LsrR-Mediated Quorum Sensing Controls Invasiveness of Salmonella typhimurium by Regulating SPI-1 and Flagella Genes

Bacterial cell-to-cell communication, termed quorum sensing (QS), controls bacterial behavior by using various signal molecules. Despite the fact that the LuxS/autoinducer-2 (AI-2) QS system is necessary for normal expression of Salmonella pathogenicity island-1 (SPI-1), the mechanism remains unknown. Here, we report that the LsrR protein, a transcriptional regulator known to be involved in LuxS/AI-2-mediated QS, is also associated with the regulation of SPI-1-mediated Salmonella virulence. We determined that LsrR negatively controls SPI-1 and flagella gene expressions. As phosphorylated AI-2 binds to and inactivates LsrR, LsrR remains active and decreases expression of SPI-1 and flagella genes in the luxS mutant. The reduced expression of those genes resulted in impaired invasion of Salmonella into epithelial cells. Expression of SPI-1 and flagella genes was also reduced by overexpression of the LsrR regulator from a plasmid, but was relieved by exogenous AI-2, which binds to and inactivates LsrR. These results imply that LsrR plays an important role in selecting infectious niche of Salmonella in QS dependent mode.


Introduction
Bacteria control gene expression patterns in response to changes in their population density through a process called quorum sensing (QS). In QS, small signaling molecules called autoinducers are synthesized and released from the bacterial cells, and accumulate in the external environment [1]. When cell density reaches a certain level, producing autoinducer concentrations over a minimal threshold, the autoinducers bind to cognate receptors to promote changes in gene expression [2]. QS systems regulate a large number of physiological processes in bacteria including biofilm formation, virulence factor production, bioluminescence, sporulation, motility, and antibiotic production [3,4,5,6,7].
Several types of QS systems have been described in various species of bacteria. Many Gram-negative bacteria employ acyl homoserine lactones for intraspecies communication [8], whereas Gram-positive autoinducers are typically peptides [9]. Another QS pathway, in addition, in which signaling is mediated by the LuxSproduced autoinducer-2 (AI-2), has been found in both Gramnegative and Gram-positive bacteria [10]. The LuxS protein is found in over 55 bacterial species [11], and catalyzes the conversion of S-ribosylhomocysteine to 4,5-dihydroxy-2,3-pentanedione (DPD), which spontaneously cyclizes to form the signaling molecule AI-2 [10].
S. Typhimurium causes gastroenteritis and diarrhea in humans due to acute intestinal inflammation, and also causes a typhoid-like disease in mice. During infection of animal hosts, Salmonella invade epithelial cells of the small intestine. Invasion is mediated by a type three secretion system encoded in Salmonella pathogenicity island 1 (SPI-1). The SPI-1 type three secretion system forms a needlelike complex through which a number of effector proteins are translocated into host cells and mediate modification of the actin cytoskeleton [14]. Expression of SPI-1 is controlled by various environmental cues including oxygen tension, osmolarity, pH, and nutrients, reflecting the complex conditions in the intestinal lumen [15,16,17].
Previously, we reported that LuxS-mediated QS is required for normal expression of a subset of genes within SPI-1 and contributes to virulence of S. Typhimurium because deletion of the luxS gene decreased the transcription of SPI-1 genes and impaired invasion of Salmonella [18]. In contrast, another study recently reported that mutation of the luxS gene has no effects on the expression of SPI-1 and Salmonella virulence in a mouse infection model [19].
LuxS could have different effects on SPI-1 expression depending on the exact experimental conditions because of its pleiotropic functions [20]. Thus, the contradictory results may be resolved by a more complete understanding of LuxS effects on SPI-1 expression. We sought to clarify the role of LuxS by finding a factor that links LuxS/AI-2-mediated QS and SPI-1 expression. In the present study, we report that LsrR, a DNA-binding repressor of the lsr operon [21], negatively controls expression of SPI-1 and flagella genes and also regulates the ability of Salmonella to invade host cells.

