Regulation of the bacterial phage-shock-protein (Psp) system involves communication between integral (PspBC) and peripheral (PspA) cytoplasmic membrane proteins and a soluble transcriptional activator (PspF). In this study protein subcellular localization studies were used to distinguish between spatial models for this putative signal transduction pathway in Yersinia enterocolitica. In non-inducing conditions PspA and PspF were almost exclusively in the soluble fraction, consistent with them forming an inhibitory complex in the cytoplasm. However, upon induction PspA, but not PspF, mainly associated with the membrane fraction. This membrane association was dependent on PspBC but independent of increased PspA concentration. Analysis of psp null, overexpression and altered function mutants further supported a model where PspA is predominantly membrane associated only when the system is induced. Activation of the Psp system normally leads to a large increase in PspA concentration and we found that this provided a second mechanism for its membrane association, which did not require PspBC. These data suggest that basal PspFABC protein levels constitute a regulatory switch that moves some PspA to the membrane when an inducing trigger is encountered. Once this switch is activated PspA concentration increases, which might then allow it to directly contact the membrane for its physiological function.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
pspA operon expression is induced by heat shock, osmotic shock, ethanol, defects in protein transport across the cytoplasmic membrane and by the overproduction of some integral membrane proteins, including secretins (reviewed by Model et al., 1997; Darwin, 2005; Joly et al., 2010). A common feature of these events is thought to be compromise of cytoplasmic membrane integrity. It has been suggested that the Psp system functions to counteract this and perhaps specifically to prevent subsequent dissipation of proton motive force (PMF). Consistent with this, the E. coli PspA protein has been specifically linked to maintenance of the PMF both in vivo (Kleerebezem et al., 1996) and in vitro (Kobayashi et al., 2007).
The regulation of pspA operon expression has been extensively studied with important roles being assigned to PspF, PspA, PspB and PspC. PspF is an enhancer binding protein required for expression of the pspAσ54-dependent promoter (Jovanovic et al., 1996). PspA has been described as a peripheral cytoplasmic membrane protein that inhibits PspF via a direct physical interaction (Dworkin et al., 2000; Elderkin et al., 2002). Finally, PspB and PspC are integral cytoplasmic membrane proteins required for stress-dependent induction of the pspA operon (Weiner et al., 1991; Kleerebezem et al., 1996; Maxson and Darwin, 2006a; Gueguen et al., 2009). Therefore, it has been speculated that PspBC somehow modulate the PspA–PspF inhibitory complex depending on the presence or absence of an inducing trigger.
One deficiency in the current understanding of Psp regulation is that there is no clear spatial explanation of how integral (PspBC) or peripheral (PspA) cytoplasmic membrane proteins communicate with each other and with a soluble transcriptional activator (PspF). One possibility is that a PspA–F inhibitory complex is formed at the membrane and detection of an inducing trigger by PspBC causes PspA to release PspF to the cytoplasm. Perhaps a less likely possibility is that upon encounter of an inducing stimulus PspF remains associated with PspA at the inner surface of the membrane but the nature of the interaction is changed so that PspF can activate transcription. However, when PspA was originally discovered, subcellular localization experiments led the authors to describe it as cytoplasmic, about half of which is associated with the cytoplasmic membrane (Brissette et al., 1990). Therefore, another spatial regulatory model is that the PspA–PspF complex forms in the cytoplasm. When an inducing trigger is encountered some PspA moves to the membrane, which frees some PspF to activate transcription.
In this study, we attempted to distinguish between the possibilities described above by investigating the subcellular location of the endogenous Y. enterocolitica PspF and PspA proteins in both Psp-uninducing and inducing conditions. Our data suggest that PspA changes its predominant location from the cytoplasm to the peripheral cytoplasmic membrane upon induction of the Psp system. This model stands up to interrogation by Psp proteins with altered regulatory function. Our data also suggest two distinct mechanisms by which PspA associates with the cytoplasmic membrane, with potential regulatory and physiological implications.
Endogenous PspA is predominantly membrane associated only when the Psp response is induced
The goal of this work was to investigate and distinguish between spatial models for the putative signal transduction system that regulates pspA operon expression (see Introduction). Our approach was designed to examine the subcellular location of the wild-type PspF and PspA proteins encoded by their native chromosomal genes. This was important because we wanted to examine PspA localization when the protein was present at its natural uninduced and induced concentrations. Furthermore, artificially changing pspA or pspF expression (e.g. by encoding them on multicopy plasmids) can interfere with normal regulation of the Psp system. Another advantage of studying the native proteins is that it avoids any risk of aberrant effects and/or altered function resulting from the proteins being fused to reporter domains (e.g. epitope tags or fluorescent protein domains). Subcellular fractionation and immunoblotting is a well-established method used to monitor changes in bacterial protein localization between cytoplasm and membrane in several systems (e.g. Tanaka et al., 2000; Coutts et al., 2002; Klopprogge et al., 2002). Therefore, we chose this approach to monitor PspF and PspA localization in cells with a wild-type pspF–pspABCDycjXF locus. The strain also had a single copy Φ(pspA–lacZ) operon fusion located at the ara locus (Maxson and Darwin, 2005) so that pspA promoter activation status could be monitored by β-galactosidase activity.
pspA operon expression is specifically induced by secretin protein production, either endogenously (Darwin and Miller, 2001) or by overproduction (e.g. Seo et al., 2007). Therefore, we began by determining the location of PspF and PspA in bacterial cells containing a tacp–ysaC secretin expression plasmid or the empty vector control. β-Galactosidase assays revealed that Φ(pspA–lacZ) expression was induced over 10-fold by YsaC secretin production, which confirmed activation of the Psp system (Fig. 1). Then, subcellular fractions were analysed by immunoblotting. Detection of the integral cytoplasmic membrane protein FtsH and the soluble cytoplasmic protein DnaK was used to validate the fractionation procedure (Fig. 1; for all subsequent experiments these controls were done but are not shown). This analysis showed that PspF and PspA were exclusively or predominantly, respectively, in the soluble fraction in the absence of YsaC secretin production (Fig. 1). However, when YsaC was produced most of the PspA was now found in the membrane fraction, whereas the location of PspF did not change. These data support the model in which the predominant location of PspA changes from cytoplasm to membrane upon induction of the Psp system.
