Phage shock proteins B and C prevent lethal cytoplasmic membrane permeability in Yersinia enterocolitica

Summary

The bacterial phage shock protein (Psp) stress response system is activated by events affecting the cytoplasmic membrane. In response, Psp protein levels increase, including PspA, which has been implicated as the master effector of stress tolerance. Yersinia enterocolitica and related bacteria with a defective Psp system are highly sensitive to the mislocalization of pore-forming secretin proteins. However, why secretins are toxic to psp null strains, whereas some other Psp-inducers are not, has not been explained. Furthermore, previous work has led to the confounding and disputable suggestion that PspA is not involved in mitigating secretin toxicity. Here we have established a correlation between the amount of secretin toxicity in a psp null strain and the extent of cytoplasmic membrane permeability to large molecules. This leads to a morphological change resembling cells undergoing plasmolysis. Furthermore, using novel strains with dis-regulated Psp proteins has allowed us to obtain unequivocal evidence that PspA is not required for secretin-stress tolerance. Together, our data suggest that the mechanism by which secretin multimers kill psp null cells is by causing a profound defect in the cytoplasmic membrane permeability barrier. This allows lethal molecular exchange with the environment, which the PspB and PspC proteins can prevent.

Introduction

The human pathogen Yersinia enterocolitica employs various virulence factors, including two type III secretion systems (T3SSs; Cornelis et al., 1998, Revell and Miller, 2001). The virulence plasmid-encoded Ysc-Yop T3SS is essential for virulence due to its role in innate immune response inhibition (reviewed by Cornelis et al., 1998, Cornelis, 2002, Shao, 2008). The Ysa-Ysp T3SS is encoded on the chromosome of the highly pathogenic 1B Y. enterocolitica biogroup (Foultier et al., 2002). The precise role of the Ysa-Ysp system is unknown, but it has a small impact on virulence in mice (Haller et al., 2000, Matsumoto and Young, 2006).

Outer membrane components of all T3SSs are secretin proteins, which form a channel of 12-15 homomeric subunits arranged in a ring (reviewed by Korotkov et al., 2011). Secretins are also part of filamentous phage export, type IV pili, and type II secretion systems. In some cases “pilot” proteins assist secretin multimerization and/or insertion into the outer membrane (reviewed by Koo et al., 2012). An example of a secretin/pilot pair is the YscC/YscW proteins of the Ysc-Yop T3SS. The absence of YscW increases the likelihood of YscC secretin multimers inserting into the incorrect (cytoplasmic) membrane (Koster et al., 1997, Burghout et al., 2004).

Secretin mislocalization to the cytoplasmic membrane induces the phage shock protein (Psp) extracytoplasmic stress response (reviewed by Darwin, 2005, Joly et al., 2010, Yamaguchi and Darwin, 2012). Furthermore, in Y. enterocolitica, the Psp response is required for viability when the YscC secretin is produced, either alone or as part of the native Ysc-Yop T3SS (Darwin and Miller, 2001). As a result, the Y. enterocolitica Psp system is essential for virulence (Darwin and Miller, 1999, Darwin and Miller, 2001). The Psp system is present in many Gram-negative bacteria and has been studied extensively in Y. enterocolitica and E. coli. Homologs of some of its components are also found in Gram-positive bacteria, archaea and plants (Brissette et al., 1990, Darwin, 2005, Huvet et al., 2011). Other inducers of the Psp response in E. coli include heat, high osmolarity, ethanol, and proteins trapped in the cytoplasmic membrane translocation machinery (e.g. Brissette et al., 1990, Kleerebezem and Tommassen, 1993, Weiner and Model, 1994, Jones et al., 2003, Wang et al., 2010). From this list of inducers, it seems likely that the Psp system is activated in response to problems with the cytoplasmic membrane.

The Psp system is encoded by pspF-pspABCDycjXF and pspG in Y. enterocolitica and by pspF-pspABCDE and pspG in E. coli. Only pspF, -A, -B, and -C have been linked to robust phenotypes. PspF is a DNA-binding protein that activates the σ⁵⁴-dependent pspA and pspG promoters (Jovanovic et al., 1996, Green and Darwin, 2004). PspA inhibits PspF by binding to it in the cytoplasm (Dworkin et al., 2000, Elderkin et al., 2002, Yamaguchi et al., 2010). PspB and PspC are integral cytoplasmic membrane proteins hypothesized to detect an inducing stimulus and then sequester PspA (Adams et al., 2003, Yamaguchi et al., 2010). This frees PspF to activate its target promoters. This leads to a large increase in Psp protein concentrations, particularly PspA, which has been proposed to counter stress by maintaining the proton motive force (PMF; e.g. Kleerebezem et al., 1996, Kobayashi et al., 2007).

Many conditions that induce the Psp response also induce other systems, including the σ^E (RpoE) and Cpx extracytoplasmic stress responses. In contrast, secretin production only induces psp gene expression (Lloyd et al., 2004, Seo et al., 2007). This remarkably specific secretin-Psp relationship is further emphasized by the fact that in Y. enterocolitica only psp null mutants are specifically sensitive to secretin production (Seo et al., 2007). Furthermore, a random screen for overexpression inducers of the Y. enterocolitica Psp system found various secretins and just three other (non-secretin) cytoplasmic membrane proteins of unknown function (Maxson and Darwin, 2004). Interestingly, although all of these proteins induced the Psp response, and not the σ^E and Cpx responses, only the secretins were toxic to a psp null strain. To date there has been no explanation for this.

Another unresolved issue is the Psp protein(s) that prevent secretin toxicity. PspA is considered to be the master effector protein involved in countering Psp-inducing stress. This is based on several observations, including the wide conservation of PspA homologues (Huvet et al., 2011), the high level of PspA when the system is induced (Brissette et al., 1990), that deletion of pspA decreased PMF during production of improperly translocated PhoE (Kleerebezem et al., 1996), and that PspA prevented proton leakage from damaged membrane vesicles in vitro (Kobayashi et al., 2007). Confounding this, deletion of pspA does not cause sensitivity to secretin production in Y. enterocolitica (Darwin and Miller, 2001, Maxson and Darwin, 2006). This result suggests that PspA is not involved in countering secretin-induced stress. However, an important caveat to that conclusion is that deletion of pspA causes massive overexpression of the remaining Psp regulon. This might mask any physiological consequence of losing PspA, especially as it is known that overexpression of pspB and/or pspC promotes secretin-stress tolerance in a strain lacking all other psp genes (Maxson and Darwin, 2006).

Here, we have addressed some of these outstanding issues. What potentially lethal physiological defects correlate with secretin production in a psp null strain? Why do secretins kill a psp null strain whereas non-secretin inducers do not? Which Psp protein(s) promote secretin-stress tolerance, and was the surprising suggestion that PspA is not involved correct?

