|Home | About | Journals | Submit | Contact Us | Français|
Bacteria utilize phosphorelay systems to respond to environmental or intracellular stimuli. Salmonella enterica encodes a four-step phosphorelay system that involves two sensor kinase proteins, RcsC and RcsD, and a response regulator, RcsB. The physiological stimulus for Rcs phosphorelay activation is unknown; however, Rcs-regulated genes can be induced in vitro by osmotic shock, low temperature and antimicrobial peptide exposure. In this report we investigate the role of the Rcs pathway using phylogenetic analysis and experimental techniques. Phylogenetic analysis determined that full-length RcsC-and RcsD-like proteins are generally restricted to Enterobacteriaceae species that have an enteric pathogenic or commensal relationship with the host. Experimental data show that RcsD and RcsB, in addition to RcsC, are important for systemic infection in mice and polymyxin B resistance in vitro. To identify Rcs-regulated genes that confer these phenotypes, we took advantage of our observation that RcsA, a transcription factor and binding partner of RcsB, is not required for polymyxin B resistance or survival in mice. S. enterica serovar Typhimurium oligonucleotide microarrays were used to identify 18 loci that are activated by RcsC, RcsD and RcsB but not RcsA. Five of the 18 loci encode genes that contribute to polymyxin B resistance. One of these genes, ydeI, was shown by quantitative real-time PCR to be regulated by the Rcs pathway independently of RcsA. Additionally, the stationary-phase sigma factor, RpoS (sigmaS), regulates ydeI transcription. In vivo infections show that ydeI mutants are out-competed by wild type 10- to 100-fold after oral inoculation, but are only modestly attenuated after intraperitoneal inoculation. These data indicate that ydeI is an Rcs-activated gene that plays an important role in persistent infection of mice, possibly by increasing bacterial resistance to antimicrobial peptides.
The Enterobacteriaceae Salmonella enterica infect hosts via an oral route. In a given host species, one of two types of infection occur depending on the bacterial serotype. The first type of infection is a non-systemic, self-limiting gastroenteritis that is referred to as a non-typhoidal infection. The second type of infection causes enteric fever and is a systemic infection that occurs when S. enterica traverses the gastrointestinal epithelia and colonizes deeper tissues, specifically the mesenteric lymph nodes (MLNs), spleen and liver. In humans, S. enterica serovar Typhi causes systemic infection (typhoid fever) while S. enterica serovar Typhimurium (S. Typhimurium) causes non-systemic gastroenteritis (food poisoning). In mice, S. Typhimurium causes a systemic infection that resembles human typhoid fever. In both humans and mice, S. enterica reside within macrophage vacuoles during systemic infection. Systemic infections can resolve into asymptomatic and persistent infections, resulting in carriage and intermittent shedding of bacteria in the faeces or urine for months to years (Tsolis et al., 1999).
Bacteria typically utilize phosphorelay signalling cascades in the form of two-component systems to respond to environmental and intracellular stimuli. Typically, two-component systems have two phosphorelay steps and consist of a membrane-associated sensor kinase and a cytoplasmic response regulator. Hybrid sensor kinases have four phosphorelay steps. The sensor protein generally has a kinase (HisKA) domain, a receiver (REC) domain and a histidine-phosphotransfer (HPT) domain. The response regulator has a second REC domain. The biological need for the four-step phosphorelay is poorly understood (Takeda et al., 2001). Certain Enterobacteriaceae, including S. Typhimurium, express an unusual type of hybrid sensor kinase that utilizes two distinct sensor kinase proteins. The Rcs (Regulator of capsule synthesis) phosphorelay system consists of three proteins, RcsC, RcsD (also called YojN) and RcsB (Fig. 1). RcsC and RcsD are periplasmic membrane proteins that heterodimerize to function as a ‘modified hybrid sensor kinase’ and RcsB is a cytoplasmic response regulator (Takeda et al., 2001; Majdalani and Gottesman, 2005). The large periplasmic domains of RcsC and RcsD share weak amino acid identity (< 8%) and therefore each may function as a distinct sensor domain. RcsC contains the HisKA and REC domains while RcsD includes the HPT domain. Although not proven biochemically, it is thought that upon stimulation, RcsC transfers a phosphate from its HisKA domain to its REC domain and then to the HPT domain of RcsD (Majdalani and Gottesman, 2005). The final phosphorelay step involves phospho-transfer from the RcsD HPT domain to the REC domain on RcsB (Takeda et al., 2001).
