INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Enterohemorrhagic Escherichia coli (EHEC) O157:H7 is the causative agent of several outbreaks of bloody diarrhea and hemolytic-uremic syndrome throughout the world. EHEC colonizes the large intestine, where it causes attaching and effacing (AE) lesions and produces the potent Shiga toxins (Stx), which are responsible for the major symptoms of hemorrhagic colitis and hemolytic-uremic syndrome (reviewed in references 23, 24, and 33). The AE lesion is characterized by effacement of the intestinal epithelial cell microvilli and the rearrangement of the cytoskeleton to form a pedestal-like structure that cups the bacteria individually (reviewed in references 23, 24, and 33).

The genes involved in the formation of the AE lesion are encoded within a chromosomal pathogenicity island named the locus of enterocyte effacement (LEE) (31). The LEE encodes a type III secretion system, effector proteins, and a bacterial adhesin (16). We have recently reported the regulation of LEE genes by quorum sensing (45). Quorum sensing is a mechanism of cell-to-cell signaling involving the production of hormonelike compounds called autoinducers. Through the accumulation of these autoinducers, the bacteria "sense" their own population as well as the population of other bacteria in a given environment. When these molecules reach a certain concentration threshold, they interact with bacterial regulatory proteins, thereby controlling gene expression. This phenomenon was first described in the regulation of bioluminescence in Vibrio fischeri (34) and since then has been shown to be a widespread gene regulation mechanism in both gram-negative and gram-positive bacteria.

Gram-negative bacteria usually produce acyl-homoserine lactones (AHLs) as autoinducers, while peptides are produced by gram-positive organisms (reviewed in reference 14). Vibrio harveyi is another marine, bioluminescent, gram-negative bacterium that regulates luminescence by quorum sensing. However, it produces two types of autoinducers; one is an AHL that is referred to as autoinducer-1 (AI-1) (8), and the other, whose biochemical nature has not yet been reported, is referred to as AI-2 (5, 6). AI-1 is primarily involved in intraspecies communication, while AI-2 is involved not only in intraspecies but also interspecies communication (46). Unlike AHLs or autoinducing peptides, AI-2 is found in both gram-positive and -negative bacteria, including E. coli, salmonellae, Vibrio cholerae, enterococci, Mycobacterium tuberculosis, and Helicobacter pylori, among others (46, 47). The gene encoding the AI-2 synthetase was cloned, sequenced, and named luxS by Surette et al. (47).

Besides production of light, quorum-sensing mechanisms have been demonstrated to regulate competence in Streptococcus (reviewed in references 11 and 18), production of hemolysins and other virulence genes in Staphylococcus (4), production of elastase and biofilm formation in Pseudomonas aeruginosa (9, 19, 36), iron acquisition in V. harveyi (28), and type III secretion in EHEC (45). Using gene array technology, we now demonstrate that quorum-sensing regulation in EHEC is far more pleiotropic and regulates a number of basic physiological functions, including cell division and motility.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains and plasmids. EHEC O157:H7 strain 86-24 was isolated from an outbreak of bloody diarrhea (21). Strain VS94 is a luxS isogenic mutant of strain 86-24. We initially constructed plasmid pVS68 by amplifying the luxS gene from EHEC O157:H7 strain 86-24 with Pwo polymerase (Boehringer Manheim) using primers K1663 (5'-GTCGACGCCGCTGATACCGAACCG-3') and K1664 (5'-GTCGACGCGGTGCGCACTAAGTACAA-3') and cloning into the EcoRV site of pBluescript KS II (Stratagene) (45). We cloned a tetracycline resistance cassette derived from pBR322 into an EcoRV site in the middle of luxS; this construct was then cloned into the suicide vector pCVD442, which contains an R6K origin of replication (13), generating plasmid pVS72. The EHEC luxS mutant strain, named VS94, was generated by allelic exchange of the luxS gene in pVS72 using Tc and sucrose selection as previously described (13, 45). VS95 is VS94 complemented with pVS84, which is wild-type luxS cloned into pACYC177 (Table 1).

