Dual regulation by phospho-OmpR of ssrA/B gene expression in Salmonella pathogenicity island 2

Authors


Summary

Expression of genes located on Salmonella pathogenicity island 2 (SPI-2) is required for systemic infection in mice. This region encodes a type III secretion system, secreted effectors and the two-component regulatory system SsrA/B (also referred to as SpiR), as well as additional uncharacterized genes. In the present work, we demonstrate that phospho-OmpR (OmpR-P) functions as an activator at the spiC–ssrA/B locus. There are two promoters at spiR; one is upstream of ssrA and the other upstream of ssrB. Our results indicate that, in contrast to many two-component regulatory systems, regulation of the sensor kinase SsrA appears to be uncoupled and distinct from regulation of the response regulator SsrB. OmpR regulation of ssrA/B is one of only a few examples known in which a two-component response regulator directly regulates the expression of another two-component regulatory system.

Introduction

Two-component regulatory systems mediate adaptive responses in prokaryotes, including the Archaea and lower eukaryotes. The first component is a sensor kinase, often a membrane protein, that senses a change in the environment and is autophosphorylated on a conserved histidine residue by cytoplasmic ATP. In a few examples, the signals that stimulate phosphorylation of the sensor kinase are known and include iron (PmrA) (Wosten et al., 2000), Mg2+/Ca2+ (PhoP) (Garcia Vescovi et al., 1997), quinone (ArcA) (Georgellis et al., 2001) and acetosyringone and α-hydroxy-acetosyringone (VirA) (Stachel et al., 1985). Once phosphorylated, the phosphoryl group is transferred from the sensor kinase to the second component, the response regulator. Phosphorylation of the response regulator at a conserved aspartic acid residue alters its output, which is most often an increase in its affinity for DNA.

The OmpR/EnvZ two-component regulatory system regulates the porin genes ompF and ompC in response to changes in osmolarity. OmpR is a two-domain response regulator containing an N-terminal phosphorylation domain and a C-terminal DNA-binding domain, separated by a linker that is protease sensitive (Kenney et al., 1995). A crystal structure of the C-terminus (OmpRc) identifies OmpR as belonging to a family of winged helix–turn–helix (HTH) proteins (Kondo et al., 1997; Martinez-Hackert and Stock, 1997). Phosphorylation of OmpR increases its affinity for DNA (Aiba et al., 1989; Huang and Igo, 1996) by 10- to 30-fold (Head et al., 1998). Communication between domains is bidirectional, as DNA binding in the C-terminus stimulates OmpR phosphorylation in the N-terminus (Ames et al., 1999; Buckler et al., 2000; Kenney, 2000). Emerging evidence indicates that OmpR regulates many additional genes outside the porin repertoire (Oshima et al., 2002), including regulation of the flagellar operon flhDC (Shin and Park, 1995), curli fimbrial expression (Vidal et al., 1998) and cryptic porins ompS1 and ompS2 in S. typhi (Fernandez-Mora et al., 1995; Oropeza et al., 1999). In Salmonella, Shigella and Yersinia, ompR mutants are attenuated for virulence (Dorman et al., 1989; Bernardini et al., 1990; 1993; Lindgren et al., 1996; Dorrell et al., 1998). Most recently, OmpR was implicated in the stationary phase acid tolerance response of S. typhimurium (Bang et al., 2000; 2002).

Salmonella enterica serovar Typhimurium contains large blocks of horizontally acquired genes required for virulence, termed pathogenicity islands. Salmonella pathogenicity islands SPI-1 and SPI-2 encode their own type III secretion systems, which function to export bacterial effector proteins into the host (Ochman et al., 1996). Type III secretion systems differ from traditional secretion systems in that the exported proteins lack Sec-dependent secretion signals and the export apparatus spans the bacterial inner and outer membrane as well as the host plasma membrane or vacuolar membrane (for reviews, see Hueck, 1998; Galan and Collmer, 1999; Aizawa, 2001). SPI-1 plays a role in early infection (Mills et al., 1995; Rakeman and Miller, 1999), whereas SPI-2 is required for systemic infection in mice and is essential for Salmonella replication within macrophages (Ochman et al., 1996; Shea et al., 1996; Cirillo et al., 1998; Hensel, 2000). Recently, it was shown that genes on SPI-2 allow Salmonella to avoid NADPH oxidase-dependent killing by macrophages (Vazquez-Torres et al., 2000).

The ssrA/B locus on SPI-2 encodes a putative two-component regulatory system that regulates the expression of the type III secretion system of SPI-2 and many other genes (Ochman et al., 1996). Upstream of ssrA/B, and, divergently expressed, is spiC (see Fig. 1). SpiC was initially reported to be a secreted effector (Uchiya et al., 1999), but more recent studies suggest that it is required for the assembly of a functional secretory apparatus (Freeman et al., 2002; Yu et al., 2002).

Figure 1.

A. Gene arrangement of spiC–ssrA/B in SPI-2. The hatched box A1 indicates the OmpR binding site originally identified by Lee et al. (2000) located −67 to −31 nucleotides upstream from the transcriptional start site of ssrA. The other boxes (A2–5) indicate new OmpR-P binding sites identified in the present work (see Fig. 6). Boxes B1–3 indicate OmpR-P binding sites in ssrB. The transcriptional start sites of ssrA and ssrB are indicated by arrows and labelled +1. The arrows above denote the translational start sites for ssrA and ssrB. The +164 indicates there is a 164 nucleotide untranslated region in ssrA, and +150 is the untranslated region of ssrB.
B. Construction of transcriptional lacZ fusions used in the present study. The regulatory regions of ssrA (253 to +209), spiC (444 to +18) and ssrB (3045 to +228) were amplified by PCR using primers as described in Experimental procedures and cloned into vector pKLC-11 to yield plasmids pKF13B, pKF13A and pKF12 as lacZ transcriptional fusions. The XbaI fragments containing these lacZ fusions were subcloned into the single-copy plasmid vector pPP2-6 to yield plasmids pKF86, pKF93 and pKF87.