Results
Salmonella LuxS Mutant is Indeed Defective for InvFdependent Expression of SPI-1 Genes and Attenuated for Virulence in Mice luxS deletion mutants have widely been used to reveal the roles of QS in bacterial physiology. However, the role of LuxS in Salmonella pathogenesis is controversial. We reported that a Salmonella strain lacking the luxS gene displays defects in virulence phenotypes associated with SPI-1 expression [18], but these findings were not reproduced by others [19]. These contradictory results could be due to the pleiotropic effects of luxS mutation [20], or reflect a recently described non-quorum sensing function of LuxS [22]. To clarify the role of LuxS in Salmonella virulence, it is necessary to identify the factor that directly connects LuxSdependent QS to regulation of SPI-1 virulence genes.
Before searching for such a factor, we confirmed that luxS is required for normal SPI-1-mediated Salmonella virulence. We showed that luxS deletion causes reduced expression of the invF gene, which in turn decreases expression of the InvF-regulated SPI-1 genes sicA, sigD, and sopE [18]. By conducting qRT-PCR, we determined that the mRNA levels of the sicA, sopB, and sopE genes decrease in both the DinvF strain and the DluxS DinvF double mutant strain (Fig. 1A). Expression of invF from a plasmid restored transcription of sicA, sopB and sopE to wild-type levels in the DluxS DinvF and the DluxS strains as well as the invF deletion strain (Fig. 1A). In contrast, the luxS expression from a plasmid failed to do so in both the DluxS DinvF and the DinvF strains ( Fig. 1A and data not shown). These data indicate that the reduced transcription levels of a subset of SPI-1 genes in the luxS deletion mutant were due to a decrease in InvF levels.
Our previous finding showed that the luxS mutant is attenuated for virulence in mice was based on bacterial numbers in liver and spleen [18], whereas Perrett et al. [19] reported that the luxS mutant did not display any significant differences in virulence compared to wild-type was based on number of live mice at day 12 post infection [19]. We decided to re-evaluate the virulence of the luxS mutant strain with a detailed monitoring (i.e. monitoring for every 12 or 24 h) of the survival of mice infected with Salmonella. Our data also show that all mice infected with either the wild-type or the luxS mutant strain died by 12 days after oral or intraperitoneal challenge ( Fig. 1B and C), but the survival of mice infected with the luxS mutant was significantly delayed compared to those infected with wild-type Salmonella ( Fig. 1B and C, p,0.0001). Taken together, these data emphasize that LuxS is necessary for normal expression of SPI-1 by controlling InvF and also for normal virulence of Salmonella.

LsrR Negatively Controls Expression of SPI-1 Genes in the Absence of LuxS
Because LuxS is not a DNA-binding protein, we hypothesize that there must be a transcription factor that links QS and SPI-1 regulation as described in other bacteria that regulate gene expression in response to QS [11]. In this sense, we focused on the LsrR protein, the repressor of the lsr operon, because it is the only known regulator controlled by Salmonella QS [1].
Phospho-AI-2 binds to LsrR inactivating its ability to act as a repressor and thereby inducing expression of the lsr operon [13]. Because the luxS deletion mutant cannot produce AI-2, LsrR is constitutively active. If LsrR negatively regulates the expression of SPI-1 genes, then deletion of the luxS gene would be expected to decrease expression of SPI-1 genes as we reported previously [18]. To test the hypothesis that LsrR lowers expression of SPI-1 genes, we constructed DlsrR and DlsrR DluxS strains carrying a chromosomal invF-lacZ fusion. b-galactosidase assay determined that deletion of lsrR had no effect on transcription of the invF-lacZ fusion ( Fig. 2A). However, the absence of LsrR compensated for the transcriptional defect in invF-lacZ expression in the luxS deletion mutant ( Fig. 2A). The regulatory effects of LsrR in the absence of luxS gene were analyzed in cultures with ( Fig.2A) or without shaking (data not shown). This result was confirmed to be due to the absence of LsrR function because expression of LsrR from a heterologous promoter reduced invF transcription in the DluxS DlsrR double mutant strain (Fig. S2). These results suggest that in the absence of LuxS-catalyzed AI-2 production, LsrR represses InvF expression, which in turn reduces transcription levels of the InvF-regulated genes within SPI-1.
The internalization of Salmonella into epithelial tissues of animal hosts is mediated by a number of gene products expressed from the SPI-1 locus [14], suggesting a role for LsrR in regulating internalization. We next examined whether LsrR regulates the invasiveness of Salmonella. Deletion of luxS impaired the ability of Salmonella to invade HEp-2 epithelial cells (Fig. 2B), as we reported previously [18]. If both lsrR and luxS are deleted, however, invasion of HEp-2 cells is comparable to that of the parental lsrR + luxS + strain (Fig. 2B). These results are in good agreement with the finding that lsrR deletion restored expression levels of InvF in the luxS deletion strain ( Fig. 2A). Taken together, our results suggest that, when activated by the absence of AI-2, the LsrR protein represses SPI-1 expression, which impairs Salmonella's invasiveness.

