Enterohaemorrhagic Escherichia coli haemolysin is cleaved and inactivated by serine protease EspPα

The haemolysin from enterohaemorrhagic Escherichia coli (EHEC-Hly) and the serine protease EspPα are putative virulence factors of EHEC. We investigated the interplay between these secreted factors and demonstrate that EspPα cleaves the 107 kDa large EHEC-Hly. Degradation was observed when purified EspPα was added to a growing culture of an EHEC-Hly-expressing strain, with isolated proteins and with coexpressing strains, and was independent of the EHEC serotype. EHEC-Hly breakdown occurred as a multistage process with the formation of characteristic fragments with relative molecular masses of ∼82 kDa and/or ∼84 kDa and ∼34 kDa. The initial cleavage occurred in the N-terminal hydrophobic domain of EHEC-Hly between Leu235 and Ser236 and abolished its haemolytic activity. In a cellular infection system, the cytolytic potential of EHEC-Hly-secreting recombinant strains was abolished when EspPα was coexpressed. EHEC in contact with human intestinal epithelial cells simultaneously upregulated their EHEC-Hly and EspP indicating that both molecules might interact under physiological conditions. We propose the concept of bacterial effector molecule interference (BEMI), reflecting the concerted interplay of virulence factors. Interference between effector molecules might be an additional way to regulate virulence functions and increases the complexity of monomolecular phenotypes.

Shiga toxins (Stx), the ribosome-inactivating AB 5 proteins with rRNA N-glycosidase activity (Sandvig, 2001), are regarded as the major virulence factors of EHEC responsible for microvascular endothelial damage which underlies the pathogenesis of HUS Tarr et al., 2005;Müthing et al., 2009). However, additional EHEC factors might also contribute to the pathogenesis of EHEC-mediated diseases (Brunder et al., 1997;Paton et al., 2004;Aldick et al., 2007;Gyles, 2007). Two such candidates are the serine protease EspP (Brunder et al., 1997) and the EHEC haemolysin (EHEC-Hly) (Schmidt et al., 1995;Aldick et al., 2007). EspP, a serine protease autotransporter of Enterobacteriaceae (SPATE), is among the most abundant secreted proteins of EHEC (Henderson and Nataro, 2001). This protein interacts with the coagulation cascade by cleaving factor V (Brunder et al., 1997) and with the complement system by degrading C3 and C5 (Orth et al., 2010). Besides this, a recent publication indicates that EspP might be involved in biofilm formation (Xicohtencatl-Cortes et al., 2010). The four recently identified EspP subtypes (a, b, g and d) differ in their transport and proteolytic capacities ; notably, the proteolytically active EspPa is produced by EHEC associated with severe human disease Khan et al., 2009). EHEC-Hly is a pore-forming cytolysin, which belongs to the RTX (repeat-in-toxin) family (Schmidt et al., 1995 and. It lyses erythrocytes from different species and bovine lymphocytes (Schmidt et al., 1995;Bauer and Welch, 1996). Moreover, EHEC-Hly injures microvascular endothelial cells, suggesting a possible role in the pathogenesis of HUS (Aldick et al., 2007). Recently, we showed that EHEC-Hly are associated with outer membrane vesicles (OMVs) shed by EHEC. This mediates an up to 80 times increased stability and prolonged activity of the toxin as compared with its free, vesicle-unbound form (Aldick et al., 2009).
It has been shown that the interaction with bacterial proteases modulates biological activity of other members of the RTX family (Nagamune et al., 1996). We observed that EHEC strains of different serotypes harbouring the structural gene for EHEC-Hly differ by the presence and amount of EHEC-Hly in culture supernatants. To gain deeper insight into this phenomenon, we sought potential interaction partners of EHEC-Hly among the EHEC virulence factors. Here, we demonstrate that EspPa degrades and inactivates EHEC-Hly.

Different expression of EHEC-Hly in wild-type EHEC strains
We observed substantial differences between EHEC strains in the amount of EHEC-Hly present in culture supernatants. Specifically, we detected full-length EHEC-Hly via immunoblot in supernatants of several EHEC serotypes, namely O6 : HNT, ONT : H25, O22 : H8, O84 : HNT, O113 : H21, O156 : NM and O163 : H19, but not EHEC O157 : H7 or O26 : H11/NM (Table 1). Interestingly, all strains containing detectable amounts of EHEC-Hly in supernatants harboured espPb, encoding the nonproteolytic EspPb subtype, whereas the strains lacking EHEC-Hly in culture supernatants harboured espPa, encoding the highly proteolytic EspPa   (Table 1). We therefore hypothesized that the observed differences in the amount of EHEC-Hly present in these culture supernatants might be caused by an interaction of EHEC-Hly with EspPa, in particular that EspPa cleaves and thereby degrades EHEC-Hly. To test this hypothesis, we used recombinant EHEC-Hly and/or EspPa in order to eliminate the effect of other wild-type EHEC-secreted molecules that could influence their interactions.

