Salmonella enterica Serovar Typhimurium Can Detect Acyl Homoserine Lactone Production by Yersinia enterocolitica in Mice

ABSTRACT LuxR-type transcription factors detect acyl homoserine lactones (AHLs) and are typically used by bacteria to determine the population density of their own species. Escherichia coli and Salmonella enterica serovar Typhimurium cannot synthesize AHLs but can detect the AHLs produced by other bacterial species using the LuxR homolog, SdiA. Previously we determined that S. Typhimurium did not detect AHLs during transit through the gastrointestinal tract of a guinea pig, a rabbit, a cow, 5 mice, 6 pigs, or 12 chickens. However, SdiA was activated during transit through turtles colonized by Aeromonas hydrophila, leading to the hypothesis that SdiA is used for detecting the AHL production of other pathogens. In this report, we determined that SdiA is activated during the transit of S. Typhimurium through mice infected with the AHL-producing pathogen Yersinia enterocolitica. SdiA is not activated during transit through mice infected with a yenI mutant of Y. enterocolitica that cannot synthesize AHLs. However, activation of SdiA did not confer a fitness advantage in Yersinia-infected mice. We hypothesized that this is due to infrequent or short interactions between S. Typhimurium and Y. enterocolitica or that the SdiA regulon members do not function in mice. To test these hypotheses, we constructed an S. Typhimurium strain that synthesizes AHLs to mimic a constant interaction with Y. enterocolitica. In this background, sdiA + S. Typhimurium rapidly outcompetes the sdiA mutant in mice. All known members of the sdiA regulon are required for this phenotype. Thus, all members of the sdiA regulon are functional in mice.

There are more than 2,500 serovars of Salmonella enterica, each with a different host range and disease manifestation. The two most common manifestations are enteric fever (typhoid) and gastroenteritis (14,40,42). Typhoid fever is caused by Salmonella enterica serovar Typhi, which is strictly host adapted to humans and higher primates and does not cause disease in mice. In contrast, Salmonella enterica serovar Typhimurium causes gastroenteritis in a wide range of animals (humans, cattle, pigs, poultry, horses, and sheep). S. Typhimurium is quite common in humans, accounting for 26% of all U.S. isolates (10). However, in susceptible mouse strains (slc11A1 Ϫ/Ϫ strains, such as BALB/c and C57/Bl6), S. Typhimurium does not cause gastroenteritis but instead causes a systemic typhoid-like disease, thus providing an extensively used animal model of human infection by S. Typhi. Some Salmonella-resistant mouse strains, such as CBA/J and 129X1/ SvJ, can be chronically colonized by S. Typhimurium, thus providing a model of intestinal persistence (24,25).
S. Typhimurium encodes a transcription factor of the LuxR family, named SdiA (3). This transcription factor detects and responds to acyl homoserine lactones (AHLs) produced by other species of bacteria (1,22,28,35). To date, SdiA is known to activate two loci, containing a total of seven genes (3,35). One locus, the rck operon, is on the virulence plasmid and contains six genes. Rck (resistance to complement killing) is predicted to be an eight-stranded ␤-barrel protein that resides in the outer membrane (13,18). Rck promotes adherence to epithelial cells and the extracellular matrix proteins fibronectin and laminin (15) and also inhibits the polymerization of complement component C9 in the bacterial membrane (19). The pef operon is located directly upstream of the rck operon and encodes "plasmid-encoded fimbriae" that specifically bind the trisaccharide Gal␤1-4(Fuc␣1-3)GlcNAc, also known as the Lewis X blood group antigen (12). Two genes in the rck operon, pefI and srgA (sdiA-regulated gene), appear to affect the expression and function of the pef operon. The pefI gene encodes a transcriptional regulator of the pef operon, and the srgA gene is a dsbA paralog that specifically catalyzes the oxidation of a disulfide bond in the PefA fimbrial subunit (6,30). Therefore, sdiA may indirectly affect the expression and assembly of plasmid-encoded fimbriae on the surface of the bacterium. The remaining genes in the rck operon have unknown functions but include a putative lipoprotein, srgB, and two putative transcription factors, srgC and srgD. The second locus regulated by SdiA is at 33.6 centisomes of the S. Typhimurium chromosome and encodes a single gene, named srgE. Very little is known about srgE except that it appears to be a singlegene horizontal acquisition and the predicted product contains a putative coiled-coil domain (35). Recently a computational approach predicted that SrgE is a secreted substrate of a type III secretion system (srgE is referred to as STM1554 in the study) (31). This has not been experimentally verified.
