INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Many bacteria have evolved the ability to condition culture medium by secreting low-molecular-weight signaling pheromones in association with growth phase to control expression of specific genes, a process termed quorum sensing (19). Physiological processes controlled by quorum sensing occur in diverse species of bacteria and include bioluminescence (17), antibiotic biosynthesis (4), pathogenicity (34), and plasmid conjugal transfer (18). While acyl-homoserine lactones (HSL) appear to be the predominant quorum signal (or autoinducer [AI]) used by host-associated gram-negative bacteria, discovery of a second signaling pathway in the marine bacterium Vibrio harveyi (6, 8, 41) revealed an alternate AI, termed AI-2, which regulates bioluminescence in conjunction with AI-1 (N-(3-hydroxybutanoyl)-L-homoserine lactone) (7).

Importantly, AI-2 (or AI-2-like) activity has been observed in virtually all strains of pathogenic and nonpathogenic Escherichia coli and Salmonella enterica serovar Typhimurium (16, 40-42), requiring the luxS gene for synthesis (43). The physiological role of AI-2 in E. coli has not been clearly elucidated, but initial findings indicate that inhibition of chromosomal replication was subject to a quorum sensing mechanism (52). More recently, quorum sensing in E. coli has been implicated in regulating the expression and activity of SdiA, a LuxR-type transcriptional activator of the cell division genes ftsQAZ, through AI-2 (15, 39). In addition, extracellular factors which accumulate in enterohemorrhagic E. coli O157:H7 culture supernatants bind to the N-terminal region of SdiA for controlling the expression of virulence factors in a quorum-dependent fashion (25). Besides possible roles in cell division and pathogenesis, quorum sensing in E. coli was postulated to play a role in stationary phase gene expression (23, 27, 39), perhaps in a bimodal fashion with the stationary phase sigma factor rpoS or with other yet-to-be-determined quorum signals.

Recently, the application of global identification methodologies (e.g., DNA microarrays) has resulted in identification of quorum-regulated processes as well as the characterization of quorum circuit architecture in Streptococcus pneumoniae and Pseudomonas aeruginosa (16, 51). Therefore, a systematic investigation of native, quorum-mediated genes in E. coli was performed here to quantitatively analyze the global transcriptional pattern in response to the extracellular AI-2 signal molecule. To this end, DNA microarray analysis was utilized to quantify changes in transcription for every open reading frame (ORF) of E. coli strain W3110 in response to AI-2 signaling molecule.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains, plasmids, and media. E. coli strains used in this study were W3110 (F IN(rrnD-rrnE) rph-1) (22), E. coli Genetic Stock Center, New Haven, Conn., and MDAI2, a luxS::Tc^r derivative of W3110 (16). V. harveyi strains BB152 (luxL::Tn5 AI-1, AI-2⁺) and BB170 (luxN::Tn5 sensor 1, sensor 2⁺) for determination of AI-2 activity (41) were kindly provided by B. L. Bassler. Plasmid pGFPuv-ftsQ2p for quantifying quorum-regulated ftsQp2 expression is described elsewhere (15). Luria-Bertani (LB) medium contained 5 g of yeast extract (Sigma Chemical Co.) per liter, 10 g of Bacto tryptone (Difco) per liter, and 10 g of NaCl per liter and was supplemented with 50 mM glucose. Autoinducer bioassay medium and LM medium (L-marine) are given in detail elsewhere (8, 37).

Growth conditions. Overnight E. coli MDAI2 cultures grown aerobically in LB broth plus supplemental glucose (50 mM) were subcultured into 200 ml of LB plus 50 mM glucose (1% [vol/vol] inoculum). Cultures were grown aerobically at 30°C to an optical density at 600 nm (OD₆₀₀) of 1.0 followed by centrifugation (2,500 × g at 4°C) and gentle resuspension in ~1 ml of fresh LB. Resuspended cells were split equally into two parallel flasks each containing 100 ml of conditioned medium (positive or negative for AI-2) plus 50 mM glucose (prepared as outlined below) such that the culture OD₆₀₀ was maintained at ~1.0. Aerobic growth ensued for 20 min, at which time 5-ml samples were collected for total RNA extraction.

