*For correspondence. E-mail firstname.lastname@example.org; Tel. (+44) 121 333 8746; Fax (+44) 121 333 8701. †The first two authors contributed equally to this paper.
Type III secretion allows bacteria to inject effector proteins into host cells. In enteropathogenic Escherichia coli (EPEC), three type III secreted proteins, EspA, EspB and EspD, have been shown to be required for translocation of the Tir effector protein into host cells. EspB and EspD have been proposed to form a pore in the host cell membrane, whereas EspA, which forms a large filamentous structure bridging bacterial and host cell surfaces, is thought to provide a conduit for translocation of effector proteins between pores in the bacterial and host cell membranes. Type III secretion has been correlated with an ability to cause contact-dependent haemolysis of red blood cells (RBCs) in vitro. As EspA filaments link bacteria and the host cell, we predicted that intimate bacteria–RBC contact would not be required for EPEC-induced haemolysis and, therefore, in this study we investigated the interaction of EPEC with monolayers of RBCs attached to polylysine-coated cell culture dishes. EPEC caused total RBC haemolysis in the absence of centrifugation and osmoprotection studies were consistent with the insertion of a hydrophilic pore into the RBC membrane. Cell attachment and haemolysis involved interaction between EspA filaments and the RBC membrane and was dependent upon a functional type III secretion system and on EspD, whereas EPEC lacking EspB still caused some haemolysis. Following haemolysis, only EspD was consistently detected in the RBC membrane. This study shows that intimate bacteria–RBC membrane contact is not a requirement for EPEC-induced haemolysis; it also provides further evidence that EspA filaments are a conduit for protein translocation and that EspD may be the major component of a translocation pore in the host cell membrane.
Most bacterial virulence factors are either located at the bacterial surface or are secreted from the bacterium, sometimes directly into host cells. Translocation of virulence proteins into host cells is dependent upon possession of a specialized secretion system, a type III secretion system, which has been found in many Gram-negative bacteria (Hueck, 1998). Enteropathogenic Escherichia coli (EPEC) is a bacterial pathogen that uses a multistage infection strategy involving type III secretion and protein translocation into host cells (Frankel et al., 1998). An established aetiological agent of human infantile diarrhoea, EPEC subverts intestinal epithelial cell function to produce distinctive ‘attaching and effacing’ (A/E) lesions, characterized by localized destruction (effacement) of brush border microvilli, intimate bacterial attachment to the host cell membrane and formation of an actin-rich cytoskeletal structure beneath intimately attached bacteria (Frankel et al., 1998); similar lesions are produced in a variety of cultured epithelial cell lines (Knutton et al., 1987). All the genes required for A/E lesion formation are encoded by the LEE (locus of enterocyte effacement) pathogenicity island (McDaniel and Kaper, 1997), which encodes the type III secretion apparatus (Jarvis et al., 1995), the intimate EPEC adhesin, intimin (Jerse et al., 1990a), secreted proteins EspA, EspB and EspD (Frankel et al., 1998), and a translocated intimin receptor, Tir (Kenny et al., 1997).
Type III secretion systems demonstrate a broad functional conservation across bacterial species and many of the components show similarity to proteins involved in flagellar biosynthesis (Hueck, 1998). Type III secretion systems (secretons) have recently been visualized using electron microscopy (Kubori et al., 1998; Blocker et al., 1999; Tamano et al., 2000); they comprise a macromolecular complex spanning both bacterial membranes with an external needle structure (Blocker et al., 1999; Tamano et al., 2000). In addition to a type III secreton, injection of proteins into host cells is thought to require insertion of pore-forming proteins into the host cell membrane (Hueck, 1998). In Yersinia, these are the type III secreted YopB/D proteins (Neyt and Cornelis, 1999), and in Shigella, these are the IpaB/C proteins (Blocker et al., 1999). Injection of proteins into host cells by Shigella and Yersinia spp. has been correlated with their ability to cause contact-dependent haemolysis of red blood cells (RBCs) in vitro when bacteria are brought into close contact with the RBC membrane by centrifugation (Clerc et al., 1986; Hakansson et al., 1996).
