Binding to The Target Cell Surface Is The Crucial Step in Pore Formation of Hemolysin BL from Bacillus cereus

A major virulence factor involved in Bacillus cereus food poisoning is the three-component enterotoxin hemolysin BL. It consists of the binding component B and the two lytic components L1 and L2. Studying its mode of action has been challenging, as natural culture supernatants additionally contain Nhe, the second three-component enterotoxin, and purification of recombinant (r) Hbl components has been difficult. In this study, we report on pore-forming, cytotoxic, cell binding and hemolytic activity of recently generated rHbl components expressed in E. coli. It is known that all three Hbl components are necessary for cytotoxicity and pore formation. Here we show that an excess of rHbl B enhances, while an excess of rHbl L1 hinders, the velocity of pore formation. Most rapid pore formation was observed with ratios L2:L1:B = 1:1:10 and 10:1:10. It was further verified that Hbl activity is due to sequential binding of the components B - L1 - L2. Accordingly, all bioassays proved that binding of Hbl B to the cell surface is the crucial step for pore formation and cytotoxic activity. Binding of Hbl B took place within minutes, while apposition of the following L1 and L2 occurred immediately. Further on, applying toxin components simultaneously, it seemed that Hbl L1 enhanced binding of B to the target cell surface. Overall, these data contribute significantly to the elucidation of the mode of action of Hbl, and suggest that its mechanism of pore formation differs substantially from that of Nhe, although both enterotoxin complexes are sequentially highly related.


Introduction
Bacillus cereus is considered an emerging pathogen relevant for food poisoning worldwide [1][2][3]. While the emetic toxin cereulide causes vomiting, enterotoxins are responsible for the diarrheal form [4]. The enterotoxin complexes Nhe (non-haemolytic enterotoxin [5]) and Hbl (haemolysin BL [6]) are composed of three individual components each, which is a unique feature of the Bacillus cereus group [4,7]. The genes encoding Nhe can be detected in nearly all enteropathogenic B. cereus strains, the hbl genes in about 45-65%. Strains bearing only hbl have not yet been found [8][9][10].

All Three Hbl Components Are Necessary for Cytotoxicity and Pore Formation
It is controversially discussed in literature whether or not two Hbl components alone might be able to exhibit toxic activity, or form a pore, as has been shown for the structurally-related Nhe components B and C [31]. In early spectrophotometric hemolysis assays, 37% hemolysis was detected with only Hbl B and L 2 ; and vascular permeability tests showed minor edema with only Hbl B and L 1 [18]. Other studies didn't detect any activity with only two Hbl components [25]. Figure 1 shows our results of WST-1 bioassays and propidium iodide (PI) influx tests on Vero cells with two rHbl components each. When combinations of rHbl L 1 and B, L 2 and B, or L 2 and L 1 were applied to the cells, neither toxic activity nor pore formation could be observed. We gained the same results when PI influx tests were performed under consecutive conditions, i.e., for 1 h each ( Figure 1B). These data strongly indicate that any combination of only two of the three Hbl components is not able to induce pore formation, and thus, no toxic activity can be achieved on the tested target cells.
Toxins 2019, 11, x FOR PEER REVIEW 3 of 16 Figure 1 shows our results of WST-1 bioassays and propidium iodide (PI) influx tests on Vero cells with two rHbl components each. When combinations of rHbl L1 and B, L2 and B, or L2 and L1 were applied to the cells, neither toxic activity nor pore formation could be observed. We gained the same results when PI influx tests were performed under consecutive conditions, i.e., for 1 h each ( Figure 1B). These data strongly indicate that any combination of only two of the three Hbl components is not able to induce pore formation, and thus, no toxic activity can be achieved on the tested target cells. The rHbl components were mixed in 1:1:1 ratios and applied as serial dilution to the cells, starting with 75 pmol/mL. Shown are reciprocal titers, which were defined as the dilution necessary to gain 50% dead cells after 24 h. Titers below 20 indicate no specific toxic activity. (B) PI influx test. The rHbl components were mixed in 1:1:1 ratios and applied to the cells (37.5 pmol/mL each). Alternatively, cells were incubated for 1 h with one component, washed and then incubated for another h with the second component. Only for the sample with rHbl L2+L1+B, PI influx and thus, pore formation was measured.