The arrangement of the ssrA/B system differs from many two-component systems in that the sensor kinase is a hybrid kinase that also contains a response regulator receiver domain and a histidine phosphotransfer domain (HPt) at its C-terminus, placing it in the BvgS family with similarity to RcsC, LuxN, LuxQ, etc. The response regulator ssrB is located downstream of the sensor kinase, separated by a 30 bp intergenic region. This region has been largely overlooked as a regulatory region and is one of the subjects of this report. A previous study discovered that transcription of SPI-2 genes, including ssrA and ssaH, was induced in macrophages (Cirillo et al., 1998), whereas a more recent study reported that OmpR was the regulator responsible for activation of ssrA (Lee et al., 2000). A single binding site was identified by DNase I footprinting, consisting of 36 bp located −67 to −31 nucleotides upstream from the transcriptional start site (identified in the present work). This single site represents a departure from the DNA binding pattern observed previously for OmpR regulation of the porin genes, which typically contain three adjacent OmpR binding sites (Maeda and Mizuno, 1990; Rampersaud et al., 1994; Huang and Igo, 1996). As unphosphorylated OmpR appears to have no role in porin gene regulation (Slauch and Silhavy, 1989), we investigated the role of phosphorylation in stimulating OmpR binding at the ssrA/B locus. Phosphorylation of OmpR has a large effect on DNA binding at ssrA, and OmpR-P also binds to the ssrB region. Our results identify multiple OmpR-P binding sites and suggest for the first time a role for OmpR-P and SsrB in activation of SPI-2-encoded genes.

Results

Cloning of the spiC-ssrA/B regulatory region and identification of the regulatory regions

The regulatory regions of ssrA, ssrB and spiC were amplified by polymerase chain reaction (PCR), cloned into pKLC-11 and subcloned into single-copy plasmids (pPP2-6) as transcriptional fusions to lacZ (see Fig. 1 and Experimental procedures). To characterize further the regulation of these genes, we transformed Salmonella wild-type strain 14028s, ompR, envZ and ssrB strains with the plasmids and measured their β-galactosidase activity in batch cultures. We also infected J774 macrophages with the Salmonella strains containing the plasmid-encoded fusions. The activities of the fusions in macrophages were similar to the activities we measured in Salmonella strains in batch cultures (data not shown). The one exception was the ssrA–lacZ transcriptional fusion, which was not OmpR dependent in vitro (see Fig. 7 and Discussion).

Figure 7.

β-Galactosidase activity of ssrA–lacZ in various growth conditions in wild-type Salmonella 14028s and ompR mutant strains. Salmonella containing the ssrA–lacZ fusion was grown in LB broth overnight and subcultured in various growth media (indicated below columns): CFA is 155 mOsm kg−1, CFA300 is 700 mOsm kg−1, LB is 448 mOsm kg−1, M9 minimal medium at pH 4.5 (281 mOsm kg−1), and M9 minimal medium at pH 4.5 plus 0.5 M NaCl (1168 mOsm kg−1). β-Galactosidase assays were performed as described in Experimental procedures. The grey columns indicate the wild-type strain, and the hatched columns indicate the ompR strain. The columns indicate the mean ± the standard error (n = 3). The β-galactosidase activity of ssrA–lacZ in J774 mouse macrophages (from Fig. 2A) is also included in the rightmost columns (n = 6) for a comparison of OmpR-dependent ssrA activation observed in Salmonella-infected macrophages but not in Salmonella cultures in vitro. The activity is not directly comparable to the other columns.

The results from ssrA–lacZ transcriptional fusions in Salmonella-infected macrophages are shown in Fig. 2A. In the wild-type strain, there is a very robust level (>6000 units) of β-galactosidase activity. In the absence of ompR, transcriptional activity is reduced by 80%. The OmpR-dependent stimulation of ssrA–lacZ requires the sensor kinase EnvZ, as the activity of an envZ deletion strain is similar to that of the ompR deletion strain. Deletion of ssrB reduces the ssrA–lacZ transcriptional activity to 50% of the wild-type strain. Thus, OmpR appears to function as an activator of ssrA, as reported previously (Lee et al., 2000), and SsrB also appears to activate ssrA, although the mechanism of regulation remains to be determined.

Figure 2.

β-Galactosidase assays of lacZ transcriptional fusions in macrophages. The fusions were constructed as described in Experimental procedures (see also Fig. 1B). Plasmids containing the fusions were transformed by electroporation into the wild-type strain 14028s, ompR, envZ and ssrB strains. The cells containing each lacZ fusion were grown and used to infect macrophages as described in Experimental procedures. The strain background is indicated below the graph. The columns indicate the mean ± the standard error.
A. ssrA–lacZ transcriptional fusion (pKF86, pKF 91), n = 6.
B. ssrB–lacZ transcriptional fusion (pKF87, pKF92), n = 5.
C. spiC–lacZ transcriptional fusion (pKF93, pKF94), n = 4.

