Activity of a Bacterial Cell Envelope Stress Response Is Controlled by the Interaction of a Protein Binding Domain with Different Partners*

Background: Regulation of the bacterial phage shock protein (Psp) stress response is poorly understood. Results: The C-terminal domain of PspC (PspCCT) can interact with PspA or PspB. Conclusion: Induction of psp gene expression involves a PspCCT binding partner switch from PspB to PspA. Significance: Understanding the activation mechanism is critical because this system is essential for virulence in Yersinia enterocolitica and Salmonella enterica. The bacterial phage shock protein (Psp) system is a highly conserved cell envelope stress response required for virulence in Yersinia enterocolitica and Salmonella enterica. In non-inducing conditions the transcription factor PspF is inhibited by an interaction with PspA. In contrast, PspA associates with the cytoplasmic membrane proteins PspBC during inducing conditions. This has led to the proposal that PspBC exists in an OFF state, which cannot recruit PspA, or an ON state, which can. However, nothing was known about the difference between these two states. Here, we provide evidence that it is the C-terminal domain of Y. enterocolitica PspC (PspCCT) that interacts directly with PspA, both in vivo and in vitro. Site-specific photocross-linking revealed that this interaction occurred only during Psp-inducing conditions in vivo. Importantly, we have also discovered that PspCCT can interact with the C-terminal domain of PspB (PspCCT·PspBCT). However, the PspCCT·PspBCT and PspCCT·PspA interactions were mutually exclusive in vitro. Furthermore, in vivo, PspCCT contacted PspBCT in the OFF state, whereas it contacted PspA in the ON state. These findings provide the first description of the previously proposed PspBC OFF and ON states and reveal that the regulatory switch is centered on a PspCCT partner-switching mechanism.

Many cells monitor their cell envelope and respond to hostile conditions and events that can compromise it. Bacteria achieve this with specialized signaling pathways composed of envelopeassociated sensory components and cytoplasmic transcriptional regulators, which work together to alter gene expression to combat the threat (1)(2)(3)(4). These signaling pathways are known as extracytoplasmic/envelope stress responses. One is the phage shock protein (Psp) 2 system, so-named because it is induced when Escherichia coli is infected with a filamentous phage (5). However, the Psp system is not linked to phage infection exclusively. Instead, it is thought to promote bacterial survival during conditions that threaten the integrity of the cytoplasmic membrane (6 -8). The Psp system is conserved in many Gram-negative bacteria, including several pathogens. It is essential for the virulence of the enteric pathogens Yersinia enterocolitica and Salmonella enterica serovar Typhimurium (9,10). Furthermore, the psp genes are up-regulated during biofilm formation, macrophage infection, and the establishment of persister cells (11)(12)(13)(14). Homologues of some Psp system components are also present in Gram-positive bacteria, archaea, and plant chloroplasts (1,15).
The Psp system has been studied most in E. coli and Y. enterocolitica, where the psp genes are expressed from 54 dependent promoters that require the DNA-binding protein, PspF, for activity (16,17). In both species the known PspF regulon consists of only the pspA operon and the unlinked pspG gene (9,16,18). However, the pspA operons of these two species differ at the distal ends, pspABCDE in E. coli and pspABCDy-cjXF in Y. enterocolitica. Furthermore, only pspA, -B, -C, and -F have been linked to robust phenotypes, with PspA, -B, and -C each having dual regulatory and stress-mitigating functions (6 -8, 19, 20). Thus, PspA, -B, -C, and -F are considered the core components of the Psp system. Expression of the pspA operon is induced by stimuli that include extreme heat and hyperosmotic shock, high ethanol concentration, and the mislocalization of outer membrane secretin proteins into the cytoplasmic membrane (6,8,21). Of these, only secretin mislocalization induces psp gene expression specifically in both E. coli and Y. enterocolitica (18,22). Secretins are multimeric proteins with significance because they normally form the outer membrane channel of several types of bacterial protein export systems, many of which are involved in pathogenesis (23).
Regulation of pspA operon expression centers upon controlling the activity of the PspF transcription factor. In uninduced Y. enterocolitica cells, PspA inhibits PspF by interacting with it in the cytoplasm (24,25). However, an inducing stimulus causes PspA to redistribute from the cytoplasm to the cytoplasmic membrane, releasing PspF to activate the pspA promoter (24).
The integral cytoplasmic membrane protein complex PspBC is thought to dissociate the PspA-PspF complex by sequestering PspA (25). Therefore, we proposed a model in which an inducing stimulus causes PspBC to switch from an OFF to an ON state, with only the latter able to interact with PspA (26). The C-terminal domain of PspC (PspC CT ) has been proposed to mediate the interaction with PspA (26,27). For example, PspC CT pulls PspA down from a complex cell lysate in vitro, although it is not known if PspC CT contacts PspA directly (26). Nor is it known what might prevent PspA from associating with PspBC in non-inducing conditions. In other words, what are the differences between the OFF and ON states of PspBC? In this work we have developed a combination of in vitro and in vivo approaches to investigate these questions. Our findings provide the first description of PspBC OFF and ON states and show that the regulatory switch is a PspC CT partner-switching mechanism.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Routine Growth-Bacterial strains and plasmids are listed in Table 1, and primer sequences are listed in Table 2. Y. enterocolitica strains were grown at 26°C or 37°C as noted. E. coli strains were grown at 37°C. Strains were grown in Luria-Bertani (LB) broth or on LB agar plates. Antibiotics were used as described previously (28) except that kanamycin was used at 40 g/ml for E. coli strain MC3 and its derivatives.
