Pseudomonas aeruginosa uses a type III secretion system to inject protein effectors into a targeted host cell. Effector secretion is triggered by host cell contact. How effector secretion is prevented prior to cell contact is not well understood. In all secretion systems studied to date, the needle tip protein is required for controlling effector secretion, but the mechanism by which needle tip proteins control effector secretion is unclear. Here we present data that the P. aeruginosa needle tip protein, PcrV, controls effector secretion by assembling into a functional needle tip complex. PcrV likely does not simply obstruct the secretion channel because the pore-forming translocator proteins can still be secreted while effector secretion is repressed. This finding suggests that PcrV controls effector secretion by affecting the conformation of the apparatus, shifting it from the default, effector secretion ‘on’ conformation, to the effector secretion ‘off’ conformation. We also present evidence that PcrG, which can bind to PcrV and is also involved in controlling effector export, is cytoplasmic and that the interaction between PcrG and PcrV is not required for effector secretion control by either protein. Taken together, these data allow us to propose a working model for control of effector secretion by PcrG and PcrV.
The observation that deletion of the Shigella flexneri needle tip protein gene ipaD, or the gene for the pore-forming translocator IpaB, results in deregulation of effector secretion led to the first model of effector secretion control, the ‘plug’ model (Menard et al., 1993). Here it was proposed that IpaD and IpaB form a structure at the tip of the type III secretion needle, which physically blocks secretion via the needle. Cell contact would then result in a conformational change in the needle tip that opens the secretion channel. Consistent with this model, deletion of the export signal of IpaD results in a non-secreted protein that also fails to control effector export (Picking et al., 2005). Some support for the plug hypothesis also comes from modelling of the needle tip based on the structure of an IpaD dimer found in one of the crystals used to solve the IpaD structure. The model predicts that five IpaD monomers assemble into a needle tip structure that only has a small aperture at its centre, much too small to allow the passage of proteins (Johnson et al., 2007). An earlier model in which the structure of LcrV had been inserted at the top of a T3SS needle resulted in a needle tip with a large aperture (Deane et al., 2006). These two models have been interpreted to represent ‘open’ and ‘closed’ versions of the needle tip (Deane et al., 2006; Blocker et al., 2008). Recent evidence suggests that the S. flexneri needle tip may in fact consist of four IpaD molecules and one molecule of the pore-forming translocator protein IpaB (Johnson et al., 2007; Veenendaal et al., 2007). While some have only found IpaB associated with the needle tip after treating bacteria with agents such as bile salts (Olive et al., 2007; Stensrud et al., 2008), it has been argued that IpaB may be queued within the needle and associated with the needle tip even in the absence of these treatments and that bile-salts simply reconfigure the needle tip to display a larger portion of IpaB (Espina et al., 2006). The great attraction of this model is that it immediately explains why in Shigella deletion of either ipaB or ipaD results in upregulation of effector secretion. It also results in one of the clearest models for triggering of effector secretion on cell contact, where insertion of needle tip-associated IpaB into the host cell membrane results in the conformational change that opens the needle tip and allows secretion of the second pore-forming translocator, IpaC, as well as effectors to commence (Veenendaal et al., 2007). It is unclear, however, how far this model can be applied to other T3SSs. For one, no pore-forming translocators have been detected associated with the needle tip of Salmonella typhimurium, Yersinia enterocolitica or P. aeruginosa prior to cell contact (Mueller et al., 2005; Broz et al., 2007; Lara-Tejero and Galan, 2009; Rietsch, A., unpublished observation). Moreover, deletion of the pore-forming translocator proteins in P. aeruginosa or S. typhimurium does not result in loss of effector secretion control (Kaniga et al., 1995b; Broms et al., 2003; Cisz et al., 2008).
Another variation of the plug model has been proposed, where a ‘sensor’ protein has been inserted into the type III secretion channel (Blocker et al., 2008). Its export is inhibited by the type III secretion needle tip and its C-terminal portion binds to a part of the cytoplasmic face of the T3SS, effectively occluding an acceptor site for effector proteins. Once the T3SS has encountered the host cell, a conformational change in the needle tip permits export of the sensor protein, which frees an acceptor site for effectors and results in effector export. A possible candidate for such a sensor molecule are proteins of the YopN/MxiC family, which are required for preventing export of effector proteins prior to cell contact (Forsberg et al., 1991; Sundin et al., 2004; Botteaux et al., 2009). The best studied of these proteins is the Yersinia pestis YopN protein. YopN is thought to be tethered to the base of the apparatus via a C-terminal interaction with the small regulatory protein TyeA (Cheng and Schneewind, 2000; Day et al., 2003; Sundberg and Forsberg, 2003). Because YopN requires an intact type III secretion signal, as well as its export chaperone, to control effector secretion, it has been proposed that YopN is partially inserted into the T3SS before effector secretion is triggered (Day and Plano, 1998; Ferracci et al., 2005). The analogy here is the ruler model for type III secretion needle-length control, where YscP has been proposed to connect the growing needle tip and the base of the apparatus (Journet et al., 2003; Wagner et al., 2009). One problem with the plug/sensor models, however, is that growing evidence from a variety of organisms suggests that translocator proteins can be exported prior to cell contact, before effector secretion has been triggered (Lee et al., 2001; Olive et al., 2007; Cisz et al., 2008). In the case of the YscP ruler model, it has been proposed that the secretion channel is wide enough to accommodate two proteins in an extended conformation (Wagner et al., 2009). This line of reasoning could also be extended to the sensor model. Here the sensor would be inserted into the secretion channel in an extended conformation, allowing the simultaneous export of translocator proteins.
