PspG, a New Member of the Yersinia enterocolitica Phage Shock Protein Regulon

Abstract

The Yersinia enterocolitica phage shock protein (Psp) system is induced when the Ysc type III secretion system is produced or when only the YscC secretin component is synthesized. Some psp null mutants have a growth defect when YscC is produced and a severe virulence defect in animals. The Y. enterocolitica psp locus is made up of two divergently transcribed cistrons, pspF and pspABCDycjXF. pspA operon expression is dependent on RpoN (σ⁵⁴) and the enhancer-binding protein PspF. Previous data indicated that PspF also controls at least one gene that is not part of the psp locus. In this study we describe the identification of pspG, a new member of the PspF regulon. Predicted RpoN-binding sites upstream of the pspA genes from different bacteria have a common divergence from the consensus sequence, which may be a signature of PspF-dependent promoters. The Y. enterocolitica pspG gene was identified because its promoter also has this signature. Like the pspA operon, pspG is positively regulated by PspF, negatively regulated by PspA, and induced in response to the production of secretins. Purified His₆-PspF protein specifically interacts with the pspA and pspG control regions. A pspA operon deletion mutant has a growth defect when the YscC secretin is produced and a virulence defect in a mouse model of infection. These phenotypes were exacerbated by a pspG null mutation. Therefore, PspG is the missing component of the Y. enterocolitica Psp regulon that was previously predicted to exist.

The phage shock protein (Psp) system was first described in Escherichia coli (6). Synthesis of E. coli PspA protein (the Psp response) is induced by the mislocalization of some outer membrane proteins, especially secretins (reviewed in reference 28), and by mutations that cause secretion defects (9, 19). The E. coli Psp response is also induced by depletion of the YidC protein (19, 30), which has pleiotropic effects on many cytoplasmic membrane proteins. Heat, osmotic, and ethanol shock also induce the Psp response (28). It is thought that a common event resulting from all of these inducing conditions may be dissipation of the proton motive force. The PspA protein apparently helps to maintain the proton motive force under inducing conditions (23).

The E. coli psp locus is made up of two divergently transcribed cistrons, pspABCDE and pspF. pspA promoter activity is dependent on the enhancer-binding protein (EBP) PspF (21) and the RpoN (σ⁵⁴) sigma factor of RNA polymerase. PspF activity is negatively controlled by a direct interaction with the peripheral inner membrane protein PspA (5, 12). When the Psp response is active, the cytoplasmic membrane proteins PspB and PspC may detect an inducing signal and interact with PspA. This interaction may mean that PspA can no longer interfere with PspF, and PspF-dependent gene expression is induced. Much of this model awaits direct experimental investigation, although interactions between the E. coli PspA, PspB, and PspC proteins do appear to occur (1).

The gastrointestinal human pathogen Yersinia enterocolitica has a homologous psp locus, although the pspA operon (pspABCDycjXF) has a different structure than that of E. coli (pspABCDE). The Y. enterocolitica psp locus is essential for virulence (7, 8). In the laboratory, expression of the pspA operon is induced when the Ysc type III secretion system is produced and, in this case, the inducing trigger appears to be mislocalization of the YscC secretin (8). YscC mislocalization causes a growth arrest in some psp null mutants. The severity of this growth defect varies between different psp null mutants, and this correlates well with their virulence defects (8). Therefore, during host infection it was hypothesized that a growth defect caused by secretin production is the likely explanation for the attenuation of some psp null mutants. Another conclusion from this genetic analysis of the Y. enterocolitica psp system was that PspF might regulate at least one other locus in addition to the pspA operon (8). The most compelling evidence for this was that the YscC-induced growth defect of a complete pspA operon deletion mutant was significantly exacerbated by a pspF null mutation. Furthermore, a ΔpspF ΔpspA operon mutant had a significantly more severe virulence defect than a strain with the ΔpspA operon mutation alone. Both strains lacked all genes of the pspA operon, and the only difference between them was the presence or absence of the PspF transcriptional activator.

In this study we investigated the hypothesis that at least one other gene is coordinately regulated with the pspA operon and that this gene plays a role in response to secretin-induced stress. We report the identification of pspG, a new member of the Y. enterocolitica phage shock protein regulon. As expected, a pspG null mutation exacerbates the phenotypes of a pspA operon deletion mutant.

MATERIALS AND METHODS

Bacterial strains, plasmids, and routine growth conditions.

Bacterial strains and plasmids used in this study are shown in Table 1. For routine plasmid manipulations, the host strain was E. coli DH5α. Plasmids with an R6K ori were maintained in E. coli CC118 λpir and conjugated into Y. enterocolitica from either E. coli S17-1 λpir or E. coli SM10 λpir. E. coli strains were grown at 37°C, and Y. enterocolitica strains were grown at 26 or 37°C as noted. Strains were routinely grown in Luria-Bertani (LB) broth or on LB agar plates (26). Antibiotics were used at the following concentrations: ampicillin (200 μg ml⁻¹), streptomycin (50 μg ml⁻¹), spectinomycin (50 μg ml⁻¹), nalidixic acid (20 μg ml⁻¹), trimethoprim (100 μg ml⁻¹), kanamycin (75 μg ml⁻¹ for E. coli; 100 μg ml⁻¹ for Y. enterocolitica), chloramphenicol (25 μg ml⁻¹ for E. coli; 12.5 μg ml⁻¹ for Y. enterocolitica).

TABLE 1.

