Integration of AI-2 Based Cell-Cell Signaling with Metabolic Cues in Escherichia coli

The quorum sensing molecule Autoinducer-2 (AI-2) is generated as a byproduct of activated methyl cycle by the action of LuxS in Escherichia coli. AI-2 is synthesized, released and later internalized in a cell-density dependent manner. Here, by mutational analysis of the genes, uvrY and csrA, we describe a regulatory circuit of accumulation and uptake of AI-2. We constructed a single-copy chromosomal luxS-lacZ fusion in a luxS + merodiploid strain and evaluated its relative expression in uvrY and csrA mutants. At the entry of stationary phase, the expression of the fusion and AI-2 accumulation was positively regulated by uvrY and negatively regulated by csrA respectively. A deletion of csrA altered message stability of the luxS transcript and CsrA protein exhibited weak binding to 5’ luxS regulatory region. DNA protein interaction and chromatin immunoprecipitation analysis confirmed direct interaction of UvrY with the luxS promoter. Additionally, reduced expression of the fusion in hfq deletion mutant suggested involvement of small RNA interactions in luxS regulation. In contrast, the expression of lsrA operon involved in AI-2 uptake, is negatively regulated by uvrY and positively by csrA in a cell-density dependent manner. The dual role of csrA in AI-2 synthesis and uptake suggested a regulatory crosstalk of cell signaling with carbon regulation in Escherichia coli. We found that the cAMP-CRP mediated catabolite repression of luxS expression was uvrY dependent. This study suggests that luxS expression is complex and regulated at the level of transcription and translation. The multifactorial regulation supports the notion that cell-cell communication requires interaction and integration of multiple metabolic signals.


Introduction
Quorum sensing is a process of cell-to-cell communication in bacteria via freely diffusible molecules called autoinducers, which modulates gene expression in a population density-dependent manner [1,2]. Many physiological processes and group behaviors such as motility, swarming, exopolysaccharide synthesis, stress survival, biofilm formation and virulence in bacteria are mediated by quorum sensing (QS) [2][3][4][5][6][7][8]. Interference with quorum sensing by quorum sensing inhibitors (QSI) can block infection processes and consequently have the potential to tackle infectious disease caused by antibiotic resistant pathogens [9]. E. coli is known to synthesize at least three types of autoinducers of which autoinducer-2 (AI-2) is generated as a byproduct of the activated methyl cycle and requires the action of enzyme, LuxS in a key step of the process [10,11]. The gene luxS encodes S-ribosyl homocysteinase that interconverts S-ribosyl homocysteine to homocysteine generating a furnanone borate ester, the active autoinducer, AI-2. AI-2 is thought to be a metabolic cue as it is generated from a central metabolic pathway. Homologs of luxI, the AI-1 synthase, are not found in E. coli and consequently Acyl-homoserine lactone (AI-1) is not detected in cell-free supernatant from cultures of E. coli, however a AI-1 receptor is present [11]. The luxS gene is conserved across many gram-positive and gram-negative bacterial pathogens and it is thought to be acquired by horizontal transfer million years ago [12][13][14]. Consequently synthesis of AI-2 is thought to be universal across both gram-positive and gram-negative bacterial species and thus AI-2 is considered as universal signaling molecule. Accumulation of AI-2 is controlled by a homolog of ribose uptake transport system, Lsr, (luxS regulated), which imports AI-2 from the external environment. Induction of the Lsr system at high cell density minimizes the levels of AI-2 from the extracellular milieu [15,16]. Several quorum sensing circuits are well established in many clinically important pathogens, many of which integrate with two-component regulatory systems [17][18][19][20][21].
Two-component regulatory systems (TCS) are unique bacterial signaling systems that facilitate adaptation in a rapidly changing environment [22][23][24][25]. TCS consists of a sensor kinase that senses and transmits external signals to its cognate response regulator by phosphorelay; the response regulator upon phosphorylation regulates gene expression usually by transcription activation [26]. TCS are attractive choice as drug targets for pathogenic bacteria as several of them are strongly are associated with virulence [27]. E. coli harbors more than thirty two-component regulatory systems, many of which are linked with pathogenesis [28]. The fimbrial gene regulations are an important contributing factor in virulence and establishing an infection. Previously, we have elucidated regulation of biofilm formation, adhesion, motility and virulence genes by an important two-component regulatory system, the BarA/UvrY/CsrA pathway in extra-intestinal pathogenic Escherichia coli. Particularly, we found that in avian pathogenic Escherichia coli (APEC) and uropathogenic Escherichia coli (UPEC), uvrY stimulates transcription of fimbrial and virulence genes [29][30][31]. Because the BarA/UvrY/CsrA pathway is strongly associated with virulence, we hypothesized that QS might be one of the mechanism by which virulence and other pleiotropic roles could be mediated through the pathway in E. coli.
Several other observations led us to hypothesis about the association of this TCS and AI-2 based quorum sensing. Similar to the association of virulence with the BarA/UvrY/CsrA TCS pathway, the luxS gene is also linked to pathogenesis of Enteropathogenic and Enterohemarrhagic E. coli [32,33]. Our previous studies have also demonstrated that luxS contributes to pathogenicity in APEC [34]. Furthermore, UvrY is a LuxR type transcriptional regulator which is commonly associated with quorum sensing [35]. Expression of small RNA CsrB and CsrC small RNAs are under the positive control of the BarA-UvrY-CsrA pathway in a cell density mediated manner. Phenotypes such as biofilm formation, swarming motility and virulence associated with the BarA-UvrY-CsrA pathway are dependent on community associations of microbes [31]. Most importantly, both the BarA/UvrY/CsrA pathway and AI-2 based quorum sensing contributes to biofilm formation in E. coli [36]. These observations led us to study the regulatory effect of the pathway on luxS based quorum sensing. In this study, we investigated the role of the uvrY and csrA genes in regulation of synthesis and uptake of AI-2 and luxS gene expression. Based on our results, we propose a regulatory circuit that controls quorum sensing at the transcriptional level via uvrY and post transcriptional level via CsrA. These findings suggest a principal role of BarA-UvrY-CsrA system in establishing early infection in pathogenic proteobacteria via quorum sensing and mediating a switch from a planktonic state to biofilm mode of persistence.