We recently tested the hypothesis that S. Typhimurium uses SdiA to detect the AHL production of the normal intestinal microbiota of the host (36). Surprisingly, we found that S.
Typhimurium did not detect AHLs during transit through the intestinal tract of a guinea pig, a rabbit, a cow, 5 mice, 6 pigs, or twelve chickens, indicating that the normal microbiota of these animals did not produce AHLs of the correct type or at a sufficient concentration to activate SdiA (36). However, S. Typhimurium did detect AHLs in an sdiA-dependent manner during transit through turtles (36). The turtles were found to be colonized with the AHL-producing organism Aeromonas hydrophila, suggesting that this organism was the source of the AHLs. Because A. hydrophila is a significant pathogen in aquaculture, this led us to a revised hypothesis that S. Typhimurium uses SdiA to detect the AHL production of other bacterial pathogens in animals. In this report, we test this hypothesis by determining if S. Typhimurium can detect the AHL production of Yersinia enterocolitica in mice.
Yersinia enterocolitica is a significant cause of food-borne disease in humans (29). Like Salmonella, this organism preferentially invades the lymphatic tissues of Peyer's patches (8,9,37). Y. enterocolitica has a quorum sensing system composed of yenI, which encodes an AHL synthase, and yenR, which encodes a LuxR-type AHL receptor (4,5,39). AHL produced by yenI has been detected in mouse tissues following infection with Y. enterocolitica (21). The genes regulated by this quorum sensing system have not been identified systematically, but it is known to regulate fleB, which encodes the flagellin subunit (4,5). Swimming motility is temporally delayed and swarming motility is eliminated in a yenI mutant (4,5).

MATERIALS AND METHODS
Bacterial strains and media. All bacterial strains and plasmids used are listed in Table 1. E. coli and Salmonella were grown in Luria-Bertani (LB) broth at 37°C unless otherwise indicated (EMD Chemicals, Gibbstown, NJ). LB agar plates and LB motility agar were made by adding agar to LB to 1.2% or 0.25%, respectively (EMD Chemicals). Yersinia enterocolitica was grown in LB broth at 26°C to 28°C. For recovery of Salmonella and Yersinia from animal tissues or feces, xylose-lysine-desoxycholate (XLD) agar plates were used (EMD Chemicals). The antibiotic ampicillin (Amp), kanamycin, nalidixic acid (Nal), or tetracycline (Tet) was added to a concentration of 150, 100, 20, or 20 g/ml respectively, when needed (Sigma-Aldrich, St. Louis, MO).
Construction of a yenI::kan mutation in Yersinia enterocolitica JB580v. The yenI::kan mutation was introduced into Y. enterocolitica strain JB580v by allelic exchange. The suicide plasmid pKNG::yenI::kan (39) was mobilized from E. coli S17-1pir to Y. enterocolitica JB580v. Mutants were selected on LB with Kan and Nal containing 5% sucrose and confirmed to be AHL negative. One isolate was named GY4493 and saved for future use.
Construction of S. Typhimurium containing the yenI gene of Yersinia enterocolitica. The lambda P R promoter was removed from pIA230 by digestion with BamHI and NotI and cloned into p2795 digested with the same enzymes, resulting in pJLD500. The yenI gene of Yersinia enterocolitica JB580v was amplified by PCR using primers BA1552 and BA1553 ( Table 2). The resulting PCR product was cloned into pCR2.1 (Invitrogen) and then removed using BamHI and EcoRI. The fragment was cloned into pJLD500 digested with BamHI and EcoRI, resulting in pJLD1600, which contains the complete P R -yenI-FRT-kan-FRT cassette. PCR was used to amplify the P R -yenI-FRT-kan-FRT cassette from pJLD1600, and sequences spanning the desired integration site downstream of pagC were appended to the primers (BA1539 and BA1540) ( Table 2). The resulting PCR product was electroporated into strain 14028 containing the arabinose-inducible, temperature-sensitive plasmid pKD46, which encodes the Red recombinase, followed by selection on LB plus Kan (LB Kan) at 37°C (16). The presence of the insertion was confirmed by PCR with a combination of primers BA1554, BA1555, c1, and c2 (16). The resulting strain was found to produce AHLs by cross-streaking against the biosensor strains 14028(pJNS25) and BA612(pJNS25) on LB plates at 37°C (35). One isolate was saved as JLD1200. We were unable to cure pKD46 from this strain for unknown reasons, so the P R -yenI-FRT-kan-FRT cassette was transduced from JLD1200 into 14028 using phage P22HTint (26). The insertion site and production of AHLs in this new strain, JLD1201, were confirmed as before. The sdiA1::mTn3 mutation was introduced into JLD1201 from BA612 using P22HTint, resulting in strain JLD1203.