Serine protease EspPa cleaves EHEC-Hly
To investigate possible interactions of EHEC-Hly with EspPa, we combined the two toxins with each other as isolated proteins, expressed them simultaneously or supplemented cultures containing EHEC-Hly with EspPa. We also studied if serotype-specific differences influence the interaction. Experimental conditions are summarized in Table 2 and strain constructs are described in Table 3.
In a first approach, we supplemented an early logphase culture of clone TA48 producing recombinant EHEC-Hly from EHEC O157 : H7 with 5 mg ml -1 purified, recombinant EspPa from EHEC O157 : H7 or with an EspPa-buffer control ( Table 2, panel A-O157 : H7-I) and continued incubation at 37°C for 2 h. Immunoblot analysis of trichloroacetic acid (TCA)-precipitated sterile supernatants using anti-EHEC-Hly polyclonal antibody demonstrated the occurrence of two immunoreactive breakdown products with relative molecular masses (Mr) of 84 Ϯ 4 kDa and 34 Ϯ 3 kDa in TA48 treated with EspPa (Fig. 1A, lane 2). These fragments were not present in TA48 supernatant treated with EspPa-buffer control (Fig. 1A, lane 1), indicating proteolytic cleavage of the 107 kDa large EHEC-Hly by EspPa. EspPa remained unaffected by EHEC-Hly as determined by immunoblot using an anti-EspP antibody. In addition, EspPa did not cross-react with the anti-EHEC-Hly antibody (data not shown).
To study the EspPa-mediated proteolytic cleavage of EHEC-Hly in more detail, we moved to the level of isolated recombinant proteins. Ammonium sulfateprecipitated EHEC-Hly from clone TA48 was incubated with purified, recombinant EspPa for 4 h at 37°C and analysed by immunoblot using anti-EHEC-Hly antibody ( Table 2, panel B-O157 : H7-I). EHEC-Hly incubated with EspPa displayed two immunoreactive breakdown products of Mr~84 kDa and~82 kDa (Fig. 1C, lane 2). The Mr 34 kDa breakdown product was not observed at this time point (data not shown). Notably, prolonged incubation of EHEC-Hly with EspPa (> 18 h) led to the loss of the EHEC-Hly degradation bands. This indicated that after the initial formation of the Mr~84 kDa and~82 kDa breakdown products, there was further degradation (at least of the immunoreactive portion). Again, EspPa remained unaffected during the incubation with EHEC-Hly confirming the former observation.
We further sought to determine if the interaction of EspPa with EHEC-Hly is limited to isolated proteins or if it also occurs in cultures that coexpress both proteins ( Table 2, panel C-O157 : H7-I). To this aim, sterile TCAprecipitated supernatants of 11 h and 13 h cultures of strains TA144 and TA145 were analysed by immunoblot using EHEC-Hly and EspPa antibodies. The supernatant of clone TA145, where proteolytically active EspPa was coexpressed with EHEC-Hly displayed, besides the 107 kDa band of the intact EHEC-Hly, again two breakdown products of Mr~84 kDa and~82 kDa (Fig. 1D, lane 2), while the Mr~34 kDa fragment was not observed at this time point. This is in accordance with the results observed for coincubated recombinant pro-teins (Fig. 1C). The specific breakdown products were not present in supernatant of clone TA144, in which the non-proteolytic EspPa mutant S263A (see Experimental procedures) was secreted simultaneously with EHEC-Hly (Fig. 1D, lane 1). Together, these data further affirm that the immunoreactive breakdown products resulted from the cleavage of EHEC-Hly by proteolytically active EspPa. The expression and secretion of EspPa and the non-proteolytic EspPa mutant in constructs TA145 and TA144, respectively, were confirmed by immunoblotting.

EspPa cleavage cuts off haemolytic activity of EHEC-Hly
To assess the functional consequences of EspPamediated cleavage for the biological activity of EHEC-Hly, we analysed the haemolytic activity of the sterile culture supernatant of EHEC-Hly-producing clone TA48 after the growing culture was supplemented with EspPa for 2 h (Table 2, panel A-O157 : H7-II). The EspPa treatment reduced the haemolytic activity of supernatant TA48 to 67%, compared to the EspPa-buffer controltreated sample (Fig. 1B), indicating that cleavage of EHEC-Hly eliminates its haemolytic activity. To further a. X indicates each particular set of experimental conditions used for studying the interaction of EspPa and EHEC-Hly. b. n.a.: not applicable due to experimental conditions. c. Freshly prepared culture supernatants of clones TA48 and TA50 were used in this experiment to ensure full haemolytic activity of EHEC-Hly. d. In addition, EspPa from O26 : H11 was used to perform the respective experiment. EspPa cleaves EHEC-Hly in bacterial culture. EHEC-Hly-producing strain TA48 was grown to early log phase and supplemented either with EspPa or with the EspPa-buffer control and incubated for further 2 h. (A) Sterile supernatants were TCA-precipitated, separated in SDS-PAGE and analysed in immunoblot using anti-EHEC-Hly antibody. The arrows indicate the 107 kDa band of EHEC-Hly (white arrow) or the specific Mr~84 kDa and~34 kDa breakdown fragments of EHEC-Hly (black arrows). The very weak immunoreactive band with a slightly higher Mr than that of the~84 kDa specific cleavage product (#) was present in all EHEC-Hly control preparations (see also C and D) and was therefore considered a background signal. (B) The sterile culture supernatants were assayed for their haemolytic activity, which was calculated as percentage of haemolysis (see Experimental procedures). Data in (B) are presented as means Ϯ standard deviations of three independent assays. Statistically significant differences (P < 0.01, Student's t-test) are indicated by asterisk. C-E. Cleavage and inactivation of recombinant EHEC-Hly from EHEC O157 via EspPa from EHEC O157. Immunoblot analysis using anti-EHEC-Hly antibody of (C) recombinant isolated EHEC-Hly after incubation with buffer (control) or with recombinant purified EspPa, and (D) TCA-precipitated supernatants of clones TA144 and TA145 coexpressing recombinant EHEC-Hly and either EspPa (TA145) or the non-proteolytic EspPa mutant S263A (TA144). The arrows indicate the 107 kDa band of intact EHEC-Hly (white arrow) and the Mr~84 kDa and the~82 kDa EHEC-Hly cleavage products (black arrows). (E) Haemolytic activity of sterile culture supernatants of clones TA144 and TA145 after 11 h and 13 h of growth calculated as percentage of haemolysis. Data are presented as means Ϯ standard deviations of at least three independent assays. Statistically significant differences between haemolytic activity of TA144 and TA145 (P < 0.01, Student's t-test) are indicated by asterisks.
confirm this result, we tested culture supernatants of clones coexpressing EHEC-Hly and either EspPa (TA145) or the non-proteolytic mutant EspPa (TA144) for their haemolytic activities ( Table 2, panel C-O157 : H7-II). Clone TA144 produced at different time points (11 h and 13 h) approximately fourfold higher haemolysis (57% versus 16% and 100% versus 27% respectively) than clone TA145 (Fig. 1E), demonstrating that the cleavage of EHEC-Hly by EspPa results in a significantly (P < 0.01) reduced haemolytic activity. In addition, sterile supernatant of clone TA48 containing EHEC-Hly was supplemented with 4 mg ml -1 recombinant EspPa or the respective buffer control, incubated at 37°C for different time intervals (0-120 min) and then tested for residual haemolytic activity (Table 2, panel B-O157 : H7-II). The activity of the EspPa-treated sample was reduced to 50% after about 35 min and totally abolished (1%) after 2 h as compared to its initial activity ( Fig. 2A).
The EHEC-Hly-containing supernatant exposed to the EspPa-buffer control also showed a reduction of haemolytic activity (from 100% to 62%) over time ( Fig. 2A). This loss of activity in the control sample is due to the relatively short half-life time of the free EHEC-Hly itself, as reported previously (Aldick et al., 2009), the phenomenon that is thought to be caused by the irreversible self-aggregation of the toxin. Taken together, the rapid and complete loss of haemolytic activity of the EspPa-treated EHEC-Hly further underlines the capacity of EspPa to abolish the biological activity of EHEC-Hly by endoproteolytic cleavage.