The third, ‘allosteric’, model of effector secretion control also circumvents the problem of translocator secretion in the absence of effector export. In this model the needle tip controls effector secretion by affecting a conformational change in the apparatus (Kenjale et al., 2005; Torruellas et al., 2005; Blocker et al., 2008). Here, cell contact results in a conformational change in the needle tip, which in turn results in a change in needle conformation that propagates down the needle and changes the conformation of the base of the apparatus to permit effector export. In support of this model, needle-protein mutations have been isolated both in Shigella and Yersinia, that result in deregulated effector secretion (Kenjale et al., 2005; Torruellas et al., 2005; Davis and Mecsas, 2007). While some of these mutations disrupt the assembly of the needle tip, and are therefore compatible with all of the proposed models, others still allow assembly of the needle tip, suggesting that the mutations lock the apparatus in the effector secretion ‘on’ conformation (Kenjale et al., 2005; Zhang et al., 2007). Analysis of the helical pitch and axial rise of the needle-subunits in needles isolated from wild-type Shigella or locked ‘on’ mutants (that either still responded to Congo red or were Congo red-insensitive), however, could not discern a difference between the wild-type needles or the mutant needles (Cordes et al., 2005). These data were taken to mean that triggering of effector secretion cannot be associated with a large-scale conformational change of the needle, akin to the change in flagellar packing that has been detected upon changing the direction of rotation (Cordes et al., 2005; Blocker et al., 2008; Mueller et al., 2008). Instead, it was proposed that the signal is propagated by subtle and/or rapid changes in the interactions between head and tail portions of adjacent needle-subunits (Deane et al., 2006).
Among the well-studied T3SSs, the Yersinia sp. LcrV protein is the needle tip protein that is most closely related to P. aeruginosa PcrV (Troisfontaines and Cornelis, 2005). However, as mentioned above, LcrV represents the curious exception to how needle tip proteins control effector secretion. Unlike other needle tip proteins, the deletion of lcrV results in a defect in effector secretion. This is not the case for P. aeruginosa PcrV (Mccaw et al., 2002; Rietsch et al., 2005). Yersinia LcrV interacts with a small export chaperone, LcrG, that is also required for effector secretion control (Nilles et al., 1997; Debord et al., 2001; Matson and Nilles, 2001; 2002; Lawton et al., 2002). Deletion of lcrG results in constitutive effector secretion. LcrG is cytoplasmic in all species of Yersinia, except Y. pestis, where a portion of it appears to be exported (Debord et al., 2001; Matson and Nilles, 2001; Reina et al., 2008). However, fusions that prevent export do not abolish its activity, suggesting that it controls effector secretion in the cytoplasm in Y. pestis as well (Reina et al., 2008). Mutations that prevent the interaction between LcrV and LcrG prevent upregulation of effector secretion (Matson and Nilles, 2001; Hamad and Nilles, 2007) and overexpression of LcrV stimulates effector secretion (Lee et al., 2000). Based on these data, it was proposed that LcrG blocks effector secretion at the cytoplasmic face of the T3SS. Upon cell contact, LcrV expression is upregulated and titrates LcrG from the apparatus, thereby negating the block imposed by LcrG (titration model). While this model explains some of the mutant data, differential upregulation of LcrV over LcrG (lcrG and lcrV are part of the lcrGVHyopBD operon and are coordinately expressed) has, to our knowledge, never been demonstrated. Moreover, effector secretion in Yersinia pseudotuberculosis can be triggered in the presence of chloramphenicol, suggesting that no de novo protein synthesis is needed to upregulate effector export (Lloyd et al., 2001).
Given the differences in effector secretion control by PcrV when compared with its nearest relative, LcrV, as well as the wealth of competing models of effector secretion control, we decided to examine the role of PcrV and PcrG in control of effector secretion by P. aeruginosa. We confirmed that deletion of pcrG or pcrV results in effector secretion even in the absence of cell contact. Moreover, the deregulation caused by the individual deletion mutations appears to be additive. Next we determined that PcrV export is required in order to control effector secretion. Specific mutants of PcrV and PcrG allowed us to determine that the interaction of PcrG and PcrV is not required for regulation of effector secretion, but rather that effector secretion control by PcrV is tied to its ability to form a functional needle tip complex. A physical block of the secretion apparatus by PcrV seems unlikely, because translocator proteins are secreted prior to cell contact even in wild-type bacteria. PcrG, while required for efficient export of PcrV, controls effector secretion from the cytoplasm in a PcrV-independent manner. Taken together these results are consistent with a model in which PcrV acts as an allosteric regulator that helps shift the needle complex from its default, effector secretion ‘on’ conformation, to the effector secretion ‘off’ conformation.
Deletion of pcrG and pcrV results in premature effector secretion
In P. aeruginosa, as in Yersinia sp., effector secretion can be triggered in vitro by removing calcium from the growth medium. As had been previously reported, deletion of pcrG or pcrV results in effector secretion in the presence of calcium (Mccaw et al., 2002; Sundin et al., 2004; Rietsch et al., 2005). We decided to test these phenotypes in our strain of P. aeruginosa using two assays: indirectly, by monitoring control of exoS transcription, and directly, by assaying effector secretion by Western blot.
In P. aeruginosa, expression of all type III secretion-related genes (both apparatus and effector genes) is controlled by the master regulator ExsA (Frank and Iglewski, 1991; Yahr and Frank, 1994; Wolfgang et al., 2003). The activity of ExsA is related to triggering of effector secretion by virtue of the small type III secreted protein, ExsE (Rietsch et al., 2005; Urbanowski et al., 2005; 2007). When effector secretion is upregulated (either by cell contact or, in vitro, by removing calcium from the medium), ExsE is exported via the type III secretion apparatus, which, through a series of protein–protein interactions, results in the release of ExsA from an inactive complex with the negative regulator ExsD, thereby allowing ExsA to bind to its cognate promoters and to activate transcription (Mccaw et al., 2002; Brutinel et al., 2009; Thibault et al., 2009). As a result, transcription of T3SS related genes (in our case, the effector gene, exoS) can be used as a convenient readout for detecting effector export.
Our second assay involves a modified secretion assay. As gene expression is co-regulated with effector secretion, we designed our secretion experiment to first upregulate gene expression (and secretion) by growing the bacteria in the absence of calcium and then pelleting the cells and resuspending them in medium with or without calcium to determine if re-addition of calcium can block effector export (reestablishing calcium control, RECC assay) (Cisz et al., 2008).