Plasmid Constructions-All PCR-generated fragments were verified by DNA sequencing. Plasmids encoding GST-PspC CT with or without the V125D mutation or GST-PspB CT were A plasmid encoding PspB and PspC-FLAG was constructed by amplifying pspBC from plasmid pAJD1136 with primers 396/2085 and cloning it as a SacI/XbaI fragment into pWSK129 to generate plasmid pAJD2510. pAJD2510 derivatives containing an amber codon (TAG) at different positions within pspC, the PspB-W23R mutation, or the pspC-V125D mutation were constructed by sewing overlap extension (SOE) PCR (29). Briefly, the pAJD2510 insert was amplified in two sections as separate PCR fragments with primers that generated a short overlap between them, encoding the desired mutation. To generate plasmid pAJD2519 (codon 131 of pspC changed to TAG), the upstream fragment was amplified with primers 396/2084, and the downstream fragment was amplified with primers 2083/2085, both from pAJD1136 template. To generate plasmid pAJD2513 (codon 109 of pspC changed to TAG), the upstream fragment was amplified with primers 396/2074, and the downstream fragment was amplified with primers 2073/ 2085, both from pAJD1136. To generate plasmid pAJD2534 (encoding PspB-W23R and with codon 109 of pspC changed to TAG), the upstream fragment was amplified from pAJD1340 with primers 396/942, and the downstream fragment was amplified with primers 941/2085 from pAJD2513. To generate plasmid pAJD2600 (encoding PspC-V125D and with codon 109 of pspC changed to TAG), the upstream fragment was ampli-fied with primers 396/2074, and the downstream fragment was amplified with primers 2073/2085, both from pAJD1659. To generate plasmid pAJD2602 (encoding PspB-W23R, PspC-V125D and with codon 109 of pspC changed to TAG), the upstream fragment was amplified with primers 396/2074 from pAJD2534, and the downstream fragment was amplified with primers 2073/2085 from pAJD1659. In all cases, each pair of upstream and downstream fragments was joined by SOE PCR and cloned into pWSK129 as SacI/XbaI fragments. Plasmid pAJD2601, encoding PspC-V125D and with codon 131 of pspC changed to TAG, was made by amplifying pspBC from plasmid pAJD1136 with primers 396/2198 and cloning it as a SacI/XbaI fragment into pWSK129. A plasmid encoding E. coli PspB and PspC was constructed by amplifying pspBC from the E. coli chromosome with primers 1573/1575 and cloning it as a SacI/ XbaI fragment into pWSK129 to generate pAJD2524.
Strain Constructions-To make Y. enterocolitica strains encoding PspB-3ϫFLAG or PspC-3ϫFLAG, for PspB two ϳ0.6-kb fragments flanking the 3Ј end of pspB were amplified by PCR to have a common 20-bp overlap. The upstream fragment was amplified from the Y. enterocolitica chromosome with primers 856/857, whereas the downstream fragment was amplified from a plasmid encoding PspB-3ϫFLAG (30) with primers 858/397. These fragments were joined by SOE PCR with primers 856/397 to generate an ϳ1.2-kb fragment with the 3ϫFLAG encoding region surrounded by ϳ 0.6 kb on each side, and this fragment was cloned into sacB ϩ suicide plasmid pRE112. For PspC, two ϳ0.5-kb fragments flanking the 3Ј end of pspC were amplified by PCR to have with a common 20-bp overlap. The upstream fragment was amplified from a plasmid encoding PspC-3ϫFLAG (30) with primers 396/1557, whereas the downstream fragment was amplified from the Y. enterocolitica chromosome with primers 1562/881. These fragments were joined by SOE PCR with primers 396/881 to generate an ϳ1-kb fragment with the 3ϫFLAG-encoding region surrounded by ϳ 0.5 kb on each side, and this fragment was cloned into sacB ϩ suicide plasmid pRE112. The suicide constructs were used to fuse the 3ϫFLAG-encoding regions to the chromosomal Y. enterocolitica pspB or pspC genes by integration, selection for sucrose-resistant segregants, and confirmation by colony PCR and DNA sequencing. The ⌬(pspF-pspE) derivative of E. coli strain BL21 (AJDE1770) was made by introducing the ⌬(pspF-pspE)::kan mutation (20) into strain BL21 by phage P1 vir transduction and confirming by colony PCR. To construct a ⌬pspBC derivative of E. coli MC3 that also encodes PspA-His 8 (AJDE2892), three fragments were amplified from E. coli genomic DNA by PCR. The first fragment encoding PspA without its first 40 codons (ЈpspA) was amplified with primers 2026/2067, which incorporated a SacI site upstream of ЈpspA and a His 8 sequence immediately upstream of the pspA stop codon. The second fragment was amplified with primers 2068/2027 so that it had a BglII site preceded by the first 7 codons of E. coli pspB, the pspA-pspB intergenic region, and then a His 8 sequence that matched the end of the first fragment. The third fragment was amplified with primers 2028/2029 so that it had a BglII site followed by the last 7 codons of E. coli pspC and ϳ550 bp of downstream DNA followed by an XbaI site. Then the first two fragments were joined by a SOE PCR reaction via their overlapping His 8 sequence using primers 2026/2027, and the resulting fragment was ligated to the third fragment via their BglII sites to generate the in-frame pspBC deletion. The resulting fragment was digested with SacI/XbaI and ligated into sacB ϩ allelic exchange plasmid pSR47S to generate pAJD2509. Plasmid pAJD2509 was used to exchange the E. coli MC3 chromosomal ЈpspABC region with ЈpspA-his 8 ⌬pspBC by integration, selection for sucrose-resistant segregants, and confirmation by colony PCR and DNA sequencing.
Native Co-immunoprecipitation Assay-Strains were grown to saturation, diluted to an approximate absorbance of 0.08 (600 nm) in 500 ml of LB broth and shaken at 250 rpm for 2 h at 37°C. Then 0.02% arabinose (final concentration) was added to induce ysaC expression, and growth continued at 37°C for an additional 2 h. Cells from the equivalent of 300 ml at an absorbance of 1 (600 nm) were harvested by centrifugation and resuspended in PBS (137 mM NaCl, 2.7 mM KCl, 10.1 mM Na 2 HPO 4 , 1.4 mM KH 2 PO 4 ) containing 2 mM phenylmethylsulfonyl fluoride (PMSF) and 2ϫ complete protease inhibitors (Roche Applied Science). After storage overnight at Ϫ20°C, 1 ml of 1.67 mg/ml DNase I per 5 g wet weight of cells was added, and cells were disrupted by sonication. Unbroken cells were removed by centrifugation at 20,000 ϫ g for 20 min, and then the supernatant was centrifuged at 100,000 ϫ g for 2.5 h to isolate the membrane fraction. Membranes were resuspended in membrane solubilization-buffer (50 mM Tris-HCl, pH 7.4, 300 mM NaCl, 5 mM EDTA, 2 mM PMSF, and 2ϫ complete protease inhibitors) (10 ml/g of membrane pellet) and homogenized with a tissue grinder. Membrane proteins were solubilized by adding 1% n-dodecyl ␤-D-maltoside (DDM) (final concentration) and inverting slowly for 2 h at 4°C. The insoluble material was removed by centrifuging for 30 min at 100,000 ϫ g. Then, the DDM-solubilized membranes were precleared by adding 30 l of 50% protein A-Sepharose slurry equilibrated in PBS, rotating for 1 h at 4°C, followed by removal of the protein A-Sepharose by centrifugation.
