Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Mol Microbiol. Author manuscript; available in PMC 2010 November 1.
Published in final edited form as:
PMCID: PMC2782453

Analysis of the Yersinia enterocolitica PspBC proteins defines functional domains, essential amino acids and new roles within the phage-shock-protein response


The Yersinia enterocolitica phage-shock-protein (Psp) stress response system is activated by mislocalized outer membrane secretin components of protein export systems and is essential for virulence. The cytoplasmic membrane proteins PspB and PspC were proposed to be dual function components of the system, acting both as positive regulators of psp gene expression and to support survival during secretin-induced stress. In this study we have uncoupled the regulatory and physiological functions of PspBC and discovered unexpected new roles, functional domains and essential amino acids. First, we showed that PspB controls PspC concentration by both pre- and post-transcriptional mechanisms. We then screened for PspBC mutants with altered transcriptional regulatory function. Unexpectedly, we identified PspB and PspC mutants that activated psp gene expression in the absence of secretin-induced stress. Together with a subsequent truncation analysis, this revealed that the PspC cytoplasmic domain plays an unforeseen role in negatively regulating psp gene expression. Conversely, mutations within the PspC periplasmic domain abolished its ability to activate psp gene expression. Significantly, PspC mutants unable to activate psp gene expression retained their ability to support survival during secretin-induced stress. These data provide compelling support for the proposal that these two functions are independent.

Keywords: Yersinia, phage-shock-protein, regulation, secretin, stress response


The phage-shock-protein (Psp) system is an extracytoplasmic stress response thought to aid bacterial survival when the ion-permeability barrier of the cytoplasmic membrane is compromised (reviewed by Darwin, 2005). It was originally discovered in Escherichia coli (Brissette et al., 1990), but all or at least some of its components are widely conserved amongst both bacterial and non-bacterial species (e.g. Darwin, 2005, Bidle et al., 2008, Vrancken et al., 2008). The expression and/or function of the Psp system has been linked to important bacterial phenotypes including virulence in animals (Darwin and Miller, 2001), survival within macrophages (Eriksson et al., 2003) and biofilm formation (Beloin et al., 2004).

The Psp system has been studied most in E. coli and Yersinia enterocolitica, where it is encoded by two unlinked loci (pspF-pspABCDE and pspG in E. coli and pspF-pspABCDycjXF and pspG in Y. enterocolitica). pspA promoter activity is induced by heat shock, osmotic shock, ethanol and the overproduction of specific cell envelope proteins (reviewed by Model et al., 1997, Darwin, 2005). These events probably induce the Psp system because of a common ability to compromise the cytoplasmic membrane ion barrier, which has consequences that include decreased proton motive force (PMF). In vivo and in vitro work has led to the proposal that the Psp system helps maintain PMF during stress conditions (e.g. Kleerebezem et al., 1996, Kobayashi et al., 2007). However, it is not known if PMF decline is the precise inducing signal for the Psp system, or if its primary function is to help maintain PMF. The possibility remains that in some circumstances differences in PMF simply coincide with the Psp-inducing event(s) and the physiological function of the Psp system itself.

psp gene expression is strongly induced by a filamentous phage secretin protein, which is how the phage-shock-protein system got its name (Brissette et al., 1990). Secretins are outer membrane proteins involved in type 2 and 3 secretion systems and type IV pili, as well as filamentous phage extrusion (Genin and Boucher, 1994). Whereas most Psp-inducing conditions affect the expression of many other genes, secretin production is a highly specific inducer of only psp gene expression (Lloyd et al., 2004, Seo et al., 2007). Evidence suggests that secretins induce the Psp system by mislocalizing into the cytoplasmic membrane and forming a pore (Guilvout et al., 1993). Some Y. enterocolitica psp deletion mutants cannot survive when secretins are overproduced. Secretin-sensitivity probably explains why these psp null mutants are also sensitive to native Ysc type 3 secretion system synthesis (which includes the YscC secretin) and avirulent in mice (Darwin and Miller, 2001, Green and Darwin, 2004).

Regulation of psp gene expression is mediated by PspF, A, B and C. PspF is a DNA-binding protein that activates the σ54-dependent promoters of pspA and pspG (Jovanovic et al., 1996, Green and Darwin, 2004, Lloyd et al., 2004). PspA is a peripheral cytoplasmic membrane protein that binds to PspF and inhibits its activity (Dworkin et al., 2000, Elderkin et al., 2002). PspB and PspC are integral cytoplasmic membrane proteins required for induction of psp gene expression in response to most triggers (Weiner et al., 1991, Kleerebezem et al., 1996, Maxson and Darwin, 2006). The current model for activation of the system is that an extracytoplasmic inducing signal is detected by PspB and/or PspC that, via their own interaction with PspA (Adams et al., 2003), prevent PspA from inhibiting PspF. Therefore, these four proteins appear to form a signal transduction system with PspA acting as a link between its membrane (PspBC) and cytoplasmic (PspF) components.

In E. coli it was reported that the ArcB redox sensor is also an essential signal transduction component for secretin-dependent activation of the pspA promoter (Jovanovic et al., 2006). However, more recent work indicates that ArcB is not essential and any involvement it has is growth condition dependent (Jovanovic et al., 2009). Under the conditions we use to study the Y. enterocolitica Psp system an arcB null mutation has no effect on secretin-dependent activation of the pspA promoter and does not cause any secretin-sensitivity, whereas pspB and pspC null mutations do (Maxson and Darwin, 2006, Seo et al., 2007). Importantly, the magnitude of secretin-sensitivity caused by various psp null mutations correlates with the reduction in virulence (e.g. Darwin and Miller, 2001, Green and Darwin, 2004). Therefore, our growth conditions and secretin-sensitivity phenotype provide a good model to study the Psp response without any involvement of ArcB, but an absolute requirement for PspB and PspC.

In addition to its role as a negative regulator of psp gene expression, evidence from E. coli suggests that PspA acts as a physiological effector of the Psp system (e.g. Kleerebezem et al., 1996, Kobayashi et al., 2007). However, a Y. enterocolitica pspA in frame deletion mutant is not secretin-sensitive or severely attenuated for virulence, whereas pspB and pspC null mutants are (Darwin and Miller, 2001, Maxson and Darwin, 2006). We showed that, like PspA, Y. enterocolitica PspB and PspC are also dual function proteins, acting both as positive regulators of psp gene expression and separately to allow survival during secretin-induced stress. For example, in a strain lacking the pspF-pspABCDycjXF and pspG genes, production of PspB or PspC alone is sufficient for survival during secretin-induced stress (Maxson and Darwin, 2006). Remarkably, even production of Y. enterocolitica PspBC in P. aeruginosa can alleviate toxicity caused by a mislocalized secretin (Seo et al., 2007).