Preparation of cell-free culture fluids and conditioned medium. W3110 (luxS⁺) and MDAI2 (luxS) overnight cultures, grown aerobically at 30°C in LB plus 50 mM glucose, were used to inoculate 500 ml (1% inoculum) of fresh LB plus 50 mM glucose. Cultures were grown to an OD₆₀₀ of 3.0 (~8 h), and glucose analysis (YSI glucose analyzer model 2700) was used to confirm identical growth patterns in addition to growth rate calculations based on culture OD₆₀₀. Cell-free culture fluids or conditioned medium (CM) was prepared by centrifugation of 500-ml E. coli cultures for 10 min (10,000 × g at 4°C). Cleared supernatants were passed through 0.22-µm vacuum-driven Millipore filters and were stored at 20°C. Prior to use in AI-2 signaling experiments, CM was supplemented with 50 mM glucose and was assayed for AI-2 activity to confirm signaling conditions (positive or negative for AI-2). V. harveyi BB152 cell-free culture fluids were prepared analogously to obtain positive (+AI-2) control samples as reported previously (41).

AI activity assay. E. coli cell-free culture fluids were tested for the presence of AI-2 using the V. harveyi reporter strain BB170, which responds only to AI-2 (41). Luminescence assays were performed as outlined elsewhere (41), and luminescence was measured as a function of V. harveyi cell density by quantitating light production with a luminometer (EG & G Berthold). Data reported as fold activation were obtained by dividing the light produced by the reporter after addition of E. coli culture fluid by the light output of the reporter when growth medium alone was added (15).

Growth stimulation assays. Overnight cultures of W3110 and MDAI2 grown in LB were used to inoculate (1%, vol/vol) one of the following: LB plus 10% CM (+AI-2); LB plus 10% CM (+AI-2) plus 0.8% glucose; LB plus 10% CM (AI-2); or LB plus 10% CM (AI-2) plus 0.8% glucose. These experiments were performed in triplicate. OD₆₀₀ measurements were taken every 60 min over a 9-h period and used to calculate the specific growth rate for exponentially growing batch cultures. The specific growth rates were determined, with most accuracy for the first four data points.

RNA isolation and labeling. Harvested cells were centrifuged (5,000 × g at 4°C) and resuspended in lysozyme-containing buffer (5 µg µl¹ of lysozyme (Sigma) in 1× Tris-EDTA Buffer [Sigma], pH 8.0) at room temperature for 5 min. Total RNA was purified from 3 × 10⁹ to 5 × 10⁹ cells using a Qiagen RNeasy mini kit. RNA was eluted with diethyl pyrocarbonate (DEPC) water (Sigma) and quantified by measuring absorbance (A₂₆₀) with a spectrophotometer (Beckman DU 640). RNA samples (~75 to 80 µg) were concentrated to ~15 µl using a Microcon YM-30 filter (Millipore). Total purified RNA was labeled with either Cy3-dUTP or Cy5-dUTP as outline previously (49). Briefly, total RNA (~15 µl) was mixed with random hexamer primers (Amersham Pharmacia) and combined with 1X labeling mixture (first-strand buffer [Gibco-BRL], deoxynucleoside triphosphates (low TTP; Pharmacia), RNAsin (Promega), and DEPC water) and incubated at room temperature for 10 min. Labeling with Cy-3-dUTP (or Cy-5-dUTP) during a reverse transcriptase reaction using Superscript II (Gibco-BRL) was at 42°C for 1 h in the dark. After labeling, NaOH was added to the sample to hydrolyze RNA template and was incubated (65°C for 15 min) followed by neutralization with HC1 and Tris (pH 7.6). Following Microcon MY-30 filter purification, control and experimental samples were combined, pulse centrifuged (<1 min), washed, and concentrated by filter centrifugation prior to direct hybridization to glass microarrays.

