Salmonella enterica Serovar Typhimurium Periplasmic Superoxide Dismutase SodCI Is a Member of the PhoPQ Regulon and Is Induced in Macrophages

Abstract

Salmonella enterica serovar Typhimurium replicates within host macrophages during the systemic stage of infection. In the macrophage, the bacteria must survive the respiratory burst that produces superoxide. Serovar Typhimurium strain 14028 produces two periplasmic superoxide dismutases, SodCI and SodCII, but only SodCI contributes to virulence. Although we have shown that this is primarily due to differences in the two proteins, evidence suggests differential regulation of the two genes. Using transcriptional sodCI- and sodCII-lac fusions, we show that sodCII is under the control of the RpoS sigma factor, as was known for the Escherichia coli ortholog, sodC. In contrast, we show that sodCI is transcriptionally controlled by the PhoPQ two-component regulatory system, which regulates an array of virulence genes required for macrophage survival. Introduction of a phoP-null mutation into the sodCI fusion strain resulted in a decrease in transcription and loss of regulation. The sodCI-lac fusion showed high-level expression in a background containing a phoQ constitutive allele. The sodCI gene is induced 15-fold in bacteria recovered from either the tissue culture macrophages or the spleens of infected mice. Induction in macrophages is dependent on PhoP. The sodCII fusion was induced three- to fourfold in macrophages and animals; this induction was unaffected by loss of PhoP. Thus, sodCI, which is horizontally transferred by the Gifsy-2 phage, is regulated by PhoPQ such that it is induced at the appropriate time and place to combat phagocytic superoxide.

Salmonella enterica serovar Typhimurium is an enteric pathogenic bacterium capable of infecting and causing disease in humans and animals. As an intracellular pathogen, serovar Typhimurium encounters a variety of host defense mechanisms and must adapt to different conditions within the host organism. An important aspect of serovar Typhimurium pathogenesis is survival within host macrophages (12), which is partly dependent on the ability of the bacterium to protect itself from the phagocytic respiratory burst that generates reactive oxygen species such as superoxide (7, 11).

Serovar Typhimurium strain 14028 produces two periplasmic Cu/Zn superoxide dismutases, SodCI and SodCII. SodCI is encoded on the functional lambdoid bacteriophage Gifsy-2, embedded within but transcribed in the opposite orientation to the late phage operon (10, 13, 22). SodCII is encoded on the chromosome and is the ortholog of Escherichia coli SodC. Strains containing a sodCI-null mutation are attenuated in mouse time-to-death assays (7, 11). In intraperitoneal competition assays against the isogenic wild-type strain, sodCI mutants show 7- to 10-fold attenuation (23, 27). In contrast, sodCII-null mutations in strain 14028 do not confer a virulence phenotype, even in the absence of SodCI (27, 45).

Krishnakumar et al. (27) provided evidence that both SodCI and SodCII are expressed during infection and inherent differences in the two proteins primarily explain the differential role in virulence. However, data also suggest that the two genes are differentially regulated. Uzzau et al. (45) showed that epitope tagged SodCI accumulated to significantly higher levels than epitope tagged SodCII in bacteria recovered from macrophages and animals. Eriksson et al. (9) noted transcriptional induction of sodCI but not sodCII in microarray analysis of serovar Typhimurium grown in tissue culture macrophages. We noted that producing SodCII under the control of the sodCI promoter not only failed to complement a sodCI virulence defect but actually attenuated the bacterium further. One explanation for this phenomenon was that the nonfunctional SodCII was being overproduced (27).

A number of enzymes involved in defense to oxidative stress are RpoS regulated, and E. coli sodC is a member of this regulon (15). Serovar Typhimurium sodCII was also shown to be controlled by RpoS (10). For sodCI, stationary-phase regulation had been reported, but it was apparently independent of RpoS (10).

PhoPQ is a two-component regulatory system that is critical for serovar Typhimurium adaptation to intracellular growth (16). It consists of a sensor kinase, PhoQ, which phosphorylates PhoP, the response regulator. PhoP then binds to the promoters of target genes resulting in activation or repression (14, 17, 35, 53). In vitro, the system is activated by low cation concentrations (14) and low pH (3). However, more recent evidence suggests that low pH (31) and direct detection of antimicrobial peptides produced by macrophages (1, 2) are the critical signals in the Salmonella-containing vacuole responsible for inducing PhoP phosphorylation. Several PhoPQ-regulated genes encode regulatory proteins that directly or indirectly control expression of a subset of the PhoPQ regulon. The PmrAB two-component regulatory system controls genes responsible for lipopolysaccharide (LPS) modification and resistance to certain antimicrobial peptides (19, 26). This system is induced by PhoPQ via pmrD (26). PmrAB can also be activated independently of PhoPQ by high-iron conditions or by mildly acidic pH (51). PhoPQ is also known to regulate rstAB, encoding a two-component system of unknown function (37, 52); ssrAB, the two-component system that controls expression of the Salmonella pathogenicity island 2 (SPI2) type three secretion system (4); and slyA, which controls a large stress response regulon (38, 39, 41).