The activity of the ssrB–lacZ transcriptional fusion is shown in Fig. 2B. The absence of ompR and ssrB substantially reduced the activity of the ssrB–lacZ fusion (40-fold and 60-fold respectively). OmpR activates the expression of ssrB, and ssrB is autoregulated. As we observed with OmpR activation of ssrA, activation of ssrB by OmpR also depends upon the kinase EnvZ.

In the case of the spiC–lacZ fusion (shown in Fig. 2C), deletion of ssrB had a significantly greater effect than deletion of ompR, reducing activity by 60-fold compared with 2.4-fold. Thus, results from the fusions shown in Fig. 2 suggest that OmpR activates ssrA and, to a lesser extent, SsrB also activates ssrA, whereas both OmpR and SsrB have a significant effect on ssrB. In contrast, spiC expression is strongly affected by SsrB and less so by OmpR.

Primer extension of ssrA and ssrB

Our results from β-galactosidase assays strongly suggest that there are two promoters at ssrA/B (Fig. 2A and B), a regulatory region upstream of ssrA and a regulatory region upstream of ssrB. In order to test this hypothesis directly, we performed primer extensions; the results are shown in Fig. 3. In the wild-type strain, a robust signal for ssrA is detected (Fig. 3, lane 1) and a signal for ssrB is also detected (Fig. 3, lane 4). The level of product is substantially reduced in the absence of ompR (Fig. 3, lanes 2 and 5) or ssrB (Fig. 3, lanes 3 and 6). In the absence of ompR, the reduction in ssrA product is greater than when ssrB is eliminated (compare Fig. 3, lanes 2 and 3), consistent with the greater effect of OmpR in stimulating ssrA–lacZ activity (Fig. 2A). The primer extension analysis places the transcriptional start site of ssrA at 164 nucleotides upstream of the translational start and the transcriptional start site of ssrB 150 nucleotides upstream of the ssrB initiating codon (see Fig. 1). The results from our β-galactosidase assays (Fig. 2) and the primer extension analysis (Fig. 3) clearly indicate that OmpR functions to activate expression of ssrA and ssrB (see Discussion).

Figure 3.

Identification of the ssrA and ssrB transcriptional start sites by primer extension. In the left and centre four lanes, the sequencing reaction is labelled TACG at the top of the gel. In the lane labelled 1 (bottom of gel), the product of the ssrA reaction in wild-type strain 14028s is shown, in lane 2 is the ssrA product from an ompR deletion strain and, in lane 3, the ssrA product from an ssrB deletion strain is shown. The lanes labelled 4–6 are identical to lanes 1–3, except the reaction is for ssrB. A sample of 5.5 µl of the 11 µl reaction volume was loaded onto the gel for the ssrA wild-type primer extension product, whereas the entire product was loaded in the other lanes. The arrow denotes the transcriptional start site.

ssrA-1 DNA stimulates OmpR phosphorylation

A previous study identified a lone OmpR binding site upstream of ssrA (Lee et al., 2000). The effect of phosphorylation of OmpR on binding to this site was not reported. However, at other OmpR-regulated loci, when OmpR functions as an activator, OmpR-P binds to multiple sites from approximately −100 to −40 upstream from the transcriptional start site (Maeda and Mizuno, 1990; Harlocker et al., 1995; Huang and Igo, 1996). Inspection of the binding site identified by Lee et al. (2000) and comparison with known OmpR binding sites indicate that it is ≈ 50% identical to the high-affinity sites C1 and F1. These are sites at which OmpR binding is greatly enhanced, ≈ 25-fold, upon phosphorylation (Head et al., 1998).

We have shown previously that the presence of high-affinity OmpR binding sites can stimulate OmpR phosphorylation via reverse signalling (Ames et al., 1999). In this situation, high-affinity binding to DNA by the C-terminus of OmpR alters the N-terminal phosphorylation domain and stimulates phosphorylation. Non-specific DNA or low-affinity sites fail to stimulate phosphorylation (Ames et al., 1999). To determine whether the ssrA-1 binding site identified previously was a high-affinity site (Lee et al., 2000), we incubated OmpR with acetyl phosphate in the presence of ssrA-1. The stimulation is easily apparent as an increase in the area of an OmpR-P peak and a decrease in the area of the OmpR peak after separation on a C4 column using reversed phase high-performance liquid chromatography (HPLC) (Fig. 4). In Fig. 4A, the results of a phosphorylation reaction in the absence of ssrA-1 DNA are shown. In the absence of DNA, 52% of the OmpR protein is phosphorylated. Under identical conditions, the addition of ssrA-1 DNA stimulates OmpR phosphorylation, to 91% (Fig. 4B). This result indicates that OmpR binds to the ssrA-1 site with reasonably high affinity (Kd < 300 nM), and that binding results in a conformational change that stimulates phosphorylation. It also suggests that phosphorylation should have a large stimulatory effect on the affinity of OmpR for DNA at the ssrA-1 site.

Figure 4.

ssrA-1 stimulates OmpR phosphorylation. OmpR and OmpR-P are separated by reversed phase HPLC. Approximately 200 µl of OmpR (7.5 µM) was injected onto a C4 column.
A. The profile of OmpR and OmpR-P is shown in the absence of ssrA-1 DNA.
B. The profile of OmpR and OmpR-P in the presence of ssrA-1 DNA at a 2:1 (DNA–protein) molar ratio is shown. OmpR is incubated with acetyl phosphate for 120 min at room temperature, and the samples are separated on a C4 column (Head et al., 1998). The two peaks corresponding to OmpR-P and OmpR are labelled. The double-stranded ssrA-1 oligonucleotide used in this experiment is (5′−3′): GCATTGA CATAAAAACTTACAATTTGAAAAATTATTT. Only the sequence of the non-template strand is shown.