Microarray hybridization procedures. Glass DNA microarrays (University of Wisconsin Gene Expression Center), consisting of full-length PCR products (spotted one time) from all E. coli ORFs according to Blattner et al. (9), were used to quantify relative mRNA levels by parallel two-color hybridization according to protocols described elsewhere (49). The number of features on the slide, therefore, consisted of the 4,290 annotated ORFs plus ~200 control spots, including fragmented E. coli genomic DNA, nonspecific salmon sperm DNA, 3× SSC (1× SSC is 0.15 M NaCl plus 0.015 sodium citrate), and six different yeast ORF PCR products. Slides were 36 spots by 36 spots by 4 panels, with each spot averaging 100 µ in diameter. Briefly, labeled probes were mixed with salmon sperm DNA, yeast tRNA, and PerfectHyb buffer (Sigma) prior to overnight hybridization to a DNA microarray (~12 h at 50°C). Arrays were washed with 0.2X SSC and 0.1% Sodium dodecyl sulfate (Sigma) for 2 min. Subsequent 2-min washes (three times) were with 0.2X SSC followed by dipping (C9. 10 times) in 0.05X SSC. Arrays were centrifuged to dry the surface and then scanned immediately using a GMS418 scanner (Genetic Microsystems) at 10-µm resolution and Arrayscan software (Genetic Microsystems). The resulting 16-bit TIFF images were analyzed using SCANALYZE software, publicly available at http://rana.stanford.edu/software/. RNA samples from experimental flasks (+AI-2 cultures) were first labeled with Cy5, and control RNA (AI-2 culture) was labeled with Cy3. Reverse labeling of the RNA (experiment, Cy3; control, Cy5) was then performed on a second array and used to verify all transcriptional induction ratios. Analogously to procedures of LaRossa and colleagues (50), further validation of microarray data was made via independent measures of RNA transcript level by RNA dot blots as outlined previously (15).

Data selection and analysis. Microarray data whose intensities were reproducibly higher than that of the background level were selected for analysis to eliminate expression ratios that were extremely high or low due to undetectable signal in control or experimental samples (26). Induction ratios (Cy5 for experimental RNA relative to Cy3 for control RNA) as well as background intensities for each dye were independently distributed in approximately normal fashion. Induction ratios were obtained by dividing background-corrected signal intensities of experimental samples by background-corrected intensities of control samples. Intensities used to calculate induction ratios were normalized as a percentage of the total of intensities of all the spots (Cy5 or Cy3) on the array, thereby accounting for the specific activity of the probes. That is, all Cy5 spots were normalized to the average Cy5 signal and vice versa for Cy3. Signals that were higher under control conditions were inverted to permit direct comparison between induction and repression ratios. Correlation among reverse-labeled repeat hybridizations with forward-labeled samples ranged from 0.756 to 0.998, and the standard deviation of duplicate (but reverse-labeled) induction ratios ranged from 0.08 to 0.45, providing a measure of reproducibility.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Generation of AI-2 signaling differential. Our overall objective was to identify all of the E. coli ORFs that exhibit a significant increase or decrease (more than twice the standard deviation [SD] of the mean induction ratio) in mRNA abundance caused by a differential in AI-2 signaling activity. MDAI2 (luxS) cells, which were unable to produce AI-2 as confirmed using a V. harveyi AI-2 activity assay (41), were grown to an OD₆₀₀ of 1.0 (ca. 3 to 4 h), split evenly, and resuspended in either CM exhibiting high AI-2 activity (+AI-2) or identically generated CM deficient in AI-2 signaling activity (AI-2) (see Materials and Methods). This procedure resulted in a greater than 300-fold difference in AI-2 signaling activity experienced by experimental cultures compared to that for negative controls (Table 1). Comparative measurements of transcript abundance were made by extracting RNA from cells immediately following 20 min of exposure to CM (+AI-2) or identically generated CM (AI-2). The final cell densities (at 20 min) of experimental and control cultures were identical (OD₆₀₀ of ca. 1.2; 17% change), suggesting that transcriptional changes were not substantially influenced by growth rate or cell density differences of control and experimental samples. Lastly, we performed growth stimulatory assays (see Materials and Methods) to determine whether the culture growth rates were affected by the presence of AI-2. The results (not shown) confirmed that there was no significant difference in growth rate between cultures exposed to CM (+AI-2) or CM (AI-2). Thus, it was concluded that under the conditions tested here, AI-2 does not stimulate or inhibit cell growth.

