Interactions between the Cytoplasmic Domains of PspB and PspC Silence the Yersinia enterocolitica Phage Shock Protein Response

ABSTRACT

The phage shock protein (Psp) system is a widely conserved cell envelope stress response that is essential for the virulence of some bacteria, including Yersinia enterocolitica. Recruitment of PspA by the inner membrane PspB-PspC complex characterizes the activated state of this response. The PspB-PspC complex has been proposed to be a stress-responsive switch, changing from an OFF to an ON state in response to an inducing stimulus. In the OFF state, PspA cannot access its binding site in the C-terminal cytoplasmic domain of PspC (PspC^CT), because this site is bound to PspB. PspC has another cytoplasmic domain at its N-terminal end (PspC^NT), which has been thought to play a role in maintaining the OFF state, because its removal causes constitutive activation. However, until now, this role has proved recalcitrant to experimental investigation. Here, we developed a combination of approaches to investigate the role of PspC^NT in Y. enterocolitica. Pulldown assays provided evidence that PspC^NT mediates the interaction of PspC with the C-terminal cytoplasmic domain of PspB (PspB^CT) in vitro. Furthermore, site-specific oxidative cross-linking suggested that a PspC^NT-PspB^CT interaction occurs only under noninducing conditions in vivo. Additional experiments indicated that mutations in pspC might cause constitutive activation by compromising this PspC^NT binding site or by causing a conformational disturbance that repositions PspC^NT in vivo. These findings have provided the first insight into the regulatory function of the N-terminal cytoplasmic domain of PspC, revealing that its ability to participate in an inhibitory complex is essential to silencing the Psp response.

IMPORTANCE The phage shock protein (Psp) response has generated widespread interest because it is linked to important phenotypes, including antibiotic resistance, biofilm formation, and virulence in a diverse group of bacteria. Therefore, achieving a comprehensive understanding of how this response is controlled at the molecular level has obvious significance. An integral inner membrane protein complex is believed to be a critical regulatory component that acts as a stress-responsive switch, but some essential characteristics of the switch states are poorly understood. This study provides an important advance by uncovering a new protein interaction domain within this membrane protein complex that is essential to silencing the Psp response in the absence of an inducing stimulus.

INTRODUCTION

Extracytoplasmic stress responses (ESRs) are a widely used strategy to protect bacterial cell envelopes from conditions and external agents that could compromise them. They monitor features within the cell envelope and direct a transcriptional response to potentially deleterious changes (1–4). One widely conserved ESR is the phage shock protein (Psp) response, which protects the cytoplasmic membrane from events that could increase its permeability (recently reviewed in references 5 and 6). Although it was discovered and initially studied in Escherichia coli (7), the Psp response is conserved in many Gram-negative bacteria, including the human pathogens Yersinia enterocolitica and Salmonella enterica serovar Typhimurium, where it is essential for virulence (8–11). Sequence and/or functional homologues of some Psp components also occur in Gram-positive bacteria, mycobacteria, archaea, and plant chloroplasts (e.g., references 1, 12, and 13). Furthermore, in addition to virulence, the Psp response has been linked to other important processes in bacteria, including the formation of biofilms and antibiotic-tolerant persister cells (14, 15).

Most research to date has focused on the Psp responses of E. coli and Y. enterocolitica, in which the core components are encoded by pspF and the adjacent divergently transcribed pspABC genes. There are a small number of additional genes in the Psp regulons of these two species, but pspF, pspA, pspB, and pspC are considered to encode the core components because only they are essential for regulation and/or stress tolerance (5, 6, 16, 17). PspF is a DNA-binding protein that activates pspA operon expression in response to conditions that might adversely affect the inner membrane (18, 19). These conditions include extreme heat, exposure to ethanol, high osmolarity, and the mislocalization of outer membrane secretin proteins into the inner membrane of Gram-negative bacteria (reviewed in reference 20). Secretins are multimeric proteins that form the outer membrane pores of various export systems, including type II and III secretion systems (21). However, in some circumstances, secretins mislocalize into the inner membrane, which specifically induces the PspF regulon but is lethal to psp-null strains (16, 22–24). Indeed, it was induction of the E. coli Psp response by the mislocalization of a filamentous phage-encoded secretin that led to the phage shock name (7).

One role of PspA, PspB, and PspC is to form a signal transduction system that controls PspF activity. Under noninducing conditions, PspA forms an inhibitory complex with PspF in the cytoplasm (25–27). However, under inducing conditions, PspA instead associates with the integral inner membrane complex PspB-PspC, and active PspF is released (26–28). This change in binding partners of PspA from PspF to PspB-PspC is controlled by the conformational state of the PspB-PspC complex, indicating that PspB-PspC functions as a stress-responsive switch. Specifically, an inducing stimulus causes a conformational change in the PspB-PspC complex that unmasks a PspA-binding site in the C-terminal domain of PspC (28). However, under noninducing conditions, this PspA-binding site is occluded, because it is instead bound to the cytoplasmic domain of PspB (28).

PspC has two transmembrane domains, with both of its termini in the cytoplasm (29). Our previous work uncovered the role played by the PspC C-terminal domain (PspC^CT) in binding to PspB or to PspA, depending on the inducing status (28). However, that study did not provide any information about the role played by the N-terminal cytoplasmic domain of PspC (PspC^NT), mostly because the isolated PspC^NT domain proved to be unsuitable for in vitro analysis. Nevertheless, PspC^NT appears to play a crucial role, because its removal causes constitutive recruitment of PspA to the PspB-PspC complex in vivo and activation of the Psp response (26, 27, 30). This suggests that PspC^NT helps to maintain the OFF state of the PspB-PspC complex under noninducing conditions. In this study, we have used a combination of in vivo and in vitro approaches to investigate this role further. Our findings support a model in which protein-protein interaction and the adoption of a precise position by PspC^NT are likely to be critical to maintain the PspB-PspC regulatory switch in its OFF position.