Phosphorylation of OmpR stimulates DNA binding to ssrA

We next measured equilibrium binding of OmpR and OmpR-P to the ssrA-1 binding site (Fig. 5). For these experiments, we used fluorescence anisotropy, which we have shown previously to be a sensitive and effective method for measuring OmpR binding to the upstream regulatory regions of the porin genes (Head et al., 1998; Tran et al., 2000). The apparent dissociation constant (Kd) for OmpR binding to ssrA-1 is 260 nM (Fig. 5, triangles and inset). When OmpR is phosphorylated, it results in an eightfold stimulation in binding affinity, and the Kd for OmpR-P is 32 nM (Fig. 5, circles). This result extends the previous study, which did not report the role of phosphorylation in the stimulation of ssrA/B binding by OmpR (Lee et al., 2000).

Figure 5.

OmpR and OmpR-P binding to ssrA-1. OmpR (triangles) or OmpR-P (circles) was titrated into a binding reaction mixture containing 3 nM fluorescein-labelled ssrA-1 oligonucleotide in a buffer described in Experimental procedures. The mean of five separate measurements performed at each titration point is plotted. The curves shown are a representative of four separate binding curves; the dissociation constant for the OmpR-P curve is 32 nM. The dissociation constant for the OmpR curve shown (triangles) is 260 nM; the full-length curve is shown in the inset. The ssrA-1 oligonucleotide used in this experiment was the same as that indicated in Fig. 4, except that it was labelled with fluorescein at the 5′ end of the top strand.

DNase I footprinting with OmpR and OmpR-P at the ssrA/ssrB locus

As phosphorylation had such a large effect on OmpR binding at ssrA-1, it seemed worthwhile to examine and compare the footprinting reactions using OmpR and OmpR-P (shown in Fig. 6). In Fig. 6A, the ssrA regulatory region is shown. OmpR protects only weakly a region from −67 to −31 (Fig. 6A, lanes 10–16), as reported previously (Lee et al., 2000). This site is upstream from the determined transcriptional start site shown in Fig. 3. In the presence of OmpR-P, a large region encompassing −83 to +6 is protected (Fig. 6, A1, lanes 2–8). This result emphasizes the importance of phosphorylation in promoting OmpR binding to the ssrA region (sites A2–A5 in Figs 1 and 6). Surprisingly, OmpR-P, but not OmpR, protects three sites in the upstream region of ssrB, labelled B1, B2 and B3 (+116 to +204, Fig. 6B, lanes 10–16). B1 is located in the intergenic region between ssrA and ssrB, whereas the B2 and B3 sites are located downstream from the translational start site at +150. The observation that OmpR-P binding overlaps the ssrB translational start site raises interesting mechanistic questions as to how OmpR activates ssrB (see Fig. 2B and Discussion). These results are summarized in Fig. 1.

Figure 6.

DNase I footprinting of ssrA (A) and ssrB (B) by OmpR and OmpR-P. Lane 1 contains 8.4 nM DNA in the absence of DNase I; lanes 2 and 10 contain similar amounts of the probe after exposure to DNase I for 2 min. Lanes 3–8 contain OmpR-P, and lanes 11–16 contain OmpR at 0.46, 1.94, 3.3, 4.6, 7.1 and 9.2 µM respectively. The OmpR binding sites were located by a sequencing reaction in adjacent lanes (indicated by TACG), using the same template. OmpR-P concentrations were determined by reverse phase HPLC (Head et al., 1998). The triangles indicate hypersensitive sites.

Is ssrA activation by OmpR osmoregulated?

The previous study by Lee et al. (2000) reported that ssrA was repressed at high osmolarity and that OmpR was required for this effect. As the location of the OmpR binding sites at ssrA suggested that OmpR might function as a repressor (Fig. 6), we examined the effect of increasing osmolality on the expression of our ssrA–lacZ fusions in batch cultures of Salmonella as described previously (Lee et al., 2000). The results are shown in Fig. 7. In CFA medium, ssrA expression was not dependent on OmpR and, furthermore, the activity was not affected by increasing osmolality (compare Fig. 7, columns 1 and 2 with 3 and 4). It is worth noting that, in M9 minimal media, the fusions had a twofold lower level of expression at high osmolality (1168 mOsm kg−1) compared with low osmolality (281 mOsm kg−1). However, the decrease in β-galactosidase activity of the ssrA–lacZ fusion strain was not dependent upon ompR (see Fig. 7, columns 7–10). In contrast to our observations in vitro, in Salmonella-infected macrophages, the activation of ssrA was substantially dependent on OmpR (>80%; Fig. 2A and Fig. 7, columns 11–12).