A protein A-Sepharose-anti-FLAG complex was made by mixing a 50% protein A-Sepharose slurry with anti-FLAG M2 monoclonal antibody (Sigma; 60 l of 50% protein A-Sepharose slurry added to 1 l of anti-FLAG antibody) and incubating for 3 h at 4°C with gentle rocking. The complex was washed twice with membrane solubilization buffer and then resuspended in the same buffer to restore the 50% protein A-Sepharose slurry concentration. The precleared DDM solubilized membranes were added to 60 l of the protein A-Sepharose-anti-FLAG complex and incubated overnight at 4°C with gentle rocking. Proteins were immunoprecipitated by centrifugation, washed 2 times with membrane-solubilization buffer and 3 times with PBS containing 0.1% Triton X-100, resuspended in 60 l of SDS-PAGE sample buffer, and then boiled for 10 min before analysis by SDS-PAGE and immunoblotting.
GST Fusion Protein Membrane Lysate Pulldown Assays-GST pulldown assays to detect interactions between the C terminus of PspC and various Psp proteins were done by purifying GST derivatives from E. coli strain AJDE1770 onto glutathione-Sepharose beads (GE Healthcare), incubating them with solubilized DDM membrane lysates from Y. enterocolitica strains, and then washing before elution by boiling in SDS-PAGE sample buffer exactly as described previously except that 0.1% (v/v) Triton X-100 was added to the wash buffer (26).
In Vitro GST and Maltose-binding Protein (MBP) Fusion Protein Interaction Assays-E. coli strain AJDE1770 transformed with plasmid pREP4-groESL (to increase protein solubility) as well as a plasmid encoding GST alone (pGEX-6P-1) or GST-PspC CT (with or without the V125D mutation) was grown to mid-exponential phase at 20°C in 1 liter of LB broth with 1 mM IPTG (final concentration). Cells were resuspended in 50 ml of PBS containing 250 g/ml lysozyme and 0.1% (v/v) Triton X-100 and disrupted by sonication. Unbroken cells were removed by centrifugation at 20,000 ϫ g for 25 min at 4°C. 1 ml of the supernatant was incubated with 40 l of glutathione-Sepharose (GE Healthcare) for 1 h at 4°C with gentle rocking. The beads were recovered by centrifugation, washed 3 times with 1 ml of PBS, and incubated with 1 ml of PBS containing 5% (w/v) bovine serum albumin. The beads were recovered by centrifugation and washed twice with 1 ml of MBP column buffer (20 mM Tris-HCl, pH 7.4, 200 mM NaCl, 1 mM EDTA).
E. coli strain AJDE1770 transformed with a plasmid encoding MBP-LacZ␣ or MBP-PspB CT was grown in LB broth supplemented with 0.2% (w/v) glucose to saturation at 30°C. The culture was diluted 100-fold in 500 ml of LB broth supplemented with 0.2% (w/v) glucose and shaken at 250 rpm for 3 h at 37°C. 0.3 mM IPTG (final concentration) was added, and growth was continued for an additional 2 h at 37°C. Cells were harvested by centrifugation and resuspended in MBP column buffer and stored overnight at Ϫ20°C. Cells were thawed and disrupted by sonication, and unbroken cells were removed by centrifugation at 20,000 ϫ g for 20 min. The supernatant was incubated with 5 ml of amylose resin (New England Biolabs) for 2 h at 4°C and then packed into a column. The column was washed with 20 volumes of MBP column buffer. Proteins were eluted with 5 ml of MBP column buffer supplemented with 10 mM maltose and ten 0.5-ml fractions were collected. Protein concentration in each fraction was determined by the bicinchronic acid assay (BCA) method (Pierce), and the three fractions with the maximum yield were pooled.
15 g of MBP-LacZ␣ or MBP-PspB CT was mixed with the GST or GST-PspC CT (with or without the V125D mutation), immobilized onto glutathione-Sepharose, and then incubated for 3 h at 4°C with gentle rocking. The beads were washed 5 times with 1 ml of 20 mM Tris-HCl, pH 7.4, 500 mM NaCl, 1 mM EDTA and collected by centrifugation. Proteins were recovered by resuspending the beads in 60 l of 2ϫ SDS-PAGE sample buffer and boiling for 10 min.
Two-phase GST Fusion Protein Membrane Lysate Pulldown Assays-GST or GST-PspC CT proteins were isolated from E. coli strain AJDE1770 transformed with plasmid pREP4-groESL and immobilized onto glutathione-Sepharose beads exactly as described above, except that 80 l of glutathione-Sepharose was used, and the final preparation was washed with PBS instead of MBP-column buffer.
Solubilized membrane lysates were made from Y. enterocolitica strains by growing them to mid-exponential phase at 26°C in 1 liter of LB broth containing 0.1 mM IPTG. Cells were harvested by centrifugation and resuspended in TBS (50 mM Tris-HCl, pH 8.0, 400 mM NaCl, 2 mM PMSF, 2ϫ complete protease inhibitors) and stored overnight at Ϫ20°C. Cells were thawed, 1 ml of 1.67 mg/ml DNase I per 5 g wet pellet weight was added, and cells were disrupted by sonication. Unbroken cells were removed by centrifugation at 20,000 ϫ g for 20 min, and then the membrane fraction was isolated from the supernatant by centrifugation at 100,000 ϫ g for 2.5 h. The membrane pellet was resuspended in the membrane solubilization buffer described above at a ratio of 10 ml/g, and membrane proteins were solubilized by adding 1% (w/v) DDM (final concentration). Insoluble material was removed by centrifuging for 30 min at 100,000 ϫ g. The supernatant was diluted 10-fold in 50 mM Tris-HCl, pH 7.4, 10% (w/v) glycerol, 150 mM NaCl, and 3 mM ␤-mercaptoethanol.