The EPEC type III secreton has yet to be visualized although, based on homology with other type III secretion system proteins, it probably has a structure similar to that described for Shigella, i.e. a macromolecular complex spanning both bacterial membranes (Blocker et al., 1999). Four proteins essential for A/E lesion formation are known to be secreted by the EPEC type III secreton, EspA, EspB, EspD and Tir. Tir is translocated and inserted into the host cell membrane in which it functions as a receptor for the intimate bacterial adhesin, intimin (Kenny et al., 1997; Deibel et al., 1998); EspA, EspB and EspD are thought to be components of the translocation apparatus. Secreted EspB (Wolff et al., 1998) and EspD (Wachter et al., 1999) have been shown to be delivered to the host cell membrane and, based on homology with Ipa B/C and Yop B/D, are proposed to form a pore complex in the host membrane (Frankel et al., 1998). EspA is the major component of a large filamentous structure that is proposed to provide a conduit between the type III translocon and a host cell membrane pore (Knutton et al., 1998).
EPEC was recently shown to exhibit a contact-dependent haemolytic activity (Warawa et al., 1999). However, because long EspA filaments form a bridge between the bacterium and the host cell, we predicted that EPEC should not require centrifugation and close bacteria–host cell contact in order to interact with RBCs and cause haemolysis. In this study, using RBC monolayers, we show that EspA filaments mediate the binding of EPEC to RBCs and that close contact is not a requirement for EPEC-induced haemolysis; haemolysis was consistent with insertion of a hydrophilic pore into the RBC membrane and, following haemolysis, EspD was associated with the RBC membrane.
EPEC-induced haemolysis does not require intimate bacteria–RBC membrane contact
Yersinia- and Shigella-induced haemolysis requires centrifugation of bacteria and RBCs in order to reduce the distance between bacterial and red cell membranes below a critical threshold. This has been termed contact-dependent haemolysis (Clerc et al., 1986; Hakansson et al., 1996) and contact-induced haemolysis was recently demonstrated with EPEC (Warawa et al., 1999). However, as long EspA filaments were shown to connect bacteria to the host cell during protein translocation (Knutton et al., 1998), we predicted that intimate bacteria–RBC contact and, thus, centrifugation should not be required for EPEC-induced haemolysis. We therefore developed a haemolysis assay equivalent to a typical cell culture adhesion assay using monolayers of RBCs attached to cell culture dishes. Wild-type EPEC strain E2348/69 was incubated with RBC monolayers for up to 6 h at 37°C and the release of haemoglobin was monitored as described in Experimental procedures. Haemoglobin release was negligible after 1 h but increased with time and was maximal (> 90% haemolysis) after 4 h (Figs 1A and 2). Attachment to polylysine-coated plates did not contribute to haemolysis as similar levels of haemolysis occurred in the absence of polylysine when RBCs were allowed to settle and attach to cell culture-treated dishes. However, the formation of tightly attached RBC monolayers made the assay much easier to perform and more reproducible. In this assay system, there was no significant additional effect of centrifuging bacteria onto RBC monolayers and haemolysis of human RBCs was not affected by blood group type (A, B, AB and O) (data not shown). Enterohaemorrhagic E. coli (EHEC) strain 85–170 (O157:H7) was also tested and shown to induce a non-contact-dependent haemolysis of human RBCs (Figs 1A and 2).
Blocker et al. (1999) recently demonstrated that haemolytic activity of Shigella could be prevented by addition of osmotic protectants to the medium, a mechanism consistent with the insertion of a hydrophilic pore into the RBC membrane; addition of molecules too large to pass through a membrane pore counterbalance the increased intracellular pressure, thereby reducing haemolysis. We performed similar experiments and Fig. 1B shows the effect on haemolysis caused by EPEC strain E2348/69 and EHEC strain 85–170 in the presence of different osmoprotectants at a concentration of 30 mM. As was the case with Shigella, molecules larger then PEG1000 yielded significant protection against lysis and protection increased with the size of the molecule.