Analysis of EHEC-Hly breakdown fragments using mass spectrometry
To verify the identity of the EHEC-Hly cleavage products and to estimate which domains might be affected, we analysed the immunoreactive breakdown fragments (Mr 84 kDa,~82 kDa and~34 kDa) by peptide mass fingerprinting using matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS). In all of the three breakdown products, EHEC-Hly-specific peptides located in the C-terminal end of the protein were identified, indicating that after the formation of the Mr 84 kDa fragment further degradation leads to the generation of the smaller EHEC-Hly breakdown products. As an example, a representative peptide mass spectrum of the Mr~82 kDa fragment after tryptic digest is shown in Fig. 3A. In this EHEC-Hly fragment we identified 14 EHEC-Hly-specific peptides ranging in length between 6 and 24 amino acids resulting in a sequence coverage of 14.5% (Table 4, 1-14). The detected EHEC-Hly-specific peptides were located between amino acids 255 and 939 of the 998-amino-acid-large EHEC-Hly molecule (Fig. 3B). Several fragments were chemically modified because of reactions during sample preparation or data acquisition or formed matrix adducts (indicated in Table 4). The potential region of proteolytic cleavage by EspPa was determined to be located in the N-terminal end of EHEC-Hly as indicated by the arrow in the schematic illustration of the toxin (Fig. 3B). This is further underlined by the theoretical mass of 80.3 kDa of the Fig. 2. A. Haemolytic activity of EHEC-Hly is cut off by EspPa cleavage. EHEC-Hly-containing sterile culture supernatant from strain TA48 was incubated for indicated time intervals (0-120 min) with EspPa ( ) or EspPa-buffer control (᭝) at 37°C and residual haemolytic activity was determined and calculated as percentage of haemolysis as described in Experimental procedures. Data are presented as means Ϯ standard deviations of three independent assays. B. EHEC-Hly associated with OMV is cleaved by EspPa. Immunoblot analysis of OMV-associated EHEC-Hly incubated with EspPa for 15 min to 26 h using anti-EHEC-Hly antibody. The white arrow indicates the 107 kDa band of intact EHEC-Hly, and the black arrows mark the Mr 84 kDa, the~82 kDa and the~34 kDa EHEC-Hly cleavage products. Incubation times of OMV-associated EHEC-Hly with EspPa are given at the top of the figure.
EHEC-Hly fragment ranging from amino acid 255 to 998, which corresponds well to the observed migration pattern in gel electrophoresis (Fig. 1A). Further analysis of non-EHEC-Hly signals revealed the presence of eight peptides specific for EspPa (Table 4, peptides E1-E8). These peptides are entirely located in the N-terminal end of the serine protease and represent the previously described autodegradation product of EspPa (Dutta et al., 2002). This EspPa-specific fragment, which is commonly found in all EspPa isolations, has a Mr of~86 kDa and accompanied also the isolated Mr~82 kDa breakdown product of EHEC-Hly due to a similar electrophoretic separation property as evidenced by immunoblotting analysis (data not shown).
Attempts to identify the potential EHEC-Hly cleavage site by N-terminal sequencing were unsuccessful, probably because of comigrating EspPa autodegradation products, leading to ambiguous sequencing results. Therefore, we employed two overlapping synthetic peptides covering the EHEC-Hly sequence from residue 191 to 250 and incubated each peptide overnight with EspPa and S263A as a negative control. The formation of peptide fragments by EspPa-mediated cleavage was analysed by MALDI-TOF-MS and displayed a specific mass signal at m/z 1475.7 corresponding to a peptide fragment where EspPa cleavage would occur after Leu 235 . Re-examination of peptide mass fingerprinting data of the Mr~82 kDa fragment showed, in addition, a signal at m/z 1862.9 which corresponds to residues Ser 236 to Lys 254 ; this further supports the above data that cleavage occurs after Leu 235 resulting in a newly formed N-terminus at Ser 236 ( Fig. 3 and Table 4).

