The pst operon of enteropathogenic Escherichia coli enhances bacterial adherence to epithelial cells

Enteropathogenic Escherichia coli (EPEC) is one of the most important bacterial causes of infant diarrhoea in developing countries. EPEC is known for its ability to cause attaching-effacing (A/E) lesions, which is histopathologically characterized by an intimate adherence of the bacteria to the brush-border microvilli, and the formation of pedestal-like structures on the epithelial tissue beneath the bacterial adherence sites (Nataro & Kaper, 1998). In the A/E lesions, bacteria adhere intimately through the adhesin intimin that binds to the intimin receptor (Tir) previously exported to the host cell (Kenny et al., 1997). Another EPEC characteristic is the formation of microcolonies on cell monolayers in vitro, a pattern known as localized adherence (LA) (Scaletsky et al., 1984).

Most genes implicated in the A/E lesions are situated in a 35 kb pathogenicity island known as the locus of enterocyte effacement (LEE). This region harbours five operons containing 41 genes divided into three functional regions. LEE1, 2 and 3 encode proteins involved in the biogenesis of a type III secretion system (TTSS), which is a multi-protein complex that spans both bacterial membranes and is used to transfer effector proteins into the host cell. LEE5 contains the eae and tir loci that encode, respectively, the adhesin intimin and Tir. LEE4 encodes the Esp proteins, which are secreted by a TTSS (Mellies et al., 1999).

Typical EPEC strains carry a large plasmid known as pEAF (plasmid of EPEC adherence factor). At least two operons (bfp and per) present in pEAF are needed to confer on EPEC the LA phenotype. The bfp operon contains 14 genes related to the biogenesis of the bundle-forming pilus (BFP), a type IV fimbria found in typical EPEC strains (Stone et al., 1996). The first gene of the operon, bfpA, encodes the main subunit of the fimbria (Donnenberg et al., 1992). PerA, encoded by the perABC operon, is responsible for the positive regulation of bfp expression (Tobe et al., 1996). PerC activates the transcription of ler, a gene present in LEE1 that encodes the main transcriptional activator of LEE (Mellies et al., 1999). In addition, expression of bfp and A/E activity is dependent on several environmental factors. Maximum transcription levels of bfp and of A/E activity occur at 37 °C and during the exponential growth phase (Puente et al., 1996; Rosenshine et al., 1996). Expression of bfp is also dependent on the presence of calcium ions and is inhibited by ammonium (Puente et al., 1996).

The PHO regulon of E. coli constitutes a group of more than 30 genes and operons whose expression is induced by phosphate limitation and that are related to the transport and assimilation of phosphorylated compounds. Some of the most well-studied PHO genes are phoA, which encodes alkaline phosphatase, the pst operon that encodes the high-affinity inorganice phosphate (P_i)-transport system Pst, and the operon phoB-phoR, which codes for a two component system that controls the expression of the PHO regulon. When P_i concentration in the medium falls below 4 µM, the histidine kinase PhoR, located in the inner membrane, auto-phosphorylates and transfers the phosphate group to the regulatory protein PhoB. Once phosphorylated, PhoB-P binds to consensus regions, known as PHO boxes, present in the promoters of all PHO genes thereby activating their transcription (Makino et al., 1988). Under phosphate-excess conditions, PhoR, which also displays phosphatase activity, dephosphorylates PhoB-P and halts the transcription of the PHO regulon genes (Makino et al., 1989).

The pst operon is induced under P_i shortage conditions and is composed of five genes, pstS, pstC, pstA, pstB and phoU, which are transcribed counter-clockwise in that order. Four of these genes encode the Pst system formed by the proteins PstS, PstC, PstA and PstB. PstS is a periplasmic protein that binds P_i with high affinity and carries the P_i molecule to the channel formed by the transmembrane proteins PstC and PstA (Webb et al., 1992). PstB is an ATP-binding protein that energizes Pst transport (Chan & Torriani, 1996). The function of PhoU is still unclear, but it does not participate in the transport of P_i (Steed & Wanner, 1993).

The Pst system also functions as a negative regulator of the PHO regulon, as most mutations in any gene of the pst operon lead to the constitutive expression of the PHO regulon (Wanner, 1996). The nature of this regulatory function is still unknown, and some point mutations in pstC and pstA abolish P_i transport through Pst, without disrupting the Pst regulatory function (Cox et al., 1988, 1989). Mutations in phoR that affect the phosphatase activity of the protein also cause the constitutive expression of the PHO regulon (Carmany et al., 2003; Kreuzer et al., 1975). At the protein level, the Pst system of EPEC (strain E2348/69) is 99.2 % identical to that of E. coli K-12. At the DNA level, the pst ORFs are 97.5 % identical on average. The only significant difference is a 92 bp deletion in EPEC pst in the intergenic region between pstA and pstB.