MATERIALS AND METHODS

Bacterial strains, plasmids, and routine growth.

The bacterial strains and plasmids used are listed in Table 1. The expected DNA sequence of all PCR-generated fragments was confirmed. The standard growth medium was Luria-Bertani (LB) broth or LB agar, and antibiotics were used as previously described (31).

TABLE 1.

Strains and plasmids

Strain or plasmid	Genotype and/or featuresa	Reference or source
E. coli B strain AJDE1173	F− ompT hsdS (rB− mB−) gal dcm Δ(pspF-pspE)	28
Y. enterocolitica strains
AJD3b	ΔyenR (R− M+)	35
AJD1171	ΔyenR (R− M+) Δ(pspF-ycjF) ΔpspG	17
AJD3298	ΔyenR (R− M+) ΔpspF ΔpspAp::lacIq-tacp	26
AJD3469	ΔyenR (R− M+) ΔpspF ΔpspAp::lacIq-tacp ΔpspBC	16
AJD3490	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp	26
AJD3700	ΔyenR (R− M+) ΔpspF ΔpspAp::lacIq-tacp ΔpspAB	16
AJD4144	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-G45W	26
AJD4145	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-ΔNT	26
AJD4146	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-V125D	26
AJD4681	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp ΔpspC	This study
AJD4776	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-L69P	This study
AJD4866	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-S26C-C43S	This study
AJD4872	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-C43S	This study
AJD5791	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S32C pspC-S26C-C43S	This study
AJD5800	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-R58G	This study
AJD5805	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S32C pspC-C43S	This study
AJD5806	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-W23R-S32C pspC-S26C-C43S	This study
AJD5807	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S32C pspC-S26C-C43S-L69P	This study
AJD5809	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S32C pspC-S26C-C43S-R58G	This study
AJD5810	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspC-C43S-V125C	This study
AJD5811	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S75C pspC-C43S	This study
AJD5812	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S75C pspC-C43S-V125C	This study
AJD5813	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S75C pspC-C43S-L69P-V125C	This study
AJD5814	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-W23R-S75C pspC-C43S-L69P-V125C	This study
AJD5816	ΔyenR (R− M+)::[pspF+] ΔaraGFB::Φ(pspA-lacZY) ΔpspF ΔpspAp::lacIq-tacp pspB-S75C pspC-C43S-R58G-V125C	This study
Plasmids
pBAD33	Cmr, p15A ori araBp expression vector	36
pGEX-6P-1	Apr, pBR322 encoding GST	GE Healthcare
pRE112	Cmr R6K ori mob+ (RP4) sacB+	37
pREP4-GroESL	Kmr, p15A ori groES groEL overexpression plasmid	38
pSR47S	Kmr R6K ori mob+ (RP4) sacB+	39
pAJD1089	pBAD33 derivative encoding wild-type PspA	26
pAJD2344	pGEX-6P-1 derivative encoding GST-PspBCT	28
pAJD2465	pBAD33 derivative encoding PspC-ΔNT	This study
pAJD2466	pBAD33 derivative encoding PspC-G45W	This study
pAJD2686	pBAD33 derivative encoding wild-type PspC	This study
pAJD2688	pBAD33 derivative encoding PspC-L69P	This study
pAJD2689	pBAD33 derivative encoding PspC-R58G	This study

^aCm^r, chloramphenicol resistance; Ap^r, ampicillin resistance; Km^r, kanamycin resistance.

^bAJD3 is a virulence plasmid-cured derivative of strain JB580v (35). All other Y. enterocolitica strains listed are derivatives of AJD3.

Plasmid constructions.

araBp expression plasmids encoding PspC or its mutant derivatives were made by amplifying the pspC gene from plasmids published previously (30) with primers that incorporated SacI (upstream) and XbaI (downstream) sites. The fragments were digested with SacI-XbaI and ligated into pBAD33.

Construction of Y. enterocolitica strains.

AJD4681 was made by introducing a ΔpspC in-frame deletion mutation into strain AJD3490 by sacB⁺-facilitated allelic exchange using a plasmid pSR47s derivative from our collection with a ΔpspC insert. The deletion was confirmed by colony PCR analysis.

To make AJD4776 and AJD5800, the ΔpspC mutation in strain AJD4681 was exchanged for an intact pspC gene encoding the previously described PspC-L69P or PspC-R58G mutants (30) via sacB⁺-facilitated allelic exchange using plasmid pRE112 derivatives with pspC-L69P or pspC-R58G inserts. The mutations were confirmed by colony PCR analysis.

To make Y. enterocolitica strains encoding PspB and/or PspC cysteine substitution mutants, the pspB⁺-ΔpspC locus of strain AJD4681 was exchanged for intact pspBC genes encoding the previously described cysteine substitution mutants (32) via sacB⁺-facilitated allelic exchange using plasmid pRE112 derivatives with pspBC⁺ inserts that had the desired mutation(s). The mutations were confirmed by colony PCR analysis.

GST-PspB^CT fusion protein membrane lysate pulldown assays.

These assays were done exactly as described previously (28). Briefly, glutathione transferase (GST) (encoded by pGEX-6P-1) or GST-PspB^CT (encoded by pAJD2344) was purified on glutathione-Sepharose beads (GE Healthcare) from E. coli strain AJDE1773 that also contained plasmid pREP4-groESL to increase protein solubility. The beads were then incubated with Y. enterocolitica n-dodecyl β-d-maltoside (DDM)-solubilized membrane lysates and washed before elution of proteins by boiling in SDS-PAGE sample buffer.

Two-phase GST-PspB^CT fusion protein membrane lysate pulldown assays.

