The phage-shock-protein (Psp) system responds to extracytoplasmic stress that may reduce the energy status of the cell. It is conserved in many different bacteria and has been linked to several important phenotypes. Escherichia coli psp mutants have defects in maintenance of the proton-motive force, protein export by the sec and tat pathways, survival in stationary phase at alkaline pH, and biofilm formation. Yersinia enterocolitica psp mutants cannot grow when the secretin component of a type III secretion system is mislocalized, and have a severe virulence defect in animals. A Salmonella enterica psp mutation exacerbates some phenotypes of an rpoE null mutant and the psp genes of S. enterica and Shigella flexneri are highly induced during macrophage infection. PspA, the most abundant of the Psp proteins, is required for most of the phenotypes associated with the Psp system. Therefore, PspA is probably an effector that may play a role in maintaining cytoplasmic membrane integrity and/or the proton-motive force. However, PspA is not required for the ability to tolerate secretin mislocalization, which suggests an important physiological role for other Psp proteins. This article summarizes our current understanding of the Psp system: inducing signals, the underlying signal transduction mechanisms, the physiological roles it may play, and a genomic analysis of its conservation.
The Gram-negative bacterial cell envelope controls the passage of molecules into and out of the cell, includes structures like the cell wall, flagella and complex secretion systems, and provides an ion-permeability barrier for the establishment of the proton-motive force. Problems that might adversely affect the cell envelope are poorly tolerated, and extracytoplasmic stress response systems have evolved to deal with them. The best characterized examples are the RpoE and Cpx systems of Escherichia coli and its relatives, which respond to misfolded envelope proteins (Duguay and Silhavy, 2004). This review discusses another example of a specialized extracytoplasmic stress response: the enigmatic phage-shock-protein (Psp) system.
Fifteen years ago Peter Model's group reported that an E. coli protein was produced at high concentration during filamentous phage infection, and they named it phage-shock-protein A (PspA; Brissette et al., 1990). The Psp system has since been the subject of many studies. It is essential for the virulence of Yersinia enterocolitica and may also be important in other pathogens. This article summarizes our current understanding of the Psp system in both E. coli and Y. enterocolitica, the organisms in which it has been most studied.
Brief overview of the Psp system
Filamentous phage infection, mislocalization of some envelope proteins, extremes of temperature, osmolarity or ethanol concentration, and the presence of proton ionophores such as carbonylcyanide m-chlorophenylhydrazone (CCCP), all induce the expression of the psp genes, hereafter referred to as the Psp response (Model et al., 1997). A unifying consequence of these conditions may be dissipation of the proton-motive force (PMF), although this has not been tested in most cases. The mislocalization of secretin proteins is a particularly potent and specific inducing condition (Hardie et al., 1996; Model et al., 1997; Lloyd et al., 2004; Maxson and Darwin, 2004). Secretins are an important family of outer membrane pore-forming proteins involved in filamentous phage extrusion (explaining Psp induction by these phage), type II and type III secretion systems, and type IV pilus biogenesis (Genin and Boucher, 1994).
The Psp systems of E. coli and Y. enterocolitica have six proteins in common: PspA, B, C, D, F and G. PspF is a member of the enhancer-binding protein family of transcriptional regulators (Jovanovic et al., 1996). Primary sequences of the other Psp proteins tell us little about their functions, although genetic and biochemical analysis has begun to provide some insight.
A working model of the Psp response is presented in Fig. 1. It must be emphasized that some aspects of this model are speculative and await direct experimental tests. In the uninduced state, PspA binds to PspF and prevents it from activating transcription (Dworkin et al., 2000; Elderkin et al., 2002). One or both of the cytoplasmic membrane proteins, PspB and PspC, probably sense an inducing condition, such as secretin mislocalization and/or dissipation of the PMF. They may then interact with PspA (Adams et al., 2003), which frees PspF to activate transcription of the pspA operon and the recently identified pspG gene (Green and Darwin, 2004; Lloyd et al., 2004). As a result, the concentration of PspA increases dramatically in comparison to the other Psp proteins. Therefore, PspA probably plays an important physiological role when the Psp response is induced. Indeed, one study has revealed that PspA reduces dissipation of the PMF (Kleerebezem et al., 1996). The roles of PspD and PspG are unknown, although neither appear to be involved in regulation of the Psp response itself (Jones et al., 2003; Lloyd et al., 2004).
