Insertion of a Yop translocation pore into the macrophage plasma membrane by Yersinia enterocolitica: requirement for translocators YopB and YopD, but not LcrG

Authors

  • Cécile Neyt,

    1. Microbial Pathogenesis Unit, Christian de Duve Institute of Cellular Pathology (ICP) and Faculté de Médecine, Université Catholique de Louvain, B-1200 Brussels, Belgium.
    Search for more papers by this author
  • Guy R. Cornelis

    1. Microbial Pathogenesis Unit, Christian de Duve Institute of Cellular Pathology (ICP) and Faculté de Médecine, Université Catholique de Louvain, B-1200 Brussels, Belgium.
    Search for more papers by this author

Guy R. Cornelis. E-mail cornelis@mipa.ucl.ac.be; Tel. (+32) 2 764 74 49; Fax (+32) 2 764 74 98.

Abstract

The Yersinia survival strategy is based on its ability to inject effector Yops into the cytosol of host cells. Translocation of these effectors across the eukaryotic cell membrane requires YopB, YopD and LcrG, but the mechanism is unclear. An effector polymutant of Y. pseudotuberculosis has a YopB-dependent contact haemolytic activity, indicating that YopB participates in the formation of a pore in the cell membrane. Here, we have investigated the formation of such a pore in the plasma membrane of macrophages. Infection of PU5-1.8 macrophages with an effector polymutant Y. enterocolitica led to complete flattening of the cells, similar to treatment with the pore-forming streptolysin O from Streptococcus pyogenes. Upon infection, cells released the low-molecular-weight marker BCECF (623 Da) but not the high-molecular-weight lactate dehydrogenase, indicating that there was no membrane lysis but, rather, insertion of a pore of small size into the macrophage plasma membrane. Permeation to lucifer yellow CH (443 Da) but not to Texas red-X phalloidin (1490 Da) supported this hypothesis. All these events were found to be dependent not only on translocator YopB as expected but also on YopD, which was required equally. In contrast, LcrG was not necessary. Consistently, lysis of sheep erythrocytes was also dependent on YopB and YopD, but not on LcrG.

Introduction

Yersinia pathogenesis is based on its capacity to overcome the defence mechanism of its animal host by injecting bacterial proteins into the cytosol of the host cells, thereby disabling them (for reviews, see Straley and Perry, 1995; Cornelis and Wolf-Watz, 1997; Lee, 1997; Cornelis et al., 1998). The virulence determinants are a set of 12 proteins called Yops that are secreted across the bacterial membranes by a type III secretion apparatus called Ysc. Some Yops are intracellular effectors (YopE, YopH, YpkA/YopO, YopP/YopJ, YopM, YopT) and are translocated across the eukaryotic cell plasma membrane to reach their intracellular targets (Rosqvist et al., 1994; Sory and Cornelis, 1994; Persson et al., 1995; Boland et al., 1996; Håkansson et al., 1996a; Iriarte and Cornelis, 1998). Release of the Yop proteins occurs only upon contact between the bacterium and the eukaryotic cell. Secretion before contact is presumably prevented by a plug made of a few proteins including YopN, which is itself secreted and translocated into the eukaryotic cell (Rosqvist et al., 1994; Persson et al., 1995; Boland et al., 1996; Lee et al., 1998).

Translocation across the cell membrane requires other secreted proteins, namely YopB, YopD, LcrV and the mainly bacterium-associated LcrG protein (Rosqvist et al., 1994; Sory and Cornelis, 1994; Persson et al., 1995; Boland et al., 1996; Håkansson et al., 1996b; Nilles et al., 1997; Boyd et al., 1998; Francis and Wolf-Watz, 1998; Sarker et al., 1998a,b). These four proteins are encoded by the same large operon lcrGVsycDyopBD, which also encodes the cytosolic SycD chaperone, required for the secretion of YopB and YopD (Wattiau et al., 1994; Neyt and Cornelis, 1999). The fact that these four proteins are required for translocation and are encoded by the same operon suggests that they could act together and form a kind of translocation apparatus in the eukaryotic cell membrane.

Håkansson et al. (1996b) observed that Yersinia has a YopB- and contact-dependent lytic activity on sheep erythrocytes, suggesting that the translocation apparatus involves some kind of a pore in the target cell membrane by which the Yop effectors pass through into the cytosol. This YopB-dependent lytic activity is higher when the effector yop genes are deleted, suggesting that the pore is normally filled with effectors (Håkansson et al., 1996b). The presence of osmoprotectants of a given size can inhibit YopB-mediated sheep erythrocyte lysis. Using two different sugars, the inner diameter of the putative pore was estimated to be between 1.2 nm and 3.5 nm (Håkansson et al., 1996b). The YopB protein contains two central hydrophobic domains and, according to the Eisenberg plot (Eisenberg, 1984), YopB is a transmembrane protein (Håkansson et al., 1993), suggesting that it could be a main structural element of the putative pore. YopK (YopQ in Y. enterocolitica) is proposed to modulate the size of the YopB-induced pore, thereby controlling Yop effector translocation (Holmström et al., 1997).