Recombinant DNA techniques
Standard molecular techniques were used for cloning [37]. Amplifications for cloning were performed by Tgo polymerase and other amplifications by Taq or Pfx polymerase. PCR products were cloned into pCR2.1 using the TOPO-TA cloning system (Invitrogen, Carlsbad, CA) and few clones were verified by sequence analysis. The uvrY gene was cloned in pBR322 generating pSM2 (p-uvrY) as described earlier [31]. Similarly the luxS gene was amplified using OSM34 and OSM35 (Table 2) and cloned into pCR2.1. A 700 bp EcoRI fragment was subsequently cloned into EcoRI site of pBR322 creating pSM3 (p-luxS). Both the luxS and uvrY open reading frame were oriented in the same direction as the tet gene in the vector.

Construction of chromosomal deletion insertion mutants
The uvrY and luxS genes in MG1655 were disrupted by lambda red recombination method [38]. The uvrY gene was deleted and replaced with a chloramphenicol cassette by using the primers OSM43 and OSM44. The luxS gene was similarly deleted with a kanamycin cassette by using the primers OSM49 and OSM50. P1vir transductions were performed as earlier described [39]. Mutations were transduced into relevant background whenever necessary and characterized for known phenotypes.

Construction of chromosomal luxS-lacZ transcriptional fusion
Since a disruption of the luxS gene causes growth-defect, we constructed a merodiploid strain with a single-copy luxS-lacZ transcriptional fusion incorporating upstream sequence from the luxS ATG codon. A 469 bp fragment incorporating 290 bp upstream regulatory sequences region and 59 codons of luxS gene were PCR amplified with Tgo polymerase from MG1655 chromosomal DNA using primers OSM53 which includes a SalI restriction site and OSM54 which includes a SmaI restriction site. The amplified fragment was cloned within the SalI-SmaI site of promoterless lacZ transcriptional fusion vector pSP417, a modified pRS415 vector with extended multiple cloning sites [40,41]. The clones were sequenced to check the integrity of the amplified fragment and the fusion junction. The plasmid-borne fusion was transferred to λRS45. The resulting recombinant phage, λPluxS-lacZ (λSM001) was used to transfer the fusion into MG1655Δlac, creating a merodiploid luxS + luxS-lacZ fusion (SM105). Single-copy fusions were isolated and verified by a Ter assay followed by measuring β-galactosidase activity. A single copy fusion integrated within the λ att site of the E. coli chromosome was selected to study luxS expression under various experimental conditions. We also observed that when a csrA::kan mutation was transduced from the parent strain TR1-5 MG1655, the resultant phenotype was that of a very slow growing strain as reported in S. typhimurium [42]. Because normal growing suppressors could be easily isolated after prolonged growth with aeration in LB broth, we selected single copy λF(luxS'-'lacZ) (hereafter referred as luxS-lacZ) fusion in TR1-5 MG1655.

Enzymatic assays
The extracellular AI-2 in cell-free supernatant was assayed using V. harveyi strain BB170 as described [43]. The reporter strain BB170, a luxN mutant of BB120 was chosen because of its sensitivity to AI-2 but not to AI-1. The positive controls were either BB152 (AI-1 -, AI-2 + ) or BB120 (AI-1 + AI-2 + ) and the negative control was Escherichia coli DH5α, a luxS mutant which was unable to synthesize AI-2. V. harveyi was cultured in autoinducer bioassay (AB) medium The V. harveyi reporter strains were grown overnight (~16 h) at 30°C on rotating wheels in AB medium, diluted 1:2500 into fresh medium and 180 μl of the diluted cells were added to  Serial dilutions of V. harveyi BB120 (wild type) and DH5α 13h-old culture supernatant were used as a positive and negative control respectively. The relative AI-2 activity was reported in relative light units (RLU) where background reading of media and surface is subtracted from the actual reading as directly reported by the instrument for each plate. β-galactosidase assay was determined as described [39]. All assays were performed in duplicate and repeated three times.