Construction of mutations in sdiA-regulated genes. Lambda Red mutagenesis (16) was used to make in-frame deletions of srgA, srgB, srgC, srgD, srgE, pefC, pefI, and rck. An additional deletion of the entire rck operon (pefI-srgD-srgA-srgB-rck-srgC) was also constructed. Oligonucleotides containing 40 nucleotides with homology to the recombination target site, including the first or last 30 nucleotides of that particular target gene, were appended to sequences that bind the priming regions of pCLF3 (32). The primer sequences are listed in Table 2.
The primer pairs to make each mutation were as follows: for srgA, BA1605-BA1606; srgB, BA1607-BA1608; srgC, BA1611-BA1612; srgD, BA1633-BA1634; srgE, BA1581-BA1582; pefC, BA1583-BA1584; pefI, BA1604-BA1641; rck, BA1609-BA1610; ⌬pefI-srgC, BA1585-BA1586. After these primers are in a PCR with pCLF3 as the template, a product is generated that includes a FRT-cam-FRT cassette flanked by 40-nucleotide (nt) regions with DNA sequence identity to the gene to be disrupted. This PCR product was electroporated into strain 14028(pKD46), followed by selection for homologous recombination on LB Kan at 37°C (16). The presence of the insertion was confirmed by PCR. The resulting mutations were transduced into 14028 using phage P22HTint. The antibiotic resistance marker was then deleted by electroporating the temperature-sensitive plasmid pCP20, which encodes FLP recombinase, into the strain. The electroporated cells were plated on LB plus Amp at 30°C, and single colonies were streaked for isolation at 37°C. Colonies were then screened for loss of chloramphenicol resistance and loss of pCP20. Each knockout strain was then transduced with the P R -yenI-FRT-kan-FRT cassette and/or the sdiA::mTn3 mutation using phage P22HTInt. Each strain was confirmed to make AHLs by a cross-streak assay with the 14028(pJNS25) and BA612(pJNS25) reporter strains. Measurement of resolvase activity in mice. Female CBA/J and BALB/c mice (8 to 10 weeks old) were obtained from Jackson Laboratories and Harlan, respectively. Overnight cultures of the wild-type and sdiA mutant RIVET (recombination-based in vivo expression technology (7, 34)) strains (JNS3206 and JNS3226) were grown in LB Kan Tet at 37°C. The next morning, the overnight cultures were centrifuged at 10,000 ϫ g and resuspended in fresh LB lacking antibiotics. Mice were inoculated intragastrically with 200 l of a 1:1 mixture of JNS3206 and JNS3226 (approximately 10 9 total CFU). Dilution plating of the inoculum was used to determine the actual dose administered. In some experiments, the mice were inoculated 24 h earlier with Yersinia enterocolitica JB580v or the isogenic yenI::kan mutant GY4493 or mock infected with 200 l LB. The Yersinia cultures were centrifuged at 10,000 ϫ g and resuspended in fresh LB before inoculation. Dilution plating of the inoculum was used to determine the Fresh fecal samples were collected at various time points and resuspended in phosphate-buffered saline, followed by serial dilution and plating on XLD plus Kan (XLD Kan) agar. At various time points, animals were euthanized; organs were recovered, homogenized, and dilution plated on XLD Kan agar. Isolated colonies were then screened for Amp and Tet resistance. The Yersinia yenI::kan mutant also grew on these plates but was easily distinguished from the Salmonella bacteria by color; the Salmonella colonies are black, and the Yersinia colonies are yellow. All animal experiments were performed at Ohio State University with IACUC approved protocol no. 2006A0037.