V. harveyi luminescence assay. The presence of AI-2 in the preconditioned media was assayed using the V. harveyi BB170 (luxN::Tn5) reporter strain, which responds only to AI-2 (46). The luminescence assays were performed as described (46), and the assays were read in a Wallac 1420 multilabel counter.

Gene array. RNA was isolated from strains 86-24 and VS94 grown at 37°C in Dulbecco modified Eagle medium (DMEM) (catalog no. 11054-020; Gibco BRL) to an optical density at 600 nm (OD₆₀₀) of 1.0 according to standard procedures (42). Synthesis of cDNA was performed using avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim), and cDNA was labeled with [-³³P]CTP using DNA polymerase and E. coli random primers (Genosys). Unincorporated nucleotides were removed with a G-25 Sephadex column, and the efficiency of incorporation was measured by comparing the samples before and after the chromatography. The radionucleotide incorporation was 69% for 86-24 cDNA and 72% for VS94 cDNA. The cDNAs were then hybridized with the Panorama E. coli K-12 gene array (Genosys), according to the manufacturer's instructions. The array was scanned using a Storm phosphorimager with a pixel size of 100 µm (10,000 dots/cm²), and the resulting TIFF image was analyzed with the Quantarray software from GSI Lumonics to determine the differences in pixel intensity. The software marked each spot with an ellipse, generating a grid map locating every single spot, and the background was automatically subtracted by the program. Each gene was spotted in duplicate on each array, and the final ratios shown are the average of the two readings. The averaged spot intensity was expressed as a percentage of the total of intensities of all spots on the DNA array, allowing direct comparison of the two arrays by normalizing with regard to the specific activity of the probes used. The data were quantified using the quantification adaptive method (according to the Quantarray software operating manual), which takes into consideration any irregularity in the spots' shapes and calculates the standard deviations between the pixels read per spot. The ratio between each point of the array hybridized with the luxS mutant and the corresponding point in a duplicated set hybridized with the wild-type strain was calculated. The log of the absolute value of the expression ratio was positive for percent intensities that were higher than those for the array hybridized with the luxS mutant and was negative for percent intensities that were lower than those for the array hybridized with the luxS mutant.

Northern blots. Total RNA was isolated from strains 86-24 and VS94 grown in DMEM to an OD₆₀₀ of 1.0 at 37°C, and Northern blotting was performed following standard procedures (42). The membranes were hybridized at high stringency (65°C) with probes for fliF, fliA, and rpoE labeled with [-³²P]dCTP by random priming (42). These probes were generated by amplifying DNA fragments internal to these genes in strain 86-24 using Taq DNA polymerase and primers K2267 and K2268 for fliF, K2263 and K2264 for fliA, and K2256 and K2255 for rpoE (Table 2).

Construction of operon fusions with lacZ. Operon fusions with lacZ were constructed by amplifying the regulatory regions of the genes in strain 86-24 by PCR using Pwo polymerase and by cloning into the EcoRI and BamHI sites of plasmid pRS551, which contains a promoterless lac operon (43). The following primer pairs were used to produce specific lacZ fusions: primers K2254 and K2253 for pyk::lacZ (pVS166); primers K2246 and K2245 for ptsN::lacZ (pVS167); primers K2302 and K2301 for flhD::lacZ (pVS182); primers K2304 and K2303 for motA::lacZ (pVS176); primers K2308 and K2307 for fliC::lacZ (pVS175); primers K2309 and K2310 for fliA::lacZ (pVS177); primers K2422 and K2423 for sulA::lacZ (pVS189); and primers K2419 and K1429 for stx::lacZ (pVS188) (Tables 1 and 2).

-Galactosidase assays. The bacteria containing the operon::lacZ fusions were grown with shaking at 250 rpm for 18 h at 37°C in DMEM or Luria broth (LB) as indicated, diluted 1:100 in fresh DMEM or LB, and grown at 37°C to an OD₆₀₀ of 1.0. These cultures were diluted 1:10 in Z buffer (Na₂HPO₄ [0.06 M], NaH₂PO₄ [0.04 M], KCl [0.01 M], MgSO₄ [0.001 M], and -mercaptoethanol [0.05 M]) and were assayed for -galactosidase activity using o-nitrophenyl--D-galactopyranoside (ONPG) as the substrate as previously described (32).