Several studies have described a positive correlation between some PHO genes and the virulence of E. coli and of other bacterial species as well. Mutations in pstC, phoU and partial deletions of the pst operon reduced the virulence of extra-intestinal E. coli strains (Buckles et al., 2006; Daigle et al., 1995; Lamarche et al., 2005). Likewise, insertions in the pstS gene of a porcine EPEC strain also negatively interfered with bacterial virulence by reducing attachment to the host cells (Batisson et al., 2003) and in Vibrio cholerae, a mutation in phoB partially impaired the colonization of rabbit intestines (von Krüger et al., 1999).

In the present study, we analysed the influence of the regulatory genes of the PHO regulon, namely the pst operon, phoB and phoR, on the adherence of EPEC to epithelial cells in vitro. We show that pst is needed for the full expression of EPEC adhesins and that the lack of pst is detrimental to bacterial adherence.

Bacterial strains, plasmids and oligonucleotides.
These are listed in Table 1.

Table 1. Strains, plasmids and oligonucleotides used in this study

Media and growth conditions.
Medium A is a semi-rich low-phosphate medium (Levinthal et al., 1962) which was called medium A+P_i when supplemented with 1 mM KH₂PO₄ and medium A–P_i when not supplemented. LB medium was used as described by Miller (1992). Dulbecco's Modified Eagle's Medium (DMEM) for epithelial cells (Cultilab-Brazil) was supplemented with 40 µg proline ml^–1 when used for the Δpst phoB mutant.

Transduction of PHO mutations.
The mutations Δpst : : Km^R and phoB519 : : Tn5 were transferred to LRT9 by transduction with phage P1 and selection with kanamycin. phoB23 proC : : Tn10 was transduced from strain BS1 into GMF195 (LRT9 Δpst : : Km^R) and selected for tetracycline resistance. The phoR69 mutation was transduced into LRT9 by a two-step procedure. First, proC : : Tn10 from strain BS8 was transduced into LRT9 and selected for tetracycline resistance and proline auxotrophy. Next, a P1 lysate of strain C3T (phoR69 proC⁺) was used to transduce LRT9 proC : : Tn10, and the transductants were selected for growth on minimal medium. The Δpst and phoR mutants were tested for alkaline phosphatase (AP) constitutivity and the phoB mutants were tested for the lack of AP induction under P_i limitation. All transductions were performed as described by Miller (1992).

Plasmid construction.
Plasmid pGM2 was constructed by ligating a PstI- and EcoRI-digested PCR product corresponding to the entire pst operon of E2348/69 to the low-copy vector pGB2 digested with the same restriction enzymes. The pst fragment of E2348/69 was amplified using the oligonucleotides pst1 and pst2, which adds PstI and EcoRI restriction sites to the ends of the DNA molecule.

To construct a transcriptional fusion between the perABC promoter and lacZ, a fragment of 645 bp containing the promoter of per was amplified by PCR using plasmid pEAF as a template and the oligonucleotides Per-P1 and Per-P2. The PCR product was first cloned in pGEM T-easy (Promega) and the recombinant plasmid was digested with EcoRI. The digested fragment containing the perABC promoter region was ligated to the EcoRI-digested vector pRKlacZ290 (Gober & Shapiro, 1992) that harbours a promoterless lacZ gene, resulting in plasmid pGM17. The orientation of the cloned per promoter with respect to lacZ was checked by amplifying the region encompassing the fusion per-lacZ by PCR with the oligonucleotides Per-P1 and lacZ-1391 and sequencing.

Strain GMF247 (LRT9 Δpst phoB) is resistant to tetracycline and could not be transformed with pGM17. A spectinomycin-resistance cassette was therefore inserted into pGM17, resulting in plasmid pGM29. Briefly, pGM17 was partially digested with 0.2 units of EcoRV (which digests inside the tetracycline-resistance gene and elsewhere) for 2 h. The spectinomycin-resistance cassette was obtained by digesting pJL74 (LeDeaux & Grossman, 1995) with PstI and BamHI, followed by treatment with Klenow to blunt the fragment ends. The fragments were then ligated and colonies were selected based on their resistance to spectinomycin and sensitivity to tetracycline.

Enzyme assays.
AP activity was assayed as described by Spira et al. (1995). Bacteria were grown in medium A±P_i. p-Nitrophenyl-phosphate (p-NPP) was used as a substrate and Na₂HPO₄ was used to terminate the reaction. AP specific activity was calculated according to the following formula: A⁴¹⁰/time (min)/cell density (OD₅₄₀).

β-Galactosidase was assayed as described by Miller (1992). Briefly, 800 µl buffer Z (16.1 g Na₂HPO₄ l^–1, 5.5 g NaH₂PO₄ l^–1,0.75 g KCl l^–1, 0.25 g MgSO₄.7H₂O l^–1 and 2.7 ml β-mercaptoethanol l^–1) was added to 200 µl of permeabilized cells and the reaction was started with the addition of 200 µl 4 mg ONPG ml^–1. Samples were incubated at 32 °C until a yellow colour developed and 500 µl 1 M Na₂CO₃ was added to terminate the reaction. The reaction product was read at 420 nm and Miller units were calculated.