Next, we tested whether the change in PspA localization was a secretin-specific phenomenon. YE0566 is a small cytoplasmic membrane protein of unknown function that was identified as a specific inducer of Y. enterocolitica pspA operon expression (Maxson and Darwin, 2004). Therefore, we repeated the experiment using a tacp–YE0566 expression plasmid. The results were very similar to those obtained using YsaC secretin production (Fig. 1). Φ(pspA–lacZ) expression was induced over 10-fold by YE0566 and most of the PspA associated with the membrane fraction. Therefore, the change in PspA location is likely to be a general feature of Psp system induction.
The change in PspA localization is independent of PspF and of any increase in PspA concentration
The preceding experiments showed that PspA associates with the membrane upon induction of the Psp system. Next we wanted to test the effect of some psp null mutations on this phenomenon. For these experiments the YE0566 inducer was valuable because, unlike secretins, it is not specifically toxic to psp null strains (Maxson and Darwin, 2004). Therefore, both wild-type and psp null strains grow at the same rate when using this inducer.
We hypothesize that PspB and/or PspC detect an inducing signal and communicate with the cytoplasmic PspA protein by recruiting it to the membrane, which leaves some PspF free in the cytoplasm. If this is correct then the first part of this signal transduction pathway might still function without PspF present. Therefore, we first tested the effect of a pspF null mutation. This experiment is possible due to low-level PspF-independent expression of the Y. enterocolitica pspA operon (Maxson and Darwin, 2006b). The reduced amount of PspA in a pspF null strain is still detectable with our polyclonal antiserum.
As expected, Φ(pspA–lacZ) expression was reduced by a ΔpspF mutation, and was unaffected by YE0566 production (Fig. 2). However, in both pspF+ and ΔpspF cells the PspA protein was predominantly in the soluble fraction in the absence of YE0566 but became membrane associated in the presence of YE0566 (Fig. 2). This suggests that the PspBC–PspA components of a putative signal transduction system are still functional when the terminal component, PspF, is absent. Furthermore, in a ΔpspF strain, the PspA concentration cannot increase when an inducing signal is encountered (Fig. 2). Therefore, we can conclude that membrane association of PspA is not simply a consequence of an increase in its concentration.
The change in PspA localization is PspBC-dependent
PspBC form an integral inner membrane complex required for stress-dependent activation of the pspA promoter (Maxson and Darwin, 2006a). If PspBC detect an inducing signal and recruit PspA to the membrane, then a ΔpspBC mutation should prevent PspA from associating with the membrane fraction. Therefore, we next tested the effect of a ΔpspBC mutation.
Consistent with published data (Maxson and Darwin, 2006a) the ΔpspBC mutation abolished induction of Φ(pspA–lacZ) expression in response to YE0566 (Fig. 3). Deletion of pspBC also abolished the localization of PspA into the membrane fraction when YE0566 was produced (Fig. 3). Therefore, PspBC are required for stress-responsive PspA membrane association. This experiment also rules out the possibility of YE0566 overproduction distorting/altering membrane properties in such a way that PspA now directly associates with it non-physiologically. Rather, it is a PspBC-dependent event that is probably part of the normal PspFABC signal transduction pathway.
Overproduction of PspBC can activate Φ(pspA–lacZ) expression in the absence of a Psp system inducer (Maxson and Darwin, 2006a). Therefore, we extended our analysis of PspBC involvement by using an araBp–pspBC expression plasmid to test the effect of PspBC overproduction on PspA localization. This experiment was carried out in a ΔpspF strain to prevent PspBC overproduction from massively elevating PspA concentration. When PspBC were not overproduced, PspA only associated with the membrane fraction when the YE0566 Psp inducer was present, which also served to confirm that this event is PspF-independent (compare lanes 3 and 9 in Fig. 3B). However, when PspBC were overproduced, PspA associated with the membrane fraction both with and without YE0566 production (lanes 6 and 12 in Fig. 3B). Therefore, PspBC overproduction is sufficient to cause PspA membrane association (Fig. 3) as well as activation of the pspA promoter (Maxson and Darwin, 2006a). One possible explanation is that PspBC are required for PspA membrane association because of a direct protein–protein interaction (see Discussion). We also tested the effect of overexpressing pspB or pspC alone from the pBAD33 plasmid, but neither was sufficient to localize PspA to the membrane (data not shown). This suggests that PspB is not sufficient. However, the results for PspC are inconclusive because PspC protein is apparently not stable when overproduced without PspB (Gueguen et al., 2009).