EspPa from EHEC O26 : H11 cleaves and inactivates EHEC-Hly from EHEC O26 : H11
To assess whether or not the cleavage and inactivation of EHEC-Hly by EspPa is limited to serotype O157 : H7, we isolated both toxins from EHEC of serotype O26 : H11. EHEC-HlyO26 (originating from clone TA50) and EspPaO26 (originating from wild-type EHEC O26 : H11 strain 5631/ 96) were coincubated for 4 h and the mixture was assayed for haemolytic activity. Similar to observations for the proteins from EHEC O157 : H7, EHEC-HlyO26 was inactivated by EspPaO26 as demonstrated by loss of its haemolytic activity (data not shown). Hence, cleavage and inactivation of EHEC-Hly by EspPa is not restricted to the O157 : H7 serotype.

Interaction of EspPa from EHEC O157 : H7 and EHEC-Hly from EHEC O26 : H11
To determine if cleavage of EHEC-Hly by EspPa is limited to proteins derived from the same serotype (i.e. O157 : H7 or O26 : H11) or if this effect is also observed between proteins from different serotypes, we investigated the interaction of EspPa from EHEC O157 : H7 with EHEC-Hly from EHEC O26 : H11. Similar to the above experiments, the interaction was studied by supplementation of growing cultures containing EHEC-Hly with purified EspPa (Table 2, panel A-O26 : H11-I and II), co-incubation of isolated recombinant proteins (Table 2, panel B-O26 : H11-I and II) and coexpression of recombinant plasmids using clones TA143 (expressing EHEC-Hly and EspPa) and TA142 (expressing EHEC-Hly and the mutant, non-proteolytic, EspPa) ( Table 2, panel C-O26 : H11-I and II). These experiments demonstrated that EHEC-Hly from EHEC O26 : H11 is cleaved and its haemolytic activity is abolished via EspPa from EHEC O157 : H7 (Fig. S1). The results were comparable to those shown above for the respective proteins from the same serotype ( Fig. 1A-C). This indicates that the cleavage and functional inactivation of EHEC-Hly via EspPa is serotype-independent and suggests a general principle underlying the interaction of these two proteins.

EspPa cleaves and inactivates EHEC-Hly associated with OMVs
Recently, we reported that EHEC-Hly secreted by EHEC rapidly binds to OMVs shed by the bacteria and that this OMV association stabilizes the toxin and significantly prolongs its haemolytic activity (Aldick et al., 2009). This prompted us to investigate whether OMV-associated EHEC-Hly might be protected from cleavage by EspPa. OMVs carrying EHEC-Hly freshly isolated from strain TA50 were incubated with 2 mg ml -1 EspPa for different time intervals (15 min to 26 h) at 37°C and subsequently analysed in immunoblot using anti-EHEC-Hly antibody ( Table 2, panel D-O26 : H11-I). Notably, already 15 min after exposure to EspPa the characteristic Mr~84 kDa immunoreactive breakdown product appeared and remained visible up to 19 h, while the band corresponding to the intact OMV-associated EHEC-Hly steadily diminished (Fig. 2B). The fragments of Mr~34 kDa and 82 kDa, the latter of which remained less defined, were present after 1 h of incubation and diminished over time. After > 19 h, all four immunoreactive bands completely disappeared (Fig. 2B). This indicates that the association with OMV does not protect EHEC-Hly from proteolysis by EspPa.
In accordance with these immunoblot data, exposure to EspPa for 20 h (Table 2, panel D-O26 : H11-II) completely ablated the haemolytic activity of OMV-associated EHEC-Hly. Outer membrane vesicle-associated EHEC-Hly causes haemolysis first after 15 h of incubation with erythrocytes probably because of the necessary maturation process of vesicles (Aldick et al., 2009). Because EspPa rapidly cleaves OMV-associated EHEC-Hly (within 4 h) (Fig. 2B), complete absence of haemolytic activity of EspPa-treated OMV-associated EHEC-Hly results from degradation of EHEC-Hly before vesicle maturation.

Chronology of EHEC-hlyA and espP expression
To determine if EHEC-Hly and EspP could theoretically come into direct contact during infection, we studied the chronology of EHEC-hlyA and espP expression in four selected EHEC strains producing EHEC-Hly together with EspPa (O157 : H7 strain EDL933 and O26 : H11 strain 5236/96) or EspPb (O6 : HNT strain 3503/98 and O163 : H19 strain 36/03; Table 1) under laboratory and in vivo mimicking conditions. All four strains were cultured in LB broth under standard laboratory conditions (37°C, rotary shaker) for 2 h, 4 h, 8 h, 12 h, 16 h, 20 h and 24 h, respectively, and mRNA transcription level of EHEC-hlyA and espP were determined relative to the housekeeping gene gapA (used as a reference) in quantitative reverse transcription (RT)-PCR. The strains showed in general slightly elevated expression of both EHEC-hlyA and espP with an increase of up to threefold relative to the 2 h time point (Fig. S2), except for strain 5236/96 where EHEC-hlyA expression was not increased. Upregulation peaked at the high cell density (> 16 h); in strain 3503/98 (O6 : HNT) the espP expression was increased up to eightfold. No significant difference was observed between strains expressing proteolytically active EspPa (EDL933, 5236/96) or inactive EspPb (3503/98, 36/03).
To investigate the expression of both virulence factors under conditions mimicking the situation during human infection, we developed an intestinal cell infection assay. Human ileo-caecal epithelial cells  were infected with overnight cultures of the same set of EHEC strains tested above and incubated at 37°C under static conditions for 2 h, 4 h, 8 h, 12 h, 16 h, 20 h and 24 h. Relative transcription of EHEC-hlyA and espP mRNA was again normalized to gapA (Fig. 4). In the presence of intestinal epithelial cells, expression of both EHEC-hlyA and espP was significantly upregulated compared with LB media. After 12 h, EHEC-hlyA expression increased by two-to fivefold relative to the starting conditions (2 h time point). Similarly, espP was significantly upregulated after 12 h leading to 15-to 20-fold increase; in strain 5236/96 (O26 : H11) upregulation reached even > 35-fold (Fig. 4). Again, no significant difference between strains expressing proteolytically active EspPa and inactive EspPb was observed.
Together, these data demonstrate that EHEC-hlyA and espP are expressed simultaneously both under laboratory conditions and under in vivo mimicking conditions, suggesting that EHEC-Hly and EspP could come together during infection. The more pronounced and simultaneous upregulation of both virulence factors after contact of EHEC with intestinal epithelial cells, the first barrier encountered by the organisms during human disease, indicates that both molecules may be effectors during infection.