Preconditioned media. To prepare preconditioned media containing high levels of AI-2, an initial inoculum of strain 86-24 or VS94 was grown with shaking in DMEM for 18 h at 37°C. This culture was diluted 1:100 in fresh DMEM and was grown to an OD₆₀₀ of 0.2 when it was once again diluted 1:100 and grown to an OD₆₀₀ of ca. 1.0. Following centrifugation (12,000 × g, 4 min, 25°C) and filtration (0.2- µm-pore-size filter), the pH of the supernatant was adjusted to 7.0.

Growth curves. Strains 86-24, VS94, VS94(pACYC177), and VS95 were grown for 18 h in DMEM at 37°C, diluted 1:500 in fresh DMEM, and grown at 37°C with shaking at 250 rpm. OD₆₀₀ measurements were taken every hour, and 100 µl of the cultures was diluted and plated on LB plates to obtain CFU counts.

Western blotting. Total proteins were extracted from strains 86-24, VS94, and VS95 grown in DMEM to an OD₆₀₀ of 1.0. In brief, 1 ml of culture was pelleted (12,000 × g for 5 min at 4°C), resuspended in 400 µl of PBS and 100 µl of 5× sample buffer (20% sodium dodecyl sulfate, 20% glycerol, 200 mM Tris base, pH 6.8, and 0.001% bromophenol blue). The protein concentration was measured using the Lowry assay (30). Equal amounts of total proteins were electrophoresed in sodium dodecyl sulfate-12% polyacrylamide gel electrophoresis (26). Western blotting was performed as previously described (42), and the blots were probed with polyclonal antisera directed against the A subunit of Stx2, kindly provided by Alison O'Brien (Uniformed Services University of the Health Sciences), and with antisera directed against the H7 flagellin (J. A. Girón and J. B. Kaper, unpublished data).

Motility assays. Motility assays were performed at 37°C on 0.3% agar plates containing DMEM, DMEM plus 10% DMEM preconditioned by growth of EHEC 86-24 or VS94 (86-24 luxS mutant), or tryptone media (1% tryptone and 0.25% NaCl). The motility halo was measured at 16, 24, and 48 h.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of EHEC luxS mutant. We constructed an isogenic luxS mutation in EHEC strain 86-24 to generate strain VS94. The luxS mutation was confirmed by PCR and Southern blotting (data not shown), and the ability of VS94 to produce AI-2 was assayed using the V. harveyi AI-2 detection test described by Surette and Bassler (46). Wild-type EHEC strain 86-24 produces AI-2, and media preconditioned with this strain can induce a 60-fold increase in the production of light by V. harveyi strain BB170 after 3 h, compared to media alone (46). Preconditioned media prepared with strain VS94 did not induce production of light in V. harveyi strain BB170 (data not shown). The luxS mutation was complemented with the luxS gene cloned in vector pACYC177(pVS84), and the resulting strain was designated VS95 (Table 1). The complemented strain produces twice the amount of AI-2 produced by the wild-type strain due to the presence of luxS on a multicopy plasmid (Sperandio and Kaper, unpublished). Preconditioned media prepared with strain VS95 induce a ca. 130-fold induction in the production of light by V. harveyi strain BB170. As a positive control to the AI-2 detection test, we used V. harveyi strain BB152, which produces only AI-2 (Table 1).