Results

A correlation between secretin production and the reduced growth yield in a psp null strain

Secretins inhibit the growth of Y. enterocolitica, E. coli and Salmonella enterica serovar Typhimurium strains with defective Psp systems (e.g. Seo et al., 2007, Seo et al., 2009). However, it has not been determined whether the magnitude of this growth defect correlates with the level of secretin production. We began by investigating this in Y. enterocolitica.

First, we used the YsaC secretin of the Ysa-Ysp T3SS produced from an arabinose-inducible expression plasmid. Growth of psp⁺ and Δpsp strains strain containing an empty vector plasmid, and of the psp⁺ strain with the plasmid encoding YsaC-His₆, was similar regardless of arabinose concentration (Fig. 1A). However, the growth yield of the Δpsp strain with YsaC-His₆ decreased as arabinose concentration increased. Immunoblot analysis confirmed more YsaC-His₆ production as arabinose concentration increased (Fig. 1B; note that all YsaC-His₆ multimers dissociated after heating prior to SDS-PAGE, data not shown). Experiments with the YscC secretin of the Ysc-Yop T3SS, produced from an IPTG-inducible (tacp) plasmid, led to similar conclusions (Fig. 1 panels C and D, which shows monomers and boiling-resistant multimers characteristic of YscC, e.g. Koster et al., 1997). These data provide a correlation between secretin level and the reduced growth yield of a Δpsp strain.

Fig. 1.

Correlation between secretin production and reduced growth yield of a psp null strain

A. YsaC-dependent growth inhibition depends on the amount of YsaC produced. psp⁺ (AJD3) and Δpsp (AJD1171) strains containing pBAD33 (− YsaC) or pAJD935 (+ YsaC) were grown with the indicated arabinose concentrations. Optical density was measured every hour.

B. Anti-His₆ (for YsaC-His₆) immunoblot analysis of cells grown as in panel A, harvested at the 3 h time point. Ponceau S stain of the YsaC region of the nitrocellulose membrane is shown as a loading control.

C. YscC-dependent growth inhibition depends on the amount of YscC produced but is alleviated by the pilot protein YscW. psp⁺ (AJD3) and Δpsp (AJD1171) strains containing pVLT35 (− YscC), pAJD126 (+ YscC), or pAJD136 (+ YscC YscW) were grown with the indicated amounts of IPTG. Optical density was measured every hour.

D. Anti-YscC immunoblot analysis of cells grown as in panel C, harvested at the 3 h time point. Ponceau S stain of the YscC monomer region of the nitrocellulose membrane is shown as a loading control. Multimers and monomers were detected on the same nitrocellulose membrane but the intervening region, which was blank, has been excised.

YscC has a known pilot protein, YscW (Burghout et al., 2004). Without YscW, fewer YscC multimers form, but the multimers that do form predominantly mislocalize to the cytoplasmic membrane (Koster et al., 1997, Burghout et al., 2004). In contrast, when YscW is present the more abundant YscC multimers localize mostly into the outer membrane (Burghout et al., 2004). We extended the YscC titration experiments by including a tacp-yscCyscW co-expression plasmid. YscW increased the abundance of YscC multimers but alleviated the reduced growth yield of the Δpsp strain (Fig. 1 panels C and D). This suggests that secretin mislocalization is the toxic event, which is consistent with work in E. coli (Guilvout et al., 2006).

Secretins decrease the membrane potential of a psp null strain whereas non-secretin Psp inducers do not

Next, we began our investigation into why secretins reduce the growth of a psp null strain, and why non-secretin Psp inducers do not (Maxson and Darwin, 2004). The Psp system has been proposed to maintain proton motive force (PMF; e.g. Kleerebezem et al., 1996, Kobayashi et al., 2007). The gold standard to monitor the membrane potential component of the PMF in E. coli K-12 uses uptake of the lipophilic cation tetraphenylphosphonium, which is proportional to the membrane potential (TPP⁺; Eisenbach, 1982). However, TPP⁺ was not taken up by Y. enterocolitica unless CCCP was added, and then uptake did not increase over time (data not shown). This is evidence for an energy dependent efflux pump that prevents TPP⁺ from accumulating in the cell (e.g. De Rossi et al., 1998). Therefore, we used E. coli as a surrogate host. First, we tested whether the Y. enterocolitica Psp inducers behaved similarly in E. coli. Production of the YsaC or YscC secretins, or the non-secretin YE0566 and AmpE proteins, induced Φ(pspA-lacZ) operon fusion expression in E. coli (Fig. 2A). However, only the secretins reduced the growth of a Δpsp strain (Fig. 2B). This is exactly how these proteins behave in Y. enterocolitica (Maxson and Darwin, 2004). Interestingly, YE0566 reduced the growth of the psp⁺ strain but not the Δpsp strain (Fig. 2B). The reason for this is not yet known.

Fig. 2.

Evidence that secretins decrease the membrane potential in an E. coli psp null strain but non-secretin Psp-inducers do not. E. coli psp⁺ (MC3) and Δ(pspF-pspE) (AJDE2419) strains contained pVLT35-based vectors encoding Y. enterocolitica Psp-inducers. For all panels IPTG was added to cultures at 200 μM for empty pVLT35 vector (−) and YscC, none for YsaC, 25 μM for AmpE, and 5 μM for YE0566.

A. Φ(pspA-lacZ) operon fusion expression in the psp⁺ strain. Cultures were grown as described in Experimental procedures. Error bars represent one positive standard deviation.

B. Growth of the psp⁺ and Δ(pspF-pspE) stains. Optical density was determined hourly. C. [³H]-TPP⁺ accumulation by whole cells grown as in panel A. Error bars represent one positive standard deviation. CPM = counts per minute.

Next we measured TPP⁺ uptake. psp⁺ and Δpsp strains containing an empty tacp expression plasmid accumulated [³H]-TPP⁺ similarly, indicating an intact membrane potential (Fig 2C). Treatment with CCCP abolished [³H]-TPP⁺ uptake, supporting its dependence on the membrane potential. Production of the YsaC or YscC secretins did not affect [³H]-TPP⁺ uptake in the psp⁺ strain, but almost completely abolished it in the Δpsp strain. In contrast, AmpE or YE0566 production did not abolish [³H]-TPP⁺ uptake. Therefore, reduced membrane potential offers a potential explanation of why secretins are toxic in a psp null strain but non-secretin inducers are not.