RcsC is important for systemic disease in two different mouse models of Salmonella infection. Mouse strains that are ‘sensitive’ to S. Typhimurium infection (e.g. BALB/c or C57BL/6) harbour a mutation in the Nramp1 cation transporter (Nramp1−/−) and succumb to infection within days. Mouse strains that are ‘resistant’ to Salmonella infection (i.e. 129Sv6) are similar to outbred mice in that some mice clear the bacteria while others become asymptomatic carriers of S. Typhimurium for up to a year after infection (Tsolis et al., 1999; Monack et al., 2004). Nramp1−/− mice are used to model acute, systemic S. Typhimurium infection whereas wild-type (Nramp1+/+) mice are used to study persistent S. Typhimurium infection (Tsolis et al., 1999). rcsC null mutants poorly colonize the MLNs and spleen of Nramp1−/− mice relative to wild-type bacteria by 11 days after inoculation. Likewise, rcsC null mutants are out-competed by wild-type S. Typhimurium 3 weeks after infection of Nramp1+/+ mice (Detweiler et al., 2003). The phenotype of rcsC mutants in mice may be partially explained by recent evidence that implicates the Rcs pathway in O-antigen production (Delgado et al., 2006). A report from Mouslim et al. found no survival differences in mice infected with rcsB or rcsC loss-of-function mutants. However, as these experiments used high bacterial doses that kill mice within 5 days of infection, it is likely that their assays were not sufficiently sensitive to determine a role for the Rcs pathway in systemic infection (Mouslim et al., 2004). S. Typhimurium does not normally kill mice in the wild (Tsolis et al., 1999). Assays that mimic persistent infection, where the majority of mice survive acute infection and become carriers of S. Typhimurium, are better indicators for the bacterial genes involved in persistence.
In addition to a role in systemic infection in vivo, RcsC is important for antimicrobial peptide (AMP) resistance in vitro, specifically for resistance to the cationic peptide polymyxin B (Detweiler et al., 2003). AMPs represent a broad spectrum defence mechanism produced prophylactically by host cells in the absence of infection. AMP expression levels are increased in response to cytokines, LPS and bacteria. For instance, levels of a cathelicidin-derived antimicrobial peptide (CRAMP) increase in macrophages infected with Salmonella (Rosenberger et al., 2004). As macrophages are the major niche in which S. enterica reside during systemic infection (Tsolis et al., 1999; Monack et al., 2004), the bacteria must encode defence mechanisms that enable them to withstand AMP insult. The PhoQ-PhoP phosphorelay system mediates AMP resistance by activating genes that lead to LPS modification. By reducing the negative charge of LPS, S. enterica reduces the affinity of many AMPs for its bacterial membranes. While the role of the PhoQ-PhoP pathway in AMP resistance is well characterized (Ernst et al., 2001), the genes and mechanisms by which the Rcs phosphorelay pathway mediates AMP resistance and causes disease are unknown.
In this report, we establish the phylogenetic distribution of the Rcs signalling pathway, determine the roles of the different Rcs signalling components in disease, and show that ydeI, an Rcs-regulated gene, is important for systemic infection. First, comprehensive BLAST and phylogenetic analyses of the Rcs sensor proteins, RcsC and RcsD, were performed. Second, the programme of gene expression in Rcs mutants exposed to AMPs was determined with oligonucleotide microarrays. Third, genes identified in the microarray analyses were validated using conventional bacterial genetics and mouse infections.
It has been reported previously that proteins with amino acid identity to RcsC and RcsD are found in Enterobacteriaceae species (Majdalani and Gottesman, 2005). To determine the distribution of RcsC and RcsD within Enterobacteriaceae, protein–protein searches were performed using the Basic Local Alignment Search Tool (BLASTP) at the National Center for Biotechnology Information (NCBI) (Altschul et al., 1997). The query consisted of RcsC and RcsD protein sequences from S. Typhimurium strain LT2. Search results were limited to proteins with amino acid similarity over at least 80% of the full length of RcsC or RcsD. Thus, proteins with identity limited to the cytoplasmic signalling domains (< 80% of the full-length protein) of either RcsC or RcsD were excluded from the final list (Table 1 and Tables S1 and S2 in Supplementary material). To complement the BLASTP analysis, the Simple Modular Architecture Research Tool (SMART) database (Schultz et al., 1998; Letunic et al., 2006) was used to identify proteins with domain organizations (Fig. 1) similar to that of either RcsC or RcsD (see Experimental procedures). All of the bacterial species listed in Table 1 for the BLASTP searches were identified as encoding RcsC- and RcsD-like proteins in the SMART architecture analysis. In addition, only the species listed in Table 1 encode proteins homologous to RcsC or RcsD beyond the common signalling domains. This confirms that while the signalling domains are preserved across bacterial families, proteins with amino acid identity over the entire length of S. Typhimurium RcsC and RcsD are exclusive to Enterobacteriaceae (Majdalani and Gottes-man, 2005). Moreover, RcsC- and RcsD-like proteins are primarily limited to genera that have intimate enteric relationships with animal hosts, the exception being the plant pathogen Erwinia carotovora. The Enterobacteriaceae that are animal pathogens but do not encode RcsC- or RcsD-like proteins are commonly associated with respiratory or urinary tract infections. These observations are consistent with a role for the Rcs signalling pathway in the establishment or maintenance of an enteric bacteria–host association.