Chloramphenicol acetyltransferase (CAT) activity of the pstS-cat fusion was measured using a modification of the biochemical assay devised by Shaw (1975). Bacteria were grown overnight in LB and resuspended in LB, medium A±P_i, DMEM supplemented with 0 %, 2 % or 10 % fetal calf serum (FCS), DMEM supplemented with 2 % FCS and 1 mM KH₂PO₄ and DMEM medium without FCS. Cells were grown to the mid-exponential phase (OD₅₄₀ 0.4–0.5), washed with 0.1 M Tris/HCl pH 8 and disrupted by incubation in lysis buffer (100 mM potassium phosphate buffer pH 8.0, 2.0 mM EDTA, 1 % Triton X-100, 5.0 mg BSA ml^–1, 1.0 mM DTT, 5.0 mg lysozyme ml^–1) for 15 min at room temperature. The lysate was then centrifuged to remove cell debris and added to a cuvette containing 0.4 mg dithio-bis(2-nitrobenzoic acid) (DTNB) ml^–1 and 0.1 mM acetyl-CoA, both diluted in 0.1 M Tris/HCl pH 7.8. The reaction was started by the addition of 0.1 mM chloramphenicol. The enzyme specific activity was calculated using the following formula: (ΔA⁴¹²xdilution factor)/(13.6xOD₅₄₀xcell volume).

Qualitative adherence assay.
The adherence of EPEC to Hep-2 cells (larynx carcinoma established culture) was essentially performed as described by Cravioto et al. (1979). A suspension containing 10⁵ Hep-2 cells in 1 ml DMEM supplemented with 10 % FCS was added to each well of a 24-well tissue plate and grown for 48 h at 37 °C with 5 % CO₂. The medium was then removed from the cell monolayer and replaced with 1 ml fresh DMEM supplemented with 2 % FCS and 1 % mannose. At this point, 5x10⁷ bacteria previously grown for 18 h in LB medium at 37 °C were added to each well. After 3 h of incubation, the cell monolayer was washed six times with PBS to remove the non-adherent bacteria. Cells were fixed with methanol and stained with Giemsa and May–Grunwald stains. Adherent bacteria were observed under an optical microscope and photographed.

Quantitative adherence assay.
To quantify the level of bacterial adherence, a modification of the method of Minami et al. (1987) was employed. This method makes use of the endogenous β-galactosidase produced by the bacteria as a reporter. Hep-2 and bacterial cells were grown as described for the qualitative adherence assay, except that 1 mM IPTG was added to the media to induce expression of the lacZ operon. After washing with PBS, the monolayer containing the adhered bacteria was treated with 200 µl lysis buffer (0.1 M Tris/HCl pH 8, 2 mM EDTA, 1 % Triton X-100, 1 mM DTT and 5 mg lysozyme ml^–1) for 20 min at room temperature; β-galactosidase activity was assayed as described above.

RNA extraction and Northern-blot analysis.
Bacteria were grown in DMEM supplemented with 2 % FCS without agitation at 37 °C and harvested at an OD₅₄₀ of 0.4–0.6. They were then homogenized with 4 ml acid phenol/guanidine thiocyanate and incubated at 60 °C for 5 min. Following the addition of 1.6 ml chloroform, samples were centrifuged and 1 vol. 2-propanol was added to the supernatant. The precipitated RNA was centrifuged, resuspended in formamide and quantified by spectrophotometry. Twenty micrograms total RNA were electrophoresed in a 1 % agarose gel containing 7 % formaldehyde and transferred to a nylon membrane by capillarity. Radioactively labelled DNA probes for bfpA, eae and rpoD were synthesized by random primer labelling using [³²P]-dCTP. The DNA templates were amplified by PCR using the oligonucleotide pairs bfp-A/bfp-B, eae-A/eae-B and rpoD+/rpoD–. Membranes were hybridized with the labelled probes at 42 °C in hybridization solution (MRC-HS114F) for at least 16 h, washed and exposed to X-ray films. To hybridize the RNA with rpoD, membranes were stripped of the labelled bfpA or eae probes and rehybridized with a radioactively labelled rpoD DNA probe.

Western-blot analysis.
Cells grown overnight in LB were diluted in 10 ml DMEM containing 2 % FCS to a final OD₅₄₀ of 0.025 and incubated at 37 °C without agitation. After 4 h, the cultures were centrifuged and resuspended in 1 ml PBS . OD₅₄₀ was measured, and 2x10⁹ bacteria from each culture were resuspended in 100 µl loading buffer (0.1 M Tris/HCl pH 6.8, 2 % SDS, 5 % β-mercaptoethanol, 10 % glycerol, 0.002 % bromophenol) and boiled for 3 min. Ten microlitres of each sample was resolved by SDS-PAGE and transferred to a nitrocellulose membrane (Hybond-ECL; Amershan Biosciences) by capillarity. The membrane was blocked for 1 h in PBS supplemented with 0.05 % Tween-20 (PBS-T) and 5 % skim milk, and incubated with the monoclonal antibody anti-intimin (diluted 1 : 500) and the polyclonal antibody anti-BFP (diluted 1 : 2000) for 1 h. Peroxidase conjugates of anti-mouse immunoglobulin G and anti-rabbit immunoglobulin G (Promega) were used as secondary antibodies. Detection was performed with the SuperSignal West Pico Chemiluminescent system (Pierce), as recommended by the manufacturer.