Global gene regulation by quorum sensing. We have previously reported that genes expressed within the LEE pathogenicity island in EHEC that are involved in the formation of the AE lesions are regulated by quorum sensing through AI-2 (45). To determine the extent of regulation via AI-2-mediated quorum sensing in EHEC, we used a gene array approach. Although gene arrays for EHEC are not yet commercially available, we hybridized the K-12 gene array (Genosys-Sigma) with cDNA synthesized from RNA extracted from wild-type EHEC strain 86-24 and its luxS isogenic mutant VS94 to identify those genes shared by K-12 and EHEC that are regulated at least fivefold by quorum sensing via AI-2. Both strains were grown in DMEM at 37°C to an OD₆₀₀ of 1.0, conditions previously found to maximize production of AI-2 in this strain (45). We observed that 404 of the 4,290 genes on the array were regulated at least fivefold by quorum sensing, which comprises ca. 10% of the K-12 gene array, suggesting that quorum sensing is a global regulatory mechanism in E. coli. Of the 4,290 genes on the array, 144 were not expressed at detectable levels and 3,886 genes were not affected by the luxS mutation. Of the 404 genes regulated by luxS, 235 were up-regulated and 169 were down-regulated in the wild-type strain compared to in the luxS mutant. Functions have not been assigned to 138 (34%) of these 404 genes (data not shown). When a less stringent twofold cutoff was used, 736 genes (ca. 17% of the array) were regulated by luxS. We were able to confirm the regulation of genes such as ptsN (2.5-fold) and motAB (3.5-fold) using operon::lacZ fusions (see below), which we would otherwise have overlooked with a fivefold cutoff.

View larger version (40K): [in this window] [in a new window]   FIG. 1.   (A) Northern blots of total RNA from strains 86-24 and VS94 (luxS mutant) with probes for fliF, fliA, and rpoE. (B) -Galactosidase assays of plasmids pVS166 (pyk::lacZ) and pVS167 (ptsN::lacZ) in 86-24, VS94 (luxS), and VS95 (luxS +) strain backgrounds in DMEM to an OD600 of 1.0. (C) DNA array spots of flgI, fliH, fliF, ptsN, and rpoE. The numbers in parentheses represent fold inductions; +, the gene is up-regulated in the wild-type (WT) strain compared to in the luxS mutant; , the gene is down-regulated in the wild-type strain compared to in the luxS mutant.

Regulation of growth and cell division. Several genes involved in cell division, such as ftsQA (2.5-fold), ftsE (6-fold), and hlfB (4-fold), were down-regulated in the wild-type strain compared to in the luxS mutant. The minE gene, which is involved in cell division topology (10), was up-regulated 13-fold, and sulA, which encodes a cell division inhibitor that blocks formation of the FtsZ ring (7), was up-regulated 25-fold in the wild-type strain compared to in the mutant (Table 3). We generated an operon fusion of the sulA promoter with a reporter lacZ gene (plasmid pVS189) and observed a 2.5-fold decrease in the expression of -galactosidase in strain VS94 (luxS mutant) compared to in strains 86-24 and VS95 (complemented strain) (Fig. 2A). The 2.5-fold difference in the expression of sulA::lacZ seen with plasmid pVS189 compared to the 25-fold difference observed with the array may be due to a copy number effect of the cloned sulA gene on a multicopy plasmid. Another explanation could be due to the fact that the transcriptional level of sulA::lacZ in pVS189 in VS94 is already very high (15,000 Miller units) and that -galactosidase activities greater than 25,000 Miller units have a toxic effect on bacteria (2). For a multicopy sulA::lacZ fusion, therefore, we may not be able to detect fold differences much higher than 2.5-fold. Given these data, we examined the growth of the luxS mutant and observed that in DMEM, which produces optimal levels of AI-2 in this strain, VS94 has an average generation time of 31.6 min, which is considerably shorter than the 55-min average generation time seen with the wild-type strain (Fig. 3A). Complementation with the cloned wild-type luxS gene on a multicopy plasmid (strain VS95) resulted in an average generation time of 101 min (Fig. 3A). The luxS mutant containing the plasmid vector pACYC177 has the same generation time as the luxS mutant alone, indicating that the presence of this plasmid does not have any effect on growth. The slow growth rate of the complemented mutant compared to that of the wild type is presumably due to overproduction of AI-2 from the presence of the luxS gene on a multicopy plasmid. The growth of the luxS mutant (VS94) can also be slowed by the addition of media containing AI-2 (DMEM plus 10% preconditioned media prepared with 86-24) (Fig. 3B). As a control for nutrient availability, growth of VS94 (luxS mutant) in DMEM plus 10% preconditioned media prepared with VS94 was compared to growth of VS94 in fresh DMEM. Growth rates were equivalent (Fig. 3B).