In this study we have followed up on this previous work by showing that a single amino acid substitution is sufficient to uncouple the regulatory and secretin-stress tolerance functions of PspC. Our experiments also define the domains within PspC that facilitate its regulatory functions and uncover previously unforeseen roles for both PspB and PspC within the Y. enterocolitica Psp response.


Analysis of lacZp-pspB/C complementation plasmids reveals that PspB controls PspC concentration by both pre- and post-transcriptional mechanisms

Our original goal was to isolate mutant PspB/C proteins defective for only one of their proposed functions. However, studying PspBC is complicated by their positive regulatory role; both are required for induction of psp gene expression. Therefore, destroying their positive regulatory function would reduce their own production. To bypass this complication we constructed low copy number plasmids in which pspB/C were expressed from the E. coli lacZ promoter (Fig. 1). Therefore, any pspB/C mutants defective for activating the pspA promoter will still be expressed from lacZp at the same level as the wild type pspB/C genes. These lacZp-pspB/C plasmids restored wild type Φ(pspA-lacZ) regulation and secretin-stress tolerance to ΔpspB/C mutants and did not cause significant over or under-expression of the Psp regulon (data not shown).

Fig. 1
Analysis of lacZp-pspB/C+ complementation plasmids.

When pspBC were both expressed from the plasmid, they restored wild type regulatory and growth phenotypes to a ΔpspBC mutant, but if either pspB or pspC was removed these functions were lost (Fig. 1 and later figures). This manifested as obvious MacConkey lactose agar phenotypes (Fig. 1). Therefore, the lacZp-pspBC+ plasmid can be used to simultaneously screen for loss of function mutations in either one of the genes. Furthermore, the tacp-yscC secretin expression plasmid induces Φ(pspA-lacZ) expression at both 26°C and 37°C, but only inhibits growth of a ΔpspBC null strain at 37°C (Darwin and Miller, 2001 and Fig. 1). Therefore, regulatory phenotypes can be examined at 26°C and stress tolerance phenotypes at 37°C.

To further characterize the lacZp-pspB/C plasmids we monitored PspBC protein production by immunoblotting. In a ΔpspBC strain, similar PspB protein levels were produced from the lacZp-pspB+ and lacZp-pspBC+ plasmids (Fig. 1). However, PspC was barely detectable from the lacZp-pspC+ plasmid, whereas it was abundantly produced from the lacZp-pspBC+ plasmid. This suggested that PspC protein concentration is PspB-dependent even when PspB no longer regulates pspC transcription. To test this conclusion we investigated whether the amount of PspC produced from the lacZp-pspC+ plasmid could be elevated if PspB was produced from a separate plasmid (Fig. 1). The results showed that in a ΔpspBC strain the PspC produced from the lacZp-pspC+ plasmid was greatly elevated by the introduction of pspB+ on a separate plasmid. Therefore, these data reveal a new function for PspB in controlling PspC concentration post-transcriptionally.

Alanine substitution of conserved PspBC amino acids produces an unexpected phenotype

As a first attempt to isolate altered function mutants, predicted PspB and PspC protein sequences from diverse species were aligned (Fig. S1) and invariant amino acids were identified as mutagenesis targets. lacZp-pspB+ plasmid derivatives were made encoding PspB with W21A, H25A, Y26A, L47A, R57A, E62A, or P70A mutations. Derivatives of the lacZp-pspC+ plasmid were made encoding PspC with R32A, G37A, G41A, V42A, C43A, G45A, R58A, Y77A, E122A, S127A or L139A mutations.

In a ΔpspB Φ(pspA-lacZ) strain most of the PspB mutants had phenotypes indistinguishable from wild type (data not shown). Exceptions were R57A and E62A mutants, which had slightly reduced functions. However, immunoblot analysis revealed that these proteins were less abundant than the wild type (data not shown).

In a ΔpspC Φ(pspA-lacZ) strain the PspC E122A and L139A mutants had slightly diminished functions that were probably due to reduced protein abundance (data not shown). However, the protein level of the PspC S127A mutant was similar to the wild type, but it reduced Φ(pspA-lacZ) expression in the presence of YscC (Fig. 2; 120 and 170 Miller units for S127A and wild type, respectively). The much stronger effect of a S127P mutation described later makes it likely that this subtle phenotype is significant.

Fig. 2
Characterization of selected PspC alanine substitution mutants.

Most of the remaining PspC mutants had phenotypes indistinguishable from wild type. However, on MacConkey lactose agar the G41A, G45A, R58A and Y77A mutants had a Lac+ (pink or red) phenotype in the absence of YscC, whereas the wild type appeared Lac (white). Enzyme assays confirmed elevated β-galactosidase activity ranging from 66 (G45A) to 960 (Y77A) Miller units, whereas the wild type had 55 Miller units (Fig. 2). The stronger effect of a G45W mutation described later makes it likely that the subtle G45A phenotype is significant.

The three mutant PspC proteins that most increased Φ(pspA-lacZ) expression were more abundant than wild type, even though all were expressed from lacZp (Fig. 2). We do not think they are inherently more stable, which would be a trivial explanation for the phenotype. Rather, because the ΔpspC host strain still has the native pspB gene on the chromosome, the constitutive PspC mutants will elevate chromosomal pspB expression. More PspB increases PspC concentration post-transcriptionally (Fig. 1). Immunoblotting confirmed elevated PspB protein in these three mutants (Fig. 2). Consistent with all of this, in work described below with pspBC both expressed from lacZp, PspC mutants with constitutive activity were not more abundant than wild type, presumably because they cannot elevate pspB expression from lacZp.

Random identification of PspBC mutants that increase Φ(pspA-lacZ) expression

To probe the constitutive phenotype further, and to determine if such mutations could also be isolated in PspB, the pspBC+ insert of the lacZp-pspBC+ plasmid was randomly mutagenized by error prone PCR. A ΔpspBC Φ(pspA-lacZ) strain, containing the empty tacp expression plasmid pVLT35 (i.e. −YscC), was transformed with five independent mutant plasmid libraries. Transformants were recovered on MacConkey lactose agar incubated at 37°C. Under these conditions the wild type lacZp-pspBC+ plasmid causes white colony formation, whereas mutants with potentially increased Φ(pspA-lacZ) expression formed red colonies.

