INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

In natural environments, bacteria are often found as sessile communities known as biofilms. Biofilms are defined as matrix-enclosed communities of microorganisms tightly interacting with each other and, in most cases, supported by an abiotic surface (7). Biofilms can become hundreds of micrometers in depth and display complex structural and functional architecture (8, 28, 40). Bacteria growing as a biofilm develop significant phenotypical, biochemical, and morphological differences from their planktonic counterparts. In particular, cells growing in biofilm are extremely resistant to treatment with biocides and to prolonged antibiotic therapy in human infections (16, 23). This different biochemical and phenotypical behavior reflects different patterns of gene expression compared with planktonic cells (39, 41). Such reprogramming of gene expression is likely to be due to changes in environmental physicochemical conditions and to involve two-component systems such as EnvZ/OmpR (41) and CpxA/CpxR (15).

In previous reports we described the isolation of a mutant from a continuous culture of Escherichia coli K-12 that is more efficient in biofilm formation. Using both genetic analysis and electron microscopy observations, we showed that this increased efficiency was due to augmented production of curli (40, 50). The genes necessary for curli production are clustered in the csgBA and csgDEFG operons, which encode the curli subunits and regulate their transcription and transport, respectively. The csgBA and csgDEFG operons appear to be expressed in environmental and clinical isolates of E. coli, as well as in Salmonella strains, in which the homologous genes are called agfBA and agfDEFG.

However, some E. coli K-12 laboratory strains do not express curli, although functional copies of the genes are still present (5, 19, 35). The reason for this different behavior in E. coli K-12 strains is not yet fully understood. Curli fibers are highly conserved between Salmonella species as well as in E. coli with respect to curlin amino acid sequence, genetic organization, and operon regulation (6, 44). Curlin, the product of csgA, is the major component of curli, while CsgB acts as a nucleator which primes the polymerization of curlin on the cell surface (20). The first gene of the csgDEFG operon encodes the CsgD protein, a putative transcription factor belonging to the luxR family and required for the transcription of csgBA (19). The csgEFG genes encode three curlin assembly factors, probably involved in export of the curli subunits (19).

Curli biogenesis is subject to tight and complex regulation: in E. coli K-12 and in Salmonella enterica serovar Typhimurium, they are only produced at temperatures below 30°C, at low osmolarity, and in stationary phase (19, 30, 35). The stationary-phase-induced transcription of genes required for curli biogenesis is dependent on the  factor RpoS (35). However, in the absence of the H-NS histone-like protein, transcription from the csgDEFG promoter becomes independent of RpoS (3), suggesting that H-NS might selectively repress ⁷⁰-dependent transcription of csgD. The two-component systems OmpR/EnvZ (44, 50) and CpxA/CpxR are also implicated in the regulation of curli biogenesis (15). The OmpR/EnvZ system constitutes a signal transduction pathway that senses external osmolarity and regulates the transcription of several genes, including the porin-encoding genes ompF and ompC (38). There is genetic evidence that curli-encoding genes are members of the OmpR regulon (44, 50), but the existence of direct transcriptional control has never been demonstrated.

In this report we show that curli biosynthesis is subject to a complex regulatory network: the OmpR protein positively regulates curli expression by binding the csgDEFG promoter region at position 49.5 relative to the transcriptional start site and by activating its transcription. However, csgDEFG and csgBA expression is also subject to negative regulation by the rpoS gene. Negative control by rpoS appears to be mediated by direct interaction between the CpxR protein and both the csgD and csgB promoters. The CpxRA pathway is induced in response to damage of envelope proteins, such as during exposure to elevated pH (10, 33), and to alteration of the inner membrane lipid composition (11, 31). Activation of the Cpx pathway results in the production of factors involved in protein folding and degradation, such as the two peptidyl-prolyl-isomerases PpiA (9) and PpiD (12), DsbA, and the protease DegP (9). However, CpxR also represses motility and chemotaxis genes (14) and is involved in regulation of P pili (26), indicating that the Cpx pathway could play a role in other cellular processes as well. We propose a new model that integrates the complex regulatory networks controlling curli biogenesis.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, and media. The E. coli strains and plasmids used in this work are listed in Table 1. Media used were Luria-Bertani broth (LB), minimal M63 medium supplemented with glucose (0.2%) (32), and M63/2, a low-osmolarity medium obtained by twofold dilution of the M63 medium and supplementation with glucose (0.2%). Congo red indicator plates were prepared as described by Hammar et al. (19); on these plates, curli-producing bacteria form red colonies, whereas non-curli-producing cells produce white colonies.