Overexpression of the LsrR Protein Lowers Expression of SPI-1 Genes
The down-regulation of invF expression by LsrR was only observed in the absence of LuxS (Fig. 2). This may be because the levels of active LsrR protein (i.e., LsrR not associated with AI-2) were elevated in the absence of LuxS. In contrast, LsrR could be present in a phospho-AI-2-bound inactive form in a luxS + strain, which could explain why deletion of lsrR gene alone did not alter invF expression ( Fig. 2A). If this were the case, overexpression of the LsrR protein could increase levels of active LsrR protein to a level sufficient to suppress expression of SPI-1 genes even in a strain carrying the luxS gene.
To test this idea, we constructed a strain carrying a chromosomal invF-lacZ fusion and harboring the plasmid pJH1 in which expression of LsrR is under control of the lac promoter. b-galactosidase assays show that IPTG induction of LsrR protein lowers the transcription levels of invF by about 4-fold (Fig. 3A). qRT-PCR analysis confirms that overexpression of LsrR decreases transcription of several SPI-1 genes: the transcript levels of the invF, sicA, sopB, and sopE genes were reduced by approximately 3-to 4-fold when the wild-type strain harboring pJH1 was grown with IPTG ( Fig. 3B). Next, we used Western blot analysis to investigate whether levels of a SPI-1 protein were affected by LsrR overexpression. Western blot analysis determined that overexpression of LsrR in the wild-type strain dramatically reduced levels of the SopB protein (Fig. 3C). Although the mRNA fold changes and protein levels were different in the LsrR-overexpression and wildtype strains, the negative regulatory effects of LsrR were consistent. The data presented here demonstrate that the LsrR regulator negatively controls expression of SPI-1 genes.

Overexpression of LsrR Inhibits Invasion into Epithelial Cells
We next investigated whether the reduced expression of SPI-1 genes resulting from LsrR overexpression affects the ability of Salmonella to invade mammalian cells. When HEp-2 epithelial cells  were incubated with the wild-type strain carrying pJH1 that had been grown with IPTG to induce LsrR expression, invasion was greatly impaired (i.e., 17% invasion compared to that of the wildtype strain; Fig. 4).
Expression of SPI-1 and motility are both required for Salmonella cells to gain entry into epithelial cells [14,23], but the requirement for motility can be overcome through the use of mild centrifugation [23]. To examine whether the defective invasion phenotype of LsrR-overexpressing Salmonella was solely due to the decrease in SPI-1 expression, or whether motility might also be affected, a mild centrifugation was applied to promote bacterium-host cell contacts during the invasion assay [23]. Although invasion was still decreased in the LsrR overexpressing strain when adherence of Salmonella cells to the epithelial cells was promoted using centrifugation ( Fig. 4), invasion was only 62% that of the wildtype strain (Fig. 4). Thus, centrifugation partially restored invasion of HEp-2 cells by Salmonella overexpressing LsrR.