View larger version (24K): [in this window] [in a new window]   FIG. 2.   Up-regulation of Stx production by quorum sensing. (A) -Galactosidase assays of a sulA::lacZ fusion in 86-24, VS94, and VS95 in DMEM at an OD600 of 1.0. (B) -Galactosidase assays of an stx::lacZ fusion in 86-24, VS94(pACYC177), and VS95 in DMEM at an OD600 of 1.0. (C) Western blotting of whole-cell lysates of strains 86-24 (WT), VS94(pACYC177) (luxS), and VS95 (luxS+) with anti-Stx2A subunit polyclonal antiserum.

View larger version (26K): [in this window] [in a new window]   FIG. 3.   (A) Growth curves of EHEC wild-type strain 86-24, VS94 (86-24 [luxS]), VS94(pACYC177), and VS95 (VS94 complemented with pVS84 [luxS +]) in DMEM at 37°C. (B) Growth curves at 37°C of VS94 (luxS) in fresh DMEM; DMEM plus 10% media preconditioned (PC) by growth of strain VS94 (luxS mutant); and DMEM plus 10% media preconditioned by growth of wild-type strain 86-24 (producing AI-2). Growth curves were performed in triplicate.

Quorum sensing controls Stx expression. Recently, Neely and Friedman (35) and Plunkett et al. (39) demonstrated that the genes encoding Stx1 and Stx2 are located within the late genes of a -like phage. The genes encoding Stx2 (which is the only Stx produced by strain 86-24) are transcribed when the phage enters the lytic cycle. Upon the induction of an SOS response, the cI repressor is cleaved, allowing transcription of the middle and late genes to proceed from the -like Q-dependent promoter, and together with late genes the stx2 is also transcribed (35). Using a stx::lacZ fusion (containing the Q-dependent promoter), we observed a threefold increase in the transcription of stx in the wild-type strain compared to in the luxS mutant, and this transcriptional increase could be complemented with the cloned luxS gene (VS95) (Fig. 2B). Western blots also demonstrated that the wild-type strain produces more Stx2 than does the luxS mutant (6.7-fold-higher pixel volume by densitometry) and that this phenotype could also be complemented with the cloned luxS gene (Fig. 2C). We previously reported that transcription of the stx genes was not controlled by quorum sensing (45). However, the fusion used on those experiments did not contain the -like Q-dependent promoter. Genes involved in an SOS response were up-regulated in the wild-type strain compared to in the luxS mutant in the array (recA, 20-fold; uvrA, 20-fold; and sulA, 25-fold). As noted above, decreased sulA expression in the luxS mutant compared to that in the wild-type strain was confirmed using a sulA::lacZ fusion (Fig. 2A), consistent with the suggestion of an SOS response induction by the array data. These results suggest that the induction of an SOS response by quorum sensing may have a role in the induction of Stx2 production.