Approximately 500 of 10,000 colonies screened were red. We selected 10 red colonies from each of the five mutant libraries, isolated the lacZp-pspBC+ plasmid DNA, checked it’s integrity by restriction digest, and reintroduced it into the Y. enterocolitica strain to confirm that it caused the mutant phenotype. DNA sequencing revealed that most encoded multiple amino acid substitutions (ranging from 2-5). These were not characterized further and indicate that the PCR reactions were highly mutagenic. The remaining plasmids had inserts encoding single amino acid substitutions in PspB or PspC. Several of the same substitutions were identified more than once, either alone or as one change in the multiple substitution mutants. Furthermore, of the 4 alanine substitutions in PspC that elevated Φ(pspA-lacZ) expression (Fig. 2), substitutions at 3 of these positions were identified in the random screen. Quantitative β-galactosidase assays revealed that some mutants could still respond to YscC production, but most had similar Φ(pspA-lacZ) expression +/− YscC (Fig. 3). Immunoblotting showed that none of the mutant proteins was more abundant than wild type (Fig. 3). For the PspB-A7D mutant two distinct protein species were detected, possibly indicative of proteolysis.

Fig. 3
Characterization of constitutive PspB and PspC mutants.

This screen revealed that substitutions at multiple positions within both PspB and PspC increased Φ(pspA-lacZ) expression. The mutations were in the predicted transmembrane and cytoplasmic domains of both proteins, which suggested that these domains are important for a negative regulatory function. Notably, none of the mutations was in the periplasmic domain of PspC. Due to their obvious structural importance we did not further investigate the role of the transmembrane domains here, but in the next set of experiments we decided to probe the role of the cytoplasmic domains by deletion analysis.

The cytoplasmic domain of PspC is required for negative regulation

To begin to investigate the importance of the PspB/C cytoplasmic domains we constructed strains with chromosomal in frame deletion mutations that removed the region encoding most of the PspB cytoplasmic domain (amino acids 32-67; strain AJD2446) or most of the PspC cytoplasmic domain (amino acids 7-51; strain AJD2444). Deletion of the PspB cytoplasmic domain resulted in a pspB null phenotype, whereas deletion of the PspC cytoplasmic domain caused constitutive high-level Φ(pspA-lacZ) expression (data not shown). This implicated the PspC cytoplasmic domain in negative regulation. To investigate this further we constructed derivatives of the lacZ-pspBC+ plasmid encoding PspCs with progressive N-terminal (cytoplasmic domain) deletions and tested the effect on Φ(pspA-lacZ) expression (Fig. 3).

The Y. enterocolitica PspC protein is predicted to have an N-terminal extension that is absent from PspC proteins of other species (Fig. S1). Deletion of this region (amino acids 2-21) had no effect on regulation of Φ(pspA-lacZ) expression (Fig. 3). However, deletion of amino acids 2-32 resulted in constitutive Φ(pspA-lacZ) expression. A similar phenotype was obtained with larger deletions (Fig. 3). Immunoblot analysis indicated that PspCs with the Δ2-32 and larger deletions were much less abundant than the wild type, which is not surprising for a truncation. However, it is also possible that the deletions remove a major epitope recognized by our PspC polyclonal antiserum. Nevertheless, these data demonstrate that an intact region 2-32 of the PspC cytoplasmic domain is essential for negatively regulating Φ(pspA-lacZ) expression. This would be equivalent to a very small deletion (amino acids 2-11) in E. coli PspC (Fig. S1).

Mutant PspB/C proteins with constitutive regulatory activity do not act independently

Secretin-dependent induction of Φ(pspA-lacZ) expression requires that PspB and PspC are both present (Fig. 1 and Maxson and Darwin, 2006). However, we considered the possibility that constitutive PspB/C mutants might now be able to activate the pspA promoter independently. The randomly identified constitutive PspB/C mutants were isolated as lacZp-pspBC+ derivatives. Therefore, to determine if they could work independently we had to separate the pspB and pspC genes. To do this we constructed lacZp-pspB+ plasmids encoding PspB with L6H, L6R, A7D, W23R or E68D mutations. In a Φ(pspA-lacZ) strain with a ΔpspB pspC+ genotype they formed red colonies on MacConkey lactose agar, whereas in a ΔpspBC strain the colonies were white (data not shown), indicating that PspC was required. Similar analysis of lacZp-pspC+ plasmid derivatives encoding PspC with V33E, G45W, R58G, or L69P mutations in pspB+ ΔpspC and ΔpspBC strains indicated that they still required PspB to activate Φ(pspA-lacZ) expression (data not shown). Protein stabilization may explain the letter, but not the former (Fig. 1). β-galactosidase assays of selected PspB and PspC mutants confirmed their failure to activate Φ(pspA-lacZ) expression in a ΔpspBC strain (Fig. S2).

Identification of PspBC mutants with defective positive functions

In addition to activating psp gene expression, PspBC are also required to support growth during YscC overproduction. Our previous work suggested that those two functions are independent (Maxson and Darwin, 2006). Next, we wanted to test that hypothesis by determining if it is possible to isolate mutations in pspB/C that reduce either positive regulatory or stress tolerance function alone.

The randomly mutagenized lacZp-pspBC+ plasmid libraries were used to transform a ΔpspBC Φ(pspA-lacZ) strain containing the tacp-yscC+ expression plasmid pAJD126. Transformants were patched onto two MacConkey lactose agar plates. One plate was incubated at 26°C and the other at 37°C. In contrast to the wild type phenotype (Fig. 1), mutants that retain the ability to activate Φ(pspA-lacZ) expression, but lose YscC-stress tolerance, should be red on the 26°C plate but fail to grow at 37°C. Conversely, mutants that lose the ability to activate Φ(pspA-lacZ) expression, but retain YscC-stress tolerance, will grow and be white (or pink) at both temperatures (e.g. Fig. 4).

Fig. 4
PspC mutants defective for positive regulation or secretin stress-tolerance.