Discussion

Phosphorylation of OmpR increases its affinity for ssrA

The equilibrium dissociation constant (Kd) for OmpR binding to the ssrA-1 binding site is 260 nM. For OmpR-P binding to this site, the Kd is 32 nM. This places the ssrA-1 binding site just below the highest affinity OmpR binding sites known. F1 and C1 are located approximately between −100 and −80 upstream from the ompF and ompC genes respectively. OmpR-P binds to these sites with a Kd of about 6 nM (Head et al., 1998). In the present work, we have shown that phosphorylation of OmpR increases its affinity for ssrA-1 nearly 10-fold. Our results extend the findings from a previous study on DNA binding by OmpR that did not examine the effect of OmpR phosphorylation on DNA binding at ssrA (Lee et al., 2000). This increase in affinity is further demonstrated by our observation that ssrA-1 DNA promotes reverse signalling and stimulates OmpR phosphorylation (Fig. 4; Ames et al., 1999). At the porin genes, OmpR-P binds with an affinity that is 10- to 30-fold higher than the affinity for OmpR, depending on the particular site (Head et al., 1998). OmpR binding sites are poorly conserved, and comparison of the regions of protection by DNase I in the present study with the well-characterized ompF and ompC binding sites indicate that the A1 site is 50% identical to the high-affinity F1 and C1 binding sites, whereas the A2–A5 sites are between 22% and 55% identical, depending on the site boundaries used. Interestingly, the B1 and B2 sites are 55–67% identical to the low-affinity sites C2 and C3. The consequences of this binding site arrangement, in terms of regulation by OmpR, remain to be determined.

From our studies of the ssrA–, ssrB– and spiC–lacZ transcriptional fusions, we found that the level of β-galactosidase activity in the ompR deletion strain was essentially equivalent to the level of activity in an envZ deletion strain, indicating that OmpR-P is the form of the protein required for activity (Fig. 2A–C). This result also indicates that EnvZ is signalling through OmpR to stimulate the expression of SPI-2. Furthermore, with the exception of the ssrA-1 site, only OmpR-P (and not OmpR) protected the regions upstream of ssrA and ssrB (Fig. 6). Previous studies used maltose binding protein fused to the cytoplasmic domain of EnvZ (MBP–EnvZ) to phosphorylate OmpR (Huang and Igo, 1996), and this construct is not particularly effective at generating high concentrations of OmpR-P (our unpublished observations). Because we used the small molecule phosphodonor acetyl phosphate to phosphorylate OmpR, we were able to produce high concentrations of phospho-OmpR for the present study. In addition, we can easily determine the concentration of OmpR-P present in our assay by separating it from OmpR on a C4 column using reversed phase HPLC (Fig. 4).

A two-component regulatory system regulates a two-component regulatory system

We are aware of two previous reports of regulation of one two-component system by another. These are regulation of ResD/E by PhoP/R in Bacillus subtilis (Birkey et al., 1998) and regulation of PmrA/B by PhoP/Q in Salmonella (Soncini and Groisman, 1996). In the OmpR regulation of SsrA/B reported in the present work, the response regulator OmpR functions to activate separately the transcription of both the kinase and the response regulator. In the case of ResD/E, PhoP functions as a dual regulator, activating transcription of the resABCDE operon under phosphate-depleted conditions and repressing the internal resDE promoter during phosphate-limited growth (Birkey et al., 1998). Thus, these complex interactions are part of a regulatory network that integrates and balances cellular functions of respiration/energy production with phosphate starvation. In the PhoP regulation of resDE, the phosphoresponse regulator is the key species that is involved in repression (Birkey et al., 1998). In the case of PmrA/B regulation, the response regulator PhoP activates pmrD. It was proposed that PmrD acts post-transcriptionally to activate PmrA/B (Kox et al., 2000). With OmpR regulation of SPI-2 gene expression, it remains to be determined how OmpR-P functions mechanistically as an activator and which binding sites are occupied.

The ssrA/B region contains two promoters

The results from β-galactosidase activity assays of the ssrA–lacZ and ssrB–lacZ transcriptional fusions and the primer extension analysis clearly indicate two distinct regulatory regions at ssrA and ssrB (Figs 2A and B and 3). The effect of OmpR appears to be direct, based on our observation that OmpR-P binds upstream of ssrA and to the intergenic region between ssrA and ssrB (Fig. 6). Furthermore, ssrB is autoregulated, as ssrB–lacZ activity was reduced in an ssrB deletion strain (64-fold), whereas deletion of ssrB had a considerably lower effect on ssrA expression (Figs 2A and 3). Thus, the regulation of ssrA appears to be uncoupled from ssrB regulation. We are presently raising antibodies to purified SsrA and SsrB proteins in order to examine their protein levels in vivo.

spiC is regulated by both OmpR and SsrB

From the data shown in Fig. 2C, it is apparent that both OmpR and SsrB activate the transcriptional fusions to spiC. Clearly, SsrB has the greater effect, activating spiC expression 60-fold, whereas the stimulation by OmpR is 2.4-fold. DNase I protection assays with OmpR and SsrB will be informative in determining whether these effects are direct (i.e. OmpR binding directly to the regulatory region of spiC) or indirect, possibly via OmpR effects on ssrB expression (Fig. 2B). It will also be of interest to determine the location of the OmpR binding sites with respect to the SsrB binding sites at spiC. There is emerging evidence that diverse response regulators can pair in distinct combinations and interact at different promoters, suggesting an increase in the complexity of transcriptional regulation more commonly observed in eukaryotic systems (Pruss et al., 2003; R. Oropeza, X. Feng and L. J. Kenney, unpublished observations).

What are the signals that determine the expression of SPI-2 genes?