Both deletion of pcrG and pcrV resulted in upregulation of effector secretion in the presence of calcium (Fig. 1A and B). Interestingly, the effect of deleting pcrG and pcrV appears to be additive, in that the pcrGV double mutant has a more severe regulatory defect than either of the single mutants. This observation was evident both by monitoring exoS transcription (which is dependent on export of the negative regulator ExsE) and by monitoring the export of the effector protein ExoT directly. Indeed, using the exoS transcriptional reporter, we were able to recapitulate the single mutant phenotypes by expressing either pcrV or pcrG from a plasmid in the ΔpcrGV mutant background, indicating that this is not an aberrant side effect of the pcrGV double null mutant strain, but rather a property of the two regulatory proteins (Fig. 1C). The remaining low-calcium dependent upregulation of exoS expression seen in these experiments is likely not the result of increased secretion. This type of upregulation can even be seen in mutants where the secretion-dependent regulatory system has been inactivated, such as exsE or exsD null mutants (Mccaw et al., 2002; Rietsch et al., 2005; Urbanowski et al., 2005). It is likely the result of global changes in gene expression [e.g. low-calcium dependent upregulation of cAMP (Wolfgang et al., 2003), which is a positive regulator of type III secretion gene expression]. However, the presence of additional regulatory factors that work in conjunction with PcrG and PcrV cannot be ruled out entirely.
It should be noted that the start codon for the pcrG open reading frame was mis-annotated in the genome (Stover et al., 2000). Selective inactivation of the annotated ATG (ATG1) or the ATG four codons downstream of the annotated ATG (ATG2) demonstrated that, the correct translation initiation codon of pcrG is ATG2 (Fig. S1). This also places the start codon in a better context with regard to the proposed pcrG promoter (ATG1 overlaps with the proposed –10 box) (Brutinel et al., 2008). Indeed, mutating ATG1 to GGG (as opposed to ATC) results in a slight deregulation of the system, suggesting that the proposed site of the pcrG promoter is correct (data not shown).
PcrV export is required to control effector secretion
Current models in the Yersinia field suggest that LcrV controls effector secretion when located in the cytoplasm through its interaction with LcrG. Overexpression of LcrV, even a non-secreted form of LcrV, can upregulate effector export in this organism. We decided to test whether PcrV must be exported in order to control effector secretion by expressing a truncated version of the protein in which the signal sequence had been deleted, PcrV(Δ3–21). To create this protein, we relied on experiments performed in Y. pseudotuberculosis in which deletion of amino acids 3–20 of LcrV prevented its export (Broms et al., 2007).
The mutant protein, lacking the secretion signal, was expressed but not secreted (Fig. 2B). Consistent with these data, expression of the deletion mutant was also not able to restore cytotoxicity to a pcrV null mutant strain (Fig. 2C). Cytotoxicity relies on the presence of PcrV at the needle tip to help form and presumably dock to the translocation pore formed by PopB and PopD (Sawa et al., 1999; Dacheux et al., 2001; Goure et al., 2004). The signal sequence deletion mutant also failed to control effector export (Fig. 2A and B). Replacement of the signal sequence of PcrV with that of the effector ExoS was able to restore export, cytotoxicity and effector secretion control, demonstrating that PcrV requires an intact export signal in order to control effector secretion (Fig. 2). The signal sequence deletion mutant was expressed at a somewhat lower level than the wild-type protein in our experiment (with 10 µM IPTG induction across the board). However, performing the experiment at a higher concentration of IPTG (25 µM) yielded the same result (data not shown), suggesting that the level of expression cannot explain the mutant phenotype.
We considered two possible scenarios to explain the lack of secretion control by PcrV(Δ3–21). On the one hand, PcrV could exert its influence on effector secretion control by assembling at the needle tip (perhaps controlling effector secretion by influencing the conformation of the apparatus). On the other hand, PcrV could require its secretion signal in order to be targeted to the secretion apparatus and control effector export from the cytoplasm (perhaps in a complex with PcrG, or independently of PcrG via a second interaction with the apparatus). In Y. enterocolitica, overexpression of wild-type or a non-secreted form of LcrV resulted in deregulation of effector secretion (Lee et al., 2000). This observation was interpreted to mean that LcrV can relieve the block of secretion imposed by LcrG by titrating it from the apparatus. We decided to test if overexpression of signal-sequenceless PcrV is dominant negative. We reasoned that if PcrV exerts its influence in the cytoplasm, the signal-sequence deletion mutant could be dominant negative by either titrating PcrG into an inactive complex or forming a non-productive interaction with the apparatus that precludes the interaction of wild-type PcrV with the T3SS.
We overexpressed either wild-type or signal-sequenceless PcrV in an otherwise wild-type strain of P. aeruginosa. Neither overexpression of wild-type PcrV nor the signal-sequenceless mutant resulted in deregulation of effector secretion (Fig. 3). We considered the possibility that PcrV(Δ3–21) could be defective in its interaction with PcrG. To this end we modified the chromosomal copy of pcrG to express an internally VSV-G (Vesicular Stomatitis Virus G-protein epitope) tagged version of PcrG in order to be able to track the protein. Importantly this tagged version of the protein is essentially fully functional (94% active, Fig. S2). Both versions of PcrV could be co-precipitated with PcrG from cytoplasmic extracts, demonstrating that the two proteins interact in P. aeruginosa and that the interaction is not perturbed by removal of the signal sequence of PcrV (Fig. 3C). The lack of overexpression phenotype, together with the inability of signal-sequenceless PcrV to control effector secretion, suggested to us that PcrV has to be exported to control effector secretion. However, as mentioned above, the absence of a dominant negative phenotype could also be explained by arguing that the regulatory interaction of PcrG or wild-type PcrV with the apparatus is too strong to be perturbed by excess non-secreted PcrV.
Effector secretion control correlates with needle tip function, not the interaction between PcrG and PcrV
The interaction between LcrG and LcrV is crucial for effector secretion control in the homologous Yersinia T3SS. The interaction of PcrG and PcrV has been demonstrated in vitro (Nanao et al., 2003), and we have now demonstrated it in vivo. We therefore set out to determine if the PcrG–PcrV interaction is important for effector secretion control in P. aeruginosa.
As mentioned in the introduction, the data regarding the localization of LcrG in Yersinia sp. is somewhat confusing. While experiments performed in Y. pseudotuberculosis and Y. enterocolitica demonstrate that LcrG is cytoplasmic, evidence gathered in Y. pestis suggests that the protein is exported. This export is not, however, required for effector secretion control because a non-secreted Gal4-LcrG fusion was still capable of complementing a lcrG null mutant (Reina et al., 2008).