A second mechanism for PspA membrane association
Stress-responsive PspA association with the membrane requires PspBC and is independent of any increase in PspA concentration (Figs 1–3). Therefore, basal Psp protein levels appear to constitute a regulatory switch that controls PspF activity in response to stress. As outlined above, this switch might depend on a direct PspBC–PspA interaction recruiting PspA to the membrane. However, once the regulatory switch has been activated the normal consequence in a wild-type cell is activation of the pspA promoter and a massive increase in PspA protein concentration relative to PspBC (because of probable partial transcription termination in the pspA–pspB intergenic region; e.g. Brissette et al., 1991; Seo et al., 2007). It seems unlikely that PspBC could then provide sufficient binding capacity to recruit most of this PspA to the membrane (unless initial interaction of PspA with PspBC serves as a nucleation event for the formation of higher-order PspA oligomers, stabilized by the first PspABC complex). Therefore, we hypothesized that PspA might have an inherent but relatively weak affinity for the membrane, or some membrane component, that only functions at elevated PspA concentration. The next series of experiments were designed to investigate this possibility.
We first tested whether increasing the concentration of PspA was sufficient to localize PspA to the membrane in the absence of a Psp inducer. Strains contained an araBp–pspA expression plasmid (increased PspA concentration) or the empty vector control (endogenous PspA). Consistent with previous results, endogenous PspA only associated with the membrane fraction when the YE0566 Psp inducer was present (Fig. 4A). However, with increased PspA concentration the majority of it associated with the membrane fraction both with and without YE0566 production (Fig. 4A). Therefore, increasing PspA concentration is sufficient to cause its membrane association in the absence of a normal Psp-inducing signal.
Next, we tested if increasing the concentration of PspA allowed it to associate with the membrane in the absence of PspBC. For this experiment a ΔpspBC strain contained the araBp–pspA expression plasmid or the empty vector control. The results showed that increasing PspA concentration still allowed it to associate with the membrane in a ΔpspBC mutant (Fig. 4B). Therefore, this mechanism of PspA membrane association does not require PspBC.
These and preceding data suggest two independent mechanisms by which PspA can associate with the membrane. The first requires a Psp-inducing signal (stress), the presence of PspBC and only basal levels of PspA protein (Figs 1–3). We refer to this as the regulatory switch. The second requires elevated PspA concentration and is independent of PspBC (Fig. 4), possibly indicating direct membrane contact. This allows us to propose a model of temporal events during and after activation of the pspA promoter (Fig. 5). The basal levels of PspA and PspBC serve as a regulatory switch, with PspBC in the membrane and PspA inactivating PspF in the cytoplasm. When PspBC detects an inducing stress they recruit some PspA to the membrane, which frees active PspF (Fig. 5, parts 1 and 2). In turn, this leads to activation of the pspA promoter and a large increase in PspA concentration, which then allows it to directly contact the membrane (or some membrane component; Fig. 5, part 3). This direct contact might then allow PspA to perform its physiological function.
Properties of PspA in the membrane fraction are consistent with a peripheral membrane protein
The subcellular fractionation procedure relies on centrifugation to separate membrane and soluble fractions. Therefore, we wanted to confirm that the properties of PspA in the membrane fraction are consistent with those of a peripheral membrane protein. This was most important for PspA overproduced from the araBp–pspA plasmid in the preceding experiments. To address this the standard fractionation procedure was repeated on strains containing the tacp–YE0566 plasmid alone (endogenous PspA) or containing the araBp–pspA plasmid (increased PspA concentration). Then, the membrane pellets were extracted with pH 11 buffer to release peripheral membrane proteins (Steck and Yu, 1973; Russel and Model, 1982; Brissette et al., 1990). Extraction with pH 7 buffer served as a negative control.
This analysis showed that both endogenous and overproduced PspA was released from the membrane by alkaline extraction (Fig. 6). The integral membrane protein FtsH was unaffected by this treatment. Therefore, PspA in the membrane fraction has properties consistent with a peripheral cytoplasmic membrane protein. Furthermore, extraction of the membrane preparations with 1% Triton X-100 solubilized PspA, which indicates that PspA is not forming inclusion bodies (data not shown; Russel and Kazmierczak, 1993).
Isolation and analysis of PspA altered function mutants
Up to this point our data had led us to propose a model whereby basal levels of the PspABC proteins constitute a regulatory switch mechanism that moves PspA to the membrane when an inducing trigger is encountered (Fig. 5). This PspA movement is predicted to be critical for activation of the pspA promoter. Next, we wanted to rigorously test this model by determining the behaviour of mutant strains with altered Psp regulatory function. The regulatory switch model makes clear predictions about PspA localization based on the regulatory phenotype of the mutation being studied. For mutations that prevent activation of the pspA promoter, PspA should always be soluble, whereas for mutations that cause constitutive pspA promoter activity, PspA should be constantly membrane associated. We started by analysing the behaviour of PspA proteins with altered function mutations.
Ongoing work in our laboratory has identified a panel of pspA single amino acid substitution mutations that cause constitutive repression of Φ(pspA–lacZ) expression (i.e. an inability to activate in response to an inducing trigger). The identification of these mutants is outlined in Experimental procedures and most of them will be described elsewhere. For the purposes of this study we arbitrarily selected two mutants with amino acid substitutions in different parts of the primary sequence (PspA–N14D and PspA–L200S).
First, to characterize the regulatory phenotypes of these mutant PspA proteins in a physiologically relevant context, the wild-type chromosomal pspA gene in the pspABCDycjXF operon was exchanged for the mutant gene in strains that also had a Φ(pspA–lacZ) operon fusion. β-Galactosidase activity was determined either with or without YE0566 production. As in the preceding experiments, with a wild-type pspA gene Φ(pspA–lacZ) expression was induced well over 10-fold by YE0566 production (Fig. 7). However, pspA genes encoding either the N14D or L200S mutants essentially abolished this induction.