Velocity of Pore Formation Depends on rHbl Concentration
To determine the influence of the amount of Hbl on pore formation, rHbl components were mixed in 1:1:1 ratios and applied in specific molar concentrations (68. 25, 37.5, 18.75, 9.38 and 4.69 pmol/mL each) to Vero cells. Increasing fluorescence was measured representing influx of propidium iodide ( Figure 2). Supernatant of B. cereus strain F837/76, which causes rapid PI influx, was used as control. It was shown that the start point of PI influx and thus, pore formation, depends significantly on the Hbl concentration. The rHbl components were mixed in 1:1:1 ratios and applied as serial dilution to the cells, starting with 75 pmol/mL. Shown are reciprocal titers, which were defined as the dilution necessary to gain 50% dead cells after 24 h. Titers below 20 indicate no specific toxic activity. (B) PI influx test. The rHbl components were mixed in 1:1:1 ratios and applied to the cells (37.5 pmol/mL each). Alternatively, cells were incubated for 1 h with one component, washed and then incubated for another h with the second component. Only for the sample with rHbl L 2 +L 1 +B, PI influx and thus, pore formation was measured.

Velocity of Pore Formation Depends on rHbl Concentration
To determine the influence of the amount of Hbl on pore formation, rHbl components were mixed in 1:1:1 ratios and applied in specific molar concentrations (68. 25, 37.5, 18.75, 9.38 and 4.69 pmol/mL each) to Vero cells. Increasing fluorescence was measured representing influx of propidium iodide ( Figure 2). Supernatant of B. cereus strain F837/76, which causes rapid PI influx, was used as control. It was shown that the start point of PI influx and thus, pore formation, depends significantly on the Hbl concentration.  Figure 1 shows our results of WST-1 bioassays and propidium iodide (PI) influx tests on Vero cells with two rHbl components each. When combinations of rHbl L1 and B, L2 and B, or L2 and L1 were applied to the cells, neither toxic activity nor pore formation could be observed. We gained the same results when PI influx tests were performed under consecutive conditions, i.e., for 1 h each ( Figure 1B). These data strongly indicate that any combination of only two of the three Hbl components is not able to induce pore formation, and thus, no toxic activity can be achieved on the tested target cells. The rHbl components were mixed in 1:1:1 ratios and applied as serial dilution to the cells, starting with 75 pmol/mL. Shown are reciprocal titers, which were defined as the dilution necessary to gain 50% dead cells after 24 h. Titers below 20 indicate no specific toxic activity. (B) PI influx test. The rHbl components were mixed in 1:1:1 ratios and applied to the cells (37.5 pmol/mL each). Alternatively, cells were incubated for 1 h with one component, washed and then incubated for another h with the second component. Only for the sample with rHbl L2+L1+B, PI influx and thus, pore formation was measured.

Velocity of Pore Formation Depends on rHbl Concentration
To determine the influence of the amount of Hbl on pore formation, rHbl components were mixed in 1:1:1 ratios and applied in specific molar concentrations (68. 25, 37.5, 18.75, 9.38 and 4.69 pmol/mL each) to Vero cells. Increasing fluorescence was measured representing influx of propidium iodide ( Figure 2). Supernatant of B. cereus strain F837/76, which causes rapid PI influx, was used as control. It was shown that the start point of PI influx and thus, pore formation, depends significantly on the Hbl concentration.