Regulation of flagella, chemotaxis, and motility. The array data also revealed that expression of numerous genes involved in production and assembly of flagella, chemotaxis, and motility were up-regulated in the wild-type strain compared to in the luxS mutant (Fig. 1C and 4A; Table 4). Figure 4A presents a representation of the flagellar class I, II, and III genes and of the fold activation by quorum sensing observed for each gene or operon (in the case of operons, the average activation among the genes within that operon). Northern blots using RNA extracted from the wild-type strain and the luxS mutant confirmed that the genes fliA (which regulates expression of class III flagellar genes) and fliF (which encodes the basal-body flagellar protein) were up-regulated in the wild-type strain, compared to in the luxS mutant (Fig. 1A). We examined the luxS mutant for production of flagella and found that it produces considerably less flagellin than the wild-type strain, as shown by Western blotting (Fig. 4B), and produces fewer flagella, as shown by electron microscopy (data not shown). To further confirm these results, we generated operon::lacZ fusions with the promoter regions of flhD (class I), fliA (class II), fliC (class III), and motA (class III) and monitored their transcription in 86-24, VS94, and VS95 backgrounds in both LB and DMEM. The transcriptional levels of all of the flagellar genes studied were always higher in LB than in DMEM, with the exception of flhD (Fig. 4C and D). Transcription of flhD was up-regulated for 86-24 and VS95 twofold in LB and fourfold in DMEM, compared to that for VS94 (luxS mutant). Transcription of fliA for strain 86-24 was up-regulated threefold in LB and twofold in DMEM, compared to that for VS94. Transcription of fliA for VS95 was up-regulated eightfold in LB and twofold in DMEM, compared to that for VS94. Transcription of fliC for 86-24 was up-regulated eightfold in LB and twofold in DMEM; for VS95 it was up-regulated fivefold in LB and twofold in DMEM when compared to that for VS94. Finally, transcription of motA for 86-24 was up-regulated 16-fold in LB and 20-fold in DMEM and for VS95 was up-regulated 13-fold in LB and 2.5-fold in DMEM compared to that for VS94 (Fig. 4C and D).

View larger version (32K): [in this window] [in a new window]   FIG. 4.   Regulation of genes involved in expression and assembly of flagella and motility by quorum sensing. (A) Organization of flagellar genes and the classes of regulation; the fold increase of up-regulation in the wild-type strain compared to in the luxS mutant in the array experiment is indicated in parentheses (numbers given for polycistronic operons represent the average of the individual fold increase of each gene in an operon). (B) Western blot of whole-cell lysates of strains 86-24 (WT), VS94 (luxS), and VS95 (VS94 with pVS84) (luxS+) with antiflagellin polyclonal antiserum. (C) -Galactosidase assays of flhD::lacZ, fliA::lacZ, fliC::lacZ, and motA::lacZ fusions in 86-24, VS94(pACYC177), and VS95 backgrounds in LB at an OD600 of 1.0. (D) -Galactosidase assays of flhD::lacZ, fliA::lacZ, fliC::lacZ, and motA::lacZ fusions in 86-24, VS94(pACYC177), and VS95 backgrounds in DMEM at an OD600 of 1.0.

View larger version (69K): [in this window] [in a new window]   FIG. 5.   (A) Motility test of strains 86-24, VS94 (luxS mutant), and VS95 (VS94 complemented with pVS84) in DMEM plus 0.3% agar after 48 h of incubation at 37°C. (B) Motility test of strains 86-24, VS94 (luxS ), and VS95 (VS94 with pVS84) in tryptone medium plus 0.3% agar after 16 h of incubation at 37°C. (C) Motility tests of 86-24, VS94, and VS95 in DMEM plus 10% media preconditioned with wild-type strain 86-24 (containing AI-2) or DMEM plus 10% media preconditioned with the luxS mutant strainVS94 after 48 h of incubation at 37°C. (D) -Galactosidase assays of motA::lacZ in VS94 background in DMEM (no preconditioning [PC]); DMEM plus 10% media preconditioned with wild-type strain 86-24 containing AI-2 [VS94 (PC 86-24)]; DMEM plus 10% media preconditioned with the luxS mutant strainVS94 (PC VS94); and VS95 (VS94 complemented with pVS84) at an OD600 of 1.0.