In the present work, we have demonstrated that OmpR binds directly to the ssrA and ssrB regulatory regions to affect their expression and that regulation requires the sensor kinase EnvZ. It is not known at present what the stimulus is to which EnvZ is responding. In the previous study, Lee et al. (2000) reported that ssrA expression was osmoregulated. At low osmolarity, ssrA was activated, and it was repressed at high osmolarity. The effect of osmolarity on ssrB expression was not examined. Given the location of the OmpR binding sites reported in the present study, we would predict that, at low osmolarity, the ssrA-1 binding site(s) upstream of the transcriptional start site would be occupied, leading to activation of ssrA (see Fig. 1). At high osmolarity, an increase in OmpR-P (Slauch and Silhavy, 1989; Russo et al., 1993) would result in occupancy of the ssrA binding sites located downstream of the transcriptional start site, repressing ssrA. However, in the present study, we did not find conditions under which either ssrA or ssrB was repressed by OmpR (see below). The activation of ssrA by OmpR and its osmoregulated expression in vitro reported in the previous study could result from differences between Salmonella strains SL1344 and 14028s (Lee et al., 2000). It is of interest, however, that OmpR activation of ssrA in the present study was only observed in Salmonella-infected macrophages (Figs 2A and 7) and not in Salmonella cultures in vitro. Thus, the signal that EnvZ is sensing must be directly or indirectly related to the macrophage environment.

How does OmpR activate expression of ssrA and ssrB?

A summary of our footprinting results and primer extension analysis is shown in Fig. 1. Altogether, our data indicate that OmpR-P binds to the upstream region of ssrA and around the intergenic region between ssrA and ssrB (Fig. 6). At ssrA, the highest affinity site for OmpR-P lies just upstream of the transcriptional start site (A1), but the remaining OmpR-P binding sites are downstream, suggesting a role for OmpR-P as a repressor. This configuration of binding sites has also been observed at the bipA gene in Bordetella, in which BvgA-P binds to high-affinity activating sites upstream and low-affinity repressing sites located downstream of the transcriptional start (Deora et al., 2001). When OmpR-P levels are low, the highest affinity sites would be occupied (ssrA-1), leading to activation of ssrA. Depending on the environmental signal, ssrB could be expressed as a polycistronic transcript from ssrA or from its own promoter. We have shown that there is a transcriptional start site 150 nucleotides upstream from the ATG codon. An increase in OmpR-P levels would enable OmpR-P to bind to lower affinity sites, repressing ssrA and ssrB. Thus, SpiC levels might also decrease, as the major effect on spiC is via SsrB rather than OmpR (Fig. 2C). However, to date, we have not observed OmpR-dependent repression of these genes.

More recently, it has become apparent that binding sites for transcriptional activators can be located downstream of the transcriptional start site and still play a role in activation. For example, Rns, an AraC family activator, regulates the expression of Cs1 and Cs2 pili in enterotoxigenic Escherichia coli. Rns is also autoregulated, and the Rns binding sites are upstream and downstream of the rns promoter (Munson and Scott, 2000). A similar arrangement of binding sites has been observed in the regulation of pstS by PhoP in B. subtilis (Qi and Hulett, 1998). How this is accomplished, in terms of interactions with the activator and RNA polymerase, is not understood. Clearly, an enhanced understanding of the molecular mechanism of OmpR and SsrB activation and the underlying complexity in gene regulation at SPI-2 will probably reveal new insights into transcriptional activation in general.

Experimental procedures

Growth conditions of bacterial cells

The bacterial strains and plasmids used in this study are listed in Table 1. Bacteria were grown with shaking at 37°C in LB medium. Antibiotics were used as needed at the following concentrations: ampicillin, 100 µg ml−1; kanamycin, 50 µg ml−1; tetracycline, 12.5 µg ml−1; chloramphenicol, 30 µg ml−1. Salmonella were transformed with various plasmids by electroporation. Bacteria used for macrophage infection were grown with gentle shaking at 37°C in LB medium overnight according to previous studies (Lee et al., 2000). For experiments at low and high osmolarity, the experiments were performed as described previously (Lee et al., 2000). M9 minimal medium (without CaCl2) was supplemented with 0.004% (w/v) histidine and 0.01 M glucose; the pH was adjusted with concentrated HCl to 4.5. For high osmolarity M9 medium, 0.5 M NaCl was added. Measurements of the osmolality of the medium were conducted in a Wescor 5100c vapour pressure osmometer according to the manufacturer's instructions, using 290 and 1000 mOsm kg−1 standards. J774 macrophages were routinely maintained in Dulbecco's modified Eagle medium (DMEM) containing sodium pyruvate, essential amino acids and 10% (v/v) fetal bovine serum (Gibco) at 37°C and 5% CO2.

