Alternative sigma factor interactions in Salmonella: σE and σH promote antioxidant defences by enhancing σS levels



Hierarchical interactions between alternative sigma factors control sequential gene expression in Gram-positive bacteria, whereas alternative sigma factors in Gram-negative bacteria are generally regarded to direct expression of discrete gene subsets. In Salmonella enterica serovar Typhimurium (S. Typhimurium), σE responds to extracytoplasmic stress, whereas σH responds to heat shock and σS is induced during nutrient limitation. Deficiency of σE, σH or σS increases S. Typhimurium susceptibility to oxidative stress, but an analysis of double and triple mutants suggested that antioxidant actions of σE and σH might be dependent on σS. Transcriptional profiling of mutant Salmonella lacking σE revealed reduced expression of genes dependent on σH and σS in addition to σE. Further investigation demonstrated that σE augments σS levels during stationary phase via enhanced expression of σH and the RNA-binding protein Hfq, leading to increased expression of σS-dependent genes and enhanced resistance to oxidative stress. Maximal expression of the σS-regulated gene katE required σE in Salmonella-infected macrophages as well as stationary-phase cultures. Interactions between alternative sigma factors permit the integration of diverse stress signals to produce coordinated genetic responses.


Bacterial sigma factors interact with RNA polymerase core enzyme to initiate transcription. In enteric bacteria such as Escherichia coli and Salmonella enterica, σD/70 (encoded by rpoD) is responsible for expression of most housekeeping genes required for normal cellular metabolism during exponential growth. To maintain cellular homeostasis in a variety of stress environments, bacteria can utilize alternative sigma factors to redirect RNA polymerase and selectively express discrete subsets of genes. S. enterica serovar Typhimurium (S. Typhimurium) has five alternative sigma factors: σE/24 (rpoE), σH/32 (rpoH), σS/38 (rpoS), σN/54 (rpoN) and σ28 (fliA) (Gruber and Gross, 2003).

In response to environmental cues, the cellular levels of alternative sigma factors available for interaction with core RNA polymerase is controlled by the actions of anti-sigma factors (Hughes and Mathee, 1998) and proteases (Bertani et al., 2001; Hengge-Aronis, 2002; Walsh et al., 2003). σE is activated by the presence of unfolded proteins in the cell envelope, whereas σH is activated by unfolded cytoplasmic proteins (Gruber and Gross, 2003). In E. coli, phosphorylation of anti-σE factor RseA and σH has been reported to modulate the activities of σE and σH (Klein et al., 2003). A variety of stress conditions can stimulate the σS regulon, most classically nutrient deprivation or stationary phase (Hengge-Aronis, 2002). Both σE and σS are essential for Salmonella virulence (Fang et al., 1992; Nickerson and Curtiss, 1997; Humphreys et al., 1999; Testerman et al., 2002), reflecting the diverse stresses imposed by the host environment upon pathogenic bacteria.

Sporulation in Gram-positive Bacillus and Streptomyces sp. is controlled by a cascade of sigma factors whose sequential activation by the preceding sigma factor allows the orderly expression of developmental genes (Stragier and Losick, 1990; Kormanec et al., 1996; Rudner and Losick, 2001; Gruber and Gross, 2003). In contrast, the regulons controlled by alternative sigma factors in Gram-negative enteric bacteria are generally conceived of as non-overlapping subsets of genes, although several phenotypes of sigma-deficient mutant strains such as resistance to heat shock, oxidative stress or nutrient deprivation have been shown to functionally overlap (Jenkins et al., 1991; McCann et al., 1991; Kogoma and Yura, 1992; Rouviere et al., 1995; Spector, 1998; Matin et al., 1999; Kenyon et al., 2002; Testerman et al., 2002).

However, the possibility of important regulatory interactions linking the alternative sigma factors in enteric bacteria has also been suggested. Stress conditions such as hyperosmolarity or stationary phase can simultaneously induce σE, σH and σS (Jenkins et al., 1991; Bianchi and Baneyx, 1999; Nitta et al., 2000; Kenyon et al., 2002; Testerman et al., 2002). One of the three rpoH promoters is activated by σE (Erickson and Gross, 1989; Wang and Kaguni, 1989). σH can activate expression of the hfq gene encoding an RNA-binding regulatory protein (Tsui et al., 1996), and this protein in turn is an important determinant of rpoS translation (Muffler et al., 1996).

In microarray analyses of transcription during stationary phase, an rpoE mutant strain of S. Typhimurium was found to exhibit reduced expression of a number of σH- and σS-regulated genes in addition to reduced expression of loci known to be directly regulated by σE. This observation suggests that sigma factor cascades may occur in Gram-negative as well as in Gram-positive bacteria. This report describes the results of an investigation to determine whether regulatory interactions between σE, σH and σS have important functional consequences for bacterial stress resistance and to elucidate the mechanism of interaction.


Salmonella requires σE, σH and σS for resistance to oxidative stress

Levels of σS increase when carbon sources are limited, thereby enhancing the resistance of E. coli or Salmonella to the effects of prolonged starvation and oxidative stress (Hengge-Aronis, 2002). A similar role for σE has also been suggested from studies in Salmonella (Kenyon et al., 2002; Testerman et al., 2002). In E. coli, σH levels rise during starvation (Jenkins et al., 1991), and rpoH mutant bacteria have increased susceptibility to oxidative stress (Kogoma and Yura, 1992). We therefore compared the effects of rpoE, rpoH and rpoS mutations on the resistance of Salmonella to oxidative stress. The viability of rpoH mutant S. Typhimurium was decreased approximately 10-fold relative to wild-type cells after 1 h challenge with 4 mM H2O2(Fig. 1A), and was similar to the effects of an rpoE mutation. As observed previously (Fang et al., 1992), rpoS mutants were extremely susceptible to challenge with 4 mM H2O2 (no detectable survivors after 30 min), confirming a dominant role of σS in oxidative stress resistance. Using a lower concentration of H2O2 (1 mM), strains carrying both rpoE and rpoH mutations were more susceptible than strains carrying either mutation alone (Fig. 1B). However, double (rpoE rpoS, rpoH rpoS) and triple (rpoE rpoH rpoS) mutants exhibited no additional impairment in survival compared with strains carrying only an rpoS mutation, suggesting that the promotion of oxidative stress resistance by σE and σH might require σS

Figure 1.