Other genes regulated by quorum sensing. Among other genes shown to be regulated by quorum sensing in the initial array analysis were genes involved in energy metabolism, cell structure, transport, DNA replication, prophage integration, transposases, carbon compound catabolism, central intermediary metabolism, nucleotide biosynthesis and metabolism, amino acid biosynthesis and metabolism, transcription, RNA processing and degradation, fatty acid and phospholipid metabolism, biosynthesis of cofactors, prosthetic groups, and carriers (data not shown). Confirmatory genetic analysis or phenotypic correlations of the DNA array data for these genes have not yet been investigated.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Many bacterial species possess the luxS gene and produce AI-2, including salmonellae, H. pylori, streptococci, E. coli, and others (47). Given our recent findings that the genes encoding the type III secretion system in EHEC are under the control of quorum sensing through AI-2, we wanted to identify other EHEC genes that are regulated by this mechanism (45). In the absence of the complete EHEC genome sequence and of DNA arrays for this pathogen, we initiated our investigations by hybridizing an E. coli K-12 gene array with cDNA synthesized from a wild-type EHEC strain and its isogenic luxS mutant. During the preparation of this paper, the complete EHEC genome sequence was published and revealed that the EHEC genome contained 1.34 Mb of DNA not found in K-12 strain MG1655 and that the K-12 MG1655 genome contained 0.53 Mb of DNA not found in EHEC (37). Therefore, there might be additional K-12 and EHEC genes that are regulated by quorum sensing. Thus, although this analysis provides a starting point to examine which genes common to K-12 and EHEC are regulated by quorum sensing, this study cannot be regarded as the definitive analysis on quorum sensing in either EHEC or K-12. The K-12 gene array contains 4,290 genes, and 404 of those (ca. 10% of the array genes) were found to be regulated at least fivefold by quorum sensing. From these 404 genes, about 58% were up-regulated and 42% were down-regulated in the wild-type strain compared to what was found in the luxS mutant (data not shown). Selected genes shown to be regulated by quorum sensing in the array were further characterized by operon fusions, Northern analysis, Western blots, and phenotypic assays to confirm the array data.

The results of the array indicate that genes involved in cell growth and division are generally down-regulated by quorum sensing. The observation that the luxS mutant has a shorter generation time than the wild-type strain (Fig. 3) is consistent with the array data showing increased transcription of cell division genes, such as ftsQA, ftsE, and hflB, in the luxS mutant. The luxS mutant generation time can be slowed not only by complementation of the luxS mutation but also by addition of AI-2 as preconditioned media (Fig. 3), suggesting that this phenotype is due to the absence of AI-2 and not to other possible pleiotropic effects of a luxS mutation. Quorum sensing has been previously described as being involved in controlling bacterial growth and entry into stationary phase in Bacillus subtilis (reviewed in reference 27), which is consistent with cell-to-cell communication serving as a way to sense bacterial population density. In E. coli, Baca-DeLancey et al. (3) also reported a subset of genes that function in the uptake, synthesis, or degradation of amino acids that yield pyruvate and succinate and are regulated by quorum sensing using signals different from AI-2. In most organisms where quorum-sensing regulation has been studied, there is usually more than one signaling molecule and cross-talk between the signaling pathways often occurs, sometimes working synergistically and sometimes antagonistically (reviewed in references 14 and 27). Our results suggest that AI-2 is involved in metabolic processes that slow growth and that the AI-2 system probably interacts with other components important in the stationary phase to "sense" the growth conditions of the population.

Recently DeLisa et al. (12) reported that transcriptional fusions containing only the promoter Q2p of ftsQ were up-regulated threefold in response to quorum sensing through AI-2. Promoter Q2p is controlled by SdiA; these results would reflect an increase in SdiA production, which is consistent with our observation that there is a ca. 11-fold induction in sdiA transcription (Table 3) by AI-2. However, we observed a 2.5-fold decrease in the expression of ftsQA, which might be due to the fact that regulation of ftsQA transcription is also controlled by RpoS at the Q1p promoter in addition to SdiA at the Q2p promoter (44). In our experiments, the array data should reflect the RNA state of our culture at a given time point, and so these differences may be due to the bimodal regulation of ftsQA by RpoS and SdiA. Another factor that might also affect the growth rate of the luxS mutant is down-regulation of flhD in this mutant compared to in the wild-type strain (Fig. 4); flhD mutants have previously been reported to divide more rapidly as they enter the stationary phase (40).