Genetic methods. Phage P1 transductions were carried out as described by Miller (32). The ompR234 mutation was transferred by using its genetic linkage (50% cotransduction) with malA, followed by screening of adherent transductants in 24-well microtitration plates. Transduction of the rpoS::Tn10 allele was obtained by selection on tetracycline plates and verification of the loss of catalase activity (29).

Enzyme assays. -Glucuronidase specific activity in toluene-treated samples was measured by spectrophotometrically monitoring the hydrolysis of p-nitrophenyl--D-glucuronide into p-nitrophenol at 405 nm (4). Specific activity was expressed as units per milligram of protein, where 1 U corresponds to 1 nmol of product liberated per min (40). -Galactosidase activity was measured by following the degradation of o-nitrophenyl--D-galactoside into o-nitrophenol, which absorbs at 420 nm (32). Specific activity was expressed as nanomoles of product liberated per minute per milligram (dry weight) of bacteria. A minimum of four independent assays were performed, and the results were averaged to obtain the indicated activities. Error bars indicate the standard deviation.

Adhesion and biofilm formation assays. Determination of biofilm thickness in microtiter plates was carried out as described by Dorel et al. (15). The ratio between surface-attached and unattached bacteria was estimated by measuring the optical density at 600 nm (OD₆₀₀). At least three independent assays were performed and averages were calculated. To determine initial attachment to a solid surface, we used the sand column method described by Simoni et al. (48). Bacteria were grown in the appropriate medium, harvested, washed, and resuspended in phosphate-buffered saline (PBS) to an A₂₈₀ of 1.0 (corresponding to ca. 5 × 10⁸ bacteria); however, the adhesion properties of bacteria in the sand column assay depend strictly on the growth medium used (48) (data not shown). The suspension was loaded onto a fine sea sand grain column (9 g of sand). The bacterial concentration in the fractions collected at the column outlet was determined spectrophotometrically and used to calculate the percentage of attaching bacteria. The accuracy of spectrophotometric measurements was confirmed by direct plate counts (data not shown). Microscopic analysis of the column sand grains shows that bacteria attach as single cells in the conditions used in our experiments and that the cell sizes of the different strains are comparable (data not shown).

Construction of a csgD::uidA fusion. To obtain a csgD::uidA chromosomal fusion, a 3,642-bp DNA fragment corresponding to the whole csg region was amplified by PCR from MG1655 chromosomal DNA as the template and by using primers C1 (5'-CGA ATA ATC TTG CGG TCG ACA AGC AGG-3') and C2 (5'-GAA AGT GCC GCA AGG AGC TCT AAC G-3'), which contain, respectively, SalI and SacI restriction sites (italic sequences). The PCR fragment digested with SalI and SacI was cloned into the corresponding sites of the vector pBC (Stratagene) to give pBCcsg. The 3.8-kb SmaI fragment containing a uidA-kan cassette (pN496) (25) was cloned into the unique EcoRV site of the csgD gene carried by pBCcsg, producing pBCcsgD::uidA. The correct csgD::uidA orientation was confirmed by restriction digestion.

