ChIP-seq and transcriptome analysis of the OmpR regulon of Salmonella enterica serovars Typhi and Typhimurium reveals accessory genes implicated in host colonization

Summary OmpR is a multifunctional DNA binding regulator with orthologues in many enteric bacteria that exhibits classical regulator activity as well as nucleoid-associated protein-like characteristics. In the enteric pathogen Salmonella enterica, using chromatin immunoprecipitation of OmpR:FLAG and nucleotide sequencing, 43 putative OmpR binding sites were identified in S. enterica serovar Typhi, 22 of which were associated with OmpR-regulated genes. Mutation of a sequence motif (TGTWACAW) that was associated with the putative OmpR binding sites abrogated binding of OmpR:6×His to the tviA upstream region. A core set of 31 orthologous genes were found to exhibit OmpR-dependent expression in both S. Typhi and S. Typhimurium. S. Typhimurium-encoded orthologues of two divergently transcribed OmpR-regulated operons (SL1068–71 and SL1066–67) had a putative OmpR binding site in the inter-operon region in S. Typhi, and were characterized using in vitro and in vivo assays. These operons are widely distributed within S. enterica but absent from the closely related Escherichia coli. SL1066 and SL1067 were required for growth on N-acetylmuramic acid as a sole carbon source. SL1068–71 exhibited sequence similarity to sialic acid uptake systems and contributed to colonization of the ileum and caecum in the streptomycin-pretreated mouse model of colitis.


Introduction
OmpR is a DNA binding protein that, with the cognate sensor EnvZ, co-ordinates transcriptional response to environmental factors including osmotic stress in many enteric bacteria (Forst and Roberts, 1994). OmpR/EnvZ are central to the adaptive response to the intestinal environment (Giraud et al., 2008), in part because of the distinct osmolyte composition of the lumen. As many as 125 genes in Escherichia coli (Oshima et al., 2002) and 305 genes in Salmonella Typhi (Perkins et al., 2009) have been implicated in OmpR/EnvZ-dependent expression. The OmpR regulon includes genes from the 'ancestral' core genome shared with many enteric bacteria as well as genes of the accessory genome. The latter include virulence-associated loci such as the viaB locus that encodes Vi polysaccharide biosynthesis genes, and genes encoded on Salmonella pathogenicity island 2 (SPI-2) via its regulation of ssrAB (Pickard et al., 1994;Feng et al., 2003;Perkins et al., 2009). OmpR-regulated orthologues in diverse enteric bacteria define the ancestral regulon and include porin genes such as ompF and ompC. However, the OmpR regulon exhibits considerable plasticity and can include genes of the ancillary genome acquired by horizontal gene transfer, many of which are involved in hostpathogen interactions. The acquisition of such genes and the ability to express them appropriately on moving from the intestinal lumen to the intracellular compartment were likely key features in the evolution of Salmonella (Bäumler, 1997;Groisman and Ochman, 1997).
The genus Salmonella consists of more than 2500 serotypes that exhibit diverse host range and pathogenicity Popoff et al., 2004). Most of the > 2500 serovars of Salmonella enterica have a relatively broad host range and are typically associated with gastroenteritis in human (Santos et al., 2001). In contrast, S. enterica serovar Typhi (S. Typhi) is highly host-adapted to cause the systemic disease typhoid specifically in human. S. Typhi can invade the intestinal mucosa but colonization of the intestine is relatively transient and rapid systemic dissemination can follow leading to typhoid. This distinct pathogenesis is driven at least in part by horizontally acquired genes that are required for virulence, including the OmpR-regulated viaB, encoding the Vi polysaccharide antigen (Pickard et al., 1994). The OmpR regulon includes both Salmonella pathogenicity island 1 (SPI-1) and SPI-2, mediated through ssrAB expression. The integration of such horizontally acquired genes into existing regulons is a recurring theme in the evolution of pathogenesis. The mechanism by which OmpR regulates gene expression is not fully understood. It has been proposed that OmpR has only weak specificity for DNA binding (Head et al., 1998;Rhee et al., 2008) and that it may have both a classical site-specific impact on gene expression through recruitment of RNA polymerase and additional nucleoid-associated protein (NAP)-like properties that may also impact global gene expression (Cameron and Dorman, 2012).
In this study we combine RNA-seq and ChIP-seq together with in vitro and in vivo phenotyping to define the interaction of OmpR with the chromosome and characterize two novel OmpR-regulated operons that are part of the S. enterica ancillary genome.