Excess of rHbl B Accelerates, Excess of rHbl L 1 Hinders Pore Formation
To determine the ratio of rHbl L 2 :L 1 :B required for maximum activity, PI influx tests on Vero cells were carried out. The three components were pre-mixed in the ratio 1:1:1 (Figure 3, shown in purple as control in every diagram) or with 2×, 5× and 10× excess (Figure 3 left) or depletion (Figure 3 right) of each single component. Excess of rHbl L 2 led to a delayed PI influx, while depletion of rHbl L 2 up to factor 5 seemed not to influence the speed of pore formation (Figure 3 upper row). On the other hand, a 2× excess of rHbl L 1 delayed PI influx significantly, at 5× and 10× excess, even no pore formation was detectable within 4 h. Depletion of L 1 accelerated pore formation compared to the 1:1:1 control (Figure 3 middle row). With regards to rHbl B, pore formation started earlier when this component was applied in excess, while 2× depletion of B delayed PI influx significantly (Figure 3 lower row). At ratios of 5:5:1 or 10:10:1 no more pore formation was detectable within 4 h. Altogether, pore formation by Hbl seems to occur faster the more Hbl B is present compared to L 1 and L 2 . On the contrary, excess of Hbl L 1 seems to hinder pore formation. Reducing the amount of Hbl L 2 seems not to be crucial, but high excess of that component also retards pore formation.

Excess of rHbl B Accelerates, Excess of rHbl L1 Hinders Pore Formation
To determine the ratio of rHbl L2:L1:B required for maximum activity, PI influx tests on Vero cells were carried out. The three components were pre-mixed in the ratio 1:1:1 (Figure 3, shown in purple as control in every diagram) or with 2×, 5× and 10× excess (Figure 3 left) or depletion ( Figure  3 right) of each single component. Excess of rHbl L2 led to a delayed PI influx, while depletion of rHbl L2 up to factor 5 seemed not to influence the speed of pore formation (Figure 3 upper row). On the other hand, a 2× excess of rHbl L1 delayed PI influx significantly, at 5× and 10× excess, even no pore formation was detectable within 4 h. Depletion of L1 accelerated pore formation compared to the 1:1:1 control (Figure 3 middle row). With regards to rHbl B, pore formation started earlier when this component was applied in excess, while 2× depletion of B delayed PI influx significantly (Figure 3 lower row). At ratios of 5:5:1 or 10:10:1 no more pore formation was detectable within 4 h. Altogether, pore formation by Hbl seems to occur faster the more Hbl B is present compared to L1 and L2. On the contrary, excess of Hbl L1 seems to hinder pore formation. Reducing the amount of Hbl L2 seems not to be crucial, but high excess of that component also retards pore formation.