To test whether the PHO regulon affects EPEC adherence to epithelial cells, mutations in PHO genes that have regulatory functions were transferred to the wild-type EPEC strain LRT9 by P1 transduction. Unlike many natural E. coli isolates, LRT9 is relatively sensitive to P1, which facilitates the transfer of gene markers or mutations from E. coli K-12 to EPEC strains, although with a considerably lower efficiency. Mutations in phoB, phoR and a deletion of the entire pst operon were transduced to LRT9. To check the phenotype of the transductants, bacteria were grown overnight in medium A under P_i excess (+P_i) or P_i limitation (–P_i) and tested for AP activity (Table 2). As expected, mutations in pst and phoR resulted in the constitutive expression of AP. The level of AP under P_i-sufficiency was higher in LRT9 Δpst than in LRT9 phoR, confirming that mutations in phoR result in less AP production than mutations in pst (Kreuzer et al., 1975). Plasmids carrying the respective wild-type genes complemented the constitutive phenotype. Two different null mutations in phoB (alleles phoB23 and phoB519) abolished the expression of AP even under P_i-limiting conditions and normal levels of AP expression were restored by complementing in trans with a wild-type phoB gene. In conclusion, mutations introduced into the PHO regulatory genes of LRT9 caused the expected phenotypes.

Table 2. AP level of overnight cultures Values represent the mean±SE of three independent experiments. +Pi, medium A supplemented with 1 mM KH2PO4; –Pi, non-supplemented medium A. See Methods for units.

Effect of PHO mutations on in vitro EPEC adherence
To test the effect of the PHO mutations on EPEC adherence, qualitative and quantitative in vitro assays with an established epithelial culture (Hep-2 cells) were performed. As happens with typical wild-type EPEC strains, LRT9 adhered in a localized fashion and formed microcolonies on Hep-2 cell surfaces after 3 h of incubation (Fig. 1a). Deletion of the pst operon negatively affected the adherence of LRT9, as the microcolonies formed by strain GMF195 (LRT9 Δpst) were more scattered and with fewer bacteria adhered (Fig. 1b). For a quantitative assessment of bacterial adherence, the β-galactosidase activity of the adhered bacteria was used as a reporter (Minami et al., 1987). The adherence level of strain GMF195 was more than twofold lower than that of the wild-type strain (Fig. 2, compare bars B and C). C.f.u. ml^–1 counting of adhered LRT9 and of its Δpst mutant confirmed the deleterious effect of Δpst on adherence (not shown). Complementation in trans of the Δpst mutation with the low-copy plasmid pGM2 (ppst⁺) restored the normal adherence phenotype (Figs 1d and 2, D). pGB2 was used as a negative control and, as expected, did not improve the adherence level of GMF195 (Fig. 2, E). A phoB null mutation did not affect the adherence pattern of LRT9 (Fig. 1c), and a 20 % increase in the adherence level of the phoB mutant was observed, but it was not statistically significant (Fig. 2, F), indicating that the presence of PhoB is not necessary to obtain normal levels of adherence.

(154K): Fig. 1. Qualitative in vitro adherence assay of PHO mutants. Bacteria (5x107 c.f.u. ml–1) were added to a monolayer of Hep-2 cells and incubated for 3 h. Following the incubation period, the cell monolayer was washed and stained. (a) EPEC wild-type strain LRT9; (b) GMF195 (LRT9 Δpst); (c) JC1 (LRT9 phoB519); (d) GMF195 transformed with pGM2 (ppst+); (e) GMF247 (Δpst phoB23); (f) GMF257 (LRT9 phoR69).

(12K): Fig. 2. Quantitative adherence assay of PHO mutants. Bacteria (5x107 c.f.u. ml–1) grown in the presence of 1 mM IPTG were added to a Hep-2 cell monolayer. After 3 h, the cells were washed to remove non-adherent bacteria, and the activity of β-galactosidase was measured. A value of 100 % was assigned to the adherence level of the wild-type LRT9 strain. Bars represent the means±SE of at least three independent experiments. **, Values significantly different (P<0.05) from the parent strain LRT9 by Student's t-test. Bars: A, K-12 strain MG1655; B, LRT9; C, GMF195 (LRT9 Δpst); D, GMF195 transformed with pGM2 (ppst+); E, GMF195 transformed with pGB2 (cloning vector); F, JC1 (LRT9 phoB519); G, GMF247 (Δpst phoB23); H, GMF257 (LRT9 phoR69).