50 l of a solubilized membrane lysate was mixed with the GST or GST-PspC CT immobilized on glutathione-Sepharose beads and rocked gently for 2 h at 4°C. The beads were washed 5 times with membrane solubilization buffer containing 500 mM NaCl and separated into two equal samples. One sample was mixed with a second DDM-solubilized membrane lysate and rocked gently for 2 h at 4°C. Then both samples were washed 5 times with membrane solubilization buffer containing 500 mM NaCl, resuspended in 60 l of 2ϫ SDS-PAGE sample buffer, and boiled to recover all proteins. Equal volumes of the elution samples were analyzed by SDS-PAGE and immunoblotting.
In Vivo Photocross-linking-E. coli AJDE2892 was transformed with plasmid pEVOL-pBpF and either plasmid pAJD2510 or one of its derivatives. Each strain also contained tacp-yscC expression plasmid pAJD126 or the empty tacp expression vector pVLT35. The strains were grown in LB broth to saturation. They were then diluted to an absorbance of ϳ0.04 (600 nm) in 5 ml of LB broth supplemented with 0.2 mM p-benzoyl-L-phenylalanine (pBpa; Fisher) and grown for 2 h at 37°C.
Then 200 M IPTG (final concentration) was added to induce pspBC and yscC expression, and growth was continued at 37°C for an additional 2 h. Cells from the equivalent of 10 ml at an absorbance of 1 (600 nm) were harvested by centrifugation, washed with 1 ml of PBS, pH 7.3 (140 mM NaCl, 2.7 mM KCl, 10.1 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 adjusted to a pH of 7.3), and then resuspended in 0.45 ml of PBS, pH 7.3, supplemented with 200 M IPTG (final concentration). The resuspended cells were divided into 2 samples of 0.2 ml each. One sample was transferred to a 96-well microtiter plate and irradiated with ultra violet light (UV) at a wavelength of 365 nm using a EA-240 lamp (Spectroline) for 30 min on ice. The second sample was treated similarly, except that it was not exposed to UV light. The cells were then harvested by centrifugation, resuspended in 80 l of SDS-PAGE sample buffer, and boiled for 10 min before analysis by SDS-PAGE and immunoblotting.
␤-Galactosidase Assays-Y. enterocolitica strains were grown to saturation and diluted into 5 ml of LB broth in 18-mm-diameter test tubes to an optical density (600 nm) of ϳ0.04. The cultures were grown on a roller drum at 37°C for 2 h. Then, 200 M IPTG or 0.02% arabinose (final concentration) was added to induce yscC and pspBC or to induce ysaC, respectively. Cells were grown for another 2 h at 37°C before harvesting. ␤-Galactosidase enzyme activity was determined at room temperature in permeabilized cells as described (31). Activities are expressed in arbitrary Miller units (32). Individual cultures were assayed in duplicate, and average values from three independent cultures are reported. However, to monitor ⌽(pspA-lacZ) expression for the native co-immunoprecipitation and photocross-linking experiments, samples were taken directly from the experimental cultures before harvesting and assayed in duplicate with the average reported.
Polyclonal Antisera and Immunoblotting-Proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes by semidry electroblotting. Enhanced chemiluminescent detection followed sequential incubation with polyclonal antiserum or monoclonal antibodies and then goat antirabbit IgG or goat anti-mouse IgG (Bio-Rad) horseradish peroxidase conjugate used at 1 in 5000 dilution. Dilutions of polyclonal antisera were 1 in 10,000 for anti-PspA, anti-PspC, and anti-FtsH and 1 in 20,000 for anti-PspB (20,24,33). Monoclonal antibodies were used at a 1 in 10,000 dilution for anti-DnaK (Assay Designs), 1 in 5,000 for anti-FLAG (Sigma) and anti-His (GenScript), and 1 in 50,000 for anti-MBP (New England Biolabs) and anti-GST (GenScript).

RESULTS
The C-terminal Domain of Y. enterocolitica PspC Can Bind to PspA in Vitro Independently of Other Core Psp Proteins-Activation of the Psp system coincides with PspA switching its binding partner from PspF to PspBC (Refs. 24 and 25 and Fig.  1). The C-terminal cytoplasmic domain of PspC (PspC CT ) is critical for PspA recruitment, but it is not known if PspC CT and PspA contact each other directly or if instead PspB bridges their association (26,27,33). Therefore, we began this work with an extensive PspC CT interaction analysis.
GST, GST-PspC CT , or GST-PspC CT-V125D proteins were bound to glutathione-Sepharose and used as bait to capture proteins from a detergent-solubilized Y. enterocolitica Psp ϩ membrane lysate (Fig. 2). The PspC V125D mutation abolishes induction of psp gene expression in vivo and is in the hydrophobic face of a predicted amphipathic helix (Ref. 33; 125 is the amino acid position in full-length PspC). Consistent with our previous report, PspA was captured by GST-PspC CT , but this was almost abolished by the V125D mutation ( Fig. 2 and Ref. 26). Next, we used a membrane lysate from an isogenic ⌬pspBC strain in which PspA was the only core Psp protein present. PspA was still captured by GST-PspC CT (Fig. 2). Therefore, the association between PspA and GST-PspC CT does not require bridging by PspB or by the full-length PspC, that were both in the Psp ϩ membrane lysate. Although these experiments cannot rule out bridging by another protein(s), they are consistent with a direct GST-PspC CT -PspA interaction, which is compromised by the V125D mutation.
The C-terminal Domain of PspC Can Bind to the Cytoplasmic Domain of PspB Directly in Vitro-During the preceding GST-PspC CT pulldown experiments with a Psp ϩ membrane lysate, we also analyzed the samples by anti-PspB immunoblot. This revealed that PspB was retained by the glutathione-Sepharose-GST-PspC CT complex in addition to PspA (Fig. 2). We repeated the experiment with a membrane lysate from an isogenic strain in which PspB was the only core Psp protein present, and PspB was still bound by GST-PspC CT (Fig. 2). Furthermore, introduction of the V125D mutation into GST-PspC CT prevented PspB binding, implicating the hydrophobic face of the predicted PspC CT amphipathic helix as a binding determinant for PspB as well as for PspA. A likely explanation for these results is that PspC CT can interact with the cytoplasmic C-terminal domain of PspB (PspB CT ) directly. Therefore, we tested this hypothesis using a purified Y. enterocolitica PspB CT fusion protein instead of a complex membrane lysate as the prey in our pulldown assay.
