg ml−1 sonicated, denatured salmon testes DNA] for at least 1 h. The solution was discarded, and 3 ml of hybridization solution containing the denatured 33P-labelled cDNA probe was added. After hybridization overnight at 65°C, the membrane was washed three times at room temperature in a 0.5× SSPE-2% (w/v) SDS solution followed by three washes in the same solution at 65°C for 20 min. The membrane was finally wrapped in plastic film and exposed overnight to a phosphorimager screen (Fujifilm imaging plate BAS-MS).
The exposed phosphorimager screen was scanned at a 50 µm pixel resolution on a Fujifilm FLA-5000 analyser, and the signal intensity of each spot was quantified with the imageneTM software (BioDiscovery). Each gene was spotted, in duplicate, on the PanoramaTME. coli gene arrays, defining two independent populations of 4290 spots. The pixel density of each spot was expressed as a percentage of the sum of the density of all the spots from the same population, providing two values for each gene in a single experiment. In this study, we hybridized each of two RNA samples twice to two different membranes and thus obtained eight independent data sets for each gene; these were compiled for analysis. The normalized pixel densities were imported into, and analysed with, the Microsoft®excel workbooks designed for the analysis of Sigma-Genosys PanoramaTME. coli gene array data in a Microsoft® Windows environment (http://www.ou.edu/microarray; Conway et al., 2002). Briefly, the eight normalized signal values obtained for each gene were averaged for both conditions (rcsC+ or rcsC– backgrounds), and the average values were used for calculation of the induction ratio (IR) for each gene. The IR values were calculated so that, if expression levels were higher in the rcsC+ background, the IR value was positive and, if the expression level was higher in the rcsC– background, the IR value was negative. Two statistical approaches were then combined to identify the genes that are expressed differently in the conditions tested. First, the mean and standard deviation of the whole population of log10-transformed IR were calculated, and only the genes for which the log10(IR) absolute value was greater than the mean plus 3 standard deviations were selected. These genes (outliers) were removed from the population, and the mean and standard deviation were calculated again. This procedure was repeated six times until a single outlier appeared (Loos et al., 2001). The Student t-test was then applied to compare the sets of natural log-transformed expression values obtained in both conditions. The outlier genes displaying a P-value of < 0.05 were selected as RcsC-regulated genes.
Screening of a promoter library
In order to identify the RcsC-dependent promoters, chromosomal DNA fragments of MC4100 were cloned in the pFZY1 vector as follows: 50 µg of genomic DNA was partially digested with 0.08 U µl−1Sau 3A by incubation at 37°C for 30 min. The reaction was stopped by the addition of 50 mM EDTA and loaded on a 1.2% (w/v) agarose gel. DNA fragments ranging from 0.5 kb to 1 kb were purified and introduced into the BamHI site of pFZY1, upstream from the promoterless lacZ gene. This promoter library was then transformed into strain PSG1066 carrying pPSG961-31, and the cells were plated out on selective LB. Approximately 1000 transformants were streaked on LB agar plates containing Xgal in the presence or absence of 0.2% (w/v) arabinose and incubated overnight at 30°C, and colonies displaying the relevant phenotype (Lac+ on one plate, Lac– on the other plate) were purified. In order to confirm RcsC-dependent regulation, the pFZY1-derived vector containing the candidate promoter (pUP) was isolated and transformed into isogenic rcsC+ (PSG1032) and rcsC– (PSG1049) backgrounds. The chromosomal fragment inserted in each pUP plasmids was amplified by PCR using the primers DC172 (5′-ATAGCGAATTCGAGCTCGGTACC) and DC173 (GCGTT AAGCTTGCATGCCTGCAGG) and sequenced.