Identification of candidate S. Typhi genes regulated by OmpR using ChIP-seq
We recently characterized the OmpR regulon of S. Typhi BRD948 using DNA microarray and RNA-seq (Perkins et al., 2009) (Table S1) identifying 208 genes by microarray and 305 genes by RNA-seq, that exhibited OmpRdependent transcription during mid-log phase culture in rich media. In order to further characterize the OmpR regulon we used ChIP-seq to identify candidate genome regions that are preferentially associated with the OmpR protein in the S. Typhi genome. To this end a S. Typhi BRD948 derivative TT53.8 was constructed in which the 3′ end of ompR harboured an in-frame fusion with sequence encoding three repeats of the FLAG epitope (3¥FLAG tag). TT53.8 expressed the fusion protein (OmpR::3¥FLAG) in place of the wild-type OmpR protein from the native chromosomal location at single copy. To assess if this fusion protein had comparable function to wild-type OmpR, we indirectly monitored the expression of the ompRdependent viaB locus in TT53 (Pickard et al., 1994). Agglutination of S. Typhi TT53 with anti-Vi antiserum was indistinguishable to that of S. Typhi BRD948 in low-salt and high-salt culture media (data not shown).
Salmonella Typhi TT53.8 or BRD948 were grown to mid-log phase (OD 600 = 0.6) and ChIP-seq was performed on DNA precipitated by anti-FLAG antibody. The normalized sequence depth at each base of the reference genome sequence was plotted as the number of standard deviations from the mean (z-score) to identify regions of significantly enriched sequence coverage (z-score > 3) and 43 ChIP-enriched peaks that were within intergenic regions were studied further (15 lay within annotated coding sequence and were excluded from further analysis) (Fig. 1, Table S2).
Twenty-two of the genes with a sequence enrichment peak in their upstream region also exhibited OmpRdependent expression as determined by RNA-seq or microarray analysis (Perkins et al., 2009) (Fig. 1, Table 1). These included many previously identified OmpRregulated genes such as tviA and ompS1 (Fig. 2). Some genes associated with aerobic lifestyle such as citrate synthase (gltA) and succinate dehydrogenase C (sdhC) also contained enrichment peaks. A ChIP peak was identified in the intergenic region of the divergently transcribed operons encoding stdA and dppA. Another within the intragenic region of two divergently transcribed putative operons encoding genes t1787-1790 and t1791-93 (Table 2). Surprisingly, statistically significant peaks with a z-score > 3 were not identified in the well-characterized OmpR-regulated genes ompF and ompC, although a minor peak that fell just short of the statistical cut-off, mapped extensively to motifs (C1-3) implicated in OmpR binding (Fig. 2C). To determine if the C-terminal FLAG tag of OmpR impacted binding to the ompC or ompF promoter region we compared expression of these genes in the wild-type (BRD948) and ompR::FLAG strains (TT53.8) (Fig. S1). Expression of tviB and ompF was similar in these two strains but ompC was expressed at a significantly lower level in TT53.8. The degree to which ompC expression was decreased in TT53.8 compared to wild-type BRD948 was not as pronounced as that in a strain in which ompR was deleted (TT10) suggesting that some OmpR activity for the ompC promoter was retained in the epitopetagged protein.

Identification of nucleotide sequence motifs associated with OmpR binding
To identify sequence motifs within the 43 intergenic ChIPenriched sequence coverage peaks that may be involved in OmpR binding, nucleotide sequences were compared using the YMF algorithm (Sinha and Tompa, 2003), which identifies candidate binding sites by searching for statistically overrepresented motifs. Five eight-nucleotide motifs (z-score > 9.8, Table S2) were identified in 14 separate loci, with some loci containing multiple motifs. The motif TGTWACAW occurred 21 times, in 12 ChIP-enriched sequences including the viaB locus (5′ tviA) where it precisely coincided with the peak of sequence enrichment Concentric circular tracks indicate: (a) plot of the sequence coverage expressed as the number of standard deviations from the mean (z-score) with values > 0 (red) or < 0 (blue); (b) differentially regulated genes identified by microarray (Perkins et al., 2009); (c) significant ChIP-seq peaks defined as spanning Ն 36 bp and with sequence read depth > 3 standard deviations (z-score) above the mean; (d) S. Typhi genes with colour coding indicating function: dark blue, pathogenicity/adaptation; black, energy metabolism; red, information transfer; dark green, membranes/surface structures; cyan, degradation of macromolecules; purple, degradation of small molecules; yellow, central/intermediary metabolism; light blue, regulators; pink, phage/IS elements; orange, conserved hypothetical; pale green, unknown function; brown, pseudogenes; (e) genes exhibiting OmpR-dependent expression that have a ChIP-seq peak in 5′ UTR region; (f) ChIP-seq peaks associated with differentially regulated genes; and (g) S. Typhi genes that exhibited OmpR-dependent expression in both microarray and RNA-seq experiments (Perkins et al., 2009 (Perkins et al., 2009). b. Normalized ratio (WT : DompR) of microarray (n = 3) fluorescent intensities from three biological replicates hybridized on four microarrays (Perkins et al., 2009). NS, not significantly different.