The LsrR Protein Negatively Regulates Expression of Flagella Genes
Because the enhanced contact of bacterium-host cell by centrifugation partially recovered the defective invasiveness of LsrR-overexpressing Salmonella, we reasoned that motility of Salmonella might also be regulated by LsrR or LuxS. We examined the effect of luxS and lsrR deletions on expression of fliC which encodes flagellin, the major component of the flagellum [24], by using chromosomal fliC-lacZ fusion strains. b-galactosidase assays  We next investigated the effect of LsrR overexpression on flagella gene expression. b-galactosidase assays revealed that in the fliC-lacZ fusion strain carrying pJH1, the transcription levels of fliC decreased by approximately 4-fold upon IPTG addition (Fig. 5B). Moreover, qRT-PCR determined that the mRNA levels of the flhD and fliA, which encode key regulators of flagella synthesis [24], were reduced by approximately 3-fold when the wild-type strain harboring pJH1 was grown with IPTG ( Fig. 3D). Analysis of secreted proteins showed that LsrR overexpression in wild-type Salmonella greatly reduced levels of the flagella proteins, FliC, FliD, and FlgL (Fig. 5C). Levels of the SPI-1 effector proteins SipA, SipC, and SipD, were also lowered upon LsrR overexpression, consistent with the regulatory role of LsrR for SPI-1 genes (Fig. 5C). These results indicate that, in addition to regulating the SPI-1 genes, the LsrR protein also negatively regulates flagella genes.

The LsrR Protein Suppresses Motility of Salmonella
Because expression levels of flagella decreased in the LsrRoverexpressing Salmonella strain ( Fig. 5B and C), we asked whether the flagella structure is affected by LsrR levels. Transmission electron microscopy using a negative stain allows visualization of flagella appendages on the surface of the S. Typhimurium cells [24]. Using transmission electron microscopy, we were able to observe flagella on the Salmonella cells harboring the vector pUHE21-2lacI q (Fig. 5D), but not on cells carrying pJH1 grown in the presence of IPTG to induce LsrR overexpression (Fig. 5D).
We also compared the motility phenotypes of the wild-type Salmonella strains harboring pUHE21-2lacI q and pJH1. We found that both strains were motile in the absence of IPTG; however, motility of the strain carrying pJH1 was abolished in the presence of IPTG (Fig. 5E). These results show that the overexpression of LsrR suppresses Salmonella motility.

LsrR-mediated QS Controls Expression of SPI-1 and Flagella Genes
As LsrR is a repressor of an operon comprising AI-2 transporter, overexpression of LsrR turns off the QS system of Salmonella and results in impaired invasiveness with reduced expression of SPI-1 and flagella as described above. Those effects may be associated with AI-2 signaling and QS or be an artifact resulting from ectopic expression of LsrR without any relevance to QS. We reasoned that we could distinguish between these two possibilities by modulating the levels of AI-2 in the overexpression strain because the phospho-AI-2 binds to LsrR, inactivating it [13]. If LsrR regulates SPI-1 and flagella genes in response to QS, addition of exogenous AI-2 should result in derepression of those genes. This experiment is complicated by the fact that the strain overexpressing LsrR cannot import and phosphorylate AI-2 due to repression of the lsr operon and lsrK [25]. To circumvent this problem, the promoters of the lsr operon and the lsrK gene were replaced with the promoter of the chloramphenicol resistance gene (Pcat) which is constitutively expressed [26], and not repressed by LsrR (Fig. 6A).
To test whether LsrR regulates SP1-1 and flagella genes in response to QS, we examined the expression of chromosomal invF-lacZ and fliC-lacZ fusions in Pcat-lsr strains (containing Pcat-lsrA and Pcat-lsrK) harboring pJH1 that were grown in the presence or absence of IPTG or exogenous AI-2. As expected, overexpression of LsrR reduced expression levels of the invF and fliC genes ( Fig. 6B and C). Exogenous AI-2 restored expression of both the invF and fliC genes in dose dependent manner ( Fig. 6B and C). In agreement with the transcriptional changes, the decreased production of flagella and the SopB protein observed upon LsrR overexpression was restored upon the addition of AI-2 ( Fig. 6D and E). In WT with backbone plasmid, however, exogenous AI-2 with or without IPTG did not affect the expression of both invF and fliC genes ( Fig. S3A and B) and the production of SopB protein (Fig. S3C). We also asked whether addition of AI-2 to the Pcat-lsr Pcat-lsrK strain could restore the defective invasiveness resulting from LsrR overexpression. Overexpression of LsrR inhibited invasion of both the wild-type and the Pcat-lsr strains into epithelial cells (17% and 25% invasion relative to the wild-type strain, respectively; Fig. 6F). However, addition of exogenous AI-2 restored the invasion phenotype of the Pcat-lsr strain, but not the wild-type strain (Fig. 6F). Taken together, these findings suggest that the LsrRmediated QS signaling system modulates expression of SPI-1 and flagella, and consequently invasion of Salmonella into host cells.