It has been shown previously that RcsC contributes to the persistent infection of mice (Detweiler et al., 2003). To determine whether RcsD and RcsB contribute to persistent infection, we constructed strains containing marked deletions of rcsD or rcsB (see Experimental procedures). Additionally, a mutant lacking all three genes (rcsC, rcsD, rcsB) was constructed. The ability of each mutant to survive in mice was tested by competition with wild type in mixed infection assays. Briefly, 129Sv6 mice (Nramp1+/+) were intraperitoneally (i.p.) inoculated with equivalent numbers of differentially marked wild-type and mutant bacteria. Mice were sacrificed 3 weeks later and bacterial loads of wild-type and mutant strains in harvested tissues were determined by plating for colony-forming units (cfu) on selective media. The competitive index (CI) of each mutant was ascertained by determining the log ratio of cfumutant:cfuwild type for each tissue (Fig. 2). As previously reported for rcsC mutants (Detweiler et al., 2003), both rcsD and rcsB mutants were out-competed by wild-type S. Typhimurium. Furthermore, the triple deletion mutant (rcsC rcsD rcsB) was attenuated to the same degree as each of the single deletion mutants (data not shown), consistent with the idea that these proteins function in the same pathway. These data indicate that RcsC, RcsD and RcsB contribute to systemic infection in mice.
RcsC null mutants are sensitive to polymyxin B exposure (Detweiler et al., 2003; Fig. 3). To determine whether RcsD and RcsB are also important for AMP resistance in vitro, mutant strains were tested for sensitivity to polymyxin B exposure and the per cent survival of each mutant after polymyxin B exposure was calculated relative to wild type. Consistent with previous results, rcsC mutants were sensitive to polymyxin B when compared with wild-type S. Typhimurium. Likewise, rcsD and rcsB mutants were sensitive to polymyxin B exposure (Fig. 3). As expected, a phoP mutant control strain was extremely sensitive to polymyxin B (< 0.1% relative survival) (Gunn and Miller, 1996). These data indicate that the Rcs signalling pathway significantly contributes to AMP resistance in vitro.
RcsB can form a heterodimer with the transcription factor RcsA, and RcsB/RcsA heterodimers bind to the RcsAB DNA box within promoters to activate or repress transcription (Wehland and Bernhard, 2000). To determine whether RcsA contributes to systemic infection and AMP resistance, rcsA deletion mutants were tested for virulence in mice and polymyxin B sensitivity in vitro. Unlike rcsC, rcsD and rcsB mutants, rcsA mutants competed as well as wild type in mixed infections of 129Sv6 mice (Fig. 2). These data indicate that RcsA is not essential for persistent, systemic infection. Likewise, rcsA mutants, as well as null mutants of wza (Gottesman and Stout, 1991), a gene activated by RcsAB heterodimers, displayed polymyxin B sensitivity similar to wild type (Fig. 3). Taken together, these data suggest that the Rcs signalling pathway regulates genes important for polymyxin B resistance independently of RcsA.
Rcs-regulated genes were identified using custom oligo-nucleotide microarrays (Agilent) that covered >98% of the S. Typhimurium LT2 genome. Total RNA from wild type and Rcs mutant strains (rcsC, rcsD, rcsB and rcsA) exposed to a sublethal concentration of polymyxin B (1 μg ml−1) was analysed as described in Experimental procedures. The concentration of polymyxin B used has been shown to induce expression of Rcs-activated genes (Bader et al., 2003). A two-class Significance Analysis for Microarrays (SAM) analysis (Tusher et al., 2001) identified 26 genes with expression that was RcsC-, RcsD- and RcsB-dependent but RcsA-independent (Table 2). Of these 26 genes, five were negatively regulated and 21 positively regulated by the Rcs pathway. The five negatively regulated genes were flagellar genes (flgD, flgB), protein synthesis genes (tufA, rplP) and a hydrolase (cdh) (Table 2). Eleven of the positively regulated genes represented eight putative operons (see Fig. S1 in Supplementary material). Many of the identified genes have either putative or unknown functions. Genes from two distinct putative ABC transporter operons were identified (STM1491, STM1492 and STM1951), one of which (STM1491-1494) is Salmonella spp.-specific.