Secretin toxicity correlates with increased cytoplasmic membrane permeability to ONPG

In typical growth experiments like those in Figure 1 we have found that once secretin-induced growth inhibition is evident, psp null cells die rapidly (colonies cannot be recovered; data not shown). However, several studies indicate that bacteria (E. coli) can remain viable for some time in the absence of a PMF and, if glucose is present, even resume growth after an adaptive lag period (e.g. Kinoshita et al., 1984, Ohyama et al., 1992, Gage and Neidhardt, 1993). Consequently, we considered it unlikely that reduced PMF was the sole explanation for the rapid secretin-induced cell death of a Y. enterocolitica psp null strain. Therefore, we investigated the possibility that secretins might increase cytoplasmic membrane permeability to molecules much larger than protons. For this we adapted a method used to monitor cytoplasmic membrane permeabilization by antimicrobial peptides (e.g. Lehrer et al., 1989, Barker et al., 2000, Arcidiacono et al., 2009). This takes advantage of a permeabilized cytoplasmic membrane allowing ortho-nitrophenyl-β-galactoside (ONPG) to enter the cytoplasm, where it can be hydrolyzed by cytoplasmic β-galactosidase. To adapt this to Y. enterocolitica we introduced a constitutively expressed lacZ gene, but the LacY permease was absent.

A Δpsp strain with constitutively expressed lacZ and the tacp-yscC expression plasmid was grown with various amounts of IPTG as in Figure 1. A corresponding psp⁺ strain was grown with the highest concentration of IPTG (200 μM). All were harvested at the 4 h time point, incubated in phosphate buffered saline containing ONPG, and the absorbance at 420 nm was measured over time. When YscC was uninduced in the Δpsp strain (no IPTG), or was induced at low level (10 μM IPTG), there was no ONPG hydrolysis above the basal level (Fig. 3A). However, there was a small increase in ONPG hydrolysis in the sample that had been grown with 25 μM IPTG, and a substantial increase following growth with 200 μM IPTG. These results correlate perfectly with the effects of YscC on growth yield at the 4 h time point, with the no IPTG and 10 μM IPTG Δpsp cultures having no growth defect, the 25 μM culture having a minor defect and the 200 μM culture having a substantial defect (Fig. 1C). Furthermore, when these assays were extended to strains co-producing the YscW pilot, ONPG hydrolysis was essentially abolished, which correlates with YscW alleviating the growth defect (Fig. 1C). Finally, cells that were grown overproducing the non-secretin Psp inducers AmpE and YE0566 did not exhibit ONPG hydrolysis above the basal level (Fig. 3A). This is consistent with these proteins failing to inhibit the growth of a psp null strain compared to a psp⁺ strain (Maxson and Darwin, 2004).

Fig. 3.

The YscC secretin causes cytoplasmic membrane permeability to ONPG in a Y. enterocolitica psp null strain.

A. ONPG hydrolysis by whole cells of psp⁺ (AJD3) and Δpsp (AJD1171) strains containing Φ(catp-lacZ) plasmid pAJD2049 as well as tacp expression plasmids pAJD126 (YscC), pAJD136 (YscC YscW), pAJD633 (AmpE), or pAJD634 (YE0566). Strains had been grown with the indicated concentrations of IPTG for the YscC and YscC YscW panels, with 25 μM IPTG for the AmpE panel, or with 10 μM IPTG for the YE0566 panel. ONPG hydrolysis was monitored by increased absorbance at 420 nm. Readings were normalized so absorbance at 2 min was zero for all samples. Total β-galactosidase activities of cells permeabilized with SDS and chloroform were not significantly different, indicating no differences in β-galactosidase enzyme content (data not shown).

B. Φ(pspAp-lacZ) operon fusion expression in psp⁺ strain AJD977 containing either pVLT35 (−), pAJD126 (YscC), pAJD136 (YscCW), pAJD633 (AmpE), or pAJD634 (YE0566). Cultures were grown as described in Experimental procedures with 200 μM IPTG (-, YscC and YscCW), 25 μM IPTG (AmpE) or 10μM IPTG (YE0566). Error bars represent one positive standard deviation.

To confirm induction of the Psp response by the levels of YscC, AmpE and YE0566 used in these experiments, a separate set of psp⁺ strains with a Φ(pspA-lacZ) operon fusion was used. The strains were grown with the same concentrations of IPTG that had been used for the ONPG hydrolysis assays (only 200 μM IPTG was used for strains producing YscC +/− YscW). Consistent with published data (e.g. Darwin and Miller, 2001, Maxson and Darwin, 2004), YscC induced the pspA promoter weakly, YscC + YscW did not cause any significant induction, and AmpE and YE0566 caused strong induction (Fig. 3B). Therefore, even though AmpE and YE0566 were much stronger inducers of the Psp response than YscC, they did not cause cytoplasmic membrane permeability when the Psp system was defective.

ONPG might be hydrolyzed by β-galactosidase leaking from the cytoplasm (e.g. by cell lysis). Therefore, a mock ONPG hydrolysis assay was set up with Δpsp +YscC, 200 μM IPTG cells, but without adding ONPG. At the end of the incubation period bacteria were removed by filtration so that any extracellular β-galactosidase would be retained in the filtrate. However, when ONPG was added to the filtrate it was not hydrolyzed (data not shown). Therefore, although YscC permeabilized the Δpsp cytoplasmic membrane to ONPG, the membrane did not become permeable to the much larger β-galactosidase enzyme.

These results suggest that secretins are toxic to a Δpsp strain, but the non-secretin Psp inducers are not, because only secretins cause lethal cytoplasmic membrane permeability to molecules at least as large as ONPG (~ 300 Da).

YscC secretin production in a Δpsp strain causes a phenotype that resembles plasmolysis

Secretin production causes major cytoplasmic membrane permeability in a psp null strain. This prompted us to examine psp⁺ and Δpsp cells +/− tacp-yscC by transmission electron microscopy. The cells were harvested after three hours grown as in Figure 1C, when the growth defect of the psp null strain was first evident, but the culture OD was still increasing. When YscC was produced in the Δpsp strain some cells appeared to have a physical separation between layers of their cell envelope, which most frequently occurred at the pole (Fig. 4). In two independent experiments this phenotype was observed in approximately 13% of the cells (out of ~ 200) in the first experiment and 3% (out of ~ 350) in the second. The relatively low frequency is probably due to the fact that cells were harvested before growth had completely arrested. Significantly, we did not observe this phenotype in any psp⁺ or Δpsp cells without YscC production, or in any psp⁺ cells overproducing YscC. Furthermore, it was also not observed in any strain overproducing the non-secretin inducers AmpE and YE0566 (data not shown). Therefore, it is a phenotype specific to the psp null strain during secretin production. These sites of cell envelope separation resemble plasmolytic bays that have been observed at the pole of E. coli cells undergoing mild plasmolysis in hypertonic conditions (reviewed by Koch, 1998). It is linked to cytoplasmic shrinkage caused by loss of water, and so is consistent with increased cytoplasmic membrane permeability. In fact, a correlation between cytoplasmic membrane permeability and plasmolytic bays has also been established recently in E. coli (Reimann and Wolfe, 2011). Therefore, secretin production in a psp null strain correlates with growth inhibition, decreased membrane potential, increased cytoplasmic membrane permeability to large molecules and a plasmolysis-like phenotype.