Hbl Activity Is Due to Sequential Binding of The Components B -L 1 -L 2 as Well as Their Concerted Action
The binding mechanism of Hbl to the target cell surface is still not clear. Individual binding of each component to erythrocytes has been suggested [24], as well as sequential binding to Chinese hamster ovary cells in the order B -L 1 -L 2 [25]. To investigate this, PI influx tests on Vero cells were set up in the first place. Initially, Vero cells were incubated for 1 h with two rHbl components and after that, the third component was added. PI influx, and thus pore formation, started earlier, the earlier rHbl B was added. Exchange of L 2 and L 1 as the last component had only minor influence ( Figure 4A). In the next setup, Vero cells were washed three times in medium before addition of the third component. In this sequential approach, pore formation was only observed when rHbl B and L 1 were added before, and L 2 after washing ( Figure 4B). In the final experiment the single components were added individually, with three washing steps in between. It became obvious that the Hbl components indeed assemble sequentially, and that the binding order B -L 1 -L 2 is obligatory for pore formation ( Figure 4C).
In WST-1 bioassays rHbl was also only toxic when applied in the order B -L 1 -L 2 (see Figure S1A). To assess the significance of the individual components for the toxic activity of Hbl, rHbl L 2 , L 1 and B were either applied in a concentration of 3.75 pmol/mL per well, or as a dilution series of 1:2, starting with 3.75 pmol/mL. Each component was incubated for 45 min with the cells before washing and the addition of the next component. After the third component, cells were again washed two times and incubated with WST-1 for 1.5 h before measuring. Application of all three components individually as a dilution series resulted in a low reciprocal titer (136) compared to the dilution series of pre-mixed L 2 , L 1 and B (titer 1:248; see Figure 1). When only rHbl B was diluted, the titer was 107. Dilution of rHbl L 1 resulted in a titer of 1700, dilution of L 2 in a titer of 1183 ( Figure 4D). These data suggest that binding of Hbl B to the target cell surface is the crucial step for Hbl pore formation. High amounts of L 2 or L 1 do not advance pore formation and toxic activity, if there is not enough B bound to the cells. On the contrary, dilution of L 2 and especially L 1 even increase toxic activity, which supports the data from PI influx tests in Figure 3.
To further corroborate the conclusions about the sequential binding order, flow cytometry analyses were carried out. First, Vero cells were incubated successively with rHbl B, the correspondent mAb 1G8 and Alexa Fluor ® 488 goat anti mouse IgG. A significant increase of fluorescence was visible, compared to the negative control ( Figure 4E and Table 1), indicating that rHbl B specifically binds to the cell surface. In a subsequent approach, rHbl B and L 1 or L 2 were mixed 1:1 and applied to the cells. For detection, mAb 1B8 and Alexa Fluor ® 488 goat anti mouse IgG were used. The mAb 1G8 (see above), which is best suited for detecting Hbl B on cell surfaces, was replaced, as it cross-reacts with L 1 [30]. Interestingly, the number of fluorescence-positive cells was increased when rHbl B was chaperoned by L 1 ( Figure 4F and Table 1). To demonstrate the impact of rHbl L 2 , the neutralizing mAb 1H9 (anti L 2 ) was used, together with different combinations of the rHbl components. Fluorescence was barely detected when Vero cells were incubated with only mAb 1H9 (negative control), rHbl L 2 +1H9, rHbl B and L 2 +1H9 or rHbl L 1 and L 2 +1H9 ( Figure 4G and Table 1). However, a strong increase of fluorescence was observed when the cells were first incubated with rHbl B+L 1 and subsequently with rHbl L 2 +1H9 ( Figure 4G and Table 1).
3.02 ± 0.06 3.33 ± 0.09 93.66 ± 0.55 Repeating these experiments, PI was added to the cells. Only in the samples where mAb 1H9, L 2 , L 1 and B were used, approximately 71% of the Vero cells were PI positive (see Figure S1B). Nevertheless, cells were morphologically still intact, firstly indicating that mAb 1H9 does not hinder assembly of the Hbl pore and PI influx, but presumably a quick lysis of the cells. Secondly, this approach confirmed again that Hbl B alone is not able to form pores.
Altogether, these data show that rHbl B binds specifically to Vero cells and that the presence of rHbl L 1 enhances this binding. rHbl L 2 can only be detected at the cell surface when B and L 1 are already present.

Binding of Hbl B to The Cell Surface Is The Crucial Step for Pore Formation
To determine the time necessary for each Hbl component to bind to the target cells, further PI influx tests were performed. The rHbl components were successively applied to the cells in concentrations of 37.5 pmol/mL each. Application order was B -L 1 -L 2 . After adding the B and L 1 components, the cells were washed twice in medium. After addition of L 2 , measurement was immediately started. When B as well as L 1 were applied for 1 h, 30 or 15 min each, rapid PI influx was detected (data not shown).
Shorter incubation times were first tested for the binding component B, with rHbl L 1 being constantly applied for 10 min. Full PI influx was observed for 10 min incubation.
At 5, 4 and 3 min incubation time, PI influx decreased significantly, and at incubation times of less than three minutes, no more PI influx was detected ( Figure 5A). A similar pattern was observed when rHbl B as well as rHbl L 1 were applied with decreasing incubation times ( Figure 5B). On the contrary, adding rHbl B constantly for 10 min and decreasing the incubation times of L 1 led to only slightly reduced PI influx ( Figure 5C). This leads to the assumption that the binding of rHbl L 1 to B is quite rapid, as incubation times as low as 1 min were sufficient for pore formation as long as B was added for 10 min. Binding of rHbl B to the cell surface seemed to be more complex, but also possible and very efficient in 5-10 min. Due to immediate PI influx after the addition of L 2 , it is assumed that binding of this component is as rapid and efficient as the binding of L 1 to B. These data confirm that binding of rHbl B to the target cell surface is indeed the crucial step in Hbl pore formation.
slightly reduced PI influx ( Figure 5C). This leads to the assumption that the binding of rHbl L1 to B is quite rapid, as incubation times as low as 1 min were sufficient for pore formation as long as B was added for 10 min. Binding of rHbl B to the cell surface seemed to be more complex, but also possible and very efficient in 5-10 min. Due to immediate PI influx after the addition of L2, it is assumed that binding of this component is as rapid and efficient as the binding of L1 to B. These data confirm that binding of rHbl B to the target cell surface is indeed the crucial step in Hbl pore formation.