Strains and plasmids

Strain or plasmid	Genotype or features	Reference or source
E. coli K-12 strains
DH5α	φ80d Δ(lacZ)M15 Δ(argF-lac)U169 endA1 recA1 hsdR17(rK− mK+) deoR thi-1 supE44 gyrA96 relA1	Gibco BRL
CC118 λpir	Δ(ara-leu) araD ΔlacX74 galE galK phoA20 thi-1 rpsE rpoB argE (Am) recA1 λpir lysogen	18
S17-1 λpir	recA thi pro hsdR−M+ RP4::2-Tc::Mu::Km Tn7 λpir lysogen	27
SM10 λpir	thi-1 thr leu tonA lacY supE recA::RP4-2-Tc::Mu λpir lysogen	27
Y. enterocolitica strainsa
JB580v	ΔyenR (R− M+) pYV+	22
AJD451	ΔyenR::kan (R− M+) pYV+	This study
YVM576	φ(pspA-lacZYA)	8
YVM761	φ(pspA-lacZYA) ΔpspA	8
AJD85	φ(pspA-lacZYA) ΔpspF::kan	—c
AJD344	φ(YPO2615b-lacZYA)	This study
AJD345	φ(YPO2615b-lacZYA) ΔpspA	This study
AJD346	φ(YPO2615b-lacZYA) ΔpspF::kan	This study
AJD347	φ(YPO2615b-lacZYA) ΔpspA ΔpspF::kan	This study
AJD384	φ(pspG-lacZYA)	This study
AJD385	φ(pspG-lacZYA) ΔpspA	This study
AJD386	φ(pspG-lacZYA) ΔpspF::kan	This study
AJD387	φ(pspG-lacZYA) ΔpspA ΔpspF::kan	This study
YVM712	ΔpspA	8
YVM790	ΔpspF	—c
AJD568	ΔpspG::tp	This study
YVM713	Δ(pspA-ycjF)::kan	8
YVM824	Δ(pspA-ycjF)::kan ΔpspF	This study
AJD571	Δ(pspA-ycjF)::kan ΔpspG::tmp	This study
AJD573	ΔpspF ΔpspG::tmp	This study
AJD574	Δ(pspA-ycjF)::kan ΔpspF ΔpspG::tmp	This study
Plasmids
pBAD18-Km	Kmr, araBp expression vector, Col E1 ori	16
pQE30	Apr, Col E1 ori, T5p expression vector for His6 proteins	Qiagen
pFUSE	Cmr, mob+ (RP4), lacZYA+ operon fusion suicide vector R6K ori	3
pKN8	Bgl/II linker in SmaI site of pFUSE	14
p34S-Tp	Apr Tpr, pUC ori	11
p34S-Km	Apr Kanr, pUC ori	11
pVLT35	Smr/Spr, tacp expression vector, RSF1010 ori	10
pAJD110	∼0.5-kb pspA′ fragment in pFUSE	8
pAJD651	∼0.4-kb YPO2615′b XbaI-BamHI fragment in pKN8	This study
pAJD651	∼0.4-kb pspG′ XbaI-BamHI fragment in pKN8	This study
pAJD126	tacp-yscC+ in pVLT35	8
pAJD555	tacp-ysaC+ in pVLT35	25
pAJD298	pspF BamHI PCR fragment in pQE30	This study
pAJD624	His6-pspF EcoRI PCR fragment of pAJD298 in pBAD18-Km	This study

^aAll Y. enterocolitica strains shown have the virulence plasmid (pYV). Some pVY⁻ derivatives of these strains were made, as described previously (8), but are not listed in the table.

^bYPO numbers refer to the Y. enterocolitica ortholog of the indicated Y. pestis CO92 gene.

^cMaxson and Darwin, unpublished.

β-Galactosidase assays.

To determine the effect of pspA and pspF null mutations on lacZ operon fusion expression, strains were grown on a roller drum at 26°C overnight in LB broth. They were diluted into 4 ml of LB broth in 18-mm-diameter test tubes so that the optical density at 600 nm was ∼0.08 and grown on a roller drum at 26°C to mid-exponential phase. Cells were collected by centrifugation and washed with 0.88% NaCl prior to enzyme activity assays.

To determine the effects of YscC and YsaC production on Φ(pspA-lacZ) and Φ(pspG-lacZ) expression, strains were grown on a roller drum at 26°C overnight in LB broth containing streptomycin and spectinomycin. They were diluted into 5 ml of the same medium in 18-mm-diameter test tubes so that the optical density at 600 nm was ∼0.04. The cultures were grown on a roller drum at 37°C for 2 h. Then, for strains containing the yscC expression plasmid, 0.2 mM (final concentration) isopropyl-β-d-thiogalactopyranoside (IPTG) was added to induce tacp promoter expression. Growth continued at 37°C for two more hours prior to harvest as described above. IPTG was not added to strains containing the ysaC expression plasmid because it caused a growth arrest in pspF null strains. However, even without IPTG ysaC was expressed sufficiently to induce Φ(pspA-lacZ) and Φ(pspG-lacZ) expression. The presence or absence of IPTG had no effect on Φ(pspA-lacZ) and Φ(pspG-lacZ) expression in strains containing the vector plasmid pVLT35 (data not shown).

β-Galactosidase enzyme activity was determined at room temperature (approximately 22°C) in permeabilized cells as described previously (24). Activities are expressed in arbitrary units, which were determined according to the formula of Miller (26). Individual cultures were assayed in duplicate, and reported values were averaged from at least three independent cultures.

Construction of lacZYA operon fusion strains.

Control region fragments were amplified from chromosomal DNA by PCR. These fragments included the complete intergenic region upstream of each gene. For each primer pair, one primer incorporated an XbaI site and the other incorporated a BglII site. The fragments were cloned into the pFUSE derivative pKN8 (Table 1), and the DNA sequence of each fragment was checked. These fusions were integrated onto the Y. enterocolitica chromosome by homologous recombination. Following integration, merodiploid strains result that encode intact genes under the control of their native promoters, in addition to the lacZYA operon fusions. Correct integration was checked by Southern hybridization analysis (data not shown).

Purification of His₆-PspF.