These mutant phenotypes were rare but candidates were identified after screening several thousand colonies. Most had multiple amino acid substitutions (data not shown). Therefore, site-directed mutagenesis was used to construct plasmids encoding the various mutations in isolation. No PspB mutants were found as a result of this analysis. Three mutations in PspC reduced or abolished activation of Φ(pspA-lacZ) expression in response to YscC (V125D, S127P or F130S; Fig. 4). However, during YscC production these mutants still supported growth (Fig. 4), suggesting that they retain at least some secretin stress tolerance function. A more rigorous assessment of secretin stress-tolerance is presented later in Results.

The V125D, S127P and F130S mutations are all located in the periplasmic domain of PspC. This region is predicted to form a leucine zipper-like amphipathic helix characterized by a repeating pattern of leucines/valines each separated by six amino acids. To further test the importance of this region we introduced a potentially helix-breaking L118P mutation, which also abolished YscC-dependent induction of Φ(pspA-lacZ) expression (Fig. 4). PspC (L118P) was less abundant than the wild type, but the same was not true for the V125D, S127P or F130S mutants (Fig. 4).

One PspC mutation (M60K) caused a YscC stress-tolerance defect without affecting Φ(pspA-lacZ) expression (Fig. 4). However, the PspC (M60K) protein was much less abundant than wild type in strains grown with YscC (Fig. 4). We suspect that this is due to an experimental technicality. The PspC (M60K) strain cannot grow at all during YscC overproduction in liquid culture. Therefore, most of the cells analyzed arise from the culture inoculum in which the lacZ promoter producing PspC (M60K) was uninduced. For this reason, we did immunoblot analysis of strains grown without YscC production. Consistent with our suspicions, this revealed the PspC (M60K) level to be similar to the wild type (Fig. 4). Growth rate/yield assays revealed that the YscC-induced growth defect of the PspC (M60K) mutant was much worse than a pspC null mutant (data not shown). This suggests that the PspC (M60K) phenotype has a more complex explanation than simple loss of function, possibly involving a negative effect on other Psp components.

The periplasmic domain of PspC is required for positive regulation

The preceding data suggested that the periplasmic domain of PspC is essential for secretin-dependent induction of pspA operon expression. To further test this conclusion we constructed derivatives of the lacZp-pspBC+ plasmid encoding PspCs with progressive C-terminal (periplasmic domain) deletions and tested the effect on Φ(pspA-lacZ) expression. The results revealed that deletion of amino acids 123-139 was sufficient to abolish YscC-dependent induction of Φ(pspA-lacZ) expression (Fig. 5).

Fig. 5
PspC periplasmic domain deletion analysis.

A PspC mutant with increased Φ(pspA-lacZ) expression still requires the positive function of the periplasmic domain

It was speculated that the PspC periplasmic domain might be involved in signal detection (Adams et al., 2003). This is an attractive hypothesis in view of its location and is consistent with our discovery that periplasmic domain mutations abolish YscC-dependent induction of Φ(pspA-lacZ) expression (Fig. 4). Therefore, we hypothesize that periplasmic domain mutations prevent signal detection and/or signal transduction to the cytoplasmic domain. Conversely, PspC cytoplasmic domain (constitutive) mutants might increase Φ(pspA-lacZ) expression independent of the normal inducing signal. In that case, the introduction of a periplasmic domain mutation should not prevent activation of the pspA promoter by a cytoplasmic domain mutant.

To investigate this we constructed a derivative of the lacZp-pspBC+ plasmid encoding PspC with two mutations: the V33E substitution in the cytoplasmic domain, which alone increases Φ(pspA-lacZ) expression (Fig. 3), and the S127P mutation in the periplasmic domain, which alone abolishes YscC-dependent induction of Φ(pspA-lacZ) expression (Fig. 4). We then compared the ability of PspC wild type, V33E, S127P and V33E S127P proteins to activate Φ(pspA-lacZ) expression. As expected, the PspC V33E mutant showed a several fold increase in Φ(pspA-lacZ) expression +/− YscC compared to wild type (Fig. 6). However, this phenotype was completely reversed by the addition of the S127P mutation (Fig. 6). Therefore, the ability of this PspC cytoplasmic domain mutant to activate the pspA promoter still requires the positive function of the periplasmic domain. One possible explanation is that a constitutive PspC mutant still needs the input of an inducing signal via the periplasmic domain, but perhaps responds to a lower threshold of this signal (see Discussion).

Fig. 6
A PspC mutant with increased Φ(pspA-lacZ) expression still requires the positive function of the periplasmic domain.

Successful separation of PspC regulatory and secretin-stress tolerance functions

Previously we published data suggesting that PspC has two functions: regulation of psp gene expression and supporting cell survival during secretin-induced stress (Maxson and Darwin, 2006). If these two functions are independent then it might be possible to isolate mutants defective in one function but not the other. The periplasmic domain of PspC is essential for it’s positive regulatory function (Figs. (Figs.44 and and5),5), but these periplasmic domain mutants retain at least some stress-tolerance function (Fig. 4). However, in the ΔpspBC strain, aberrant regulation of remaining chromosomal psp genes can potentially complicate interpretation of stress-tolerance phenotypes. Therefore, to more rigorously test for survival during secretin production we used a Δ(pspF-ycjF) ΔpspG host strain, which lacks all chromosomal psp genes.

We used a plate dilution assay, which we have found to be a highly sensitive measure of survival/growth and is also more direct than monitoring optical densities of liquid cultures. The Δ(pspF-ycjF) ΔpspG strain contained a tacp-yscC+ plasmid, or the empty vector control, and the various lacZp-pspBC+ plasmid derivatives. Dilutions were spotted onto LB agar containing IPTG and incubated at 37°C. The results confirmed that in this experimental system PspC was critical for survival during YscC production (compare B+ and B+C+ strains in Fig. 7). Significantly, the three PspC single amino acid substitution mutants that have completely lost positive regulatory function still supported essentially wild type levels of survival during YscC production. Therefore, the periplasmic domain mutations have specifically abolished regulatory function whereas the secretin-stress tolerance function remains intact. We extended this analysis to some PspC constitutive mutants (Fig. S3), most of which grew normally in the ΔpspBC strain in which they were isolated. In the Δ(pspF-ycjF) ΔpspG host strain, most retained secretin-stress tolerance well above that of a pspC null mutant (V33E, C43Y, L69P, and T73N). Those that did have defects (L65R and L80P) were also less abundant than the wild type protein (Fig. 3), which probably impacts, and certainly complicates interpretation of, their negative growth phenotype.