Electrophoretic Mobility Shift Assay
Since many response regulators can be phosphorylated by acetyl phosphate, we wanted to determine whether UvrY requires phosphorylation in order to interact with luxS promoter. The luxS promoter DNA was radiolabeled and various concentration of purified UvrY protein ranging from 1 to 2.5μM was used for gel-shift analysis. Purified UvrY was phosphorylated with 20 mM acetyl phosphate. Cold DNA was added to determine the strength of the interaction of the protein with the promoter.

Chromatin Immunoprecipitation assay
In vitro binding of UvrY protein to various promoters was determined by methods described previously [44,45]

Real time RT-PCR
Total RNA was isolated from wild type and relevant mutants at early exponential phase, midexponential phase and stationary phases and then subjected to qRT-PCR analysis with rrnA transcript as an internal control as described earlier [31]. Quantitative polymerase chain reaction (qPCR) and quantitative real-time polymerase chain reaction (qRT-PCR) were performed as per the manufacturer's recommendations. For qPCR, the first-strand cDNA was synthesized from 5μg of total RNA using Moloney Murine Leukemia Virus Reverse Transcriptase, Superscript II RNase H -(Invitrogen, Carlsbad, CA) and 50 ng of random hexamers (Invitrogen, Carlsbad, CA). For the PCR reaction, 10 ng of first-strand cDNA was amplified separately with 10 μM each of gene-specific primer and 16S rrnA gene-specific primers in a 25 μl total reaction volume with Taq polymerase in a Biometra T-Gradient PCR instrument (Biometra, Horsham, PA) for 30 cycles. At the end of several cycles, a gene-specific and an rrnA-specific reaction tube was removed. Five μl of the reaction products were resolved separately in a 1.2% agarose gel and the product intensities were quantitated by a BioRad Gel Documentation system (BioRad, Hercules, CA). The linear range of amplification for the rrnA gene was from 5-15 cycles in all backgrounds, while that of the luxS were from 12-22 cycles in the wild-type strain, and appeared later in the mutants. A qRT-PCR reaction was performed on the above set of samples under identical reaction conditions in a LightCycler (Roche, Indianapolis, IN) with SYBR Green-1 PCR Master Mix. The fluorescence signal from SYBR Green intercalation was monitored to quantify double-stranded DNA product formed after each PCR cycle. The threshold cycle for which a statistically significant increase in the amount of the PCR product is detected is denoted as Ct. Starting with individual cDNA pools from various genetic backgrounds, Ct values were determined for rrnA and luxS amplification products. The ΔCt values between samples derived from various strains were normalized with the rrnA product, as ΔCt = − (Ct rrnA − Ct wild-type ) − (Ct rrnA − Ct mutant ). Since PCR products double with each amplification cycle, the fold difference in the initial concentration of each transcript is determined by 2 ΔCt . The results derived from gel-based experiment indicated slightly less difference than the SYBR Green fluorescent method.