Hydrogen peroxide susceptibility of sigma factor mutants. Cells grown overnight in LB media were washed and resuspended in M9 minimal media without glucose at the concentration of 105 cfu (colony forming unit) ml−1. Cells were subsequently challenged with 4 mM (A) or 1 mM (B) hydrogen peroxide. Viable counts were determined at indicated time points after challenge by serially diluting aliquots in phosphate-buffered saline (PBS) and plating. Data represent the means of triplicate experiments with standard deviation. The asterisk indicates that no detectable surviving rpoS single, double or triple mutant bacteria were detected after 30 min exposure to 4 mM hydrogen peroxide.

An rpoE null mutation reduces σS levels

As part of a comprehensive analysis of the effects of σE on gene expression, microarray analyses were performed using cDNA from E. coli K-12, S. Typhimurium 14028s and isogenic rpoE null mutant derivatives grown to stationary phase. E. coli rpoE mutant strain PND818 carries an uncharacterized suppressor mutation(s) that allows growth in the absence of σE, which is otherwise essential in E. coli (De Las Penas et al., 1997a), but rpoE mutant S. Typhimurium strains without a second-site suppressor mutation are viable (Testerman et al., 2002). In the microarray analysis of E. coli, several genes known to belong to the σS (e.g. katE, poxB) or σE (e.g. nlpB, surA) regulons were downregulated in the rpoE mutant strain (Fig. 2A). The microarray analysis was subsequently expanded to include Salmonella rpoS mutants as well. A large number of σS-dependent genes were found to require σE for maximal expression (Table 1). Among these were loci directly related to antioxidant defence, including katE (catalase), sodC (periplasmic superoxide dismutase) and adhE (Fe-activated alcohol dehydrogenase) (Fang et al., 1999; Melov et al., 2000; Echave et al., 2003). These results suggested that σE promotes resistance to oxidative stress by augmenting expression of the σS regulon, which could account for the failure of an rpoE mutation to further increase the H2O2 susceptibility of an rpoS mutant strain (Fig. 1B).

Figure 2.

σE-dependent σS expression.
A. Sections of microarray containing nlpB, surA, katE and poxB mRNA from E. coli MC1061 and isogenic rpoE null mutant strain PND818. cDNA was synthesized from total RNA of stationary-phase cells grown overnight in LB broth and hybridized onto a membrane as described in Experimental procedures.
B. Effect of rpoE mutation on katE and spvA transcription in S. Typhimurium. Top: slot blot analysis. RNA extracted from wild-type S. Typhimurium 14028s and an isogenic rpoE mutant strain grown overnight were hybridized with a katE-specific DIG-labelled DNA probe. An rpoD probe was used as an internal control for the RNA level of each strain. Bottom: the same RNA samples were used to measure katE and spvA mRNA levels by quantitative real-time RT-PCR.
C. Effect of rpoE mutation on σS expression. Total protein isolated from wild-type, rpoE, hfq and rpoS mutant S. Typhimurium cells grown overnight in LB was processed for immunoblot analysis with an anti-σS antibody.

Table 1. Relative expression of σS-dependent genes in rpoE and rpoS mutant S. Typhimurium.
GeneFunctionExpression ratio compared with wild type
rpoE rpoS
spvA Salmonella plasmid virulence: outer membrane protein0.240.11
spvB Salmonella plasmid virulence: hydrophilic protein0.450.15
spvC Salmonella plasmid virulence: hydrophilic protein0.240.41
grxB Glutaredoxin 20.560.47
astD Succinylglutamic semialdehyde dehydrogenase0.280.58
katE Catalase; hydroperoxidase HPII(III), RpoS dependent0.490.04
sodC Copper/zinc superoxide dismutase0.320.15
adhP Alcohol dehydrogenase, propanol preferring0.200.59
narZ Nitrate reductase 2, alpha subunit0.480.12
narY Nitrate reductase 2, beta subunit0.630.15
narW Nitrate reductase 2, delta subunit0.480.23
yciG Putative cytoplasmic protein0.390.04
yciE Putative cytoplasmic protein0.220.01
STM1731Putative catalase0.150.01
adhE Iron-dependent alcohol dehydrogenase of the multifunctional alcohol dehydrogenase AdhE0.440.16
yohF Putative oxidoreductase0.530.20
gabD Succinate-semialdehyde dehydrogenase I, NADP dependent0.290.45
gabT 4-Aminobutyrate aminotransferase0.230.37
gabP APC family, gamma-aminobutyrate transport protein0.410.28
aldB Aldehyde dehydrogenase B (lactaldehyde dehydrogenase)0.450.13