Fig. 7
PspC mutants defective for positive regulation retain their secretin-stress tolerance function. Δ(pspB-ycjF) ΔpspG strains contained lacZp-pspB+ plasmid pAJD1134 (B+), lacZp-pspBC+ plasmid pAJD1136 (B+C+), or pAJD1136 derivatives encoding ...


We have found two new phenomena for PspB and PspC: post-transcriptional control of PspC by PspB and negative regulation of psp gene expression, which involves both PspB and PspC. Furthermore, mutational and truncation analysis has allowed us to define PspC domains required for negative (cytoplasmic domain) and positive (periplasmic domain) regulation of psp gene expression. We have also fulfilled a prediction about the independence of the regulatory and stress-tolerance functions of PspC by isolating mutations that destroy only one.

PspB and PspC are well known as positive regulators of pspA operon expression in E. coli and Y. enterocolitica (e.g. Weiner et al., 1991, Kleerebezem et al., 1996, Maxson and Darwin, 2006). However, our analysis demonstrates that even when pspC is expressed from a promoter that PspB does not regulate, the PspC protein level remains PspB-dependent (Fig.1). We did not investigate the mechanism here, but speculate that the previously described PspB-PspC interaction (Maxson and Darwin, 2006) might stabilize PspC. Our data also suggest that an equimolar PspB-PspC stoichiometry might be important for any stabilization function. The lacZp-pspC+ plasmid does not significantly increase the steady state level of PspC when PspB is only produced from the chromosomal pspB gene, but it does when PspB is also produced from the lacZp plasmid (Fig. 1). All of this raises the question of whether PspC stability plays a role in the Psp response. We note that the difference in endogenous PspC level +/− YscC appears much larger than the difference in PspB level (Fig. 1 panel D(i), the first two lanes on the left hand side). When the Psp response is uninduced, the PspC level might decline more due to both reduced pspC gene expression and protein destabilization caused by lower PspB concentration. Tight control of PspC concentration might be particularly important.

Our unbiased molecular genetic approach to study PspBC function yielded a surprise with the isolation of mutants that increased pspA promoter expression +/− YscC (Figs. (Figs.22 and and3).3). To speculate a mechanistic explanation we must accommodate the following: (1) Constitutive mutations occur in PspB and PspC at various locations with high frequency, suggesting disruption of a function; (2) pspB/C null mutations do not cause a constitutive phenotype; (3) Deletion of the PspC cytoplasmic domain causes a constitutive phenotype, but not deletion of the PspB cytoplasmic domain; (4) Constitutive PspB mutants require PspC for activity and constitutive PspC mutants require PspB (protein stabilization may explain the latter, but not the former). Add to this the current model that PspBC control pspA promoter activity by modulating the inhibitory PspA-PspF interaction and that E. coli PspA interacts with PspBC (Adams et al., 2003). One model that accommodates all of this information is that the PspC cytoplasmic domain prevents a PspA-PspBC interaction in non-inducing conditions, leaving PspA free to inhibit PspF (Fig. 8). Deletion of the cytoplasmic domain would remove this inhibition, allowing the PspA-PspBC complex to form. Transmembrane/cytoplasmic domain mutations in PspBC might cause a conformational change in the PspC cytoplasmic region that has a similar effect.

Fig. 8
Basic working model for regulation of the Psp system by PspBC. In non-inducing conditions (OFF) the level of the Psp-inducing signal(s), represented by the stars, is below the inducing threshold. PspC remains in the “off” state such that ...

It is possible that the constitutive mutant PspBC proteins might interact differently with the membrane to cause a Psp-inducing stress. This is unlikely because some of the mutations are in cytoplasmic domain regions far from the transmembrane domains (e.g. PspB-E68D/G, PspC-V33E). Furthermore, only a very few cytoplasmic membrane proteins induce the Psp response when overproduced (Maxson and Darwin, 2004). Therefore, the nature of any inducing membrane stress is highly specific and unlikely to be caused by a variety of PspB/C mutants. In addition, if the mutant proteins were causing stress, they might affect growth of the Δ(pspF-ycjF) ΔpspG strain even without YscC, but they did not (Fig. S3). Regardless, and even if increased stress did occur, our data still show that normal PspBC sequence (probably conformation) plays a role in negatively controlling pspA operon expression. Specifically, the PspC cytoplasmic domain is required for reduced Psp system production. Its removal, even by a deletion within the native chromosomal pspC gene, activates the pspA promoter.

The major goal of our work was to isolate PspB/C mutants with decreased pspA promoter activating or stress-tolerance ability. This led to us finding that the PspC periplasmic domain is involved in secretin-dependent activation of the pspA promoter, with V125D and S127P substitutions completely abrogating this function (Fig. 4). This domain is predicted to form a leucine zipper-like amphipathic helix made up of six leucines/valines, each separated by six amino acids. V125D affects the fourth of the repeating leucines/valines on the predicted hydrophobic face of the helix. Replacement of the third leucine with proline (L118P), which like S127P might also disrupt the helix, also abolished positive regulatory function (although, unlike the V125D and S127P mutations, this change might also destabilize the protein; Fig. 4). The periplasmic location of this domain makes it a good candidate for sensing extracytoplasmic stress. One possibility, previously proposed by others (Adams et al., 2003), is that dimerization of the PspC periplasmic domain is modulated by the Psp-inducing signal. Another possibility is that the hydrophobic face of the amphipathic helix interacts with the membrane to detect stress. L118P, V125D and S127P mutations might blind PspC to the inducing signal or lock it into an “off” state for activation even if a signal is detected.

A bacterial two hybrid (BACTH) system based on bringing together T18 and T25 fragments of adenylate cyclase (Karimova et al., 1998) has been used to study Psp protein-protein interactions (Maxson and Darwin, 2006, Engl et al., 2009, Jovanovic et al., 2009). Encouragingly, we have found that the PspC V125D and S127P mutations significantly reduce the apparent interaction between PspA-T18 and T25-PspC proteins, which is consistent with our model (Fig. 8; E. Gueguen and A. J. Darwin, unpubl. data). However, the artificial expression of these fusion proteins is unsuitable to monitor the interaction dynamics, which our model demands. Therefore, we are trying to develop assays to monitor dynamic interactions between endogenous Psp proteins to permit a future rigorous test of our model.