The regulatory switch model predicts that these mutant PspA proteins should remain in the cytoplasm in the presence of a Psp-inducing signal. Therefore, we tested the effects of the pspA mutations on localization of the PspA protein. To focus on the regulatory switch mechanism of PspA membrane association, rather than that caused by elevated PspA protein concentration (Fig. 5), the experiment was carried out in a ΔpspF strain background. This prevented elevation of PspA concentration in response to YE0566, which ensured similar PspA levels in wild-type and mutant strains (Fig. S1).
As expected, the wild-type PspA protein associated with the membrane when YE0566 was produced, but was soluble in the absence of YE0566 (Fig. 7). In contrast, the PspA–N14D and PspA–L200S proteins did not localize to the membrane fraction when YE0566 was produced (Fig. 7). These results showed that PspA mutants unable to allow activation of the pspA promoter also fail to associate the membrane in response to an inducing trigger. This is fully consistent with the regulatory switch model.
Analysis of PspC altered function mutants
We previously reported the isolation and characterization of pspC mutants with contrasting regulatory phenotypes (Gueguen et al., 2009). Some failed to activate Φ(pspA–lacZ) expression in response to secretin protein production (activation defect), whereas others resulted in high Φ(pspA–lacZ) expression with or without secretin production (constitutive). The regulatory switch model predicts that the activation defect mutants should cause PspA to remain soluble +/− secretin, whereas the constitutive mutants should cause PspA to be membrane associated +/− secretin. The next experiments were designed to test these predictions. For this we chose the constitutive PspC–G45W and PspC–Δcyto (cytoplasmic domain deletion) mutants, and the activation defect PspC–V125D mutant (Gueguen et al., 2009).
We tried the same strategy used for the pspA mutants, designed to focus on the regulatory switch mechanism of PspA membrane association. The chromosomal pspC gene was replaced with a mutant pspC gene in a ΔpspF strain. When PspA localization was analysed the activation defect PspC–V125D mutation caused PspA to remain soluble +/− secretin as expected (data not shown). However, the constitutive pspC mutants did not cause PspA to associate with the membrane (it remained soluble +/− secretin; data not shown). This appeared to invalidate our model. However, we realized that this might be an artefact of using a ΔpspF strain in this specific case. We do not yet know the molecular mechanism for the constitutive PspC mutants, but one possibility is increased affinity for the PspA protein. Therefore, because the ΔpspF mutation significantly lowers basal level expression of the pspA operon, the reduced PspC concentration might now be insufficient for an affinity-mediated constitutive phenotype to manifest. To circumvent this we needed an approach that kept PspA protein levels constant +/− secretin but, unlike a ΔpspF mutation, did not reduce PspABC below their normal basal levels. Other ongoing work in our laboratory has developed a strain that meets these requirements.
The pspF–pspAp region of the chromosome was replaced with a lacIq–tacp fragment, placing the pspA operon under tac promoter control. Therefore, the pspABC expression level is now unaffected by Psp-inducing signals. In order to be able to study regulation of the wild-type pspA promoter the strain also had the Φ(pspA–lacZ) operon fusion at the ara locus, and pspF expressed from its own promoter was reintroduced at the yenR deletion site (because the native pspF gene and the common pspF–pspA control had been replaced by lacIq–tacp). More details on this strain construction are in the Experimental procedures and full characterization of this system will be described elsewhere. Without IPTG the leaky tac promoter leads to PspABC levels that are higher than basal levels in a wild-type strain, but lower than those in a wild-type strain when the Psp system is induced (N.K. Horstman and A.J. Darwin, unpubl. data). Importantly, this level of PspABC proteins still function as a regulatory switch that allows induction of Φ(pspA–lacZ) expression in response to secretin (Fig. 8).
The chromosomal pspC gene was replaced by the mutant genes in this tacp–pspA operon strain background. β-Galactosidase assays revealed that the mutant PspC proteins maintained their constitutive or activation defect phenotypes with regard to Φ(pspA–lacZ) expression (Fig. 8). With wild-type PspC the amount of membrane associated PspA in the absence of induction was more than in strains with the pspA operon controlled by its native promoter (e.g. compare Fig. 8 with Fig. 1). This is probably due to the elevated PspA concentration in the tacp strain compared with the basal level in the wild type, causing some of it to directly contact the membrane (Figs 4 and 5). However, in the presence of secretin the ratio of membrane associated to soluble PspA increased as expected (Fig. 8). Compared with the wild type, both constitutive PspC mutants elevated the ratio of membrane associated to soluble PspA in the absence of secretin. In contrast, the activation defect PspC–V125D mutation reduced the ratio of membrane associated to soluble PspA, and prevented any change in response to secretin. The Φ(pspA–lacZ) expression and PspA localization profiles of these mutants are fully consistent with predictions made by the regulatory switch model. Therefore, activation of the pspA promoter correlates with increased membrane association of PspA, independent of an elevation in its steady state concentration.
Bacterial two-hybrid analysis supports PspA–PspF, PspA–PspB and PspA–PspC associations
The regulatory switch model predicts that PspA can associate with PspF as well as with PspB and/or PspC (Fig. 5). This has not previously been tested for the Y. enterocolitica proteins. Therefore, to investigate this we used a bacterial two-hybrid (BACTH) system based on the reconstitution of a cyclic AMP signalling cascade (Karimova et al., 1998). This probes for proximity of two fragments (T18 and T25) of the catalytic domain of Bordetella pertussis adenylate cyclase (Cya). Test proteins are fused to T18 and T25, restoring Cya activity if they associate. In an E. coliΔcya mutant this leads to restored activation of cAMP–CRP-dependent genes such as those required for maltose catabolism, which manifests as red colonies on MacConkey-maltose agar.