To confirm the microarray results, levels of specific mRNAs were compared in wild-type and mutant strains. By slot blot analysis, markedly reduced levels of katE mRNA were found in an rpoE mutant strain in comparison to mRNA from the housekeeping σ70 gene rpoD (Fig. 2B, top). Quantitative real-time reverse transcription polymerase chain reaction (RT-PCR) analysis also showed σE-dependent transcription of katE and another gene belonging to the σS regulon (spvA) which is essential for Salmonella virulence in mice (Fig. 2B, bottom). σE and σS recognize distinct −35 and −10 promoter elements (Raina et al., 1995; Rouviere et al., 1995; Becker and Hengge-Aronis, 2001), and a computational analysis using a weighted matrix of σE promoter consensus sequences indicated that most σS-dependent genes lack a canonical σE promoter element (data not shown). Therefore, an effect of σE on σS itself was considered. Immunoblotting of whole-cell lysates from stationary-phase cultures was performed to compare σS levels in wild-type and rpoE mutant strains. σS levels were reduced approximately twofold in rpoE mutant S. Typhimurium (Fig. 2C), indicating that σE influences expression of the σS regulon by modulating the concentration of σS protein. Control of σS expression is extremely complex, including transcriptional, translational and post-translational controls (Hengge-Aronis, 2002). To determine whether σE controls rpoS transcription, the β-galactosidase activity expressed by an rpoS–lacZ transcriptional fusion was determined in wild-type and rpoE mutant bacteria. No significant difference in β-galactosidase activity was detected between the two strains (data not shown), suggesting that σE controls σS expression at a post-transcriptional level.

σE is required for transcription of rpoH and hfq in stationary phase

Next, the mechanism by which σE post-transcriptionally regulates σS expression was investigated. In the microarray experiments described above, hfq mRNA levels were significantly reduced in an rpoE mutant strain (Fig. 3A). Hfq (also known as HF-1), originally described as a factor essential for the replication of phage Qβ RNA, is required for efficient translation of rpoS mRNA. Although no canonical σE promoter was found upstream of hfq, one of the three hfq promoters in E. coli is σH dependent (Tsui et al., 1996). Sequence alignment of the E. coli and S. Typhimurium promoter regions revealed absolute conservation (not shown). Reduced hfq transcription in rpoE and rpoH mutant strains was confirmed with an hfq–lacZ transcriptional fusion (Fig. 3A), comparable to the reduction in katE–lacZ expression observed in rpoH mutants (Fig. 4A). These results suggested that σE might regulate hfq transcription via σH. To confirm the predicted effects of σE on rpoH expression, rpoH transcription was determined by quantitative real-time polymerase chain reaction (PCR) measurement of rpoH mRNA and β-galactosidase assay of a lacZ transcriptional fusion to the rpoH promoter. Both assays confirmed that S. Typhimurium expression is regulated by σE during stationary phase (Fig. 3B). Furthermore, an rpoE mutation resulted in reduced β-galactosidase expression from a dnaK–lacZ transcriptional fusion whose promoter is absolutely dependent on σH, confirming that maximal expression of the σH regulon in stationary phase requires σE (Fig. 3B). Collectively, these results suggested that σE indirectly regulates hfq transcription, rpoS translation, and therefore expression of σS-dependent genes via direct effects on rpoH transcription.

Figure 3.

Regulatory cascade of σE–σH–Hfq modulates σS expression.
A. σE- and σH-dependent hfq transcription. Left: a section of microarray comparing hfq mRNA in wild type and rpoE mutant E. coli. Right: β-galactosidase activity expressed from an hfq–lacZ transcriptional fusion in overnight LB broth cultures of wild-type 14028s, rpoE and rpoH mutant strains.
B. σE-dependent rpoH transcription. Left: rpoH mRNA was measured by quantitative real-time RT-PCR analysis of total RNA isolated from 14028s wild type and isogenic rpoE mutant strains. Right: β-galactosidase activity expressed by the rpoH–lacZ and dnaK–lacZ transcriptional fusions was measured as described for (A).

Figure 4.

Effect of σE overexpression on downstream genes.
A. σH- and Hfq-dependent katE transcription in rseA mutant Salmonella. β-Galactosidase activity of the P1/P3rpoE–lacZ or katE–lacZ transcriptional fusions was measured in cells under the same growth conditions as described in Fig. 2. Loss of the RseA anti-sigma results in specific activation of the P3 rpoE promoter, demonstrating an increase in σE activity. Expression of the σS-dependent katE gene is dependent on rpoE, rpoH and hfq, and increases in an rseA mutant strain.
B. Effect of OmpC overexpression on rpoE, rpoH, hfq and katE transcription. Stationary-phase cells containing either pBAD::ompC or the pBAD30 vector grown in LB broth were incubated for 1 h after the addition of l-arabinose (0.2%) to the culture medium. Total RNA was purified and subjected to quantitative real-time RT-PCR. Increased σE expression leads to increases in rpoH, hfq and katE expression as well.

Mutational analysis of the −10 region of the σE- and σH-dependent rpoH and hfq promoters

Previous studies in E. coli have demonstrated multiple promoters  controlling  expression  of  the  rpoH and hfq genes.  The  σE-dependent  rpoH promoter  and σH-dependent hfq promoter are conserved in S. Typhimurium. To examine the contribution of σE-dependent rpoH transcription during stationary-phase growth, we mutated the σE-specific TCTGA −10 element (corresponding to nucleotides 3736864–3736868 of the LT2 genome; ftp:ftp.ncbi.nih.govgenomesBacteriaSalmonellatyphimuriumLT2NC003197.gbk) to a non-functional GGGGG sequence and used the mutant promoter to drive expression of a lacZ reporter. The mutant construct expressed approximately 50% of wild-type β-galactosidase activity (Table 2), comparable to the effects of an rpoE mutation, suggesting that normal levels of stationary-phase rpoH transcription require σE. Similarly, the σH-specific CCTCATTTA −10 region (corresponding to nucleotides 4603673–4603681 of the LT2 genome) of the hfq promoter was mutated to AAAACGGGG. This mutation decreased hfq transcription in wild-type bacteria but had little effect on an rpoE mutant (Table 2). Together, these results are consistent with the sequential activation of rpoH and hfq transcription by σE and σH.