These assays were done as described previously (28). Briefly, GST or GST-PspB^CT was purified onto glutathione-Sepharose beads from E. coli strain AJDE1773 that also contained plasmid pREP4-groESL. The beads were then incubated with 1 ml of 137 mM NaCl, 2.7 mM KCl, 10.1 mM Na₂HPO₄, and 1.4 mM KH₂PO₄ (phosphate-buffered saline [PBS]) containing 5% (wt/vol) bovine serum albumin, recovered by centrifugation, and washed with PBS. DDM-solubilized membrane prey lysates were prepared from Y. enterocolitica strain AJD1171 containing an arabinose-inducible expression plasmid encoding PspA, PspC, or a PspC mutant protein, which had been grown to mid-exponential phase at 26°C in 1 liter of LB broth containing 0.2% (wt/vol) arabinose.

A DDM-solubilized membrane lysate from a strain with an expression plasmid encoding PspC, or a PspC mutant protein, was incubated with the GST or GST-PspB^CT immobilized on glutathione-Sepharose beads. The beads were then washed and separated into two equal samples. One of the samples was incubated with a second DDM-solubilized membrane lysate from the strain with an expression plasmid encoding PspA, and then both samples were washed, resuspended in SDS-PAGE sample buffer, and boiled to recover all proteins. Equal volumes of the elution samples were analyzed by SDS-PAGE and immunoblotting.

Analysis of PspBC complex formation in vivo by disulfide cross-linking.

Approximately 8 × 10⁸ cells in exponential-growth phase (optical density at 600 nm, ∼0.6) were harvested by centrifugation, washed with 10 ml of 20 mM sodium phosphate buffer (pH 6.8) (NaP), and resuspended in 1.6 ml of NaP containing 0.02% (wt/vol) arabinose to maintain induction of araBp-ysaC expression. The cells were divided into three samples of 0.5 ml each (samples A to C). Sample A was incubated with 2.5 mM N-ethyl-maleimide (NEM; Sigma-Aldrich) for 10 min on ice. Samples B and C were incubated first with 0.3 mM dichloro(1,10-phenanthroline) copper(II) (Cu-oP; Sigma-Aldrich) for 20 min and then with 2.5 mM NEM for 10 min on ice. After centrifugation, cell pellets (samples A and B) were resuspended in SDS-PAGE sample buffer without reducing agent. Sample C was resuspended in SDS-PAGE sample buffer containing 10 mM dithiothreitol (DTT). Proteins were denatured by boiling and then analyzed by SDS-PAGE and anti-PspB or anti-PspC immunoblotting.

β-Galactosidase assays.

Saturated cultures were diluted into 5 ml of LB with appropriate antibiotics in 18-mm-diameter test tubes to an optical density at 600 nm of approximately 0.04. The cultures were grown on a roller drum at 37°C for 2 h, and then 0.02% (wt/vol) arabinose (final concentration) was added to induce ysaC expression. Incubation then continued for a further 2 h at 37°C. β-Galactosidase enzyme activity was determined at room temperature in cells permeabilized with SDS-chloroform, as described previously (33). Activity is expressed in arbitrary Miller units. Individual cultures were assayed in duplicate, and average values from three independent cultures are reported. However, for in vivo disulfide cross-linking experiments, samples for the β-galactosidase enzyme assays were taken directly from the cultures prior to harvest and assayed in duplicate.

Polyclonal antisera and immunoblotting.

Proteins were separated by SDS-PAGE and then transferred to a nitrocellulose membrane by semidry electroblotting. For analysis of total-cell lysates, approximately equal loading and transfer were confirmed by total-protein staining of the nitrocellulose membrane with Ponceau S (Amresco). Enhanced chemiluminescent detection followed sequential incubation with a polyclonal antiserum, and then goat anti-rabbit IgG horseradish peroxidase conjugate (Sigma-Aldrich) was added at the manufacturer's recommended dilution. Polyclonal antisera were diluted 1:10,000 for anti-PspC and anti-PspA and 1:20,000 for anti-PspB (17, 26, 32).

RESULTS

Pulldowns suggest that the N-terminal domain of PspC plays a role in protein-protein interactions.

An inducing stimulus causes the C-terminal domain of PspC (PspC^CT) to switch its binding partner from PspB to PspA (28) (Fig. 1). However, removal of the N-terminal domain of PspC (PspC^NT) causes constitutive PspBC-PspA interaction and induction of the Psp response in vivo (26, 27, 30) (Fig. 1). This suggests that PspC^NT plays an important role to silence the Psp response under noninducing conditions. We hypothesized that this involves the participation of PspC^NT in protein-protein interaction(s). Previously, we used a GST-PspC^CT fusion protein to capture interacting partners from Y. enterocolitica membrane lysates (28, 29). However, this approach was unsuccessful with various PspC^NT fusion proteins, most of which could not be produced stably (28). Therefore, as an alternative approach to further investigate interactions between the cytoplasmic domains of PspB and PspC with each other, and with PspA, we explored the use of the C-terminal cytoplasmic domain of PspB as a bait protein by using a GST-PspB^CT fusion protein.

FIG 1.

Explanatory diagrams. (A) Model and rationale for this study. In the absence of an inducing signal, PspBC adopt a conformation in a wild-type strain with their C-terminal cytoplasmic domains interacting, and PspA cannot access its binding site in the C-terminal domain of PspC (white rectangle). An inducing signal, such as the mislocalization of a secretin protein into the inner membrane, causes a conformational shift to make the PspA-binding site available. PspA is then sequestered away from PspF, freeing PspF to induce the pspA promoter. In a strain where the N-terminal cytoplasmic domain of PspC has been removed (PspC-ΔNT), PspA interacts with the C-terminal domain of PspC regardless of the presence or absence of an inducing signal, rendering the Psp response constitutively active. This suggests that the N-terminal domain of PspC plays a key role in maintaining the uninduced conformation of PspBC. (B) Approximate locations of amino acid substitution mutations used in this study. In both panels, the N and C termini of PspB and PspC are labeled N and C, respectively, and TM indicates transmembrane domains.