Cross-streak assay. Cross-streak assays were performed previously as described (2,35). Yersinia bacteria were grown overnight at 28°C, and then 20 l was dripped down an LB plate. Yersinia plates were incubated at room temperature overnight. Salmonella strains were grown overnight at 37°C, and then the strains to be tested for AHL production were dripped on an LB plate and allowed to dry. Twelve microliters of each of the reporter strains, 14028(pJNS25) and BA612(pJNS25), was then dripped perpendicular to the test strain. Plates were incubated at 37°C for 7 h and then imaged using a C2400-32 intensified charge-coupled-device camera with an Argus 20 image processor (Hamamatsu Photonics).
Competitive index measurements. The competitive index equals the output ratio (CFU of mutant/CFU of wild type) divided by the input ratio (CFU of mutant/CFU of wild type). The log of the competitive index represents a normal distribution, so this value was calculated and plotted. A value of zero indicates that the wild type and the sdiA mutant had equal fitness during transit through the animal. A negative value indicates that the wild type was more fit than the mutant.

RESULTS
Previously we determined that the normal gastrointestinal microbiota of several animal species did not synthesize AHLs of the correct type or in sufficient quantities to activate SdiA of Salmonella (36). However, SdiA was activated during the transit of Salmonella through turtles that were colonized with the known AHL producer Aeromonas hydrophila (36). Since Aeromonas hydrophila is typically thought of as a pathogen, these results led us to hypothesize that the function of SdiA is to detect the AHL production of other bacterial pathogens in animals rather than the normal microbiota.
As a first step toward testing this hypothesis, we wanted to determine if Salmonella can detect the AHL production of another pathogen in a more typical animal model. Therefore, we tested the ability of Salmonella to detect Yersinia enterocolitica in mice. As in the previous study (36), we used the RIVET method, utilizing an isogenic pair of Salmonella RIVET strains: JNS3206 and JNS3226 (Table 1). JNS3206 is sdiA ϩ , while JNS3226 carries sdiA::mTn3. Both strains contain a chromosomal srgE-tnpR fusion and a res1-tetRA-res1 cassette. The srgE gene is regulated by sdiA. The tnpR gene encodes a site-specific recombinase or "resolvase" that catalyzes the permanent deletion of the res1-tetRA-res1 cassette, leaving just a single res1 site behind (hereafter referred to as "resolution"). Thus, loss of Tet resistance in the sdiA ϩ strain but not the sdiA::mTn3 strain indicates sdiA-dependent fusion activity. This pair of strains can be incubated together in any environment of interest, followed by recovery of the strains by dilution plating. The ratio of sdiA ϩ strain to sdiA mutant can be determined by screening individual colonies for Amp resistance (the sdiA::mTn3 mutation confers Amp resistance). This competition experiment determines whether the sdiA ϩ strain had a fitness advantage over the sdiA mutant during the incubation. The percentage in which the Tet resistance gene was deleted or "resolved" is determined by screening the colonies for Tet sensitivity. SdiA activity is indicated by resolution occurring more often in the sdiA ϩ strain than in the sdiA mutant strain.
Several preliminary experiments were performed using various Yersinia and Salmonella inoculum sizes and different mouse strains (CBA/J and BALB/c). CBA/J mice are slc11A1 ϩ/ϩ and resistant to systemic Salmonella infection (24). These mice provide a model of Salmonella intestinal persistence. BALB/c mice are slc11A1 Ϫ/Ϫ and susceptible to systemic Salmonella infection. The overall conclusion of the preliminary experiments was that in mice that have been preinfected with Yersinia, resolution occurs in the sdiA ϩ Salmonella strain but not the sdiA mutant strain (data not shown). However, there were advantages and disadvantages to each strain of mouse. With both mouse strains, the highest bacterial inoculums yielded the most consistent results (200 l of an overnight culture intragastrically, which is between 10 8 and 10 9 CFU). Lower inoculums yield the same overall conclusion, but the presence or absence of bacteria in each sample becomes sporadic. CBA/J mice were the best for taking fecal samples over long periods of time. The disadvantage of CBA/J mice was that obtaining Salmonella or Yersinia from intestinal tissues and systemic sites was sporadic and the numbers recovered were low. BALB/c mice provided highly consistent numbers of Salmonella and Yersinia CFU from all sites, and Salmonella resolution could be observed just 1 day postinfection. This is presumably because in BALB/c mice, the battle between host and pathogens is consistently being won by the pathogens, thus providing more-consistent interactions between Yersinia and Salmonella. The disadvantage of BALB/c for these studies is that the mice become ill very quickly after inoculation with either Yersinia or Salmonella and immediately stop defecating. They also succumb to the infections within a week, hindering the examination of long-term persistence phenotypes. Below, we describe a representative persistence experiment using CBA/J mice and a tissue distribution experiment using BALB/c mice.