15 g of MBP-PspB CT or MBP-LacZ␣ (negative control) fusion protein was mixed with GST, GST-PspC CT , or GST-PspC CT-V125D proteins bound to glutathione-Sepharose. After washing, all proteins were eluted in SDS-PAGE sample buffer and analyzed by immunoblot. Consistent with our direct interaction hypothesis, MBP-PspB CT bound to GST-PspC CT but not to GST or to GST-PspC CT-V125D (Fig. 3). In contrast, MBP-LacZ␣ did not bind to any of the GST proteins. Therefore, PspC CT can interact directly with PspB CT in vitro. Furthermore, the observation that the V125D mutation prevents Psp-C CT from interacting with both PspB CT and PspA suggests that their binding sites might overlap. This is consistent with a model in which PspC CT interacts with PspB CT in the OFF state, whereas it interacts with PspA in the ON state (Fig. 1). If all of this is correct, the binding of PspB CT or PspA to PspC CT might be mutually exclusive. Next, we tested this prediction using a modification of our in vitro pulldown assay.
Evidence That the Interaction of the C-terminal Domain of PspC with PspA or PspB Is Mutually Exclusive in Vitro-The GST pulldown assay was modified to include a second phase. In  phase one, a membrane lysate from a strain with PspA as the only core Psp protein (Fig. 4A, prey lysate 1) was mixed with glutathione-Sepharose-GST-PspC CT followed by washing. This served to saturate the GST-PspC CT with bound PspA, which was monitored by taking half of the Sepharose beads and eluting all proteins in SDS-PAGE sample buffer (Fig. 4A, Elution 1). The other half of the glutathione-Sepharose-GST-PspC CT ⅐PspA beads were used for phase two by mixing them with a second membrane lysate containing PspB ( Fig. 4A; prey lysate 2) followed by washing and elution (Fig. 4A, Elution 2). Equal amounts of elutions 1 and 2 were analyzed by SDS-PAGE and immunoblotting. We reasoned that if the binding of PspA and PspB to GST-PspC CT was not mutually exclusive, then in phase two, incubation of a GST-PspC CT ⅐PspA complex with a PspB-containing membrane lysate should result in retention of both PspA and PspB. However, if binding is mutually exclusive, then PspB should fail to bind or should displace PspA from the GST-PspC CT ⅐PspA complex.
In the first experiment PspA was bound to GST-PspC CT in phase one (Fig. 4B, Experiment A, Elution 1). Then, when this complex was incubated with a membrane lysate containing PspB, PspA was displaced, and PspB was retained (Fig. 4B,  Experiment A, Elution 2). Displacement of PspA was not a nonspecific consequence of mixing the GST-PspC CT ⅐PspA complex with a membrane lysate because PspA was not displaced when GST-PspC CT ⅐PspA was mixed with a lysate that did not contain PspB (data not shown). We also did an experiment where the order of addition was reversed (Fig. 4B, Experiment  B). In this case PspA displaced PspB from a GST-PspC CT ⅐PspB complex (some PspB remained, but the majority was displaced; Fig. 4B, Experiment B, compare Elutions 1 and 2). These results support the hypothesis that the binding of PspA or PspB to PspC CT is mutually exclusive in vitro.
The Isolated C-terminal Domain of PspC Can Activate the Psp System in Vivo-The preceding data supported a model in which a Psp-inducing signal might cause PspC CT to sequester PspA perhaps by switching its interaction partner from PspB CT (Figs. 1-4). However, all of the data were from in vitro experiments. Therefore, we designed a series of in vivo experiments to interrogate this model further.
If PspC CT normally sequesters PspA away from PspF (Fig.  1B), then overproducing only this domain of PspC might be sufficient to activate the Psp system in vivo. To test this, we introduced IPTG-inducible expression plasmids encoding GST-PspC CT , with or without the V125D mutation, into a Y. enterocolitica strain with a single copy ⌽(pspA-lacZ) operon fusion. As negative controls, we included plasmids encoding GST only or GST-PspB CT (which does not pull down PspA from a membrane lysate in which PspA is the only core Psp protein present; data not shown and see "Discussion"). Cells were grown in non-Psp inducing conditions, and ␤-galactosidase activities were determined.
When IPTG was not included in the growth medium, leaky expression was sufficient for the GST protein derivatives to be detected by immunoblot (Fig. 5). However, only GST-PspC CT induced ⌽(pspA-lacZ) expression (Fig. 5). When IPTG was added, the level of the GST protein derivatives increased, but once again only GST-PspC CT induced ⌽(pspA-lacZ) expression. Furthermore, as the GST-PspC CT level increased, ⌽(pspA-lacZ) expression also increased, showing that the effect was concentration-dependent. This is consistent with GST-PspC CT sequestering the PspA protein away from PspF. The failure of GST-PspC CT-V125D to induce ⌽(pspA-lacZ) expression suggests that the V125D mutation prevents PspA sequestration in vivo. This agrees with the V125D mutation preventing the GST-PspC CT ⅐PspA interaction in vitro (Fig. 2). Therefore, these results support the model that PspC CT is directly involved in sequestering PspA from PspF.
The Activation Mechanism Does Not Involve Complete Dissolution of the PspBC Complex in Vivo-If the PspC CT ⅐PspB CT interaction we detected in vitro occurs in the OFF state of fulllength PspBC in vivo, then it must dissociate in inducing conditions so that PspA can bind to PspC CT . A simple hypothesis is that the entire PspB⅐PspC complex separates when an inducing signal is encountered. PspB and PspC interact in vivo, and it has been suggested that this complex is present in both OFF and ON conditions (20,30). However, in those experiments the PspBC complex was identified only after treating cells with formaldehyde to cross-link proteins. Therefore, we were concerned that it might have been possible to cross-link PspB and PspC, which are close together but no longer in a stable complex. To address this we developed a co-immunoprecipitation assay to isolate a PspBC complex without using any cross-linker.