The array data also suggested an induction of an SOS response by quorum sensing, with genes like recA, uvrA, and sulA being up-regulated in the wild-type strain compared to in the luxS mutant. The SOS response is triggered by a diverse set of treatments that damage DNA or inhibit DNA replication (reviewed in reference 29), and it was previously reported that quorum sensing inhibits chromosomal replication in E. coli (48). In the SOS response the RecA protease is activated, leading to the cleavage of the LexA repressor protein. Among the large number of genes whose expression is normally repressed by LexA and derepressed by the SOS response is sulA (reviewed in reference 41). Using a sulA::lacZ reporter fusion, we were able to confirm that transcription of sulA is up-regulated by quorum sensing through AI-2 (Fig. 2A). The sulA gene product inhibits formation of the FtsZ ring, which is essential for cell division, and in most strains is solely responsible for the filamentation that occurs as part of the SOS response inhibiting cell division (reviewed in reference 41). The onset of an SOS response and up-regulation of sulA transcription may also account for the generation time of the wild-type strain being longer than that of the luxS mutant (Fig. 2 and 3).

The Stx toxins are encoded within the late genes of -like phages and are expressed only when these phages enter the lytic cycle (35, 39). Recently Kimmitt et al. (25) reported that induction of an SOS response in EHEC induces the production of Stx2. Moreover, Fuchs et al. (17) reported that recA induction in vivo is involved in increased production of Stx2 due to Stx2 phage induction, which is controlled by RecA. We observed increased transcription of stx, using a stx::lacZ fusion, and increased production of Stx2 by Western blots in the wild-type and complemented strains compared to that in the luxS mutant (Fig. 2). Together with the increased transcription of recA and uvrA revealed by the array analysis, these results indicate that there is an effect on Stx2 production by luxS and suggest that this might be the result of an induction of an SOS response by quorum sensing through AI-2. We previously reported that an stx::lacZ fusion was not activated by quorum sensing (45). However, the lacZ fusion in the previous study contained only the proximal promoter immediately upstream of the stx structural gene and did not contain the Q-dependent promoter, which was contained in the fusion used in the present study.

Several genes involved in expression and assembly of flagella, as well as motility and chemotaxis, were up-regulated by quorum sensing. We were able to confirm the array data using transcriptional fusions to flagellar class I, II, and III genes; Western blots of flagellin protein; electron microscopy; and motility tests (Fig. 4 and 5 and data not shown). The luxS mutant produces fewer flagella than does the wild-type strain, which is consistent with the down-regulation of flhD, fliA, and fliC in this strain (Fig. 4). There is also decreased transcription of motA in the luxS mutant (Fig. 4), which is restored by complementing with the cloned luxS gene or by adding preconditioned media containing AI-2 (Fig. 4 and 5). Consistent with the transcriptional data, the luxS mutant shows smaller motility halos in semisolid agar and a slower swimming speed than does the wild-type strain (Fig. 5; video at www.medschool.umaryland.edu/CVD/kaper.html). One explanation for the decreased swimming speed of the luxS mutant might be the observation that it tumbles more than the wild-type strain, taking a longer time to swim the same distance. Regulation of flagellar genes by quorum sensing has been described in other bacteria, such as Yersinia pseudotuberculosis (1). The proteins comprising the type III secretion systems are homologous to the flagellar basal-body proteins, and since quorum sensing regulates the type III secretion system in EHEC, it is perhaps not too surprising that it also regulates flagellar expression, assembly, and motility. Moreover, coupled regulation of type III secretion and flagellar genes has been recently described in Salmonella (15, 20). In enteropathogenic E. coli (Girón and Kaper, unpublished), flagella are also involved in bacterial adherence and are essential for the formation of microcolonies, which is one of the steps in biofilm development.

This study expands the knowledge on quorum-sensing regulation in E. coli, showing that quorum sensing is a global regulatory system that controls not only genes involved in pathogenesis but also genes involved in bacterial metabolism, DNA repair, nucleotide and protein biosynthesis, and cell growth and division, among other functions. Our results suggest that quorum sensing is a very important regulatory mechanism through which E. coli senses and adapts to a given environment.