( Fig. 2). The maximum z-score of peaks associated with this motif was significantly greater (P = 0.0013, Student's t-test) than that of peaks without an identifiable motif (P = 0.0013, Student's t-test) (Fig. S2). The TGTWACAW motif was present in two copies in the upstream region of seven genes: ompS1, csgD, sdhC, galP, dppA, pckA and t4357. t4357 encodes a putative integrase encoded on a prophage~3.6 kbp upstream of tviA. A second motif AYG-GCCTA was present in single copy in the upstream region of four loci: t0528, t1320, dppA and tviA. There was also a significant difference in the magnitude of ChIP sequence peaks containing motif AYGGCCTA compared with those with no identifiable motif (Student's t-test, P < 0.0001) (Fig. S2), suggesting a link between this sequence and the avidity of OmpR binding. None of the sequence motifs are present in the previously described C1-3 or F1-4 OmpR binding sites in ompC and ompF promoter regions respectively. The largest enrichment peak (z-score = 16.23) was found upstream of tviA of the viaB locus, encoding four different candidate motifs: TGTWACAW, CTAGACTA, AYGGCCTA and AACTAACW (Table S2a). To find if the TGTWACAW motif was involved in binding of OmpR to the tviA upstream region, we used an electrophoresis mobility shift assay (EMSA) with phosphorylated recombinant OmpR::6¥His protein and oligonucleotide probes. The probes comprised either tviA -133 to -460 or tviA -303 to -377 of the tviA upstream region. Arbitrary mutation of the TGTTACAA motif at the -341 to -348 position to GCTCG-GAC resulted in abrogation of OmpR::6¥His binding to either probe (Fig. 3). No binding of OmpR::6¥His was observed with a probe containing the mutant motif sequence, suggesting that this motif is important for OmpR binding. Significantly, the TGTTACAA motif in the tviA upstream region also coincides with the genome sequence most highly overrepresented following ChIP enrichment (z-score = 30, Fig. 2).
The OmpR regulons of S. Typhi and S. Typhimurium contain a core set of shared orthologous genes We next compared the previously unreported OmpR regulon of the broad host range S. Typhimurium strain SL1344 with that of the human restricted S. Typhi Ty2 (Perkins et al., 2009) to identify previously uncharacterized genes, controlled by OmpR in both pathogens. A total of 208 genes and 329 genes were expressed in an OmpRdependent manner in S. Typhi and S. Typhimurium respectively. Of these, 31 orthologous genes were expressed in an OmpR-dependent manner in both serotypes (Table S3). OmpR-dependent expression levels of genes that were OmpR-regulated in both S. Typhi and S. Typhimurium showed a high degree of correlation ( Fig. 4; R 2 = 0.73) indicating strongly conserved regulation between the two serovars in this cohort of genes. OmpRdependent genes found in both serotypes included ompS1, ompC, sprB and ompR, all of which showed decreased expression in the absence of OmpR. A number of genes were upregulated in the absence of OmpR, including the succinate dehydrogenase genes sdhCDA (Cunningham and Guest, 1998), fatty acid dehydrogenase genes fadABI (Campbell et al., 2003), narK (Rowe et al., 1994), required for nitrite extrusion, and the nitrite reductase gene nrfA (Clarke et al., 2008).
Potentially important differences in the expression of SPI-1, SPI-2 and flagellin secretion apparatus in S. Typhi and S. Typhimurium were also revealed by the transcriptomic data in S. Typhimurium; 28 SPI-1-associated genes exhibited decreased expression in the absence of OmpR, including the sprB gene. Furthermore, several genes associated with the flagella type III secretion system (fliGHJLMOPR) and 10 apparatus genes encoded on SPI-2 (ssaA, ssaB, ssaGHIJKLT and STM1410) also exhibited up to fivefold greater expression in the absence of a functional OmpR (Table S4). In contrast, the only known SPI-1 gene that was OmpR-dependent in S. Typhi was sprB, and this encodes a transcriptional regulator (Golubeva et al., 2012). However, it is important to note that the culture conditions employed in our studies are known to result in low expression of SPI-1 and SPI-2 genes and therefore the biological impact of the observed differences in OmpR regulation in these conditions is not clear.