The low level of PspBC in uninduced cells makes complex detection difficult, and the massive increase in their concentrations when the Psp system is induced would complicate interpretation. Therefore, we used our Y. enterocolitica strain with the chromosomal pspA operon expressed from a tacp promoter (24). Leaky tacp expression leads to higher levels of PspBC than their basal levels in a wild type strain, but they can still regulate ⌽(pspA-lacZ) expression normally. However, Psp protein levels remain constant ( Fig. 6; Refs. 19, 24, and 25). To facilitate the co-immunoprecipitation, a sequence encoding the 3ϫFLAG  MAY 1, 2015 • VOLUME 290 • NUMBER 18 epitope was fused to the end of the chromosomal pspB or pspC gene. The FLAG tags did not compromise regulation of ⌽(pspA-lacZ) expression (Fig. 6A).

Signal Transduction in the Psp Stress Response
Strains were grown in inducing (ϩYsaC secretin) or noninducing (ϪYsaC secretin) conditions, and PspB-3ϫFLAG or PspC-3ϫFLAG was isolated from detergent-solubilized membranes by anti-FLAG immunoprecipitation. PspC co-immunoprecipitated with PspB-3ϫFLAG regardless of the induction status (Fig. 6B). Similarly, PspB co-immunoprecipitated with PspC-3ϫFLAG. The control integral inner membrane protein FtsH did not co-immunoprecipitate with PspB-3ϫFLAG or with PspC-3ϫFLAG, and when PspB and PspC were not FLAG-tagged neither of them was present in the immunoprecipitates (Fig. 6B). These results suggest that PspB and PspC remain in complex when an inducing signal is encountered, perhaps via interactions between their transmembrane domains as has been suggested (27). If so, an inducing signal might dissociate or alter the interaction between only their C-terminal domains (Fig. 1).

Detection of a Contact between PspA and the C-terminal Domain of PspC upon Activation of the Psp Response in Vivo-
Next, we wanted an in vivo test of the hypothesis that PspC CT interacts directly with PspA but only when an inducing signal is present. Photo-reactive site-specific cross-linking has been used to map interacting protein domains (e.g. Ref. 34 and 35). The photo-reactive amino acid analog pBpa forms cross-links when cells are exposed to UV light (36,37). pBpa is incorporated at a specific location in a bacterial protein by introducing an internal amber codon into the gene and decoding it with an orthogonal aminoacyl-tRNA synthetase/suppressor tRNA pair

. Evidence that interaction of the PspC C-terminal domain with PspA or PspB is mutually exclusive in vitro.
A, summary of the protocol used for the GST fusion protein two-phase membrane lysate pulldown assay. B, immunoblot analysis. In Experiment A, GST or GST fused to the PspC C-terminal domain (GST-PspC CT ) was bound to glutathione-Sepharose (beads) and incubated with a detergent-solubilized membrane lysate from a Y. enterocolitica strain in which the only core Psp protein present was PspA (prey lysate 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 core Psp protein present was PspB (prey lysate 2). After washing, proteins were recovered by boiling in SDS-PAGE sample buffer (Elution 2). Experiment B was done similarly, except that the order of incubation with the PspA and PspB membrane lysates was reversed. For each experiment, membrane lysates (Inputs) and recovered proteins (Elutions) were analyzed by SDS-PAGE and immunoblotting with PspA or PspB antiserum. The GST fusion protein in each elution was detected by Ponceau S staining of the immunoblot membrane (for experiments A and B the elution samples for Ponceau S staining were run on the same gels, but irrelevant lanes between elutions 1 and 2 have been removed). (38). This system has been used in E. coli but did not work in even closely related bacteria without extensive modification (39). Indeed, the aminoacyl-tRNA synthetase/suppressor tRNA-encoding plasmid pEVOL-pBpF was extremely toxic to Y. enterocolitica (data not shown). Others have dealt with this by using E. coli as a surrogate host to study their proteins (40). Therefore, we decided to use E. coli as the host but to continue our analysis of the Y. enterocolitica PspBC proteins used in all preceding experiments. To facilitate this we modified E. coli ⌽(pspA-lacZ) operon fusion strain MC3 by introducing a ⌬pspBC in-frame deletion and fusing a His 8 tag-encoding region to its chromosomal pspA gene to enhance PspA detection. To test this strain we introduced lacZp expression plasmids encoding E. coli PspBC, Y. enterocolitica PspBC, or the empty vector. Cells were grown ϩ/ϪYscC secretin, and ⌽(pspA-lacZ) expression was monitored by measuring ␤-galactosidase activity. Miller unit activities were 430 ϪYscC and 510 ϩYscC with the empty vector, 480 ϪYscC and 2400 ϩYscC with E. coli PspBC, and 510 ϪYscC and 2,600 ϩYscC with Y. enterocolitica PspBC. Therefore, Y. enterocolitica PspBC complemented the ⌬pspBC mutation and communicated with E. coli PspA, and the His 8 tag on PspA did not prevent normal regulation.
Next, we constructed derivatives of the Y. enterocolitica pspBC plasmid in which pspC had a single amber codon (denoted as X) at various positions within the region encoding its predicted C-terminal amphipathic helix. PspC also had a FLAG tag added to its C terminus for specific detection of the pBpa-suppressed protein (explained below). These plasmids were used to transform the E. coli host strain containing pEVOL-pBpF and screened for PspC functionality after growth with pBpa. In several cases, ⌽(pspA-lacZ) expression was not induced by YscC, suggesting that the pBpa substitution destroyed PspC regulatory function or that suppression of the amber codon was insufficient (data not shown). However, a plasmid with an amber codon at position 131 (G131X) allowed induction of ⌽(pspA-lacZ) expression in response to YscC, which was enhanced to the wild type level when pBpa was present (Fig. 7A). Suppression was also tested by immunoblot anal-ysis. Without pBpa, the FLAG antibody did not detect any PspC-G131X (Fig. 7A). This was expected because the protein should be truncated at position 131, which precedes the region encoding the C-terminal FLAG tag. However, PspC-G131X was detected by FLAG antibody when pBpa was included in the growth medium (Fig. 7A). Furthermore, when anti-PspC polyclonal antiserum was used instead of FLAG antibody, only truncated PspC-G131X was detected without pBpa, whereas both truncated and full-length PspC-G131X were detected with pBpa (Fig. 7A).