Role of the type III secretion system, type III secreted proteins and other virulence proteins in EPEC-induced haemolysis
An association of the EPEC type III secretion system with haemolytic activity was confirmed using a type III secretion mutant (EscC) of E2348/69; no significant haemoglobin release was detected with this strain (Fig. 1C). The same assay was used to assess the involvement in haemolysis of EPEC type III secreted proteins. EPEC with single mutations in genes encoding secreted proteins were examined and, as previously reported using a contact-dependent haemolysis assay (Warawa et al., 1999), tir and espF mutants induced levels of haemolysis similar to the wild-type strain, whereas no lysis was observed with espA and espD mutants. In each case, haemolysis was restored, albeit not to levels of the wild type, when the deleted gene was reintroduced on a plasmid (Fig. 1C). Interestingly, an espB mutant reproducibly caused some (≈25%) haemolysis that increased to ≈40% when the espB gene was reintroduced (Fig. 1C). We have previously shown that an espD mutant makes barely detectable EspA filaments that could theoretically be functional if brought into intimate contact with the host cell membrane. However, centrifugation of this strain onto RBC monolayers did not induce haemolysis (data not shown). EPEC strains lacking the EPEC adherence factor (EAF) plasmid (strain JPN15) and intimin (strain CVD206) induced levels of haemolysis similar to wild-type E2348/69 (Fig. 1C).
We have previously shown that the coiled-coil domain of EspA is important for assembly of functional EspA filaments on the surface of EPEC; disruption of the coiled-coil domain of EspA by a double radical amino acid substitution (strain UMD872(pICC27)) prevented EspA filament formation, protein translocation and A/E lesion formation (Delahay et al., 1999). However, a single radical amino acid substitution (strain UMD872(pICC25)) was insufficient to totally disrupt the coiled-coil domain and A/E lesion formation, but resulted in greatly shortened EspA filaments similar to those produced by an espD mutant. Both strains were tested for their ability to induce haemolysis. No lysis was observed with strain UMD872 (pICC27), whereas strain UMD872(pICC25) routinely produced low levels (≈10%) of lysis; interestingly, this figure increased to ≈40% when this strain was centrifuged onto the RBC monolayer. Centrifugation of the EspA-minus strain (UMD872) onto RBC monolayers did not induce haemolysis.
EspA filaments mediate EPEC binding to RBCs
Phase-contrast microscopy revealed bacterial attachment to RBCs of all the haemolytic strains (wild type, EAF plasmid mutant, eae mutant, tir mutant, espF mutant, espB mutant), but no adhesion of the non-haemolytic strains (espA mutant, espD mutant); bacterial adhesion was primarily as individual bacteria and not bacterial microcolonies, as is typical of adhesion to cultured cells (Fig. 2). The plasmid-encoded bundle-forming pilus (Bfp) (Giron et al., 1991), intimin and EspA filaments are known to play a role in EPEC adhesion to cultured cells. In order to determine which adhesin was important in EPEC adhesion to RBCs, we performed quantitative adhesion assays using wild-type (E2348/69), intimin-minus (CVD206), Bfp-minus (JPN15) and EspA-minus (UMD872) strains; lack of intimin or Bfp had no effect on EPEC adhesion to RBCs, whereas lack of EspA filaments resulted in a total loss of bacterial adherence (Fig. 3).
The importance of EspA filaments in RBC adhesion was confirmed using scanning (SEM) and transmission electron microscopy (TEM), which revealed bacterial attachment to both intact and lysed RBCs mediated by large filamentous surface appendages characteristic of EspA filaments; immunofluorescence and immunogold labelling confirmed these structures as EspA filaments (Fig. 4). Typical EspA filaments promoted RBC attachment of all the haemolytic strains including the wild-type EPEC (Fig. 4B) and tir, espB (Fig. 4E), and espF mutants, whereas the espA and espD mutants did not adhere to RBCs and were non-haemolytic. Strain UMD872(pICC25), which produces barely detectable EspA filaments and low levels of haemolysis, showed a more intimate attachment of bacteria to the RBC membrane, but vestigial EspA filaments detected by immunofluorescence (Delahay et al., 1999) were not seen using scanning electron microscopy (Fig. 4F). RBC attachment of EHEC strain 85–170 was also promoted by morphologically similar filaments (Fig. 4G). Interestingly, using gold-labelling TEM, there appeared to be a short ≈50 nm section of the EspA filament structure adjacent to the bacterial surface that did not label with the EspA antiserum (Fig. 4D, inset).