Identification of quorum-regulated genes in E. coli. Using DNA microarrays, we identified 242 genes representing approximately 5.6% of the entire genome that were upregulated (154 total genes) or repressed (88 total genes) more than 2.3-fold (corresponding to 2 SD above the mean induction ratio over the entire array) in the presence of AI-2. A total of 139 genes changed more than 2.9-fold (3 SD), and 23 genes changed more than 5-fold, including frwC (33.0-fold), yeiK (25.4-fold), and yidS (21.3-fold). On the contrary, 25 genes were more than 5-fold repressed, notably b2650 (27.8-fold), thiH (19.2-fold), and b2247 (15.2-fold). The entire dataset of induction ratios for all 4,290 annotated E. coli ORFs can be accessed at http://www.umbi.umd.edu/~cab/bentley/AI-2_array.html.

Involvement of AI-2 in multiple physiological processes. Importantly, we observed that a large number of the responding genes comprised three broad functional categories, according to Riley and Labedan (35). To be specific, we identified 22 genes involved in cell division, DNA processing, and morphological (cell shape) processes (Table 2), which was consistent with previous findings that a quorum sensing mechanism regulates DNA replication and cell division (48, 52). Twenty-three genes were involved in processes known to be quorum-regulated in other gram-negative bacteria, such as virulence, biofilm and exopolysaccharide formation, cell motility, and other surface-associated phenomena (Table 3) (14, 21, 31, 32, 45). Lastly, a cluster of 28 genes involved in small-molecule metabolism included genes which, to date, have not been implicated in cell-cell communication but may provide a link between central metabolism and quorum signaling, perhaps for the production and degradation of AI-2 itself (Table 4). Interestingly, addition of AI-2 did not significantly affect any of the currently known quorum sensing genes in E. coli under the timing and conditions studied here (Table 5) but did significantly alter expression of 10 putative signal transduction-associated genes (Table 6). Also, a large percentage (ca. 60%) of the genes exhibiting 2.9-fold changes (3 SD) were putative ORFs with no currently known function (Table 7).

View larger version (74K): [in this window] [in a new window]   FIG. 1.   Alignment of LuxO of V. harveyi with a putative 54-dependent transcriptional activator of E. coli (A), YgeV ORF95 of V. harveyi with YhbH of E. coli (B), both putative 54-modulating proteins, and LuxP of V. harveyi with the ribose periplasmic binding protein of E. coli, RbsB (C). Amino acids that match the consensus generated for the two sequences are boxed in black. The glycine-rich region encoding the nucleotide binding domain common of 54-interacting proteins is underlined, while the putative HTH DNA binding domains for LuxO and YgeV are boxed by a dashed line.

View larger version (5K): [in this window] [in a new window]   FIG. 2.   Genetic organization of the E. coli rpoN chromosomal region. The genetic organization of this region is similar to that described for the rpoN region of Vibrio cholerae and V. harveyi. yhbG encodes a probable ATP-binding cassette (ABC) transporter protein, yhbH encodes a putative 54 regulatory protein much like orf95 of V. harveyi, ptsN encodes a PTS system nitrogen regulator, and yhbJ has no known function (24).

View larger version (21K): [in this window] [in a new window]   FIG. 3.   Induction of ftsQA expression through the p2 promoter by W3110/pGFPuv-ftsQ2p reporter cells in LB medium plus 50 mM glucose or LB medium plus 50 mM glucose and supplemented with L-ribose (2 g/liter). AU, arbitrary units.