A virulence plasmid-free derivative of Y. enterocolitica strain YVM712 containing plasmid pAJD624 was grown to mid-exponential phase at 26°C (1-liter culture volume). Arabinose was added (0.2% final concentration), and incubation continued for a further 3 h. Bacterial cells were collected by centrifugation, frozen at −20°C, and then resuspended in 20 ml of 50 mM NaH₂PO₄, 300 mM NaCl, and 10 mM imidazole (pH 8.0) containing 1 mg of lysozyme/ml. After 30 min on ice, the cells were disrupted by sonication and the soluble and insoluble fractions were separated by centrifugation. The soluble crude extract (supernatant) was incubated with 5 ml of Ni-nitrilotriacetic acid-agarose (QIAGEN) for 1 h at 4°C and then poured into a column. The column was washed with 20 ml of 50 mM NaH₂PO₄, 300 mM NaCl, and 20 mM imidazole (pH 8.0). The His₆-PspF protein was eluted with 10 ml of 50 mM NaH₂PO₄, 300 mM NaCl, and 250 mM imidazole (pH 8.0), collected in 1-ml fractions, and used directly in gel mobility assays. Protein concentrations were estimated with a Bio-Rad protein assay kit (Bradford method) using bovine serum albumin as the standard.

Gel mobility shift assays.

The pspA and pspG control regions were amplified by PCR as approximately 150-bp fragments, which included the predicted RpoN-binding site and all of the upstream noncoding DNA. Gel mobility shift assays were done with the DIG gel shift kit (Roche) according to the manufacturer's instructions. Briefly, 1.5 ng of digoxigenin (DIG)-labeled control region fragment was incubated with the indicated concentration of His₆-PspF for 15 min at room temperature. The binding buffer contained 20 mM HEPES (pH 7.6), 1 mM EDTA, 10 mM (NH₄)₂SO₄, 1 mM dithiothreitol, 0.2% (wt/vol) Tween 20, 30 mM KCl, and 50 ng of poly(l-lysine)/μl. Each reaction also contained 1,000 ng of poly(dI-dC) as nonspecific competitor DNA. Reaction mixtures were separated by nondenaturing 6% polyacrylamide electrophoresis, and DIG-labeled DNA was detected according to the manufacturer's instructions.

Construction of pspG null mutant strains.

Two fragments were amplified by PCR. One fragment had a BglII site followed by the first four codons of pspG and approximately 0.5 kb of upstream DNA. The other fragment had a BglII site followed by the last two codons of pspG and approximately 0.5 kb of downstream DNA. The DNA sequences of both PCR fragments were confirmed. These fragments were ligated at the BglII site and cloned into the pEP185.2 suicide vector. The ∼0.6-kb trimethoprim resistance fragment of p34s-Tp was then cloned into the BglII site so that it was in the same orientation as the deleted pspG gene (however, polarity is not a concern, because the gene downstream of pspG is in the opposite orientation). This plasmid was transferred to Y. enterocolitica strains by conjugation, and Tp^r Cm^s exconjugants were isolated. The presence of the ΔpspG::tmp mutation was confirmed by Southern hybridization analysis (data not shown).

Construction of ΔyenR::kan strain AJD451.

Two fragments were amplified by PCR. One fragment had a BglII site followed by approximately 0.5 kb of DNA upstream of the ΔyenR mutation in strain JB580v. The other fragment had a BglII site followed by approximately 0.5 kb of DNA downstream of the ΔyenR mutation in JB580v. The DNA sequences of both PCR fragments were confirmed. These fragments were ligated at the BglII site and cloned into the pEP185.2 suicide vector. The ∼1.3-kb kanamycin resistance fragment of p34s-Km was then cloned into the BglII site so that it was in the same orientation as the deleted yenR gene. This plasmid was transferred to Y. enterocolitica strains by conjugation, and Kan^r Cm^s exconjugants were isolated. The presence of the ΔyenR::kan mutation was confirmed by Southern hybridization analysis (data not shown).

Growth curves.

The growth curve experiments were done as described previously (8). Briefly, saturated cultures were diluted into 5 ml of LB broth in 18-mm-diameter test tubes so that the initial optical density (600 nm) was approximately 0.1. Culture medium also contained streptomycin, spectinomycin, and 0.2 mM IPTG to induce yscC expression. The cultures were grown on a roller drum at 37°C, and a 0.1-ml aliquot was removed at hourly intervals for optical density determination. All experiments were done on at least three separate occasions to ensure reproducibility.

Mouse competition assays.

Female BALB/c mice were obtained from Charles River Laboratories and used when 6 to 7 weeks old. Bacterial strains were grown to saturation in LB broth and washed twice in phosphate-buffered saline. They were then diluted to the desired optical density prior to animal infections. Mice were infected by intraperitoneal injection with 0.2 ml of a 10⁷-CFU/ml bacterial suspension that consisted of an equal mixture of the two strains being tested. A dilution of the inoculum was also spread onto an LB agar plate, and at least 100 colonies were patched onto agar containing kanamycin to determine the actual ratio of the two strains. Mice were euthanized 48 h after infection, and spleens were recovered and homogenized in phosphate-buffered saline. Dilutions were plated onto LB with nalidixic acid only (to measure total CFU) and then replica plated or patched to LB with nalidixic acid and kanamycin (to determine mutant CFU). Alternatively, dilutions were plated separately on LB with nalidixic acid only and LB with nalidixic acid and kanamycin. The competitive index (CI) is defined as the output ratio (mutant/wild type) divided by the input ratio (mutant/wild type).

One of two different psp⁺ (wild-type) strains was used in each of these assays, JB580v or AJD451. The only difference between these two strains is that JB580v has the ΔyenR deletion (which eliminates a restriction endonuclease), whereas AJD451 has exactly the same deletion but with a kanamycin resistance gene inserted at the deletion point. There is no difference in the virulence of these two strains (see Fig. 6 and data not shown). When the psp mutant strain had the Δ(pspA-ycjF)::kan mutation, the JB580v wild-type strain was used. In this case, the mutant colonies could be identified by selection for kanamycin resistance. Therefore, the limit of detection was one mutant colony per spleen (usually approximately 1 in 10⁵ total recovered bacteria). However, when the psp mutant strain did not have the Δ(pspA-ycjF)::kan mutation, the kanamycin-resistant AJD451 wild-type strain was used. In this case, the kanamycin-sensitive mutant colonies had to be identified by replica plating or patching colonies onto plates with and without kanamycin. Therefore, the limit of detection was not as sensitive (usually 1 in 200 total recovered bacteria). In cases where no mutant bacteria were recovered from an animal, the CI was calculated assuming that one mutant bacterium had been recovered. To assess the statistical significance of the results a two-tailed, unpaired Student's t test was used to compare the CI data sets from the two experiments under consideration.