rHbl B Needs A Free C-Terminus for Optimum Cell Binding
To investigate the role of the N-and C-termini in Hbl activity, rHbl L2, L1 and B were used in PI influx tests, each supplied with a C-terminal or N-terminal strep-tag, respectively. Figure 6A shows that exchanging the tags on rHbl L1 and L2 did not lead to any significant differences in the velocity of pore formation compared to the initially used combination (purple). Interestingly, moving the tag on rHbl B from the N-to the C-terminus resulted in a significant delay of pore formation. That pore formation is decelerated became even more obvious when the Hbl components were applied consecutively with decreasing incubation times for rHbl B. While for rHbl B with an N-terminal tag (free C-terminus), incubation times as low as 4-5 min were sufficient for cell binding ( Figure 6B; see also Figure 5A), no activity was observed for rHbl B with C-terminal tag (free N-terminus) with incubation times lower than 20 min ( Figure 6B). On the other hand, no reduction of cytotoxic activity was observed in WST-1 bioassays after 24 h incubation ( Figure 6C). In subsequent flow cytometry analyses, the two different rHbl B proteins were applied to Vero cells with increasing incubation

rHbl B Needs A Free C-Terminus for Optimum Cell Binding
To investigate the role of the N-and C-termini in Hbl activity, rHbl L 2 , L 1 and B were used in PI influx tests, each supplied with a C-terminal or N-terminal strep-tag, respectively. Figure 6A shows that exchanging the tags on rHbl L 1 and L 2 did not lead to any significant differences in the velocity of pore formation compared to the initially used combination (purple). Interestingly, moving the tag on rHbl B from the N-to the C-terminus resulted in a significant delay of pore formation. That pore formation is decelerated became even more obvious when the Hbl components were applied consecutively with decreasing incubation times for rHbl B. While for rHbl B with an N-terminal tag (free C-terminus), incubation times as low as 4-5 min were sufficient for cell binding ( Figure 6B; see also Figure 5A), no activity was observed for rHbl B with C-terminal tag (free N-terminus) with incubation times lower than 20 min ( Figure 6B). On the other hand, no reduction of cytotoxic activity was observed in WST-1 bioassays after 24 h incubation ( Figure 6C). In subsequent flow cytometry analyses, the two different rHbl B proteins were applied to Vero cells with increasing incubation times ( Table 2). After only 5 min of incubation, it became obvious that rHbl B with N-terminal tag (free C-terminus) binds quicker and more effectively to the target cells than rHbl B with C-terminal tag (free N-terminus) ( Table 2 and Figure 6D). times ( Table 2). After only 5 min of incubation, it became obvious that rHbl B with N-terminal tag (free C-terminus) binds quicker and more effectively to the target cells than rHbl B with C-terminal tag (free N-terminus) (Table 2 and Figure 6D).
Altogether, pore formation is clearly delayed when the C-terminus of rHbl B is blocked. This is due to decelerated binding capacity to the target cell surface. Toxic activity after 24 h is not affected.    Altogether, pore formation is clearly delayed when the C-terminus of rHbl B is blocked. This is due to decelerated binding capacity to the target cell surface. Toxic activity after 24 h is not affected.