Rifampicin chase assay
RNA stability assay was performed as described earlier [31]. Briefly total RNA was isolated from cells at the entry of stationary phase when csrA is maximally expressed. Rifampicin was added to inhibit transcription initiation and before and after 2.5, 7.5 and 10 minutes post addition of rifampicin, total RNA was isolated and RT-PCR was performed. Rifmapicin was added at a final concentration of 500 μg/ml. A housekeeping control icd was kept as the internal control. Gene specific primers were used to amplify luxS and icd message ( Table 2).

Effect of extracellular AI-2 accumulation in uvrY mutant
The AI-2 accumulation of E. coli grown in LB broth was growth phase-dependent in consistence with previous studies [16,46]. The accumulation of AI-2 in the extracellular milieu peaks at the mid logarithmic phase, declines at the entry of stationary phase and vanishes once the cells shift deep within stationary phase of the growth cycle. No secondary peak of AI-2 was observed within a span of 24 hours. A mutation in the uvrY gene reduced AI-2 accumulation in the extracellular milieu compared to the isogenic wild-type (Fig 1A). The difference was pronounced in mid-exponential and in early stationary phase. In the complemented strain, the extracellular AI-2 accumulation was similar to the wild type. A very low level of AI-2 was detected in the luxS::kan mutant from cell free supernatant, and plasmid complementation in the mutant increased AI-2 levels to that of the wild type (S1 Fig).

Expression of luxS-lacZ transcriptional fusion in the uvrY mutant
The growth dependent AI-2 expression correlates with luxS expression until the culture reaches the stationary phase ( Fig 1B). The expression of the luxS-lacZ fusion in the uvrY::cam mutant was significantly lower in mid-exponential and early stationary phase as compared to the wild type. The basal level of expression of the fusion in uvrY::cam mutant was around 2-fold lower than wild type in both mid-exponential and stationary-phase and plasmid complementation of uvrY in the mutant partly restores the wild type expression levels ( Fig 1B).

Mutation in uvrY reduced expression of luxS transcript
Transcription of luxS was evaluated by quantitative reverse transcription PCR analysis and Northern analysis (Table 3 and    Relative Light Units. b The fold down-regulation is calculated as 2 ΔCt , Where ΔC t = (C t wt -C t rrnA )-(C t mutant -C t rrnA ).
c Standard deviation of three independent experiments. d Fold-difference of the luxS transcript transcript normalized with rrnA levels and compared to the wild-type strain.

In vitro and in vivo interaction of UvrY with luxS promoter
Purified UvrY interacts with luxS promoter without and with acetyl phosphate (Fig 2). The shift in the probe was shown by an arrow. The interaction of UvrY with luxS promoter is relatively weak compared to that with csrB promoter. Furthermore, interaction was also modulated by addition of negative and positive competitor DNA. Addition of positive competitor releases the binding whereas that of negative competitor tightens it. Since the interaction of uvrY with luxS promoter was relatively weak indicating there may be yet other factors associated with the interaction, we performed a in vivo chromatin immunoprecipitation of the luxS promoter fragment to determine the in vivo interaction using anti-uvrY antibody. Our result confirms the in-vivo binding of UvrY with luxS promoter (Fig 3). For the control reactions, we found the in vivo interaction of UvrY with csrB is relatively strong, while that with csrA was very weak, as expected from previous studies.