Strains with deletions in each of the 21 positively regulated RcsC/RcsD/RcsB genes listed in Table 2 were constructed using the lambda red recombinase system (Datsenko and Wanner, 2000). If genes were adjacent or potentially part of the same operon, then the entire operon was replaced with a chloramphenicol-resistance gene, resulting in 18 mutant strains. After exposure to polymyxin B, four of the 18 mutants displayed polymyxin B sensitivity (Fig. 4 and Fig. S1 in Supplementary material). Three of the four mutants were intermediate in their polymyxin B sensitivity, while one mutant (STM1284-1285) was considerably more sensitive to polymyxin B exposure (Fig. 4). The remainder of this report focuses on analysis of the putative periplasmic protein YdeI (STM1515), which contributes to polymyxin B resistance and is restricted to Salmonella, Escherichia and Shigella spp. (Table S3 in Supplementary material).
Real-time quantitative PCR analysis confirmed that ydeI is regulated by the Rcs pathway. Wild-type and Rcs pathway null mutants (rcsC, rcsD, rcsB, rcsA) were exposed to polymyxin B to activate the Rcs pathway and induce ydeI expression. As expected, ydeI mRNA was transcribed in wild-type S. Typhimurium and rcsA mutants, but not in any other Rcs mutants after polymyxin B exposure (Fig. 5A). These data corroborate the DNA microarray data and indicate that YdeI is an RcsA-independent effector of the Rcs signalling pathway that is induced upon exposure to AMPs in vitro.
A recent report suggests that regulation of gene expression by the Rcs pathway in Escherichia coli acts through the stationary/stress sigma factor, RpoS (Hengge-Aronis, 2002; Peterson et al., 2006). To determine whether RpoS regulates ydeI transcription, real-time quantitative PCR analysis was performed with RNA isolated from rpoS deletion mutants (Coynault et al., 1996) with and without polymyxin B exposure. The rcsB mutant was included for comparison. ydeI RNA levels were 15- to 18-fold lower in rpoS mutants than in wild-type bacteria, with or without polymyxin B (Fig. 5B). However, ydeI RNA levels were threefold higher in rpoS single mutants than in rpoS rcsB double mutants, with or without polymyxin B treatment. (Fig. 5B). Collectively these data indicate that RpoS is the major sigma factor for ydeI but that RcsB can promote low levels of ydeI transcription when the RpoS is absent.
We tested mutants lacking the ydeI gene for their ability to colonize tissues during persistent infection in competition with wild-type S. Typhimurium in 129Sv6 mice. Data were expressed as the log ratio of mutant:wild type cfu for each tissue (Fig. 6). Three weeks following i.p. inoculation, a phenotype was observed only in the spleen (Fig. 6A). The CI for ydeI mutants in the spleen (average = −0.69, P < 0.005) was similar to the CI for Rcs mutants (average log CI ranged from −0.51 to −0.70; Fig. 2). In contrast, 3 weeks following oral inoculation, ydeI mutants were out-competed 10- to 100-fold by wild type in the Peyer’s Patches, MLN, spleen and liver (Fig. 6C). The attenuation of ydeI mutant colonization observed at 3 weeks after oral inoculation was not apparent 1 week after oral inoculation (Fig. 6B). These data indicate that YdeI is important for persistent infection in mice after oral inoculation and suggest there is a defect in survival but not in colonization. Rescue of the ydeI mutant with the wild-type ydeI gene (see Experimental procedures) restored polymyxin B resistance to wild-type levels and abrogated the survival defect in mice (Fig. 7).