GST-PspB^CT was immobilized onto glutathione-Sepharose and incubated with lysates from Y. enterocolitica strains encoding either all core Psp proteins (Psp⁺), or only PspC or PspA. PspC was captured from the Psp⁺ and PspC-only lysates, consistent with the direct PspB^CT-PspC^CT interaction we reported previously (Fig. 2) (28). In contrast, GST-PspB^CT captured PspA from a Psp⁺ lysate but not from a PspA-only lysate (Fig. 2). This is consistent with PspB^CT being unable to bind to PspA directly (Fig. 1), as well as work with GST-PspB^CT from E. coli suggesting that a complex containing PspB and PspA depended on another Psp protein, which was predicted to be PspC (34).

FIG 2.

GST-PspB^CT fusion protein pulldown assay. GST or GST fused to the C-terminal domain of PspB (GST-PspB^CT) bound to glutathione-Sepharose (bait/beads) was incubated with a detergent-solubilized membrane lysate from a Y. enterocolitica strain with all core Psp proteins (Psp⁺) or in which the only core Psp proteins present were PspC or PspA as indicated (prey/lysate). After washing, proteins were recovered by boiling in SDS-PAGE sample buffer. Membrane lysates (inputs) and recovered proteins (elutions) were analyzed by SDS-PAGE and anti-PspA or anti-PspC immunoblotting. The GST fusion protein in each elution was detected by Ponceau S staining of the immunoblot membrane. The diagrams at the bottom show the proposed GST-PspB^CT-containing complexes present in each elution, with the N- and C-terminal domains of PspC shown as gray boxes.

If GST-PspB^CT cannot bind to PspA directly, its ability to capture PspC and PspA from a Psp⁺ lysate suggests that it might bind to PspC in a GST-PspB^CT–PspC-PspA complex (Fig. 2). The only known binding site for PspB^CT is within PspC^CT, but PspA also binds to that site, and these interactions cannot occur simultaneously (28) (Fig. 1). Therefore, if PspA occupies PspC^CT, a GST-PspB^CT–PspC-PspA complex could only form if PspB^CT has another binding site in PspC, most likely in its other cytoplasmic domain, PspC^NT (Fig. 2).

Both ends of PspC are important for interactions with the cytoplasmic domain of PspB in vitro.

Our preceding hypothesis is that PspB^CT contacts PspC^NT in a GST-PspB^CT–PspC-PspA complex (Fig. 2). If this is correct, GST-PspB^CT should not be able to capture PspC-PspA from a strain in which the N-terminal domain of PspC is deleted (PspC-ΔNT). Indeed, a PspC-ΔNT mutation prevented both PspA and PspC from being captured by GST-PspB^CT (Fig. 3). Furthermore, the failure of GST-PspB^CT to capture PspC-ΔNT, despite the PspC^CT binding site being intact, can be explained by the ΔNT mutation causing constitutive activation of the Psp response in vivo (Fig. 3A) (30). This would cause PspC^CT to be fully occupied by PspA in this strain/lysate and so unavailable to bind to GST-PspB^CT during the in vitro pulldown procedure (Fig. 1 and 3).

FIG 3.

Effect of PspC mutations on the capture of PspA and PspC by GST-PspB^CT. (A) Φ(pspA-lacZ) operon fusion expression analysis. Strains with the chromosomal pspA operon expressed from the tac promoter encoded wild-type PspABC (PspABC) or PspC mutant derivatives. Strains also contained an empty araBp expression plasmid pBAD33 (−) or the ysaC⁺ derivative pAJD935 (YsaC). Cultures were grown in medium containing 0.02% (wt/vol) arabinose, and β-galactosidase activities were determined. Error bars indicate the positive standard deviations from the means. (B) GST-PspB^CT fusion protein pulldown assay. GST or GST fused to the C-terminal domain of PspB (GST-PspB^CT) was bound to glutathione-Sepharose (bait/beads) and incubated with a detergent-solubilized membrane lysate from a Y. enterocolitica strain with all core Psp proteins (PspABC) or PspC mutant derivatives (prey/lysate). After washing, proteins were recovered by boiling in SDS-PAGE sample buffer. Membrane lysates (inputs) and recovered proteins (elutions) were analyzed by SDS-PAGE and immunoblotting with PspC or PspA antiserum. The GST fusion protein in each elution was detected by Ponceau S staining of the immunoblot membrane. The diagrams at the bottom show the proposed GST-PspB^CT-containing complexes present in each elution, as well as the proposed proteins/complexes that were not be captured. The N- and C-terminal domains of PspC shown as gray boxes, and asterisks show the approximate location of each amino acid substitution mutation.

Previously, we isolated PspC amino acid substitution mutations that cause constitutive activation of the Psp response in vivo (30). They all mapped to PspC^NT or to the PspC transmembrane (TM) domains (30). Next, we tested the effect of representative PspC constitutive mutations in our GST-PspB^CT pulldown assay. Like PspC-ΔNT, a G45W mutation within PspC^NT prevented PspC-PspA from being captured by GST-PspB^CT (Fig. 3B). This is consistent with the G45W mutation disrupting the PspC^NT binding site for PspB^CT. In contrast, substitutions in the first (R58G) or second (L69P) PspC TM domains did not prevent the capture of PspC-PspA by GST-PspB^CT (Fig. 3B). This can be explained by these TM domain mutations leaving the fidelity of the PspC^NT binding site intact and available to interact with GST-PspB^CT in vitro (Fig. 3B). However, in vivo, these PspC TM domain mutants might be constitutively active, because they force a conformational shift to the ON state (investigated in later experiments).