t1787-t1790 and t1791-1793 encode putative nutrient-scavenging systems
We next characterized the function of t1787-t1790 and t1791-1793 using a number of in vitro and in vivo assays. In S. Typhimurium SL1344, genes SL1071-SL1068 (STM1133-STM1130) and SL1067-SL1066 (STM1129-STM1128) are orthologues of the S. Typhi t1787-t1790 and t1791-1793 genes respectively (Fig. 5). However, t1792 and t1793 of S. Typhi are present as a single open reading frame (SL1066) in S. Typhimurium, suggesting that these genes may represent fragments of a pseudogene in S. Typhi.
We considered that these operons may be involved in host-pathogen interactions since they were absent from the closely related species E. coli (strain K12) (Blattner et al., 1997) but were highly conserved within S. enterica serotypes exhibiting > 99% identity at the amino acid level in these pathogens (Fig. 5). The transcriptomic data showed that expression of at least two of the S. Typhimurium orthologues was OmpR-dependent, namely SL1068 (4.75-fold increase, P < 0.05, orthologue of t1791) and SL1069 (9.59-fold increase, P < 0.05, orthologue of t1789, Fig. 4). Proteins encoded by several of the genes in these operons exhibit similarity to proteins that have previously been implicated in sialic acid uptake and metabolism. An orthologue of SL1066 encoded by S. Typhimurium LT2 (STM1128) is a sodium solute symporter (SSS) family transporter of sialic acid and shares 44-48% identity at the amino acid level with similar transport systems in Lactobacillus spp. and Staphylococcus spp. (Severi et al., 2010). The STM1128 gene complemented a mutant E. coli lacking the sialic acid transporter NanT for growth on sialic acid as the sole source of carbon (Severi et al., 2010). SL1067 is a homologue of nanE (SL3309) that is encoded elsewhere on the SL1344 chromosome, sharing 69% identity at the amino acid level. SL1068 to SL1071 have previously been proposed to be orthologues of genes encoded by E. coli K12 (nanM, nanC, yjhB and yjhC respectively), some of which have been implicated in sialic acid metabolism (Severi et al., 2008). However, the genomic context for these genes is quite different in E. coli compared to S. Typhimurium and amino acid sequence identity ranges from just 22% for NanC to 62% for YhjC, considerably less than observed for orthologous proteins of E. coli and S. Typhimurium of approximately 90% (Parkhill et al., 2001). We therefore tested the hypothesis that genes within these operons were involved in utilization of acetylated amino sugars including sialic acid as a sole carbon source, and colonization of the murine host.  To this end, we determined the ability of S. Typhimurium SL1344 and the isogenic mutant derivatives RAK103 (DSL1067/SL1066) and RAK105 (DSL1068-SL1071) to grow on M9 minimal media supplemented with one of several alternative acetylated amino sugars as the sole source of carbon. RAK103 (DSL1067/SL1066) was deficient in growth on N-acetylmuramic acid, a component of peptidoglycan, but grew normally on all other sole carbon sources tested (Table 3). A strain (SW738) in which SL1067/SL1066 were reintroduced onto the chromosome of strain RAK103 (DSL1067/SL1066) by phage-mediated transduction grew normally on N-acetylmuramic acid as a sole carbon source. No defect in growth on any of the carbon sources tested was observed for RAK105 including N-acetylneuraminic acid, a common sialic acid, possibly because this strain encodes a second sialic transport protein, NanT.
Experimental infections of mice with S. Typhimurium are a common surrogate model for typhoid fever (Tsolis et al., 1999). To determine if these OmpR-regulated genes contribute to the pathogenesis of S. Typhimurium in mice we carried out competitive infection experiments between the fully virulent S. Typhimurium RAK113 and mutant derivatives harbouring deletions within each operon (RAK105 DSL1068-SL1071, orthologues of t1787-t1790 and RAK103 DSL1067/SL1066, orthologues of t1791-3) (Table 4). Initially, groups of C57BL/6 mice were inoculated orally with 1 ¥ 10 8 cfu of an equal mixture of RAK103 or RAK105 and RAK113. No significant difference in the ratio of each derivative was observed in the caecum, ileum, Peyers patch, spleen or liver (Table 4).