Fig. 4.

Transmission electron microscopy of psp⁺ (AJD3) and Δpsp (AJD1171) cells containing pVLT35 (− YscC) or pAJD126 (+ YscC). Samples were grown for 3 hr with 200 μM IPTG. Arrows point to areas of apparent cell envelope layer separation that resembles a plasmolysis phenotype. Bars in the lower left hand corner of each panel represent 500 nm.

A secretin that cannot multimerize does not affect the growth of a psp null strain

Our experiments had suggested that secretin toxicity is the result of major cytoplasmic membrane permeability caused by secretin multimers. A recent insightful review article grouped secretins into five classes depending on their behavior in the absence of pilot or other accessory proteins (Koo et al., 2012). Secretins in classes 1-3 still form multimers whereas those in classes 4-5 do not. All secretins we have tested to date are toxic to a Y. enterocolitica psp null strain: YscC and YsaC from Y. enterocolitica, InvG from S. enterica serovar Typhimurium, XcpQ from P. aeruginosa and pIV from filamentous phage f1 (e.g. Seo et al., 2007, Seo et al., 2009 and this study). All of those secretins can multimerize in the absence of helper proteins (Seo et al., 2009, Koo et al., 2012 and literature cited within; data not shown for YsaC). Therefore, their toxicity to a psp null strain is consistent with the proposal that it is only secretin multimers that impact the Psp system (e.g. Guilvout et al., 2006). However, if this is true, we reasoned that a secretin unable to multimerize should not be toxic to a psp null strain.

To test this hypothesis we used PilQ from the P. aeruginosa type IV pilus, which is a class 5 secretin (Koo et al., 2012). Without its pilot protein, PilQ mislocalizes to the cytoplasmic membrane but does not form the heat/SDS-resistant multimers detected when its pilot protein is present (Koo et al., 2008). Therefore, PilQ produced in isolation in a Y. enterocolitica psp null mutant should not be toxic. This was tested by introducing a tacp-pilQhis₆ expression plasmid, or the empty vector control, into Y. enterocolitica psp⁺ and Δpsp strains. The strains were grown with concentrations of IPTG up to 1 mM and optical density was measured over time. Regardless of the IPTG concentration there was no evidence of PilQ-dependent toxicity in either strain (Fig. 5A and data not shown). Immunoblot analysis revealed robust PilQ production (Fig. 5B). However, all of the PilQ was in monomeric form, consistent with the situation when PilQ was produced in P. aeruginosa without its pilot protein (Koo et al., 2008). We also failed to detect any evidence of PilQ multimers after prolonged exposure of the immunoblots and regardless of whether samples were heated or not prior to SDS-PAGE (data not shown). Finally, we found that PilQ production also failed to induce Φ(pspA-lacZ) expression in a psp⁺ Y. enterocolitica strain (Fig. 5C).

Fig. 5.

A secretin that does not multimerize has no impact on the Psp system.

A. PilQ production is not toxic in Y. enterocolitica psp⁺ or psp null strains. psp⁺ (AJD3) and Δpsp (AJD1171) strains containing pVLT35 (− PilQ) or pAJD1806 (+ PilQ) were grown in medium containing 1 mM IPTG. Optical density was measured hourly. B. Anti-His₆ (for PilQ-His₆) immunoblot analysis of cells grown as in panel A and harvested at the 4 h time point. Ponceau S stain of the PilQ monomer region of the nitrocellulose membrane is shown as a loading control. The region corresponding to the top of the original 4-12% polyacrylamide gradient gel is indicated (any secretin multimers would be slightly below the top). C. PilQ does not induce the Psp response. Φ(pspA-lacZ) operon fusion expression in psp⁺ strain AJD977 containing pVLT35 (− PilQ) or pAJD1806 (+ PilQ) grown with 1mM IPTG. Error bars represent one positive standard deviation

These results are notable because PilQ is the first secretin protein we have found to be unable to induce the Psp response and to be non-toxic to a psp null strain. It is also the first secretin we have tested that does not form any multimers under the experimental conditions we used. Therefore, all of this is consistent with the contention that secretins are toxic to a psp strain when in their multimeric forms.

A physiologically relevant system where the regulatory link between Psp proteins is uncoupled

Next, we established a system to unequivocally identify which Psp protein(s) prevent multimeric secretin toxicity when present at a normal level. Deletion of pspA does not cause secretin-sensitivity (Darwin and Miller, 2001, Maxson and Darwin, 2006), suggesting that the proposed master effector PspA is not involved in secretin-tolerance. However, loss of PspA causes non-physiological overexpression of the remaining Psp regulon. In fact, because PspA, -B and -C all regulate pspA operon expression, removing any of them changes the levels of those remaining. Any phenotype could be due to loss of regulatory function, loss of physiological function, or both. Therefore, to carefully examine the contribution of Psp components to secretin tolerance we have established a chromosomal expression system where the proteins are produced at physiological levels, but without regulatory links between them. For this, a region encompassing the Y. enterocolitica pspF gene and the pspA promoter was replaced by a fragment encoding E. coli LacI repressor and the tac promoter (Fig. 6A). As a result, psp gene expression can be controlled by IPTG. We reported a preliminary use of this system to investigate PspC regulatory mutants (Yamaguchi et al., 2010) but here we characterized it in detail.

Fig. 6.

Characterization of a strain with a tac promoter-controlled pspA operon.

A. Diagrams of the wild type pspA locus (pspAp-pspA) and the derivative with the pspA operon expressed from the tac promoter (tacp-pspA). In the tacp-psp strain, the region encompassing pspF and the pspA promoter was replaced by the E. coli lacI^q gene and the tac promoter.

B. Growth of pspAp-pspA and tacp-pspA operon strains. A derivative of the pspAp-pspA strain with all psp genes deleted was included as a secretin-sensitivity control (Δpsp; strain AJD1171). Strains contained pBAD33 (− YsaC) or pAJD935 (+ YsaC) and were grown in the presence of 0.02% arabinose. 10 μM IPTG was included in all cultures except to one set of the tacp-pspA strains as indicated (− IPTG). Optical density was measured hourly.

C. Anti-PspA, -PspB and –PspC immunoblot analysis of samples taken at the 4 h time point from the cultures in panel B. Ponceau S stain of the PspA-PspC region of the nitrocellulose membrane is shown as a loading control.