We show here that sodCI and sodCII are differentially regulated. The sodCI gene is regulated by the PhoPQ two-component regulatory system in vitro and in vivo. In contrast, sodCII is under the control of RpoS. Both genes are induced in bacteria recovered from tissue culture macrophages or the spleens of infected mice, with sodCI expressed at a higher level than sodCII.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

All Salmonella strains used in the present study are isogenic derivatives of Salmonella enterica serovar Typhimurium 14028 (American Type Culture Collection) and were constructed by using P22 HT105/1 int-201 (P22)-mediated transduction (30). Luria-Bertani (LB) medium and modified N-minimal medium (18) were used for the growth of bacteria. Bacterial strains were grown at 37°C except for the strains containing the temperature-sensitive plasmids pCP20, pKD46, and pINT-ts (CRIM) (6, 20), which were grown at 30°C. Antibiotics were used at the following concentrations: 100 μg of ampicillin/ml, 20 μg of chloramphenicol (Cm)/ml, 12.5 μg of gentamicin/ml, 50 μg of kanamycin/ml, 25 μg of tetracycline/ml, and 50 μg of apramycin/ml. The β-galactosidase chromogenic indicator X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) was used at a concentration of 80 μg/ml. Enzymes were purchased from Invitrogen or New England Biolabs and used according to the manufacturer's recommendations. Primers were purchased from IDT, Inc.

Deletion of various genes and concomitant insertion of an antibiotic resistance cassette was carried out by using Lambda Red-mediated recombination (6, 54) as described previously (8). The endpoints of each deletion are indicated in Table 1. The appropriate insertion of the antibiotic resistance marker was checked by P22 linkage to known markers and/or PCR analysis. In each case, the constructs resulting from this procedure were moved into a clean wild-type background (strain 14028) by P22 transduction. In some cases, antibiotic resistance cassettes were removed by using the temperature sensitive plasmid pCP20 carrying the FLP recombinase (5).

TABLE 1.

S. enterica serovar Typhimurium strains used in this study

Strain or plasmid	Genotypea	Deletion or cloned endpointsb	Source or referencec
Strains
14028	Wild type		ATCCd
JS538	Φ(sodCI+-lac+)2917	1098178-1130040
JS531	Φ(sodCII+-lac+)110	1516586-1516598
JS539	ΔrpoS1191::Tet	3065506-3066484
JS540	Φ(sodCI+-lac+)2917 ΔrpoS1191::tet
JS541	Φ(sodCII+-lac+)110 ΔrpoS1191::tet
JS542	phoQ24		36
JS543	Φ(sodCI+-lac+)2917 phoQ24
JS544	Φ(sodCII+-lac+)110 phoQ24
JS545	phoP102::Tn10d-Cm		36
JS546	Φ(sodCI+-lac+)2917 phoP102::Tn10d-Cm
JS547	Φ(sodCII+-lac+)110 phoP102::Tn10d-Cm
JS548	ΔpmrA621::Cm	4533696-4534407
JS549	Φ(sodCI+-lac+)2917 ΔpmrA621::Cm
JS550	Φ(sodCI+-lac+)2917 phoQ24 ΔpmrA621::Cm
JSG1050	pmrI::MudJ		J. S. Gunn
JS551	pmrI::MudJ ΔpmrA621::Cm
JS552	ΔrstA1::Cm	1551266-1552007
JS553	Φ(sodCI+-lac+)2917 ΔrstA1::Cm
JS560	Φ(sodCI+-lac+)2917 phoQ24 ΔrstA1::Cm
JS554	ΔssrB101::Cm	1476102-1476851
JS555	Φ(sodCI+-lac+)2917 ΔssrB101::Cm
JS561	Φ(sodCI+-lac+)2917 phoQ24 ΔssrB101::Cm
Plasmids
pDX1
pKG101	pDX1 sodCI fragment E	1130675-1130745
pKG102	pDX1 sodCI fragment D	1130635-1130745
pKG103	pDX1 sodCI fragment A	1130589-1130745
pKG104	pDX1 sodCI fragment C	1130635-1130689
pKG105	pDX1 sodCI fragment B	1130589-1130689
pKG112	pDX1 sodCI fragment F	Mutant fragment C (see Fig. 8)
pKG113	pDX1 sodCI fragment G	Mutant fragment C (see Fig. 8)
pKG114	pDX1 sodCI fragment H	Mutant fragment C (see Fig. 8)
pKD46	bla PBADgam bet exo pSC101 oriTS		6
pCP20	bla cat cI857 λPRflp pSC101 oriTS		5
pCE70	oriR6K FRT-tnpRwt-lacZY aph (Kanr)		33
pINT-ts	Intλ		20
pSC2-phoP	reppMB1 Apr T7 6His-phoP		24