We decided to test whether PcrG is exported in order to determine if it could be involved in controlling effector export in conjunction with PcrV at the needle tip. A standard fractionation experiment, however, demonstrated that, unlike PcrV, PcrG is a cytoplasmic protein and not secreted (Fig. 4). While we detected a small amount of PcrG in the membrane and periplasmic fractions in this experiment, the presence of PcrG in these fractions was not affected by the presence or absence of an intact T3SS, suggesting that this represents the resolution limit of the technique rather than an association with the T3SS (data not shown).
Next we determined if the interaction between PcrG and PcrV is required for controlling effector secretion. While PcrV appears to exert its influence at the needle tip, the interaction could still be required to stabilize PcrG or to influence its conformation in a way that allows it to interact with the apparatus.
To this end, we constructed mutants of PcrV that either disrupt the PcrG–PcrV interaction without influencing needle tip function or disrupt needle tip function and still allow binding of PcrG. The L262D and F279R mutations were originally chosen because the homologous residues are thought to be involved in the interaction with LcrG in Y. pestis (Hamad and Nilles, 2007). The L262D amino-acid substitution had been constructed previously in P. aeruginosa and was reported to allow intoxication of host cells (albeit with a slight defect) (Caroline et al., 2008). The effect of an F279R mutation on PcrV function has not been reported, nor has the effect of either mutation on binding of PcrV to PcrG.
We first examined the ability of these mutant forms of PcrV to interact with PcrG. To this end we used an Escherichia coli two-hybrid system in which one interaction partner is fused to the omega subunit of RNA polymerase, the other interaction partner is fused to the monomeric zinc-finger DNA binding domain of murine Zif268 (Zif) (Vallet-gely et al., 2005). An interaction between the two proteins results in recruitment of RNA polymerase to a test promoter, thereby activating lacZ transcription, which can be readily measured by β-galactosidase assay. PcrG interacts strongly with PcrV in this system and neither protein interacts with the unrelated nucleoid proteins MvaT and MvaU (Fig. 5A). In subsequent experiments we used a modified version of the Zif–PcrG fusion, in which a VSV-G tag was inserted between the fusion partners in order to be able to monitor the stability of the fusion protein. The insertion of the tag did not interfere with the PcrG–PcrV interaction (Fig. 5B and D). Introduction of the L262D amino acid substitution into PcrV abolished the interaction with PcrG, whereas introduction of the F279R substitution had no effect on binding in this assay. Interestingly, the Zif–PcrG fusion was unstable when co-expressed with the ω-PcrV(L262D) fusion protein, or omega alone, suggesting that PcrG is unstable in this background when not interacting with PcrV (Fig. 5C). This observation is consistent with published reports that binding of PcrG to PcrV results in greater resistance of PcrG to thermal denaturation in vitro (Nanao et al., 2003). We also tested these interactions by determining whether PcrV can be co-purified with a His-tagged version of PcrG from cytoplasmic extracts of E. coli BL21 expressing either wild-type or mutant versions of the proteins (Fig. 5F). While the F279R mutation had a deleterious effect on PcrG-binding when compared with wild-type PcrV in this assay (only barely detectable in the two-hybrid data), it was not as severe as the interaction-defect seen with PcrG(A16R) or PcrV(L262D).
We next crossed the L262D and F279R mutations into the chromosomal copy of pcrV and assayed the effect of these mutations on effector secretion control, export of the protein and cytotoxicity. All forms of PcrV were stably expressed and secreted, although secretion of the L262D mutant was somewhat reduced when compared with that of wild-type PcrV or the F279R mutant (Fig. 6B). Consistent with the proposed role in needle tip function, the F279R mutation led to a severe defect in cytotoxicity (Fig. 6C). In order to assess the ability of the mutant PcrV to assemble at the needle tip we employed a Fluorescence-Activated Cell Sorting (FACS) assay in which we stained intact P. aeruginosa with an affinity-purified antiserum directed against PcrV. Purified PcrV added to our null mutant bacteria did not interfere with the assay, suggesting that any PcrV protein secreted by these strains should not interfere with the detection of needle tip-associated PcrV (Fig. 6D). In addition, analysis of the FACS samples by immunofluorescence microscopy demonstrated a punctate localization, consistent with the detection of needle tip complexes as opposed to non-specifically associated PcrV (Fig. S3). The pcrV(F279R) mutation resulted in a ∼5-fold reduction of PcrV displayed at the bacterial cell surface as assayed by FACS (Fig. 6D). Taken together, these data demonstrate that the PcrV(F279R) mutant protein is indeed defective at assembling into a needle tip. The L262D mutation, on the other hand, did not interfere with cytotoxicity or assembly at the needle tip (Fig. 6C and D). Interestingly, the L262D mutant of PcrV, which had lost the ability to interact with PcrG, was still able to control effector secretion. The F279R mutant of PcrV, on the other hand, lost the ability to control effector secretion (Fig. 6A and B).
Taken together, these data suggest that the ability of PcrV to control effector secretion correlates with its function as the needle tip protein, and not with its ability to form a complex with PcrG. These data are consistent with our observation that PcrV must be exported to control effector secretion and suggest that assembling into a functional needle tip is required for controlling effector secretion. As the phenotype of the pcrV null mutant is deregulated effector secretion, it would suggest that the default state of the core secretion apparatus is effector secretion ‘on’ and that assembly of PcrV at the needle tip shifts the overall conformation of the apparatus to the effector secretion ‘off’ position.
PcrG controls effector secretion in a PcrV-independent manner
The above data also suggest that PcrG can exert its influence without being in a complex with PcrV. To further confirm these data we modified the chromosomal copy of pcrG to express an A16R mutant of PcrG based on the homologous mutation in LcrG that had been demonstrated to abrogate the LcrG–LcrV interaction (Matson and Nilles, 2001). This mutant also failed to interact with PcrV (Fig. 5D and F); however, it retained most of its regulatory capacity (Fig. 7A and B). Interestingly, the Zif–PcrG(A16R) fusion was also unstable, even when co-expressed with wild-type PcrV (Fig. 5E). We therefore decided to determine if PcrG also depends on PcrV binding for stability in P. aeruginosa. Indeed, introduction of the A16R mutation into our VSV-G-tagged PcrG, or co-expression of the tagged PcrG with the PcrV(L262D) mutant protein (all expressed in their native chromosomal context) resulted in destabilization of PcrG (Fig. 7C). It is interesting that the PcrG(A16R) mutant is still able to control effector export in P. aeruginosa despite this instability. In fact, what little deregulation we observed in this mutant can likely be attributed to the reduction in PcrG stability.