Discussion
QS systems have been proposed to play a role in virulence of several pathogens based on studies of luxS mutant phenotypes, including Escherichia coli O157:H7 [27,28], Streptococcus pyogenes [4,29], Porphyromonas gingivalis [30], Neisseria meningitides [31], and Salmonella Typhimurium [18]. However, because the luxS gene has also been proposed to have non-QS functions [22], it can be difficult to define a role for LuxS-mediated QS in bacterial pathogenesis. Indeed, although we have reported the role of LuxS/AI-2 QS in regulation of SPI-1 in S. Typhimurium [18], another report indicated that LuxS did not play a role in SPI-1 regulation [19].
In S. Typhimurium, the regulatory network for AI-2 signaling is comprised of an AI-2 transporter complex, LsrABCD, the LsrR repressor, and the LsrK signal kinase [13]. LsrR regulates AI-2 uptake by directly binding and repressing the promoter of the lsr operon [21]. Because LsrR is the only known transcription factor in QS in Salmonella, we hypothesized that LsrR might connect LuxS/AI-2-mediated QS to SPI-1 regulation. Here, we report that the LsrR protein can negatively regulate expression of SPI-1 and flagella genes, and impair Salmonella's invasion into mammalian cells.
Consistent with previous report [32], the deletion of lsrR gene in wild-type Salmonella did not alter invF expression. However, the deletion of lsrR in the luxS mutant strain restored the reduced expression of the invF and fliC genes in the luxS mutant strain to the wild-type levels ( Fig. 2A and Fig.5A). These results suggest that LsrR is involved in regulation of SPI-1 and flagella genes, in contrast to a recent proposal that the regulatory function of LsrR is limited to the lsr operon [32]. The luxS deletion mutant would produce constitutively active LsrR due to the lack of AI-2, which supports the idea that LsrR might link LuxS-mediated QS and SPI-1 regulation and allows us to reveal additional target of LsrR in Salmonella Typhimurium. Overexpression of LsrR in wild-type Salmonella reduced both the mRNA and protein levels of SPI-1 (Fig. 3) and flagella genes (Fig. 5A, B, and C). These results emphasize that LuxS-mediated QS controls SPI-1 and flagella expression with AI-2-regulated LsrR activity playing a pivotal role.
We tested whether LsrR might regulate expression of SPI-1 and flagella genes by directly binding to their promoters. However, electrophoretic mobility shift assays failed to reveal LsrR protein binding to DNA fragments carrying promoters of SPI-1 and flagella genes such as invF, sopB, flhDC, and fliC, while LsrR did bind to the promoter DNA of lsr operon (data not shown). It is possible that additional factor(s) may be required for LsrR binding to the target promoter or that LsrR works indirectly. Although we do not understand the molecular basis of how the LsrR protein negatively controls transcription of SPI-1 and flagella genes, our findings clearly indicate that LsrR has regulatory targets than the lsr operon in Salmonella, as reported in Escherichia coli [25].