Independent measure of RNA levels confirms AI-2 stimulatory effect on gene expression. Previously, quantitative reverse transcription-PCR (50) and lacZ transcriptional fusions (45) were used to independently validate microarray expression data. In a similar manner, total RNA dot blotting was performed as previously outlined (15) and used to quantitatively verify the expression level changes of representative genes identified by microarray analysis. For these confirmatory experiments, RNA samples were harvested from a replicate experiment (see Materials and Methods), and induction ratios were compared between the two methodologies (Table 8). Reassuringly, the expression changes of 8 genes (out of 9 tested) were confirmed using total RNA dot blotting, which represented agreement (ca. 89%) similar to that previously reported (50). The lone discrepancy, thiH, was significantly repressed (19.2-fold) according to comprehensive transcript profiling but was upregulated (2.8-fold) according to the dot blotting procedure. The source of this difference, while not definitively resolved, was possibly due to high background signal in the film-based development associated with RNA blotting.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

To determine the transcriptional response of E. coli to the AI-2 quorum signal, cells deficient in AI-2 production (W3110 luxS::Tc^r) (15) were exposed for 20 min to medium conditioned by either AI-2-producing (W3110 luxS⁺) or non-AI-2-producing (W3110 luxS) cells. Generation of a 300-fold differential in AI-2, confirmed by an AI-2 activity assay (41), led to the discovery that almost 6% of the E. coli genome (242 genes) was modulated more than 2.3-fold (2 SD) in response to AI-2-regulated quorum sensing. Consistent with these results, it has been conservatively estimated that 3 to 5% of all genes in P. aeruginosa are regulated by acyl-HSL quorum signaling, as was partially demonstrated by screening a library of lacZ promoter probes which uncovered 270 genes showing more than twofold stimulation (70 showed more than fivefold stimulation) (51).

It is now well known that V. harveyi uses a species nonspecific signaling pathway mediated by AI-2 for regulating lux gene expression (7, 8, 41). Recognition of AI-2 by LuxP, homologous to the ribose binding protein of E. coli, has been proposed to transmit the signal to the LuxU phosphorelay protein via interaction with a hybrid sensor kinase, LuxQ (7, 8, 28). In turn, the signal is relayed to a central regulator, LuxO, which, upon interaction with ⁵⁴, indirectly represses the lux operon (28). In this study it was observed that AI-2 induced a gene encoding a ⁵⁴-dependent transcriptional activator (ygeV) as well as a ⁵⁴ modulator (yhbH), leading to our postulation that E. coli may employ ⁵⁴ during quorum sensing in a fashion analogous to that of V. harveyi. The striking similarity of the rpoN chromosomal regions of V. harveyi and E. coli along with the observed increase in yhbH expression (2.5-fold) suggests similar regulatory controls exist. Finally, in addition to regulating light production, LuxO and ⁵⁴ regulate siderophore production and colony morphology, demonstrating that multiple processes are regulated by quorum sensing in V. harveyi (28). In the present study, quorum-regulated genes existed in several functional classes, and while systematic determination of quorum-controlled processes still remains, it is clear that AI-2 also affects multiple processes in E. coli and perhaps enables population-wide coordination of these events.