Our laboratory has also isolated mutations in PspC^CT (e.g., PspC-V125D), which stop activation of the Psp response by compromising the PspA-binding site and preventing the recruitment of PspA from PspF (30). The V125D mutation also prevents PspB^CT from binding to PspC^CT (28–30). We tested the effect of the PspC-V125D mutation in our GST-PspB^CT pulldown assay and found that it had no effect on the capture of PspC but almost abolished the capture of PspA (Fig. 3B). This further supports the hypothesis that PspB^CT can interact with PspC^NT, which is intact in PspC-V125D. However, PspA can only interact with PspC^CT, which is disrupted by the V125D mutation (Fig. 3B).

Displacement experiments support the hypothesis that the cytoplasmic domain of PspB can interact with either end of PspC.

Previously, we developed a two-phase pulldown assay (28). First, glutathione-Sepharose–GST-PspC^CT was incubated with a PspA-only lysate to saturate GST-PspC^CT with PspA. This complex was then incubated with a second lysate containing PspB, which displaced PspA by competing for the PspC^CT binding site (PspA displaced PspB if the lysates were added in the reverse order). A similar approach can provide a rigorous test of the hypothesis that GST-PspB^CT can bind to either end of PspC, because the version of PspC used will predict the ability of PspA to join a GST-PspB^CT–PspC complex or to displace PspC from it (Fig. 4, diagrams). For instance, in a GST-PspB^CT–PspC-G45W complex, binding has to be via PspC^CT, and so PspA should completely displace PspC by competing with GST-PspB^CT for the PspC^CT binding site. Conversely, in a GST-PspB^CT–PspC-V125D complex, binding has to be via PspC^NT, and so PspA should not displace PspC or join the complex, because its only binding site (in PspC^CT) is destroyed. Finally, in a GST-PspB^CT–PspC complex with both PspC termini intact (wild type or TM domain mutants), binding could be via PspC^NT or PspC^CT. Therefore, although PspA might displace any PspC bound via PspC^CT, it will be captured by any GST-PspB^CT–PspC complexes joined together via PspC^NT. In the next set of experiments, we tested all of these predictions.

FIG 4.

Two-phase GST-PspB^CT fusion protein pulldown assay. GST or GST fused to the C-terminal domain of PspB (GST-PspB^CT) was bound to glutathione-Sepharose (bait/beads) and incubated with a detergent-solubilized membrane lysate from a Y. enterocolitica strain in which the only Psp protein present was PspC or a mutant derivative (prey 1). After washing, proteins were recovered from half of the beads by boiling in SDS-PAGE sample buffer (elution 1). The other half of the beads was incubated with a second detergent-solubilized membrane lysate from a Y. enterocolitica strain in which the only Psp protein present was PspA (prey 2). After washing, proteins were recovered by boiling in SDS-PAGE sample buffer (elution 2). Membrane lysates (inputs) and recovered proteins (elutions 1 and 2) were analyzed by immunoblotting with PspC or PspA antiserum. The GST fusion protein in each elution was detected by Ponceau S staining of the immunoblot membrane. The diagrams at the bottom show the proposed GST-PspB^CT-containing complexes present in elutions 1 and 2. Also shown are the proposed proteins/complexes that were competed off by PspA or not captured during incubation with prey lysate 2. The N- and C-terminal domains of PspC shown as gray boxes, and asterisks show the approximate location of each amino acid substitution mutation.

To eliminate any possible complications from other Psp proteins, the prey lysates for these experiments were derived from a Y. enterocolitica strain with all psp genes deleted (AJD1171) that contained either an araBp-pspC (prey 1) or araBp-pspA (prey 2) expression plasmid. First, a lysate with wild-type PspC or one of its mutant variants was incubated with glutathione-Sepharose–GST-PspB^CT to generate the GST-PspB^CT–PspC complexes. After extensive washing, half of the Sepharose beads were resuspended in SDS-PAGE sample buffer and boiled to elute all proteins (elution 1; this sample served to confirm the binding of PspC). The other half was incubated with a lysate containing PspA, followed by washing and elution (elution 2). Equal amounts of elutions 1 and 2 were analyzed by SDS-PAGE and immunoblotting.

In the experiments with wild-type PspC, or the R58G and L69P transmembrane domain mutants, both PspC and PspA were retained by GST-PspB^CT after the second phase (Fig. 4). This is consistent with those GST-PspB^CT–PspC complexes joined together via PspC^NT in phase 1 capturing PspA via a PspC^CT–PspA interaction in phase 2. In contrast, when PspC-G45W was used, neither PspC nor PspA were retained by GST-PspB^CT after the second phase (Fig. 4). This is consistent with PspA dissociating the GST-PspB^CT–PspC-G45W complex joined via PspC^CT by successfully competing for the PspC^CT binding site. Finally, when PspC-V125D was used, only PspC was retained by GST-PspB^CT after the second phase (Fig. 4). This is consistent with PspA being unable to displace PspC-V125D from the GST-PspB^CT–PspC-V125D complex because it is connected via PspC^NT, to which PspA does not bind. It is also consistent with PspA being unable to join the GST-PspB^CT–PspC-V125D complex because its PspC^CT binding site is destroyed by the V125D mutation. The findings from all of these displacement experiments matched the predictions and are fully consistent with the hypothesis that the C-terminal cytoplasmic domain of PspB can interact with either end of PspC.

An inhibitory complex between the cytoplasmic domain of PspB and both ends of PspC dissociates upon induction of the Psp response in vivo.