Biological consequence of EHEC-Hly cleavage by EspP
To investigate if the cleavage of EHEC-Hly by EspP might have biological implications during an EHEC infection, we used a model of human brain microvascular endothelial cells (HBMECs). The brain microvasculature is a target during EHEC-mediated HUS and these cells are highly susceptible to the cytolytic effect of EHEC-Hly (Aldick et al., 2007). Cultured HBMECs were first exposed to supernatants of overnight cultures of clones TA145/TA144 and TA143/TA142 producing EHEC-Hly from EHEC O157 and EHEC O26, respectively, together with proteolytically active/inactive EspPa, and HBMEC lysis was monitored by measuring release of intracellular lactate dehydrogenase (LDH). In both pairs, the clones expressing EHEC-Hly together with inactive EspPa (TA144 and TA142) caused lysis of HBMECs which was low, but significantly higher than that caused by the clones that express EHEC-Hly together with the proteolytically active EspPa (TA145 and TA143) (Fig. 5A). To enhance the effect of EHEC-Hly on the target cells, we next infected the HBMEC monolayers with living bacteria which continuously produce the toxin, increasing therefore the probability of its interaction with the cells. After 4 h exposure of HBMECs to the bacterial cultures and subsequent post-incubation for 18 h, clones TA144 and TA142 caused a pronounced dosedependent lysis of HBMECs, whereas clones TA145 and TA143 caused only baseline and dose-independent LDH release ( Fig. 5B and C). This suggests that cleavage of EHEC-Hly by EspPa might prevent the lytic effect of the toxin on the target cells.