Isolation of λp lacMu53 insertions in the chromosome
A library of λplacMu53 insertions was constructed as described previously (Bremer et al., 1985). Briefly, 1 ml of a fresh overnight culture of PSG1032, carrying the plasmid pPSG961-31, was grown in LB broth supplemented with 10 mM MgSO4 and 0.2% (w/v) maltose, and co-infected with λplacMu53 (108 pfu) and the λpMu507 helper phage (109 pfu). After incubation at room temperature for 30 min, 5 ml of LB broth was added, and unabsorbed phage particles were removed by centrifugation. This washing step was repeated three times, and the cells were finally resuspended in 1 ml of LB broth. After phenotypic expression at 37°C for 1 h, 100 µl of a serial dilution (10−1−10−4) of the sample was plated on selective medium containing Km and Cm, and the plates were incubated overnight at 37°C. The integration of λplacMu53 in the appropriate orientation results in a transcriptional fusion between the phage-borne lacZ gene and the interrupted gene. Therefore, λplacMu53 insertions in genes regulated by DjlA overproduction were identified by streaking 1200 colonies in duplicate on Xgal-containing LB agar plates with or without 0.2% (w/v) l-arabinose. The plates were incubated at 30°C overnight, and colonies that appeared blue (Lac+) on one plate and white (Lac –) on the other were purified. To identify insertions that were occurring in genes regulated by arabinose itself (false positives), we determined the effect of arabinose on lacZ expression in the absence of the pPSG961-31 plasmid. In order to ensure that a single insertion was present in the cell, the mutation was transferred to a clean genetic background using transduction with P1cml. The rcsC52::Tn10 mutation was finally introduced in the remaining strains by P1cml transduction, and the RcsC-dependent regulation was confirmed by measuring the β-galactosidase activity in overnight cultures of the rcsC+ and rcsC– derivatives, grown in LB broth containing 0.2% (w/v) l-arabinose and Cm.
The site of λplacMu53 insertion was determined by arbitrary-primed PCR (O’Toole et al., 1999). A sample of 5 µl of an overnight culture of the mutant was added to 50 µl of a PCR mix [20 mM Tris-HCl, pH 8.8, 10 mM KCl, 10 mM (NH4)2SO4, 2 mM MgSO4, 0.1% (v/v) Triton X-100, 0.1 mg ml−1 BSA, 200 µM each dNTP, 0.04 U µl−1Pfu:Taq polymerase mixture (10:1, v/v)] containing 300 nM randomized primers LFARB1 (5′-G G C C A C G C A T CG ACTAGTACNNNNNNNNNNGATAT) and LFARB2 (5′-GGCCACGCATCGACTAGTACNNNNNN NNNNACGCC) and 600 nM LF13 (5′-AATTTGCACTACAGG CTTGC) primer, which was designed to complement the 3′ end region of λplacMu53. The amplification conditions were as follows: 5 min at 96°C; six cycles of 30 s at 94°C, 30 s at 30°C, 3 min at 72°C; 30 cycles of 30 s at 94°C, 30 s at 45°C, 3 min at 72°C; 5 min at 72°C. A second round of PCR was subsequently performed after dilution of 2.5 µl from the first PCR into 50 µl of a fresh PCR mix. Amplification (40 cycles of 30 s at 94°C, 30 s at 60°C, 3 min at 72°C; 5 min at 72°C) was carried out with 600 nM LF9 (5′-GGCCACGCATCGAC TAGTAC) and 600 nM LF12 (5′-CCCGAATAATCCAATGTC CTCCCGG) as primers. Products were purified by migration on 2% agarose gel and sequenced using the LF12 oligonucleotide as a primer.
Gene expression assay for growth on solid surface
Bacteria were grown at 30°C in LB broth until mid-exponential phase (OD600 off 0.5). Cells were harvested by centrifugation at room temperature and resuspended in PBS to a final OD600 of 1.0. This cell suspension was further diluted 10 times in PBS, and 1 ml of the dilution was filtered on sterile nitrocellulose membrane (0.45 µm, WCN type; Whatman). After two washes of the cells with 10 ml of PBS, the membrane was placed on a prewarmed LB agar plate, supplemented with antibiotics when appropriate. Two membranes per assay were prepared in this way. After incubation at 30°C, cells were recovered by vigorously washing the two membranes in 10 ml of LB broth, and the cell suspension was stored on ice. Gene expression was assayed by measuring the β-galactosidase activity.
β-Galactosidase activities were measured on CHCl3-SDS permeabilized cells using the colorimetric method described previously (Miller, 1972).
We would like to thank Roberto Kolter (Harvard Medical School) for the gift of strain ZK2686, and Tom Silhavy (Princeton University) for strains TR235 and TR238. We would also like to thank Kaymeuang Cam (CNRS, Toulouse, France) for sharing unpublished data, and other members of the D.J.C. laboratory for useful discussions. This work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC grant number 86/P16371).