FIG. 6.

Mouse competition assays between wild-type and psp null strains. Mice were infected by intraperitoneal injection with 2 × 10⁶ CFU, which consisted of an equal mixture of a wild-type and a mutant strain. Each circle represents the CI derived from a single mouse. CI = (mutant/wild type output ratio)/(mutant/wild type input ratio). Open circles indicate that the mutant was below the limit of detection (see Materials and Methods). The double line represents the geometric mean of the CI values (from four to five total mice). The wild-type strains used were either JB580v (ΔyenR) or AJD451 (ΔyenR::kan) as described in Materials and Methods. The wild type competition assay was between these two wild-type strains. ΔG, ΔpspG::tmp; ΔF, ΔpspF; ΔAop, Δ(pspA-ycjF)::kan. When comparing the CI data sets of pairs of different mutants, as described in Materials and Methods, the P values were as follows: ΔpspF Δ(pspA-ycjF)::kan compared to Δ(pspA-ycjF)::kan, P = 0.008; ΔpspG Δ(pspA-ycjF)::kan compared to Δ(pspA-ycjF)::kan, P = 0.018; ΔpspF Δ(pspA-ycjF)::kan compared to ΔpspG Δ(pspA-ycjF)::kan, P = 0.35.

The Institutional Animal Care and Use Committee of New York University School of Medicine approved all animal procedures.

RESULTS

Identification of putative RpoN-dependent promoters.

Previous data suggested that PspF positively regulates at least one unknown Y. enterocolitica gene (8). PspF is a member of the EBP family of transcriptional activators, which only activate RpoN (σ⁵⁴)-dependent promoters. Therefore, we reasoned that any unidentified PspF-dependent genes should also be RpoN dependent. The RpoN-binding site sequence, commonly referred to as a −24/−12 promoter, is highly conserved among different bacterial species (2). Therefore, we began our search by assembling a list of putative RpoN-dependent promoters encoded in the Y. enterocolitica genome.

Promscan software, developed by David Studholme, has been made freely available for use via the internet (http://www.promscan.uklinux.net/home.html). Promscan uses the RpoN-binding site consensus sequence to search for close matches in a DNA sequence of interest. The Y. enterocolitica genome sequence was obtained from the Sanger Centre (http://www.sanger.ac.uk/Projects/Y_enterocolitica/) and searched with Promscan. The closest 800 matches were selected for further analysis. We analyzed the DNA regions surrounding each of these to identify those that were in noncoding regions immediately upstream of a gene (the Y. enterocolitica genome sequence annotation had not been released at this time). This resulted in a list of 34 putative RpoN-dependent promoters (Table 2). Although this analysis may have missed some genuine RpoN-dependent promoters, the number correlates well with the predicted number of RpoN-dependent promoters in E. coli K-12, which has been estimated to be around 30 (29). The list of putative RpoN-dependent genes includes pspA and homologs of several genes that are known or predicted to be RpoN dependent in other organisms (Table 2).

TABLE 2.

Putative Y. enterocolitica RpoN-binding sites in intergenic regions^a

Sequenceb	Relative scorec	Downstream gened	Function or Commente
TGGCACGGCGATTGCG	100.0	YPO1187 ortholog	Putative Leu/Ile/Val/Thr-binding protein of ABC transporter
TGGCACAGACCTTGCA	98.8	YPO2615 ortholog	Putative amino acid-binding protein of ABC transporter
TGGCACTCCATTTGCT	97.2	EC yhdW ortholog	Putative amino acid-binding protein of ABC transporter
TGGCACGAAAGCTGCT	93.5	YP astC ortholog	Succinylornithine aminotransferase
TGGCATAGCCTTTGCT	93.2	YP glnK ortholog	Nitrogen regulatory protein P-II
TGGCACGATTACTGCA	93.0	YP hisJ ortholog	Histidine-binding protein of ABC transporter
TGGCACAGTTCATGCT	92.8	RSp1575 homolog	Putative amino acid-binding protein of ABC transporter
TGGCATTGCGCTTGCA	92.4	YP fhuC ortholog	Component of ferrichrome ABC transporter
TGGCACGACTCTTGAT	92.1	YE pspA	First gene of phage shock protein operon
TGGCAGCATTCTTGCT	91.1	YE ysrR	Putative response regulator, part of Ysa TTSSf locus
TGGCAGGTTTATTGCC	90.8	YE yspA	Putative secreted effector protein of Ysa TTSS
TGGCACAGAATATGCT	90.5	EC hyfA/hycB homolog	Putative Fe-S protein, part of a hydrogenase
TGGCAAAGTAGTTGCA	90.4	YE ompH	Outer membrane protein
TGGCATCGTCTTTGCA	89.7	EC hypA homolog	Putative hydrogenase maturation protein
TGGCATTAGCTTTGCT	89.2	EC ycdM ortholog	Putative monooxygenase
TGGCTCTCTGCTTGCT	89.1	YP maeB ortholog	NADP-dependent malate dehydrogenase
TGGCACGTTTCTTGTA	89.1	YPO0318 ortholog	Putative inner membrane protein, unknown function
TGGCCCAAGAGTTGCT	88.9	YPO1155 ortholog	Putative component of amino acid ABC transporter
TGGCACAGTTTATGCG	88.2	PS PA1730 ortholog	Putative cytoplasmic protein, unknown function
TGGCATCTTATTTGCT	88.1	YPO1355 ortholog	Putative nucleotide di-P-sugar epimerase or dehydratase
TGGCACGATCCTTGGG	86.7	YPO2585 ortholog	Putative carbohydrate kinase
TGGCACGATCTTTTCA	86.9	YP glnH ortholog	Glutamine-binding protein of ABC transporter
TGGCAATAATATTGCC	86.9	EC tnp ortholog	Putative transposase
TGGCAATTTATTTGCA	86.8	YPO2449 ortholog	Putative LuxR family transcriptional regulator
TGGCCGGTTTTTTGCT	86.1	YPO2986 ortholog	Putative inner membrane protein, unknown function
TGGCGTGTTTTTTGCA	85.5	YP asnB ortholog	Asparagine synthetase B
TGGCGTGAATTTTGCG	94.3	YP rmf ortholog	Ribosome modulation factor
TGGCACCCCGTTTACT	84.2	YP gntR ortholog	Gluconate utilization system Gnt-I transcriptional repressor
TGGAACCAGCCTTGCG	84.2	ST STM2763 ortholog	Putative prophage integrase
TGGCTTTCCTTTTGCT	84.2	YP speC ortholog	Ornithine decarboxylase
TGGCTCTTTAGTTGCG	83.9	YP rimJ ortholog	Ribosomal protein alanine acetyltransferase
TGACACAATTTTTGCA	83.5	YPO1749 ortholog	Putative cytoplasmic protein, unknown function
TGGCATAAACCATGCA	83.2	EC hydN homolog	Putative Fe-S protein involved in formate dehydrogenase H activity
TGGCGTGTGTTTTGCG	83.2	YP ggt ortholog	Gamma-glutamyltranspeptidase