Discussion
Most studies of EHEC virulence factors have focused on isolated molecules such as Shiga toxins (Hurley et al., 2001;Lee et al., 2008), cytolethal distending toxin , subtilase toxin (Paton et al., 2004;Chong et al., 2008), EHEC-Hly (Bauer and Welch, 1996;Schmidt et al., 1996;Aldick et al., 2007;) and EspP (Brunder et al., 1997). Although such an approach is a prerequisite for understanding exact mechanisms of action of particular virulence factors, its obvious drawback is that potential interactions between the various effector molecules produced by an EHEC strain are not considered. Experiments restricted to single toxins might thus lead to inadequate conclusions and rather simplified interpretations of the toxins' contributions to the biological effects observed during infection.
An example for multifactorial protein interactions is the activation of toxin precursors via proteolytic cleavage, which can be mediated either by host or by bacterial proteases. More specific, the eukaryotic cellular protease furin cleaves the Stx A subunit, generating a catalytically active A 1 fragment with rRNA N-glycosidase activity (Garred et al., 1995). Similarly, cleavage of the last two amino acids from the A2 fragment of the A subunit of Stx2d via elastase in the intestinal mucus dramatically increases cytotoxicity of this toxin (Melton-Celsa et al., 2002;Bielaszewska et al., 2006). An example of a bacterial protease with a precursor-activating effect is the Vibrio cholerae haemagglutinin/protease which cleaves, and thereby activates, V. cholerae haemolysin, another member of the RTX family (Nagamune et al., 1996).
In contrast to these examples, the cleavage and inactivation of a bacterial toxin by a bacterial protease secreted from the same organism, has, to our knowledge, been demonstrated for the first time in the present study.
The data we present are likely to reflect a genuine biologic process, in view of the different situations in which we demonstrated this effect, including purified EspPa added to a strain expressing EHEC-Hly, effector molecules co-incubated as isolated proteins, and finally, the interaction of EspPa and EHEC-Hly coexpressed by recombinants. Regardless in which form both molecules Bacteria were harvested by centrifugation, RNA was isolated and transcription levels of EHEC-hlyA and espP were determined using RT-PCR and normalized to gapA. Upregulation of each gene expression relative to 2 h time point was determined using Student's t-test with *P < 0.05 and **P < 0.001. Data are means Ϯ standard deviations from two independent assays. are brought together, EHEC-Hly is degraded and functionally inactivated by EspPa. Furthermore, this effect is also observed when EspPa and EHEC-Hly are derived from different serotypes. The cleavage appears to be a multistage process in which, depending on the specific incubation time and conditions, breakdown products of different sizes are observed. The initial cleavage results in the formation of a Mr~82 kDa and/or an~84 kDa break-down fragment and proceeds further to generate a Mr 34 kDa fragment (Figs 1 and 2B). The time-course of the formation of the Mr~84 kDa,~82 kDa and the~34 kDa fragment is well documented and exemplarily shown for the OMV-associated EHEC-Hly (Fig. 2B). Specifically, the appearance of the Mr~34 kDa band after 1 h and its strong reduction already after 4 h of exposure of EHEC-Hly to EspPa (Fig. 2B) may explain why this fragment was observed neither after 4 h of co-incubation of the recombinant proteins (Fig. 1C) nor in the supernatant of clone TA145 after overnight growth (Fig. 1D). Prolonged incubation with EspPa finally leads to a complete proteolytic degradation of OMV-associated EHEC-Hly.
Analysis using mass spectrometry confirmed the identity of the Mr~84 kDa,~82 kDa and~34 kDa breakdown products as EHEC-Hly-specific. Notably, the initial cleavage of EHEC-Hly occurs at position Leu 235 in the early N-terminal end of the toxin. In this region a domain of predominantly hydrophobic amino acids is located (between amino acid 210 and 410) (Schmidt et al., 1995). This domain is proposed to be crucial for the biological activity of RTX toxins in general, as it is involved in the interaction with target cell membranes leading to pore formation (Welch, 2001). The cleavage of EHEC-Hly via EspPa in the region of the hydrophobic domain thus plausibly explains the immediate reduction of its haemolytic activity, as the functionality of this domain is most probably abolished. Interestingly, we were not able to detect any larger fragments of the N-terminal hydrophobic domain using MALDI-TOF-MS, indicating that this region might be degraded to smaller peptides. However, we cannot exclude the possibility that this region is not recognized by the employed polyclonal anti-EHEC-Hly antibody.
Currently, the physiological significance of the cleavage and inactivation of EHEC-Hly by EspPa remains unclear. Prolonged exposure of EHEC bacteria to EHEC-Hly might, for example, directly or indirectly cause detrimental effects to the producing bacteria and EspPa might serve as a specific inactivator for EHEC-Hly. Although we were not able to detect such direct antibacterial toxic effects under laboratory conditions, we cannot exclude the possibility that EHEC are affected by their own haemolysin in a physiological environment. An animal model of acute haemorrhagic colitis, ideally with systemic thrombotic microangiopathy, would be very useful in establishing the relevance of our findings to mammalian infection. However, the human host is a singular victim of EHEC infections from STEC of the serotypes studied, and a suitable model does not, regrettably, exist, for us to perform such experiments (Mundy et al., 2006). Our results using in vivo mimicking systems substantiate the possible relevance of the findings for the in vivo situation. Analysis of the EHEC-hlyA and espP expression demon- Fig. 5. Consequences of EHEC-Hly cleavage by EspPa on cytolytic activity of EHEC-Hly. HBMEC monolayers were exposed to supernatants of overnight cultures of clones TA145, TA144, TA143 and TA142 which produce EHEC-Hly together with proteolytically active EspPa (TA145, TA143) or proteolytically inactive EspPa mutant S263A (TA144, TA142) (A) or to decreasing doses (3 ¥ 10 8 to 7.5 ¥ 10 7 cfu) of bacterial cultures of the respective clones (B and C) as described in Experimental procedures. HBMEC lysis was monitored by measuring release of intracellular LDH using the CytoTox 96 kit. The differences between HBMEC lysis caused by clones expressing proteolytically active EspPa and the non-proteolytic S263A were determined using Student's t-test with *P < 0.05 and **P < 0.001. Data are means Ϯ standard deviations from four measurements. strated that both toxins are moderately upregulated especially at higher cell densities. Intriguingly, expression of both toxins was substantial increased when EHEC were grown in direct contact with cultured epithelial cells from human colon. This suggests that both toxins might be expressed during infection, in particular after EHEC colonize the large intestinal epithelium. Since EHEC-hlyA and espP were expressed simultaneously in all four tested strains, the two toxins could stand a realistic chance to physically interact during natural infection. Abolishment of the EHEC-Hly cytolytic activity towards the brain microvascular endothelium using recombinant strains coexpressing EspPa further demonstrates that EspPa may act as an inactivator for EHEC-Hly in cells that are affected during extra-intestinal complications of EHEC infection, such as HUS. Even though cell culture models have certain limitations, our results suggest that the observed cleavage and inactivation of EHEC-Hly might also take place in vivo during infection.
Although its significance during human infection is presently unknown, the cleavage and inactivation of EHEC-Hly via EspPa suggest that pathogens can regulate their virulence phenotypes by interference of effector molecules. Therefore, we propose the concept of bacterial effector molecule interference (BEMI), reflecting the concerted interplay of proven and potential virulence factors, thus increasing the complexity of monomolecular phenotypes. Further studies addressing in detail different aspects of the interaction of the virulence factors with their target cells (e.g. the way and the kinetics of the cellular delivery) are clearly required to fully understand the significance of the interaction of EHEC-Hly and EspP (and perhaps also other virulence factors) in the context of human infections.
Fourteen EHEC isolates analysed for the presence of EHEC-Hly in culture supernatants are listed in Table 1. The presence of the structural gene for EHEC-Hly (EHEC-hlyA) and the espP subtypes in these strains (all isolated in our laboratory) were determined using PCR (Schmidt et al., 1995;Brockmeyer et al., 2007). Expression of EHEC-Hly was confirmed by the strains' ability to produce enterohaemolytic phenotype on enterohaemolysin agar (Schmidt et al., 1995). Production of EspP and its proteolytic activity were determined using an immunoblot and the ability of culture supernatants to cleave an oligopeptide substrate, respectively, as described previously .