Overproduction of LsrR from a plasmid reduced entry into epithelial cells (Fig. 4), perhaps by generating increased levels of the active form of LsrR (i.e., LsrR not associated with phospho-AI-2). This result is consistent with our previous report that the luxS mutant was defective for invasion [18]. Moreover, decreased expression of flagella genes in bacterial cells possessing active form of LsrR has also been reported in a E. coli luxS mutant [7]. It is still possible that the altered expression of SPI-1 and flagella genes and reduced invasiveness resulting from LsrR overexpression might be artifacts unrelated with QS. However, that possibility is made less likely by the finding that treatment with exogenous AI-2 restored the behaviors of the LsrR-overexpressing Salmonella (Fig. 6). Overall, the data are consistent with a role for LsrR in regulating SPI-1 and flagella in response to QS.
The LsrR regulator represses expression of the lsr operon and the lsrK gene [25]. Consistent with this, exogenous AI-2 only counteracted the effects of LsrR overexpression in the engineered Salmonella strain (Pcat-lsr) in which the promoters of the lsr operon and lsrK were replaced with the constitutive Pcat promoter. To further verify whether the regulatory functions LsrR on SPI-1 and flagella genes are controlled by phospho-AI-2, we constructed a Salmonella strain (Pcat-lsrA) carrying only Pcat-lsrA and not Pcat-lsrK (Fig. S4A). We reasoned that the Pcat-lsrA strain could uptake AI-2 but not phosphorylate it due to LsrR-mediated repression of the lsrK gene which encodes the AI-2 kinase [1,25]. The Pcat-lsrA strain failed to restore expression of the invF and fliC genes upon the addition of AI-2 (Fig. S4B), indicating that LsrK is necessary for the LsrR protein to respond to AI-2 and control expression of SPI-1 and flagella (see Fig. S1).
When ingested by a host animal, Salmonella reach the small intestine and penetrate the epithelial cells. LuxS is a highly conserved protein, and many species of bacteria produce AI-2 [5,11]. Based on our findings, we propose that Salmonella could recognize the site of infection by detecting and internalizing the AI-2 signaling molecules produced by microorganisms in the normal gut flora in addition to recognizing well known environmental signals such as high osmolarity to induce the expression of SPI-1 [33]. Moreover, studies of the known QS signaling mechanisms employing AI-2 have revealed that most QS bacteria are sensing the extracellular AI-2 rather than importing it [9,11]. Thus, AI-2 molecules produced by normal flora would accumulate in the intestinal tract. The ability of Salmonella to consume AI-2 molecules could be beneficial for their virulence because the consumption of AI-2 could interfere with communication between other bacterial species and perhaps disrupt the flora in a way that is favorable for Salmonella pathogenicity.
In summary, we have presented evidence that active LsrR can negatively regulate expression of SPI-1 as well as flagella genes, and in doing so, reduce the ability of Salmonella to invade host cells. The effects of LsrR overexpression were relieved by exogenous treatment with the QS signal molecule, AI-2. These findings suggest that S. Typhimurium may require QS for successful invasion when they have reached the proper environment. QS may also enable Salmonella cells to modulate expression of virulence factors when they reached a sufficient population level in the right niche.