UvrY is required for cAMP-CRP repression of luxS-lacZ expression
We further wanted to test any additional regulator might be involved in expression of luxS-lacZ fusion. An important regulation occurs through the cyclic AMP-cyclic AMP Receptor Protein (cAMP-CRP) system. To further determine the role of UvrY in the glucose repression of luxS-lacZ fusion, we deleted adenylate cyclase encoding gene cyaA, which blockssynthesis of cAMP from ATP. Disruption of cya gene led to a constitutive expression of the luxS-lacZ fusion (Fig 4A). Addition of 1mM cAMP reduced the expression of luxS fusion to a basal level in the mutant. In case of ΔuvrYΔcya double mutant, the expression of the fusion is constitutive and addition of 1mM cAMP reduced the expression marginally (Fig 4B). The repression was faster in a Δcya mutant (30 min) compared to a ΔuvrYΔcya mutant (~90 min). Only addition

Effect of extracellular AI-2 accumulation and expression of luxS-lacZ fusion in csrA mutant
In contrast to uvrY mutant, mutation in the csrA leads to elevated levels of AI-2 relative to the wild type and complementation of csrA in the mutant reduced extracellular accumulation of AI-2 to the wild-type level (Fig 6A). The inverse correlation of AI-2 accumulation in the uvrY and csrA deficient strains suggested a regulation both at the level of transcription and post transcription of luxS and AI-2 uptake. The uptake of AI-2 was also evaluated in the uvrY and csrA mutants and indicated an inverse relationship in the lsr operon expression. Similarly in case of the csrA mutant, the expression of the fusion was significantly higher at the entry of stationary phase as compared to the wild type while complementation of the mutant restored the expression to the wild type level (Fig 6B). Interestingly, the expression pattern was similar to that of the wild-type strain, with maximum expression being observed as cells enter the stationary phase. A mutation in the csrB gene resulted in marginally higher expression of the luxS-lacZ fusion.
Regulatory interaction of CsrA with luxS mRNA stability or luxS mRNA leader sequence We evaluated the effect of csrA in luxS transcript stability. At the entry into stationary phase, loss of csrA resulted in stabilization of luxS transcript relative to the wild-type straining harboring csrA (Fig 7A). Quantification of the DNA intensities by ImageJ suggested a three-fold reduction in the half-life of luxS transcript in the wild-type relative to the ΔcsrA mutant ( Fig  7B). CsrA is known to bind leader regions of target genes with high affinity for GGA sites. We predicted the potential binding sites for CsrA in the luxS leader region with GGA sites using Lasergene program suite (DNAstar, Madison, WI, USA), (Fig 8A). Folding of the leader RNA using RNAFold in the Vienna RNA website indicated that one of the GGA site is actually present within the ribosome binding site of luxS (Fig 8B). Our results indicate that CsrA stabilizes luxS message stability at the entry into stationary phase. We also determined the interaction of CsrA with luxS leader region. Our results indicate that CsrA binds to luxS leader region and  Regulation of luxS influences luxS message stability (Fig 8C). We also determined the effect of hfq on the luxS fusion expression. We found that loss of hfq reduced the luxS fusion expression significantly and complementation of hfq in the mutant restored luxS expression to wild-type levels (S3 Fig).
Effect of expression of Lsr promoter,lsrA and lsrk in uvrY and csrA mutant In order to determine whether uvrY and csrA play a role in the regulation of uptake of AI-2 signal, the expression of the lsrA or lsrK or the Lsr promoter was evaluated in uvrY and csrA mutant. The Lsr promoter activity increased marginally in ΔuvrY mutant and approximately four-fold reduced in the csrA mutant at the entry of stationary phase (Fig 9). Expression of the lsrR was reduced both in the uvrY and csrA mutant ( Table 4). The uptake of AI-2 was also evaluated in the uvrY and csrA mutants and indicated an inverse relationship in the lsr operon expression.
Predicted integrative model of AI-2 based cell-signaling and the BarA/ UvrY/CsrA pathway We predicted a model integrating the BarA/UvrY/CsrA pathway with AI-2 based signaling (Fig 10). The BarA/UvrY/CsrA pathway controls both the synthesis and uptake of AI-2 via