Induction of the Psp response
A review of the recent Psp literature suggests that the current consensus of most authors (including this one) is that dissipation of the PMF is probably an inducing signal for the Psp response. Whether it is a precise signal itself, or acts indirectly, is not yet known. Indeed, we do not have definitive answers to two fundamental questions: Is there a single inducing signal or are there independent signals, and what is the precise nature of the inducing signal(s)?
PMF dissipation as an inducing signal of the Psp response
It has either been shown or can be reasonably speculated that many, if not all, of the Psp-inducing conditions affect the PMF. The E. coli Psp response is induced in stationary phase, when cellular energy status is relatively low, and by the uncouplers CCCP and octanoic acid (Weiner and Model, 1994). Expression of a translocation-defective PhoE protein (an outer membrane porin) significantly reduces the membrane potential of a pspA null mutant, and induces the Psp response in a psp+ strain (Kleerebezem et al., 1996). Substrate-saturation of the PMF-dependent Tat export pathway induces the Psp response and increased pspA expression helps to relieve this saturation (DeLisa et al., 2004). Export defects such as sec null mutations or depletion of the YidC protein induce the Psp response (Jones et al., 2003; van der Laan et al., 2003). YidC depletion reduces PMF due to defects in assembly of cytochrome o oxidase and the F0F1-ATPase (van der Laan et al., 2003). In Y. enterocolitica, E. coli and Salmonella enterica, atp null mutations (encoding the F0F1-ATPase) also cause induction of the Psp response (Maxson and Darwin, 2004; Becker et al., 2005). An S. enterica rpoE null mutation decreases the PMF and induces the Psp response (Becker et al., 2005). Psp-inducing environmental cues or overproduced envelope proteins may also affect the permeability of the cytoplasmic membrane to protons.
Questions and problems remain. Psp may be induced by one of the many consequences of PMF dissipation (Gage and Neidhardt, 1993), or by something that precedes the drop in PMF, such as altered cytoplasmic membrane properties. Teasing these possibilities apart is a daunting experimental task. Another concern is that dissipation of the PMF does not specifically induce Psp, but instead induces expression of many other proteins, including members of the cytoplasmic heat-shock response (Gage and Neidhardt, 1993). We also do not know if the overproduced proteins that induce the Psp response, most notably secretins, cause a reduction in the PMF. In the one case where this was tested the results actually argue against PMF dissipation as the inducing signal. The mutant PhoE porin mentioned above does not affect PMF in a wild-type strain despite the fact that the Psp response is induced (Kleerebezem et al., 1996).
Induction by secretins
Secretins, normally localized to the outer membrane, mislocalize in the absence of specific chaperone-like pilot proteins (Hardie et al., 1996). A significant amount of the filamentous phage pIV secretin is naturally mislocalized (Brissette and Russel, 1990) because a pilot protein is not encoded in the phage genome. This mislocalization of secretins is a potent Psp-inducing signal, but it is not clear why. Secretin induction of the Psp response was discussed in depth in a previous review article (Model et al., 1997). I will summarize the major conclusions and refer the reader to that review for a more detailed discussion. One important conclusion was that the secretin-stimulated induction of the Psp response was not due to the formation of a leaky outer membrane. The pIV secretin of filamentous phage does not cause a leaky membrane, but does induce Psp. In contrast, the phage pIII protein, which is not a secretin, causes a leaky outer membrane without inducing Psp. Other conclusions were that prolonged Psp induction required continued secretin synthesis, and that induction was not caused by the ensuing instability of a mislocalized secretin.