Since S. Typhimurium normally causes inflammatory diarrhoea manifesting as gastroenteritis we determined if these genes are required for colonization and growth during a robust inflammatory response in the intestinal tract. To this end we performed mixed inoculum experi- The surface of the agar was inoculated with each strain cultured on LB agar and incubated for 48 h at 37°C. Growth was assessed and recorded as no growth (-), growth retarded relative to that with glucose as a sole carbon source (+) and comparable growth to that with glucose as a sole carbon source (++).  SE, standard error. The data were calculated from the combined data from two independent experiments each of which individually had comparable outcomes. ments in the streptomycin-pretreated mouse model of colitis (Hapfelmeier et al., 2004;Hapfelmeier and Hardt, 2005). Groups of streptomycin-pretreated mice were inoculated with 1 ¥ 10 3 cfu of an equal mixture of S. Typhimurium RAK103 or RAK105 and RAK113. Four days post inoculation RAK103 was present in similar proportion to RAK113 (Table 4). However, RAK113 was present in approximately threefold greater numbers in the caecum of mice compared with RAK105 (Table 4). This decrease in fitness specifically in the inflamed gut was statistically significant, and reproducible. Furthermore, when the SL1067/SL1066 genes were reintroduced into RAK105 by phage-mediated transduction giving rise to strain SW771, the ability to compete successfully with the wild-type RAK113 in colonization of the caecum in streptomycinpretreated mice was restored (Table 4).

Discussion
Transcriptional regulons have been defined using DNA microarrays and more recently by RNA-seq approaches.
Observed changes in transcript abundance can be directly or indirectly related to a regulator protein binding either within an operator or at a secondary regulatory site.
We have combined measurement of transcript abundance with a direct assay of OmpR binding using ChIP-seq to gain a more complete understanding of the regulon and identify novel genes within this network. Using this approach genome sequences that were enriched included many previously described OmpR-regulated genes (Fig. 1). The most highly enriched sequences were upstream of the viaB locus (tviA) (Fig. 2A). Furthermore, there was considerable enrichment in the 5′ UTR of ompS1 (Fig. 2B), ompR and between the divergently transcribed csg operons. These observations provided proof of principle that the ChIP-seq approach identified wellknown OmpR-regulated genes. Perhaps surprisingly, substantial enrichment peaks were not observed in the 5′ UTR of the ompC and ompF genes, even though these are known to be regulated by OmpR (Rhee et al., 2008). A minor peak that did map to previously identified OmpR binding sites (C1-3) was present, but fell below the criteria used for peak identification. The reason for the lack of enrichment peaks associated with the ompC and ompF genes is not known, but may be related to the specific culture conditions used in this study resulting in incomplete phosphorylation of OmpR or due to interference from the C-terminal FLAG epitope tag. The presence of a C-terminal FLAG epitope had little impact on expression of the tviB and ompF genes but the ompC gene exhibited decreased expression, suggesting that the epitope may impact binding sites differently. Therefore, it is possible that all OmpR binding sites were not identified in this study.
Specific binding of OmpR is thought to depend at least in part on short nucleotide sequence motifs in the 5′ UTR region of genes within the regulon (Huang et al., 1994;Harlocker et al., 1995;Rhee et al., 2008), although the dependence on specific sequence is markedly less pronounced than for another two-component regulator of Salmonella, PhoP (Harari et al., 2010). Specific recognition of these motifs by OmpR is dependent on the phosphorylation state of the regulator and subsequent positive regulation of transcription results from direct interaction with RNA polymerase. A number of motifs have been proposed based on sequence similarity in the 5′ UTR of the ompC and ompF genes of E. coli, and DNAase footprint analysis. However, the lack of specificity for OmpR binding to these motifs is shown by the absence from the upstream sequence of other OmpRregulated genes. We used the YMF algorithm to identify sequences that were statistically overrepresented within enriched sequences following immunoprecipitation. While no such motifs were identified in 28 of 43 enrichment peaks using this approach, the motif (TGTWACAW) was present in 12 enrichment peaks and appeared in multiple copies (two or three copies) in seven of these regions. The motif TGTTACAA was present precisely at the point of greatest ChIP enrichment in the tviA upstream region determined by sequencing. Furthermore, this sequence was critical for binding of recombinant OmpR-6¥His in vitro using an EMSA approach. A total of four additional motifs were also identified and generally where these were present they were in the 5′ UTR of genes that also contained the common motif TGTWACAW. This suggested there may be a functional link between these sequences.
A total of 31 orthologous pairs of genes showed OmpRdependent expression in both S. Typhi and S. Typhimurium. Many more genes were OmpR-dependent in either S. Typhi or S. Typhimurium. The reason for this distinction is not clear but may be related to differences in the phosphorylation state of EnvZ, the OmpR cognate sensor kinase, that has been reported between these two serotypes (Oropeza and Calva, 2009). OmpR has pleiotropic effects on the homeostasis of the bacterial cell and these may manifest differently in Typhi and Typhimurium due to the overall differences in genotype.