Since mutations in the pst operon lead to the constitutive transcription of all PHO genes, we next investigated whether the effect of Pst on EPEC adherence was related to the constitutive expression of the PHO genes. To test this, mutant strains LRT9 phoR and LRT9 Δpst phoB were obtained. We reasoned that if PHO constitutivity elicited by pst deletion is responsible for the inhibition of adherence, inactivation of PhoB would alleviate the deleterious effect of PHO constitutivity on bacterial adherence. Likewise, we reasoned that a phoR mutation, which also causes the constitutive expression of the PHO genes, would also inhibit EPEC adherence. The double mutant Δpst phoB23 displayed an adherence level similar to that shown by the Δpst mutant (Figs 1e and 2, G), indicating that although this strain ceased to express PHO constitutively, it still adhered less than the wild-type or the phoB mutant, suggesting that pst was epistatic to phoB in this case. Interestingly, the opposite occurs in the regulation of PHO, i.e. phoB is epistatic to pst. The phoR mutation did not significantly affect the adherence of LRT9 (Figs 1f and 2, H). Taken together, these results suggest that the deleterious effect of the Δpst mutation on EPEC adherence is not due to the constitutive expression of PHO.

It might be argued that the negative effect of Δpst on EPEC adherence is caused by an artefact due to a decrease in bacterial growth rate or by an unspecific inhibition of endogenous β-galactosidase expression. To address these concerns, the kinetics of β-galactosidase expression and the growth patterns of strains LRT9, GMF195 (LRT9 Δpst) and GMF195 transformed with pGM2 (ppst⁺) were determined (Fig. 3). Bacteria were suspended in DMEM supplemented with 2 % FCS and 1 mM IPTG at an initial concentration of 5x10⁷ cells ml^–1 and grown without agitation for 9 h at 37 °C. Samples were harvested every hour and assayed for β-galactosidase activity and growth. All three strains grew exponentially with a very similar growth rate for the first 5–6 h (up to OD₅₄₀ ∼0.7), when they reached the stationary phase (Fig. 3b). It could also be argued that the presence of pst would be advantageous under these conditions because a pst⁺ strain would take up P_i more efficiently. However, DMEM is rich in P_i (0.9 mM), which enables maximal growth rate and yield even in the absence of the high-affinity P_i transport system Pst. The entry of the cells into the stationary phase after 5 h of growth (OD₅₄₀ ∼0.8) was probably caused by O₂ depletion and by the presence of 1 mM IPTG, which limited bacterial growth.

(14K): Fig. 3. Endogenous β-galactosidase activity and growth curve of LRT9 and its pst mutant. Bacteria were grown in DMEM supplemented with 2 % FCS and 1 mM IPTG for 9 h. Samples were harvested every hour and assayed for (a) β-galactosidase activity and (b) growth. , LRT9; , GMF195 (LRT9 Δpst); , GMF201 (pGM2 in GMF195). Experiments were repeated twice, and representative results are presented.

The β-galactosidase activity of all three strains reached a plateau at 3 h (Fig. 3a). At this point the strain bearing pGM2 showed a slightly higher activity that slowly decreased from that point on. Overall, the growth pattern and β-galactosidase activity of all three strains was very similar and it can thus be concluded that the negative effect of Δpst on adherence is not due to a reduction in growth rate or in β-galactosidase expression.

The pst operon positively regulates the expression of bfp, eae and per
EPEC adherence depends on the expression of at least two different adhesins, BFP and intimin. Deletion of pst might affect EPEC adherence by inhibiting the expression of one of these adhesins. To test this hypothesis, Northern- and Western-blot analyses were conducted. Cells were grown in DMEM and harvested at the mid-exponential phase. Total RNA was extracted and hybridized with DNA probes corresponding to the first gene of the bfp operon, bfpA. The bfpA probe hybridized with a 0.6 kb band corresponding to the transcript of bfpA alone (Fig. 4a). Interestingly, although the bfp operon is composed of 14 genes, bands corresponding to transcripts containing the distal genes of this operon could not be observed under normal conditions. Overexposed autoradiograms, however, revealed the presence of large bands (data not shown). The presence of a stem–loop structure [ΔG –16.9 kcal mol^–1 (–70.7 kJ mol^–1)] in the intergenic region between bfpA and bfpG, might be acting as an mRNA stabilizer of the bfpA transcript or as a transcription terminator. This would explain why the mRNA of bfpA appears to be in excess in relation to the other cognate operon gene transcripts. The blot was rehybridized with a DNA probe corresponding to the housekeeping gene rpoD to normalize the RNA present in each sample. The densitometric analysis of the band intensities showed there was, on average, a twofold reduction in bfpA mRNA levels in the Δpst mutant when compared to the wild-type strain, and that this reduced level of bfpA transcript was also observed in the Δpst phoB double mutant (Fig. 4a). When the mutant was transformed with the low-copy plasmid pGM2 (ppst⁺), the level of bfpA mRNA was restored and even slightly increased, probably due to the presence of several copies of pst⁺. Similar results were observed when the production of the BFP protein was assessed by Western blotting (Fig. 4b). There was a 1.7-fold decrease in both Δpst and Δpst phoB relative to the wild-type strain, which was restored by complementation in trans with pGM2. The fact that the Δpst phoB double mutant was not able to restore expression of BFP at both transcript and protein level suggests again that the constituvity of PHO caused by the pst deletion is not responsible for the inhibition of bfp expression.