^aUnless otherwise noted, all Salmonella strains are isogenic derivatives of serovar Typhimurium strain 14028.

^bNumbers indicate the base pairs that are deleted or cloned (inclusive) as defined in the S. enterica serovar Typhimurium LT2 genome sequence in the National Center for Biotechnology Information Database NC_003197.

^cThis study, unless specified otherwise.

^dATCC, American Type Culture Collection.

Construction of the transcriptional lac fusions.

Transcriptional lac fusions were generated from the above constructs by using pCE70 (33) as described by Ellermeier et al. (8). The fusion joints are indicated in Table 1. In order to prevent the possible amplification of the fusion construct in the chromosome due to phage induction, with the resulting artifactual increase in lac activity, the sodCI fusion was made such that the Gifsy-2 phage sequences downstream of sodCI were deleted through the attachment site (see reference 22). To ensure that the strains were not attenuated in vivo, the fusions were positioned immediately downstream of the sodCI and sodCII coding region. The resulting fusion strains competed evenly with the wild-type strain in intraperitoneal competition assays performed in BALB/c mice (data not shown; see reference 22).

β-Galactosidase assays in vitro.

β-Galactosidase assays were performed by using a microtiter plate assay as previously described (43) on strains grown under the indicated conditions. β-Galactosidase activity units are defined as (μmol of orthonitrophenol [ONP] formed min⁻¹) × 10⁶/(optical density at 600 nm [OD₆₀₀] × ml of cell suspension) and are reported as mean ± the standard deviation where n = 4. For log-phase cultures, bacteria were grown overnight in LB medium, diluted 1/100 in the indicated medium and upon reaching an OD₆₀₀ of 0.2, diluted 1/4, and grown to OD₆₀₀ of 0.2 to 0.3. Strains were assayed at the same OD₆₀₀.

β-Galactosidase assays in bacteria recovered from infected macrophages.

RAW 264.7 macrophages (American Type Culture Collection) were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% l-glutamine (BioWhittaker). Macrophages were seeded into six-well plates at 1 × 10⁶ to 5 × 10⁶ cells/well. Bacteria were grown overnight in modified N-minimal medium (pH 5.6) supplemented with 10 μM magnesium chloride. Bacterial cells were washed with sterile phosphate-buffered saline (PBS) and opsonized in 50% mouse serum for 20 min at 37°C. Bacteria were then diluted in RPMI medium and used to infect macrophages at a multiplicity of infection of 20. After a 40-min incubation period, the wells were washed twice with PBS and then RPMI medium containing 12.5 μg of gentamicin/ml to kill extracellular bacteria was added to the wells. After 20 min of additional incubation, the wells were washed once with PBS. RPMI medium supplemented with 12.5 μg of gentamicin/ml was added to the wells, and this was designated time zero. After 16 h, the wells were washed twice with PBS and macrophages were lysed with 1% Triton X-100. The released bacteria were washed with PBS, and dilutions of each sample were plated on LB agar to determine the number of bacteria. β-Galactosidase activity was assayed by using the chemiluminescent substrate Lumigal 530 according to manufacturer's instructions (Lumigen, Inc.). The β-galactosidase activity of each sample was calculated per CFU of bacteria in a sample. The in vitro β-galactosidase activity of bacteria grown in RPMI medium (16 h of growth) was measured by using the same assay.

β-Galactosidase assays in bacteria recovered from infected mouse splenic tissue.