There is plenty of motivation to focus on mislocalized secretins as a Psp-inducing signal. Secretins are essential components of type II and type III secretion systems, which play a role in the virulence of some bacteria. Secretin mislocalization is likely to be a physiologically relevant inducing signal. The Y. enterocolitica Psp response is induced during the synthesis of the functional Ysc type III secretion system, which is essential for virulence (Darwin and Miller, 2001). This is probably because some of the component secretin molecules fail to insert in the outer membrane. Indeed, overproduction of the YscC secretin is toxic in Y. enterocolitica psp null mutants, and this toxicity is exacerbated in the absence of the YscC-specific pilot protein YscW (Darwin and Miller, 2001). Secretins also provide an example of a functional link between Psp induction and phenotype because secretin mislocalization inhibits the growth of some Y. enterocolitica psp null mutants (see below). One of the most compelling reasons to study secretin induction of the Psp response is the exquisite specificity. This was most dramatically demonstrated in a microarray study from Martin Buck's group. The overexpressed pIV secretin induced the pspA and pspG promoters in both E. coli and Salmonella typhimurium, whereas the expression of all other genes was not significantly affected (Lloyd et al., 2004). This raises doubt about the most likely explanation for secretin induction of the Psp response, which is that secretin mislocalization reduces the PMF. If this were the case, one might expect a more global change in the transcriptional profile. However, it is also possible that secretin mislocalization causes a relatively small reduction in PMF that is sufficient to induce Psp, but insufficient to cause global transcriptional changes.
How does secretin mislocalization induce the Psp response? Perhaps the mislocalized secretin aggregates at the periplasmic face of the cytoplasmic membrane and causes an alteration in its properties. If so, the effect of a secretin on the cytoplasmic membrane would have to be different from those of other proteins that may aggregate at the cytoplasmic membrane but do not induce Psp. Alternatively, the inefficient attempt of the secretin to insert into the outer membrane may generate the inducing signal. The latter hypothesis was previously favoured because known inducers of the Psp response were proteins with a potential to interact with the outer membrane (secretins and mutant PhoE and LamB porins; Model et al., 1997). However, we now know that the overexpression of some cytoplasmic membrane proteins specifically induces the Y. enterocolitica Psp response (Maxson and Darwin, 2004), which favours the hypothesis that a signal is generated at the cytoplasmic membrane. Furthermore, overproduction of cytoplasmic membrane-localized lipoproteins induces Psp in E. coli, whereas outer membrane-localized versions do not (Robichon et al., 2003). However, it should be noted that overexpressed secretins prevent the growth of a Y. enterocolitica pspC null mutant, whereas overexpressed cytoplasmic membrane proteins do not (Maxson and Darwin, 2004). This may indicate distinct mechanisms of induction for the two classes of protein.
Global transcriptional changes that include the psp genes
Microarray studies have revealed interesting conditions during which the Psp response is induced. One of the most exciting is the observation that the Psp responses of both S. enterica and Shigella flexneri are strongly induced during macrophage infection (Eriksson et al., 2003; Lucchini et al., 2005). Another was the observation that the E. coli Psp response is induced during biofilm formation and a pspF null mutant has a biofilm-formation defect (Beloin et al., 2004). These studies provide examples of physiologically relevant conditions that induce the Psp response, which increases the motivation and excitement to study this system.
Signal transduction within the Psp system
Whatever the precise nature of the inducing signal(s), the Psp system must apparently translate stress in the cell envelope to changes in gene expression in the cytoplasm. PspA, B, C and F play critical roles in this regulatory pathway (Fig. 1). PspF is a DNA-binding transcriptional activator of the psp genes. The favourable biochemical properties of PspF have made it an excellent model to study the enhancer-binding protein family of activators, of which it is a member. Therefore, the mechanism by which PspF activates transcription has been studied extensively, most recently in an impressive series of studies from Martin Buck and colleagues (e.g. Schumacher et al., 2004; Rappas et al., 2005).