The S127P PspC periplasmic domain mutation was dominant over the V33E cytoplasmic domain mutation (Fig. 6). Therefore, the enhanced pspA promoter activating capacity of PspC (V33E) requires the function of the periplasmic domain. This argues that the constitutive mutants still require a signal input for activity rather than becoming signal independent. We speculate that the inducing signal is always present but there is a threshold at which PspC becomes active. Mutations such as V33E might make PspC respond at a lower threshold, but the signal must still be detected/transduced by the periplasmic domain. Such a permanently present but variable signal could be the PMF, some other transmembrane ion gradient such as potassium or a particular physical property of the cytoplasmic membrane.

We did not isolate PspB mutants unable to activate the pspA promoter. However, our screen demanded that any such mutants retained secretin-stress tolerance. Separating those functions may be difficult for the 75-amino acid PspB. Nevertheless, it is possible that PspC is the direct stress-responsive regulator of psp gene expression and PspB exerts its regulatory role entirely through PspC. One way it might do this is via PspC stabilization. However, stabilization of PspC is unlikely to be the only role for PspB.

An especially significant finding was the isolation of PspC mutants that lost the ability to activate Φ(pspA-lacZ) expression but clearly retained the ability to support growth during YscC production (Figs. (Figs.44 and and7).7). This provides strong support to our contention that the ability of PspC to mitigate secretin-induced stress does not require, and is not a consequence of, its ability to function as an activator of psp gene expression (Maxson and Darwin, 2006).

The issue of Psp effector proteins is complex. A wealth of evidence supports the role of PspA as a physiological effector in E. coli (e.g. Kleerebezem et al., 1996, Kobayashi et al., 2007). Together with PspA’s abundance, there is every reason to believe that it plays an important physiological role. However, whereas Y. enterocolitica pspB and pspC null mutants are exquisitely sensitive to secretin production, and highly attenuated for virulence, a pspA in frame deletion mutant is not (Darwin and Miller, 2001, Maxson and Darwin, 2006). Overproduction of PspB or PspC alone can almost completely reverse the secretin-sensitive growth phenotype of a complete psp null strain whereas PspA overproduction cannot (Maxson and Darwin, 2006). We hypothesize that PspABC (and PspG; Green and Darwin, 2004) are all physiological effectors that have both common and distinct roles, none of which is currently understood. A pspA in frame deletion mutant overexpresses all remaining psp genes and the physiological function of overproduced PspBC (and PspG) might mask the effect of losing PspA. With the low copy number lacZp-pspB/C expression system used here both PspB and PspC were required for robust growth in a complete psp null strain (Fig. 7 and data not shown). However, the lacZp-pspBC+ plasmid does not restore full wild type growth rate/yield in liquid culture (D. C. Savitzky and A. J. Darwin, unpubl. data). A complete response to Psp-inducing stress might involve interplay between the physiological functions of at least PspABC.

This work provides a great deal of new mechanistic information about PspBC function. PspB controls PspC levels post transcriptionally, the cytoplasmic domain of PspC is required for reduced psp gene expression, and the PspC periplasmic domain is involved in pspA promoter activation. A single amino acid substitution can uncouple the regulatory and secretin-stress tolerance functions of PspC, providing compelling evidence of independent functions. More challenges lay ahead before we can fully understand the critical roles played by these two multi-functional proteins, but the new information here provides exciting and intriguing models to test.

Experimental procedures

Bacterial strains, plasmids and routine growth conditions

Bacterial strains and plasmids are shown in Table 1. Primer sequences used in this study will be supplied upon request (please contact the corresponding author). All PCR-generated fragments were verified by DNA sequencing. Y. enterocolitica strains were grown at 26°C or 37°C as noted. Strains were routinely grown in Luria-Bertani (LB) broth, or on LB agar plates (Miller, 1972). Antibiotics were used as previously (Maxson and Darwin, 2004).

Table 1
Strains and plasmids

lacZp-pspB+/C+ plasmid constructions

The lacZp-pspB/C plasmids pAJD1134-1136 were constructed by amplifying fragments from Y. enterocolitica genomic DNA using primers that incorporated SacI (upstream) and XbaI (downstream) sites. The SacI-XbaI fragments were ligated into pWSK129. To prevent potential translational polarity effects on pspC expression, the same primer pairs were used for the pspBC+ and pspC+ inserts, but the genomic DNA templates were from wild type and a ΔpspB in frame deletion mutant, respectively. Therefore, both plasmids maintain the overlapping stop and start codons between the pspB and pspC open reading frames and the putative ribosome binding sites upstream of each gene.

Identification of conserved amino acids in PspB and PspC

Clustal W ( was used to align predicted PspB or PspC protein sequences (Fig. S1). Positions with invariant amino acids were selected for alanine substitution as listed in Results.

Site directed mutagenesis

For alanine substitution of conserved PspB and PspC amino acids, plasmids containing the insert fragments from pAJD1134 and pAJD1135, respectively, were amplified by PCR with Pfu Ultra polymerase (Stratagene) and two perfectly complimentary primers with mismatches to introduce the desired mutation. The products were digested with DpnI to remove template DNA and then used to transform E. coli DH5α. The DNA sequence of the entire insert fragment was determined, which was then transferred into pWSK129 as a SacI-XbaI fragment to make pAJD1134 or pAJD1135 derivatives with a single codon mutation.

The pAJD1136 derivative encoding PspB+ PspC(L118P) was made by a PCR SOEing strategy (Heckman and Pease, 2007). Two fragments encompassing the pAJD1136 insert were amplified by PCR with primers that generated a 20 nucleotide overlapping end between the two fragments, which encoded the L118P mutation. The two fragments were joined in a PCR SOEing reaction. The final product was cloned into pWSK129 as a SacI-XbaI fragment. A similar strategy was used to make other mutant derivatives of pAJD1136 described in Results.

Random mutagenesis

The pspBC+ insert of pAJD1136 was amplified by PCR using the GeneMorph® II Random Mutagenesis kit (Stratagene). PCR reactions contained 200 μM of each dNTP, ~200 ng of plasmid template, 300 nM of each primer and 2.5 units of Mutazyme II DNA polymerase. The cycling program was 30 × [95°C 30 sec, 54°C 30 sec, 72°C 1 min]. Five independent PCR reactions were done. The products were digested with SacI and XbaI, ligated into pWSK129 and used to transform E. coli DH5α. Each transformation gave ~1500-2000 colonies. The colonies from each individual transformation were combined and plasmid DNA was isolated, resulting in five independent mutant libraries. Aliquots of each library were used to transform a pYV derivative of Y. enterocolitica strain AJD1198 containing either pVLT35 or pAJD126. Transformants were screened for mutant phenotypes as described in Results.