^aIdentified by Promscan software (http://www.promscan.uklinux.net/home.html), using the consensus sequence YTGGCACGrNNNTTGCW. Only sites in intergenic regions and upstream of genes in the correct orientation are shown.

^bDNA sequence of the putative RpoN-binding site.

^cThe “Promscan” software assigns a score to each sequence based on the relative base-frequencies at each position within known RpoN-binding sites (higher score indicates a closer match to the consensus). For this table, the highest score given was arbitrarily set to 100 and other scores are expressed relative to this.

^dYE = Yersinia enterocolitica, YP = Yersinia pestis, EC = Escherichia coli, ST = Salmonella enterica serovar Typhimurium, PS = Pseudomonas aeruginosa, RS = Ralstonia solanacearum. Ortholog = at least 65% identity, homolog = less than 65% identity (these identity values were arbitrarily chosen for the purposes of this table).

^eBased on sequence analysis or characterization of orthologs/homologs.

^fTTSS, type III secretion system.

Common signature for most PspF-dependent promoters.

The most highly conserved positions of the RpoN-binding site consensus are the GG and GC dinucleotides of the −24 and −12 regions, respectively (2). In a 1999 review article, it was noted that the C at position −12 is conserved in 96% of RpoN-dependent promoters with a mapped transcriptional start site (2). There were only four promoters without a C at this position, and one of these was the E. coli pspA promoter. Subsequently it was found that, as in E. coli, the predicted RpoN-binding site of the Y. enterocolitica pspA promoter also lacked the C at position −12 (8).

We hypothesized that a nonconsensus nucleotide at position −12 might be a feature shared by pspA operon promoters. If so, it might also be the case for other PspF-dependent promoters. Therefore, we analyzed the DNA sequence immediately upstream of the pspA genes from a number of bacteria (Fig. 1). In six out of seven cases, the predicted RpoN-binding site did not have the C at position −12. One exception was Shigella flexneri serotype 2a strain 301. However, examination of another genome sequence revealed that this was not the case for all S. flexneri strains (Fig. 1). This lack of the consensus −12 position C in pspA promoters is striking, given the high degree of conservation of this position in other RpoN-dependent promoters from various bacteria (2). It is unlikely that this has occurred by chance.

FIG. 1.

RpoN-binding sites of pspA and pspG promoters have a nonconsensus −12 dinucleotide. The core of the RpoN-binding site consensus sequence is shown at the top, with the highly conserved −24 and −12 dinucleotides overlined. Putative or known RpoN-binding sites upstream of pspA and pspG orthologs from various bacterial species are shown. The nucleotide that diverges from the −12 consensus is boxed. Sequence data were obtained from the website http://www.ncbi.nlm.nih.gov.

Next, the list of putative Y. enterocolitica RpoN-dependent promoters was analyzed to determine how many did not have a C at position −12. Besides pspA, two other putative promoters had this property (Table 2). One of these sequences is upstream of an ortholog of the Yersinia pestis CO92 gene YPO2585. However, this sequence is over 160 bp from the downstream gene. Furthermore, we could not identify a putative RpoN-binding site upstream of homologs of this gene in other bacteria, including Y. pestis. Therefore, it is unlikely that this is a genuine RpoN-dependent promoter. The second sequence is 83 bp upstream of a gene encoding an ortholog of the Y. pestis CO92 gene YPO0318 and the E. coli K-12 gene yjbO. We examined the DNA sequence upstream of orthologs of this gene in the same group of bacteria used for the pspA promoter analysis. In all seven cases we were able to identify a putative RpoN-binding site, which increased the probability that it is authentic (Fig. 1). Furthermore, all seven sequences did not have the C at position −12. Therefore, we identified the YPO0318 ortholog as a candidate to be a member of the Psp regulon. This gene is apparently monocistronic, because the downstream gene is in the opposite orientation (data not shown). It encodes a very hydrophobic 8.2-kDa protein that may be located in the cytoplasmic membrane. We named the YPO0318 ortholog pspG, because subsequent analysis demonstrated that it is in the Psp regulon.

The pspG gene is negatively regulated by PspA and positively regulated by PspF.

A characteristic of the pspA promoter is that its expression is significantly elevated in a pspA null mutant because PspA can no longer inhibit PspF activity (8). To determine whether pspG regulation followed a similar pattern, we compared the effects of pspA, pspF, and pspF pspA null mutations on expression of single-copy Φ(pspG-lacZ) and Φ(pspA-lacZ) operon fusions (Fig. 2).

FIG. 2.