Primer extension analysis of transcript. Total RNA was isolated from E. coli cells grown to an OD₆₀₀ of 0.2 (2.5 × 10⁹ CFU/ml) or 1 (10¹⁰ CFU/ml) in M9/glucose medium at 28°C, as described by Sambrook et al. (47). For csgB transcript analysis, we used the 5'-CCCAGGCGCACCCAGTATTGTT-3' primer, which anneals to the coding strand between 117 and 139 nucleotides downstream of the csgB gene transcription start. The sequence of the primer used for csgD transcript analysis was 5'-AAGATTTAGTGATCAACAATAATG-3', annealing to nucleotides +181 to +203 of csgD. The primers were labeled with the fluorescent dye IRD-800 at the 5' end. Extension products were run on a sequencing gel and densitometrically analyzed in a 4000L automated sequencer (Li-Cor Inc., Lincoln, Neb.).

Overproduction and purification of the OmpR and OmpR234 proteins. The coding regions of OmpR and OmpR234 were amplified by PCR from chromosomal DNA of, respectively, MG1655 and MG1655 ompR234 strains and by using the primers R1 (5'-AGTACAAACCATGGAAGAGAACTAC-3') and R2 (5'-CTTCGTACGCGAAAGCTTTATTAAACTG-3') carrying, respectively, an NcoI and an HindIII site (italic sequences). The presence of an NaeI cutting site in the ompR234 but not in the ompR amplified fragment (50) was checked. The 864-bp NcoI-HindIII fragments were then subcloned in the NcoI and HindIII unique sites of the plasmid cloning vector pKK233-2. The resulting plasmids, pKKompR and pKKompR234, contain fusions of the ompR and ompR234 start codons with the ATG start codon of the strong regulated trc promoter, placing ompR and ompR234 under the transcriptional and translational regulatory signals of the trc promoter.

Overproduction and purification of the CBP-CpxR protein. To construct the plasmid producing the calmodulin binding peptide (CBP)-CpxR fusion protein, a 700-bp region corresponding to the CpxR open reading frame was amplified by PCR using E. coli MC4100 genomic DNA as the template with primers cpx+ (5'-TATTTAAACCATGGATAAAATC-3') and cpxrev (5'-CTATCATGAAGCTTAAACCATC-3') carrying, respectively, an NcoI and an HindIII site. The amplified DNA was digested with NcoI and HindIII and subcloned into the corresponding restriction sites of pCAL-n (Stratagene), generating pCAL-n-cpxR, which was subsequently introduced into the BL21 strain. The pCAL vectors use the T7 lac promoter configuration and contain a copy of the lacI gene. On induction with 1 mM IPTG, the lacUV5 promoter was derepressed, allowing overexpression of T7 RNA polymerase and expression of the T7-promoted cpxR gene. Crude protein extracts were obtained as described above and incubated overnight with calmodulin affinity resin in CaCl₂ binding buffer, according to the instruction manual (Stratagene). Washes and elution were performed as recommended by the manufacturer.

Gel retardation assay. DNA probes containing either the csgDEFG promoter region (128 to +12) or the same region with deletions of the putative OmpR- and CpxR-binding sites, were obtained by PCR amplification from pCSG4. For the wild-type promoter, PCR amplification was performed by using primers D1 (5'-CCAAATGTACAAGCTTTCTATCATTTC-3') and D2 (5'-GGATTACATCTGATTTCAATCTAGCC-3'). For the promoter with a deletion of the putative OmpR binding site, a 119-bp fragment was amplified using two primers, D1 and D3 (5'-GGATTACATCTGATTTCAATCTAGCCATTACAAATCTTAAATCAAGTGTTCTCGTTATATTAAAATG-3'). The D3 primer sequence is identical to the sequence of D2 for its first 25 bases and then contains a 20-bp deletion corresponding to the putative OmpR binding site (Fig. 1). The sequence of the deleted operator was checked by automated sequencing using D1 primer (Genome Express France). A DNA probe containing the csgB promoter region was obtained by PCR amplification from pCSG4 using D4 (5'-CTGTCTGAAGCTTTTTGATAGCGGAAAACGG-3') and D5 (5'-CACCCTGGACCTGGTCGTACATTTAA-3') primers. All three operator fragments were digested with HindIII and then ³²P-labeled by the Klenow fragment of DNA polymerase.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