EspD is transferred to RBC membranes
EPEC-induced haemolysis is consistent with insertion of a pore into the red cell membrane. In order to determine which bacterial proteins might become associated with the RBC membrane during haemolysis, we isolated RBC membranes following haemolysis using sucrose density gradient centrifugation according to the method of Blocker et al. (1999). Infections were performed using wild-type EPEC (E2348/69) and also strains deficient in EspA (UMD872), EspB (UMD864) and EspD (UMD870). Membrane proteins from each of the infections were separated using sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE), blotted and probed with antibodies against EspA, EspB, EspD and Tir. No bacterial proteins were detected in membranes of RBCs infected with UMD872, UMD864 or UMD870. Only EspD was found consistently associated with membranes infected with E2348/69 (Fig. 5); EspB was also detected very faintly, although this result was inconsistent in repeated experiments (data not shown). EspD, but not EspB, was also associated with red cell membranes infected with EHEC strain 85–170 (Fig. 5).
Immunofluorescence staining of RBC membranes following infection with E2348/69 revealed no staining with the EspB antibody, but a clear punctate staining of some RBC membranes was seen with the EspD antibody (Fig. 6).
Using a static infection of RBC monolayers, this study has shown that, unlike the contact-dependent haemolysis of Yersinia and Shigella, close bacteria–RBC contact is not a requirement for EPEC- or EHEC-induced haemolysis; important for haemolysis, however, was type III secretion, bacterial attachment by EspA filaments and EspD translocation. The EAF virulence plasmid, intimin, Tir and EspF played no role in haemolytic activity.
Haemolysis of RBCs has been correlated with type III protein secretion/translocation, although the structural basis for protein translocation has yet to be fully elucidated for any type III secretion system (Hueck, 1998). In EPEC, the proposed translocation apparatus (translocon) consists of pores in the bacterial and host cell membrane connected by a hollow EspA filament (Frankel et al., 1998) that would provide a continuous channel from the bacterial to the host cell cytosol. Such a model is consistent with the observed EPEC-induced haemolysis as EspA filaments are essential for haemolysis, EspA filaments mediate the interaction of EPEC with the RBC membrane and one putative pore-forming protein, EspD, was localized to the RBC membrane following haemolysis. The involvement of long EspA filaments explains why close bacteria–RBC contact is not a requirement for EPEC-induced haemolysis and contrasts with Yersinia and Shigella which do not possess EspA filaments or related structures. The Shigella secreton consists of a macromolecular complex spanning both bacterial membranes with a 60 nm long external needle structure (Blocker et al., 1999). Hence, the close Shigella–host cell contact required for protein translocation and close Shigella–RBC contact required for haemolysis. Interestingly, EspA immunogold labelling following haemolysis revealed, in addition to EspA-mediated bacterial attachment to RBCs, a short (≈50 nm) external part of the filament structure that did not stain with the EspA antibody; this could be the equivalent external needle structure of the EPEC type III secreton onto which EspA is polymerized. We are currently investigating the protein structure of the EPEC type III secreton including this putative needle complex.
Using non-conservative amino acid substitution, we recently demonstrated the importance of a carboxy terminal EspA coiled-coil domain in EspA filament assembly and function; single substitutions generated mutants defective in filament assembly but which still retained some functional protein translocation and the ability to induce A/E lesion formation, whereas double substitutions totally abolished EspA filament assembly, protein translocation and A/E lesion formation (Delahay et al., 1999). In this study, strains with double substitutions were non-haemolytic, whereas strains with a single substitution and which produced very short EspA filaments were weakly haemolytic consistent with retained functionality. More interesting, however, was the observation that haemolysis was increased a further ≈fourfold when this strain was centrifuged onto the RBC monolayer, i.e. when very short EspA filaments were brought into intimate contact with the RBC membrane.
Interaction of EHEC strain 85–170 (O157:H7) with RBCs also involved long filamentous structures. Although these filaments could not be confirmed as EspA because they did not react with our EPEC EspA antiserum, they do have all the characteristics of EspA filaments; thus, this is the first description of EspA-like filaments produced by EHEC serotype O157:H7.