Coexpression of EHEC-Hly and EspPa
Strains TA142, TA143, TA144 and TA145 were grown overnight on enterohaemolysin agar, inoculated into 150 ml of Cleavage and inactivation of EHEC-Hly via EspPa 1337 LB broth supplemented with 100 mg ml -1 ampicillin and 30 mg ml -1 kanamycin and incubated for 11 h and 13 h at 37°C and 180 r.p.m. Cell free supernatants were either used directly in haemolytic activity assay or 1 ml was precipitated using 10% TCA (1 h, 4°C), and the precipitate was sedimented (20 000 g, 30 min, 4°C), resuspended in 30 ml of sample buffer (20 mM Tris HCl, pH 8.0) and separated electrophoretically for immunodetection.

Sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting analyses
Samples including TCA-precipitated supernatants (see above) of clones TA142, TA143, TA144 and TA145 or of wild-type EHEC strains (Table 1), and ammonium sulfateprecipitated EHEC-Hly treated with EspPa or with a buffer control were analysed using standard SDS-PAGE (Laemmli, 1970). Separated proteins were transferred to PVDF membrane (Carl Roth, Karlsruhe, Germany) using Trans-Blot SD (Bio-Rad, Munich, Germany) semi-dry blotting system (1 h, 50 mA). Different primary antibodies and a secondary antibody were used as listed above. Detection was achieved using the chromogenic substrates 5-Bromo-4-chloro-3′indolylphosphate p-toluidine salt and p-Nitro-Blue tetrazolium chloride (both from Carl Roth) and signal intensities determined by densitometry (Quantity One®, Bio-Rad, Munich, Germany).

Haemolysis assay
Haemolytic activity was quantified as described previously (Aldick et al., 2007), with slight modifications. Briefly, 950 ml of sterile culture supernatant of strain TA50 containing 5 mM CaCl2 was supplemented with 50 ml of an EspPa preparation (described above) and incubated at 37°C for indicated time points (0-120 min). Subsequently, 100 ml of an erythrocyte suspension (10% human erythrocytes washed thrice with PBS) was added and incubation was continued for 4 h at 37°C with gentle shaking. Erythrocytes were sedimented by centrifugation (400 g, 5 min) and absorbance of the clear supernatant was measured at OD570. A microtitre assay was used to quantify haemolytic activity of OMV-associated EHEC-Hly (Aldick et al., 2009). Briefly, 100 ml of OMV preparations were diluted twofold in the assay buffer (0.9% NaCl containing 10 mM CaCl2) in a 96-well plate and 50 ml of an erythrocyte suspension was added. The plate was incubated at 37°C with gentle shaking for 20 h, and the erythrocytes were sedimented by centrifugation (400 g, 5 min). Clear supernatants were transferred to a fresh plate and OD570 was measured. Either LB broth (for supernatants) or HEPES buffer (for OMVs) were used as a background and distilled water as a total lysis control. The percentage of haemolysis was calculated for both assays as follows: % haemolysis = (OD570 of sample -OD570 of background)/(OD570 of total -OD570 of background) ¥ 100.

Peptide mass fingerprinting using MALDI-TOF-MS
Samples containing the EHEC-Hly breakdown fragments were separated electrophoretically, transferred to a PVDF membrane and a stripe of the membrane was immunostained using anti-EHEC-Hly antibody. This served to identify the characteristic breakdown fragments. Electrophoretically separated proteins were incubated with 300 ng of sequencing grade trypsin (Roche, Mannheim, Germany) in 50 mM NH4HCO3 for 5 h at room temperature according to Shevchenko and colleagues (1996) with minor modifications (Müthing et al., 2004). Saturated a-cyano-4hydroxycinnammic acid in 50% acetonitril and 0.1% trifluoracetate was applied to the digest as a matrix. MALDI-TOF-MS was performed with a prototype of a prOTOF 2000 mass spectrometer by Perkin Elmer (from Sciex, Concordia, Canada) in positive ion mode. A pulsed N2-laser (l = 337 nm) was used for ultraviolet matrix-assisted laser desorption/ ionization orthogonal time-of-flight (UV-MALDI-o-TOF) mass spectrometry. Identification of trypsin-digested peptides was performed by comparing molecular ions [M+H] + (m/zexp.) with ions of theoretically digested fragments of EHEC-Hly (m/ztheor.) using ProteinProspector (UCSF Proteomic tools, v4.27.2) and the ALDENTE peptide mass fingerprinting tool (available at the ExPASy proteomics server from the SWISS Institute of Bioinformatics SIB).

Digestion of synthetic EHEC-Hly peptides
For a more detailed analysis of EHEC-Hly cleavage sites two overlapping synthetic peptides covering the EHEC-Hly sequence from residue 191 to 250 (sequence derived from EHEC O157 : H7 EDL 933) were custom synthesized (Selleck, Houston, Texas). Sequences of peptides were: Ehly1 (FSEQLNQLGS FLSSKPRLS SVGGKLQNLPD) (covering residue 191-220) and Ehly2 (QNLPDLGPLG DGLDV-VSGIL SAVSASFILG NSDAH) representing residue 216-250. One hundred microlitres of a 2 mM solution of each peptide dissolved in PBS was incubated overnight with 2.5 mg of EspPa or the negative control S263A at 37°C overnight. Peptide solutions were desalted using C18 SPE cartridges (Phenomenex, Germany, Aschaffenburg) and stepwise elution. Mass spectrometric analysis was performed on a prOTOF 2000 mass spectrometer by Perkin Elmer (from Sciex, Concordia, Canada) in positive ion mode as described above.