(24K): Fig. 4. Effect of pst on the expression of bfpA and BFP. (a) Northern-blot analysis of bfpA. RNA was extracted from mid-exponential-phase cultures of strains LRT9, GMF195 (LRT9 Δpst), GMF201 (pGM2 in GMF195) and GMF247 (Δpst phoB23) grown in DMEM and hybridized with a labelled bfpA DNA probe. Densitometric analyses of the bfpA band intensities normalized against the mRNA of rpoD are shown in the graphs. Each bar corresponds to the mean±SE of five independent experiments, except for GMF247, which corresponds to the mean±SE of three independent experiments. (b) Western-blot of BFP. Cells were grown as above and total protein extracts were assayed for BFP using a polyclonal anti-BFP serum. The bars represent the mean±SE of densitometric analysis of four independent experiments. The pictures of the blots are of typical experiments. **, Significantly different from the parent strain LRT9 by Student's t-test (P<0.05).

Fig. 5(a) shows that a DNA probe corresponding to eae, which encodes intimin, hybridized to a 3 kb band, as described by Gómez-Duarte & Kaper (1995), although minor bands of 2.8 and 5.2 kb, corresponding to cesT-eae and to the entire tir-cesT-eae operon were also observed. The blots were rehybridized with rpoD as above and the densitometric analysis of the band intensities showed there was a 2.5-fold reduction in the level of Δpst eae mRNA. pGM2 (ppst⁺) restored the level of eae mRNA. The Western blot analysis of intimin revealed that the pst deletion caused a 1.5-fold decrease in intimin level (Fig. 5b). These results indicate that pst is required for the full expression of bfpA and eae, and consequently for the maximization of synthesis of BFP and intimin.

(19K): Fig. 5. Effect of pst on the expression of eae and intimin. (a) Northern-blot analysis of eae. RNA was extracted from mid-exponential cultures of strains LRT9, GMF195 (LRT9 Δpst) and GMF201 (pGM2 in GMF195) grown in DMEM and hybridized with a labelled eae DNA probe. Densitometric analysis of the eae band intensities normalized against the mRNA of rpoD are shown in the graphs. Each bar corresponds to the mean±SE of five independent experiments (b) Western blot of intimin. Cells were grown as above and total protein extracts were assayed for intimin using a monoclonal anti-intimin serum. The bars represent the mean±SE of densitometric analysis of four independent experiments. The pictures of the blots are of typical experiments. **, Significantly different from the parent strain LRT9 by Student's t-test (P<0.05).

PerA and PerC are positive regulators of the bfp and LEE1 operons, respectively (Mellies et al., 1999; Tobe et al., 1996). LEE1 contains ler, which activates eae. perA–perB–perC form an operon on the pEAF plasmid. Several attempts to measure the effect of Δpst on per expression through Northern-blot analysis failed, because no signal could be observed (data not shown). This suggests that perA is transcribed at a very low level, and its mRNA cannot be detected with a moderately sensitive technique, such as Northern hybridization.

To measure the effect of Δpst on per expression, a transcriptional fusion between the promoter of perA and lacZ was constructed (pGM17) and transformed into strains LRT9, GMF195 (LRT9 Δpst), GMF201 [GMF195 bearing pGM2 (ppst⁺)] and JPN15 (EPEC cured of pEAF) as a negative control. Because GMF247 (LRT9 Δpst phoB) is resistant to tetracycline, a spectinomycin-resistance cassette was introduced into pGM17, resulting in plasmid pGM29. This plasmid, which behaves very similarly to pGM17 (not shown), was transformed into GMF247. All transformants were grown overnight and resuspended in DMEM supplemented with 2 % FCS and 0.4 % glucose (to repress endogenous β-galactosidase expression) to an OD₅₄₀ of 0.025, and grown for 9 h without agitation. Growth of all strains was similar; only strain JPN15 (EPEC cured of pEAF and negative control) showed a slightly higher growth rate (Fig. 6b). Samples were taken every hour and assayed for β-galactosidase activity (Fig. 6a). With the exception of JPN15, all strains displayed a gradual increase in β-galactosidase, which reached its peak at the height of the exponential phase (around 6 h), followed by a decrease until a new plateau was reached. At its highest point, the β-galactosidase activity of LRT9 was almost twice as high as that of the Δpst mutant and 2.5-fold above the β-galactosidase level of Δpst phoB. This demonstrates that the deleterious effect of Δpst on per expression cannot be alleviated by abolishing PHO constitutivity. Introduction of pGM2 (ppst⁺) into LRT9 Δpst led to a 70 % increase in the level of β-galactosidase compared to the wild-type strain, suggesting that pst has a positive role in the induction of the per operon. Since PerA is a positive regulator of bfp, pst is likely to affect bfp expression via perA. Analogously, an augmentation in PerC would positively affect the expression of ler, which in turn increases the transcription of eae. Therefore, the positive effect of pst on eae expression is likely to be through an increase in perC transcription.