The role of YopD in the translocation apparatus remains more elusive. Neyt and Cornelis (1999) showed that YopB and YopD are associated together in the bacterial cytoplasm with their chaperone SycD, suggesting that YopB and YopD are secreted as a heterocomplex and might act together at some stage in the translocation apparatus. YopD also contains a central hydrophobic domain, and the Eisenberg plot predicts that it is also a transmembrane protein (Eisenberg, 1984; Håkansson et al., 1993), suggesting that YopD could also be located in the eukaryotic cell membrane. However, recent data have shown that YopD is translocated into the eukaryotic cell (Francis and Wolf-Watz, 1998).

Secretion of YopB and YopD requires LcrV, another secreted protein, which can interact with both of them (Sarker et al., 1998a). Finally, the 11 kDa LcrG protein (Skrzypek and Straley, 1993) is also required for efficient translocation of Yersinia Yop effector proteins into the eukaryotic cells (Sarker et al., 1998b), but its role in the translocation process is still unclear. It can interact with LcrV (Sarker et al., 1998b), and it has also been shown to bind to heparan sulphate proteoglycans on the surface of HeLa cells (Boyd et al., 1998).

In this paper, we show that Yersinia is able to form pores in the macrophage cell membrane, and we address the question of the requirement for the translocators YopD and LcrG in this pore formation.

Results

ΔHOPEMN Y. enterocolitica alters the morphology of infected cells

To study pore formation in macrophages, we used the ΔHOPEMN Y. enterocolitica strain, which no longer synthesizes the translocated proteins YopH, YopO, YopP, YopE, YopM and YopN. This choice was based on the observation that a similar mutant of Y. pseudotuberculosis exhibits a clear lytic activity on red blood cells, whereas wild-type bacteria do not (Håkansson et al., 1996b), presumably because the translocated effectors obstruct the translocation channel. After infection of PU5-1.8 macrophages with these bacteria, almost all (> 90%) of the cells flattened except on the nucleus, evoking fried eggs (Fig. 1). This phenomenon could already be observed after 1 h of infection if the bacteria had been preincubated at 37°C to induce Yop synthesis. If the bacteria had not been preincubated at 37°C, flattened cells were only observed after 5 h of infection. In contrast, cells infected with the wild-type Y. enterocolitica did not flatten, but rather rounded up under the action of the cytotoxic effectors YopE (Rosqvist et al., 1990), YopT (Iriarte and Cornelis, 1998) and YopH (Rosqvist et al., 1988) (Fig. 1).

Figure 1.

. Effect of Yersinia infection on the morphology of PU5-1.8 macrophages. Cells were infected with the ΔHOPEMN or the wild-type strain. For comparison, cells were treated with streptolysin O.

The observed morphological change is not caused by lysis of the cell membrane

To investigate whether flattening of the macrophages was the result of cell lysis, we monitored lactate dehydrogenase (LDH) release from the infected cells. It appeared that macrophages infected by the ΔHOPEMN strain did not release more LDH than the uninfected cells, while a strong LDH activity was measured in the culture supernatant after total cell lysis by Triton X-100. The lack of LDH release by the flattened cells indicated that their plasma membrane had not been grossly damaged and disrupted. To determine whether formation of transmembrane pores might be responsible for the observed morphological change, cells were treated with the pore-forming streptolysin O from Streptococcus pyogenes (SLO). As shown in Fig. 1, PU5-1.8 cells treated with SLO or infected with ΔHOPEMN Y. enterocolitica bacteria showed a similar morphology, indicating that flattening could indeed result from the formation of pores that are too small to allow LDH (135 kDa) release.

Evidence for the existence of small transmembrane pores in macrophages infected with ΔHOPEMN Y. enterocolitica

As the preceding experiments suggested that ΔHOPEMN Y. enterocolitica could form pores in the macrophage cell membrane, we investigated whether small intracellular marker molecules could be released. Macrophages were first loaded with BCECF(AM) (2′,7′-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein, acetoxymethyl ester), a membrane-permeant dye that becomes fluorescent after cleavage by intracellular esterases, and then used as targets for infection. We observed that BCECF (623 Da) was released from cells infected with ΔHOPEMN bacteria (Fig. 2), supporting the idea that Yersinia indeed forms a pore in the macrophage cell membrane through which BCECF is released. However, no BCECF was released from cells infected with the wild-type bacteria, which supports the initial hypothesis that effector Yops obstruct the channel.