Determination of quorum-regulated processes in E. coli has been elusive, as discovery of a quorum signal was only recently made. However, the finding that an extracellular factor exhibited inhibitory activity during initiation of DNA replication (52) provided preliminary evidence that E. coli employed a quorum sensing mechanism. Consistent with this observation, a gene encoding the integration host factor alpha subunit (himA) involved in the replication of the E. coli chromosome, and holE, a gene encoding the  subunit of DNA polymerase III (necessary for elongation), were upregulated 3.9- and 3.1-fold, respectively, in response to AI-2. A role for quorum sensing in cell division was first demonstrated by SdiA-mediated changes in ftsQAZ expression from the p2 upstream promoter induced by CM (20, 39, 48) and later attributed to AI-2 (15). In this study, expression of sdiA, a luxR-type transcriptional regulator, was observed to increase only slightly (2.0-fold) in response to AI-2, indicating either that AI-2 does not significantly affect sdiA expression or that the effect occurs on a different time scale than that tested here. Accordingly, expression of ftsQ, ftsA, and ftsZ were relatively unchanged (1.1-fold, 1.0-fold, and 1.4-fold, respectively), although ftsE, encoding an ATP-binding component of a membrane-associated complex involved in cell division, decreased 3.1-fold. While the results presented here appear to contradict the earlier findings of increased AI-2-mediated ftsQAZ transcription through p2, we conclude that the p2 construct alone behaves differently than the p1 and p2 promoters acting in concert. This observation is supported by experiments using CM that showed that an extracellular factor stimulated ftsQA expression fivefold from P2_ftsQ but only two- to threefold from both P1_ftsQ and P2_ftsQ together (39). Bassler and colleagues also reported that fusions of both p1 and p2 promoters to lacZ were not significantly altered by the presence of AI-2. Therefore, that we earlier observed ftsQAZ expression changes stimulated by AI-2 through promoter p2 alone suggests that under physiological conditions ftsQAZ expression is influenced by overlapping regulation from the neighboring rpoS-dependent P1_ftsQ promoter. In fact, it has been documented that the two ftsQA promoters are regulated differentially, with expression from P2_ftsQ occurring throughout growth and dependent on sdiA, while that from P1_ftsQ (a gearbox promoter) increases as the growth rate declines and is dependent on rpoS (1, 39, 48).

Additional evidence that quorum sensing positively regulates cell division was the observed 3.2-fold decrease in expression of dicB, an inhibitor of the synthesis and activity of FtsZ. While sdiA is known to positively regulate cell division, inhibition of division can occur via derepression of dicB, whose gene product cooperates with MinC to inhibit FtsZ assembly, blocking septation at all potential division sites (13). Therefore, repression of dicB by the AI-2 quorum signal might exert additional positive control over cell division. Further, rcsB, another luxR-type transcriptional regulator protein known to affect colanic acid capsular polysaccaride synthesis, demonstrated a 3.5-fold increase in transcription. The role of rcsB in activating the ftsA and ftsZ genes (10), perhaps through P1_ftsQ or P2_ftsQ, coupled with its increased transcription induced by AI-2, suggests that a quorum regulatory mechanism governs these distinct processes. In further support of a role for AI-2 in exopolysaccharide biosynthesis was the increased transcription of wzb (6.2-fold increase), a gene found within the colanic acid gene cluster and, in conjunction with wzc, that is known to participate in the export of the extracellular polysaccharide colanic acid from the cell to the medium (46).

Using RegulonDB software (36) available at http://www.cifn.unam.mx/regulondb/, we obtained a predicted 81-bp promoter of rcsB, which was subsequently input to GRASP-DNA software (38) available at http://www-bioeng.ucsd.edu/~grasp/home.html and was used to identify homologous putative DNA-protein binding sites. This analysis revealed significant homologous regulatory regions upstream of rcsB and ompG, which was interesting, as ompG expression was similarly upregulated (5.1-fold). Additionally, the threonyl-tRNA synthetase, thrS, having ca. 46% identity to a short segment of luxU (data not shown), was observed to increase 3.1-fold. A putative 81-bp promoter of thrS had upstream promoter homology with rfaJ, a lipopolysaccharide biosynthesis gene which was similarly upregulated (3.7-fold). Of note, thrS mRNA has been shown to accumulate with increasing growth rate (11), which might be a consequence of its response to AI-2, which also accumulates in a growth-rate-dependent fashion (15). Similarly, operon structure could be probed using RegulonDB. All of the AI-2-responding genes were examined using the RegulonDB graphical interface, and while many of the genes occurred within predicted (as opposed to known) operons, only the cheAW-motAB, potABCD, and rfaQGPSBIJYZK operons contained multiple genes responding to AI-2. In the first two cases, two genes responded similarly (e.g., rfaY and rfaJ, and potA and potB), while in the last case the responses of cheW and motB were in different directions. In all cases putative promoters are located between the identified genes, so a differential response might be expected. Interestingly, it has been shown that mutation of the att operon of Agrobacterium tumefaciens, a 10-kb region of 9 ORFs bearing strong homology to the pot operon of gram-negative bacteria, resulted in avirulence and inability to attach to plant cells (31). However, the ability of att mutants to bind to host cells was restored by the addition of conditioned medium during incubation of the bacteria with the host, suggesting that either efflux or uptake of an extracellular factor necessary for attachment through the spermidine pathway was blocked in mutant strains. This is very interesting in light of the 3.1- and 5.1-fold decrease in potAB expression in response to E. coli CM containing AI-2. This approach demonstrates that combination of global expression data with powerful bioinformatic algorithms, such as GRASP-DNA and RegulonDB, can elucidate potential regulatory overlap from transcriptional data.