(20K): Fig. 6. Effect of Δpst on per expression. Plasmid pGM17 (perA-lacZ fusion) was transformed into LRT9 (), GMF195 (LRT9 Δpst) (), GMF201 (pGM2 in LRT9 Δpst) (), and JPN15 (pEAF-cured EPEC) (). GMF247 (Δpst phoB23) was transformed with plasmid pGM29 (). Cells were grown in DMEM for 9 h. Samples were taken every hour and assayed for (a) β-galactosidase and (b) growth. Experiments were repeated three times, and typical results are presented.

The pst operon is partially induced in DMEM
Genes of the PHO regulon are activated by P_i-shortage. Under P_i-excess conditions, expression of Pst and other PHO genes is only basal (Wanner, 1996). We asked whether a low (basal) expression of Pst would be sufficient to affect the expression of the adhesins, or if under the growth conditions used in the adherence assay (DMEM) the actual level of Pst is increased. To answer this question, the activity of the pst promoter was measured under different growth conditions. Plasmid pBS11 (pstS-cat transcriptional fusion) was transformed into LRT9 and assayed for CAT and AP activity in bacteria grown exponentially in LB, medium A±P_i and DMEM alone or supplemented with FCS and 1 mM P_i. Table 3 shows that cells grown under P_i-limiting conditions (medium A–P_i) presented an 11-fold induction in the level of pst expression when compared to cells grown in medium A+P_i or in LB. In contrast, the CAT activity in DMEM+2 % FCS (the medium used in the adherence assays) was only fivefold lower than under P_i-starvation conditions (medium A–P_i) and twofold higher than the activity of cells grown in other high-P_i media (medium A+P_i or LB). Addition of 1 mM P_i to DMEM did not significantly alter the level of pst expression, suggesting that the increased pst expression in DMEM is not due to a shortage of P_i. Likewise, addition of 10 % FCS had no effect on the ability of DMEM to increase pst transcription. These results suggest that one or more components of DMEM positively affect pst expression. This effect is specific to pst because the basal level of another PHO protein, AP, was not affected by DMEM. In conclusion, expression of pst under the growth conditions used in the adherence assay was significantly higher for P_i-replete media. The relatively high level of Pst is likely to contribute to its positive effect on the expression of EPEC adhesins.

Table 3. CAT and AP activities of strains carrying the pst-cat fusion in various growth media Values represent the mean±SE of at least three independent experiments. See Methods for units.

EPEC is characterized by its pattern of localized (primary) and intimate adherence, which are its main virulence factors, and by the lack of Shiga-like toxin. EPEC adheres to epithelial cells in a two-step fashion: the first step is dependent on BFP; followed by the intimate adherence which is dependent on intimin. Environmental factors contribute to the regulation of virulence-related genes in EPEC. For instance, a temperature of 37 °C, the presence of calcium ions and low ammonium concentrations are paramount to the development of the LA phenotype and to bfp transcription (Puente et al., 1996). Also, bfp expression is growth phase dependent, being transcribed predominantly during the exponential phase. Similarly, the A/E phenotype is also dependent on temperature and growth phase (Rosenshine et al., 1996).

Some PHO regulon genes, particularly the pst operon and phoB, have been shown to be involved in different aspects of the virulence of E. coli and other bacteria. Among pathogenic E. coli, point mutations in phoU, pstC and a pst deletion reduced the virulence of different extra-intestinal strains (Buckles et al., 2006; Daigle et al., 1995; Lamarche et al., 2005). Insertions in the pstS gene of a porcine EPEC strain resulted in reduced attachment of the bacteria to piglet ileal explants (Batisson et al., 2003). In Shigella flexneri, a pst mutant displayed a significant reduction in plaque formation, which was suppressed by a mutation in phoB and also by the introduction of a non-constitutive pstA mutation (Runyen-Janecky et al., 2005). Restoration of normal levels of plaque formation by suppression of the PHO-constitutive phenotype implied that at least in this case, the constitutive expression of the PHO genes was the cause of the inhibition of S. flexneri virulence. On the other hand, it was suggested that the lack of a functional Pst in the avian pathogenic E. coli O78 reduced its virulence due to changes in cell surface composition (Lamarche et al., 2005). However, in most cases, the exact mechanism through which pst enhances pathogenicity remains unclear.