Figure 2.

. BCECF release from macrophages upon infection with the Y. enterocolitica polymutant bacteria ΔHOPEMN, ΔHOPEMNB, ΔHOPEMNDΔ121–165, ΔHOPEMND::pMSL31 and ΔHOPEMNG, with these bacteria complemented by plasmids pCNR27 (yopB+) or pMRS74(yopD+) or the wild-type strain. Results are expressed as the percentage of Triton-lysed macrophages after subtraction of the value measured in the supernatant of uninfected cells. Data show means and standard deviations of assays performed in quadruplicate.

To strengthen the pore hypothesis, we tested whether a small molecule could gain access from the exterior to the cytoplasm of macrophages infected with ΔHOPEMN Y. enterocolitica. Infected cells were incubated with lucifer yellow CH, a membrane-impermeant dye (molecular weight 443 Da), and visualized for staining. This experiment was performed in the presence of dextran 6000, as we observed that, in these conditions, infected cells did not flatten. As shown in 3Fig. 3A, cells infected with ΔHOPEMN Y. enterocolitica were stained, while uninfected cells were not. Consistent with our previous experiment, cells infected by wild-type bacteria did not allow entry of the dye (Fig. 3A). These observations confirm that the ΔHOPEMN strain is able to form a pore in the macrophage plasma membrane. To estimate the size of this pore, we tested the entry of a larger molecule, Texas red-X phalloidin, a fixable tracer that binds to F-actin (approximate molecular weight 1490 Da). Unlike lucifer yellow CH, this dye did not stain the ΔHOPEMN-infected cells, whereas it stained macrophages treated with the pore-forming streptolysin O (Fig. 3A). Thus, the pore formed by ΔHOPEMN bacteria appears to allow the passage of molecules smaller than 623 Da (BCECF) but not larger than 1490 Da (Texas red-X phalloidin).

Figure 3.

. Entry of lucifer yellow CH or Texas red-X phalloidin into PU5-1.8 macrophages upon infection with different mutants of Y. enterocolitica in the presence of dextran 6000. Representative fluorescent and corresponding phase-contrast images are shown. A. Entry of lucifer yellow CH but not of Texas red-X phalloidin into PU5-1.8 macrophages infected with ΔHOPEMN or ΔHOPEMNG Y. enterocolitica. Uninfected cells (data not shown) and cells infected with the wild-type strain were not stained by lucifer yellow CH. As a positive control for staining by Texas red-X phalloidin, cells were treated with streptolysin O. B. Staining of PU5-1.8 macrophages by lucifer yellow CH in the presence of dextran 6000 upon infection with the Y. enterocolitica polymutant strains ΔHOPEMNB, ΔHOPEMNDΔ121–165 and ΔHOPEMND::pMSL31 or with these strains complemented by plasmids pCNR27 (yopB+) or pMRS74(yopD+).

Pore formation in the macrophage cell membrane by Yersinia requires YopB and YopD, but not LcrG

We then set out to identify the proteins required for the formation of this pore. The ΔHOPEMN bacteria used in the previous experiments secrete the translocators YopB, YopD and LcrV, but also secrete proteins that are not required for translocation, namely YopT, YopR and YopQ (Holmström et al., 1997; Iriarte and Cornelis, 1998; M.-P. Sory, unpublished). To eliminate any possibility that the observation we made could be caused by the presence of the latter three proteins, macrophages were infected with a new ΔHOPEMTQR strain, which secretes only YopB, YopD, LcrV and YopN (N. Grosdent et al., unpublished). These infected macrophages had the same flattened morphology as the ΔHOPEMN-infected cells, indicating that YopT, YopQ and YopR were not involved in pore formation (data not shown).