In Yersinia sp., LcrG serves as an export chaperone of LcrV. We decided to use the A16R mutant of PcrG to test if the interaction with PcrV is required for PcrV export. Both deletion of pcrG and introduction of the A16R mutation resulted in a significant defect in PcrV export (Fig. 7B), suggesting that the PcrG–PcrV interaction, although not absolutely required, does facilitate PcrV export. Consistent with this finding, the PcrV(L262D) mutant protein, which is impaired in its ability to interact with PcrG, also displayed a slight defect in export (Fig. 6B). Notably, in all cases, PcrV is still secreted in the presence and absence of calcium [as we had observed for all translocator proteins (Cisz et al., 2008)], demonstrating that PcrG is not specifically required for PcrV export prior to triggering of effector secretion.
In a final set of experiments we decided to attempt to completely separate the PcrV-binding and regulatory activities of PcrG. To this end we created fusions of either full-length PcrG, amino acids 2–40 or 41–95 to maltose binding protein (MBP) and assayed them for their ability to control effector secretion or mediate PcrV binding. Both full-length PcrG and PcrG(41–95) were able to control effector secretion [exoS expression was only slightly increased in the strain expressing PcrG(41–95)], while expression of the MBP-PcrG(2–40) fusion protein resulted in deregulated effector secretion similar to that observed in the vector control (Fig. 8A). However, pull-down of PcrG using an amylose resin demonstrated that only full-length PcrG and PcrG(2–40) were able to interact with PcrV (Fig. 8B). These two activities of PcrG can therefore be separated and binding of PcrV to PcrG is not required for effector secretion control, or low-calcium dependent triggering of effector export. Interestingly, both halves of PcrG are required to restore efficient export of PcrV (Fig. 8C). As the regulatory activity of PcrG likely requires a specific interaction with the T3SS, these data suggest that the interaction also serves to bring PcrV to the apparatus to facilitate its export.
Taken together, these data illustrate that the interaction between PcrG and PcrV is not required for effector secretion control. PcrV controls effector export by assembling into a functional needle tip, while PcrG controls effector secretion in the cytoplasm by an as-yet-to-be discovered interaction with the T3SS.
Cell contact-dependent delivery of effector proteins is a core feature of most virulence-related T3SSs. How this process is regulated, however, is not well understood. In most cases where it has been tested, the needle tip protein prevents premature secretion of effector proteins.
We have presented several lines of evidence that PcrV controls effector secretion by assembling at the needle tip. PcrV has to be export competent to control effector secretion. Similarly, effector secretion control was specifically abrogated in a point mutant (F279R) that interfered with the ability of PcrV to assemble into a functional needle tip (without interfering with the ability to bind to PcrG or be targeted for secretion). A complimentary mutant (L262D) that was still able to assemble at the needle tip, but had lost the ability to bind to PcrG, still controlled effector secretion. The effector secretion control defect of the non-secreted PcrV mutant, as well as the needle tip assembly mutant (F279R), is of a similar magnitude as that of the complete null. Finally, unlike Yersinia LcrV, overexpression of wild-type PcrV or a non-secreted mutant of PcrV had no effect on effector secretion control. Taken together these data argue strongly that PcrV controls effector secretion by assembling into a functional needle tip complex, and argue against models where PcrV physically obstructs access of effectors to the cytoplasmic face of the T3SS. Our data on PcrV function are consistent with published work in Shigella. Small (as little as 3 amino acids) deletions at the C-terminus of IpaD were able to prevent surface assembly and resulted in loss of effector secretion control (Espina et al., 2006). Similarly, deletion of the secretion signal of IpaD also resulted in constitutive effector secretion (Picking et al., 2005). While the data were interpreted to corroborate the plug model of effector secretion control, the authors did note that all of their internal deletion mutants resulted in loss of effector secretion control, but not, for example, the ability to invade tissue culture cells, suggesting that the needle senses the overall conformation of the needle tip protein (Picking et al., 2005).
Of the three models for effector secretion control that have been proposed, our data best fit the allosteric model of effector secretion control. An outright block of the secretion channel by the needle tip protein as proposed in the ‘plug’ model seems unlikely, because we have demonstrated that translocator proteins can be secreted even when effector protein secretion is off (Cisz et al., 2008). The sensor model also doesn't fit our data well, because the partial secretion control defect of the pcrV null mutant is difficult to reconcile with this model where loss of tethering of the sensor at the needle tip should result in unrestricted effector export. Assuming that PcrV controls effector secretion by influencing the conformation of the apparatus, then our data imply that the default state of the apparatus, in the absence of PcrG and PcrV, is the effector secretion ‘on’ conformation. Assembly of PcrV at the needle tip (and perhaps binding of PcrG to a cytoplasmic component(s) of the apparatus) shifts the conformation of the apparatus to the effector secretion ‘off’ state (Fig. 9). Cell contact, in turn, would likely result in a conformational change in the needle tip (perhaps by interacting with the assembled translocation pore) which is then propagated to the base of the apparatus resulting in release of PcrG and triggering of effector secretion.
While we favour an allosteric model of effector secretion control, some reason for caution remains. As mentioned in the Introduction, one problem with the conformational control model is that the biophysical characterization of isolated wild-type and effector secretion ‘on’ mutant needles failed to discern a large-scale conformational difference (Cordes et al., 2005). One potential explanation for this result is that the isolated wild-type needles used in this study may in fact have switched to the effector secretion ‘on’ conformation once separated from the base of the apparatus. This assumes that the Shigella T3SS has a similar propensity to be in the effector secretion ‘on’ conformation as we are proposing for the P. aeruginosa T3SS. Perhaps the analysis of mutants locked in the effector secretion ‘off’ state could be used to trap the second conformation proposed by the allosteric control model. It could also be argued, for example, that the translocator secretion we detect in our RECC assays is the result of translocator proteins being secreted from assembling T3SS, which haven't yet assembled the PcrV needle tip. However, this seems unlikely since export of the pore-forming translocator proteins in our RECC assays is quite robust. Moreover, the pore-forming translocator proteins have to be exported before effector proteins and in cis for triggering of effector secretion to occur on cell contact (Cisz et al., 2008), arguing that translocator export has to occur before effector export and likely by the same apparatus. In other systems it has been reported that translocator export requires a specific trigger [e.g. serum albumin in the case of Y. enterocolitica, or bile salts in the case of S. flexneri (Lee et al., 2001; Olive et al., 2007)], raising the possibility that three secretion states exist: no secretion, translocator secretion and effector secretion. In P. aeruginosa, however, translocator secretion appears to be constitutive and is even detectable in minimal media (Cisz et al., 2008). Relying on a low level of constitutive translocator export to allow assembly of the translocation pore upon cell contact and subsequent triggering of effector secretion seems to be the simpler model, but more work will have to be done to fully understand how needle tip proteins control secretion.