We hypothesize that the proposed PspB^CT-PspC^NT interaction is critical for negative regulation because mutations that prevent it in vitro (G45W or ΔNT mutation; see above) cause constitutive activation of the Psp response in vivo (30). If this is correct, there should be a contact between PspB^CT and PspC^NT when the PspBC complex is in its OFF state but not in its ON state. To test this, we needed an in vivo approach, such as the site-specific photocross-linking method we used previously to monitor the interactions of PspC^CT with its partners in vivo (28). However, we could not find any position in PspC^NT that could be substituted by the photoactivable amino acid analog p-benzoyl-l-phenylalanine without affecting function or stability (data not shown). Therefore, we decided to try in vivo disulfide cross-linking. We have used this to study PspB and PspC homo- and heterointeractions, but we did not investigate if it could be used to monitor changes upon Psp response induction (32). In this approach, PspB and PspC cysteine substitution mutants are treated with an oxidation catalyst (Cu-oP), and a disulfide bond forms if the cysteines are close enough to interact. PspB is naturally cysteine-less, but the cysteine substitution mutants of PspC were made in PspC-C43S, a derivative in which the only native cysteine was replaced by serine without affecting its regulatory function (32). We also used strains in which the chromosomal pspA operon is expressed at a physiological level from the tac promoter, so that PspABC levels are constant regardless of the Psp induction status, but they still respond to secretin stress to induce Φ(pspA-lacZ) expression (16, 26). This prevents the large increase in the amount of PspB and PspC that normally occurs upon induction of the Psp response, which would complicate the interpretation of changes in the levels of cross-linked complexes.

We have reported that a disulfide cross-link can form between PspB-S32C and PspC-S26C, supporting the close proximity of PspB^CT and PspC^NT (32). Here, we screened several other cysteine substitution mutants in order to find a pair that could be used to monitor the PspB^CT-PspC^CT interaction, with PspB-S75C and PspC-V125C being found to work (data not shown and Fig. 5). Then, to test if the PspB^CT-PspC^NT and PspB^CT-PspC^CT interactions depended on the induction status of the Psp response in vivo, an arabinose-inducible plasmid encoding the YsaC secretin or the empty vector control was introduced. Functional analysis of the mutant strains showed that secretin-dependent induction of Φ(pspA-lacZ) expression was maintained (Fig. 5). Cells were treated with Cu-oP, followed by the cysteine-reactive reagent N-ethyl-maleimide (NEM) to terminate the Cu-oP-catalyzed reaction. As controls, cells were treated with NEM only, or a set of Cu-oP-treated samples was reduced with dithiothreitol (DTT) to destroy disulfide bonds. Samples were analyzed by anti-PspC and anti-PspB immunoblotting (Fig. 5).

FIG 5.

Evidence that a complex between the cytoplasmic domain of PspB and both ends of PspC dissociates upon induction of the Psp response in vivo. (A) In vivo analysis of the PspB^CT-PspC^CT interaction. (B) In vivo analysis of the PspB^CT-PspC^NT interaction. (Ai and Bi) Φ(pspA-lacZ) operon fusion expression analysis. Strains with the chromosomal pspA operon expressed from the tac promoter encoded PspB, PspC-C43S (PspC), or the indicated cysteine substitution mutant derivatives. Strains also contained an empty araBp expression plasmid pBAD33 (−) or the ysaC⁺ derivative pAJD935 (YsaC). Cultures were grown in medium containing 0.02% (wt/vol) arabinose, and β-galactosidase activities were determined. Error bars indicate the positive standard deviations from the means. (Aii and Bii) Anti-PspC and anti-PspB immunoblot analysis following in vivo oxidative treatment. Cultures of the strains used in panel Ai or Bi were treated with NEM only, Cu-oP followed by NEM, or Cu-oP followed by NEM and then DTT as indicated. Approximate positions of molecular mass marker proteins (in kilodaltons) are indicated.

First, we tested the feasibility of this approach to monitor the dynamics of an interaction in vivo by focusing on the contact between PspB^CT and PspC^CT, which we have already shown to occur in the OFF but not the ON state using site-specific photocross-linking (28). In strains with either PspB-S75C or PspC-V125C alone, cross-linked PspB-PspB and PspC-PspC homodimers were detected, as occurred when we analyzed several other PspB and PspC cysteine substitution mutants (32). However, in the strain with both PspB-S75C and PspC-V125C, an additional complex of ∼23 kDa was detected by the PspB and PspC antisera, consistent with a PspB-PspC complex cross-linked via the PspB^CT and PspC^CT cysteines (Fig. 5). Importantly, this PspB-PspC complex was detected when the Psp response was uninduced but was not detected when the Psp response was switched to its ON state by YsaC production. These results validate this experimental approach, because they are consistent with our previous photocross-linking analysis showing that an inhibitory PspB^CT-PspC^CT complex dissociates when the Psp response is induced (Fig. 1) (28).

Having demonstrated the validity of the approach, we analyzed the proposed PspB^CT-PspC^NT interaction that had been suggested by our in vitro experiments. In a strain with PspB-S32C and PspC-S26C, an ∼23-kDa PspB-PspC complex was detected when the Psp response was uninduced (Fig. 5). This is consistent with our previous report and provides support for the contention that a PspB^CT-PspC^NT interaction can occur in vivo (32) (Fig. 1). Significantly, this complex was not detected when YsaC induced the Psp response (Fig. 5). This supports our hypothesis that a PspB^CT-PspC^NT interaction is a defining feature of the OFF state of the PspB-PspC complex.

Mutations that constitutively activate the Psp response compromise the association between the cytoplasmic domain of PspB and the N-terminal domain of PspC in vivo.

The preceding experiments are consistent with an interaction between PspC^NT and the cytoplasmic domain of PspB being a critical part of the inhibitory complex. Part of the support comes from the finding that the PspC ΔNT and G45W mutations prevent the PspB^CT-PspC^NT interaction in vitro and cause constitutive activation of the Psp response in vivo (Fig. 4) (30). However, there was one remaining discrepancy between this model and our experimental findings. The PspC R58G and L69P transmembrane domain mutations constitutively activate the Psp response in vivo, but they did not prevent the PspB^CT-PspC^NT interaction in vitro (Fig. 5) (30). We hypothesize that an in vitro PspB^CT-PspC^NT interaction can still occur because these transmembrane domain mutations leave the fidelity of the cytoplasmic PspC^NT binding site intact. However, in vivo, these PspC TM domain mutants might be constitutively active, because they cause a conformational shift to the ON state, which dissociates the PspB^CT-PspC^NT interaction within intact cells. Therefore, to resolve the discrepancy, we tested this hypothesis in the final set of experiments.