EspA filaments have been shown to promote attachment of EPEC and EHEC to cultured cells (Ebel et al., 1998; Knutton et al., 1998). This data also support an adhesive role for EspA filaments as bacteria–RBC attachment appears to depend solely on the presence of EspA filaments. Strains that lacked other recognized EPEC adhesins (Bfp, intimin) but which produced EspA filaments adhered to RBCs as effectively as wild-type EPEC, whereas strains that were unable to produce EspA filaments (espA, espD mutants) did not adhere to human RBCs. This contrasts with adhesion to cultured cells in which espA and espD mutants do adhere in localized aggregates, probably mediated by Bfp. If this is the case, this would indicate that Bfp receptors are lacking on RBC membranes.
Based on osmoprotection experiments, Shigella was recently shown to insert a ≈22 Å hydrophilic pore in the RBC membrane (Blocker et al., 1999); the osmoprotection experiments reported here are consistent with a similar mechanism for EPEC- and EHEC-induced haemolysis, although the detailed analysis to determine a predicted pore size has yet to be performed. Five EPEC proteins, EspA, EspB, EspD, EspF and Tir, have been shown to be secreted by the EPEC type III secretion system and EspB, EspD and Tir have been shown to be translocated to the host cell membrane (Kenny et al., 1997; Wolff et al., 1998; Wachter et al., 1999). Based on homology with the pore-forming proteins secreted by Yersinia and Shigella, respectively, EspB and EspD have been proposed as candidate pore-forming proteins secreted by EPEC (Frankel et al., 1998). In this study, EspD was required for EPEC-induced haemolysis and was the only RBC membrane-associated protein detected following haemolysis, suggesting that EspD may play an important role in pore formation. EspB, on the other hand, was detected very faintly or not at all in the RBC membrane fraction following haemolysis. In the case of Yersinia and Shigella, both YopB/YopD and IpaB/IpaC are required for pore formation (Blocker et al., 1999; Neyt and Cornelis, 1999). It could also be the case with EPEC that both EspB and EspD are involved in pore formation but that EspB was not consistently detected in this study owing to the sensitivity of our assay system. However, the observation that an espB mutant still produced normal EspA filaments, adhered to RBCs by EspA filaments and caused some haemolysis, suggests that EspB is required for full haemolytic activity but is not an absolute requirement for haemolysis. Contrary to our observations, Warawa et al. (1999), using a contact-dependent haemolysis assay, reported that an espB mutant was non-haemolytic. This difference possibly reflects our more sensitive haemolysis assay which yielded > 90% haemolysis with wild-type strains compared with a maximum of only 15% in the Warawa study.
In conclusion, this study has shown that intimate bacteria–RBC membrane contact is not a requirement for EPEC- and EHEC-induced haemolysis. The data also provides further support for a model of the EPEC and EHEC translocon in which pores in the bacterial and host cell membrane are connected by hollow EspA filaments, and suggests that EspD may be the major component of a translocation pore inserted in the host cell membrane.
Bacterial strains and plasmids
The EPEC and EHEC strains and plasmids used in this study are listed in Table 1. Stock cultures of the strains were subcultured in Luria broth (LB) or LB supplemented with kanamycin (100 μl ml−1) or ampicillin (100 μl ml−1) as appropriate and incubated aerobically for 18 h at 37°C.
Assay protocol. Human blood was obtained from laboratory volunteers. Red blood cells (RBCs) were sedimented, washed three times in phosphate-buffered saline (PBS) and a 3% suspension added to polylysine-coated 30 mm tissue culture dishes for 20 min. Non-attached RBCs were removed by further washing with PBS and the resulting RBC monolayer was covered with 2 ml of HEPES buffered Dulbecco's modified Eagle's medium (DMEM) without phenol red. An overnight broth culture (30 μl) was added to each RBC monolayer and the dishes incubated for up to 6 h at 37°C after which the culture medium was transferred to a microfuge tube and the bacteria sedimented. Supernatants were monitored for haemoglobin release by measuring the optical density at 543 nm. Supernatants from uninfected RBC monolayers incubated under the same conditions were used to provide a baseline level of haemolysis (B); total haemolysis (T) was obtained from monolayers incubated with a 30-fold dilution of PBS. Percentage haemolysis (P) was calculated from: P = [(X–B)/(T–B)] × 100, in which X is the optical density of the sample analysed. Contact-dependent haemolysis was performed essentially as above except dishes were incubated for 2 h, centrifuged (2000 g × 5 min) and then incubated for a further 2 h. The results are the mean of three independent experiments. Errors given are standard deviations.