This analysis supported close PspA–PspF, PspA–PspB and PspA–PspC associations (Fig. 9). As in previous studies, MalF and MalG fusion proteins were used as membrane protein specificity controls (Karimova et al., 2005; Maxson and Darwin, 2006a). A couple of other interesting observations were revealed by this BACTH analysis. First, the PspA–PspB association was detected when T18 was fused to the C-terminus of PspA (Fig. 9) but not when T18 was fused to the N-terminus of PspA (data not shown), suggesting that the latter fusion interferes with the association. Second, in order to detect the PspA–PspC association the plasmid encoding the T25–PspC fusion also had to encode the native pspB gene. We speculate that this is because PspC protein needs a similar concentration of PspB for stability (Gueguen et al., 2009), although we cannot rule out that PspB bridges the PspA–PspC association. For this reason, we are careful to conclude that this BACTH analysis only reveals close association between two proteins rather than direct molecular interaction.
We report for the first time that the Y. enterocolitica PspA protein is almost entirely soluble when the Psp system is uninduced. However, once an inducing signal is encountered PspA associates with the membrane via a mechanism that requires the PspBC proteins. This provides a spatial separation model consistent with PspA having its maximum repressive activity against the cytoplasmic PspF protein in non-inducing conditions. Analysis of PspA and PspC altered regulatory function mutants was consistent with this model, which predicts that increased membrane association of basal PspA correlates with elevated expression of the pspA operon. We also found that the significant increase of PspA concentration that normally occurs after induction of the pspA promoter might also allow PspA to contact the membrane independent of PspBC, which could be important for its proposed physiological function.
This study adds PspA to other examples involving membrane sequestration of regulatory proteins in bacteria. For example, the S. enterica serovar Typhimurium PutA protein, which is a DNA-binding protein that represses the proline utilization operon, shuttles between the membrane and cytoplasm depending on intracellular proline concentration (Muro-Pastor et al., 1997). The Klebsiella pneumoniae NifL protein, which antagonizes activity of the NifA transcriptional activator, moves from the cytoplasm to the membrane in nitrogen-limiting conditions (Klopprogge et al., 2002). Mlc is an E. coli repressor that is inactivated in the absence of glucose by being recruited to the membrane via an interaction with the unphosphorylated IICBGlc glucose transporter (Lee et al., 2000; Tanaka et al., 2000; Nam et al., 2001). PspA has some other striking parallels with these examples. Like PspA, all three are negative regulators of gene expression when in the cytoplasm. NifL, like PspA, antagonizes the activity of a transcriptional regulator. PutA, which represses gene expression when in the cytoplasm, has a physiological function when bound to the membrane (proline to glutamate conversion). PspA is also proposed to play a physiological role at the membrane, although the mechanistic details are not understood (e.g. Kobayashi et al., 2007).
We propose a model with two distinct mechanisms of PspA membrane interaction (Fig. 5). The first does not require any elevation in PspA concentration and is mediated by the low levels of PspABC proteins in an uninduced cell. Here, the basal levels of PspABC constitute a regulatory mechanism controlling PspF activity. PspBC are integral membrane proteins, one or both of which might act as a sensor for a Psp-inducing trigger. Once the signal is encountered there might be a conformational change in PspB and/or PspC that permits an interaction with the basal PspA protein, spatially separating it from PspF. We refer to this as the Psp regulatory switch and its independence from changes in Psp protein concentrations is demonstrated by the fact that we observed it in a pspF null strain (Fig. 2) and when PspABC levels are held constant by pspA operon expression from the tac promoter (Fig. 8).
If it is cytoplasmic PspA that negatively regulates PspF then the question arises of why there is still a significant amount of soluble PspA upon induction (e.g. Fig. 1). Importantly, our data are consistent with reports of a substantial amount of soluble endogenous PspA protein in E. coli under Psp-inducing conditions (Brissette et al., 1990; Kleerebezem and Tommassen, 1993; those authors did not examine PspA location in non-inducing conditions). At least part of the answer is that this is exactly what would be expected. In the presence of YsaC or YE0566, PspA interference with PspF is apparently reduced, as evidenced by increased pspA promoter activity. However, the negative role of PspA is not abolished because a pspA null mutation causes much higher Φ(pspA–lacZ) expression than does the production of a Psp inducer such as YE0566 (Fig. 7). Furthermore, although PspA levels in the soluble fraction appear similar +/− induction (e.g. Fig. 1) it is possible that the oligomeric state of some of the PspA has changed into that which is less able to bind to PspF (Joly et al., 2009). Alternatively, a trivial explanation is that the massive increase in PspA concentration upon induction causes some of it to detach to the soluble fraction during cellular fractionation. In fact, when PspA concentration was held constant as a result of a pspF null mutation, or the pspA operon being under tac promoter control, the amount of soluble PspA did decrease in the presence of an inducing signal (e.g. Figs 2 and 8).