Table 1. . Bacterial strains and plasmids.
Strain or plasmidDescriptionReference or source
Strains
 14028s Salmonella enterica serovar TyphimuriumM. J. Worley
 MGWP72714028s ompR1009::Tn10d-TetM. J. Worley
 MJW11214028s ssrB::KmM. J. Worley
 NK18214028s envZ::CmM. J. Worley
Plasmids
 pBS IIpUC18 ori Apr Sambrook et al. (1989)
 pKLC-11pBR322 ori promoterless lacZ cat AprThis study
 pMC1871pBR322 ori promoterless lacZ TetrPharmacia
 pPP2-6Mini-F replicon CmrJ. Duncan
 pUC4KpBR322 ori Kmr AprPharmacia
 pKF13870 bp fragment spanning promoter region of ssrA, coding region of ssrA, part of spiC, promoter region of ssrB, and coding region of ssrB cloned into pBS II vector EcoRI siteThis study
 pKF6A462 bp fragment spanning promoter region of ssrA cloned into pMC1871 SmaI siteThis study
 pKF8A393 bp fragment spanning promoter region of ssrB cloned into pMC1871 SmaI siteThis study
 pKF13B462 bp fragment spanning promoter region of ssrA cloned into pKLC-11 SmaI siteThis study
 pKF123273 bp fragment spanning promoter region of ssrA, coding region of ssrA, and promoter region of ssrB cloned into pKLC-11 SmaI siteThis study
 pKF13A462 bp fragment spanning promoter region of spiC cloned into pKLC-11 SmaI siteThis study
 pKF863462 bp fragment of ssrA–lacZ fusion cloned into pPP2-6 XbaI siteThis study
 pKF876273 bp fragment of ssrB–lacZ fusion cloned into pPP2-6 XbaI siteThis study
 pKF933462 bp fragment of spiC–lacZ fusion cloned into pPP2-6 XbaI siteThis study
 pKF91Kmr gene from pUC4K cloned into pKF86 SalI siteThis study
 pKF92Kmr gene from pUC4K cloned into pKF87 SalI siteThis study
 pKF94Kmr gene from pUC4K cloned into pKF93 SalI siteThis study

Molecular biology techniques

All enzymatic manipulations of DNA were performed using established techniques (Sambrook et al., 1989) with reagents purchased from Boehringer Mannheim, Gibco BRL (restriction endonucleases and alkaline phosphatase), New England Biolabs (T4 DNA ligase) and Stratagene (Pfu DNA polymerase). Plasmid DNA used for sequencing was isolated using ion exchange columns from Qiagen. Oligonucleotide synthesis was performed by Applied Biosystems automated solid-phase synthesis at the OHSU Microbiology Core Facility. DNA sequence determination was performed at the Vollum Institute (OHSU) using an Applied Biosystems 377 fluorescent DNA sequencer with chain termination chemistry.

Construction of transcriptional lacZ fusions

Standard molecular cloning techniques were used to construct all the fusions listed in Table 1. The entire 3870 bp fragment spanning the promoter region of ssrA, the coding region of ssrA, the promoter region of ssrB, the coding region of ssrB and part of spiC was amplified using PCR with Pfu DNA polymerase from the Salmonella 14028s genome with primers STMF1 (5′-CTGCCAGCATGAATTCCTCCTCAGAC-3′) and STMF3 (5′-GACCAATGCTGAATTCCATCGGACGC CCCTGG-3′), including an EcoRI site, and ligated into the cloning vector pBluescript-II KS, yielding pKF1. Plasmid pKLC-11 was used for construction of transcriptional lacZ fusions. The following oligonucleotides with desired restriction sites were synthesized and used: the reverse primer from vector pBluescript-II KS (5′-GGAAACAGCTATGACCATC-3′) and forward primer STMF4 (5′-GCCTGATATCTAAAGAT GTTTGCAGCG-3′) including an EcoRV site were used to construct pKF13A and pKF13B; reverse primer STMF16 (5′-GAATTCCTCCTCAGGCCTAAATGGGAG-3′) and for-ward primer STMF14 (5′-GCTCTACAATTTTAAAATGAGGC CTGG-3′) including a StuI site were used to construct pKF12. The ssrA–lacZ (253 to +209), ssrB–lacZ (−3045 to +228) and spiC–lacZ (444 to +18) fusions were subcloned into the XbaI site of the single-copy plasmid vector pPP2-6 to yield plasmids pKF86, pKF87 and pKF93. Plasmids pKF91, pKF92 and pKF94 contained a kanamycin gene insertion. The regulatory regions of ssrA (253 to +209) and ssrB (90 to +303) were amplified by PCR and cloned into vector pMC1871 to yield plasmids pKF6A and pKF8A as lacZ translational fusions. These two plasmids were used as templates for footprinting experiments.

Measurement of β-galactosidase activity in vitro

β-Galactosidase activity assays were performed in culture tubes. Cell culture (100 µl) was mixed with 900 µl of Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, pH 7.0); two drops of chloroform and one drop of 0.1% SDS was added followed by vigorous vortexing to lyse the bacteria. An aliquot of 200 µl of ONPG (4 mg ml−1) was added, and the reaction was stopped by the addition of 0.5 ml of 1 M Na2CO3, and the time was recorded. The reaction was centrifuged, and 0.8 ml of supernatant was used to measure the absorbance at 420 nm in a spectrophotometer. The β-galactosidase activity was calculated according to OD420 × 1000/t (min) × volume (ml) × OD590.

β-Galactosidase activity in macrophages

Strains containing the fusions were inoculated into 3 ml of LB medium with the appropriate antibiotics and incubated at 37°C with gentle shaking overnight. Approximately 0.5 × 109 cells were harvested and resuspended in 1 ml of DMEM, and 100 µl of cell suspension was used to inoculate each well of a six-well plate containing J774 mouse macrophages at a multiplicity of infection (MOI) of 100. J774 macrophages were cultured in DMEM as described previously (Alpuche Aranda et al., 1992). The plates were centrifuged at 148 g for 5 min to enable the cells to adhere and then incubated at 37°C for 20 min to infect. J774 cells were washed three times with Dulbecco's phosphate-buffered saline (DPBS) buffer to remove non-adherent cells, then 2 ml of DMEM containing 50 µg ml−1 gentamicin was added to each well, and the plates were incubated at 37°C for 7 h. The macrophages were harvested, collected into a tube, centrifuged at 4300 g for 10 min and washed with 1 ml of DPBS. After resuspension in 100 µl of DPBS, 0.9 ml of Z buffer, three drops of chloroform and two drops of 0.1% SDS were added followed by vigorous vortexing to lyse the macrophages and release intracellular bacteria. An aliquot of 200 µl of ONPG (4 mg ml−1) was added, and the reaction was stopped by the addition of 0.5 ml of 1 M Na2CO3, and the time was recorded. The reaction was centrifuged, and 0.8 ml of supernatant was transferred to a cuvette, and the OD at 420 nm was measured. The β-galactosidase activity was calculated according to: OD420 × 1000/t (min) × volume (ml). Controls using non-infected macrophages showed no measurable β-galactosidase activity.