To test if pBpa at position 131 of PspC would cross-link to PspA, cells were grown with or without YscC in media supplemented with pBpa and then exposed to UV light. Whole cell lysates were analyzed by SDS-PAGE and immunoblot (Fig. 7B). When YscC was overproduced (ϩYscC), two UV-dependent slower-migrating species were observed for PspC-G131X. One was also detected when PspC did not have the G131X mutation (but at a much lower level), which suggests that it was pBpaindependent. A similar UV-specific, but pBpa-independent phenomenon has been observed in photocross-linking analysis of another protein (41). Interestingly, the size of this complex was consistent with a PspC dimer (see "Discussion"). Regardless, the other UV-dependent band was specific to PspC-G131X, and its migration suggested that it might be a PspC⅐PspA complex (predicted mass of ϳ43 kDa). To test this, the samples were also analyzed by an anti-His 6 immunoblot to detect PspA, and a PspC-G131X/UV-specific band of the same size was also detected (Fig. 7B). Therefore, this band corresponds to a PspC⅐PspA cross-linked complex. This is the first demonstration of a direct contact between PspC and PspA. Furthermore, because pBpa was within the C-terminal domain of PspC, it supports this as a PspA contact site in vivo. Finally, this cross-link was observed only in the ON state, which provides in vivo support for the model in which activation depends on the availability of a PspA-binding site within PspC.
We extended this analysis by testing the effect of the PspC V125D mutation. Consistent with previous data, incorporation of the V125D mutation into PspC-G131X abolished secretindependent induction of ⌽(pspA-lacZ) expression (Fig. 7A). Importantly, it also prevented the PspC-G131X-PspA crosslink (Fig. 7B). This corroborates the effect of the V125D mutation on the GST-PspC CT -PspA interaction in vitro (Fig. 2) and further demonstrates that a direct contact between PspC CT and PspA is essential for induction of the Psp response.
A Single Position within the C-terminal Domain of PspC Cross-links to PspB in the OFF State but to PspA in the ON State-Our in vitro experiments suggested that PspC CT interacted with PspB CT or PspA in a mutually exclusive manner (Fig.  4). This led to the hypothesis that the PspC CT ⅐PspB CT interaction might represent the OFF state, whereas the PspC CT ⅐PspA interaction is the ON state. In our final experiment we used photocross-linking to investigate if this predicted switch in the PspC CT interaction partner could be observed in vivo.
pBpa at position 131 in PspC cross-linked to PspA but not to PspB, making it unsuitable to test our hypothesis ( Fig. 7 and data not shown). Therefore, we constructed additional PspC CT pBpa substitution mutants and screened them for regulatory function and for photocross-linked complexes corresponding in size to both PspC⅐PspA and PspC⅐PspB (data not shown). pBpa at position 109 met these criteria. However, the amount of the PspC⅐PspB complex was not very different in OFF (ϪYscC) and ON (ϩYscC) conditions (data not shown). Although this could be interpreted to mean that our model is incorrect, we noticed a phenomenon that might make such a conclusion unsafe. pspBC were expressed from the lacZp promoter to maintain constant expression ϩ/ϪYscC. Despite this, the amount of PspB and PspC protein was higher in ϩYscC cells (e.g. Fig. 7Aii, and data not shown). We do not understand why, and this did not occur when the lacZp-pspBC plasmid was in Y. enterocolitica (data not shown). Regardless, a change in the total amount of PspBC complicates interpretation of any differences in the PspB⅐PspC complex level ϩ/ϪYscC. To circumvent this we took advantage of our previous discovery of PspB and PspC mutations that cause constitutive ⌽(pspA-lacZ) expression (33). Combining one of these mutations with PspC-Y109X should lock the PspBC regulatory complex into the ON state. Then, PspC-Y109X cross-linked complexes could be compared in the OFF state (i.e. without a constitutive mutation) to the ON state (i.e. with a constitutive mutation) but always without YscC so that PspBC protein levels would be similar in all strains. We chose the W23R mutation in PspB (33) so that the PspC protein would be identical in the strains being compared.
As expected, in cells with PspC-Y109X grown with pBpa, the PspB-W23R mutation increased ⌽(pspA-lacZ) expression compared with cells with wild type PspB (Fig. 8A). These strains were exposed to UV light, and cell lysates were analyzed by SDS-PAGE and immunoblot (Fig. 8B). In cells with PspC-Y109X and wild type PspB (OFF state), a pBpa/UV-dependent complex was recognized by anti-FLAG (PspC) and by anti-PspB antibodies, and it corresponded in size to PspC⅐PspB (ϳ24 kDa; Fig. 8B). However, in PspC-Y109X PspB-W23R cells (ON state), the PspC⅐PspB complex was reduced to a barely detectable level, whereas a pBpa/UV-dependent complex that corresponded in size to PspC⅐PspA was recognized by anti-FLAG (PspC) and by anti-His 6 (PspA) antibodies. Once again, we extended the experiment to test the effect of the PspC V125D mutation, which abolished both the PspC-Y109X-PspA and PspC-Y109X-PspB cross-links (Fig. 8B). This is completely consistent with the in vitro binding experiments (Figs. 2 and 3).
These experiments show that the same residue within PspC CT cross-links predominantly to PspB in the absence of a Psp-inducing signal but to PspA upon induction. The PspC CT ⅐PspB interaction might be part of an inhibitory mechanism that occludes the PspA-binding site in non-Psp-inducing conditions. Most significantly, these data provide the first demonstration that PspBC do exist in different OFF and ON states in vivo.

DISCUSSION
Activation of the Psp response requires PspA to interact with the PspBC complex at the cytoplasmic membrane (24,25). Therefore, we proposed that PspBC alternate between OFF and ON states, with only the latter able to bind PspA (26). Here, our findings from both in vitro and in vivo approaches have supported this model and revealed that regulation of the Psp response is centered around a PspC CT partnerswitching mechanism.