Transcription of several other exopolysaccharide (rcsB, rfaD, rfaJ, rfaY, and rnk)- and virulence (hha and evgS)-related genes responded to AI-2. Outer surface polysaccharides are important components in the virulence of many pathogens, as they mediate direct interaction between bacteria and their immediate environment. While the E. coli strain studied here was not virulent, this was not entirely surprising, as many pathogenic gram-negative bacteria regulate virulence via quorum sensing (14). For example, hha, encoding the regulator of the hemolysin operon and reported to mediate the environmental regulation of virulence factors in P. aeruginosa (51), increased 11.1-fold in response to AI-2. Interestingly, ompA expression increased 3.2-fold, consistent with existing evidence that OmpA, in addition to maintaining outer membrane integrity, might play an important role in virulence of Pasteurella haemolytica (30). Of note, ompA expression was reported to decrease by 59% in an E. coli hha mutant (5). Also, the putative E. coli virulence gene, evgS, which constitutes a two-component system with the luxR-type regulator evgA that is structurally and functionally similar to the bvgAS two-component regulator of virulence factors in Bordatella pertussis (44), was repressed 2.8-fold. Finally, csrA, a global repressor of glycogen biosynthesis that alters stability of specific mRNA targets (29), increased 2.8-fold. While csrA has not been directly associated with quorum sensing in E. coli, structural and functional homologues regulate invasion genes in S. enterica serovar Typhimurium (2) and extracellular enzymes and N-(3-oxohexanoyl)-L-homoserine lactone quorum signals and pathogenicity in Erwinia carotovora (rsmA) (12). Additionally, csrA has been documented to affect cell size and surface properties, which is in agreement with the transcriptional changes of several murein sacculus-associated morphological genes. These genes include bolA (10.2-fold), an ftsZ-dependent regulator of the murein genes (1), and mreD (3.6-fold), encoding a rod-shape-determining protein as well as several exoskeletal (fimbriae, flagella, and curli surface fibers) genes, such as yadK (3.8-fold), yadN (3.5-fold), crl (3.5-fold), b1502 (3.0-fold), yehA (2.7-fold), fliP (2.7-fold), and flgN (3.7-fold). The coupling of morphological gene expression to AI-2 quorum signaling might ensure that the cytoskeletal framework be temporally regulated in association with growth phase and cell cycle progression.

Overall, our results yield significant insight into possible AI-2-coordinated changes in gene regulation that might temporally and spatially unify processes such as cell division, morphogenesis, and cell surface architecture. Interestingly, as many as 10 known sensors and/or transcriptional regulators as well as 10 putative signal transduction genes responded to increased AI-2 signaling, which, along with several other candidate genes and processes, warrant further study in the context of AI-2-stimulated quorum regulation. It is clear that quorum sensing is a complex signaling circuit that is built upon transducing elements that allow integration and channeling of multiple environmental cues, and elucidation of AI-2-controlled genes is a critical first step in mapping the metabolic pathways that define the E. coli quorum circuit.