Enterobacteriaceae represents a family of bacteria that exist in intimate relationships with their hosts. These bacteria can be extracellular commensals or intracellular pathogens that cause a range of diseases including urinary tract, respiratory and gastrointestinal tract infections. We determined the distribution of the Rcs sensor kinases, RcsC and RcsD, within Enterobacteriaceae by BLAST and architectural analyses. Organisms that have RcsC- and RcsD-like proteins (amino acid similarity over at least >80% of the full-length sequence) are primarily associated with gut colonization or infection. An exception is the plant pathogen E. carotovora, which causes potato rot. We speculate that this species uses the Rcs pathway differently in plants and/or it colonizes the gut of an animal. Yersinia pestis and Yersinia enterocolitica both contain RcsD proteins that are truncated, raising the possibility that these organisms use a modified version of the pathway. Additional phylogenetic analysis of RcsC and RcsD with PSI-BLAST (Altschul et al., 1997) revealed that the putative periplasmic domains of RcsC in Erwinia, Yersinia, Sodalis and Photorhabdus spp. shared only 33–43% identity with the S. enterica RcsC periplasmic domain (amino acids 39–312). Only the closely related Escherichia and Shigella spp. had RcsC and RcsD (amino acids 39–309) periplasmic domains with > 80% identity to the corresponding domains in S. enterica. As the periplasmic domains are predicted to initiate signalling by sensing change in the bacterial environment, these results suggest that different bacteria may use the Rcs pathway to respond to different signals.
Pathogens are exposed to AMPs throughout infection, and genes that confer resistance to AMPs often are important for survival within the host (Ernst et al., 2001). For instance, the PhoQ/PhoP sensor kinase system is required for AMP resistance and is activated by polymyxin B (Gunn and Miller, 1996; Bader et al., 2003). However, phoP mutant bacteria are exquisitely sensitive to polymyxin B (Gunn and Miller, 1996), whereas mutants in the Rcs pathway are only moderately sensitive (Fig. 3). This is consistent with the notion that the PhoQ/PhoP system serves as a master regulator of AMP resistance, whereas the Rcs pathway modulates bacterial resistance to AMPs. For example, the RcsC/RcsD/RcsB system may regulate responses to specific classes or concentrations of AMPs encountered within host tissues. Alternatively, if the Rcs pathway is one of multiple activators of particular AMP-resistance genes, significant levels of AMP-resistance proteins may still accumulate in an Rcs null mutant.
The RcsB-binding partner, RcsA, contributes to neither polymyxin B resistance in vitro nor infection in mice (Figs 2 and and3).3). Therefore, Rcs-controlled genes that contribute to AMP resistance should be regulated independently of RcsA. A custom-designed high-fidelity oligonucleotide microarray was used to identify Rcs-regulated, RcsA-independent genes in polymyxin B at 37°C. This screen identified 21 loci and strains harbouring mutations in four of the loci were sensitive to polymyxin B exposure (Fig. 4). The remaining genes may not play a role in AMP resistance or may function in redundant pathways such that polymyxin B sensitivity would only be observed in strains with deletions in multiple genes. Previous studies using microarray screens to identify Rcs-regulated genes have been performed in E. coli K12 by isolating total RNA from bacteria incubated at either 20°C (Hagiwara et al., 2003) or 30°C (Ferrieres and Clarke, 2003). The Rcs pathway was activated in these screens with either zinc treatment (Hagiwara et al., 2003) or exogenously expressed djlA (Ferrieres and Clarke, 2003). While these studies revealed interesting Rcs-regulated genes, they were performed in non-pathogenic E. coli and neither study focused on RcsA-independent genes. Of the 21 Rcs-activated, RcsA-independent genes identified in our microarray screen (Table 2), 16 were found in a previous Salmonella microarray analysis of polymyxin B-induced genes (Bader et al., 2003). The Bader et al. microarray analysis focused on genes regulated by the PhoQ/PhoP two-component signalling system, a master regulator of AMP-resistance genes. It is not clear whether the genes reported here or in previous studies (Bader et al., 2003; Ferrieres and Clarke, 2003; Hagiwara et al., 2003) are exclusively Rcs- (or PhoP-) dependent or whether they are regulated by both pathways. One possibility is that Rcs indirectly regulates expression of some genes by affecting the PhoQ/PhoP signalling system. If this were the case, it would suggest cross-talk between the RcsC/RcsD/RcsB and PhoQ/PhoP systems and underscores the complexity of bacterial signalling.