The GST-PspC CT membrane lysate pulldown experiments revealed that a PspC CT ⅐PspA complex could form in vitro even in the absence of PspB (Fig. 2). This suggests a direct interaction, but intriguingly we have found that purified PspA is no longer captured by GST-PspC CT . 3 The purification process might have inactivated its ability to bind to PspC CT , because purification is known to affect the physical state of PspA (42). Alternatively, PspA might require contact with membrane fragments in the lysate, in addition to PspC CT , because PspA can bind to phospholipid membranes directly (24,43). Regardless, we confirmed a direct PspC CT ⅐PspA interaction by in vivo photocross-linking (Fig. 7). This is the first demonstration of a direct contact between PspA and PspC. Other observations indicate that this interaction is essential to activate the Psp response. First, the V125D mutation prevents full-length PspC 3 J. Flores-Kim and A. J. Darwin, unpublished data. from inducing the Psp response in vivo, prevents the PspC CT ⅐PspA interaction in vitro, and also prevents PspC CT from cross-linking to PspA in vivo (Figs. 2 and 7-8 and Refs. 26 and 33). Second, overproduction of only the C-terminal domain of Y. enterocolitica PspC in vivo was sufficient to activate the Psp response in a concentration-dependent manner, and this was also prevented by the V125D mutation (Fig. 5). Consistent with our findings, it has also been reported that the C-terminal domain of E. coli PspC induced ⌽(pspA-lacZ) operon fusion expression but only if pspB was co-expressed from the same plasmid (27). It is not clear why pspB co-expression was required in that case.
We also detected a direct interaction between PspB CT and PspC CT in vitro as well as a photocross-link between PspC CT and PspB in vivo (Figs. 3 and 8). Therefore, PspC CT can interact directly with both PspA and PspB. These interactions appeared to be mutually exclusive in vitro (Fig. 4) and occurred in different induction states in vivo (Fig. 8). This supports a model where the OFF state of PspBC involves an interaction between their C-terminal domains, which might help to block the access of PspA to its binding site. In other words, the PspC CT ⅐PspB CT interaction might be an inhibitory complex. This means that negative regulation of the Psp system might involve two different inhibitory interactions, one between PspA and PspF and the other between PspB CT and PspC CT . Therefore, activation of the Psp system involves two partner switching mechanisms: PspA switching its partner from PspF to PspC and PspC CT switching its partner from PspB to PspA (Fig. 1B).
An alternative model for negative regulation has been proposed in which PspC is a bitopic membrane protein with its C-terminal domain in the periplasm in non-inducing conditions (27). An inducing signal flips the topology, positioning the C terminus in the cytoplasm to bind PspA. This is an exciting model, but so far several observations argue against it. First, analysis of Y. enterocolitica PspC showed that it is a polytopic membrane protein with both termini in the cytoplasm regardless of the Psp system induction status (26). Second, cysteine substitutions throughout the C-terminal domain of PspC did not form disulfide bonds without an oxidative catalyst, suggesting they were not in the oxidizing environment of the periplasm (33). Third, our finding that PspC CT photocross-links to PspB in non-inducing conditions suggests that it is not in the periplasm (Fig. 8).
PspC CT is predicted to form a conserved leucine zipper-like amphipathic helix (30,44). If it does, the V125D mutation would place a charged amino acid into its hydrophobic face. The fact that this mutation prevents the binding of both PspA and PspB suggests that the hydrophobic face is an important binding determinant for both proteins. An obvious possibility is that the hydrophobic face is a common binding site for PspA and PspB. However, another possibility is that the hydrophobic property is important for dimerization of adjacent PspC CT domains. Indeed, previous oxidative cross-linking suggested that the hydrophobic face could be a PspC dimerization interface (30). In this scenario, PspB or PspA might only be able to bind to PspC CT dimers, which the V125D mutation might disrupt. Interestingly, our in vivo cross-linking detected a complex that might be a PspC dimer (Figs. 7B and 8B). A band of the same size was also barely detectable when PspC did not contain the cross-likable pBpa (Fig. 7B). SDS-resistant PspC dimers have been reported in E. coli (45). Regardless, if the PspC-G131X complex we detected is a dimer, then dimerization of PspC CT might increase when the Psp system is induced (Fig.  7B). Whether or not this has any implications for the regulatory mechanism is not yet known.
Here, we have shown that one role played by PspB might be negative regulation, which is achieved by tethering PspC CT . However, PspB is also essential to activate the Psp system in response to most stimuli (e.g. Refs. 20, 46, and 47). A GST-PspB CT fusion protein does not pull PspA down from a membrane lysate in which PspA is the only core protein (45). 3 This suggests that PspB and PspA might not interact directly. The requirement of PspB for activation might have a trivial explanation because it protects PspC from degradation by the FtsH protease (48). However, it remains possible that PspB plays a more active role in inducing the Psp system.
Another intriguing area for future investigation is the role of the N-terminal cytoplasmic domain of PspC (PspC NT ). Deletion of PspC NT causes constitutive activation (33). This implies that PspC NT is required for negative regulation. Therefore, PspC NT could be part of the inhibitory complex containing PspB CT and PspC CT . We have attempted to investigate if PspC NT can interact with other Psp proteins. 3 However, attempts to overproduce GST-PspC NT , MBP-PspC NT , and PspC NT -MBP fusion proteins in vivo were unsuccessful. Only a PspC NT -GST fusion protein could be produced and purified, but it did not capture any core Psp proteins in pulldown assays. Of course, more experiments are needed to investigate if these negative data are meaningful. Nevertheless, if PspC NT does not form stable interactions with other Psp proteins, how could it be involved in negative regulation? One possibility is steric hindrance, with PspC NT preventing PspA from approaching the PspC CT ⅐PspB CT complex. An inducing signal might cause PspC NT to shift its position, allowing PspA to invade and displace PspB CT from PspC CT . This might explain why a random screen isolated numerous mutations throughout PspB and PspC (except for within PspC CT ) that caused constitutive activation (33). Perhaps any mutation that causes even a slightly aberrant conformation of PspB or PspC interferes with the precise relative positioning of PspC NT .
In summary, we have developed in vitro and in vivo approaches to provide important new insights into how the Psp response is controlled, including the first indication into how PspBC might be held in their OFF state. First, we have demonstrated a direct contact between PspA and PspC, which is essential for activation of the Psp response. Second, we have discovered a direct contact between the C-terminal domains of PspB and PspC. Third, our discoveries that these interactions are mutually exclusive in vitro and occur in different activation states in vivo provide the first description of OFF and ON states for the PspBC complex. Importantly, our findings have revealed that a PspC CT partner-switching mechanism controls the activation status of the Psp response.