We thus investigated the requirement for the translocators YopB, YopD and LcrG for pore formation. Håkansson et al. (1996b) have already shown that the contact-dependent haemolytic activity of Yersinia is dependent on YopB. We deleted yopB or lcrG from the ΔHOPEMN strain, giving the ΔHOPEMNB and ΔHOPEMNG strains respectively. Two different yopD mutations were also introduced into the ΔHOPEMN strain, one leading to the production of a truncated YopD protein missing its hydrophobic domain (ΔHOPEMNYopDΔ121–165) and one completely abolishing the production of YopD (ΔHOPEMNYopD::pMSL31). The absence of one of the translocators did not affect the synthesis and secretion of the others (Fig. 4). Cells infected by the strains lacking YopB or YopD looked like uninfected cells, while the morphology of cells infected with the ΔHOPEMNG bacteria was identical to that of cells infected with ΔHOPEMN bacteria, irrespective of the growth conditions of the bacteria before infection. Complementation of the yopB and yopD mutations with pCNR27 and pMRS74 encoding yopB and yopD, respectively, from the pyopE promoter restored the phenotype of the ΔHOPEMN strain, either completely or partially, depending upon whether the bacteria were preinduced or not before infection (Fig. 5). This indicates that both YopB and YopD are required for pore formation, while LcrG is not necessary.

Figure 4.

. SDS–PAGE analysis (Coomassie blue staining) of the Yop proteins (identified at left) secreted by wild-type Y. enterocolitica MRS40(pYV40) (lane 1), ΔHOPEMN mutant MRS40(pIM417) (lane 2), ΔHOPEMNB mutant MRS40(pCN4008) (lane 3), ΔHOPEMNDΔ121–165 mutant MRS40(pCN4007) (lane 4), ΔHOPEMND::pMSL31 mutant MRS40(pCNK4009) (lane 5) and ΔHOPEMNG mutant MRS40(pCN4006) (lane 6).

Figure 5.

. Requirement of YopB and YopD but not LcrG in flattening of PU5-1.8 macrophages upon infection by Y. enterocolitica. Percentage of flattened macrophages after 5 h infection with bacteria that were not preincubated at 37°C to induce Yop synthesis. Cells were infected with Y. enterocoliticaΔHOPEMN, ΔHOPEMNB, ΔHOPEMNDΔ121–165, ΔHOPEMND::pMSL31 and ΔHOPEMNG or with the same mutant bacteria complemented by plasmids pCNR27 (yopB+) or pMRS74 (yopD+) or the wild-type strain.

To confirm this observation, we monitored the release of BCECF from infected cells. As shown in Fig. 2, release of BCECF required the presence of both YopB and YopD, but not LcrG. We also assessed entry of lucifer yellow into the infected cells and, as shown 3Fig. 3A and B, entry of lucifer yellow only occurred when YopB and YopD were present, but the presence of LcrG was not required. This confirms that pore formation by Yersinia in the macrophage cell membrane requires YopB and YopD, but not LcrG.

The absence of LcrG does not affect the size of the pore in the macrophage cell membrane

We then wondered whether the absence of LcrG could affect the size of the pore. In this attempt, ΔHOPEMNG-infected cells were incubated with lucifer yellow CH (molecular weight 443) and Texas red-X phalloidin (approximate molecular weight 1490). As seen for the ΔHOPEMN-infected cells (see above), lucifer yellow stained the ΔHOPEMNG-infected cells but Texas red-X phalloidin did not (Fig. 3A). As BCECF (623 Da) was still released (see above), we conclude that the absence of LcrG does not affect the size of the pore, reinforcing the idea that LcrG is not involved in pore formation, although it is required for translocation of the effectors (Sarker et al., 1998b).

YopD is required for Yersinia-mediated contact haemolysis, but LcrG is not

As YopB and YopD, but not LcrG, are required for pore formation in the macrophage cell membrane, we expected that the same would apply for Yersinia-mediated lysis of erythrocytes. Yersinia-mediated haemolysis is known to be YopB dependent (Håkansson et al., 1996b). Here, we investigated the importance of YopD and LcrG in haemolysis. Sheep erythrocytes were mixed with the ΔHOPEMN, ΔHOPEMNB, ΔHOPEMNDΔ121–165, ΔHOPEMND::pMSL31 and ΔHOPEMNG bacteria. As expected, ΔHOPEMNB, ΔHOPEMNDΔ121–165 and ΔHOPEMND::pMSL31 bacteria were not haemolytic, and complementation of the mutations restored the activity (Fig. 6A). The ΔHOPEMNG bacteria, however, were haemolytic, just like ΔHOPEMN bacteria. In agreement with the result of Håkansson et al. (1996b), the wild-type strain caused little haemolysis.

Figure 6.

. Yersinia-induced haemolysis. A. Haemolytic activity of Y. enterocoliticaΔHOPEMN, ΔHOPEMNB, ΔHOPEMNDΔ121–165, ΔHOPEMND::pMSL31 and ΔHOPEMNG and of the strains complemented by plasmids pCNR27 (yopB+) or pMRS74(yopD+) and the wild-type strain. Results are expressed as the percentage of water-lysed erythrocytes. Data show means and standard deviations of assays performed in quadruplicate. B. Carbohydrate inhibition of haemolytic activity. Haemolysis was monitored upon infection with the polymutant strains ΔHOPEMN and ΔHOPEMNG in the presence of dextran 6000 or sucrose. Results are expressed as the percentage of Triton-lysed macrophages. Data show means and standard deviations of assays performed in quadruplicate.