Table 2. Effect of promoter mutations on σE- and σH-dependent expression of rpoH– and hfq–lacZ.
StrainGenotype lacZ fusionβ-Galactosidase activity (Miller units)
IB152Wild typePrpoH–lacZ11 889 ± 1836
IB562Wild typePrpoHσE −10 mutation–lacZ 6 631 ± 330
IB153 rpoE PrpoH–lacZ 4 799 ± 1838
IB564 rpoE PrpoHσE −10 mutation–lacZ 3 532 ± 640
IB130Wild typePhfq–lacZ13 376 ± 1595
IB566Wild typePhfqσH −10 mutation–lacZ 4 306 ± 1305
IB131 rpoE Phfq–lacZ 6 707 ± 903
IB568 rpoE PhfqσH −10 mutation–lacZ 7 809 ± 1837

Enhanced σE expression increases σS levels via σH and Hfq

To explore the regulatory effect of σE on σS expression when σE is overexpressed, we constructed a null mutation in the rseA gene that encodes the anti-σE factor RseA. Mutants lacking rseA exhibit enhanced transcription from σE-dependent promoters in E. coli (De Las Penas et al., 1997b). To compare expression of σE and the housekeeping sigma σD in an rseA mutant strain, lacZ transcriptional fusions to the σD-specific P1 promoter and σE-specific P3 promoter of rpoE were used (Testerman et al., 2002). [Note: since our previous report on rpoE expression in S. Typhimurium, a recent study identified an additional σD-specific rpoE promoter in S. Typhimurium designated P2 (Miticka et al., 2003). To maintain uniform nomenclature with regard to the S. Typhimurium promoters, we have renamed the σE-dependent lacZ fusion P3rpoE–lacZ.] An rseA mutation in S. Typhimurium elevated β-galactosidase activity of P3rpoE–lacZ approximately 10-fold in stationary-phase culture, whereas the activity of σD-dependent P1rpoE–lacZ was unaffected (Fig. 4A, left), demonstrating increased specific expression of the σE regulon in the absence of rseA. To monitor σS expression in the rseA mutant strain, a katE–lacZ transcriptional fusion was employed (Brown and Elliott, 1996). The katE promoters of both E. coli and S. Typhimurium have cytosine at the −13 position which is essential for σS-dependent expression (Becker and Hengge-Aronis, 2001). A computational search using a weighted matrix of σE or σH consensus sequences failed to identify a σE- or σH-dependent promoter upstream of katE (not shown). β-Galactosidase activity expressed by katE–lacZ increased significantly in the presence of an rseA mutation, and this increase was largely dependent on functional σH and Hfq (Fig. 4A, right), highlighting the importance of the σE–σH–Hfq regulatory cascade in modulation of the σS regulon. Importantly, plasmid-driven overexpression of Hfq in an rpoE hfq mutant strain could restore increased katE–lacZ expression to levels exceeding those observed in wild-type cells, demonstrating that Hfq is limiting for expression of the σS regulon under native conditions in stationary phase. Induction of σE by overexpression of the OmpC porin, which sequentially activates proteases DegS and YaeL to degrade RseA and release σE (Walsh et al., 2003), enhanced transcription of rpoH, hfq and katE, as well as rpoE, compared with control cells harbouring an empty vector plasmid (Fig. 4B), and this increase was completely abrogated in an rpoE mutant strain. Together with the observations in rpoE mutant strains, these results demonstrate that activation of σE significantly enhances expression of the σS regulon via σH and Hfq during stationary phase.

σE enhances σS expression during nutrient limitation via σH and Hfq

Previously, it has been demonstrated that Salmonella experiences nutrient limitation following uptake by macrophages (Abshire and Neidhardt, 1993). As σE, σH and σS are each induced by carbon starvation in vitro (Jenkins et al., 1991; Hengge-Aronis, 2002; Kenyon et al., 2002), we investigated whether σE is required for expression of σS during nutrient limitation. Cells grown to early log phase in nutrient-rich LB medium were shifted to nutrient-poor M9 medium, and β-galactosidase activity was expressed by rpoE P3 (Kenyon et al., 2002; Testerman et al., 2002) and katE transcriptional fusions with lacZ was measured at timed intervals. Expression of P3rpoE–lacZ dramatically increased upon nutrient deprivation, in parallel with katE–lacZ expression. Mutations in rpoS or rpoE blunted the increase in katE–lacZ expression (Fig. 5A), and immunoblotting with an anti-σS antibody confirmed σE-dependent σS expression, demonstrating that σS regulation by σE is physiologically relevant during nutrient limitation.

Figure 5.

Effects of σE, σH and Hfq on σS expression during nutrient limitation.
A. σE-dependent σS expression. Top: log-phase (OD600∼0.3) cells grown in LB were washed with M9 broth containing 0.002% glucose and resuspended in the same broth at identical cell density. Cell aliquots were collected at indicated time points for measuring β-galactosidase activity expressed by P3rpoE–lacZ and katE–lacZ transcriptional fusions (top) and for immunoblot analysis with an anti-σS antibody (bottom). Solid and open symbols are used for β-galactosidase activity and growth rate of cells respectively: wild type, squares; rpoE mutant, diamonds; rpoS mutant, triangles. Data represent three independent experiments.
B. Effect of σH and Hfq on katE–lacZ expression. β-Galactosidase activity of katE–lacZ was measured 4 h after changing the media as described in (A). Values shown represent fold-increase calculated as a ratio of β-galactosidase activity before and after media change.

Induction of katE–lacZ expression following a shift to nutrient-poor medium was virtually abolished in rpoE, rpoH or hfq mutant strains. Attempted complementation with the rpoE gene cloned on a plasmid restored katE–lacZ expression in an rpoE mutant but not an hfq mutant strain (Fig. 5B). rpoE overexpression from a multicopy plasmid was able to partially restore katE–lacZ expression in an rpoH mutant, suggesting that σE may also influence the expression of σS-regulated genes by a σH-independent mechanism.