Ethics Statement
This study was carried out according to the recommended protocol for the care and use of laboratory animals from the Institute of Laboratory Animal Resource in Seoul National University, which is based on the Korean Animal Protection Law and Korea Food and Drug Administration regulation on the laboratory animals. The protocol was approved by the Committee on the Ethics of animal experiments of Seoul National University (Institutional Animal Care and Use Committee permit number: SNU-090717-3).

Bacterial Strains, Plasmids and Growth Conditions
Bacterial strains and plasmids used in this study are listed in Table 1. Salmonella enterica serovar Typhimurium strains used in this study are derived from strain SL1344. Phage P22-mediated transductions were performed as described [34,35]. All Salmonella strains were grown with (aerobically) or without shaking (anaerobically) at 37uC in Luria-Bertani broth (LB). For shaking cultures, 1.5 ml of media for glass tubes or 50 ml of media for flasks were used. For static cultures, 50 ml of media in sealed tubes were used. Antibiotics were used in the following concentrations; ampicillin 50 mg/ml, chloramphenicol 25 mg/ ml, kanamycin 50 mg/ml.

Construction of Strains
The method of Datsenko and Wanner [36] was used for chromosomal gene deletion, chromosomal lacZ fusion, and epitope tagging. For construction of a Salmonella strain carrying a chromosomal fliC-lacZ fusion, the Kan R cassette from plasmid pKD13 was amplified using primers fliC-lacZ-F and FliC-lacZ-R (see Table 2 for primer sequences) and the resulting polymerase chain reaction (PCR) products were introduced into the SL1344 strain containing the plasmid pKD46, followed by selection for fliC::Kan transformants. The Kan R cassette was removed using plasmid pCP20 [36] leaving an FRT site for introduction of lacZY using pCE70 [37]. Other chromosomal lacZ fusions were constructed following these procedures, and deletion mutants or epitope tagging mutants were constructed similarly including removal of the Kan R cassette.

Construction of Plasmids
An invF + plasmid for complementation was constructed as follows. The invF gene from SL1344 was amplified by PCR using primers invF-F2 and invF-R2, and the resulting product was cloned between HindIII and SphI sites of vector pACYC184 [38] to generate plasmid pJJ13. To construct the lsrR + plasmid pJH1 the lsrR gene from SL1344 was amplified by PCR using primers lsrR-pF1 and lsrR-pR1, and the resulting product was cloned between EcoRI and BanHI sites of vector pUHE21-2lacI q [39]. Primers used for plasmid construction are listed in Table. 2.
b-galactosidase Assay b-galactosidase assays were carried out in triplicate and enzyme activity (Miller units) was determined as described [40].

Quantitative Real-time Reverse Transcription (RT)-PCR Analysis
Salmonella strains were grown in LB media aerobically or anaerobically to desired growth phase, and total RNA was isolated using RNaesy Mini Kit (Qiagen). After DNase treatment of the RNA solution, cDNA was synthesized using Omnitranscript Reverse Transcription reagents (Qiagen) and random hexamers. Quantification of cDNA was carried out using 26iQ SYBR Green Supermix (Bio-Rad Laboratories), and real-time amplification of PCR products was performed using the iCycler iQ real-time detection system (Bio-Rad Laboratories). The relative amount of cDNA was calculated using a standard curve obtained from PCR using serially diluted genomic DNA as templates. The mRNA expression level of the target gene was normalized to the 16S rRNA expression level. Finally the values were normalized by those of wild-type. The sequences of the primers used are presented in Table 2.

Secreted Protein Analysis by Sodium Dodecyl Sulfatepolyacrylamide Gel Electrophoresis (SDS-PAGE)
The proteins secreted into media were analyzed by SDS-PAGE. Bacterial overnight cultures were diluted (1:100) in fresh LB and incubated aerobically or anaerobically. Cell-free spent media was obtained by centrifuging bacterial cultures (13,200 6  g for 20 min), collecting the supernatant, and then removing bacterial cells from the supernatant by filtration (0.22-mm pore size, Millipore). Cell-free spent media were mixed with prechilled trichloroacetic acid (TCA) to final concentration of 10%, chilled on ice for 2 h, and then centrifuged at 15,000 6 g for 15 min. The pellets were washed with acetone twice to remove all the TCA from the precipitates. After air drying, the pellets were dissolved in SDS sample buffer [41], and SDS-PAGE was performed using a 15% acrylamide gels from a mini-gel kit (Bio-Rad Laboratories, Inc). Proteins were visualized by staining with colloidal blue.

Transmission Electron Microscopic Analysis
Bacterial cells grown in LB for 16 h at 37uC without shaking were deposited on carbon-film grids, partially dried, and stained with 2.0% uranyl acetate. The negatively stained samples were observed using a 2000EX transmission electron microscope (JEOL, Ltd) at an acceleration voltage of 100 kV.