To ensure that haemolysis caused by bacteria lacking LcrG was indeed the result of a pore, we tested the inhibitory effect of carbohydrates acting as osmoprotectants. Dextran 6000 (6000 Da) prevented haemolysis whether the bacteria produced LcrG or not; however, sucrose (molecular weight 342.3) did not (Fig. 6B). This indicates that haemolysis in the absence of LcrG is caused by pore formation and, thus, that LcrG is not required for pore formation as observed with macrophages.

Discussion

We observed that polymutant Y. enterocolitica generate pores into the membrane of the eukaryotic target cell that allow the release of small molecules such as BCECF (623 Da) but not large molecules such as lactate dehydrogenase (LDH). These pores also allowed entry of a small dye such as lucifer yellow (443 Da) but not of the larger Texas red-phalloidin (1490 Da). The molecular radius of BCECF and Texas red-phalloidin is not available, but one can refer to molecules of a similar mass, i.e. polyethylene glycol 600 [molecular weight 600; rES 0.8 nm (Einstein–Stokes radius)] and dextran (molecular weight 1500; rES1.15 nm) or PEG E1450 (molecular weight 1450; rES1.2 nm) (Scherrer and Gerhardt, 1971) to estimate the size of the pore. By inference, the inner diameter of the pore can be estimated to be between 1.6 and 2.3 nm, a value that is consistent with the prediction that Håkansson et al. (1996b) made using erythrocytes (1.2–3.5 nm). The pore inserted by Yersinia thus has a size similar to that estimated for pores formed by the Escherichia coli haemolysin (≈ 2 nm) and Staphylococcus aureusα-toxin (1–2 nm) (Bhakdi and Tranum-Jensen, 1987).

No effector Yop was necessary for pore formation, indicating that this pore would be constituted of translocator proteins that allow the passage of the Yop effectors across the eukaryotic cell membrane. Three proteins, YopB, YopD and LcrG, have been shown to be required for translocation (Rosqvist et al., 1994; Sory et al., 1995; Boland et al., 1996; Francis and Wolf-Watz, 1998; Sarker et al., 1998b), but this view has recently been challenged (Lee and Schneewind, 1999). We tested various genetic constructs to determine whether these proteins are required for pore formation in macrophages. As expected from the results of Håkansson et al. (1996b), YopB turned out to be required. However, in contradiction of observations reported by Håkansson et al. (1996b), YopD turned out to be equally essential for pore formation, and this was confirmed in the contact haemolysis assay. The use of a mutant lacking only the hydrophobic domain of YopD demonstrated that this hydrophobic domain has a crucial role in pore formation.

As YopB and YopD are both putative transmembrane proteins (Håkansson et al., 1993) and the hydrophobic domain of YopD is crucial for pore formation, we suggest that these two proteins are the structural components of the pore constituting a heterogeneous transmembrane complex, which is part of the translocation machinery. The fact that YopB and YopD act together to form the pore is consistent with our previous work (Neyt and Cornelis, 1999), which suggested that they are secreted in an associated form. After secretion, the pore could be formed by polymerization of the YopB–YopD complexes, but this needs to be confirmed by the analysis of the pore purified from the eukaryotic cell membrane. YopD could be a more dynamic element of the channel, as it is thought to interact with the translocated effector Yops (Hartland and Robins-Browne, 1998). Moreover, YopD has been shown to be recovered in the eukaryotic cell cytoplasm (Francis and Wolf-Watz, 1998), suggesting that it could be internalized with the Yop effectors during the translocation process. In addition to this role in pore formation, YopD is also known to exert a negative regulatory effect on the Yop virulon, suggesting that it has a dual role (Francis and Wolf-Watz, 1998; Williams and Straley, 1998).

In contrast to yopB and yopD mutant bacteria, lcrG mutants formed pores that were indistinguishable from those formed by the isogenic lcrG+ bacteria. LcrG is thus the only protein that is necessary for translocation of all the effector Yops, without being required for pore formation. This observation allows us to rule out the possibility that LcrG is necessary for the insertion of the pore, but it shows that the pore constitutes only a part of the translocation machinery, as other components are required to constitute the functional apparatus. In addition, some proteins, such as LcrV and SycD for example, are indirectly involved in pore formation, as they are required for the secretion of YopB and YopD (Wattiau et al., 1994; Sarker et al., 1998a). As LcrV is also secreted, we could hypothesize that it is part of the translocation apparatus as well.