BALB/c mice were infected intraperitoneally with 10⁴ cells of the sodCI⁺-lac⁺ or sodCII⁺-lac⁺ fusion strain. After 4 days of infection, mice were sacrificed, and the spleens were homogenized to release the bacteria. Bacterial cells were extracted from splenic tissue as described previously (42). Τhe isolated bacteria were resuspended in PBS, and dilutions of each sample were plated on LB agar to determine the number of viable bacteria in a sample. β-Galactosidase activity was assayed using the chemiluminescent substrate Lumigal 530 according to the manufacturer's instructions (Lumigen, Inc.). The β-galactosidase activity of each sample was calculated per CFU of bacteria in a sample. The in vitro β-galactosidase activity of bacteria in the inoculum grown in LB medium (16 h of growth) was measured by using the same assay.

Primer extension analysis.

Bacterial cells were grown in LB medium to an OD₆₀₀ of ∼1.0, and the total RNA was isolated according to the manufacturer's instructions (QIAGEN RNEasy Mini). Omniscript Reverse Transcriptase (QIAGEN, Inc.) was used to reverse transcribe the sodCI using extension primer sodCIext3 (CCAGCTACCAGCGACAATATTGTG). The primer was radioactively labeled with [γ-³²P]ATP by using Optikinase (USB Corp.). A DNA fragment of 400 bp corresponding to the sequence immediately upstream of sodCI coding region was amplified by PCR using Platinum Pfx DNA polymerase (Invitrogen, Inc.). The resulting PCR product was used to generate a sequence ladder with the same primer that was used for the extension reaction. The extension product was resolved on an 8% polyacrylamide-7 M urea gel alongside the sequence ladders. The Sequenase 2.0 kit used for the sequencing reactions, polyacrylamide gel mix, and GTG running buffer were purchased from the USB Corp.

Deletion analysis of the sodCI promoter.

The sodCI promoter was analyzed by amplifying and cloning portions of the promoter region 5′ to a promoterless lacZ gene in pDX1, an apramycin-resistant plasmid derived from pAH125 (20; D. Lin and J. M. Slauch, unpublished data). All constructs were confirmed by DNA sequence analysis. The resulting fusion plasmids were integrated into the serovar Typhimurium chromosome at the Lambda attachment site using λInt produced from CRIM helper plasmid pINT-ts (20). Each integrant was tested by PCR to ensure that only a single copy of the plasmid was present (20).

RESULTS

The sodCI and sodCII genes are transcriptionally induced in stationary phase.

Previous studies (27, 45) have shown that SodCI, but not SodCII, contributes to serovar Typhimurium strain 14028 virulence in mice. These studies also suggested that sodCI is expressed at a higher level than sodCII in vivo, although our results (27) indicated that sodCII is expressed during infection. Both enzymes are produced in laboratory culture (27, 45) and induced in the stationary phase of growth (10, 27). We used single-copy chromosomal transcriptional lac fusions to sodCI and sodCII to study the expression of both genes in vitro and in vivo. We constructed our fusions by inserting promoterless lacZY genes (8) just downstream of the sodCI or sodCII stop codons. In the case of sodCI, this insertion is associated with a deletion of all Gifsy-2 phage genes downstream of sodCI through attL, including int, the immunity region, and the genes encoding the putative replication proteins. This ensured that phage induction did not artifactually increase lac activity. Thus, both fusion constructs produced SodCI or SodCII equivalent to that produced by the wild type (data not shown). Moreover, the strain containing the sodCI⁺-lac⁺ fusion and the associated deletion of the Gifsy-2 phage genes was fully virulent (data not shown), which is consistent with our previous data showing that the phage genes are not required for virulence and that SodCI is produced from the lysogenic phage (22).

To analyze expression, we first determined the β-galactosidase activity produced from the sodCI-lac and sodCII-lac fusion strains in different stages of growth in LB medium. Figure 1 shows that sodCI and sodCII genes are induced 3.75- and 5.3-fold in stationary phase, respectively. Genes that are induced in the stationary phase are often under the control of RpoS (21). Indeed, E. coli sodC (the sodCII ortholog) is known to be regulated by RpoS (15). Introduction of an rpoS deletion mutation into the fusion-bearing strains resulted in the decreased expression of sodCII. In contrast, sodCI expression was not affected by the loss of RpoS. These data are in agreement with the previous results of Fang et al. (10). The enzymatic activity of SodCI and SodCII in wild-type (wt) and rpoS backgrounds in vitro correlated with the lac fusion results (data not shown). These results show that sodCII is regulated by RpoS in vitro. In contrast, sodCI expression in vitro is apparently under the control of some other regulator.

FIG. 1.