D. ONPG hydrolysis by derivatives of the pspAp-pspA and tacp-pspA operon strains with a Φ(catp-lacZ) operon fusion on the chromosome. Strains contained either pBAD33 (− YsaC) or pAJD935 (+ YsaC). Arabinose was added to all cultures at 0.02% and 10 μM IPTG was also included where indicated. ONPG hydrolysis was monitored by increased absorbance at 420 nm. Readings were normalized so absorbance at 2 min was zero for all samples. Total β-galactosidase activities of cells permeabilized with SDS and chloroform were not significantly different, indicating no differences in β-galactosidase enzyme content (data not shown).

YsaC production reduced the growth yield of the tacp-psp strain when IPTG was omitted from the medium, but the inclusion of 10 μM IPTG prevented this (Fig. 6B). This is consistent with IPTG-dependent induction of the pspA operon, which was confirmed by immunoblot analysis of PspA, -B and -C protein levels (Fig. 6C). Importantly, the Psp protein levels induced by 10 μM IPTG were comparable to the YsaC-induced levels in the wild type strain. Therefore, the proteins were produced at a physiological level. We also used our ONPG hydrolysis assay to investigate YsaC-induced cytoplasmic membrane permeability. YsaC caused severe cytoplasmic membrane permeability in a complete psp deletion strain, but not in a wild type psp locus strain, which was similar to the effect of the YscC secretin (compare Figs. 3A and 6D). In the tacp-psp strain, YsaC caused cytoplasmic membrane permeability in cells grown without IPTG, but not in cells grown with 10 μM IPTG (Fig. 6D). Finally, these experiments also revealed that in the absence of IPTG, the reduced growth yield and cytoplasmic membrane permeability of the tacp-psp strain were not as severe as in a complete psp deletion strain (Fig. 6). This can be explained by low but detectable Psp protein levels in the tacp-psp strain without IPTG (Fig. 6C), probably due to so called leaky expression from the tac promoter.

PspB and PspC are both necessary to prevent secretin-induced toxicity

The tacp-psp strain produced physiological levels of PspA, -B and -C that allowed tolerance to secretin-induced stress (Fig. 6). We could now use this system to delete members of the pspA operon without affecting the expression of those remaining. This approach would allow a definitive determination of which Psp proteins are needed to prevent secretin-induced toxicity when produced at physiological levels. Therefore, derivatives of the tacp-psp strain were constructed with ΔpspA, ΔpspB, ΔpspC, ΔpspBC or Δ(pspD-ycjXF) in frame deletions.

Immunoblot analysis was used to test the effect of each deletion on the levels of PspA, −B, and −C during growth with 10 μM IPTG (Fig. 7). The Δ(pspD-ycjXF) deletion did not affect PspA, −B, or −C. With only one exception, the ΔpspA, ΔpspB, ΔpspC, ΔpspBC deletions abolished production of the protein(s) encoded by the deleted gene(s) but did not significantly affect any others (Fig. 7A). The exception was ΔpspB, which also abolished detection of PspC. This was expected, because PspB is required to protect PspC from FtsH-dependent degradation (Singh and Darwin, 2011).

Fig. 7.

Only PspB and PspC are necessary to prevent secretin-induced toxicity and cytoplasmic membrane permeability to ONPG.

A. Anti-PspA, -PspB, and -PspC immunoblot analysis of tacp-pspA strain derivatives with the indicated gene deletions. Samples were taken at the 4 h time point from the + YsaC cultures in panel B. Ponceau S staining of the PspC region of the nitrocellulose membrane is shown as a loading control.

B. Growth of tacp-pspA operon strain derivatives with the indicated gene deletions. Strains contained either pBAD33 (− YsaC) or pAJD935 (+ YsaC). All culture media contained 0.02% arabinose. 10 μM IPTG was also included in the medium as indicated above each graph. Optical density was measured hourly.

C. ONPG hydrolysis by tacp-pspA operon strain derivatives with a Φ(catp-lacZ) operon fusion on the chromosome. Strains contained pBAD33 (− YsaC) or pAJD935 (+ YsaC). All were grown with 0.02% arabinose and 10 μM IPTG. ONPG hydrolysis was monitored by increased absorbance at 420 nm. Readings were normalized so absorbance at 2 min was zero for all samples. Total β-galactosidase activities of cells permeabilized with SDS and chloroform were not significantly different, indicating no differences in β-galactosidase enzyme content (data not shown).

Next we determined the effect of the psp deletions on growth during YsaC secretin production (Fig. 7B). As expected, the Δ(pspD-ycjXF) deletion had no effect. Importantly, the same was true for ΔpspA, which demonstrates that a physiological level of PspA is not required for secretin stress tolerance. Furthermore, it shows that the lack of secretin sensitivity when pspA is deleted in a wild type strain (Darwin and Miller, 2001, Maxson and Darwin, 2006) cannot be explained by compensatory overproduction of PspBC, because this does not occur in the tacp-pspA operon strain (Fig. 7A). In contrast, the ΔpspB, ΔpspC and ΔpspBC deletions all caused secretin-sensitivity (Fig. 7), and comparison of their phenotypes suggested that both PspB and PspC contribute to secretin-stress tolerance. First, ΔpspC caused secretin sensitivity, which demonstrates a role for PspC because this mutation only abolishes PspC protein production. Second, the ΔpspB mutation caused a larger reduction in growth yield than ΔpspC, which indicates a contribution of PspB towards secretin-tolerance. Finally, results from ONPG hydrolysis assays correlated perfectly with the effects on growth yield, with only deletion of pspB and/or pspC causing detectable permeability to ONPG (Fig. 7C)

Taken together these data provide unequivocal evidence that of the six proteins encoded by the Y. enterocolitica pspA operon, only PspB and PspC are required to prevent the lethal cytoplasmic membrane permeability caused by secretin production.

Discussion

A Y. enterocolitica psp null mutant is avirulent in a mouse model of infection (Darwin and Miller, 1999). This phenotype correlates with bacterial cell death caused by mislocalization of the YscC secretin from the Ysc-Yop T3SS (Darwin and Miller, 2001). Secretin-sensitivity also occurs in E. coli and S. enterica psp null mutants (Seo et al., 2007, Seo et al., 2009). However, there has been no explanation for this secretin-sensitivity phenotype that might account for the rapid cell death. Here, we report that secretins probably kill psp null cells by causing a defect in the cytoplasmic membrane permeability barrier that is more profound than previously demonstrated. Our data also show unequivocally that PspA is not required to prevent this, whereas PspBC are.