The evidence that the pore described here is part of the translocation machinery is as follows. First, the two proteins required for pore formation, namely YopB and YopD, have been shown by several laboratories to be required for translocation of the effector Yops (Sory and Cornelis, 1994; Boland et al., 1996; Håkansson et al., 1996b; Francis and Wolf-Watz, 1998) and, up until now, these two proteins (and LcrG) are the only proteins that have been shown to be required directly for the translocation phenomenon. No mutation other than in the lcrGVsycDyopBD operon or in tyeA has been shown to affect only the translocation step, and we find it unlikely that the Ysc syringe would be sufficient to cross the eukaryotic cell plasma membrane. Secondly, the fact that the pore is only detectable with effector polymutant bacteria and not wild-type bacteria strongly suggests a link between the traffic of effectors and leakage through the pore, as already suggested by Håkansson et al. (1996b). The view of a Yop translocation pore constituted from YopB and YopD, however, does not fit with the results of Lee and Schneewind (1999), who presented data showing that secreted YopB and YopD are dispensable for the translocation process.

The choice of the polymutant ΔHOPEMN strain to study pore formation in macrophages deserves further comment. With wild-type bacteria, pores could not be detected, indicating that our observations could only be made under artificial conditions. Our interpretation is that the formation of a pore is part of the natural attack system but that passage of dyes through the pore does not occur with the wild-type strain. As mentioned earlier, we believe that, under normal conditions, the pore is filled with translocating effectors, but we would like to point out that, even in the absence of Yop effectors, the passage of dyes is surprising. Indeed, if secretion and translocation of the effectors occur through a continuous channel connecting one cytosol to the other (‘the injectisome’), there should be no direct communication between the eukaryotic cell cytosol and the extracellular medium, irrespective of any Yop effector traffic. Our observations of entry and leakage of dyes show, however, that there is direct communication between these two compartments. This suggests that the association between the pore and the injectisome is quite loose. This looseness, however, does not affect translocation, as the effector polymutant strain has been shown by two laboratories (Boland et al., 1996; Håkansson et al., 1996b) to be fully proficient for translocation. The idea that the pore is not tightly associated with the injectisome is consistent with the results of Pettersson et al. (1996) showing that infection of HeLa cells by yopB mutant bacteria leads to upregulation of YopE synthesis, although YopE cannot be translocated. This upregulation indicates that contact-induced secretion occurs even when translocation does not. This hypothesis of a loose association between the injectisome and the pore is also supported by the fact that the pore constituents are only secreted upon contact with the target, while the Ysc secretion machinery is deployed before contact (Cornelis, 1998). In spite of the fact that the pore seems to be loosely associated with the injectisome, it has been observed that effector Yops must be secreted by the same bacterium as the one producing YopB and YopD in order to be translocated (Sory et al., 1995). Similar observations have been made with the closely related type III system of Pseudomonas aeruginosa: P. aeruginosa mutant bacteria proficient for type III secretion but deficient in ExoS synthesis cannot translocate exogenous soluble ExoS added in the culture supernatant (McGuffie et al., 1998).

As discussed above, YopB and YopD probably constitute a heterogeneous transmembrane complex. However, they differ from classical pore-forming toxins in several aspects. Pore-forming toxins are secreted by bacteria in primarily water-soluble conformation and, after diffusion and binding to target cells, they usually oligomerize and insert spontaneously into membranes, creating water-filled channels that will destroy the target cell (Bhakdi and Tranum-Jensen, 1987; Song et al., 1996). In contrast, the YopBD pore is inserted in the eukaryotic cell membrane upon adherence of Yersinia to a target cell. The effector Yops are then injected through this pore into the eukaryotic cytoplasm where they act on their targets. Pore formation is thus used to translocate macromolecules from the bacterial cytoplasm to the host cell. Conceptually, the Yop virulon is thus closer to intracellularly acting toxins such as anthrax toxins (Petosa et al., 1997). The Yops could be viewed as some kind of A–B toxins in which the effector Yops would be the A subunits, while the YopBD pore would be the B subunit. This concept presumably applies to all type III systems.

Experimental procedures

Bacterial strains, plasmids, growth conditions and genetic conjugations

Y. enterocolitica MRS40 is a blaA mutant of strain E40, in which the gene encoding β-lactamase A was replaced by the luxAB genes (Sarker et al., 1998b). XL1 blue (Stratagene) was used for standard genetic manipulations. E. coli CJ236 was used for site-directed mutagenesis (Kunkel et al., 1987). E. coli SM10 lambda pir+ constructed by Miller and Mekalanos (1988) was used to deliver the mobilizable plasmids into Y. enterocolitica. The plasmids used in this study are listed in Table 1.