ydeI is one of the array-identified genes that contributes to polymyxin B resistance (Fig. 4). Real-time quantitative PCR analyses confirm that ydeI RNA accumulation is induced by the Rcs pathway independently of RcsA (Fig. 5A), and is dependent on the sigma factor RpoS (Fig. 5B). However, ydeI RNA accumulates in the absence of RpoS, although at much lower levels than in wild type (Fig. 5B). One explanation for this observation is that RcsB can recruit a second sigma factor, such as RpoD (sigma70), to the ydeI promoter but the second sigma factor is much less efficient than RpoS at activating ydeI transcription. In addition to ydeI, other genes identified in our screen are RpoS-regulated. These genes include ecnB (our unpublished data), yeaG (Ibanez-Ruiz et al., 2000), osmY (Bader et al., 2003), bfr, yjbJ (Lacour and Landini, 2004), elaB, yeaH, ygaU, yhbO and yncB (Weber et al., 2005). The 5′ untranslated region of ydeI reveals a degenerate −35 region, consistent with regulation by RpoS (Typas and Hengge, 2006). A recent report suggests that the Rcs pathway in E. coli promotes RpoS protein accumulation by blocking the activity of an RpoS repressor, LrhA (Peterson et al., 2006). It is possible that RcsB promotes ydeI transcription via RpoS by inhibiting LrhA. However, as RpoS is a global sigma factor, specific transcription factors, such as RcsB, may determine the activation of specific genes under certain conditions (Hengge-Aronis, 2002). For instance, RcsB, a proven DNA-binding protein and transcriptional regulator (Wehland and Bernhard, 2000), may determine the transcription of ydeI in the presence of AMPs by binding to the promoter of ydeI and recruiting RpoS. Alternatively, an unidentified RcsB-activated transcription factor may bind the ydeI promoter and recruit RpoS. Thus, the mechanism of transcriptional activation for ydeI and other polymyxin B-induced genes after induction of the Rcs pathway remains to be determined.
In competitive infections, ydeI null mutants were out-competed by wild-type S. Typhimurium in 129Sv6 mice. After i.p. inoculation, ydeI mutants had only a subtle deep tissue phenotype, comparable to that of Rcs mutant strains (Figs 2 and and6A).6A). However, after oral inoculation, 10- to 100-fold more wild type than ydeI mutant bacteria were recovered from both intestinal and deep tissues (Fig. 6C). The dependence of phenotype severity on inoculation route may be informative. Oral inoculation delivers the bacteria into the stomach. S. Typhimurium must survive transit through to the small intestine where the bacteria breach the intestinal wall and access deep tissues, including the MLNs, spleen and liver. In contrast, i.p. inoculation by-passes the gastrointestinal tract and results in the rapid distribution of bacteria to the spleen and liver via blood-borne phagocytes. Several S. Typhimurium mutants are attenuated for virulence after oral but not i.p. inoculation. For instance, mutants lacking a functional Salmonella pathogenicity island-1 (SPI1) type three secretion system (T3SS) are out-competed by wild type 10-fold in the MLN and spleen only after oral inoculation (Galan and Curtiss, 1989; Jones and Falkow, 1994; Baumler et al., 1997). SPI1 mediates epithelial cell invasion, suggesting that invasion of intestinal epithelial cells is important for bacterial access to deep tissues (Baumler et al., 1997). These experiments were performed in a mouse model of acute infection (BALB/c, Nramp1−/− mice). In a model of persistent infection (129Sv6, Nramp1+/+ mice), ydeI mutants show a similar pattern in that they are strongly out-competed by wild type in deep tissues only after oral inoculation (Fig. 6). We think it unlikely that YdeI regulates the SPI1 T3SS because cell invasion defects were not observed for ydeI mutants in tissue culture experiments where SPI1 mutant (invA) strains were used as a control (data not shown). Perhaps YdeI is needed after oral but not i.p. inoculation because it plays a role in traversing the intestinal barrier to allow for colonization of deep tissues, an event that may occur repeatedly during persistent infection as the bacteria re-seed the Peyer’s Patches or MLNs from the gastrointestinal tract.
The molecular mechanism of YdeI is unknown. YdeI is a putative periplasmic protein that could interact directly with the periplasmic loops of either RcsC or RcsD to regulate Rcs signal transduction. It is possible that YdeI is activated by the Rcs pathway to regulate another signalling pathway, such as PhoQ/PhoP or PmrA/PmrB. Alternatively, YdeI could contribute to the stability of the cell envelope by directly or indirectly modifying outer membrane proteins, lipids, or polysaccharides. By understanding the mechanisms of YdeI and other genes regulated by the Rcs pathway we will gain insights into how pathogens adapt to changes in their host microenvironments.
All bacteria were grown and maintained at 37°C with agitation in Luria–Bertani (LB) broth or on LB agar plates. The following antibiotic concentrations were used: streptomycin (Str) 200 μg ml−1, kanamycin (Kan) 30 μg ml−1, chloramphenicol (Cm) 10 μg ml−1. Strains described in this study were derived from wild-type serovar Typhimurium strain SL1344 (source ATCC) (Smith et al., 1984). Deletion mutants (Table 3) were generated using the methods of Datsenko and Wanner (2000). Oligos used for generating the deletion mutants are detailed in Table S4 in Supplementary material. Mutations were made in the 14028 strain background (source ATCC), verified by PCR and transduced into SL1344 using standard P22 phage transduction. Transductants were verified by growth on LB agar containing Str and the appropriate deletion-marked antibiotic resistance cassette (14028 is not Str-resistant whereas SL1344 is Str-resistant) as well as by PCR.