Effect of pspF and pspA null mutations on expression of the pspA, pspG, and YPO2615 ortholog promoters. Strains were grown and β-galactosidase activities were determined as described in Materials and Methods. Sp. Act., specific activity; ND, not determined. WT, PspA⁺ PspF⁺; F⁻, PspA⁺ PspF⁻; A⁻, PspA⁻ PspF⁺; A⁻ F⁻, PspA⁻ PspF⁻. The Φ(YPO2615-lacZ) fusion strain provides an example of a fusion that was not affected by pspA or pspF null mutations. Other PspA/PspF-independent fusions behaved similarly (data not shown).

The Φ(pspG-lacZ) and Φ(pspA-lacZ) operon fusions were regulated similarly by PspA and PspF (Fig. 2). Expression was significantly elevated in a pspA null mutant, although the magnitude of this effect was somewhat lower for Φ(pspG-lacZ) than for Φ(pspA-lacZ). For the Φ(pspG-lacZ) fusion, this increased expression did not occur in a pspA pspF double null strain, indicating that it required a functional PspF protein. Φ(pspA-lacZ) expression was not measured in a pspA pspF double null strain because the double deletion makes it difficult to integrate the Φ(pspA-lacZ) fusion by homologous recombination.

These results indicated that pspG expression is negatively regulated by PspA in a PspF-dependent manner. This was the first indication that pspG might be coordinately regulated with the pspA operon.

Most putative RpoN-dependent promoters are not regulated by PspA or PspF.

Most PspF-dependent promoters apparently have a nonconsensus nucleotide at position −12 (Fig. 1). However, S. flexneri 2a strain 301 proves that there may be exceptions to this rule (although the regulation of that pspA promoter has not been studied). It is certainly possible that some PspF-dependent promoters have a C at position −12. Therefore, we also tested the effect of pspA and pspF null mutations on the expression of some other putative RpoN-dependent promoters. Many of the promoters in Table 2 have homologs in other species that are regulated by EBPs other than PspF. We focused our efforts on promoters for which the regulatory EBP could not be predicted and especially on putative promoters upstream of predicted cell envelope components. We also concentrated on those putative RpoN-binding sites that were most likely to be part of functional RpoN-dependent promoters, using two criteria: first, those RpoN-binding sites with a noncoding region upstream that was 100 bp or more (to permit binding of an EBP), and second, those RpoN-binding sites with 100 bp or less between the site and the downstream gene. Putative promoter regions were amplified by PCR and used to construct single-copy or multicopy lacZ operon fusions. We then measured the expression of these fusions in pspA null and pspA pspF double null strains. Besides Φ(pspA-lacZ) and Φ(pspG-lacZ), none of the other fusions tested was affected by pspA and pspF null mutations (data not shown). This allowed us to eliminate several other promoters as members of the Psp regulon (promoters upstream of orthologs of YPO1355, YPO2986, YPO2449, YPO1749, PA1730, ycdM, rmf, and rimJ) (Table 2).

We had also originally attempted to identify PspF-dependent genes with a transposon mutagenesis screen to find lacZ operon fusions that responded to overproduced PspF. This screen was discontinued because it identified putative RpoN-dependent promoters that responded to overexpressed PspF but not to physiological concentrations of PspF (data not shown). However, the results of this screen also allowed us to eliminate the putative promoters upstream of orthologs of YPO2615 and hyfA as members of the Psp regulon, because their expression was unaffected by pspA and pspF null mutations. The analysis of the Φ(YPO2615-lacZ) fusion strains is included in Fig. 2 for comparative purposes.

pspG expression is induced by secretins.

Expression of the pspA operon is induced by overproduction of the YscC secretin protein (8). We have also found that several other Y. enterocolitica secretins can also induce Φ(pspA-lacZ) expression and that the YsaC secretin is particularly potent (25). YsaC is a component of a chromosomally encoded type III secretion system (17). Furthermore, secretin overexpression might be a specific signal for the Psp regulon, because it does not induce two other extracytoplasmic stress responses, the RpoE and Cpx systems (25). Therefore, we compared the effects of the YscC and YsaC secretins on Φ(pspA-lacZ) and Φ(pspG-lacZ) expression in pspF⁺ and pspF null strains (Fig. 3).

FIG. 3.

The pspA and pspG promoters are induced by secretin overexpression. The tac promoter expression plasmid pVLT35 (−) or derivatives encoding the yscC or ysaC genes were transferred into PspF⁺ (black bars) or PspF⁻ (white bars) Φ(pspA-lacZ) and Φ(pspG-lacZ) operon fusion strains. Strains were grown and β-galactosidase activities were determined as described in Materials and Methods. Sp. Act., specific activity.

The pspA and pspG promoters showed a similar response to secretin overexpression. There was modest induction by yscC and much more significant induction by ysaC. In both cases, a pspF null mutation abolished the response to secretin overexpression (Fig. 3) and an rpoN null mutation had a similar effect (data not shown). These results provide further evidence that pspG is coordinately regulated with pspA and that this regulation only occurs in a pspF⁺ strain.

PspF interacts with the pspA and pspG control regions.

Our data indicated that pspG is a member of the Psp regulon. This was most likely due to direct control by PspF. To confirm this, we purified His₆-PspF and tested its ability to bind to pspA and pspG control region fragments in a gel mobility shift assay. Each control region fragment was approximately 150 bp in length and included the predicted RpoN-binding site and all noncoding DNA upstream. As shown in Fig. 4, 50 nM His₆-PspF was sufficient to retard the mobility of both the pspA and pspG control region fragments in the presence of a 700-fold excess of nonspecific competitor DNA. Less than 50 nM His₆-PspF did not produce a mobility shift for either control region (data not shown). In both cases, the interaction was inhibited by the addition of a 150-fold excess of the unlabeled control region fragment (specific competitor) (Fig. 4). In addition, the DNA sequences of the pspA and pspG control regions share some similarity in the region predicted to be bound by PspF, identified by alignment with the PspF-binding site of the E. coli pspA control region (data not shown and reference 20). Therefore, we conclude that pspA and pspG are both members of the PspF regulon.