In considering how a mislocalized secretin kills a psp null cell, it has long been hypothesized that the Psp system functions to maintain the PMF (Model et al., 1997, Darwin, 2005, Joly et al., 2010). The failure of Y. enterocolitica to accumulate the lipophilic cation TPP⁺ prevented us from monitoring its membrane potential. However, secretin production significantly reduced uptake of TPP⁺ in E. coli psp null cells (Fig. 2C), which is consistent with a reduced membrane potential, and with other studies (e.g. Guilvout et al., 2006). Even so, reduced PMF is unlikely to explain abrupt death of psp null cells. E. coli can survive for a prolonged period and even divide in some situations after the PMF has been destroyed (Kinoshita et al., 1984, Ohyama et al., 1992, Gage and Neidhardt, 1993). In fact, our own data revealed that secretin production was sufficient to abolish the ability of E. coli psp null cells to take up TPP⁺ at the 3 hour time point, suggesting a dissipated PMF (Fig. 2C). Nevertheless, these cells continued to divide long after 3 h (Fig. 2B).

Secretin production allowed externally added ONPG to be hydrolyzed by cytoplasmic β-galactosidase in Y. enterocolitica psp null cells (e.g. Fig. 3A). Therefore, a mislocalized secretin permeabilizes the cytoplasmic membrane to molecules much larger than protons and other small ions. Importantly, the rate at which cells hydrolyzed ONPG correlated perfectly with the magnitude of the growth yield reduction. First, overproduction of YsaC (Fig. 1A, 0.02% arabinose) reduced the growth yield of a psp null strain more than overproduction of YscC (Fig. 1C, 200 μM IPTG) and also led to a faster rate of ONPG hydrolysis (compare Figs. 3A and 6D, note the different y-axis scales). Second, increasing the level of YscC production correlated with increasingly reduced growth yield and an increasing rate of ONPG-hydrolysis (Figs. 1C and 3A). Third, YsaC production reduced the growth yield of a psp null strain more than the tacp-pspA operon strain without IPTG, and the former strain also had an increased rate of ONPG-hydrolysis (Fig. 6). Finally, using the tacp-pspA operon strain, the relative effect of pspB and/or pspC deletions on growth yield correlated exactly with their relative rate of ONPG hydrolysis (Fig. 7). Therefore, we suggest that secretins kill psp null cells by permeabilizing the cytoplasmic membrane and allowing exchange of molecules at least as large as ONPG.

We reported that overproduction of secretins and some non-secretin cytoplasmic membrane proteins (e.g. AmpE and YE0566) induced Y. enterocolitica pspA operon expression (Maxson and Darwin, 2004). However, in a psp null strain, secretins were toxic whereas AmpE and YE0566 were not. This can now be explained because only secretins permeabilize the cytoplasmic membrane of a psp null strain (Fig. 3A). Still unexplained is what secretin, AmpE and YE0566 production have in common that generates a Psp-inducing signal. It has been hypothesized that the Psp system is induced by dissipated PMF (Model et al., 1997, Darwin, 2005, Joly et al., 2010). This is unlikely because the PMF does not dissipate unless the Psp system is defective (Kleerebezem et al., 1996, Guilvout et al., 2006, Jovanovic et al., 2006, Engl et al., 2009, Jovanovic et al., 2009, Jovanovic et al., 2010). We also found this to be true for secretins, which induced Φ(pspA-lacZ) expression in psp⁺ cells without affecting TPP⁺ uptake (Fig. 2). Furthermore, AmpE or YE0566 production did not abolish TPP⁺ uptake in psp⁺ or psp null cells (Fig. 2). It is possible that all of these Psp inducers reduce the PMF in psp⁺ cells but the difference is below the detection limit. However, a recent study also showed that reduced PMF is not sufficient to induce E. coli psp gene expression (Engl et al., 2011).

Perhaps the Psp system responds to increased cytoplasmic membrane permeability, rather than specifically to reduced PMF. However, secretins, AmpE and YE0566 all induced Φ(pspA-lacZ) expression in psp⁺ cells but did not increase cytoplasmic membrane permeability to ONPG (e.g. Figs. 3A and 6D). This suggests that a Psp-inducing signal is generated in the absence of increased cytoplasmic membrane permeability to ONPG. Nevertheless, it remains possible that some altered physical property of the cytoplasmic membrane is sufficient to generate the Psp-inducing stimulus.

Another important part of this work was the question of whether PspA is involved in mitigating secretin toxicity. Formally, we have shown that it is not required because a pspA in frame deletion mutant is not secretin sensitive (Darwin and Miller, 2001, Maxson and Darwin, 2006). However, this does not rule out its involvement because the remaining psp genes are massively overexpressed without PspA, which might have provided non-physiological compensation for its loss. We engineered Y. enterocolitica to produce Psp proteins at near-physiological levels but no longer regulated by the Psp proteins themselves. This allowed us to show unequivocally that only PspB and PspC play a role in secretin stress tolerance. The removal of PspA had no effect, even though other psp genes could not be overexpressed to compensate. We still think that PspA is involved in countering adverse effects associated with activation of the Psp response. In particular, the genetics of one study convincingly linked PspA to maintenance of the PMF during production of a translocation-defective PhoE protein (Kleerebezem et al., 1996). Perhaps PspA counters mild membrane permeabilization that might lead to leakage of small ions and/or as yet undetected defects caused by many Psp inducers, including YE0566 and AmpE, but secretins cause much more severe damage that demands the function(s) of PspB and PspC to counteract.

An obvious possibility is that cytoplasmic membrane permeability is caused by the passage of molecules through the mislocalized multimeric secretin channel itself. However, it is also possible that secretin multimers disrupt the integrity of the phospholipid bilayer around them, allowing passage of molecules through the membrane rather than through the secretin channel. Regardless of how secretins permeabilize the cytoplasmic membrane, a focus of future studies will be to understand how PspB and PspC prevent it. We have not found any evidence to support a model that Psp proteins simply inhibit secretin mislocalization into the cytoplasmic membrane. Differential detergent solubilization experiments did not reveal any difference in the distribution of YscC between inner and outer membrane fractions in psp⁺ and Δpsp cells (N.K. Horstman and A.J. Darwin, unpubl. data). However, the severe toxicity of secretins to the Δpsp strain makes these fractionation experiments difficult and it is possible that changes in cell envelope properties could impact the fractionation procedure. Nevertheless, this is consistent with mislocalization of PulD into the E. coli cytoplasmic membrane in the presence of a functional Psp system (Guilvout et al., 2006). Also, a pspA insertion mutation did not affect pullulanase secretion in E. coli with a plasmid encoding the Klebsiella oxytoca Pul type 2 secretion system, suggesting that PulD secretin function/localization was similar with or without an intact pspA operon (Possot et al., 1992). Similarly, a pspC null mutation did not affect Yop export by the Y. enterocolitica Ysc-Yop T3SS, suggesting that YscC secretin function/localization is not affected by the Psp system (Darwin and Miller, 1999). Other possibilities include PspBC preventing leakage through the secretin channel or physically modifying membrane properties to prevent permeability.