PspA is the pivotal component of the signal transduction pathway, facilitating signal transfer between the integral membrane proteins PspB and PspC and the cytoplasmic transcriptional regulator, PspF (Fig. 1). A pspA null mutation causes constitutive expression of the Psp response in both E. coli and Y. enterocolitica (Weiner et al., 1991; Darwin and Miller, 2001). This is because PspA normally inhibits PspF activity (Dworkin et al., 2000) via direct interaction of PspA with the ATPase domain of PspF (Elderkin et al., 2002). Presumably, basal levels of PspA in the uninduced state are sufficient to prevent significant PspF-dependent expression of the psp genes (Fig. 1).
The E. coli PspB and PspC proteins also interact with PspA (Adams et al., 2003). An attractive model is that PspB and PspC interact with PspA only when the Psp response is induced (Fig. 1). This could be mediated by a change in PspB/C properties upon induction, a change in PspA or even a change in PspF. The first possibility seems most likely because the cytoplasmic membrane location of PspB and PspC makes them obvious candidates as sensors of extracytoplasmic stress. Consistent with this, pspB or pspC null mutations prevent or reduce induction of the Psp response in E. coli and Y. enterocolitica (Weiner et al., 1991; Darwin and Miller, 2001). PspC, which has a periplasmic domain, is perhaps a prime candidate for sensing an inducing signal. However, although a pspC null mutation severely reduces psp gene induction, it does not completely abolish it (Kleerebezem et al., 1996; M.E. Maxson and A.J. Darwin, unpubl. data). Therefore, perhaps PspC plays an important role in signal detection, but is either nonessential, or has a partially redundant function with another protein.
The current signal transduction model is compelling and supported by the experimental data described above. However, a notable gap in the data concerns the dynamics of the protein–protein interactions. We do not know if these interactions change depending on the activation status of the Psp response, as the model predicts (Fig. 1). Another question is raised by the dramatic increase in PspA concentration when the Psp response is induced. PspA binds to PspF in vitro and inhibits its activity (Elderkin et al., 2002). When PspA concentration is dramatically elevated during induction, why is it apparently less able to inhibit PspF activity? One possibility is that the conformation of PspA or PspF changes so that their capacity to interact is diminished. An alternative hypothesis is that although the total concentration of PspA increases upon induction, most of it now associates with the cytoplasmic membrane and the soluble cytoplasmic concentration decreases. Only the cytoplasmic PspA might be competent to bind to PspF. The reduction in cytoplasmic PspA could be mediated in part by its interaction with PspB/C in the membrane (Fig. 1). However, the low abundance of PspBC compared to PspA makes it likely that PspA also associates with some other membrane component.
The physiological role of the Psp response
In relative terms we know a lot about the induction and regulation of the Psp response, but little about the physiological role it plays. This is unfortunate because the most important questions are what does it do and how does it do it? PspA is the most abundant Psp protein upon induction and has widely been assumed to be an effector of the system (Model et al., 1997). In E. coli, several phenotypes have been solely attributed to PspA: survival in stationary phase at alkaline pH, maintenance of the PMF under inducing conditions, and efficient export of proteins by the sec and tat pathways (Kleerebezem and Tommassen, 1993; Weiner and Model, 1994; Kleerebezem et al., 1996; DeLisa et al., 2004). In S. enterica, a pspA null mutation reduces resistance to the protonophore CCCP, and exacerbates the stationary phase survival defect of an rpoE null mutant (Becker et al., 2005). PMF maintenance may explain all these phenotypes. Survival in stationary phase represents a low energy state of the cell, and the sec and tat pathways directly require adenosine triphosphate (ATP) and PMF respectively. If this is the function of PspA, then how is it achieved? It may help to maintain cytoplasmic membrane integrity, interact with other proteins that use the proton gradient or ATP, or even play a direct role in proton transport. An advance came with the publication of a low resolution structure of E. coli PspA (Hankamer et al., 2004). PspA forms a symmetrical oligomeric ring structure. The authors postulated how this structure might interact with PspF in non-inducing conditions and with cytoplasmic membrane components during induction. The authors also speculated that the ring-like PspA structure might interact with the F1 subunit of the F0F1-ATPase. PspA homologues are widespread in bacteria and higher organisms. In cyanobacteria and plants these homologues, which are not part of an analogous Psp response system, have a role in thylakoid biogenesis (Li et al., 1994; Westphal et al., 2001). Therefore, like PspA, they have a function that involves interaction with membrane components and possibly an effect on membrane integrity. Understanding PspA function could shed light on a highly conserved family of proteins.