β-Galactosidase activity of sodCI- and sodCII-lac fusion strains. (A) Cells were assayed in the log phase or stationary phase of growth. (B) Effect of ΔrpoS::Tet mutation on sodCI and sodCII expression in stationary phase of growth. β-Galactosidase activity units are defined as (μmol of ONP formed min⁻¹) × 10⁶/(OD₆₀₀ × ml of cell suspension) and are reported as means ± the standard deviation where number of measurements (n = 4).

Expression of sodCI is controlled by PhoPQ in vitro.

Microarray analysis suggested that sodCI was a member of the PhoPQ two-component regulon (38). To examine whether sodCI was under the control of PhoPQ in vitro, we monitored the expression of the sodCI-lac fusion in modified N-minimal medium (pH 5.6), supplemented with 10 mM or 10 μM MgCl₂; divalent cations serve as a signal for the PhoPQ system in vitro (2, 14, 40). As shown in Fig. 2, sodCI was induced twofold in low-Mg²⁺ conditions (10 μM). Introduction of a phoP-null mutation abolished this induction. Introduction of the phoQ24 allele (25, 36, 49) resulted in high-level constitutive expression of sodCI. The expression of the sodCII-lac fusion was analyzed under the same conditions. In contrast to sodCI, the expression of the sodCII was not significantly affected by either the concentration of magnesium or introduction of the phoP or phoQ24 mutations (Fig. 2). These results indicate that sodCI is regulated by the PhoPQ two-component regulatory system in vitro, whereas sodCII is not.

FIG. 2.

Effect of phoP::Tn10dCm and phoQ24 mutations on the expression of sodCI and sodCII in vitro. Cultures were grown in N-min modified medium (pH 5.6) supplemented with 10 μM or 10 mM magnesium chloride (inducing and repressing conditions, respectively) to an OD₆₀₀ of 0.3. β-Galactosidase activity units are defined as (μmol of ONP formed min⁻¹) × 10⁶/(OD₆₀₀ × ml of cell suspension) and are reported as means ± the standard deviation (n = 4).

Transcription of sodCI is induced 10- to 15-fold in tissue culture macrophages and in mice.

Both sodCI and sodCII are expressed in the host (27), but several previous studies suggested that sodCI is expressed at a higher level than sodCII (9, 27, 45). In order to study the in vivo expression of both genes, the strains bearing sodCI⁺- and sodCII⁺-transcriptional lac fusions were used to infect RAW 264.7 macrophages. The β-galactosidase activity of bacteria isolated from macrophages was compared to that of bacteria grown in a laboratory culture. As seen in Fig. 3A, expression of sodCI was 10-fold induced in macrophages, whereas sodCII expression was induced 4-fold. To study the expression levels of both genes during infection, BALB/c mice were infected with sodCI⁺ or sodCII⁺ fusion strains, and bacteria were isolated from the spleen after 4 days of infection. The β-galactosidase activities of the bacteria isolated from mice and those grown in laboratory medium are shown in Fig. 3B. Strikingly similar to our macrophage expression data, sodCI was induced 15-fold in BALB/c mice compared to the laboratory culture, whereas sodCII was induced 3.5-fold. This correlation between the tissue culture macrophage and mouse data is consistent with the concept that the majority of bacteria are found within macrophages in the host (29). These data also suggest that our macrophage experiments are relevant for the study of the in vivo regulation of sodCI and sodCII.

FIG. 3.

In vivo expression of sodCI and sodCII. (A) Expression of the sodCI-lac and sodCII-lac fusions in cells recovered from RAW 264.7 macrophages or grown in RPMI medium. Expression of sodCI in macrophages was considered 100%. (B) Expression of the sodCI-lac and sodCII-lac fusions in cells isolated from BALB/c mice or grown in LB medium. Expression of sodCI in BALB/c mice was considered 100%. β-Galactosidase activity was determined by using a chemiluminescence assay.

PhoPQ regulates sodCI in vivo.

To test whether PhoPQ is responsible for sodCI induction in vivo, RAW 264.7 macrophages were infected with strains harboring the sodCI- or sodCII-transcriptional fusions in wt, phoP, or phoQ24 backgrounds. As above, expression of sodCI was induced 15-fold in macrophages (Fig. 4A). This induction was completely dependent on PhoP; introduction of the phoP-null mutation abolished sodCI expression in macrophages. Expression of sodCI was elevated 3-fold in tissue culture medium in the phoQ24 background compared to the wild type, and expression was induced an additional 22-fold in macrophages. This is likely due to the effect of lowering pH on the activation of the PhoQ24 protein; a similar effect was seen in vitro when we compared growth media at pH 7.4 versus pH 5.6 (data not shown). The sodCII fusion was induced two- to threefold in macrophages, and this induction was not affected in the presence of the phoP-null or phoQ24 mutations (Fig. 4A). These experiments are potentially complicated by the reduced viability of the phoP and phoQ mutants in macrophages (12, 35, 36). Indeed, we recovered approximately 3- to 10-fold less phoP or phoQ24 mutant compared to wild-type cells after 16 h of incubation. However, the fact that the specific activity of the sodCII fusion was not affected in the pho mutants shows that any survival defect was irrelevant to these measurements. These results show that PhoPQ regulates sodCI expression in vivo but does not affect sodCII expression, which correlates well with our in vitro data.