Maximal katE expression in Salmonella-infected macrophages requires σE

RpoS translation rapidly increases in S. Typhimurium immediately after phagocytosis by macrophages (Chen et al., 1996), but the mechanism of RpoS induction is unclear. To analyse the effect of σE on RpoS expression during phagocytosis, we measured katE mRNA in RAW 264.7 murine macrophages infected with wild-type, rpoE or rpoS mutant Salmonella grown to log phase or stationary phase in LB medium. Quantitative real-time RT-PCR was employed as a sensitive method to monitor Salmonella gene transcription inside RAW cells. During the initial 40 min after uptake by RAW cells, levels of katE mRNA rapidly increased in both log-phase and stationary-phase S. Typhimurium (Fig. 6). An rpoE mutation substantially reduced katE mRNA levels, although to a lesser extent than an rpoS mutation, demonstrating that the σE–σS regulatory interaction is relevant during Salmonella–phagocyte interactions. Compared with log-phase cells, stationary-phase Salmonella cells exhibited an approximately sixfold increase in katE expression inside RAW cells, suggesting that σE activation in stationary phase can further enhance σS expression after phagocytosis.

Figure 6.

Effect of σE on σS-dependent gene expression in RAW 264.7 macrophages. Salmonella cells grown to log phase (left) or stationary phase (right) were opsonized and phagocytosed by RAW 264.7 macrophages. Salmonella were isolated 40 min after infection and subjected to RNA isolation and quantitative real-time RT-PCR. Values represent fold increase calculated as a ratio of katE mRNA levels before and after infection of macrophages. The data are representative of three independent experiments.


The ability to withstand oxidative stress is a critical element of Salmonella physiology (Foster and Spector, 1995) that allows this pathogenic bacterium to resist the anti-microbial actions of phagocyte-derived reactive oxygen species (ROS) (Mastroeni et al., 2000; Vazquez-Torres et al., 2000). Many Salmonella antioxidant defences are regulated by the stationary-phase sigma factor σS (Spector, 1998). Therefore, σS is appreciated to play a central role in the survival of Salmonella during oxidative stress, and σS-deficient mutants are profoundly hypersusceptible to a hydrogen peroxide challenge (Fig. 1). Other alternative sigma factors, specifically σE and σH, have also been implicated in the resistance of enteric bacteria to oxidative stress (Jenkins et al., 1991; Testerman et al., 2002). In the present study, we demonstrate that this largely results from the ability of σE and σH to enhance σS expression.

Proteolytic turnover or inactivation of cognate anti-sigma factors govern the expression or activity of alternative sigma factors in bacterial cells (Hughes and Mathee, 1998). Environmental signals or metabolic abnormalities regulate the activity of anti-sigma factors (Fig. 7): RseA in the case of σE (Missiakas et al., 1997; Walsh et al., 2003), the molecular chaperones DnaJ and DnaK in the case of σH (Hughes and Mathee, 1998), and RssB (also called MviA/SprE) in the case of σS (Hengge-Aronis, 2002). Overproduction of outer membrane proteins induces RseA proteolysis by inner membrane proteases DegS and YaeL, with subsequent release of σE (Walsh et al., 2003). Although the mechanism has not yet been determined, σE levels also increase upon entry into stationary phase. It has been proposed that misfolding of periplasmic proteins or perturbation of the extracytoplasmic redox environment during stationary phase might trigger proteolysis or destabilization of RseA (Testerman et al., 2002). A study in E. coli has suggested an alternative regulatory mechanism for controlling the activity of RseA and σH by a putative tyrosine kinase (Etk; YccY) and phosphatase (Etp; YccC) in response to high temperature (Klein et al., 2003). However, YccY/C homologues are absent from any sequenced Salmonella genome, suggesting some differences in sigma factor regulation between E. coli and S. Typhimurium. σH remains inactive in complexes with DnaK and DnaJ in the cytoplasm unless thermally denatured proteins accumulating in the cytoplasm recruit DnaK from σH (Hughes and Mathee, 1998); activation of the heat shock regulon by oxidative stress (Dukan and Nystrom, 1998; Estruch, 2000) suggests that stationary-phase or oxidative stress may also modulate σH via DnaJ and DnaK. As a chaperonin, DnaK can also protect σS from ClpXP-mediated proteolysis during heat shock or carbon starvation (Muffler et al., 1997a), suggesting a pivotal role of DnaK in regulating levels of σH and σS. Turnover of σS by the ClpXP protease is faciliated by the RssB response regulator (Hengge-Aronis, 2002), but the mechanism is uncertain (Peterson et al., 2004).

Figure 7.

Model of SalmonellaσS regulon modulation by σE and σH. Activation of σE resulting from proteolysis of RseA leads to enhanced rpoH transcription. σH in turn leads to transcription of hfq, and Hfq (HF-1) protein promotes σS translation. In this manner, signals leading to activation of σE or σH can augment expression of the σS stress regulon.

Expression of an alternative sigma factor reflects a net balance between production and turnover. As the σH protein is highly labile (t1/2 = 1–8 min), probably as a result of degradation by the FtsH protease (Bertani et al., 2001), continuous synthesis of σH is required for bacterial survival (Straus et al., 1987). The current study demonstrates that σE is required for rpoH transcription in stationary phase (Fig. 3), presumably acting at the P3 rpoH promoter (Erickson and Gross, 1989; Nakahigashi et al., 1995; Ramirez-Santos et al., 2001).