Although many things point to PspA as an essential effector of the Psp response, there is one important psp mutant phenotype in which PspA is apparently not involved. A Y. enterocolitica pspC null mutant cannot grow when a secretin protein is mislocalized, either by overexpression or when produced as part of a functional type III secretion system (Darwin and Miller, 2001). This almost certainly explains why the pspC mutant is attenuated in a mouse model of infection. This is perhaps the most severe phenotype reported for any psp null mutant. Growth of a pspC null mutant stops 3–4 h after initiation of secretin synthesis, and the cells die rapidly thereafter. However, this secretin-induced growth defect does not occur in a pspA null mutant (Darwin and Miller, 2001). Therefore, PspA is dispensable for growth during secretin mislocalization. We do not understand the reason for this phenotype. The fact that PspA is not involved raises the possibility that secretin mislocalization does not deplete the PMF, and that the mechanism by which secretins induce the Psp response is distinct from some other inducing cues. It also suggests that at least one of the Psp proteins, besides PspA, has a physiological role distinct from its role in the signal transduction pathway. The secretin-induced growth defect has not been reported for E. coli psp mutants. However, in unpublished experiments my group has found that it does occur in E. coliΔ(pspFABCDE) and S. typhimuriumΔ(pspFABC) mutants.
A new component of the Psp system
Induction of the Psp response allows PspF to activate transcription, but until recently the only promoter it was known to activate was that of the pspA operon itself. In 2001 evidence emerged that PspF-dependent expression of at least one unknown gene alleviated the secretin-induced growth defect of some Y. enterocolitica psp null mutants (Darwin and Miller, 2001). This gene (pspG) has now been identified in both Y. enterocolitica and E. coli (Green and Darwin, 2004; Lloyd et al., 2004).
PspF, an enhancer-binding protein family member, should only activate promoters that use the RpoN sigma factor (σ54), such as the pspA promoter itself. The C nucleotide at position −12 is highly conserved in a large majority of RpoN-dependent promoters from several species, with the striking exception of most pspA promoters (Green and Darwin, 2004). Working with a hypothesis that this feature may be a characteristic of PspF-dependent promoters, a bioinformatics approach was used to identify similar Y. enterocolitica RpoN-dependent promoters (Green and Darwin, 2004). One other promoter was identified, upstream of a previously uncharacterized gene, which was designated pspG. pspG expression is positively regulated by PspF, negatively regulated by PspA and induced by secretin mislocalization. Increased expression of pspG, in the absence of PspA, alleviates the secretin-induced growth defect of a ΔpspA operon mutant. pspG is not linked to the pspA operon and is predicted to encode a small hydrophobic protein that traverses the cytoplasmic membrane. In an otherwise wild-type cell, a pspG null mutation does not cause a secretin-induced growth defect or affect the Psp signal transduction pathway (Green and Darwin, 2004; R.C. Green and A.J. Darwin, unpubl. data).
Around the same time, pspG was also identified in E. coli and S. typhimurium. In this case a microarray approach was used to find promoters induced by the pIV secretin (Lloyd et al., 2004). Remarkably, the only promoters that responded to pIV were the pspA and pspG promoters. Therefore, these two promoters are probably the only members of the Psp regulon. As in Y. enterocolitica, a pspG null mutation does not affect regulation of the Psp response. Interestingly, the authors reported that the overexpression of pspG caused a significant reduction in motility, which may reflect an effect on the PMF.