FIG. 4.

Effect of regulatory mutations on sodCI and sodCII expression in vivo. The activity of sodCI- and sodCII-lac fusions in phoP::Tn10dCm and phoQ24 backgrounds (A) or an rpoS background (B) is presented. After 16 h of incubation, the β-galactosidase activity of the bacteria released from macrophages, and those grown in RPMI medium were determined by using the chemiluminescence assay. Expression of sodCI in a wt background in macrophages is considered 100%.

In vitro, sodCII is primarily under the control of RpoS. We therefore tested the effect of an rpoS-null mutation on sodCII and sodCI expression in macrophages. As described above, sodCII was induced almost fourfold in bacteria recovered from macrophages compared to those propagated in tissue culture medium (Fig. 4B). The loss of RpoS abrogated sodCII induction. In contrast, sodCI expression was unaffected. Although not significant in this experiment, there might be residual RpoS-independent induction of sodCII, which again would be consistent with our in vitro data.

PhoPQ regulation of sodCI is apparently direct.

PhoPQ is known to activate a number of regulatory systems: PmrAB, SlyA, SsrAB, and RstAB (4, 26, 37-39, 41, 52). It is possible that sodCI is not directly regulated by PhoP but rather indirectly via one of these systems. In order to test this hypothesis, we inactivated these systems in both the wild-type and the phoQ24 backgrounds and assayed the resulting β-galactosidase activity from the sodCI-lac fusion. Navarre et al. (38) previously concluded that sodCI was not under the control of SlyA. Our data show that the expression of sodCI is not affected by the deletion of pmrA, ssrB, or rstA in either the wild-type or the phoQ24 background (Fig. 5). These results are consistent with the hypothesis that PhoP directly activates sodCI expression.

FIG. 5.

Expression of sodCI is unaffected by mutations in PhoP-dependent regulatory systems. (A) Effect of a ΔpmrA::Cm mutation on the expression of sodCI. The PmrA-regulated gene, pmrI, was monitored as a control. (B) Effect of ΔrstA::Cm and ΔssrB::Cm mutations on the expression of sodCI. All strains were grown in N-min modified medium (pH 5.6) supplemented with 10 μM or 10 mM magnesium chloride (inducing and repressing conditions, respectively) to an OD₆₀₀ of 0.3. β-Galactosidase activity units are defined as (μmol of ONP formed min⁻¹) × 10⁶/(OD₆₀₀ × ml of cell suspension) and are reported as means ± the standard deviation (n = 4).

Analysis of the sodCI promoter region.

To precisely map the sodCI promoter, we used primer extension analysis (Fig. 6). The results revealed that the transcription start site of sodCI is 61 bp upstream of the ATG start codon. To find the promoter region that is necessary for PhoP activation, we cloned various fragments of the sodCI promoter 5′ to a promoterless lacZ gene in pDX1, an apramycin-resistant plasmid derived from pAH125 (20; Lin and Slauch, unpublished). These constructs were integrated into the chromosome at the Lambda attachment site in both wild-type and phoQ24 backgrounds. The results shown in Fig. 7 indicate that the smallest fragment that was cloned, corresponding to positions −42 to +12 of the sodCI promoter (fragment C), conferred regulation in response the phoQ24 allele. The fragment corresponding to positions −98 to −29 was not significantly regulated and serves as a negative control.

FIG. 6.

Mapping of the transcription start site of sodCI by primer extension. The phoQ24 mutant cells were grown in LB medium to an OD₆₀₀ of ∼1.0 for isolation of total RNA. The transcription start site on the corresponding sequence is underlined.

FIG. 7.