Studies in E. coli have demonstrated that hfq is also transcribed from multiple promoters including two transcribed by σD and one by σH (Tsui et al., 1996; Muffler et al., 1997b). Our observations indicate that hfq transcription in Salmonella is dependent on σE and σH during stationary phase (Fig. 3) and that this regulatory link is required for σS synthesis (Figs 3 and 4). Moreover, the loss of thermotolerance conferred by mutations in rpoE, rpoS or hfq (Rouviere et al., 1995; Muffler et al., 1997b) is consistent with a regulatory cascade involving σE, σH, Hfq and σS. The central importance of Hfq for σS regulation is underscored by the discovery of several small non-coding RNAs (e.g. DsrA, RprA, OxyS) that fine-tune RpoS translation in an Hfq-dependent manner (Repoila et al., 2003). The large number of small regulatory RNAs interacting with Hfq (Zhang et al., 2003) suggests that there may be considerable competition for Hfq binding. The increase in katE transcription beyond wild-type levels as a result of hfq overexpression (Fig. 4A, right) confirms that Hfq availability is a critical factor in the modulation of σS expression.

Many bacterial genes, like rpoH and hfq, possess multiple promoters (Collado-Vides et al., 1991) that allow their transcription in response to a variety of stimuli. Furthermore, many sigma factor genes contain promoters recognized by the sigma factors themselves. SigmaM, W and X of Bacillus subtilis, sigmaE of Rhodobacter sphaeroides, sigmaE of Pseudomonas putida and sigmaE of E. coli and Salmonella can each recognize promoters upstream of their own genes (Newman et al., 1999; Dartigalongue et al., 2001; Asai et al., 2003). A positive feedback loop involving σE is likely to be physiologically relevant, because overexpression of σEin trans can dramatically increase expression of the σE-dependent rpoE P3 promoter (data not shown). Autoinduction of sigma factors in both Gram-positive and Gram-negative bacteria may have evolved to allow the rapid acceleration of sigma factor synthesis upon sensing of the appropriate environmental triggers (Missiakas and Raina, 1998).

The present study along with previous observations by other groups suggests a model (Fig. 7) in which extracytoplasmic signals to σE and cytoplasmic signals acting on σH can lead to stimulation of σS synthesis. Release of σE into the cytoplasm by proteolysis of the anti-sigma factor RseA leads to increased transcription of σE-activated genes including rpoE itself and rpoH. Newly synthesized σH can lead to increased hfq transcription, and Hfq facilitates rpoS translation, ultimately resulting in increased levels of σS and enhanced production of antioxidant enzymes such as catalase hydroperoxidase, superoxide dismutase and glutaredoxin, and the ferritin-like Dps protein.

In summary, we have demonstrated that σE and σH are required for the resistance of Salmonella to oxidative stress during stationary phase in a manner that appears to be epistatic to σS. A regulatory cascade involving σE, σH, Hfq and σS is proposed to control the expression of the σS stress regulon. This pathway appears to be functionally important, as the loss of rpoE, rpoH or hfq results in a significant reduction in the expression of σS-activated genes both during stationary phase in vitro and within Salmonella-infected macrophages. Recent studies of functional genomics have provided new insights into regulatory networks of bacterial gene expression and provide evidence of substantial overlap between regulons (Martinez-Antonio and Collado-Vides, 2003). Interactions between alternative sigma factors provide a means by which bacteria can integrate diverse environmental signals to trigger and fine-tune a common protective stress response.

Experimental procedures

Bacterial strains and culture conditions

Bacterial strains and plasmids used in this study are listed in Table 3. DNA oligonucleotides are listed in Table 4. Deletion mutations were constructed in S. Typhimurium by the λRed recombinase method (Datsenko and Wanner, 2000). An rpoH deletion mutation was constructed by using primers ♯131 and ♯132, and the desired mutation was confirmed by PCR amplification with primers ♯58 and ♯237 followed by DNA sequencing. To construct an rseA deletion, primers ♯213 and ♯214 were used for mutagenesis, and primers ♯215 and ♯216 were used to confirm the mutation by PCR and DNA sequencing.

Table 3. Strains and plasmids.
  1. [O], operon (transcriptional) fusion; [Pr], protein (translational) fusion.