Construction of pspB and pspC cytoplasmic domain in frame deletion mutants

Two ~650 bp fragments corresponding to the regions immediately upstream and downstream of the deletion site were amplified by PCR with primers that incorporated a BglII restriction site at one end of each fragment. The fragments were ligated at this BglII site to form an in frame deletion within pspB or pspC, cloned into plasmid pEP185.2 (Miller, 1972) and the DNA sequence was confirmed. The plasmids were integrated into the Y. enterocolitica chromosome following conjugation from E. coli and then chloramphenicol-sensitive segregants were isolated by cycloserine enrichment as described previously (Darwin and Miller, 2001). The deletion mutations were confirmed by colony PCR and Southern hybridization analysis.

PspC truncations

Derivatives of pAJD1136 encoding PspC with cytoplasmic domain deletions (N-terminus) were generated by a PCR SOEing strategy as described above. The constructs had internal deletions within pspC so that the native ATG was retained as the initiation codon. This also maintained the overlapping stop and start codons of pspB and pspC, respectively, and the correct positioning of a putative ribosome binding upstream of pspC. Derivatives of pAJD1136 encoding PspC with periplasmic domain deletions (C-terminus) were generated by PCR amplification of the pAJD1136 insert using a common upstream primer and various downstream primers that annealed at different distances from the end of pspC and incorporated a stop codon. Products were cloned into pWSK129 as SacI-XbaI fragments.

β-galactosidase assays

Saturated cultures were diluted into 5 ml of LB broth in 18 mm diameter test tubes. The initial optical density (600 nm) was approximately 0.04. The cultures were grown on a roller drum at 37°C for two hours. Then 0.2 mM IPTG (final concentration) was added and growth continued at 37°C for two more hours prior to harvest. β-galactosidase enzyme activity was determined at room temperature (approximately 22°C) in permeabilized cells as described previously (Maloy et al., 1996). Activities are expressed in arbitrary Miller units (Miller, 1972). Individual cultures were assayed in duplicate, and values were averaged from at least three independent cultures.

YscC-induced stress tolerance assay

Strains were grown to saturation at 26°C in LB broth containing appropriate antibiotics. Optical densities (600 nm) were determined and then adjusted to be equivalent for all strains by bacterial cell concentration (centrifugation and resuspension in appropriate volumes of culture medium). Then 5 μl of undiluted and serial 10-fold dilutions (10−1 - 10−7) of each sample were spotted onto the surface of LB agar containing appropriate antibiotics and 0.2 mM IPTG followed by incubation at 37°C for 24 h.

Polyclonal antisera and immunoblotting

The anti-PspC polyclonal antiserum was described previously (Maxson and Darwin, 2006). An anti-PspB polyclonal antiserum which had higher avidity than that described previously (Maxson and Darwin, 2006) was generated for this study. The region of pspB encoding its predicted cytoplasmic domain was amplified by PCR and cloned into plasmid pMalc (New England Biolabs) so that it encoded a MBP-‘PspB fusion protein (MBP = maltose binding protein). This plasmid was transferred into E. coli strain M15 (Qiagen) and the resulting strain was grown at 37°C in LB broth containing 0.3 mM IPTG. The MBP-‘PspB protein was purified from the soluble fraction of a cell lysate by amylose resin affinity chromatography as described by the manufacturer (New England Biolabs). A polyclonal rabbit antiserum was raised against the intact MBP-‘PspB fusion protein at Covance Research Products Inc., Denver, PA.

For each immunoblot experiment equal amounts of bacterial cells (as assessed by optical density determination) were loaded into each lane of the gel. Proteins were separated by SDS-PAGE on gels containing 15% polyacrylamide and then transferred to nitrocellulose by electroblotting. Equal loading was confirmed by total protein staining with Ponceau S. Then, enhanced chemiluminescent detection followed sequential incubation with raw anti-PspB or anti-PspC antiserum used at 1 in 20,000 and 1 in 10,000 dilution, respectively, and then goat anti-rabbit IgG horseradish peroxidase conjugate (BioRad) used at 1 in 5,000 dilution.

Supplementary Material



We thank Heran Darwin for valuable discussions and for critically reviewing draft versions of the manuscript. This study was supported by Award Number R01AI052148 from the National Institute of Allergy and Infectious Diseases (NIAID). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIAID or the National Institutes of Health.