Hemolytic Activity of rHbl Confirms Results on Vero Cells
Analogous to previous experiments, the properties of rHbl on sheep blood agar were investigated. As expected, hemolytic activity was only detectable when all three rHbl components were present ( Figure 7A). Interestingly, no discontinuous hemolysis could be observed in any experimental setup with the recombinant proteins. In the next step, the rHbl components were applied to two or three stamp holes with approximately 3 mm distance to investigate diffusion properties. Each time, the hemolysis zone was shifted towards rHbl B ( Figure 7B). When rHbl was applied in 3 mm distance, or directly in the same stamp hole as the supernatant of strain F837/76, rHbl L 1 disrupted or prevented the outer hemolysis ring, while rHbl B indeed enhanced it, and rHbl L 2 had no detectable influence (ratios of 1:1; Figure 7C). Additionally, also sequential application of the rHbl components was tested. 10 µL of each component were successively applied to the stamp hole with an incubation time of 1 h each. The hemolysis zone was enlarged the earlier rHbl B was applied ( Figure 7D). This showed that-in accordance with our earlier results-rHbl B is the key to hemolytic activity.

Hemolytic Activity of rHbl Confirms Results on Vero Cells
Analogous to previous experiments, the properties of rHbl on sheep blood agar were investigated. As expected, hemolytic activity was only detectable when all three rHbl components were present ( Figure 7A). Interestingly, no discontinuous hemolysis could be observed in any experimental setup with the recombinant proteins. In the next step, the rHbl components were applied to two or three stamp holes with approximately 3 mm distance to investigate diffusion properties. Each time, the hemolysis zone was shifted towards rHbl B ( Figure 7B). When rHbl was applied in 3 mm distance, or directly in the same stamp hole as the supernatant of strain F837/76, rHbl L1 disrupted or prevented the outer hemolysis ring, while rHbl B indeed enhanced it, and rHbl L2 had no detectable influence (ratios of 1:1; Figure 7C). Additionally, also sequential application of the rHbl components was tested. 10 µ L of each component were successively applied to the stamp hole with an incubation time of 1 h each. The hemolysis zone was enlarged the earlier rHbl B was applied ( Figure 7D). This showed that -in accordance with our earlier results -rHbl B is the key to hemolytic activity.

Discussion
Studies on the mode of action of the three-component hemolysin BL from B. cereus have always been difficult. This is due to the facts that Hbl never appears in natural culture supernatants without Nhe [8][9][10], and that for a long time the generation of functional recombinant Hbl components in E. coli has not been possible [11]. With our recently purified rHbl proteins [30], we ultimately prove in the present study that all three Hbl components are necessary for both pore formation and cytotoxicity. One or two rHbl components did not exhibit any toxic activity, neither in WST-1 bioassays and PI influx tests nor on sheep blood agar plates. On the contrary, earlier