Rescued ydeI strains were generated by insertion of a Cm-resistance cassette in the intergenic region downstream from the ydeI stop codon (STM nucleotides 1594016–1594315; forward primer: 5′-GCGTGC ATTCGGATTTTTCTACTTATTTTTCCGTGGTGGCGTGTAG GCTGGAGCTGCTCC-3′; reverse primer: 5′-CAGCGAAGC GGTAATAAATCTTTGGCAATCAAGCCATGGGAATTAGCC ATGGTCC-3′; underline indicates P1 and P2 priming sites respectively). The Cm-marked, wild-type ydeI strain (kde605) was phage transduced into kde445 in which the Cm-resistance cassette had been removed using the pCP20 plasmid (kde602) (Datsenko and Wanner, 2000). Cm-resistant colonies were checked by PCR for the wild-type ydeI gene and the intergenic Cm-resistance cassette. Positive colonies were transformed with pCP20 to remove the Cm-resistance cassette from the ydeI intergenic region. Plasmid-cured colonies were checked by PCR for loss of the Cm-resistance cassette (kde608).
All phylogenetic work used S. Typhimurium LT2 RcsC and RcsD (Accession No. P58662 and Q8ZNH2 respectively) as the query sequences.
BLASTP analysis was performed at http://www.ncbi.nlm.nih.gov/BLAST. Search results (Altschul et al., 1997) were limited to proteins that were at least 35% identical to at least 80% of the full length of RcsC or RcsD. This eliminated proteins with amino acid identity restricted to the cytoplasmic signalling domains (< 75% of the full-length protein) of either RcsC or RcsD. Databases included were NCBI and the Sanger Center Y. enterocolitica 8081 sequence.
SMART analysis was performed at http://smart.embl-heidelberg.de/. The SMART search (Schultz et al., 1998; Letunic et al., 2006) for RcsC-like proteins identified proteins that contained a HisKA, histidine ATPase (HATPase_c) and REC domain but lacked a HPT domain. This search returned a total of 1434 protein sequences in Bacteria, including 361 in Gammaproteobacteria and 28 in Enterobacteriales. The SMART search for RcsD-like proteins identified those that contained HATPase_c and HPT domains but lacked HisKA, REC or CheW domains. Proteins with REC or CheW domains were excluded to eliminate chemotaxis proteins. This search returned 23 proteins in the Kingdom Bacteria, 20 of which were in the Proteobacteria phylum. Enterobacteriales includes 11 of the 20 Proteobacteria RcsD-like proteins. The non-Gammaproteobacteria proteins identified with RcsD-like architecture encode the HPT domain before the HATPase_c domain, while in all enterobacteria these domains are reversed. All of the bacterial species listed in Table 1 for the BLASTP searches were identified as encoding RcsC- and RcsD-like proteins in the SMART architecture analysis. Only proteins from the species in Table 1 were homologous to S. enterica RcsC or RcsD beyond the common signalling domains, which supports the BLASTP results.
For microarray and real-time PCR analysis, overnight bacterial cultures were diluted 1:100 into 50 ml of selective LB media and grown for 3 h at 37°C with shaking. The cultures were split in half and polymyxin B was added to a final concentration of 1 μg ml−1 to one set of samples while the other set was left untreated. All cultures were incubated at 37°C for 40 min with shaking. Bacteria were collected by centrifugation and processed for RNA.
For sensitivity assays (Gunn and Miller, 1996), 1 ml of overnight bacterial cultures was spun in a microcentrifuge and washed with PBS three times. After the final wash, the bacterial pellet was resuspended in 900 μl of MGM media and diluted 1:100 into 3 ml of MGM media. Cultures were grown for 3 h with agitation at 37°C. After 3 h, the OD600 for each of the cultures was determined. One thousand five hundred bacteria were inoculated into 3 ml of LB plus 2–2.5 μg ml−1 polymyxin B (freshly prepared stock) and incubated at 37°C. After 1 h incubation, samples were placed on ice and 100 μl was spread onto selective LB agar plates to determine cfuplus polymyxin B. Input samples were determined by inoculating 3 ml of LB without polymyxin B with 1000 bacteria and plating immediately (cfuinput). Per cent survival was determined by dividing the cfuplus polymyxin B by cfuinput. Each polymyxin B-treated sample was performed in triplicate.