We were able to demonstrate that PcrG is not secreted and therefore must control effector secretion in the cytoplasm. This observation fits well with the most recent data regarding LcrG function, which also argue that LcrG controls effector secretion at the cytoplasmic face of the apparatus (Reina et al., 2008). Whether PcrG controls effector secretion by stabilizing the effector secretion ‘off’ state of the secretion apparatus, or whether it controls effector secretion by directly blocking access of effectors to the T3SS is unclear. Interestingly, unlike the situation in Yersinia sp., our data demonstrate that effector secretion control by PcrG does not require the ability to interact with PcrV. Both a mutant of PcrV (L262D) that has lost the ability to interact with PcrG and a reciprocal mutant of PcrG (A16R) that had lost the ability to interact with PcrV could still prevent premature effector secretion. Moreover, using MBP fusions of truncated versions of PcrG we were able to demonstrate that the regulatory activity and the PcrV-binding activity of PcrG can be separated. The ability to control effector export is tied to the C-terminal part of the protein. While the interaction between PcrG and PcrV is not required for effector secretion control, it is required for efficient secretion of PcrV, which suggests that PcrG serves as export chaperone for PcrV. A similar LcrV secretion defect had also been noticed in the context of an LcrG mutant (A16R) that had lost the ability to bind to LcrV (Matson and Nilles, 2001).
Reconciling control of effector secretion by PcrV with the mode of regulation exhibited by its nearest neighbour, LcrV, is more challenging. In some instances, a slight deregulation of effector secretion has been observed, even in an lcrV null mutant (Lee et al., 2000; Davis and Mecsas, 2007). However, in most instances, it would appear that deletion of lcrV downregulates secretion, even when expression has been artificially upregulated by deleting the negative regulator gene lcrQ (Sarker et al., 1998; Fields et al., 1999; Pettersson et al., 1999; Broms et al., 2007). It may well be that LcrV performs multiple functions, such as control of T3SS gene expression and effector secretion control and that the regulatory function of LcrV has to be taken out of the equation before its role in effector secretion control can be fully appreciated.
Taken together, we have presented data that PcrV controls effector secretion in P. aeruginosa by assembling at the needle tip of the T3SS. This control is likely mediated allosterically by stabilizing the effector secretion ‘off’ conformation of the apparatus. PcrG, which facilitates PcrV export, does not have to bind to PcrV to control effector export and likely does so independently through an interaction with the T3SS at the cytoplasmic face of the apparatus.
Media and culture conditions
All E. coli strains were routinely grown at 37°C in LB medium containing 10 g l−1 NaCl. P. aeruginosa was grown at 37°C in a modified LB medium (LB-MC) containing 200 mM NaCl, 0.5 mM CaCl2 and 10 mM MgCl2. Strains and plasmids used in this study are listed in Table 1.
Table 1. Strains and Plasmids.
BL21(DE3) Codon+– RP
E. coli B F–ompT hsdS(rB– mB–) dcm+ Tetrgalλ (DE3) endA Hte [argU proL Camr]
Two-hybrid analysis strain lacking rpoZ (encoding ω) and harbouring a test promoter-lacZ fusion to detect Zif-dependent two-hybrid interactions
Two-hybrid plasmid encoding a PcrG-Zif fusion protein
Two-hybrid plasmid encoding a PcrV-ω fusion protein
Two-hybrid plasmid encoding a PcrV(L262D)-ω fusion protein
Two-hybrid plasmid encoding a PcrV(F279R)-ω fusion protein
Two-hybrid plasmid encoding a PcrG-VSV-G tag-Zif fusion protein
Two-hybrid plasmid encoding a PcrG(A16R)-VSV-G tag-Zif fusion protein
T7 promoter expression vector for the expression of an amino-terminally His-tagged version of PcrG
T7 promoter expression vector for the expression of an untagged version of PcrV
T7 promoter expression vector for the concomitant expression of an amino-terminally His-tagged version of PcrG and an untagged version of PcrV
T7 promoter expression vector for the concomitant expression of an amino-terminally His-tagged version of PcrG(A16R) and an untagged version of PcrV
T7 promoter expression vector for the concomitant expression of an amino-terminally His-tagged version of PcrG and an untagged version of PcrV(L262D)
T7 promoter expression vector for the concomitant expression of an amino-terminally His-tagged version of PcrG and an untagged version of PcrV(F279R)
Plasmid encoding an MBP-PcrG fusion protein under control of a lacUV5 promoter
Plasmid encoding an MBP-PcrG(aa 2–40) fusion protein under control of a lacUV5 promoter
Plasmid encoding an MBP-PcrG(aa 41–95) fusion protein under control of a lacUV5 promoter
Primers used in this study are listed in Table S1. Plasmid pPSV37 was created in two steps from plasmid pPSV35 (Rietsch et al., 2005). First a T7 terminator was inserted after the multiple cloning site by digesting pPSV35 using enzymes HindIII and PvuI and ligating the cut vector with a linker created by annealing primers T7ter1 and T7ter2. The resultant construct was then digested with EcoRI and ligated with a linker containing stop codons in every reading frame leading up to the polylinker (created by annealing the primers stops1 and stops2) resulting in plasmid pPSV37. Plasmid pZifVG was created by digesting plasmid pACTR-AP-Zif with NotI and XhoI and ligating the cut vector with a linker created by annealing primers ZifVG-sense and ZifVG-AS. Plasmid pMal was created by amplifying a signal-sequenceless version of malE in which the stop codon had been omitted to allow the creation of C-terminal fusions to MBP. To this end, malE was amplified using primers malE-5R and malE-3K and cloned into plasmid pPSV37 as an EcoRI/KpnI fragment.