Primer extension

Total RNA was obtained from the wild-type, ompR and ssrB strains using Trizol reagent according to the manufacturer's instructions (Gibco BRL). Approximately 50 µg of total RNA was isolated from 30 ml of cell culture in magnesium minimal medium (MgM), pH 5.0 (Beuzon et al., 1999). The primers used to detect the ssrA and ssrB transcripts were STMF4 (5′-GCCTGATATCTAAAGATGTTTGCAGCG-3′) and STMF17 (5′-GCCGTTAATGATGAATTCATGATCGTC-3′) respectively. The primers were 5′ labelled with [γ-32P]-ATP using polynucleotide kinase (NEB). Each RNA sample (10 µg) was suspended in 215 mM NaCl, 32 mM Tris-HCl, pH 7.8, and then hybridized to the appropriate primer by incubation for 3 min at 90°C followed by slow cooling to 42°C. The primer extension reaction was performed at 42°C for 90 min by the addition of 100 µM dNTPs and 10 units of AMV reverse transcriptase (Roche). The final products were precipitated with isopropanol, washed with 70% ethanol, dried and resuspended in sequencing stop buffer. The products were separated by electrophoresis on a sequencing gel along with a sequencing ladder generated using the same primers and plasmid pKF1 as a template.

OmpR purification

OmpR was purified as described previously (Jo et al., 1986) with the modifications indicated by Head et al. (1998). OmpR concentrations were determined by absorbance at 280 nm in a Beckman DU640 spectrophotometer using a value of 13 490 (M−1 cm−1) for the molar extinction coefficient.

Phosphorylation of OmpR by acetyl phosphate

OmpR protein was phosphorylated using 25 mM acetyl phosphate (Sigma) (Kenney et al., 1995). For some experiments, DNA was added in a 1:2 molar ratio of protein–DNA (Ames et al., 1999).

Separation of OmpR from OmpR-P

In order to estimate the extent of phosphorylation of OmpR, a mixture containing phosphorylated OmpR and unphosphorylated OmpR was separated using C4 reversed phase HPLC (Head et al., 1998). The sample was eluted with a gradient that varied from 80% eluent A to 0% over 40 min. Eluent B started at 20% and went to 100% over the same interval. Eluent A was 20% acetonitrile, 0.1% trifluoroacetic acid; eluent B was 60% acetonitrile, 0.1% trifluoroacetic acid. The flow rate was 1 ml min−1.

DNA binding

Equilibrium binding measurements were performed using fluorescence anisotropy (Head et al., 1998). The change in anisotropy, where ΔF/Fo represents the difference in anisotropy in the presence of protein minus the anisotropy in the absence of protein divided by the anisotropy in the absence of protein, is plotted as a function of the total protein concentration. The results from the binding curves were fit by non-linear least squares regression as described previously (Head et al., 1998). The apparent dissociation constants reported in the text represent the mean.

DNase I protection assay

Footprinting reactions were conducted according to the method of Huang and Igo (1996) as follows. Plasmids pKF6A and PKF8A were used as templates for PCR amplification of the ssrA and ssrB 5′ regulatory region respectively. For ssrA, the 32P-labelled oligonucleotide complementary to nucleotides +173 to +195 and an oligonucleotide complementary to nucleotides −263 to −238 were used as primers (non-template strand shown). For ssrB, they were +293 to +319 and −90 to −63. Each assay contained 3 × 105 c.p.m. labelled template. For the binding reaction, OmpR or OmpR-P is incubated for 20 min at room temperature in a buffer containing 40 mM KCl, 4 mM Tris-HCl, pH 7.9, 1 mM EDTA, 1 mM dithiothreitol (DTT), 12% (v/v) glycerol. DNase I was added, and the reaction was stopped after 2 min by the addition of 20 mM EDTA, 360 mM Na acetate, pH 5.5 (final concentration). The final products were precipitated with isopropanol, washed with 70% ethanol, dried and resuspended in sequencing stop buffer. The products were separated by electrophoresis on a sequencing gel along with a sequencing ladder generated using the same primers and plasmids as templates.

Acknowledgements

We thank Charlotte Head for OmpR purification, and Richard Goodman (Vollum Institute) for use of the fluorometer. We thank our laboratory members Kirsten Mattison, Don Walthers and Carolyn Snarskis for advice, discussions and comments on the manuscript, and anonymous reviewers for their helpful suggestions. We are indebted to Sunghee Chai for help and advice culturing macrophages, Joanne Rue provided RNA for primer extension, and Micah J. Worley provided strains. Fred Heffron contributed helpful discussions in the early phase of this project. Thanks also to Jack H. Kaplan (OHSU), Ferric Fang and Kelly Hughes (University of Washington) for thoughtful comments on the manuscript. Supported by grants MCB-9904658 from the National Science Foundation and GM58746 from the National Institutes of Health to L.J.K.

Ancillary