Facultative anaerobic bacteria such as E. coli and Salmonella are thought to occupy a niche in the mucus layer close to the intestinal epithelium. Here they scavenge monosaccharides produced from the hydrolysis of complex polysaccharides and dietary fibre by anaerobic bacterial members of the microbiota (Chang et al., 2004). However, pathogenic bacteria such as Salmonella (Stecher et al., 2007) can induce a strong inflammatory response that results in a decrease in the population of many components of the microbiota that not only alters ChIP-seq analysis of OmpR regulon 533 the available nutrients (Stecher et al., 2008) but also available respiratory electron acceptors (Winter et al., 2010). Two divergently transcribed operons that were differentially expressed on inactivation of the ompR gene and contained a candidate OmpR binding site were predicted to be involved in scavenging and transport of alternative carbon sources. The predicted product of t1787-t1790 (SL1071-SL1068) had sequence similarity to sialic acid transport and metabolism systems. However, genetic deletion of SL1071-SL1068 did not impact on the utilization of N-acetylneuraminic acid (sialic acid) as a sole carbon source under the conditions tested, probably due to the presence of other sialic acid transport system, such as NanA/NanT in S. Typhimurium (Plumbridge and Vimr, 1999). The proteins encoded by t1791-3 (SL1067-1066) are also predicted to be involved in sialic acid metabolism. SL1067 exhibited homology with NanE, an N-acetylmannosamine-6-phosphate epimerase, and SL1066 orthologue has been reported to complement a nanT mutant of E. coli for growth in sialic acid as a sole source of carbon (Severi et al., 2010). However, deletion of these genes did not detectably impact utilization of sialic acid in vitro, presumably because of the presence of nanT and nanE. However, deletion of these genes resulted in the inability to use a related acetylated carbon compound, N-acetylmuramic acid, as a sole source of carbon during culture in vitro. Furthermore, although SL1068-SL1071 were not obviously required for colonization of the murine host in conventional mixed inoculum experiments, in the colitis model there was a reproducible and statistically significant decrease in the ability of RAK105 DSL1068-SL1071 to colonize the caecum of streptomycin-pretreated mice in competition with the wildtype parent. S. Typhimurium RAK103 was indistinguishable from the RAK105 SL1067-SL1066 locus in the ability to colonize the murine host.
Sialic acid has several potential impacts on hostpathogen interactions. It can be utilized as a carbon or nitrogen source, and is used by Haemophilus influenzae to modify LPS in order to evade detection by the host immune system (Severi et al., 2005;, although this has not been reported in enteric pathogens to date. Nutrient content of the intestine is impacted by the microbial community because of the complex interplay in catabolism of complex nutrients in the luminal contents (Bertin et al., 2012). However, it is likely that nutrient availability is altered as a result of the inflammatory response induced by Salmonella during infection, concomitant with disturbance of the normal microbiota (Stecher et al., 2007). Indeed, it was recently reported that Salmonella can use hostderived ethanolamine as a carbon source and respiratory electron acceptor following the switch to anaerobic respiration in the inflamed intestine (Thiennimitr et al., 2011). Our findings suggest that additional OmpR-regulated genes may contribute to nutrient scavenging in the inflamed intestine.

Experimental procedures
Bacterial culture and strains Salmonella Typhi was cultured routinely in LB broth with aromatic amino acids and pABA supplements as described previously (Lowe et al., 1999). Growth media were supplemented with antibiotics as appropriate at final a concentration of 0.05 mg l -1 kanamycin or 0.03 mg l -1 chloramphenicol. A strain in which the ompR gene is replaced by the aph gene encoding kanamycin resistance has been described previously (Kingsley et al., 2003). To construct a chromosomally encoding ompR::3¥FLAG, overlap extension PCR was employed to create a sequence encoding an in-frame 3¥FLAG peptide at the C-terminus of the ompR gene. This was complicated by the overlapping start-stop codon