S. Typhimurium
 IB114028S wild typeATCC
 IB2 rpoE::CmIB1 X HT(TF951)
 IB21 hfq::MudCmIB1 X HT (TE6266)
 IB39/pRS1274::P3rpoElacZIB1 pTFP2
 IB43 rpoS::Tn10dCmIB1 X HT (TE6133)
 IB66 putPA1303::Kan R -katE-lac[O]IB1 X HT (TE6153)
 IB67 putPA1303::Kan R -katE-lac[O]rpoE::CmIB66 X HT (IB2)
 IB130/pRS1274::hfq–lacZIB1 X pFB115
 IB131 rpoE::Cm/pRS1274::hfq–lacZIB2 X pFB115
 IB136 rpoS1307::MudJ [O]IB1 X HT (JF2938)
 IB137 rpoE::Cm rpoS1307::MudJ [O]IB2 X HT (JF2938)
 IB138 putPA1303::Kan R -katE-lac[O]hfq::MudCmIB66 X HT (TE6266)
 IB139 putPA1303::Kan R -katE-lac[O]rpoS::Tn10dCmIB66 X HT (TE6133)
 IB152/pRS1274::rpoH–lacZIB1 X pFB55
 IB153 rpoE::Cm/pRS1274::rpoH–lacZIB2 X pFB55
 IB162 rpoE::Cm/pRS1274::dnaK–lacZ 
 IB208 rpoE::Cm/pRB983IB2 X pRB983
 IB237 putPA1303::Kan R -katE-lac[O]/pRB983IB66 X pRB983
 IB239 putPA1303::Kan R -katE-lac[O]rpoE::Cm/pRB983IB67 X pRB983
 IB240 putPA1303::Kan R -katE-lac[O]hfq::MudCm/ pRB983IB138 X pRB983
 IB272 ΔrpoH::Km/pRS1274::hfq–lacZIB130 X HT (λRed/ΔrpoH::Km)
 IB314 ΔrpoH::APIB1 X HT (λRed/ΔrpoH::AP)
 IB316 Δhfq IB1 X HT (λRed/Δhfq::Cm) Cm gene was deleted by FLP
 IB318 Δhfq putPA1303::Kan R -katE-lac[O]IB316 X HT (IB66)
 IB320 putPA1303::Kan R -katE-lac[O]ΔrpoH::APIB66 X HT (IB314)
 IB332 ΔrseA::Cm/pRS1274::P3rpoElacZIB39 X HT (λRed/ΔrseA::Cm)
 IB339 ΔrseA::Cm putPA1303::KanR-katE-lac[O]IB66 X (λRed/ΔrseA::Cm)
 IB363 ΔrseA::Cm ΔrpoH::AP putPA1303::KanR-katE-lac[O]IB339 X HT (IB314)
 IB375 ΔrseA::Cm hfq::MudCm putPA1303::KanR-katE-lac[O]IB318 X HT (λRed/ΔrseA::Cm)
 IB393/pRS1274:P1rpoElacZIB1 X pTFP1
 IB420 rpoS::MudJ ΔrpoH::APIB136 X HT (IB314)
 IB436 rpoE::Cm ΔrpoH::APIB2 X HT (IB222)
 IB438 rpoS::MudJ rpoE::Cm ΔrpoH::APIB420 X HT (IB2)
 IB444 putPA1303::Kan R -katE-lac[O]ΔrpoH::AP/ pRB983IB320 X pRB983
 IB466/pBAD::ompCIB1 X pFB175
 IB467 rpoE::Cm/pBAD::ompCIB2 X pFB175
 IB488 rpoE::Cm Δhfq putPA1303::KanR-katE-lac[O]IB318 X HT (IB2)
 IB489 rpoE::Cm ΔrpoH::AP putPA1303::KanR-katE-lac[O]IB320 X HT (IB2)
 IB495 rpoE::Cm Δhfq putPA1303::KanR-katE-lac[O]/pFB125IB488 X pFB125
 IB501 ΔrseA::Cm/pRS1274::P1rpoElacZIB393 X HT (λRed/ΔrseA::Cm)
 IB503 rpoE::Cm rpoS::AP putPA1303::KanR-katE-lac[O]IB67 X HT (JF2690)
 IB562/pRS1274::rpoHσE −10 mutationlacZIB1 XpFB186
 IB564 rpoE::Cm/pRS1274::rpoHσE −10 mutationlacZIB2 XpFB186
 IB566/pRS1274::hfqσH −10 mutationlacZIB1 XpFB187
 IB567 rpoE::Cm/pRS1274::hfqσH −10 mutationlacZIB2 XpFB187
 TF95114028S rpoE::Cm Testerman et al. (2002)
 TE6266LT2 putPA1303::KanR-rpoS-lac[Pr]hfq::MudCm Brown and Elliott (1996)
 TE6133 rpoS::Tn10dCm Brown and Elliott (1996)
 TE6153LT2 putPA1303::KanR-katE-lac[O] Brown and Elliott (1996)
 JF2938UK1 rpoS1307::MudJ [O] Webb et al. (1999)
 JF2690UK1 rpoS::AP Webb et al. (1999)
E. coli
 MC1061Wild type 
 PND818 rpoE::Cm Danese et al. (1995)
 pBAD18ParaBAD expression vector Guzman et al. (1995)
 pBAD30ParaBAD expression vector 
 pFB55pRS1274::rpoH–lacZThis study
 pFB115pRS1274::hfq–lacZThis study
 pFB125pBAD18-hfqThis study
 pFB172pRS1274::dnaK–lacZThis study
 pFB175pBAD30::ompCThis study
 pFB186pRS1274::rpoHσE −10 mutationlacZThis study
 pFB187pRS1274::hfqσH −10 mutationlacZThis study
 pRB983pRB3–rpoE Testerman et al. (2002)
 pRS1274Transcriptional lacZ fusion vector Simons et al. (1987)
 pTFP1pRS1274::P1rpoElacZ Testerman et al. (2002)
 pTFP2pRS1274::P3rpoElacZ Testerman et al. (2002)
Table 4. DNA oligonucleotide primers.
Primer No.Sequences

To construct an hfq complementing plasmid (pFB125), the hfq gene was amplified from S. Typhimurium 14028s with oligonucleotides ♯28 and ♯112. The PCR product was cloned into pCR2.1 (Invitrogen), and the XbaI–HindIII fragment was subcloned into pBAD18.

General bacterial culture conditions were as previously described (Testerman et al., 2002). For the nutrient limitation experiment, cells grown overnight in LB were subcultured in 5 ml of fresh LB at a 1:1000 dilution and reincubated without agitation to early exponential growth phase (OD600∼0.3). Cells were subsequently harvested, washed twice in 1× M9 salts, resuspended in M9 minimal medium containing 0.002% glucose and reincubated with shaking.

Hydrogen peroxide survival assay

Salmonella grown overnight in LB broth was challenged with 4 mM H2O2 or 1 mM H2O2 at a density of 105 colony forming units (cfu) per millilitre in M9 minimal medium without glucose at 37°C. Viable counts were measured by immediate dilution in phosphate-buffered saline (PBS) and plating on LB agar at timed intervals. Percentage survival was calculated by dividing cfu at time points after challenge by cfu before challenge and multiplying by 100.

Genetic techniques and microarray analysis

Standard molecular biological techniques were performed according to protocols described by Ausubel et al. (1999). Protocols for Salmonella genetic manipulations were as described previously (Bang et al., 2000). For S. Typhimurium microarray analysis, total RNA was isolated from 10 ml of overnight cultures in LB broth by a published method (Bang et al., 2000), and 50 µg of total RNA was used as a template for cDNA synthesis. Microarray analysis was performed as described previously (Bader et al., 2003; Porwollik et al., 2003). Transcriptional profiles are provided in TableS1 as supplemental information. E. coli microarray analysis was performed using the Panorama E. coli K-12 gene array system (Sigma-Genosys, The Woodlands, TX). Total RNA was isolated as described above, and cDNA was synthesized with AMV (avian myeloblastosis virus) reverse transcriptase (Sigma-Genosys), and subsequently labelled with [α-33P]-CTP using DNA polymerase and E. coli-optimized primers (Sigma-Genosys). After unincorporated nucleotides were removed with a G-25 Sephadex column, the cDNAs were hybridized to the Panorama E. coli K-12 gene array according to protocol. The array was analysed on a Storm Imager (Molecular Dynamics, Piscataway, NJ) with comparison of spot intensities between wild-type and rpoE mutant strains.