RNA was isolated from samples using a combination of Trizol (Invitrogen) and Qiagen’s RNeasy kit. Briefly, bacterial pellets were incubated in TE/lysozyme (1 mg ml−1) at room temperature for 5 min. Buffer RLT (Qiagen) was added to lysed samples and insoluble material was removed by centrifugation. Samples were extracted with Trizol reagent by incubating at room temperature for 15 min. After chloroform extraction and centrifugation, the Trizol layer was chloroform-extracted a second time. The aqueous layers were pooled and total RNA was precipitated with ethanol and purified using RNeasy columns (Qiagen) per manufacturer’s protocol. An on-column DNase digestion was performed using DNase I (Qiagen) for 1 h at room temperature. Total RNA was eluted using RNase-free water and analysed for purity using the Agilent 2100 Bioanalyser and RNA 6000 LabChip kit (Agilent).
The microarrays covered the entire Salmonella genome, with many genes being represented at least twice on the array. Dye-swap experiments confirmed there was little dye bias. Each experiment was conducted at least twice with different biological replicates. Wild-type cDNA was labelled with Cy-3 and cDNA from each of the mutants was labelled with Cy-5. Custom oligonucleotide microarrays (Agilent) were hybridized using in situ Hybridization Kit Plus (Agilent) with a mix of wild type and mutant labelled cDNAs for 17 h at 60°C with rotation. Hybridized arrays were processed by washing in 6×SSPE/0.05% lauroylsarcosine at room temperature and 0.6×SSPE/0.005% lauroylsarcosine at room temperature. Arrays were dried by rinsing in Agilent’s Stabilization and Drying solution. Dried arrays were scanned on the Agilent Microarray Scanner System and data were compiled using Feature Extractor software (Agilent).
The log ratio of expression values (mutant-Cy-5/wild-type-Cy-3) for each replicate was compiled and analysed using a two-class, unpaired analysis in the SAM software from Stan-ford University (Tusher et al., 2001). We identified genes with expression ratios that differed between rcsC/rcsD/rcsB mutants (class one) and rcsA mutants (class two).
Seven-week-old female 129Sv6 mice (Taconic Farms) were inoculated with equivalent numbers of differentially marked wild type (hisG::kan or hisG::cm) and mutant bacteria. Overnight cultures of bacteria were diluted to appropriate densities based on OD600 in PBS. Each mouse received either 100–250 (intraperitoneal) or 109 (oral) total bacteria. The ratio of input bacterial strains was determined by plating on selective LB agar plates. Three weeks after inoculation, mice were sacrificed. Tissues (cecum, MLNs, Peyer’s patches, spleen, liver) were collected in 1 ml of PBS, homogenized with a TissueMiser (FisherBrand) and diluted in PBS for plating on selective LB agar plates. The log ratio of competitive indices was calculated as follows: log10[(cfumutant/cfuwild type)output/(cfumutant/cfuwild type)input] (Beuzon and Holden, 2001).
Bacterial strains were exposed to polymyxin B and RNA was isolated as described for microarray analysis. Total RNA was reverse transcribed into cDNA with the SuperScript III First Strand Synthesis system (Invitrogen). Gene expression was determined using SYBR green (ABI) and primers specific to ydeI (forward primer, 5′-ATTGAGGATGGTTATCGCGGTA-3′; reverse primer, 5′-CCTGTTCGATGGTCATTTTTTCT-3′) or gapA (forward primer, 5′-TGTTTTCCGTGCTGCTCAGA-3′; reverse primer, 5′-TTGATTGCAACGATCTCGATGT-3′). Reactions were run on a 7300 Real-Time PCR System (ABI) and analysed using 7000 SDS 1.1 RQ software (ABI).
Expression of the Salmonella gene, gapA, was used as an internal control and ydeI levels were normalized to gapA levels in each of the strains. A standard curve of RNA isolated from SL1344 exposed to polymyxin B was used for each PCR primer set to determine efficacy of the primers and quantitative expression level of the RNA.
We would like to thank Jennifer Martin, Brad Olwin, Norm Pace, MaryAnn DeGroote and Sara Symons for their critical reading of the manuscript, Katy Metzler for her help with mouse infections, and Yiwen He and Ryan Gill for performing the in silico analysis of the probes used on the DNA microarray. We thank William Navarre for suggesting a link between ydeI and RpoS. We thank Virginia Miller for suggestions regarding the Rcs pathway in Yersinia spp. This work was supported by the University of Colorado at Boulder, a Jane and Charlie Butcher Award (C.S.D. and Ryan Gill) and R56 AI063116-01A1 (C.S.D.).