Discussion
Studies on the mode of action of the three-component hemolysin BL from B. cereus have always been difficult. This is due to the facts that Hbl never appears in natural culture supernatants without Nhe [8][9][10], and that for a long time the generation of functional recombinant Hbl components in E. coli has not been possible [11]. With our recently purified rHbl proteins [30], we ultimately prove in the present study that all three Hbl components are necessary for both pore formation and cytotoxicity. One or two rHbl components did not exhibit any toxic activity, neither in WST-1 bioassays and PI influx tests nor on sheep blood agar plates. On the contrary, earlier spectrophotometric hemolysis assays detected 37% hemolysis after incubation with only Hbl L 2 and B; and minor edema after application of only Hbl L 1 and B in vascular permeability assays were reported [18].
With respect to concentration ratios, PI influx tests showed rapid pore formation with up to 10× excess of rHbl B, while depletion of B hindered pore formation. Excess of Hbl L 1 also hindered pore formation and even hemolysis on sheep blood agar, while the amount of Hbl L 2 seemed not to be crucial. Excess of L 2 or L 1 as well as depletion of B also decreased toxic activity in WST-1 bioassays (see Figure S1C). All our results point to a ratio of L 2 :L 1 :B = 1:1:10 for the fastest pore formation and 1:2:2 for maximum cytotoxicity. Contradictory to that, Sastalla and coworkers observed a loss of toxicity of supernatant of B. cereus strain ATCC 10876 with a decreasing amount of Hbl L 1 at the late stationary growth phase [25]. On the other hand, already in 1991 when only two Hbl components were known, it has been shown that in relation to Hbl B, only small quantities of L are necessary to cause hemolysis [18]. Turbidity measurements confirmed that hemolysis varies with varying concentrations of Hbl B, while variations of L 2 and L 1 had no significant influence [29]. This coincides with the findings of our study. We also found that pore formation of Hbl into the target cell membrane is defined by the binding of the B component.
Furthermore, we showed that pore formation of Hbl occurs within minutes. Binding of Hbl L 1 to membrane bound Hbl B happens immediately, as we were able to reduce incubation times for rHbl L 1 down to one minute. Also binding of Hbl L 2 must occur instantly, as PI influx could be measured right after addition of L 2 . In earlier spectrophotometric hemolysis assays a "slow priming reaction of the B component" (3 h at 37 • C or 1 h at 42 • C) and a "rapid lytic reaction by L 1 +L 2 " (within 3 min) have been observed [24]. Our data also suggest that the priming reaction is quite fast. Depending on the protein concentration, incubation of Vero cells with rHbl B could be decreased to a minimum of 4 min, while pore formation was still detectable. Thus, we suggest that binding of Hbl B to the target cell surface is also a rapid and optimized process. Nevertheless, our WST-1 bioassays with diluted Hbl components showed that the toxicity of Hbl is defined by the B component, depending on how much B is bound to the cell surface. Thus, priming with B is the crucial step, then the lytic components L 1 and L 2 can easily be diluted by a factor of 10-20 without affecting the toxic activity.
In our study, only Hbl B was able to bind to the target cell surface. Thus, we found no indication for a "membrane attack complex" leading to cell lysis [24]. Furthermore we showed that neither a single component nor the combination of two components are cytotoxic, or able to form a pore. This disagrees with the idea of Madegowda and coworkers, who suggested that due to its structural homology to E. coli ClyA [27,28], Hbl B alone might be able to bind to the cell surface, oligomerize and form a pore, while Hbl L 2 and L 1 induce conformational changes, stabilize B or even enter the cell similar to anthrax toxin [26].
We showed that rHbl B is able to bind the target cell surface excellently on its own, and that the binding order is definitely B -L 1 -L 2 . Nevertheless, rHbl B binding seemed to be enhanced in the presence of rHbl L 1 (see Figure 4). In an earlier study we analogically observed increased binding of rHbl B in the presence of Hbl B-specific mAb 1D7, and described it as "antibody-dependent enhancement" [30,32]. Hbl L 1 might have a similar, stabilizing effect on Hbl B binding. This observation is especially interesting considering that even before cell contact Hbl B and L 1 form complexes in solution-the recombinant proteins as well as the natural components in B. cereus culture supernatants [30]. Due to additional complex formation with Hbl L 2 , we postulated that in natural B. cereus culture supernatants, Hbl B is mainly bound in complexes [30]. Further, free Hbl B seems to be no less important, as pore formation as well as cytotoxic and hemolytic activity were enhanced with an excess of B (this study).
With K D values of 4.7 × 10 -7 M and 3.4 × 10 -6 M, the Hbl L 1 -B and Hbl B-L 2 complexes differ significantly from the one determined for NheB and C, which was 4.8 × 10 -10 M [30,33]. Although organized in similar operons-nheABC and hblC(L 2 )D(L 1 )A(B)-and structurally related [11,34], there seem to be substantial differences in the mechanism of pore formation of the non-haemolytic enterotoxin and hemolysin BL of B. cereus.
For Nhe, formation of small "pro-pores" by NheB and C (homologs to Hbl L 1 and B) has been reported [31,33]. First, highly stable NheB-C complexes are formed in solution. Then, these complexes, as well as free C but not B, are able to bind to the target cell surface. After conformational changes, free NheB can attach. In this process, a defined level of NheBC complexes, as well as a sufficient amount of free NheB, is necessary for efficient cell binding and toxicity [31,35]. NheBC forms stable transmembrane channels with a diameter of about 2 nm, increasing membrane permeability and cytotoxic effects prior to the formation of the full pore [33]. With regards to Hbl, the corresponding B-L 1 complexes show less stability [30], and so far, no evidence for the formation of small pores by only two components was found, although the presence of L 1 seems to enhance the cell binding of B (this study). Furthermore, large amounts of NheC inhibit toxicity [36,37], while in the present study large amounts of the homolog Hbl B enhanced especially pore formation.
Either way, the pore is completed by attachment of the corresponding third component, NheA [31,38,39] or Hbl L 2 [25], this study. A recent study provided evidence that NheA is highly flexible and undergoes major conformational changes during pore formation [38]. The mechanism of the Hbl L 2 attachment to B and L 1 is yet to be determined. For Nhe, it has been shown that maximum toxicity is reached at a concentration ratio of NheA:B:C = 10:10:1, but with strain-specific differences [36][37][38]. On the contrary, in the present study a molar ratio of Hbl L 2 :L 1 :B = 10:10:1 did not result in any toxic or pore-forming activity of Hbl.
According to the current knowledge, the mechanism of Hbl is modeled as follows: First, Hbl B binds to the target cell surface. This is the crucial and most complex step in Hbl pore formation, but is still completed within a few minutes. In natural culture supernatants of B. cereus, large amounts of Hbl B and L 1 are present as complexes [30]. Linked with L 1 , the binding capacity of B to the target cell surface is increased. In a subsequent step, free Hbl L 1 binds to B. This is a very efficient and quick process, as well as the adjacent binding of Hbl L 2 . Strong excess of Hbl L 1 hinders pore formation. On the other hand, it is still unclear how many molecules of each component are necessary for pore formation, if conformational changes appear, and which protein moieties are involved in the interaction.