  • Adams H, Teertstra W, Demmers J, Boesten R, Tommassen J. Interactions between phage-shock proteins in Escherichia coli. J. Bacteriol. 2003;185:1174–1180. [PMC free article] [PubMed]
  • Beloin C, Valle J, Latour-Lambert P, Faure P, Kzreminski M, Balestrino D, Haagensen JA, Molin S, Prensier G, Arbeille B, Ghigo JM. Global impact of mature biofilm lifestyle on Escherichia coli K-12 gene expression. Mol. Microbiol. 2004;51:659–674. [PubMed]
  • Bidle KA, Kirkland PA, Nannen JL, Maupin-Furlow JA. Proteomic analysis of Haloferax volcanii reveals salinity-mediated regulation of the stress response protein PspA. Microbiology. 2008;154:1436–1443. [PMC free article] [PubMed]
  • Brissette JL, Russel M, Weiner L, Model P. Phage shock protein, a stress protein of Escherichia coli. Proc. Natl. Acad. Sci. USA. 1990;87:862–866. [PubMed]
  • Darwin AJ. The phage-shock-protein response. Mol. Microbiol. 2005;57:621–628. [PubMed]
  • Darwin AJ, Miller VL. The psp locus of Yersinia enterocolitica is required for virulence and for growth in vitro when the Ysc type III secretion system is produced. Mol. Microbiol. 2001;39:429–444. [PubMed]
  • de Lorenzo V, Eltis L, Kessler B, Timmis KN. Analysis of Pseudomonas gene products using lacIq/Ptrp-lac plasmids and transposons that confer conditional phenotypes. Gene. 1993;123:17–24. [PubMed]
  • Dworkin J, Jovanovic G, Model P. The PspA protein of Escherichia coli is a negative regulator of σ54-dependent transcription. J. Bacteriol. 2000;182:311–319. [PMC free article] [PubMed]
  • Elderkin S, Jones S, Schumacher J, Studholme D, Buck M. Mechanism of action of the Escherichia coli phage shock protein PspA in repression of the AAA family transcription factor PspF. J. Mol. Biol. 2002;320:23–37. [PubMed]
  • Engl C, Jovanovic G, Lloyd LJ, Murray H, Spitaler M, Ying L, Errington J, Buck M. In vivo localizations of membrane stress controllers PspA and PspG in Escherichia coli. Mol. Microbiol. 2009;73:382–396. [PMC free article] [PubMed]
  • Eriksson S, Lucchini S, Thompson A, Rhen M, Hinton JC. Unravelling the biology of macrophage infection by gene expression profiling of intracellular Salmonella enterica. Mol. Microbiol. 2003;47:103–118. [PubMed]
  • Genin S, Boucher CA. A superfamily of proteins involved in different secretion pathways in gram-negative bacteria: modular structure and specificity of the N-terminal domain. Mol. Gen. Genet. 1994;243:112–118. [PubMed]
  • Green RC, Darwin AJ. PspG, a new member of the Yersinia enterocolitica phage shock protein regulon. J. Bacteriol. 2004;186:4910–4920. [PMC free article] [PubMed]
  • Guilvout I, Mercereau-Puijalon O, Bonnefoy S, Pugsley AP, Carniel E. High-molecular-weight protein 2 of Yersinia enterocolitica is homologous to AngR of Vibrio aguillarium and belong to a family of proteins involved in nonribosomal peptide synthesis. J. Bacteriol. 1993;175:5488–5504. [PMC free article] [PubMed]
  • Guzman L, Belin D, Carson MJ, Beckwith J. Tight regulation, modulation, and high-level expression by vectors containing the arabinose PBAD promoter. J. Bacteriol. 1995;177:4121–4130. [PMC free article] [PubMed]
  • Heckman KL, Pease LR. Gene splicing and mutagenesis by PCR-driven overlap extension. Nat. Protoc. 2007;2:924–932. [PubMed]
  • Jovanovic G, Engl C, Buck M. Physical, functional and conditional interactions between ArcAB and phage shock proteins upon secretin-induced stress in Escherichia coli. Mol. Microbiol. 2009 in press: DOI: 10.1111/j.1365-2958.2009.06809.x. [PMC free article] [PubMed]
  • Jovanovic G, Lloyd LJ, Stumpf MP, Mayhew AJ, Buck M. Induction and function of the phage shock protein extracytoplasmic stress response in Escherichia coli. J. Biol. Chem. 2006;281:21147–21161. [PubMed]
  • Jovanovic G, Weiner L, Model P. Identification, nucleotide sequence, and characterization of PspF, the transcriptional activator of the Escherichia coli stress-induced psp operon. J. Bacteriol. 1996;178:1936–1945. [PMC free article] [PubMed]
  • Karimova G, Pidoux J, Ullmann A, Ladant D. A bacterial two-hybrid system based on a reconstituted signal transduction pathway. Proc. Natl. Acad. Sci. USA. 1998;95:5752–5756. [PubMed]
  • Kinder SA, Badger JL, Bryant GO, Pepe JC, Miller VL. Cloning of the YenI restriction endonuclease and methyltransferase from Yersinia enterocolitica serotype O:8 and construction of a transformable RM+ mutant. Gene. 1993;136:271–275. [PubMed]
  • Kleerebezem M, Crielaard W, Tommassen J. Involvement of stress protein PspA (phage shock protein A) of Escherichia coli in maintenance of the protonmotive force under stress conditions. EMBO J. 1996;15:162–171. [PubMed]
  • Kobayashi R, Suzuki T, Yoshida M. Escherichia coli phage-shock protein A (PspA) binds to membrane phospholipids and repairs proton leakage of the damaged membranes. Mol. Microbiol. 2007;66:100–109. [PubMed]
  • Lloyd LJ, Jones SE, Jovanovic G, Gyaneshwar P, Rolfe MD, Thompson A, Hinton JC, Buck M. Identification of a new member of the phage shock protein response in Escherichia coli, the phage shock protein G (PspG) J. Biol. Chem. 2004;279:55707–55714. [PubMed]
  • Maloy SR, Stewart VJ, Taylor RK. Genetic analysis of pathogenic bacteria. Cold Spring Harbor Laboratory Press; Plainview, NY: 1996.
  • Maxson ME, Darwin AJ. Identification of inducers of the Yersinia enterocolitica phage shock protein system and comparison to the regulation of the RpoE and Cpx extracytoplasmic stress responses. J. Bacteriol. 2004;186:4199–4208. [PMC free article] [PubMed]
  • Maxson ME, Darwin AJ. Improved system for construction and analysis of single-copy β-galactosidase operon fusions in Yersinia enterocolitica. Appl. Environ. Microbiol. 2005;71:5614–5618. [PMC free article] [PubMed]
  • Maxson ME, Darwin AJ. PspB and PspC of Yersinia enterocolitica are dual function proteins: regulators and effectors of the phage-shock-protein response. Mol. Microbiol. 2006;59:1610–1623. [PubMed]
  • Miller JH. Experiments in molecular genetics. Cold Spring Harbor Laboratory; Cold Spring Harbor, New York: 1972.
  • Model P, Jovanovic G, Dworkin J. The Escherichia coli phage-shock-protein (psp) operon. Mol. Microbiol. 1997;24:255–261. [PubMed]
  • Seo J, Savitzky DC, Ford E, Darwin AJ. Global analysis of tolerance to secretin-induced stress in Yersinia enterocolitica suggests that the phage-shock-protein system may be a remarkably self-contained stress response. Mol. Microbiol. 2007;65:714–727. [PubMed]
  • Vrancken K, Van Mellaert L, Anne J. Characterization of the Streptomyces lividans PspA response. J. Bacteriol. 2008;190:3475–3481. [PMC free article] [PubMed]
  • Wang RF, Kushner SR. Construction of versatile low-copy-number vectors for cloning, sequencing and gene expression in Escherichia coli. Gene. 1991;100:195–199. [PubMed]
  • Weiner L, Brissette JL, Model P. Stress-induced expression of the Escherichia coli phage shock protein operon is dependent on σ54 and modulated by positive and negative feedback mechanisms. Genes Dev. 1991;5:1912–1923. [PubMed]