Derivatives of the PspB and PspC cysteine substitution mutant strains were constructed that also had the PspC-R58G or PspC-L69P mutation. We could not include the PspC-G45W mutation, because its combination with the cysteine substitutions destabilized the PspC protein (data not shown). However, we decided to include a PspB mutant that causes constitutive activation, PspB-W23R (30), to test if any constitutive mutation might dissociate the inhibitory complex between the PspB and PspC cytoplasmic domains. Analysis of Φ(pspA-lacZ) expression confirmed that each mutation still caused constitutive induction when combined with the cysteine substitutions (Fig. 6). Next, we did the disulfide cross-linking analysis as described before. In strains with PspB-S32C and PspC-S26C grown under noninducing conditions (−YsaC), the level of the cross-linked PspB-PspC complex was reduced by all of the constitutive mutations, including the R58G and L69P transmembrane mutations (Fig. 6A). This supports the hypothesis that these transmembrane domain mutations switch the PspBC complex to its ON conformation and dissociate the PspB^CT-PspC^NT complex in vivo, even though the PspC^NT binding site itself remains intact and is able to bind to PspB^CT in vitro. Finally, we also analyzed the effect of the constitutive mutations on the PspB^CT-PspC^CT interaction by using the strains with PspB-S75C and PspC-V125C. As expected, all the mutations abolished detection of the cross-linked PspB-PspC complex, which is consistent with dissociation of the PspB^CT-PspC^CT interaction when the Psp response is activated (Fig. 6) (28).

FIG 6.

Mutations that constitutively activate the Psp response compromise the association between the cytoplasmic domain of PspB and both ends of PspC in vivo. (A) In vivo analysis of the PspB^CT-PspC^NT interaction. (B) In vivo analysis of the PspB^CT-PspC^CT interaction. (Ai and Bi) Φ(pspA-lacZ) operon fusion expression analysis. Strains with the chromosomal pspA operon expressed from the tac promoter encoded PspB-S32C, PspC-C43S-S26C (PspC-S26C), PspB-S75C, PspC-C43S-V125C (PspC-V125C), or the indicated constitutively active mutant derivatives of PspB (B-W23R) or PspC (C-R58G and C-L69P). Strains also contained an empty araBp expression plasmid pBAD33 (−) or the ysaC⁺ derivative pAJD935 (YsaC). Cultures were grown in medium containing 0.02% (wt/vol) arabinose, and β-galactosidase activities were determined. Error bars indicate the positive standard deviations from the means. (Aii and Bii) Anti-PspC and anti-PspB immunoblot analysis following in vivo oxidative treatment. Cultures of the strains used in panel Ai or Bi were treated with NEM only, Cu-oP followed by NEM, or Cu-oP followed by NEM and then DTT as indicated. Asterisks in panels Aii and Bii indicate irrelevant cross-reactive bands. Approximate positions of molecular mass marker proteins (in kilodaltons) are indicated.

DISCUSSION

Signal transduction pathways that control cell envelope stress responses are generally controlled by sensory cytoplasmic membrane proteins, which alternate between inactive (OFF) and active (ON) states. The PspB-PspC complex appears to play this role in controlling the activity of the Psp system in response to most inducing conditions (Fig. 1A) (28). The OFF and ON states of the PspB-PspC complex are defined by the interaction partner of the C-terminal cytoplasmic domain of PspC (PspC^CT), with PspB^CT in the OFF state and PspA in the ON state (28) (Fig. 1A). However, nothing has been known about the underlying factors and mechanisms that control which of these interactions can form. Notably, a major omission has been any information about the role played by the N-terminal cytoplasmic domain of PspC (PspC^NT). Removal of PspC^NT causes constitutive activation of the Psp response, indicating that it is critical for maintenance of the PspC-PspB OFF state (30) (Fig. 1). Here, we have used a combination of in vitro and in vivo approaches to investigate the role of PspC^NT. Our findings suggest that PspC^NT must be precisely positioned in vivo to interact with the cytoplasmic domain of PspB in order for the OFF state to be maintained (Fig. 7).

FIG 7.

Models to explain the role played by the N-terminal domain of PspC in maintaining the PspB-PspC complex in its OFF state. Under noninducing conditions, PspA cannot access its binding site within PspC^CT (white rectangle) because it is bound to PspB^CT. PspC^NT is essential to maintain the OFF state, suggesting that it facilitates the inhibitory PspC^CT-PspB^CT interaction. One possibility, shown on the left, is that PspC^NT binds only to PspB^CT in order to correctly position it for binding to PspC^CT. Another model, shown on the right, is that PspC^NT binds to both PspB^CT and PspC^CT, acting as a scaffold to bring PspB^CT and PspC^CT together for their interaction. The N- and C-terminal cytoplasmic domains of PspB and PspC are labeled N and C, respectively.