Bacteria were grown routinely in tryptic soy broth (Oxoid) and plated on tryptic soy agar. To introduce a plasmid into Y. enterocolitica by conjugation, the plasmid was introduced in E. coli SM10 lambda pir+ by electroporation. This donor strain was then mated with the recipient Y. enterocolitica during a 3–5 h period on a plate at 28°C. For the induction of the yop regulon (see below), bacteria are grown in BHI-OX media, which is brain–heart infusion (BHI) broth supplemented with 4 mg ml−1 glucose, 20 mM MgCl2 and 20 mM sodium oxalate.

Construction of the pYV mutants

Allele yopDΔ121–165 (Mills et al., 1997) was introduced in the Y. enterocolitica yopHOPEM MRS40(pABL403) strain by allelic exchange with mutator pMSL19, yielding strain MRS40(pMSK49). This strain produces and secretes a truncated YopD protein missing the hydrophobic domain.

Allele yopN45 (Boland et al., 1996) was introduced in the Y. enterocolitica yopHOPEM mutant MRS40(pABL403), yopHOPEMG mutant MRS40(pMSK48), the yopHOPEMD Δ121–165 mutant MRS40(pMSK49) and the yopHOPEMB mutant MRS40(pAB409) by allelic exchange with the mutator pIM150 (Boland et al., 1996), yielding strains MRS40(pIM417), MRS40(pCN4006), MRS40(pCN4007) and MRS40(pCN4008) respectively.

The yopHOPEMND::pMSL31 mutant strain MRS40-(pCNK4009) was obtained by disruptive integration of pMSL31 into yopD in the Y. enterocolitica yopHOPEMN mutant strain MRS40(pIM417). This construct should produce a truncated YopD protein of 198 amino acids that still contains the hydrophobic domain but not the C-terminal amphipathic helix; however, no stable truncated product could be detected (Boland et al., 1996).

Construction of a yopB complementing plasmid

The yopB gene was cloned downstream of the yopE promoter and an optimalized Shine–Dalgarno (SD) sequence (AAGGAGGA), resulting in plasmid pCNR27. The construct was made in two steps. First, directed mutagenesis was performed according to Kunkel et al. (1987) on pCNR21, a plasmid carrying the yopE gene downstream of its own promoter (Sarker et al., 1998a). Oligonucleotide MIPA 394 (5′-AATGATGATATTTTCATGTCTAGATCCTCCTTTGGCTATTAA AACAAG-3′) introduced an XbaI site (underlined and bold) and a modified SD sequence (underlined), yielding plasmid pCNR25. The yopB gene was then extracted from plasmid pCN13 carrying gst–yopB and sycD (Neyt and Cornelis, 1999) using XbaI and HindIII restriction sites and was introduced into the corresponding sites of pCNR25 instead of yopE, giving plasmid pCNR27. The ATG of the yopB gene is 8 nucleotides downstream of the SD sequence.

Eukaryotic cell growth and infection conditions

PU5-1.8 (ATCC TIB 61) mouse monocyte macrophage cell lines were grown routinely in RPMI-1640 medium (Gibco) supplemented with 2 mM l-glutamine (Seromed), 10% fetal bovine serum (Gibco) and 100 μg ml−1 streptomycin (Sigma) at 37°C under 8% CO2. At 20 h before infection, cells (1.5 × 105 cells ml−1) were seeded in 24-well tissue culture plates (1 ml well−1).

Before infection, the cells were washed twice with RPMI-1640 medium (Gibco) supplemented only with 2 mM l-glutamine (Seromed). Bacteria were grown for 90 min at room temperature in BHI-OX and then transferred to 37°C for 2 h in order to induce yop synthesis. The bacteria were then washed and resuspended in prewarmed saline. Cells were then infected for 1 h at a multiplicity of infection (MOI) of 500. To facilitate contact between bacteria and cells, the plates were then centrifuged for 5 min at 400 × g. Alternatively, where indicated, cells were infected for up to 5 h at a MOI of 50 with bacteria that were pregrown in BHI for 2 h at room temperature but not heat induced.

Lactate dehydrogenase (LDH) release

Before infection, PU5-1.8 macrophages were washed and incubated further with RPMI-1640 without phenol red. Lactate dehydrogenase release determinations were performed using the CytoTox 96 cytotoxicity assay kit (Promega). After infection, the plate was centrifuged for 4 min at 250 × g, and aliquots of media (50 μl) were transferred to a 96-multiwell plate. The substrate was then added and, after 20–30 min of incubation, the reaction was terminated by the addition of 50 μl of stop solution, and the OD at 490 nm was determined. As a positive control for total cell-associated LDH, macrophages were lysed with 0.09% Triton X-100. As negative controls, we used uninfected cells and cells infected with a pYV strain.