OmpR protein promotes curli production by activation of csgD promoter. As shown in a previous report, the presence of the ompR234 allele favors biofilm formation by stimulating the expression of the csgBA operon and the production of curli (50). However, it is not clear if OmpR234 activates csgBA by binding to the csgB promoter or in an indirect fashion. Regulation of the csgBA operon is fairly complex and involves the transcription activator CsgD, the product of the first gene in the csgDEFG operon (3, 44). In order to establish whether CsgD alone is sufficient for csgB activation, we overexpressed the CsgD protein in an OmpR strain.

View larger version (21K): [in this window] [in a new window]   FIG. 1.   Representation of the csgD-csgBA intergenic region in E. coli strain MG1655 (AE000205). The position of the transcriptional start site is indicated by the small arrow. The 10 and 35 regions are underlined. The putative OmpR binding site is boxed. The perfect CpxR recognition consensus sequence is encircled with a bold line, whereas the consensus sequence with a 1-bp mismatch is encircled with a thin line. The consensus mismatch is indicated by a solid circle. Primers D1, D2, and D3 used for amplification of the wild-type csgD promoter and of the promoter with deletions of the OmpR and CpxR boxes are indicated by the big arrows. The positions of primers D4 and D5, used for amplification of the wild-type csgB promoter, are indicated.

View larger version (29K): [in this window] [in a new window]   FIG. 2.   Electrophoretic mobility shift analysis of the csgD promoter region with OmpR-P. (A and B) Increasing amounts of FPLC-purified OmpR (A) and OmpR234 (B) protein were incubated with the wild-type csgD promoter (D1D2): no protein (lane 1), 730 nM (lane 2), 1.8 µM (lane 3), and 3.6 µM (lane 4) OmpR protein; lane 5, the same as lane 4 with a 40-fold excess of competing unlabeled D1D2 DNA. C, OmpR-DNA (A) or OmpR234-DNA (B) complex. (C) Specific binding to the 20-bp F1-box centered at position 49.5; the deleted (D1D3) operator was used as the probe in this experiment.

ompR234 mutation allows biofilm formation at high osmolarity by transcriptional enhancement of both csg operons. Expression of the csgD promoter in E. coli and of agfD in S. enterica serovar Typhimurium is negatively affected by high osmolarity, consistent with a role of the envZ/ompR system in their regulation (35, 44). We tested the effects of the osmotic conditions in the growth medium on biofilm formation and on csgDEFG and csgBA expression levels. Osmolarity variations were achieved by adding increasing amounts of osmolytes such as NaCl (Fig. 3) and sucrose (data not shown) to low-osmolarity minimal M63/2 medium. The results of increased medium osmolarity on biofilm formation, measured as crystal violet staining and from OD₆₀₀ measurement of cells attached to microtiter plates, are shown in Figure 3.

View larger version (26K): [in this window] [in a new window]   FIG. 3.   Negative effect of osmolarity on biofilm development in microtiter plates. The osmolarity of minimal M63/2 medium supplemented with glucose at different NaCl concentrations was measured with a Fiske OS/220 osmometer. Biofilm visualization by crystal violet staining was performed as previously described (15, 50). The thickness of the biofilm was quantified as follows. For each well, two washes were pooled with the initial supernatant and are referred to as swimming cells; the biofilm was recovered in 1 ml of M63 by scraping and pipetting up and down. The numbers of surface-attached and swimming bacteria were estimated from the OD600 to give the adherence percentage corresponding to each osmolarity condition and each bacterial strain. The total growth in each well reached 1.5 OD600 U except in the presence of 0.3 M NaCl. In this particular condition, only 1.2 OD600 U could be attained. The formation of biofilm was inhibited by increasing NaCl concentrations in MG1655 (wild-type ompR and ompR234) and clinical strains. Similar trends were seen when sucrose was used as an osmotic agent (data not shown). Whatever osmolarity cells encounter, knocking out the csgA gene in MG1655 or in medical isolates results in a nonadherent phenotype (40) (data not shown).