Motility Assay
Bacterial motility was measured as described previously [42]. Briefly, 0.3% LB agar plates were stab inoculated with 1 ml of a washed overnight culture. The plates were incubated upright at 37uC, and the diameter of bacterial growth was measured every hour. Where indicated, isopropyl-b-D-thiogalactoside (IPTG) was added to the LB agar at 100 mM.

Western Blot Analysis
Salmonella strains expressing a HA epitope tagged version of SopB (SopB-HA) from the chromosome were grown in LB broth at 37uC. Bacteria were collected by centrifugation, and cell lysates were prepared using B-PER solution (Pierce). Cell lysates were resolved in 12% SDS-PAGE and the SopB protein was detected using anti-HA antibody (Sigma). Western blot was developed using anti-mouse immunoglobulin G horseradish peroxidase-linked antibody and the enhanced chemiluminescence detection system (Amersham Biosciences).

Invasion Assay
HEp-2 cells (ATCC, CCL-23) were grown in Dulbecco's modified Eagle medium (DMEM; Invitrogen) supplemented with 10% fetal bovine serum (FBS), penicillin (50 U/ml), and streptomycin (50 U/ml). Confluent monolayers for infection with bacteria were prepared in 24-well tissue culture plates. Each well was seeded with 2610 5 cells suspended in DMEM-10% FBS without antibiotics and incubated for 1 h at 37uC under 5% CO 2 . The wells were washed three times with phosphate-buffered saline (PBS) before bacterial cells were added. Bacterial cells were washed with PBS, suspended in pre-warmed DMEM medium, and then added onto the monolayer at a multiplicity of infection of 10:1. After 1 h of incubation, the wells were washed three times with pre-warmed PBS to remove extracellular bacteria and then incubated for 1 h with pre-warmed medium supplemented with 100 mg/ml of gentamicin (Gm) to kill extracellular bacteria. The wells were then washed three times with PBS to remove the Gm. The HEp-2 cells were lysed in 1% Triton X-100 for 30 min, and then diluted with PBS. Dilutions of the lysed cell suspension were plated on LB agar medium to enumerate colony forming units (CFUs). As necessary, the 24-well tissue culture plates were subjected to low-speed centrifugation (10 min at 500 6 g) [14]. This mild centrifugation promotes bacterium-epithelial cell contact, which bypasses the requirement for bacterial motility in Salmonella invasion. CFU values for mutant strains were divided by the mean CFU value for wild-type Salmonella.

Animal Experiments
The experiment was conducted as described previously [43,44]. Six-week-old female BALB/c mice were purchased from the Institute of Laboratory Animal Resources at Seoul National University. The mice were kept under pathogen-free conditions in filter-topped cages in Individually Ventilated Cage Racks (MSRS-M70S; Orient Bio Inc.) containing sterile bedding, and were fed sterile food and water ad libitum. All mice were acclimatized for at least 1 week prior to experimentation. Cohorts of ten mice were infected by oral gavage or intraperitoneally with about 10 7 CFU or 10 3 CFU of Salmonella cells in 100 mL of PBS, respectively. Water and food were withdrawn 4 h before infection, and re-supplied 2 h postinfection. Mouse mortality was recorded daily. Figure S1 Schematic diagram of LsrR-mediated QS regulatory circuit in Salmonella typhimurium. AI-2 is synthesized by LuxS and accumulates extracellularly. Lsr transporter, encoded by lsr operon, internalizes the AI-2 into the cytoplasm, where it is phosphorylated to produce phospho-AI-2 by LsrK. LsrR represses the expression of lsr operon in the absence of phospho-AI-2, while it is de-repressed in the presence of phospho-AI-2 molecules that bind and inactivate LsrR [13]. In this study, we demonstrated that LsrR negatively controls the expression of SPI-1 and flagella genes and this regulation was abolished by phospho-AI-2. (TIF) Figure S2 LsrR is required for the regulation of invF expression by LuxS. Wild-type (WT) and other mutant strains carrying invF-lacZ fusion on chromosome were diluted in LB medium and grown with shaking, and b-galactosidase activity (Miller units