Experimental procedures

Bacterial strains, plasmids, and routine growth conditions

Bacterial strains and plasmids used in this study are listed in Table 1. E. coli was grown at 37 °C and Y. enterocolitica at 26°C or 37°C as noted. Strains were grown in Luria-Bertani (LB) broth or on LB agar plates. Antibiotics were used as described previously (Maxson and Darwin, 2004) and tetracycline was used at 15 μg ml⁻¹ for E. coli and 7.5 μg ml⁻¹ for Y. enterocolitica.

Table 1.

Strains and plasmids

Name	Genotype/Features	Reference or Source
E. coli K-12 strain
MC3	F− araD139δ(argF-lac) U169 relA1 rpsL150 flbB5301 depC1 ptsF25 rbsR λ[Φ(pspAp-lacZY)]	Bergler et al., 1994
AJDE2419	MC3 Δ(pspF-E)::kanr	This study
Y. enterocolitica strains
AJD3 a	ΔyenR (R−M+) Nalr	Laboratory collection
AJD977	ΔyenR (R−M+) ΔaraGFB::[Φ(pspAp-lacZY)]	Maxson and Darwin, 2005
AJD1171	ΔyenR (R−M+) Δ(pspF-ycjF) ΔpspG	Maxson and Darwin, 2006
AJD3298	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp]	Yamaguchi et al., 2010
AJD3463	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔpspB	This study
AJD3465	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔpspC	This study
AJD3469	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔpspBC	This study
AJD3494	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] Δ(pspD-ycjF)	This study
AJD3497	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔpspA	This study
AJD4250	ΔyenR (R−M+) ΔaraGFB::[Φ(catp-lacZ)]	This study
AJD4251	ΔyenR (R−M+) Δ(pspF-ycjF) ΔpspG ΔaraGFB::[Φ(catp-lacZ)]	This study
AJD4505	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔaraGFB::[Φ(catp-lacZ)]	This study
AJD4507	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔaraGFB::[Φ(catp-lacZ)] ΔpspB	This study
AJD4508	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔaraGFB::[Φ(catp-lacZ)] ΔpspC	This study
AJD4510	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔaraGFB::[Φ(catp-lacZ)] ΔpspBC	This study
AJD4511	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔaraGFB::[Φ(catp-lacZ)] δ(pspD-ycjF)	This study
AJD4512	ΔyenR (R−M+) ΔpspF ΔpspAp::[lacIq-tacp] ΔaraGFB::[Φ(catp-lacZ)] ΔpspA	This study
Plasmids
pACYC184	Cmr Tetr, p15A ori	Chang and Cohen, 1978
pBAD33	Cmr, p15A ori, araBp expression vector	Guzman et al., 1995
pRE112	Cmr, R6K ori, mob+ (RP4), sacB+	Edwards et al., 1998
pSR47S	Kmr, R6K ori, mob+ (RP4), sacB+	Merriam et al., 1997
pVLT35	Smr, Spr, RSF1010 ori, tacp expression	vector de Lorenzo et al., 1993
pAJD126	tacp-yscC in pVLT35	Darwin and Miller, 2001
pAJD136	tacp-yscCyscW in pVLT35	Darwin and Miller, 2001
pAJD555	tacp-ysaC in pVLT35	Maxson and Darwin, 2004
pAJD633	tacp-ampE in pVLT35	Maxson and Darwin, 2004
pAJD634	tacp-YE0566 in pVLT35	Maxson and Darwin, 2004
pAJD935	araBp-ysaC-his6 in pBAD33	This study
pAJD1806	tacp-pilQ-his6 in pVLT35	This study
pAJD2049	Φ(catp-lacZ) in pACYC184	This study

^aAJD3 is a virulence plasmid cured derivative of strain JB580v (Kinder et al., 1993). All other Y. enterocolitica strains listed are derivatives of AJD3.

Strain constructions

The Δ(pspF-pspE)::kan mutation (Maxson and Darwin, 2006) was introduced into E. coli strain MC3 by phage P1 vir transduction.

The pspF-pspAp region of the Y. enterocolitica chromosome was replaced with a lacIq- tacp fragment using the sacB⁺ allelic exchange plasmid and strategy described previously (Yamaguchi et al., 2010). The same allelic exchange plasmid was used to replace the pspA promoter with the tac promoter in psp⁺ ΔpspB, ΔpspC, ΔpspBC and Δ(pspD-ycjF) strains. However, to do this in a ΔpspA strain, a new pSR47S-based allelic exchange plasmid was constructed as described (Yamaguchi et al., 2010), except that the downstream fragment flanking the pspA promoter region was derived from a ΔpspA strain.

For the ONPG hydrolysis assays, a promoterless E. coli lacZ gene was amplified from plasmid pAJD905 (Maxson and Darwin, 2005) by PCR with primers that incorporated EagI and NotI sites. This fragment was cloned into sacB⁺ suicide plasmid pAJD898 (Maxson and Darwin, 2005) at the unique NotI site, which is flanked by fragments surrounding araGFB on the Y. enterocolitica chromosome. Then the cat promoter (catp) region was amplified from plasmid pACYC184 with primers that incorporated XbaI and BglII restriction sites and cloned into these sites upstream of lacZ so that catp controlled lacZ expression. Standard sacB⁺ allelic exchange procedures were used to introduce this Φ(catp-lacZ) operon fusion into the ara locus of the Y. enterocolitica chromosome. This was confirmed by colony PCR analysis, as were all of the above-described chromosomal alterations.

Plasmid constructions

For cytoplasmic membrane permeability assays with strains that did not contain an araBp expression plasmid, a plasmid-encoded Φ(catp-lacZ) operon fusion was used (combining this plasmid with an araBp plasmid in the same strain led to toxicity). To make this plasmid, promoterless lacZ was amplified from pAJD905 with a primer that incorporated a BamHI site downstream of lacZ. The fragment was digested with EcoRI and BamHI and ligated between the EcoRI and BclI sites of pACYC184, which placed lacZ downstream of the cat promoter. The araBp-ysaC-his₆ (pAJD935) plasmid was constructed by first transferring the EcoRI-BamHI insert of pAJD801 (Maxson and Darwin, 2004) into plasmid pWSK129 (Wang and Kushner, 1991), then excising it from pWSK129 as a KpnI-XbaI fragment and cloning it into pBAD33. The tacp-pilQ-his₆ plasmid (pAJD1806) was constructed by amplifying the promoterless pilQ (PA5040) gene from the chromosome of P. aeruginosa strain PAK (Strom and Lory, 1986) with an upstream primer that incorporated an EcoRI site and a downstream primer that incorporated a His₆-encoding region immediately before the stop codon and an XbaI site downstream of the stop codon. The EcoRI-XbaI fragment was ligated between the same sites of plasmid pVLT35.