Deletion analysis of the sodCI promoter region. The serovar Typhimurium strains contained lac transcriptional fusions to the indicated fragments of the sodCI promoter region integrated at the λatt site in wt and phoQ24 strains. Cultures were grown in N-min modified medium (pH 5.6) supplemented with 10 mM MgCl₂ to an OD₆₀₀ of 0.3. β-Galactosidase activity units are defined as (μmol of ONP formed min⁻¹) × 10⁶/(OD₆₀₀ × ml of cell suspension) and are reported as means ± the standard deviation (n = 4).

Analyses of a series of PhoP-regulated promoters suggests a consensus sequence for PhoP binding: two (G/T)GTTTA(A/T) direct repeats separated by 4 bp (28, 37, 44, 55). In a significant subset of PhoP activated promoters, this binding site overlaps the −35 promoter sequence (55). A similar sequence is noted at the appropriate location in the sodCI promoter from −42 to −25 (Fig. 8). To test the relevance of this sequence, we altered specific base pairs in the context of the fragment from −42 to +12. Any change to this consensus sequence significantly affected activation in the phoQ24 background. The most striking example is fragment H (Fig. 8), in which three base pair changes, corresponding to sites shown to be critical for PhoP activation of the mgtA promoter in E. coli (53), completely abrogated the transcriptional activation in the phoQ24 background. These results show that the fragment from −42 to +12 contains all of the sequence required for PhoP activation and suggest that the identified consensus PhoP-binding site is critical for this activation.

FIG. 8.

Analysis of the PhoP consensus sequence within the sodCI promoter. The sequence of fragment C (Fig. 7) is shown. The +1 base, −10 sequences, and bases matching the PhoP consensus binding sequence are in boldface. The PhoP binding site consensus is shown above the sequence. The underlined bases are those that have been shown to be critical in PhoP binding to the mgtA promoter in E. coli (53). The N-terminal amino acid sequence of Gifsy-2 L is indicated starting at −30; the first four of five amino acids in the L open reading frame are conserved with those of Lambda L. The serovar Typhimurium strains contained lac transcriptional fusions to the indicated fragments of the sodCI promoter region integrated at the λatt site in wt and phoQ24 backgrounds. The base pairs that differ in fragments F to H are indicated. Cultures were grown in N-min modified media (pH 5.6) supplemented with 10 mM magnesium chloride to an OD₆₀₀ of 0.3. β-Galactosidase activity units are defined as (μmol of ONP formed min⁻¹) × 10⁶/(OD₆₀₀ × ml of cell suspension) and are reported as means ± the standard deviation (n = 4).

The data presented above identify a putative PhoP binding site. Together with the fact that known PhoP-regulated transcriptional activators are apparently not involved in sodCI regulation, these results suggest that PhoP acts directly at the sodCI promoter. However, our attempts to gel shift sodCI promoter fragments by using purified His-tagged PhoP under a variety of conditions were inconclusive (not shown). Zwir et al. (55) recently combined bioinformatics and gene expression analysis to identify members of the PhoP regulon. These authors used chromatin immunoprecipitation assays to confirm PhoP binding to designated sites. E. A. Groisman tested whether PhoP was cross-linked to the sodCI promoter in his assay (unpublished data). Using primers that amplify a 276-bp fragment centered on the putative PhoP site, the results (not shown) suggested that PhoP did not bind to this region under the in vitro inducing conditions. Thus, although our genetic data strongly argue that PhoP directly activates sodCI transcription, we have been unable to confirm direct binding using in vitro molecular techniques. There are two possible explanations for these inconclusive results. Perhaps there is an additional unidentified regulator that is controlled by PhoP and directly acts at the sodCI promoter. If so, then this regulator apparently binds to a site that is strikingly similar to that recognized by PhoP. Alternatively, PhoP binding to the nonconsensus sodCI site is weak and is only realized under the in vivo induced conditions. Indeed, the induction of sodCI in vivo is far greater than what is achievable in vitro under any condition that we tested (Fig. 3 and 4). At the moment, we favor the latter interpretation.

DISCUSSION

Salmonella enterica serovar Typhimurium strain 14028 produces two periplasmic superoxide dismutases, SodCI and SodCII. Although both proteins are produced during infection (27), only SodCI contributes to virulence. Although this inequality in roles is mostly due to some difference in the two proteins, several lines of evidence suggested that the two genes were differentially regulated in the host (9, 27, 38, 45). Here we show that sodCI, encoded on the Gifsy-2 bacteriophage, is a member of the PhoPQ regulon. As such, it is transcriptionally induced in the Salmonella-containing vacuole of the macrophage, ideally expressed to combat phagocytic superoxide known to be important in controlling Salmonella infection (7, 32, 46-48). The PhoPQ regulon includes a large number of genes whose products contribute to macrophage survival by conferring, for example, resistance to antimicrobial peptides and low pH (16, 55). PhoP also contributes to activation of the SPI-2 type three secretion system required for establishment of the so-called Salmonella-containing vacuole in which the bacteria replicate (4). We now add resistance to phagocytic superoxide to the PhoP repertoire of intracellular survival functions, further emphasizing the central importance of this regulon.