BCECF release

The BCECF release assay was based on the principle described by Bhakdi et al. (1989). Just before infection, cells were washed twice with 1 ml of PBS and labelled by incubation with 10 μM BCECF-AM [2′,7′-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescein, acetoxymethyl ester; Molecular Probes] for 20 min at 37°C. Cells were then washed twice with RPMI without phenol red and, after 1 h infection, they were pelleted by centrifugation at 250 × g for 4 min. The cell culture supernatant was transferred into 2 ml cuvettes, and fluorescence was measured using a spectrofluorimeter (Versafluor; Bio-Rad) with an excitation wavelength of 490 nm and emission at 520 nm. Cells treated with Triton (0.09%) for 45 min were used as a positive control. Percentage of lysis was calculated using the following formula: % lysis = (sample–uninfected)/(Triton–uninfected) × 100.

Haemolysis assays

The haemolysis assay was performed essentially as described by Håkansson et al. (1996b). Overnight Yersinia cultures were diluted to an OD of 0.1 in BHI-OX medium supplemented with 20 mM MgCl2 and 0.4% glucose and grown for 1.5 h at room temperature and for 2 h at 37°C. Sheep erythrocytes were washed and centrifuged three times for 5 min at 2000 × g with cold PBS until the supernatant was essentially colourless, counted with a haemocytometer and diluted to 4 × 109 cells ml−1. Erythrocyte suspension (50 μl) was mixed with 108 bacteria in a round-bottomed 96-well plate and, after 10 min centrifugation at 1200 × g, the plate was incubated for 1.5 h at 37°C. The pellets were then resuspended in PBS to a total volume of 250 μl, and the plate was centrifuged for 15 min at 1200 × g. Supernatants (100 μl) were then transferred to a flat-bottomed 96-well plate, and the OD at 570 nm was measured. As a positive control, we used erythrocytes lysed with 50 μl of water. The percentage of haemolysis is determined by dividing the sample value by the OD at 570 nm of water-lysed erythrocytes and multiplying by 100.

Osmoprotection assays were performed in the presence of 30 mM dextran 6000 (Fluka) or 30 mM sucrose. The washed erythrocytes were resuspended in culture medium containing 60 mM carbohydrate to yield a final concentration of 30 mM during infection.

Dye exclusion experiments

Dye exclusion experiments were performed essentially as described by Kirby et al. (1998). Bacteria that were not heat induced were diluted to an OD of 1, and 100 μl of the suspension was added to 1 ml of cells grown on coverslips (1.5 × 105 cells ml−1), previously washed with RPMI containing 30 mM dextran 6000. After incubation for 3 h at 37°C, macrophages were washed once with ice-cold PBS–dextran 6000, incubated for 10 min at 0°C and then incubated with fluorescent dyes in RPMI–dextran 6000 for 1 h on ice. Macrophages were washed seven times with ice-cold PBS–dextran 6000, and the cells were fixed for 20 min with 3% paraformaldehyde in PBS–dextran 6000. After washing of the coverslips, samples were mounted in 50% Mowiol (Polysciences) containing 100 mg ml−1 diazabicyclooctane (DABCO; Sigma). Cells were examined under a Diaplan Leitz microscope. Slides were taken using phase-contrast optics or fluorescent illumination with a Wild Leitz MPS46/52 and then scanned. Dyes were used at the following concentrations: lucifer yellow CH, lithium salt (625 μg ml−1; Molecular Probes) and Texas red-X phalloidin (1 U per coverslip).

Acknowledgements

We acknowledge A. Boyd and S. Bhakdi for helpful discussions and suggestions at various stages of the work. We also thank P. Courtoy and F. Geuijen for critical reading of the manuscript. We are grateful to N. Grosdent, M. Iriarte and M.-P. Sory for providing genetic constructs, and to J.-F. Rémy for assisting with the genetic work. This work was supported by the Belgian ‘Fonds National de la Recherche Scientifique Médicale’ (convention 3.4595.97), the ‘Direction générale de la Recherche Scientifique-Communauté Française de Belgique’ (Action de Recherche Concertée 94/99-172) and the ‘Interuniversity Poles of Attraction Program–Belgian State, Prime Minister's Office, Federal Office for Scientific, Technical and Cultural affairs' (PAI 4/03).

Ancillary