SdiA is activated during transit of Salmonella through CBA/J mice preinfected with Yersinia enterocolitica. To determine if Salmonella can detect the AHL production of Yersinia in mice, we first used oral inoculation of Salmonella RIVET strains, followed by recovery of the strains from fecal pellets. Thirty mice were split into three groups. Ten mice were inoculated with wild-type Yersinia enterocolitica JB580v, ten mice were inoculated with an isogenic yenI::kan mutant of JB580v, and ten mice were mock infected with LB. The following day, all 30 mice were inoculated with a 1:1 mixture of the two isogenic Salmonella RIVET strains (sdiA ϩ strain and sdiA mutant). Fecal samples were then collected over time and dilution plated on XLD and XLD Kan to enumerate the Yersinia and Salmonella bacteria recovered. The two could be distinguished on XLD, since Yersinia colonies are yellow whereas Salmonella colonies are black. On the XLD Kan plates, only the Salmonella RIVET strains were recovered (except in the yenI mutant group, where we also recovered the Yersinia yenI mutant). No Yersinia or Salmonella was recovered from uninfected mice. Fecal samples were collected for 27 days; however, some mice became ill and were euthanized during the course of the ex- The results indicate that resolution occurred during the transit of Salmonella through mice infected with wild-type Y. enterocolitica but not during transit through mice infected with a yenI mutant of Y. enterocolitica or mock-infected with LB alone (Fig. 1). Resolution was seen in sdiA ϩ Salmonella colonies but not in sdiA mutant colonies, indicating that the activation of the srgE-tnpR fusion was sdiA dependent. The log competitive index of sdiA ϩ versus sdiA mutant Salmonella recovered from feces over time remained at zero, indicating that sdiA does not confer a fitness advantage upon Salmonella in Yersinia-infected CBA/J mice (Fig. 1F).

SdiA is activated in intestine and Peyer's patches of BALB/c mice preinfected with Yersinia enterocolitica.
To determine when and where Salmonella and Yersinia colocalize within the mouse and to determine the time and location of sdiA-dependent gene expression, we performed a RIVET experiment using BALB/c mice. This experiment was similar to the experiment described above (Fig. 1), except that instead of collecting fecal samples over time, four mice from each of the three groups (infected with wild-type Yersinia or the yenI mutant or mock infected with LB) were sacrificed on days one, three, and four postinfection (no mice survived beyond day four). The tissues recovered were from the small intestine, cecum, large intestine, mesentery, spleen, and between three and five individual Peyer's patches. Each Peyer's patch was treated as a separate sample. On the first day post-Salmonella infection, one of four mice from the Yersinia-infected group already contained Salmonella that had resolved. All mice dissected on days three and four from the Yersinia-infected group contained Salmonella that had resolved. Resolution was dependent upon the yenI gene of Yersinia and the sdiA gene of Salmonella (Fig.  2). There was also a correlation between resolution and the number of Yersinia CFU recovered from the sample, with resolution requiring approximately 10 6 Yersinia CFU (Fig. 3). A very interesting result is that resolution was found throughout the intestines but was highest in the Peyer's patches. There was almost no resolution in the mesentery and spleen. Since resolution is irreversible, this suggests that those Salmonella bacteria that resolve (i.e., those that have detected AHL) stay in the Peyer's patches (Fig. 2). However, there was no apparent fitness advantage for sdiA ϩ Salmonella. In fact, the sdiA ϩ Salmonella bacteria were at a slight but statistically significant disadvantage in the small intestine, large intestine, and Peyer's patches (Fig. 2F). The difference is such that if the Salmonella sdiA mutant was 50% of the inoculum, it made up 78% of the recovered bacteria.