FIG. 4.

PspF interacts with the pspA and pspG control regions. Gel mobility shift assays were done as described in Materials and Methods. Each reaction mixture contained 1.5 ng of DIG-labeled pspA or pspG control region fragment and the indicated concentration of His₆-PspF. All reaction mixtures contained 1,000 ng of poly(dI-dC) as a nonspecific competitor. Additional control reaction mixtures included 225 ng of the unlabeled pspA or pspG control region fragment as a specific competitor (Sp. Comp.).

pspG plays a role in responding to stress caused by secretin production.

Production of the YscC secretin causes a growth defect in some psp null mutants (8). In a complete ΔpspA operon deletion mutant, the severity of this growth defect was exacerbated by a pspF null mutation, leading to our original hypothesis of an unidentified PspF-dependent gene. If this gene were pspG, then a pspG null mutation should also exacerbate the YscC-induced growth defect of a ΔpspA operon mutant.

To test this hypothesis, we constructed a pspG deletion mutation and introduced it into strains with psp⁺, Δ(pspA-ycjF), and ΔpspF alleles. We then determined the effect of a tacp-yscC expression plasmid on the growth of these strains (Fig. 5). A Δ(pspA-ycjF) mutant had a moderate growth defect, reaching an optical density (600 nm) of approximately 1.5, compared to over 3.0 for the wild-type strain. This growth defect was significantly exacerbated by a ΔpspF mutation, as previously described (Fig. 4) (8). As predicted, the ΔpspG mutation also exacerbated the growth defect of a Δ(pspA-ycjF) mutant, such that the phenotypes of Δ(pspA-ycjF) ΔpspF and Δ(pspA-ycjF) ΔpspG strains were indistinguishable (Fig. 5).

FIG. 5.

Effect of yscC overexpression on growth of psp null strains. The tac promoter expression plasmid pVLT35 (▪) or the pAJD126 derivative encoding yscC (⧫) was transferred into various strains (genotypes indicated above each graph). The strains were grown as described in Materials and Methods, and optical density was measured at hourly intervals. The data are from a single experiment in which all strains were tested simultaneously (however, the experiment was done on three separate occasions to ensure reproducibility).

The growth defect of a ΔpspF mutant was reproducibly not as severe as that of a Δ(pspA-ycjF) ΔpspG mutant. This was probably because of residual low-level expression of the pspA operon and/or pspG in the absence of PspF. In support of this, we have identified putative RpoN- and PspF-independent transcription initiation sites in the pspA operon (M. E. Maxson and A. J. Darwin, unpublished data).

A pspG null mutation alone did not cause a growth defect when YscC was overproduced at 37°C (Fig. 5). This is not inconsistent with our original hypothesis, which predicted that a ΔpspG mutation would cause a phenotype in a Δ(pspA-ycjF) strain, but made no prediction of a phenotype in a pspABCDycjXF⁺ strain (see Discussion). We also investigated the possibility that PspG may be more important when the chromosomal Ysa type III secretion system is active (L broth with 290 mM NaCl at 26°C [32]). However, neither Δ(pspA-ycjF) nor ΔpspG null mutants had a growth defect under these conditions or when YscC was overproduced under these conditions (data not shown).

From these experiments, we conclude that a pspF null mutation exacerbates the YscC-induced growth defect of a Δ(pspA-ycjF) mutant because it reduces pspG expression. Therefore, not only is pspG coordinately regulated with the pspA operon, but it also plays a functional role in responding to secretin-induced stress, at least in a Δ(pspA-ycjF) strain.

Analysis of psp null mutants in a systemic mouse model of infection.

It was observed previously that a pspF null mutation increased the virulence defect of a Δ(pspA-ycjF) mutant. Therefore, we investigated the virulence defects of strains with various combinations of Δ(pspA-ycjF), ΔpspF, and ΔpspG alleles. Specifically, we hypothesized that a pspG null mutation would increase the virulence defect of a Δ(pspA-ycjF) mutant. Consistent with previous Y. enterocolitica psp studies (7, 8), virulence defects were assessed by mouse competition assays. Briefly, mice were infected by intraperitoneal injection with an approximately equal mixture of the wild type and a psp mutant strain. Forty-eight hours later, the ratio of mutant to wild-type bacteria in the spleen was determined. Virulence is expressed in terms of the CI as described in Materials and Methods (a CI of less than 1.0 indicates that the test strain is less virulent than the wild type). Note that none of the psp null mutant strains had general in vitro growth defects (data not shown and Fig. 5, vector control strains).

For the psp null mutants, there was a general correlation between the relative severity of their YscC-induced growth defects (Fig. 5) and their virulence defects (Fig. 6). As a negative control, a competition assay between two different psp⁺ strains gave an average CI of approximately 1.0 (wild type in Fig. 6). The Δ(pspA-ycjF) mutant had a mean CI of 0.11 (P = 0.003). This virulence defect was significantly increased by either a pspF null mutation or a pspG null mutation. The ΔpspF Δ(pspA-ycjF) and ΔpspG Δ(pspA-ycjF) mutants had mean CIs of 0.00004 and 0.0013, respectively. In both cases, the difference from the CI of the Δ(pspA-ycjF) mutant was statistically significant (Fig. 6 legend). However, note that the difference between the CIs of the Δ(pspA-ycjF) ΔpspF and Δ(pspA-ycjF) ΔpspG mutants was not statistically significant when the data sets from those two experiments were compared (P = 0.35).

The ΔpspF in-frame deletion mutant had a more severe virulence phenotype than a pspF insertion mutant described previously (8). We suspect this is because the previous pspF null mutation was made by inserting a suicide plasmid into the pspF gene. It is possible that a promoter encoded within the suicide plasmid allowed low-level expression of the divergent pspA operon. However, even if this did occur we note that the major conclusions of the previous study are unaffected. Both pspF mutations exacerbate the growth and virulence defects of a Δ(pspA-ycjF) mutant.