Inserts cloned into the above vectors were amplified using primers listed in Table S1. Constructs designed to introduce mutations on the chromosome (cloned into the allelic exchange vector pEXG2) were created using splicing by overlap extension PCR (Warrens et al., 1997). Internal primers (defining the mutation) and external primers (defining the ends of the two flanking regions amplified, as well as the restriction sites used to cloned the spliced cross-over PCR product) are listed in Table S1.
Cells were permeabilized with chloroform/SDS, and β-galactosidase activity was assayed as described previously (Miller, 1992).
Pseudomonas aeruginosa was diluted 1:300 from overnight cultures and grown to an OD600 of ∼0.4 in LB-MC medium supplemented with the calcium chelator EGTA (5 mM). At this point, 2 × 1.5 ml of each culture were spun down in microcentrifuge tubes, the supernatants were removed and one cell pellet was resuspended in 2 ml pre-warmed LB-MC medium with calcium, the other in 2 ml pre-warmed medium with EGTA. The cultures were incubated for an additional 25 min, at which point the cultures were placed on ice. Cells from 1 ml of each culture were pelleted by centrifugation and 0.5 ml of the supernatant was transferred to a fresh microcentrifuge tube, wherein supernatant proteins were precipitated by the addition of trichloroacetic acid to a final concentration of 10%. The remainder of the culture was used to determine the OD600. Cell pellets and supernatant proteins were resuspended in 1× SDS sample buffer to correspond to an OD600 of 10. The samples were then boiled for 10 min before separating the proteins on SDS-PAGE. Proteins were transferred to PVDF membrane and probed with specific antisera. Primary antibodies were detected using horseradish peroxidase-conjugated secondary antibodies and a chemiluminescent detection reagent [SuperSignal West Pico (Pierce)]. Exposure times for cell and supernatant fractions varied. Antibodies to ExoT and PcrV were generated in rabbits using a commercial service (Covance). The PcrV antiserum was further purified using an affinity purification protocol. Antibodies directed against RpoA (Neoclone) and the VSV-G tag (Rockland) were obtained commercially.
Bacteria were diluted 1:300 into fresh LB-MC medium supplemented with 5 mM EGTA to chelate calcium. The cultures were then grown at 37°C to mid-log phase, whereupon 1 ml of culture was removed to a microcentrifuge tube (at this point 50 ng ml−1 PcrV were added to one of the ΔpcrV control samples and incubated at room temperature for 10 min). The cells were pelleted by centrifugation (4 min, 4000 r.p.m.), washed once in PBST (PBS with 0.1% triton, 10 mM MgCl2, 0.5 mM CaCl2) and resuspended in 500 µl of the same buffer. At this point 500 µl of 4% paraformaldehyde was added, mixed by inversion and incubated for 20 min at room temperature. The remaining cross-linker was quenched by the addition of 50 µl 1 M Tris.Cl (pH 7.5) and the cell suspension was again mixed by inversion, incubated for 5 min at room temperature, and cells were pelleted (3 min, 13 000 r.p.m.), washed 1× with PBST and 1× with PBS (containing Mg2+ and Ca2+, as indicated for PBST) and resuspended in PBS with 2% goat serum and 2% BSA. Blocking was performed for 30 min on ice at which point the bacteria were pelleted and resuspended in 100 µl of the blocking solution containing an affinity-purified rabbit anti-PcrV antibody (1:500 dilution). The primary antibody incubation was performed at room temperature for 1 h, at which point 1 ml of PBS was added to each sample, the cells were pelleted and washed 2× with PBS before being resuspended in 250 µl of the secondary antibody solution [anti-rabbit-APC-conjugated antibody (Invitrogen) diluted 1:2500 in blocking solution]. The tubes were then wrapped in aluminium foil and rocked at 4°C for 1 h. The bacteria were then pelleted, washed 1× with PBS and resuspended in 50 µl of PBS. For each sample, 10 µl of the labelled bacteria was diluted into 1 ml of PBS in a 5 ml polystyrene tube (BD Falcon) and analysed by flow Cytometry [Becton Dickinson LSR II, a 4 laser, 42 parameter (12 colour) bench-top flow cytometer] at the Case Comprehensive Cancer Center Flow Cytometry Core.
Bacteria were diluted 1:300 into fresh LB-MC medium and subsequently grown at 37°C for ∼3 h, at which point bacteria from 5 ml of culture were pelleted and resuspended in IPP150 buffer (10 mM Tris pH 7.5, 150 mM NaCl, 1 mM PMSF) at an OD600 of 5. The cell suspensions were kept on ice and sonicated four times for 30 s [power level 4, Sonicator Cell Disruptor (W200R) Heat System Ultrasonics]. At this point, cell debris was pelleted at 4°C (13 200 r.p.m., 10 min) and the supernatant was transferred to a new microcentrifuge tube. For each sample, 45 µl of the lysate was removed and combined with 15 µl of 4× SDS sample buffer (input control). The remainder of each lysate was pre-cleared for 15 min by the addition of 20 µl of IPP150-washed protein A/G agarose beads (Santa Cruz Biotechnology). The beads were removed by centrifugation (8000 r.p.m., 3 min) and the supernatant was transferred to a new microcentrifuge tube. Three microlitres of rabbit anti-VSV-G antiserum (Sigma) was added to each tube and the lysates were incubated on a rocker at 4°C for 45 min. Next, 30 µl of washed protein A/G agarose beads was added to each tube and the samples were rocked for an additional 45 min at 4°C. The beads were then pelleted, washed 3× with wash buffer (10 mM Tris pH 7.5, 50 mM NaCl, 1% Triton X-100, 1 mg ml−1 BSA) and resuspended in 60 µl 1× SDS sample buffer. The samples were incubated at 55°C for 10 min to dissociate the bound proteins, vortexed, the beads were pelleted, and the supernatant was removed (elution fraction).