Overproduction of PspBC localized PspA to the membrane in the absence of a Psp-inducing signal and independent of any change in PspA concentration (Fig. 3). This, together with the bacterial two-hybrid analysis, supports the idea that the regulatory switch involves a direct PspA–PspBC interaction. Even so, it is not clear why PspBC overproduction is sufficient for PspA recruitment in the absence of an inducing trigger. One possibility is that PspBC normally exist in one of two states: one in the absence of an inducing signal that has a low affinity for PspA and the other in the presence of an inducing signal that has a high affinity for PspA. According to this, overproducing PspBC without an inducing signal could compensate for their low affinity for PspA allowing them to recruit it to the membrane. Another possibility is that increasing the concentration of PspBC forces them into their active conformation. For example, PspBC overproduction might itself cause membrane stress that actually generates a Psp-inducing signal.
The bacterial two-hybrid analysis supported close associations of PspA with PspF, PspB and PspC (Fig. 9). Evidence for these interactions has also been reported for the E. coli Psp proteins (e.g. Elderkin et al., 2002; Adams et al., 2003). However, it has never been reported whether any dynamic changes in these interactions occur upon induction of the Psp system. The two-hybrid approach is inappropriate for this and we are trying to develop suitable assays for it, but we do not think that PspA–PspF and PspA–PspB/PspC interactions are ‘all or none’ depending on activation status of the Psp system. Rather, the relative amounts of these interactions might change.
Once we established the regulatory switch model for basal level Psp proteins we tested it by using Psp mutant proteins with altered regulatory functions. PspA–N14D, PspA–L200S and PspC–V125D cannot activate the Psp system and according to our model, these mutants should prevent localization of basal PspA to the membrane. As the model predicted, PspA remained predominantly soluble in the presence of an inducing signal in these three cases (Figs 7 and 8). PspC–G45W and PspC–Δcyto, which are both constitutive mutants that constantly induce the Psp system, increased PspA membrane association without an inducing signal (Fig. 8). These results are also consistent with the regulatory switch model.
Most of this work focused on events associated with pspA promoter activation, which is the regulatory switch component of our model (Fig. 5, parts 1 and 2). However, we also found that elevating PspA concentration forced its membrane association in the absence of both PspBC and a Psp-inducing signal. We hypothesize that PspA has an inherent ability to bind to the membrane directly. In an uninduced cell the concentration of PspA might be too low for this to occur. However, once the Psp system is activated the much higher PspA concentration drives its direct interaction with the membrane. In support of this is the observation that in vitro E. coli PspA binds preferentially to liposomes containing phosphatidylserine or phosphatidylglycerol, which are cytoplasmic membrane components (Kobayashi et al., 2007). Therefore, we extended our model to describe events after the regulatory switch has been activated (Fig. 5, part 3). This concentration-driven direct interaction of PspA with the membrane, after an inducing stimulus has been encountered, might be important for a proposed physiological role to improve membrane integrity and PMF (e.g. Kobayashi et al., 2007; Standar et al., 2008).
A recent study examined in vivo location of PspA in E. coli by monitoring a GFP–PspA fusion protein by fluorescence microscopy (Engl et al., 2009). GFP–PspA was observed in the membrane in the absence of a Psp-inducing signal. This contrasts with our data. However, the GFP–PspA protein was encoded on a plasmid with a non-native promoter. Importantly, the GFP–PspA protein was produced at a level similar to that of the endogenous PspA protein upon secretin-induced membrane stress (Engl et al., 2009). In other words, the GFP–PspA concentration was much higher than the normal basal level of PspA. This probably forced its direct association with the membrane in the absence of an inducing signal. Our work with the tacp–pspA operon strain showed that even a small elevation in PspA concentration causes substantial membrane association in the absence of an inducing signal (compare Figs 1 and 8). The authors of the GFP–PspA study proposed that membrane associated PspA could be part of a complex involved in signalling, sensing and release of the PspA–PspF inhibitory complex if a cytoplasmic PspA–PspF complex does not exist (Engl et al., 2009). However, our data studying the endogenous basal PspA protein in Y. enterocolitica suggest that such a cytoplasmic complex is likely to exist.
This work provides a spatial model for events involving PspA both during and after initial activation of the Psp response. Prior to activation PspA is primarily a cytoplasmic protein, most of which is probably in an inhibitory complex with PspF. An inducing signal is probably sensed by PspBC, allowing them to recruit PspA to the membrane, which frees some PspF to activate the pspA operon promoter. Subsequently, the PspA concentration increases to a level where its inherent affinity for the membrane allows a direct contact that might facilitate its physiological function. Challenges for the future include determining whether basal PspA changes its preferred binding partners depending on activation status of the Psp system and whether conformational changes govern the relative affinities of Psp proteins for one another.
Bacterial strains, plasmids and routine growth conditions
Bacterial strains and plasmids are listed 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. 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.
Reference or source
AJD3 is a virulence plasmid cured derivative of strain JB580v (Kinder et al., 1993). All other Y. enterocolitica strains listed are derivatives of AJD3.
The araBp–ysaC (pAJD922) expression plasmid was constructed by transferring the EcoRI–BamHI ysaC insert fragment of pAJD955 into plasmid pWSK129. The KpnI–XbaI ysaC fragment of the resulting plasmid was then transferred into pBAD33. The araBp–YE0566 (pAJD1001) expression plasmid was constructed by transferring the SacI–XbaI YE0566 insert fragment of pAJD634 into plasmid pBAD33. The araBp–pspA (pAJD1089) expression plasmid was constructed by transferring the SacI–SphI pspA insert fragment of pAJD240 (Darwin and Miller, 2001) into plasmid pBAD33.