Addressing the involvement of PspC^NT in protein-protein interactions has been slowed by our inability to work with the isolated domain in vitro. Previous attempts to produce PspC^NT-maltose-binding protein (MBP), MBP-PspC^NT, and GST-PspC^NT fusion proteins for purification were unsuccessful, because the proteins appeared to be unstable in vivo (28). A PspC^NT-GST protein could be produced, but only at low levels, and it did not provide any positive data in pulldown experiments in vitro (28). Therefore, we developed alternative strategies to investigate the role of the PspC^NT domain in this study. We started by finding that GST-PspB^CT could not recruit PspA in vitro unless PspC was also present (Fig. 2). This suggested that PspB^CT was interacting with PspC to recruit a PspC-PspA complex. However, prior to this work, PspB^CT and PspA were both known to bind only to PspC^CT, but they could not do so simultaneously (28). Therefore, there were two likely possibilities to explain how PspB^CT could recruit a PspC-PspA complex. First, PspA might bind to the other cytoplasmic domain of PspC (PspC^NT), leaving PspC^CT free to interact with PspB. However, this was unlikely, because PspC^NT is dispensable for the recruitment of PspA by PspC in vivo, whereas PspC^CT is essential (26, 27, 30). The more likely possibility was that PspB^CT interacts with the PspC^NT cytoplasmic domain, allowing it to capture a PspC-PspA complex in which PspA occupies PspC^CT.

All of our findings support the contention that PspC^NT can mediate a PspB^CT-PspC interaction in vitro, and that this role is a critical feature of the OFF state of the PspB-PspC complex in vivo. First, all of the in vitro pulldown experiments were consistent with PspB^CT being able to capture a PspC-PspA complex by interacting with PspC^NT (Fig. 2 to 4). Second, we were able to cross-link a cysteine within PspB^CT to one within PspC^NT, suggesting that these domains are close enough to interact in vivo (Fig. 5 and 6). Third, the abundance of this in vivo cross-linked complex was markedly reduced when the Psp response was activated (Fig. 5). Fourth, the PspC ΔNT and G45W mutations that abolish the apparent PspB^CT-PspC^NT interaction in vitro cause constitutive psp gene expression in vivo (Fig. 4 and 5) (30). Fifth, all mutations we tested that cause constitutive activation of the Psp response also reduce the proximity of PspB^CT to PspC^NT in vivo (Fig. 6). Furthermore, especially compelling support for the role of PspC^NT as an interaction interface with PspB^CT was provided by the two-phase pulldown experiments (Fig. 4). The ability of GST-PspB^CT–PspC complexes to capture PspA, or to be dissociated by it, was predicted based on whether the GST-PspB^CT was connected to PspC via an interaction with PspC^NT or with PspC^CT (see Results). Without exception, each of these predictions matched the experimental outcome (Fig. 4).

The proposed role of a PspB^CT-PspC^NT interaction in maintaining the OFF state, coupled with our analysis of PspC-altered function mutants, suggests that there might be different mechanisms by which pspC mutations can cause constitutive activation of the Psp response. The most obvious mechanism is by direct disruption of the PspC^NT binding site. Examples of this are the N-terminal domain deletion mutant (PspC-ΔNT) and the PspC-G45W substitution mutant, both of which behaved in vitro in a manner that is consistent with loss of the PspB^CT-PspC^NT interaction (Fig. 3 and 4). However, mutations in either of the PspC transmembrane domains had no effect on the PspB^CT-PspC^NT interaction in vitro (Fig. 3 and 4) but almost abolished it in vivo (Fig. 6). We hypothesize that their mechanism of action is to disrupt the conformation of the inner membrane PspB-PspC complex, forcing it into its ON state, perhaps by altering the relative positioning of their cytoplasmic domains. Therefore, even though the PspC^NT binding site is unaltered in these transmembrane domain mutants, it cannot adopt the precise positioning required to participate in the inhibitory complex. A need for precise positioning might explain why a random screen to find constitutively active mutants led to the identification of a variety of missense mutations in PspB and PspC (30). We suspect that many of these mutations simply disturb the conformation of PspBC, preventing the required interactions between their cytoplasmic domains that are needed to silence the response. Perhaps this provides a clue about how the regulatory switch normally functions. The PspB-PspC complex might act like a “hair trigger” that is poised to release in response to an inner membrane disturbance and rapidly “snap” into its ON state. In this scenario, a variety of missense mutations might be able to cause enough of a conformational disruption to release this trigger.

The mechanism by which a PspC^NT-PspB^CT interaction maintains the OFF state is not yet clear, but we can speculate (Fig. 7). A key feature in preventing activation of the Psp response is sequestration of the PspA-binding site within PspC^CT by its interaction with PspB^CT (Fig. 1A) (28). Therefore, PspC^NT must be important to maintain this PspC^CT-PspB^CT interaction in vivo. One possibility is that in addition to interacting with PspB^CT, PspC^NT also interacts with PspC^CT (i.e., both cytoplasmic domains of PspC interact with one another). This way, PspC^NT could be thought of as a scaffold that binds to both PspB^CT and to PspC^CT, bringing them close enough together for PspB^CT to occupy the PspA-binding site within PspC^CT (Fig. 7). A variation of this concept is that PspC^NT interacts only with PspB^CT and acts as a controller or guide to correctly position PspB^CT to bind to PspC^CT (Fig. 7).

In summary, in vitro and in vivo approaches have for the first time, to our knowledge, allowed us to investigate the role of a critical domain within the PspB-PspC regulatory complex. This work has provided the first mechanistic insight that can begin to offer an explanation for the essential role of the N-terminal cytoplasmic domain of PspC in silencing the Psp response. Our findings suggest that PspC^NT is a critical domain that must be precisely positioned in vivo in order to participate as an essential component of an inhibitory complex involving all three cytoplasmic domains of PspB and PspC. Some pspC-altered function mutations that render the Psp response constitutively active appear to simply disrupt the PspC^NT binding site. However, analysis of some other constitutively activating mutations suggests that a conformational disturbance might be sufficient to prevent the inhibitory complex from forming in vivo. This raises the intriguing possibility that conformational disruption of the native PspB-PspC regulatory complex, caused by Psp-inducing membrane stress, is the activation mechanism. Some of the goals for future work will be to investigate this possibility and to uncover the precise order in which protein-protein interactions form and dissociate as part of the Psp response activation mechanism.

Funding Statement

This study was supported by award 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.