Pseudomonas aeruginosa PAO1F ΔexsEΔfleQ pcrG(i3VG) harbouring pPSV18 (ampR– source of beta-lactamase) was grown overnight in high salt LB (no antibiotics). Sphaeroblast formation was carried out based on the technique of Cheng et al. (1971). The overnight culture was diluted 1:300 into fresh high salt LB (12 ml) and grown at 37°C to an OD600 of ∼0.3–0.4. At this point, 1 ml of culture was removed, the cells were pelleted and the supernatant removed. The cell pellet was resuspended in 100 µl 1× SDS sample buffer. Supernatant protein was collected by TCA precipitation (10% final TCA concentration) and also resuspended in 100 µl 1× SDS sample buffer. The bacteria from 10 ml of the remaining culture were pelleted and resuspend in 0.9 ml PEB buffer (20% sucrose, 30 mM Tris pH 8.0) to which 100 µl of a 10 mg ml−1 lysozyme solution in PEB buffer were added and 5 µl of 400 mM EDTA, pH 8.0. The suspension was incubated at room temperature for 20 min, at which point the bacteria were pelleted (5 min at 5000 g). Seventy-five microlitres of the supernatant fraction (periplasm) was removed and mixed with 25 µl of 4× sample buffer. The remaining supernatant was removed and discarded. The cell pellet was then resuspended in 1 ml of cold PBS (with 1 mM PMSF). Sphaeroblast formation was confirmed by microscopy. The cells were then lysed by sonication [4× 30 s interval, power level 4, Sonicator Cell Disruptor (W200R) Heat System Ultrasonics]. The tube was kept on ice to prevent overheating of the sample. Lysis of the cells was controlled by microscopic examination of the sample. After sonication, unlysed cells were removed by centrifugation (10 min at 13 200 r.p.m. at 4°C). The supernatant was then moved to ultracentrifuge tubes and spun for 1 h at 4°C (45 000 r.p.m.). Seventy-five microlitres of the supernatant fraction (cytoplasm) was removed and mixed with 25 µl of 4× sample buffer. The pellet (membrane fraction) was resuspended in 1 ml 1× SDS sample buffer. PcrG was detected by Western blot using a commercial rabbit anti-VSV-G antibody (Sigma). The fractionation was controlled by detecting OprH (membrane fraction), β-lactamase (periplasm), RpoA (cytoplasm) and PcrV (secreted). The antibody against β-lactamase was a kind gift from Dr Robert Bonomo (Case Western Reserve University/VA hospital). The antibody against OprH was a kind gift from Dr Robert E. Hancock (University of British Columbia).
Overexpression in E. coli and Ni-chromatography
Combinations of PcrG, PcrV and mutants of either protein were expressed in E. coli BL21(DE3) Codon+ -RP (Stratagene) using the pDuet1 co-overexpression plasmid (Novagen). Bacteria were grown overnight in 2× YT broth with 30 µg ml−1 chloramphenicol and 60 µg ml−1 carbenicillin. The overnight culture was then diluted 1:300 into fresh 2× YT and grown for 2–2.5 h, at which point expression was induced by the addition of IPTG (100 µM final). The incubation was continued for an additional 25 min, at which point the cultures were placed on ice. For each strain, bacteria from 5 ml of culture were pelleted and resuspended cold Equilibrium buffer (50 mM Na2PO4, 300 mM NaCl, 5 mM imidazole, 1 mM PMSF pH 7.0). The OD600 was normalized to 5. The bacteria were then lysed by sonication (power level 4 with 30 s interval for four times). Cellular debris was pelleted (13 200 r.p.m. for 10 min at 4°C) and discarded. Forty-five microlitres of the supernatant was removed and mixed with 15 µl of 4× SDS sample buffer (input control). Equilibrium buffer washed TALON beads (50 µl of original volume of bead suspension) were added to the remaining supernatant and the suspension was rocked for 1 h at 4°C. The beads were then pelleted (8000 r.p.m. for 3 min) and washed 4× with Equilibrium buffer. Bound proteins were eluted by adding 200 µl Elution buffer (50 mM Na2PO4, 300 mM NaCl, 300 mM imidazole, 1 mM PMSF pH 7.0) to the beads and pelleting the beads. This process was repeated once and the two elution fractions were combined. PcrV and PcrG were detected by Western blot.
MBP fusion purification
Pseudomonas aeruginosa PAO1F ΔpcrG2ΔexoS::GL3 was transformed with pMal, pMal-pcrG, pMal-pcrG(2–40) or pMal-pcrG(41–95). Overnight cultures of each strain were diluted 1:300 into fresh LB-MC with 50 µM IPTG and grown to an OD600 of ∼1 and placed on ice for 10 min. At this point, the OD600 of the cultures was normalized and 4.5 ml of cells were pelleted and resuspended in 500 µl of wash buffer (20 mM Tris (pH 7.5), 0.2 M NaCl, 10 mM β-mercaptoethanol, 1 mM EDTA, 2 mM PMSF). The cells were lysed by sonication and cellular debris was removed by centrifugation (10 min, 14 000 r.p.m., at 4°C). Forty-five microlitres of the supernatant was removed to a fresh tube and mixed with 15 µl of 4× SDS sample buffer (input). Four hundred microlitres of the remaining supernatant was removed to a separate tube and mixed with 50 µl of amylose resin (NEB) washed 1× with wash buffer. The suspension was rotated for 1 h at 4°C, at which point the beads were pelleted and washed 3× with wash buffer. Bound protein was eluted by washing the beads 4× with 100 µl elution buffer (wash buffer + 10 mM maltose) and combining the elution fractions. Forty-five microlitres l of the eluate was combined with 15 µl of 4× SDS sample buffer (eluate). Samples were separated by SDS-PAGE. PcrV and RpoA were detected by Western blot.
We would like to thank Dr Simon Dove and Dr Joseph Mougous for critical reading of the manuscript. We also gratefully acknowledge Dr Robert E. Hancock for the gift of the anti-OprH antiserum, Dr Robert Bonomo for the gift of the anti-β-lactamase antibody, as well as Dr Simon Dove for the components of the Zif-omega two-hybrid system. This research was supported by the Flow Cytometry Core Facility of the Comprehensive Cancer Center of Case Western Reserve University and University Hospitals of Cleveland (P30 CA43703). A.G.S. was supported by the Pulmonary Host Defense training grant from the National Institutes of Health (T32 HL083823). This work was also supported by an American Cancer Society Research Scholar Grant (RSG-09-198-01-MPC) to A.R. as well as a Pilot and Feasibility grant awarded to A.R. through a Cystic Fibrosis Foundation program project grant (R447-CR07).