In vitro, the PhoPQ system is induced in media with a low concentration of divalent cations, including Mg and Ca (16), and the regulon includes Mg transporters apparently required to adapt to the low-Mg environment. However, recent data suggest that Mg may not be the dominant signal that activates the system in the macrophage. Rather, a low pH seems critical (31). Also, it has been shown that PhoQ can respond to sublethal concentrations of antimicrobial peptides, a signal that is also likely to be important in vivo (1, 2). The PhoPQ regulon, partially via the PmrAB system, is responsible for resistance to these peptide antibiotics. Our data are consistent with these observations. In N-minimal medium, even at pH 5.6, lowering the Mg concentration induced sodCI only two- to threefold. In contrast, the gene was induced ∼15-fold in macrophages and in the animal. Induction was even more dramatic in the strain producing the PhoQ24 protein, which does not respond to Mg at pH 5.6 (Fig. 2).

All of the known PhoP regulon members involved in virulence were apparently acquired by horizontal gene transfer (16, 55). SodCI is expressed from the lysogenic Gifsy-2 phage. The phage encodes a number of potential virulence factors, including SseI, which is under the transcriptional control of SsrAB (50) and is secreted by the SPI2 type three secretion system (34), and GtgE, a putative cytoplasmic protein of unknown function with no known homologs. (See reference 22 for a more complete analysis of Gifsy-2.) Mutations in only sodCI and gtgE confer a phenotype in the mouse model of infection, together accounting for essentially all of the ∼150-fold decrease in virulence seen in a strain cured of Gifsy-2 phage (22). We know that both sodCI and gtgE are expressed from the lysogen and that phage induction is not required, because strains containing deletions of Gifsy-2 that block excision, immunity, late gene transcription, and replication, but which leave sodCI and gtgE intact, are fully virulent (22). Indeed, studies here were performed with strains deleted for all phage genes downstream of sodCI, analogous to the deletion mentioned above.

In the phage, the sodCI open reading frame and downstream ailT gene are inserted between the open reading frames for the minor tail proteins M and L. In Lambda, the stop codon for M overlaps the start codon for L, as is common for genes in the late operon. Presumably, sodCI evolved to be regulated by PhoPQ after it was acquired by the Gifsy-2 phage. Indeed, the start codon for the Gifsy-2 L gene is at position −30 in the sodCI promoter and much of the putative PhoP binding site is within the L open reading frame (Fig. 8). Interestingly, eight of the first nine amino acids of Gifsy-2 L (corresponding to 23 of the first 27 nucleotides) are identical to those of Lambda L and a subset of other lambdoid phage L proteins. This conservation of amino acids is certainly not the rule among the lambdoid phages, and many of the genes in the late operon show no direct homology to Lambda, although the overall layout of genes and, presumably, function are conserved. Thus, this conservation suggests that the N terminus of L is important. One of the base pairs that differs from Lambda, leading to an amino acid change from Arg to Pro in the sequence, corresponds to the first conserved G in the upstream PhoP half site. Although the overall PhoP binding site differs from consensus, this must be viewed in the context of the L open reading frame. For example, the (G/T)G in the downstream PhoP half site in the sodCI promoter is not conserved, but these nucleotides correspond to the ATG start codon for L (Fig. 8). Thus, the phage presumably evolved to regulate the acquired sodCI gene via PhoP while maintaining the amino acid sequence critical for L function.

Given these constraints, sodCI has apparently evolved an imperfect PhoP binding site. Indeed, although a few members of the PhoPQ regulon, such as phoPQ, mgtA, and slyB, possess consensus or near consensus binding sites, many members of the regulon do not (28, 55). Binding has also been observed at imperfect sites in the pmrD and pcgL promoters (28). In other cases, however, binding has not been shown in vitro, although partial sites are apparent. We have been unable to prove direct binding of PhoP to the sodCI promoter in vitro. Given the strong induction in vivo versus that obtainable in vitro, it is possible that PhoP, appropriately activated in the phagosome and itself increased in production, is capable of binding these imperfect sites and directly activating expression of this entire repertoire of genes, which is important for survival in this hostile environment.