Osmoprotection studies. Haemolysis assays were performed as described above except that 30 mM osmoprotectant (sucrose, M. Wt. 350; raffinose, M. Wt. 600; PEG, M. Wt. 1000, 2000 and 3000) was added to the assay medium.
RBC membrane isolation
The method for membrane isolation was adapted from Blocker et al. (1999). Human blood was obtained from laboratory volunteers. Tris-buffered saline (TBS) containing a protease inhibitor cocktail (Roche) was used throughout the procedure. RBCs were pelleted and washed three times with TBS. The sedimented RBCs were resuspended in TBS at approximately 5 × 108 cells ml−1. Washed RBCs (2 ml) were then mixed with 0.4 ml of a 37°C overnight stationary culture of each of E2348/69, UMD872, UMD864, UMD870 and 85–170. TBS (2.4 ml) was added and the infections incubated at 37°C for 6 h. Distilled water (0.8 ml) was then added to obtain equal lysis in each infection. The samples were vortexed and spun to pellet any unlysed RBCs. Supernatant (3.4 ml) was then mixed with 6 ml of 72% sucrose and transferred to SW40 centrifuge tubes. The mixtures were then overlaid with 2 ml of 42% sucrose and 1 ml of 25% sucrose. The sucrose cushions were then spun at 15 000 g for 16 h. The RBC membranes were collected from the 25%/42% sucrose interface and spun at 45 000 g for 30 min at 4°C. The RBC membrane pellets were finally resuspended in 40 μl of TBS.
The RBC membrane preparation (10 μl) from each of the E2348/69, UMD872, UMD864, UMD870 and 85–170 infections were separated using SDS–PAGE and transferred to nitrocellulose membrane. The membrane was blocked with 10% skimmed milk in PBS, 0.05% Tween-20 for 1 h. The membranes were then washed in PBS–Tween and incubated with either anti-EspA, anti-EspB, anti-EspD or anti-Tir monoclonal antibodies diluted 1:100 in PBS–Tween overnight at 4°C (Hartland et al., 2000). The blots were washed and detected with anti-mouse-AP conjugated antibody as previously described (Knutton et al., 1998).
Light microscopy. RBC monolayers were examined for lysis and bacterial adhesion using phase-contrast microscopy following the removal of non-adherent bacteria by washing with PBS. Quantification of adhesion was performed using phase-contrast micrographs and counting numbers of bacteria adhering to 100 RBCs. The results, expressed as numbers of bacteria/RBC, are the mean of two separate experiments; errors given are standard deviations.
Immunofluorescence. Immunofluorescence was performed as previously described on fixed and permeabilized RBCs using polyclonal EspA and monoclonal EspB and EspD antisera (Knutton et al., 1998; Wolff et al., 1998).
Electron microscopy. For immunogold labelling of cell-associated bacteria, RBC monolayers on plastic (Therminox) coverslips were briefly fixed for 10 min in 0.1% glutaraldehyde, washed and incubated with EspA or EspD antiserum (1:100) for 4 h at room temperature. Cells were washed and incubated with 10 nm gold-labelled goat anti-mouse serum for 12 h at 4°C. After further thorough washing, cells were fixed in 3% buffered glutaraldehyde and processed for thin-section electron microscopy using standard procedures (Knutton, 1995). Samples were examined in a Jeol 1200EX electron microscope operated at 80 kV.
For scanning electron microscopy RBC monolayers prepared on glass coverslips were fixed with 3% glutaraldehyde, post-fixed in 1% osmium tetroxide, dehydrated through graded acetone solutions and critical-point dried. For immunogold labelling, monolayers were briefly fixed for 10 min in 0.1% glutaraldehyde, washed, incubated with EspA antiserum (1:100) for 2 h at room temperature, washed again and incubated with 30 nm gold-labelled goat anti-rabbit serum for 2 h. After further thorough washing, cells were fixed in 3% buffered glutaraldehyde and processed as described above. Mounted specimens were sputter-coated with platinum (Polaron) and examined in a Jeol 1200EX Scanning Transmission EM.
We thank James Kaper, Michael Donnenberg and Brenden Kenny for providing E2348/69 mutants and plasmids. This work was supported by The Wellcome Trust.
†The first two authors contributed equally to this paper.