View larger version (23K): [in this window] [in a new window]   FIG. 4.   Negative effect of high medium osmolarity on transcription of the curli genes. (A) csgA::uidA gene fusion. (B) csgD::uidA gene fusion. The transcriptional level of each fusion was compared in wild-type ompR (open squares), ompR234 (solid squares), and ompR::Tn10 (open circles) backgrounds. Increasing amounts of NaCl were added to minimal M63/2 medium supplemented with glucose, as in Fig. 3. Bacterial growth was similar in all enzymatic assays (1.5 OD600 U) except in the presence of 0.3 M NaCl. In this particular condition, only 1.2 OD600 U could be attained. Results are means and standard deviations from four independent -glucuronidase assays. The compilation of data regarding adherence (Fig. 3) and expression of the csgA::uidA fusion (this figure) suggests that adherence does not occur below a transcriptional threshold of 800 U/mg of protein.

ompR234 mutation increases initial adhesion to solid substrates in a stationary-phase-dependent but rpoS-independent manner. The formation of biofilm takes place in several steps, which include initial attachment to a solid surface, formation of a microcolony, and differentiation into a complex structure (7). To test if the ompR234 mutation affects initial adhesion, we used the sand column system described by Simoni et al. (48), which provides a simple and direct method to measure attachment. Cells grown overnight in minimal medium were loaded onto sand columns after being washed and resuspended in PBS. As shown in Fig. 5, for strain MG1655 (wild-type ompR) grown to the stationary phase, only around 30% of the cells attached to the column. This shows that strain MG1655 has a weak ability to adhere to sand grains. In sharp contrast, for the otherwise isogenic strain MG1655 ompR234, more than 70% of the cells adhered to the sand grains.

View larger version (32K): [in this window] [in a new window]   FIG. 5.   Adhesion experiments in sand columns. The percentage of attaching bacteria was determined as described in Materials and Methods. Cells were grown in M9/glucose and harvested either in the exponential phase (OD600 of 0.2, corresponding to circa 2.5 × 109 CFU/ml) or in the stationary phase (OD600 of 1.0, corresponding to circa 1010 CFU/ml). Cells were resuspended in PBS to a final OD280 of 1, corresponding to 2 × 109 CFU/ml, and the same number of cells were loaded in each experiment. For each experiment, seven measurements were taken. Data shown are the averages of four independent experiments. WT, wild type.

View larger version (35K): [in this window] [in a new window]   FIG. 6.   Primer extension from csgD and csgBA mRNAs. The experiments were performed with 10 µg of total RNA. Lane 1, wild-type (MG1655). Lane 2, ompR234 (PHL628). Lane 3, rpoS (EB1.3). Lane 4, ompR234 rpoS (EB2.16). Cells were grown in M9/glucose and harvested either in the exponential phase (OD600 = 0.2, corresponding to circa 2.5 × 109 CFU/ml) or in the stationary phase (after 16 to 18 h of growth; OD600 > 1, corresponding to 1010 CFU/ml). The sizes of the transcripts from both csgD and csgB, determined using a sequencing ladder from plasmid pUC19 as a molecular weight marker, were consistent with the previously proposed start sites (44).

Negative regulation of curli operon by transcription regulator CpxR. In a previous paper, we have shown that the two-component regulatory system CpxA/CpxR negatively affects csgBA expression (15). Both the csgD and csgBA promoters display sequences with high similarity to the proposed binding site for CpxR (36). To investigate the possibility of a direct role of CpxR in curli regulation, we purified CpxR and assayed its specific DNA binding to the csgD and the csgB promoter regions. Indeed, CpxR binds specifically to both the csgD and the csgB promoters (Fig. 7), showing that CpxR is directly involved in the downregulation of curli expression.