of the ompB locus (Parkhill et al., 2001). The Shine-Dalgarno sequence of the envZ gene, predicted to be encoded in the ompR ORF, was encoded downstream of the stop codon after the 3¥FLAG sequence. This sequence was cloned into the suicide vector pWT12 and the strain TT53 made by allelic exchange (Turner et al., 2006). Primers used were as follows: CGTCAGGCAAACGAACTGCC, 5′ to 3′ ompR bases (364.383), CCGTCATGGTCTTTGTAGTCTGCTTTA GAACCGTCCGGTA (full reverse primer sequence 5′ to 3′), GACTACAAAGACCATGACGGTGATTATAAAGATCATGATA TCGATTACAAGGATGACGATGACAAGTAGGTACCGGACG GTTCTAAAGC [concatenated primers are 5′ to 3′ forward (1:69 FLAG + 1:20)], CGAAACGCAGGCGGCACG [reverse for envZ is 5′ to 3′ (213:230)]. A strain designated RAK105 in which the SL1068-SL1071 genes of SL1344 were replaced by the aph gene was constructed using oligonucleotides 5′ accataagatcactaatgatgaagctttactccaattgtatttcttcgcTGTG TAGGCTGGAGCTGCTTC 3′ and 5′ cataagcgcagcgccaccg gccaataacaccaccatccggctttcaattCATATGAATATCCTCCTTAG 3′ to amplify with the pKD4 plasmid template. A strain designated RAK103 in which the SL1067-SL1066 genes of SL1344 were replaced by the aph gene was constructed using oligonucleotides 5′ cgcgttggcgtcaccgtatgctgtgtcg-gtatagcgtggtatcatgaaaTGTGTAGGCTGGAGCTGCTTCG 3′ and 5′ agacataacataaaacggagcaaaacttcaaatatataaggcgga actggCATATGAATATCCTCCTTAG 3′ to amplify with the pKD4 plasmid template. In all cases the mutation was retransduced into S. Typhimurium SL1344 using bacteriophage P22 in order to decrease the chances of the accumulation unlinked mutations during the passaging of bacteria during mutation construction. Strains (SW738 and SW771) in which the wild-type copy of genes SL1066 and SL1067 or SL1068-1071 was replaced in strains RAK103 (DSL1066-1067::aph) and RAK105 (DSL1068-1071::aph), respectively, were constructed using phage-mediated transduction. In order to select for transductants in this region a cat gene was introduced in the intergenic region of SL1071 and SL1072 using oligonucleotide primers 5′ cgcaaagtaaaactcactgaaat-tcttggctaaaattgaaagccgGTGTAGGCTGGAGCTGCTTCG 3′ and 5′ ccggtctacataagcgcagcgccaccggccaataacaccaccatc CATATGAATATCCTCCTTAG 3′. The cat gene was then introduced into S. Typhimurium strain RAK105 by P22 transduc-tion and chloramphenicol-resistant transductants selected on LB + Cm culture medium. Transductants that were resistant to chloramphenicol but sensitive to kanamycin were identified by replica plating on culture media containing the appropriate antibiotics. One such transductant was designated SW771 and the replacement of the aph gene with the wild-type SL1068-1071 confirmed by PCR amplification.

Expression analyses using microarray data and RNA-seq
Bacterial strains were cultured to OD600 = 0.6 and immediately fixed with RNAprotect (Qiagen) and harvested. The pellet was dried and RNA isolated using SV RNA isolation kit (Promega) according to manufacturer's instructions; however, elutions were performed using DEPC-treated water (Ambion). Dye incorporation, microarray design and analysis were performed as described previously (Kelly et al., 2004). RNA-seq data were described previously (Perkins et al., 2009). For S. Typhimurium microarrays, strain SL1344 and variants were cultured shaking at 250 r.p.m. in a New Brunswick Innova 3100 water bath at 37°C in 25 ml of fresh LB medium inoculated with a 1:100 dilution from an overnight bacterial culture. Three biological replicates were performed for each strain, and RNA was extracted at an optical density at 600 nm of 0.6 (mid-exponential phase). RNA was extracted using Promega's SV 96 total RNA purification kit. RNA quality was assessed on an Agilent 2100 Bioanalyser. Transcriptomic analyses were performed on a SALSA microarray that contained the 5000 open reading frames (ORFs) identified from the sequence of S. enterica serovar Typhimurium SL1344, as described previously (Balbontin et al., 2006). Hybridization, microarray scanning and data analysis were all performed as described previously (Kelly et al., 2004), using a falsediscovery rate of 0.05. The expression data have been deposited in the NCBI GeneExpression Omnibus http://www.ncbi. nlm.nih.gov/geo/query/acc.cgi?token=pbkdfwskomsowpq& acc=GSE35938 and are accessible through GEO Series Accession Number GSE35938. All microarray data are MIAME-compliant.