Yersinia enterocolitica cells were separated into soluble and membrane fractions based on a procedure used to study changes in the subcellular location of the E. coli GlnK protein (Coutts et al., 2002). Saturated cultures were diluted into 200 ml LB broth to an optical density (600 nm) of approximately 0.1. The cultures were grown in a shaker (225 r.p.m.) at 37°C for 2 h. Then 10 µM IPTG (Figs 1–4 and 7) or 0.02% arabinose (Fig. 8) was added to induce YE0566 or YsaC production and growth continued at 225 r.p.m. and 37°C for two more hours prior to harvest. For experiments involving the araBp–pspBC and araBp–pspA plasmids the growth medium contained 0.02% or 0.005% arabinose respectively. Cells were collected by centrifugation at 6500× g for 10 min at 4°C. The cells were resuspended in 10 ml of 50 mM sodium phosphate buffer pH 7.0 containing complete protease inhibitors (Roche). Cells were lysed by 10 sonication pulses of 15 s each with a 15 s interval between each pulse (Sonic Dismembrator Model 500, Fisher Scientific, equipped with a microtip set to 70% amplitude). A 1.5 ml aliquot was transferred to a microcentrifuge tube and unbroken cells and large debris were collected by centrifugation at 16 000× g for 4 min at 4°C. A 0.2 ml aliquot of the supernatant was retained as the total cell lysate. One millilitre of the remaining supernatant was transferred to an ultracentrifuge tube and the membrane pellet was collected by centrifugation at 100 000× g for 60 min at 4°C (the upper 0.2 ml of the supernatant was retained as the soluble fraction). The membrane pellet was washed by resuspending it in 1 ml of 50 mM sodium phosphate buffer pH 7.0 followed by centrifugation at 100 000× g for 60 min at 4°C. The supernatant was discarded and the pellet was resuspended in 1 ml of 50 mM sodium phosphate buffer pH 7.0 and used as the membrane fraction. All samples were mixed with an equal volume of 2× SDS-PAGE sample buffer prior to separation by SDS-PAGE. For each individual strain, samples derived from an equivalent amount of whole cells were loaded for the soluble and membrane fractions so that the relative amount of PspA in each could be assessed. However, fractions derived from different strains were not normalized to one another. Immunoblot detection of the integral cytoplasmic membrane protein FtsH and the soluble cytoplasmic protein DnaK was used to validate the fractionation procedure (these localization controls were done for all experiments but are only shown in Figs 1 and 6). All subcellular fractionation experiments were performed from at least two independent cultures to ensure reproducibility, but each figure shows fractionation data from a single representative experiment.
Biochemical analysis of membrane proteins
Membrane pellets obtained exactly as described above were resuspended in 50 mM phosphate buffer at either pH 11 to release peripheral membrane proteins, or at pH 7 to retain them. The samples were incubated on a slowly rotating mixer for 1 h at 4°C and then centrifuged at 100 000× g for 60 min at 4°C. Both the pellet and the supernatant were analysed by SDS-PAGE and immunoblot. The supernatant contained any released/solubilized proteins and was referred to as the solubilized membrane fraction.
Cytoplasmic membrane and insoluble (aggregate) components were separated by Triton X-100 solubilization based on a previous protocol (Russel and Kazmierczak, 1993). The membrane pellet was resuspended in 1% Triton X-100, 50 mM Tris-HCl (pH 8) containing complete protease inhibitors (Roche). The sample was incubated on a slowly rotating mixer for 20 min at 37°C and then centrifuged at 16 000× g for 5 min. The supernatant contained the solubilized membrane proteins, largely from the inner membrane. The pellet, which contained the Triton X-100 non-extractable material (primarily outer membrane proteins and aggregates), was resuspended in 1 ml of 4% SDS and boiled for 2 min.
Polyclonal antisera and immunoblotting
pspA and pspF were amplified by PCR and cloned into plasmid pQE30 (Qiagen) to encode His6–PspA or His6–PspF proteins respectively. The plasmids were transferred into E. coli strain M15-[pREP4] (Qiagen). The region encoding the cytoplasmic domain of ftsH (YE0428; amino acids 159–564) was amplified by PCR and cloned into plasmid pET-24b(+) (Novagen) to encode a ‘FtsH’–His6 protein. The plasmid was transferred into E. coli strain ER2566 (NEB). The strains were grown to mid-exponential phase at 30°C in LB broth containing 1 mM IPTG. Total cell lysates were prepared, and His6-tagged proteins purified, under denaturing conditions by nickel-nitrilotriacetic acid affinity chromatography as described by the manufacturer. Polyclonal rabbit antisera were raised against the purified His6–PspA, His6–PspF and ‘FtsH’–His6 fusion proteins at Covance Research Products. Anti-DnaK mouse monoclonal antibody (8E2/2) was purchased from Assay Designs (catalogue number SPA-880).
For immunoblot analysis proteins were separated by SDS-PAGE and transferred to nitrocellulose by electroblotting. Enhanced chemiluminescent detection followed sequential incubation with a diluted polyclonal antiserum or monoclonal antibody, and then goat α-rabbit IgG or goat α-mouse IgG horseradish peroxidase conjugate (Bio-Rad) used at 1 in 5000. Dilutions of polyclonal antisera were 1 in 1000 for anti-PspF and 1 in 10 000 for anti-PspA and anti-FtsH. The anti-DnaK monoclonal antibody was used at 1 in 2000 dilution.