Chronology of EHEC-hlyA and espP expression
Chronology of the expression of EHEC-hlyA and espP in selected EHEC strains producing EHEC-Hly together with EspPa (O157 : H7 strain EDL933 and O26 : H11 strain 5236/ 96) or EspPb (strains of serotypes O6 : HNT and O163 : H19; Table 1) was investigated using quantitative RT-PCR using two different experimental conditions. First, the strains were grown in LB broth at 180 r.p.m. and 37°C for 2 h, 4 h, 8 h, 12 h, 16 h, 20 h and 24 h. At each time interval bacteria were harvested by centrifugation and RNA was isolated as described below. Second, the expression of each toxin was analysed after contact of the bacteria with human intestinal epithelial cells (HCT-8) to simulate a situation during infection. HCT-8 cells (human ileo-caecal adenocarcinoma epithelial cells; ATCC CCL-244) were grown in RPMI 1640 (Lonza, Cologne, Germany) supplemented with 10% fetal calf serum (PAA, Pasching, Austria), 2 mM L-glutamine and 1 mM sodium pyruvate (Lonza). For the experiment, cells were seeded (2.5 ¥ 10 5 cells per well) into six-well plates (Corning, Corning, NY, USA) and grown until~70% confluence. Cells were washed three times with phosphate-buffered saline (PBS), infected with 2 ml (1 ¥ 10 8 cfu) of bacterial overnight cultures in LB broth and incubated with the bacteria (37°C, 5% CO2, 0.5% mannose in cell culture medium) for 2 h, 4 h, 8 h, 12 h, 16 h, 20 h and 24 h. To recover at each time point both bacteria that adhered to the cells during the incubation period and those that remained non-adherent, medium from the cells in each well was collected and the cells were harvested by trypsinization. Both the cell culture medium and trypsinized cells were centrifuged (2000 g, 10 min, 4°C), and pellets were pooled, washed twice with PBS and used for RNA extraction.

Total RNA isolation
Total bacterial RNA was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to manufacturer's instruction. To eliminate contamination by genomic DNA, all RNA samples were treated with DNase I (Roche Diagnostics, Mannheim, Germany). The concentration of RNA was determined and its quality was checked by measuring optical density at 260 nm.

Quantitative RT-PCR
A one-step quantitative RT-PCR, performed with an iCycler iQ-5 (Bio-Rad, Munich, Germany) and 2 ¥ SensiMix™ SYBR One-Step Kit (Peqlab Biotechnologie, Erlangen, Germany), was used to measure the relative expression of EHEC-hlyA and espP mRNA. The PCR reactions were performed in a 96-well plate using a 20 ml volume containing 1 ml of total RNA (100 ng), 10 ml of 2 ¥ SensiMix™ SYBR One-Step, 0.4 ml of RNase Inhibitor and 200 nM of each primer. Primers HlyA9 (5′-ACC CCA GGA GAA GAA GTT AG-3′) and HlyA4 (5′-TCT CGC CTG ATA GTG TTT GGT A-3′), EspP-RTF (5′-GCT CCA CCC TGA AAC TAC CG-3′) and EspP-RTR (5′-CGT TCA AGT GCC TGC TGT TT-3′), and GapA_forward and GapA_reverse (Blumer et al., 2005) were used to amplify EHEC-hlyA, espP and gapA (encoding D-glyceraldehyde-3phosphate dehydrogenase A) used as a reference respectively. The one-step RT-PCR included a reverse transcription step at 42°C for 10 min, and a polymerase activation and preliminary denaturation step at 95°C for 10 min, followed by 35 cycles of denaturation at 95°C for 15 s, annealing at 57°C for 15 s and extension at 72°C for 20 s. A melting curve analysis to confirm the specificity of the amplification products was constructed with continuous fluorescence reading from 55°C to 95°C. Data were analysed using the Bio-Rad iQ5 standard edition optical system software V2.0. EHEC-hlyA and espP mRNA levels were normalized to gapA mRNA and compared with the 2 h time point. Quantitative RT-PCRs were performed three times with two independent RNA preparations.

Cell lysis assay
Cytolytic activity of clones TA142, TA143, TA144 and TA145 was determined using cultured HBMECs (Stins et al., 1997), which are sensitive to EHEC-Hly-mediated lysis (Aldick et al., 2007). Cell lysis was determined by measuring release of LDH from HBMECs exposed to bacterial cultures or supernatants as described previously (Aldick et al., 2007) with slight modifications. Briefly, HBMECs were seeded (1 ¥ 10 4 cells per well) into 96-well plates in Endothelial medium (PAA, Pasching, Austria) and grown until confluence. To investigate LDH release caused by bacteria, 100 ml of twofold dilutions (3 ¥ 10 8 to 7.5 ¥ 10 7 cfu) of overnight bacterial cultures grown in LB broth supplemented with 5 mM CaCl2, ampicillin (100 mg ml -1 ) and kanamycin (30 mg ml -1 ) were added to the cells (A) or to empty wells (D) and incubated (37°C, 5% CO2) for 4 h. Bacterial cultures were then removed and cells were incubated for additional 18 h (37°C, 5% CO2) in cell culture medium supplemented with 5 mM CaCl2 and gentamicin (100 mg ml -1 ). Untreated cells (C), cells exposed to 0.9% Triton X-100 (added to the cells 45 min before termination of the incubation) (B) and wells with 100 ml of medium without cells (E) served as controls. To test lytic activity of supernatants, HBMEC monolayers were incubated for 18 h (37°C, 5% CO2) with 100 ml of sterile-filtered supernatants of the above cultures diluted 1:2 in cell culture medium using the same controls. After incubation, the plates were centrifuged (250 g, 5 min), 50 ml of the samples were transferred into a new microtiter plate and the LDH activity was determined using the CytoTox 96 kit (Promega, Mannheim, Germany) according to the manufacturer's instructions. The absorbance was measured at 490 nm (Dynex microplate reader) and LDH release was calculated as follows: [(A -C) -(E -D)]/ (B -C) ¥ 100.

Statistical analysis
The statistical analysis of results was performed using Student's t-test; P Յ 0.05 was considered significant.