The function of pspG will obviously require further investigation. In Y. enterocolitica the pspG mutant phenotype is highly complicated. In a wild-type strain, deletion of pspG does not cause a secretin-induced growth defect. However, in a strain lacking the entire pspA operon, in which pspG is overexpressed because of the absence of PspA, subsequent deletion of pspG exacerbates the existing secretin-induced growth defect. Therefore, whatever the reason for the growth defect of a pspA operon deletion mutant, greatly increased concentrations of PspG appear to alleviate it. It is striking that pspG homologues are only found in bacteria that have a pspABC operon (see below), despite the fact that the two loci are not linked. However, pspG is not conserved in all bacteria that have a pspABC operon (see below; Fig. 2).
Conservation of the Psp system
The broad significance of the Psp response has already increased beyond E. coli, in which it was originally described. The availability of numerous complete bacterial genome sequences now facilitates an examination of how well the Psp response is conserved. As mentioned earlier, PspA homologues are present in many different organisms, but most of them are not adjacent to any other psp genes and unlikely to be part of a Psp response system. For this article, I have identified candidate Psp response systems as those where at least two adjacent psp genes are present in a completed bacterial genome sequence (Fig. 2). This analysis reveals that putative Psp systems are present in many Gammaproteobacteria, as well as one Deltaproteobacterium (Desulfovibrio vulgaris) and one Alphaproteobacterium (Zymomonas mobilis). It is immediately apparent that PspF, A, B and C are always conserved, whereas PspD, E and G have only limited conservation (Fig. 2). The implications of this are uncertain, but it may be that PspFABC constitute the minimal functional Psp response system, with PspD, E and G playing accessory roles in some organisms (for example, PspE is a rhodanese enzyme; Adams et al., 2002). Consistent with this, null mutations of the E. coli and Y. enterocolitica pspF, pspA, pspB and pspC genes have robust phenotypes, whereas disruptions of the other psp genes do not. Two additional points are noteworthy. First, a TBLASTN search with E. coli PspG reveals that homologues are only present in organisms that have the pspFpspABC genes, despite the fact that the two loci are not linked. Second, the ycjX and ycjF genes, which are in the pspA operon of Y. enterocolitica (Darwin and Miller, 2001), are also adjacent to the pspABC loci of several other species. It is not known how often ycjXF are co-transcribed with pspABC, although it is likely to occur in Idiomarina loihiensis, where the pspC and ycjX genes overlap. The functions of YcjX and YcjF are not known.
Although the Psp response remains something of an enigma, we are steadily beginning to assemble a clearer picture. Genetic and biochemical evidence has provided strong support for the current signal transduction model (Fig. 1), around which future experiments can be designed. There is a consensus opinion that a reduction in the energy status of the cell, and possibly the PMF in particular, is at least one inducing signal. The broad significance of the Psp response has increased with the revelation that it plays a role in Yersinia virulence and E. coli biofilm formation, and is activated in Salmonella and Shigella during macrophage infection. Some of the major challenges for the future include probing the details and dynamics of the Psp protein–protein interactions, understanding how secretin mislocalization induces the Psp response with such exquisite specificity, ascertaining why secretin mislocalization inhibits the growth of psp mutants, and determining the precise mechanism by which PspA may help to maintain the PMF.
I apologize to those authors whose work is not cited because of editorial limitations and a desire to primarily focus on developments that have followed a 1997 review article about the E. coli Psp system. I thank Grace Axler-DiPerte, Heran Darwin, Michelle Maxson and Michael Pearce for critical comments on the manuscript. Work in my laboratory is supported by Public Health Service grant AI-052148 from the National Institute of Allergy and Infectious Diseases, and by a grant from the Speaker's Fund for Biomedical Research: Toward the Science of Patient Care, awarded by the City of New York.