Growth curves

Saturated Y. enterocolitica cultures that had been grown at 26°C were diluted into fresh LB broth in 18 mm diameter test tubes to an optical density (600 nm) of 0.1. Arabinose and/or IPTG were included at concentrations indicated in the figures. Growth on a roller drum at 37°C was monitored hourly by removing a 0.1 ml aliquot and determining optical density (600 nm). Growth curves of E. coli strains were done similarly except that cultures were diluted to an initial optical density (600 nm) of 0.04. All growth curves were done on two or more occasions to ensure reproducibility, but the figures report data from a single experiment.

β-galactosidase assays

For Y. enterocolitica, saturated cultures that had been grown at 26°C were diluted into 5 ml of fresh medium in 18 mm diameter test tubes to an optical density (600 nm) of 0.04. For E. coli, saturated cultures grown at 37°C were diluted into 4 ml fresh media to an optical density (600nm) of 0.02. For all, after 2 hours of growth in a roller drum at 37°C, IPTG and/or arabinose was added to induce tac or araB promoter expression unless otherwise indicated in figure legends. Growth continued at 37°C for two more hours prior to harvest. β-Galactosidase activity was determined at room temperature as described previously (Maloy et al., 1996) and reported in arbitrary Miller units (Miller, 1972). The mean of the data from three independently grown cultures, each of which was assayed in duplicate, is reported.

Polyclonal antisera and immunoblotting

The region encoding the C-terminus of YscC (amino acids 321-607) was cloned into pET-32a(+) (Novagen) to encode a TrxA-His₆-’YscC-His₆ fusion protein (TrxA = thioredoxin). The plasmid was transferred into E. coli strain ER2566 (NEB) and the strain was grown to mid-exponential phase at 37°C in LB broth containing 1 mM IPTG. Total cell lysates were prepared and the His₆-tagged protein purified under denaturing conditions using nickel-nitrilotriacetic acid (Ni-NTA) affinity chromatography as recommended by the manufacturer (Qiagen). Polyclonal rabbit antiserum was raised against the purified TrxA-His₆-’YscC-His₆ fusion protein at Covance Research Products.

For immunoblots, cells were resuspended at equivalent optical densities in SDS sample buffer and heated at 90-100°C for 10 min. Proteins were separated by 10% polyacrylamide SDS-PAGE (for detection of secretin monomers only, or Psp proteins) or by 4-12% gradient polyacrylamide SDS-PAGE (Lonza; for simultaneous detection of secretin monomers and multimers) and transferred to nitrocellulose by semidry electroblotting. The nitrocellulose was stained with Ponceau S and then destained prior to immunoblotting. Enhanced chemiluminescent detection followed sequential incubation in diluted polyclonal antisera or monoclonal antibody, followed by a 1 in 5000 dilution of goat α-rabbit IgG or goat α-mouse IgG horseradish peroxidase conjugate (Bio-Rad). Dilutions of polyclonal antisera or monoclonal antibody were 1 in 5000 for His₆ antibody (GenScript), 1 in 20,000 for PspA antiserum (Yamaguchi et al., 2010), 1 in 100,000 for PspB antiserum (Gueguen et al., 2009), 1 in 20,000 for PspC antiserum (Maxson and Darwin, 2006) and 1 in 10,000 for YscC antiserum. For detection of secretin multimers, after SDS-PAGE the gels were soaked in 10% trichloroacetic acid and washed prior to semidry electroblotting as described previously (Burghout et al., 2004).

[³H]-tetraphenylphosphonium uptake assay

E. coli strains were grown as for β-galactosidase assays. Bacterial cells equivalent to 0.6 ml at an optical density (600 nm) of 2 were collected by centrifugation at room temperature, resuspended in 0.75 ml 100 mM Tris (pH 7.6) 1mM EDTA, and incubated at 37°C for 10 min to permeabilize the outer membrane. Cells were collected by centrifugation and resuspended in 0.6 ml 0.1 M sodium phosphate buffer (pH 7.6) containing 1 mM KCl and 10 mM glucose. Carbonyl cyanide m-chlorophenyl hydrazone (CCCP) in ethanol was added to aliquots of the empty pVLT35 samples to a final concentration of 80 μM and all other samples were treated with an equivalent volume of ethanol. After incubating at 37°C for 5 min, 1.25 μl of a 1:10 dilution of 1mCi/ml [³H]-tetraphenylphosphonium bromide (TPP; American Radiolabeled Chemicals) and an equivalent volume of non-radioactive TPP were added so the total TPP concentration was 2 mM. After a further 10 min at 37°C, samples were captured onto GF/F filters (Whatman) using a vacuum manifold (Millipore) and washed twice with sodium phosphate buffer. Filters were air-dried and [³H] was detected by liquid scintillation (Beckman Coulter). The mean of the data from three independently grown cultures is reported.

Cytoplasmic membrane permeability assay

The method to detect cytoplasmic membrane permeability to ONPG was adapted from a published procedure (Arcidiacono et al., 2009). Bacterial cells equivalent to 1 ml at an optical density (600 nm) of 0.5 were collected by centrifugation at room temperature and washed twice with 1 ml phosphate buffer saline (PBS; 0.01 M sodium phosphate, 0.15 M NaCl, pH 7.1). Cells were resuspended in 1 ml PBS and diluted 1:10 in PBS containing 2 mM ONPG. Absorbance at 420 nm was measured every 2 minutes for 3 hours using the kinetics program of a Beckman Coulter DU® 730 spectrophotometer. The blank was PBS with 2 mM ONPG. To control for changes in β-galactosidase enzyme production, the total β-galactosidase of cells permeabilized with SDS and chloroform (Miller, 1972) was determined in all experiments, but no significant differences were detected (data not shown).

Transmission Electron Microscopy

Cultures were grown at 37°C with 200 μM IPTG for 3 h and collected by centrifugation. Bacteria were fixed with 2.5% glutaraldehyde and 2% paraformaldehyde, washed with cacodylate buffer (50 mM, pH 7.2), then fixed with 2% osmium tetroxide and washed with Kellenberger buffer. Pellets were embedded in 2% agar, cut and stained in the dark with 0.5% (w/v) uranyl acetate. Samples were dehydrated with alcohol, transferred to propylene oxide/Epon mixtures and finally embedded in EMbed 812 (Electron Microscopy Sciences, Hatfield, PA). Thin sections were cut, adsorbed on electron microscope grids, and stained with uranyl acetate and lead citrate. Stained grids were imaged in a Philips CM12 electron microscope (FEI; Eindhoven, the Netherlands) and photographed with a Gatan (4kx2.7k) digital camera (Gatan Inc., Pleasanton, CA). Sample fixation, preparation, and microscopy were performed by the Microscopy Core at the New York University Langone Medical Center (New York, NY).