View larger version (38K): [in this window] [in a new window]   FIG. 7.   Electrophoretic mobility shift analysis of the csgD and csgB promoter regions with CBP-CpxR. Mobility shift assays were performed with pure CBP-CpxR and 32P-end-labeled csgD (left) and csgB (right) promoter fragments. (A) pcsgD. D1D2 probe without protein (lane 1); D1D2 probe with 300 nM CBP-CpxR (lane 2); same as lane 2 but challenged with unrelated binding site (100-fold excess of calf thymus DNA, lane 3); same as lane 2 but with a 40-fold excess of unlabeled competing D1D2 DNA (lane 4). pcsgD. D1D3 probe without protein (lane 5) and with 300 nM CBP-CpxR (lane 6). (B) pcsgB. D4D5 probe without protein (lane 7); D4D5 probe with 300 nM CBP-CpxR (lane 8); same as lane 8 but challenged with 100-fold excess of calf thymus DNA unrelated binding site (lane 9); same as lane 8 but with a 40-fold excess of competing unlabeled D4D5 DNA (lane 10). Arrows C, CBP-CpxR-DNA complex.

View larger version (18K): [in this window] [in a new window]   FIG. 8.   cpxP-lacZ transcription is induced by curli overproduction (A) and high osmolarity (B). (A) -Galactosidase activities were determined for strains TR50 (MC4100 RS88[cpxP-lacZ]) (lane 1) and PHL1152 (MC4100 RS88[cpxP-lacZ]) (lanes 2, 3, and 4). (A) -Galactosidase activities of strains transformed either with pUC19 (control for pCSG4, lane 3) or pCSG4 (overexpresses curlin, lane 4). All strains were grown in Luria broth at 30°C. Bacterial growth was similar in all enzymatic assays (1.5 OD600 U). Similar results were obtained in minimal M63 medium supplemented with glucose. (B) -Galactosidase activities were determined for the PHL1152 strain in the presence of increasing amounts of osmolytes (NaCl and sucrose) added to minimal M63/2 medium supplemented with glucose. The total growth in each culture reached 1.5 OD600 U except in the presence of 0.3 M NaCl. In this particular condition, only 1.2 OD600 U could be attained.

View larger version (42K): [in this window] [in a new window]   FIG. 9.   Model of the regulatory network controlling biofilm formation in E. coli. Curli production at low osmolarity results from transcriptional activation of the csgD promoter by OmpR and OmpR234, OmpR234 appearing to be a much more efficient activator than OmpR. In response to Cpx pathway activation via either RpoS, high osmolarity, curlin accumulation, or a combination of these factors, transcription of the two csg operons is repressed. The different controls are indicated by arrows for positive controls or by a line with a bar for negative controls.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The genetic organization and transcriptional regulation of curli-related operons csgBA and csgDEFG are highly conserved in E. coli and S. enterica serovar Typhimurium strains (44). Expression of these genes responds to different environmental signals and is positively regulated by ompR and rpoS (44). However, in strains MC4100 and MG1655 of E. coli, the expression of the csgBA genes is negligible despite the presence of functional ompR and rpoS genes. In a previous report, we showed that a G-to-T mutation in the ompR gene (ompR234 mutation), corresponding to a leucine-to-arginine substitution at position 43 of the OmpR protein, resulted in an increase in the expression of curli (50). In this paper, we have shown that increased curli production is mediated by OmpR234-dependent stimulation of transcription of the csgDEFG operon (Table 2). Increased production of the CsgD transcription activator results in activation of the csgBA operon; when CsgD is expressed independently of OmpR, the latter becomes dispensable for activation of the csgB promoter (Table 2).

Both genetic and biochemical evidence shows that OmpR binds to an imperfect direct repeat (5'-GTTACATTTA/GTTACATGTT-3') centered at position 49.5 relative to the transcriptional start site of the csgD promoter (Table 2 and Fig. 1 and 2). This sequence is very similar to the consensus for OmpR binding sites proposed by several groups (21, 22, 37). In vitro binding experiments show that both the OmpR wild-type and OmpR234 mutant proteins can recognize this sequence, although OmpR234 displays a slightly higher binding affinity than the wild-type protein, which might be enough to improve activation of csgD transcription. The OmpR234 protein might be better able to counteract the negative regulation of csgD by the CpxR protein (15) (Fig. 7A), either by competing for the same binding site or by inducing structural changes in the csgD promoter region. Alternatively, the leucine-to-arginine substitution in the OmpR234 protein (L43R) might improve its interaction with RNA polymerase.