These animal experiments suggest that at least one reason why a pspF null mutation exacerbates the virulence defect of a Δ(pspA-ycjF) mutant is that it reduces pspG expression during infection. Therefore, we conclude that pspG is important for virulence, but only in a Δ(pspA-ycjF) strain.

DISCUSSION

The role of the Psp system is unknown, although it has been proposed to sense and respond to dissipation of the proton motive force (23). Microarray studies have revealed that the E. coli pspA operon is induced during biofilm formation (4) and that the Salmonella enterica pspA operon is induced during macrophage infection (15). The Psp system of Y. enterocolitica is essential for virulence, apparently because it responds to stress during the assembly of the Ysc type III secretion system (8). In this study we followed up on a hypothesis that there was at least one unknown member of the Y. enterocolitica Psp regulon. We describe the pspG gene, which is a new member of the Psp stress response regulon.

Our data indicate that the pspA and pspG promoters are coordinately regulated. Both are highly expressed in a pspA null mutant (Fig. 2), and both show PspF-dependent induction in response to secretin overexpression (Fig. 3). In addition, the His₆-PspF protein specifically interacts with the pspA and pspG control regions. Some support for the coordinate regulation of pspA and pspG in other bacteria comes from the microarray analysis described above. Both loci were induced in E. coli during biofilm formation (4) and in S. enterica during macrophage infection (15). It is also interesting that null mutations within the E. coli pspA operon or of the E. coli pspG homolog (yjbO) did not affect biofilm formation, whereas a pspF null mutation did (4). Perhaps this is because the pspF null mutation simultaneously reduced expression of both the pspA operon and pspG.

PspG is predicted to be a small and extremely hydrophobic cytoplasmic membrane protein. These properties are shared with PspB and PspD, although there is no obvious sequence homology between them. A cytoplasmic membrane location makes it possible that PspG may interact with other Psp proteins and play a role in signal transduction. However, the precise role of the other cytoplasmic membrane Psp proteins (PspBCD) is not yet known, making it difficult to speculate about the role of PspG.

The first evidence that pspG might exist was that a pspA null mutation partially suppressed the severe phenotypes of a pspC null mutant (8). It was hypothesized that the overexpression of another gene, due to the absence of PspA, was responsible for this suppression. A Δ(pspA-ycjF) mutant also had less severe phenotypes than a strain with only a pspC mutation. This suggested that the absence of PspA caused overexpression of a gene that was not in the pspA operon. This overexpression was apparently PspF dependent, because a pspF null mutation exacerbated the phenotypes of a Δ(pspA-ycjF) mutant. The data presented here suggest that it is increased pspG expression that moderates the phenotype of a Δ(pspA-ycjF) mutant. However, a strain with a pspG null mutation alone does not have a phenotype under the conditions we tested (Fig. 5 and 6). Of course, it is possible that a pspG null mutant will have a phenotype under conditions not tested in this study, such as more extreme Psp-inducing conditions than YscC overproduction. However, the advantage of the YscC-induced growth defect assay is that it correlates well with the physiologically relevant virulence defects. Another possibility for the absence of a pspG null mutant phenotype is functional redundancy between PspG and one or more proteins encoded in the pspA operon. PspG is most similar in size and properties to PspD. A pspD null mutant does not have a YscC-induced growth defect (Maxson and Darwin, unpublished). Therefore, we wondered if there might be redundancy between PspD and PspG. However, a ΔpspD ΔpspG mutant also has no YscC-induced growth defect (data not shown). The possibility of functional redundancy between PspG and one or more of the other Psp proteins will be investigated in future genetic and biochemical studies.

The predicted RpoN-binding sites upstream of many pspA and pspG homologs diverge from the highly conserved consensus sequence at position −12 (Fig. 1). In most cases the C at position −12 has been replaced with an A or T. A systematic mutational analysis of the region surrounding position −12 within an E. coli glnH promoter derivative indicated that it plays multiple roles in transcription (31). These roles include modulating both basal and induced RNA levels. When the C at position −12 was mutated, a change to A had no effect on transcription, whereas replacing it with a T significantly reduced promoter strength (31). For the E. coli pspA promoter, changing the T at position −12 to the consensus C did not affect transcription in vitro (13). However, the in vivo effect was not studied. The C at position −12 is extremely well conserved in other RpoN-dependent promoters. The fact that such a large group of pspA and pspG promoters have an A or T at this position is likely to have a functional significance, possibly related to activation by PspF or modulation of uninduced activity. The lack of a consensus C at position −12 certainly appears to be a signature of most PspF-dependent promoters. However, this rule is apparently not always true, because at least one pspA operon promoter is predicted to have a C at position −12 (Fig. 1). Similarly, we are not suggesting that all bacterial RpoN-dependent promoters without a C at position −12 will be PspF dependent. Investigating the relationship between the RpoN-binding site sequence and the regulation of PspF-dependent promoters should be an interesting area for future research.

The pspA operon and pspG are not linked in any bacterial genome. It is not uncommon for coregulated genes to be physically separated. It may be that the pspA and pspG promoters have different strengths or induction levels by the Psp response. Our analysis indicates that the fold induction of Φ(pspA-lacZ) by a pspA null mutation, or secretin overexpression, is higher than that of Φ(pspG-lacZ) (Fig. 2 and 3). It is also possible that, in addition to control by the Psp proteins, one or both of these promoters is also regulated by an additional mechanism.

Are pspABCDycjXF and pspG the only genes directly activated by PspF?

We cannot yet answer this question, although we speculate that this may be the case. Apparently, there are only a small number of RpoN-dependent promoters in Y. enterocolitica and related bacteria. We have already eliminated several as members of the Psp regulon, and many of the others are predicted to be controlled by EBPs other than PspF. We also note that the pspF and pspG null mutations had indistinguishable effects on the secretin-induced growth defect of a Δ(pspA-ycjF) mutant (Fig. 5). Microarray technology is just becoming available for Y. enterocolitica, which may allow this question to be answered in the future.