Quantitative real-time RT-PCR assays

For real-time RT-PCR assays, reaction and statistical interpretation of quantitative parameters were carried out by methods described previously (Bader et al., 2003). Briefly, S. Typhimurium cells were collected from 500 µl of LB cultures grown overnight, and total RNA was prepared by using an RNeasy Mini kit (Qiagen). For each reaction, 7.8 µl of diluted RNA sample (5 µg ml−1) were mixed with 2 µl of primer sets (1 µM each) containing forward and reverse primers for each gene. Real-time RT-PCR was initiated by adding this mixture to 10 µl of SYBR Green mix containing 0.2 µl reverse transcriptase mix. gyrB encoding subunit B of DNA gyrase was used as a control for normalization of gene expression and to obtain a standard curve using control RNA from wild-type cells. Each sample was independently tested three times and assayed in triplicate during each run. Reported values represent the mean value from each assay. Primer pairs used in this assay were ♯155 and ♯156 for katE, ♯209 and ♯210 for gyrB, and ♯254 and ♯255 for rpoH (Table 3).

Immunoblot analysis

σS levels were measured by immunoblot analysis. A murine monoclonal IgG1 antibody to σS was purchased from Neoclone (Madison, WI). Immunoblotting was performed essentially as described in Bang et al. (2000). Equivalent quantities of protein were added to each lane. Data presented are representative of three independent experiments.

Construction of transcriptional lacZ fusions and β-galactosidase assays

To construct an rpoH–lacZ transcriptional fusion, a 345 bp PCR product of oligonucleotides ♯58 and ♯59 spanning from 245 bp upstream to 100 bp downstream of the rpoH translational start site was cloned into the pRS1274 operon fusion vector (Simons et al., 1987) restricted with SmaI. To confirm the desired orientation, a second PCR was performed using oligonucleotide ♯79 from lacZ and oligonucleotide ♯59. For the hfq–lacZ fusion, oligonucleotides ♯71 and ♯72 were used to amplify a PCR product spanning from 1040 bp upstream of the hfq translational start site, which was cloned as described for the rpoH–lacZ fusion. For the dnaK–lacZ fusion, oligonucleotides ♯87 and ♯88 were used to obtain a PCR product containing the σH promoter region. For lacZ transcriptional fusions containing mutated promoters, two-step PCR was performed. First, DNA fragments containing a mutated −10 element at the 3′ end were amplified with primers ♯374/♯375 for the rpoH promoter and ♯371/♯372 for the hfq promoter respectively. Corresponding DNA fragments containing the same mutated −10 element at the 5′ end were amplified with primers ♯376/♯377 for the rpoH promoter and ♯373/♯72 for the hfq promoter respectively. The DNA fragments were ligated and used as a template for full-length PCR using primers ♯374/♯377 for the rpoH promoter and ♯371/♯72 for the hfq promoter. The final products were directionally cloned into pRS1274. To decrease the effect of plasmid copy number in studies using the expression vector pRS1274, a pcnB mutation to decrease copy number to 10% of wild-type levels (Podkovyrov and Larson, 1995) was introduced for the studies shown in Table 2.

β-Galactosidase assays were performed as described by Miller (1992) and expressed as nmoles per minute per optical density unit (600 nm). To show the relative activity expressed by mutant strains, per cent activity relative to wild type is displayed in some figures.

Measurement of Salmonella gene expression in macrophages

RAW 264.7 murine macrophages were seeded at a density of 2 × 105 cells in six-well culture plates and infected with complement-opsonized S. Typhimurium as described (Vazquez-Torres et al., 2000). S. Typhimurium cells were grown to log phase (OD600 = 0.3–0.5) or stationary phase (OD600 = 2.5–3.0) in LB at 37°C, then diluted to an OD600 of 0.1 in pre-warmed RPMI medium with 10% normal mouse serum for 20 min at 37°C. To synchronize infection, plates were centrifuged at 1000 g for 5 min (defined as the starting time point for gene expression analysis). Fifteen minutes after phagocytosis, extracellular bacteria were killed with 30 µg ml−1 gentamicin. The incubation was continued for an additional 20 min before macrophages were lysed with 1% Triton X-100 (v/v) and bacterial RNA stabilized with the RNAprotect bacteria reagent (Qiagen). As an extracellular control, opsonized bacteria were added to RAW cells [10:1 multiplicity of infection (moi)] and incubated as described above with elimination of the centrifugation step. RNA purification and measurement of gene expression using quantitative real-time RT-PCR were performed by the methods described above. To ensure that comparable bacterial numbers were recovered from macrophages at each time point, aliquots of 1% Triton X-100 lysates were used to determine the number of viable bacteria by plating of serial dilutions. The data presented are representative of three independent experiments.

Supplementary material

The following material is available from

Table S1. Effects of mutations in rpoE and rpoS on S. Typhimurium gene transcription in stationary-phase culture.


We thank C. Gross for helpful comments and suggestions; T. Elliott, J. Foster and T. Silhavy for generous gifts of strains; S. Porwollik for printing microarray slides and assisting with analysis; and W. Navarre and S. Libby for technical advice on microarray experiments. This work was supported by Grants AI44486 (F.C.F.) and AI34829 (M.M. and J.G.F.) from the National Institutes of Health.