The observation that, unlike the situation in other E. coli strains and in S. enterica serovar Typhimurium (2, 44), the rpoS gene is no longer necessary for csgD transcription in the ompR234 strain (Fig. 6) might suggest that the L43R substitution results in a better interaction of OmpR with ⁷⁰ RNA polymerase. Interestingly, a single-nucleotide substitution in the putative binding site for OmpR in the S. enterica serovar Typhimurium agfD promoter region also makes agfD transcription rpoS independent (45).

The ompR gene is part of the ompR/envZ two-component regulatory system that senses osmolarity; thus, we investigated the possibility that biofilm formation might be regulated by osmolarity via the ompR/envZ pathway. Increasing NaCl or sucrose concentrations in the growth medium resulted in the inhibition of biofilm formation and decreased in vivo expression of the csgDEFG and csgBA operons, as measured with csgD::uidA and csgA::uidA chromosomal fusions (Fig. 3 and 4). Inhibition by NaCl or sucrose was observed in laboratory strains as well as in clinical isolates (Fig. 3), suggesting that osmolarity regulation of biofilm formation is broadly conserved in E. coli strains. Inhibition of curli production in the MG1655 strain PHL565 was observed at NaCl concentrations lower than those present in commonly used growth media (Fig. 3 and 4), providing an explanation for the previously reported lack of curli expression in laboratory strains (6, 19, 34).

Increasing osmolarity of the growth environment activates the sensor protein EnvZ and leads to increased phosphorylation of OmpR, resulting in enhancement of its DNA-binding ability (1, 13, 17, 24, 46). However, increased phosphorylation of OmpR results in activation of transcription only at certain promoters, such as ompC, while other genes, such as ompF, are downregulated. Repression of ompF transcription is due to the binding of phosphorylated OmpR (OmpR-P) to multiple binding sites, with the consequent inhibition of RNA polymerase binding to the ompF promoter (49). This mechanism might also apply to the regulation of csgD, although we were not able to identify multiple binding sites for OmpR (Fig. 1). It is possible that the negative regulation of csgD expression at high osmolarity is mediated by CpxR, since high osmolarity does indeed activate the Cpx pathway (Fig. 9). Control of cpxR expression by osmolarity would be consistent with the observation that increased osmolarity also results in a further decrease in the basal levels of csgBA transcription in a csgD mutant strain (Table 2).

Adhesion experiments using sand columns strongly suggest that the stage of biofilm formation positively affected by the ompR234 mutation is the initial adhesion to a solid surface (Fig. 5). Initial adhesion experiments also showed that the ompR234 mutation stimulates adhesion only in stationary-phase cells (Fig. 5), consistent with the simultaneous increase in csgB expression (Fig. 6). However, this stationary-phase-specific effect is not mediated by the master regulator RpoS, which, on the contrary, appears to negatively regulate csgBA expression in an ompR234 strain (Fig. 6).

We propose that negative regulation of csgB transcription by rpoS is due to rpoS-dependent transcription of cpxR (14). Interestingly, the CpxR protein binds both csgD and csgBA promoter regions, where it acts as a repressor (Fig. 7) (15). Lack of cpxR transcription in the rpoS-deficient strain would allow increased csgB transcription in the stationary phase. Our observations reiterate the importance of different regulatory networks in the regulation of the curli operon in E. coli and suggest that the CpxRA pathway plays a major role in the expression of virulence factors such as curli (this report) and P pili (26). The extreme complexity of the regulation mechanisms is likely to reflect the importance of finely tuning the expression of adhesion genes for survival of the bacterium in different environments.