Phenotypes are observed when Salmonella produces its own AHLs. There are at least two potential explanations for the lack of an sdiA fitness phenotype in mice infected with Yersinia. One explanation is that while SdiA is clearly activated by the AHL production of Yersinia in mice, the genes that SdiA regulates may not have a function in mice. Another explanation is that the SdiA regulon is in fact functional in mice, but the percentage of Salmonella cells interacting with Yersinia cells at any given time is too low for a phenotype to be detected among the larger pool of bacteria that are not interacting. To test both of these hypotheses, we constructed a Salmonella strain that named JLD1201. An sdiA::mTn3 mutation was transduced into JLD1201 to create strain JLD1203. From results of a crossstreak assay with sdiA biosensor strains, it is clear that both strains produce AHLs at levels similar to that for Yersinia (Fig. 4). FIG. 4. AHL biosensor assay. The AHL biosensor strains 14028 (pJNS25) and BA612(pJNS25) (sdiA ϩ and sdiA mutant, respectively) were each streaked in duplicate down an LB agar plate (sdiA ϩ strain on the left and sdiA mutant on the right of each plate). The test strain was streaked perpendicular to the biosensor strains, and light production was imaged after 7 h at 37°C. Light is pseudocolored, with blue being the weakest and red being the most intense. JLD1201 and JLD1203 were used in a competition experiment with CBA/J mice. The results are dramatic in that the sdiA ϩ strain immediately outcompetes the sdiA mutant strain (Fig. 5A). This fitness phenotype does not occur during growth in LB broth (data not shown). This experiment demonstrates that sdiA ϩ Salmonella has a competitive advantage over sdiA mutant Salmonella in the mouse environment in the presence of AHLs. Furthermore, this experiment demonstrates that at least one individual member(s) of the SdiA regulon must be functional in the mouse.
Determination of sdiA regulon members responsible for fitness phenotype. To determine which SdiA regulon members contribute to the yenI-dependent fitness phenotype observed in mice, in-frame deletions were constructed in every known member of the SdiA regulon in the JLD1201 and JLD1203 backgrounds and the competition experiments were repeated. Deletion of srgE eliminated the sdiA fitness phenotype (in a srgE mutant background, sdiA ϩ bacteria cannot outcompete sdiA mutant bacteria) (Fig. 5B). Similarly, deletion of the entire rck operon (Fig. 5C) or any individual gene of the rck operon ( Fig. 5E to J) eliminated the sdiA fitness phenotype. Because two of the genes in the rck operon, pefI and srgA, are thought to play a role in regulation of the pef operon or folding of the PefA fimbrial subunit, respectively, we tested the hypothesis that the pef operon may be required for the sdiA fitness phenotype. However, deletion of pefC had no effect, FIG. 5. Members of the sdiA regulon required for fitness in mice. (A) In a yenI ϩ Salmonella genetic background, an sdiA mutant is less fit in CBA/J mice than an sdiA ϩ strain. Five mice were inoculated with approximately 1 ϫ 10 9 CFU of a 1:1 mixture of JLD1201 (yenI ϩ sdiA ϩ ) and JLD1203 (yenI ϩ sdiA::mTn3). The strains were recovered from fecal pellets over time, and the log competitive index (logCI) was plotted. The experiment was repeated as for panel A except using pairs of strains in which the either the srgE (B), pefI-srgC (rck operon) (C), pefC (D), pefI (E), srgD (F), srgA (G), srgB (H), rck (I), or srgC (J) gene had been deleted. Error bars represent 95% confidence intervals; ‫"ء"‬ indicates a P value of Ͻ0.05, and ‫"ءء"‬ indicates a P value of Ͻ0.01. indicating that the fitness phenotype conferred by the rck operon is independent of the pef operon (Fig. 5D).

DISCUSSION
Previously SdiA activation was not observed in a guinea pig, a rabbit, a cow, 5 mice, 6 pigs, or 12 chickens (36). However, SdiA was activated in turtles that carried Aeromonas hydrophila (36). In this study, we observed SdiA activation in mice infected with Y. enterocolitica (Fig. 1 to 3). SdiA was not activated in mice infected with a yenI mutant of Yersinia even though the numbers of yenI mutant bacteria recovered from most tissues at most time points were similar to the numbers of yenI ϩ Yersinia bacteria that were recovered ( Fig. 1 and 2). The only exception was that yenI mutant Yersinia bacteria were recovered at roughly 10-fold-lower numbers than yenI ϩ bacteria for the first 7 days of CBA/J infection (Fig. 1). After that, the numbers of bacteria recovered were similar through 28 days. This is the first observation of any virulence defect for a yenI mutant of Y. enterocolitica. Also of note is that S. Typhimurium was able to colonize BALB/c mice in higher numbers, especially in the small intestine and Peyer's patches, when the mice were infected with Yersinia (Fig. 2). However, this did not require sdiA of S. Typhimurium or yenI of Y. enterocolitica.