Discussions
Quorum sensing is a bacterial population dependent cell-to-cell signaling mechanism mediated via synthesis, release and uptake of small freely diffusible molecules called autoinducers [47,48]. Apart from the requirement of QS on cell-density, other environmental cues such as pH, nutrient depletion and stress are important for quorum sensing systems to synchronize coordinated response for adaptation and survival of bacteria [49]. In Pseudomonas aeruginosa, the integration of metabolic cues with cell-cell signaling requires several regulators at the level of transcription and post transcription [50]. Metabolic adaptation through two-component regulatory system is achieved by signal detection and response by appropriate changes in gene expression by the cognate response regulator. The conserved pathways of BarA/UvrY/CsrA and its orthologs in the γ-subdivision of proteobacteria regulates virulence and secondary metabolism in E. coli, S. typhimurium, P. fluorescens and V. fischeri [51][52][53][54][55][56][57][58]. Our earlier work in E. coli demonstrated the pleiotropic role of BarA/UvrY/CsrA pathway in regulation of biofilm formation, motility and virulence gene expression [29][30][31]. Here we have shown the association of this regulatory pathway with luxS based AI-2 quorum sensing. Our data shows that multiple regulatory factors control the expression of LuxS including the BarA/UvrY/CsrA pathway, and global virulence regulator, crp and small regulatory RNAs. We found the interaction of UvrY with luxS promoter is very   [59]. These findings indicate that the BarA/UvrY/CsrA pathway integrates with AI-2 based cell-cell signaling in Escherichia coli and as a consequence the regulatory pathway play a critical role in metabolism, persistence, virulence and other community associated microbial behaviors. The control of synthesis and uptake of AI-2 suggest regulatory crosstalk of metabolic pathways with cell-cell signaling. Our results are similar to the regulation in Vibrio cholera where CsrA and three small RNA's regulate quorum sensing [60]. VarS/VarA twocomponent regulatory system in Vibrio cholera also regulates quorum sensing [20]. Furthermore luxS influences biofilm formation in both AI-2 and AI-2 independent manner in E. coli [61]. Our results further are in consistence with the role of cAMP-CRP mediated repression of luxS in AI-2 synthesis [62]. Swarming, a flagella and cell-density dependent phenomenon, is also reduced in uvrY delectation mutant and complementation of uvrY in the mutant restored swarming in UPEC [31]. This further confirms the integration of the BarA/UvrY/CsrA pathway with AI-2 based quorum sensing signaling and biofilm formation. This study indicates that the BarA-UvrY-CsrA pathway regulate E. coli metabolism, nutrient acquisition and cell-cell signaling and forms an underlying basis of bacterium-host signaling recognition and pathogenesis. Because both luxS and the BarA/UvrY/CsrA pathway are conserved in several bacterial species, such regulation and integration might be investigated in clinically relevant pathogens. The integration of the BarA/UvrY/CsrA pathway and its homologs with cell-to-cell signaling in diverse pathogenic bacteria may be explored to determine if integration of metabolic pathway with cell-to cell signaling can be viewed as common themes in biological regulation. However, many strains of E. coli lack a functional lsr gene which might lead to accumulation of AI-2 in the extracellular environment and consequently cross species signaling might be more frequent in certain niche for coordinating changes in gene expression doi:10.1371/journal.pone.0157532.g010 [63]. Depending on the polymicrobial environment, the regulatory interactions of the pathway will influence accumulation of AI-2 in the extracellular environment and consequently AI-2 signaling can have varying role in the community structure and dynamics. Targeting two-component regulatory system that integrates with cell-to-cell signaling may influence both intraspecies and interspecies microbial communication in a niche-specific manner. These strategies may be effective for development of novel therapeutics particularly for infections that are difficult to treat due to emergence of antibiotic resistance.