In the present study, we showed that a complete deletion of pst caused a reduction in the ability of EPEC to adhere to epithelial cells in vitro and to display the LA phenotype. Moreover, the Pst system was required for the full expression of the adhesins BFP and intimin, and of the regulators encoded by the per operon. It was also shown that the pst operon itself is slightly induced under the conditions of the adherence assay. Unlike in S. flexneri (Runyen-Janecky et al., 2005), the constitutivity of the PHO genes caused by Δpst was not responsible for its deleterious effect on EPEC virulence, because introduction of a phoB mutation into LRT9 Δpst eliminated PHO constitutivity, but did not restore bacterial adherence to its original levels. Furthermore, a phoR mutant, which also constitutively expresses the PHO genes, did not inhibit bacterial adherence. Also, a pstS polar mutant of LRT9 inhibited bfp expression to the same extent as Δpst, but a non-polar mutation in pstS (which is also PHO-constitutive) had no effect on bfp (data not shown), confirming the lack of effect of PHO constitutivity on cell adherence.

LRT9 and its Δpst mutant displayed a similar growth rate in DMEM, indicating that the reduction in adherence was not due to a growth disadvantage of the Δpst mutant. Measurement of the intrinsic β-galactosidase activities of the wild-type and Δpst mutant showed that they were very similar. This experiment was important to show that the quantitative method employed to determine bacterial adherence (i.e. endogenous β-galactosidase activity) was not influenced by some unrelated variation in the β-galactosidase expression pattern caused by the pst mutation. Although less common than determining c.f.u. ml^–1, the use of β-galactosidase as a reporter of EPEC adherence has already been shown to be reliable and reproducible (Minami et al., 1987). Here, we confirmed the reliability of this methodology using a different EPEC lineage (LRT9) and also by comparing the adherence levels obtained with the β-galactosidase assays with c.f.u. ml^–1 calculations of the wild-type and the Δpst mutant, and by microscopic examination of the adhered bacteria.

Northern-blot analysis of bfp and eae, Western blots of BFP and intimin, and the enzymic assays of the perA-lacZ fusion showed that pst plays a positive role in the expression of the bfp, tir–cesT–eae and per operons. Since PerA and PerC positively regulate bfp and eae, respectively (Tobe et al., 1996; Mellies et al., 1999), it seems plausible to suggest that the mechanism by which pst enhances transcription of these genes is via induction of per, whose products in turn upregulate bfp and eae.

A possible source of concern was the fact that the adherence assay was performed under conditions of P_i abundance and that the level of pst expression under these conditions would be too low to exert any significant effect. However, measurement of pst transcription level in DMEM showed that it was twofold higher than that observed for a high-P_i medium such as LB or medium A+P_i. This implies that in DMEM, expression of pst is higher than normal and more likely to have a broader effect, for instance as an adherence enhancer, as observed here. The possibility that some component of FCS, which supplements DMEM, is responsible for the increased expression of pst was ruled out, because the level of pst was even higher in non-supplemented DMEM. The nature of the pst-inducing factor in DMEM remains to be investigated.

It has recently been shown that a short RNA sequence in the intergenic region between pstA and pstB has a regulatory role in σ^S translation (Schurdell et al., 2007). It is not known whether rpoS has any effect on EPEC adherence, but in enterohaemorrhagic E. coli (EHEC), rpoS activates the expression of LEE via DsrA (Laaberki et al., 2006). It is possible that the regulatory region between pstA and pstB plays a role in per transcription via rpoS or through another mechanism. The pstA–pstB intergenic region of EPEC E2348/69 carries a deletion of 92 bp (from position 3247 to 3338 on the ECOPHOS sequence K01992), but the rpoS regulatory sequence (which comprises the 3352–3383 region) is intact in this strain.

At this stage, it is still unclear if the entire pst operon or just one of its genes is involved in the regulation of EPEC adherence. A polar pstS mutant presented a reduction in the adherence level similar to the one observed for Δpst, while a non-polar mutation in this gene did not affect the level of adherence (not shown). To test the involvement of individual pst genes in adherence, single pst mutations (polar and non-polar) should be isolated and tested.

In conclusion, this study adds to a number of reports showing that the pst operon participates in the mechanism of E. coli virulence. This function could not be previously anticipated judging from the main roles of Pst as a P_i transport system and as a negative regulator of PHO. However, it has been recently reported that Pst also plays a role in the regulation of RpoS (Schurdell et al., 2007) and in the phenomenon of bacterial persistence (Li & Zhang, 2007). It is becoming evident that Pst has a broader regulatory impact as a global (though not principal) regulator of pathogenicity in different bacterial species and even as a more general regulator beyond its role in P_i metabolism.

This work was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP). G. M. F. was supported by grant 02/08604-8 of FAPESP. The authors thank Roxane Maria Fontes Piazza (Instituto Butantan, São Paulo) for kindly supplying the antibodies against BFP and intimin, and Ezra Yagil for valuable comments.

Edited by: B. Kenny