For Fig. 8 ImageJ 1.43 software (http://rsb.info.nih.gov/ij) was used to determine the ratio of the integrated densities of the PspA bands in the membrane and soluble fractions. Furthermore, separate immunoblot analysis of the fractionation samples with infrared secondary antibodies and a LI-COR Odyssey Infrared Imaging system also confirmed the conclusions from the ImageJ analysis (data not shown).
For Figs 1–3 samples for β-galactosidase assays were taken directly from subcellular fractionation cultures prior to harvest (two independent cultures for each strain). To characterize the phenotypes of the PspA and PspC altered function mutant strains, saturated cultures were diluted into 5 ml of LB broth in 18 mm-diameter test tubes to an optical density (600 nm) of 0.1. The cultures were grown on a roller drum at 37°C for 2 h. Then 10 µM IPTG or 0.02% arabinose (final concentrations) was added to induce YE0566 or YsaC production and growth continued at 37°C for an additional 2 h prior to harvest. For bacterial two-hybrid analysis strains were grown in LB broth containing 1 mM IPTG at 30°C for 15 h. 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 or triplicate, and values were averaged from two to four independent cultures.
Isolation of PspA altered function mutants
The pspA gene 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 programme was 25 × [95°C 30 s, 45°C 30 s, 72°C 2 min]. Six independent PCR reactions were done. The products were cloned into a derivative of pWSK29 encoding kanamycin resistance, so that pspA was expressed from its native promoter, and used to transform E. coli DH5α. The colonies from each individual transformation were combined and plasmid DNA was isolated, resulting in six independent mutant libraries. Aliquots of each library were used to transform Y. enterocolitica strain AJD1200 containing a tacp–yscC secretin expression plasmid. Transformants were recovered on MacConkey agar at 26°C. Colonies that were white or pale red (the wild-type pspA phenotype was dark red colonies) were characterized further as potential constitutive repressor mutants. Only two of these mutants were used in this study (N14D and L200S; the others will be described elsewhere). The pspA–N14D and pspA–L200S mutations were cloned into the sacB+ allelic exchange plasmid pRE112 and introduced into the native chromosomal pspA operon by plasmid integration, selection for sucrose-resistant segregants and confirmation by colony PCR.
tacp–pspA operon strain constructions
Two ∼550 bp fragments flanking the pspA promoter were amplified from the chromosome of a Y. enterocoliticaΔpspF strain by PCR with primers that incorporated a SpeI site at one end of each fragment. The fragments were ligated at this SpeI site to form a fragment encoding a deletion of the transcription initiation sites upstream of pspA and cloned into sacB+ allelic exchange plasmid pSR47S. Then an ∼1.5 kb fragment encoding lacIq–tacp from plasmid pMAL-p2 (New England Biolabs) was ligated into this SpeI site so that the tac promoter was in the same orientation as the original pspA promoter. The lacIq–tacp cassette was then introduced upstream of the native chromosomal pspA operon of a ΔpspF strain by plasmid integration, selection for sucrose-resistant segregants and confirmation by colony PCR. The result was a strain with the pspF+–pspAp region replaced by lacIq–tacp, which placed the chromosomal pspA operon under the control of the tac promoter.
Next the pspF gene was introduced into the site of the yenR restriction enzyme deletion present in all of our strains, which we have used previously to insert a kanamycin resistance marker (Green and Darwin, 2004). The region flanking the yenR deletion was amplified from the Y. enterocolitica chromosome by PCR with primers that incorporated a XbaI site at the site of the yenR deletion. This fragment was cloned into sacB+ allelic exchange plasmid pRE112. Then a fragment encoding the pspF gene with its entire non-coding upstream region was cloned into the XbaI site. The primers used to amplify pspF incorporated E. coli rrnBT1 and rrnBT2 terminators upstream and downstream respectively. The pspF gene was then introduced into the chromosomal yenR deletion site by plasmid integration, selection for sucrose-resistant segregants and confirmation by colony PCR.
pspC–G45W, pspC–V125D or cytoplasmic domain deletion mutations (Gueguen et al., 2009) were cloned into the sacB+ allelic exchange plasmid pRE112 and introduced into the chromosomal pspA operon downstream of the tac promoter by plasmid integration, selection for sucrose-resistant segregants and confirmation by colony PCR.
Bacterial two-hybrid analysis
For the PspA–T18 fusion the pspA gene was amplified by PCR and cloned into the unique BamHI site of pUT18 as a BamHI–BglII fragment, to make pAJD1088. The plasmid encoding T25–PspC also encoded the native pspB gene (pspB was encoded downstream of the cyaT25–pspC gene and not fused to any other coding sequence). For this, the pspB gene was amplified by PCR and cloned into the unique BamHI site of pAJD489 (Maxson and Darwin, 2006a) as a BamHI fragment, to make pAJD1493. To make a vector only control for this plasmid (encoding only T25 and pspB) the same pspB fragment was cloned into pKT25 to make pAJD1492. Construction of the plasmids encoding PspB–T25 and the various MalF and MalG fusions was described previously (Maxson and Darwin, 2006a) and they are listed in Table 1.
Pairs of plasmids were introduced simultaneously into E. coli BTH101 by calcium chloride transformation. Transformants were streaked onto MacConkey-maltose agar and incubated at 26°C for approximately 40 h before being photographed. β-Galactosidase enzyme activity of selected transformants was determined as described above.
We thank Heran Darwin for critically reviewing a draft version of the manuscript. This study was supported by a Vilcek Endowment Fellowship Award to S.Y. and by Award Number R01AI052148 from the National Institute of Allergy and Infectious Diseases. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Allergy and Infectious Diseases or the National Institutes of Health.