The current model of S. Typhimurium pathogenesis states that the majority of bacteria enter the mesenteric lymph nodes (MLN) and spleen via the Peyer's patches (38). A smaller proportion may invade the intestinal epithelial cells or be taken up by CD18 ϩ macrophages (41). In either case, we would expect the percentage of resolved bacteria in the MLN and spleen (those that have detected AHL and become Tet sensitive) to equal or exceed the percentage that was present in their starting location. Instead, we observe almost no resolution in MLN and the spleen despite the fact that high numbers of both sdiA ϩ and sdiA mutant bacteria were recovered from those sites (Fig. 2). Therefore, we hypothesize that those bacteria in which SdiA becomes activated stay in the Peyer's patch while the sdiA mutant bacteria or the sdiA ϩ bacteria that are not activated continue to the MLN and spleen.
Although SdiA was activated, no fitness advantage was seen for sdiA ϩ S. Typhimurium compared to the sdiA mutant in Yersinia-infected mice ( Fig. 1 and 2). However, when S. Typhimurium was engineered to produce AHLs using the Y. enterocolitica yenI gene, the sdiA ϩ bacteria gained a large and immediate fitness advantage over the sdiA mutant (Fig. 5). Furthermore, every member of the sdiA regulon was required for this fitness phenotype (Fig. 5). This indicates that each member of the S. Typhimurium sdiA regulon is functional in mice. However, this begs the question of why the sdiA fitness advantage is not observed in the normal genetic background that lacks yenI during infection of mice that are infected with Yersinia. The resolution data in Fig. 2 suggests that at least 30% to 40% of the Salmonella cells in Peyer's patches detected AHL at some point during the infection. Given that the use of tnpR fusions permanently records even a transient activation, it is possible that the interaction time was sufficient to cause resolution of the Tet resistance marker but was not sufficient to cause sdiA ϩ S. Typhimurium to predominate over the sdiA mutant. When the Salmonella were engineered to synthesize AHLs, the interaction time was maximized, allowing the sdiA ϩ strain enough time to predominate over the sdiA mutant.
The mechanism by which sdiA ϩ bacteria are more fit than sdiA mutant bacteria is not known, but it is interesting that all members of the sdiA regulon are required for this fitness phenotype. Y. enterocolitica is presumably present at a high population density within the Peyer's patches, resulting in high concentrations of AHL. It is not clear if this high population density is confined to certain locations in the Peyer's patches or if the entire Peyer's patch can be considered to contain a high population density of Yersinia. Likewise, it is not clear if the AHL concentration is high throughout the Peyer's patch or only in small localized areas. This question is posed because many sdiA ϩ S. Typhimurium bacteria appear to move from the Peyer's patches to the MLN and spleen in the unactivated state. In any case, at least part of the Peyer's patch contains sufficiently high concentrations of Y. enterocolitica and AHL to activate SdiA. The genes required for the sdiA fitness phenotype include srgE and all of the rck operon. The rck gene is known to confer resistance to complement, although this function is redundant in Salmonella (19). Rck may also provide adhesion to fibronectin or laminin or have an as yet unidentified function (15). SrgE is predicted to be a type III secreted effector (referred to as STM1554 in reference 31), but it is not clear why S. Typhimurium would add one more type III effector to its arsenal in the presence of Y. enterocolitica. Further research into the function of SrgE is required. The rck operon encodes two transcription factors, srgC and srgD, that are required for the fitness phenotype. The regulon of these transcription factors needs to be determined. The rck operon also includes pefI and srgA, which are thought to play a role in regulation of the pef operon and folding of the PefA fimbrial subunit, respectively (6,30). The observation that pefI and srgA are required for the sdiA fitness phenotype while pefC is not indicates that pefI and srgA must have additional functions that are independent of the pef operon.