ChIP-seq
Salmonella Typhi ompR::3¥FLAG (strain TT53) and S. Typhi BRD948 were cultured in LB broth to OD600 = 0.6, lysed, incubated with 1% formaldehyde at 37°C for 20 min to crosslink DNA with protein then quenched with glycine (ph7) to a final concentration of 0.5 M. Cells were harvested and washed twice in TBS and lysed by osmotic shock. Genomic DNA was then sheered by sonication to an average size of 300 bp and immunoprecipitated using anti-3¥FLAG monoclonal antibody (Sigma, F3165) as previously described (Pfeiffer et al., 2007) using the Protein G Immunoprecipitation kit (Sigma). Eluates were then treated with pronase (0.8 mg ml -1 , Sigma) at 65°C overnight. The nucleotide sequence of genomic DNA fragments was determined by Illumina GAII paired-end sequencing with read length 36 bp and mapped to the S. Typhi Ty2 whole genome sequence (AE014613). Sequence data were mapped to the S. Typhi Ty2 genome using the same parameters as previously described (Perkins et al., 2009), without assigning the sequence reads to each strand. Plots were z-score-normalized, in order to indicate the number of standard deviations above or below the mean for each datum point, and the differences between the untagged S. Typhi Ty2 and ompR::3¥FLAG-tagged associated DNA sequences determined. Plots were then read into the genome browser tool Artemis (Rutherford et al., 2000). The Peakfinder function was used to determine enrichment for OmpR::3¥FLAG bound DNA sequences. The Peakfinder function (36 bp window and z-score cut-off score set to 3) identified 58 peaks. Due to the background noise of the mapped sequence data plots and low stringency of the Peakfinder conditions, identified peaks were then filtered manually. Sites of DNA enrichment present within a predicted or known CDS were ignored unless there were multiple similar sites nearby, reducing the total number of analysed peaks to 43. Enriched sequences were then input to the motif finding algorithm YMF with the length of motif set to eight nucleotides and with a maximum of two redundant bases (Sinha and Tompa, 2003).

RNA extraction, reverse transcription-PCR (RT-PCR) and real-time PCR
RNA was extracted from S. Typhi using a Fast RNA Blue Kit (MP Biomedicals) according to the instructions of the manufacturer. RNA samples (40 mg) were DNase I (Thermo Scientific) treated in a 100 ml volume and diluted to 100 ng ml -1 . RNA samples were reverse transcribed and used as the template for Real-Time PCR with Express One-Step SYBR GreenER (Invitrogen) in a 20 ml total reaction volume. Real-Time PCR was performed using a StepOnePlus Real-Time PCR System (Applied Biosystems) with the oligonucleotides (Sigma) ATATGTTGGGCTTCCTCTGG and TTCAGATAAC GAGCCTCACG (tviB), TTGATGGCCTGCACTACTTC and TGGTTGCCCTGAATCTGATA (ompC), GAAACGCAGAT TAACACCGA and ACTTCCGCGTATTTCAAACC (ompF) and TACCTGCTGGCGGAGATTA and ATACCATGCTGAT GCAGAGAA (waaY). Data were analysed by using the comparative CT method where target gene transcription of each sample was normalized to the CT of the waaY transcript.

Animal experiments
In all mouse experiments female, 7-8 week-old C57BL/6 mice (Charles River) were inoculated orally by gavage with S. Typhimurium suspended in PBS pH 7.4. For mixed inoculum experiments in order to distinguish the wild-type strain from the mutant test strains, a cat (chloramphenicol acetyltransferase, chloramphenicol resistance gene) was inserted in the S. Typhimurium SL1344 chromosome in a position that has previously been described to have no effect on colonization of the murine host (Kingsley et al., 2003;Winter et al., 2010) (phoN locus, strain RAK113). Groups of five mice were inoculated orally with a 1:1 (log10 = 0) of approximately 1 ¥ 10 8 cfu of strain RAK113 and the test strain. When mice were moribund (less than 80% body weight compared with day of inoculation) or on day 5 post inoculation, mice were culled and cfu of each strain in homogenized mesenteric lymph nodes (MLN), caecum, ileum, spleen and liver was determined by serial dilution in PBS pH 7.4 and plating on LB agar containing cloramphenicol and LB agar containing kanamycin. Serial 10-fold dilutions were plated on LB + Cm or LB + Km agar, as appropriate, to determine cfu per organ. The ratio of wild-type (strain RAK113) to test strain was transformed to log10 and to determine if these values were significantly different from the log10 of the input ratio (input ratio log10 = 0) was determined using the two-tailed Student's t-test in the Prism 4 software version 4.0c (Graph Pad). Value for P < 0.05 was considered significantly different.

Minimal media growth assays
Salmonella Typhimurium SL1344 and isogenic mutant strains were grown overnight and washed three times in PBS and then plated onto M9 minimal media supplemented with L-histidine (for SL1344 growth) and 1% agar. The glucose carbon source was substituted with amino sugars. Wild-type controls were grown concomitantly on separate plates made from the same agar mix.

Ethics statement
All animal procedures were performed in accordance with the United Kingdom Home Office Inspectorate under the Animals (Scientific Procedures) Act 1986. The Wellcome Trust Sanger Institute Ethical Review Committee granted ethical approval for these procedures.