Bacterial Strains and Culture Conditions
The B. cereus strain F837/76 (DSM 4222) was used in this study. To obtain toxin-rich culture supernatant, the strain was grown in CGY medium with 1% glucose for 6 h at 32 • C with shaking. According to [40], the sample was centrifuged (4.000× g at 4 • C for 20 min) before addition of 1 mM EDTA and filtration through a 0.2 µm sterile filter. E. coli strain BL21 (DE3) was used for overexpression of recombinant (r) Hbl proteins.
As rHbl L 2 with C-terminal strep-tag could not be purified via affinity chromatography, its concentration was estimated via enzyme immuno assay (see [43,44]) and the crude E. coli cell extract was used. The cell extract applied solely did not lead to any pore formation or toxic activity (data not shown).

Cell Line and Culture Conditions
Vero cells were obtained from ECACC (European Collection of Cell Cultures). They were cultured in 80 cm 2 culture flasks in a humidified incubator at 37 • C and 7% CO 2 in MEM Earle's medium (Biochrom GmbH, Berlin, Germany) plus supplements, as recommended by the supplier.

Cytotoxicity Assays
Cytotoxicity assays were performed as previously described [40,43,45]. Briefly, for WST-1 bioassays serial dilutions of B. cereus culture supernatant or rHbl proteins were placed into 96 well plates (0.1 mL per well) and subsequently, Vero cell suspensions (1 × 10 4 cells and 0.1 mL per well) were added. After 24 h incubation at 37 • C and 7% CO 2 , cell viability was determined by the addition of WST-1 (Roche diagnostics, Penzberg, Germany). Optical density was recorded in a Tecan photometer at 450 nm using Ridawin software. Dose-response curves were calculated to obtain 50% lethal concentrations, which are shown as reciprocal titers.
To assess pore formation in the membranes of Vero cells, propidium iodide (PI) influx tests were used. 4 × 10 4 Vero cells were seeded in 200 µL MEM Earle's medium /well in black 96 well plates and incubated for 24 h at 37 • C and 7% CO 2 . 100 µL medium were removed and 100 µL fresh medium were added containing 10 µg/mL PI (Sigma-Aldrich Biochemie GmbH, Hamburg, Germany) and B. cereus culture supernatant (1:20) or rHbl components in appropriate dilutions. Fluorescence was measured immediately in a Victor 1420 multilabel counter (Perkin Elmer, Boston, MA, USA) for 4 h every 2.5 min (excitation: 530 nm; emission: 616 nm; excitation time: 1s; excitation strength: 20,000). Fluorescence curves of three replicates were calculated using Microsoft Excel. The start point of PI incorporation and the slope of the fluorescence curves allow concluding on the efficiency and velocity of pore formation (see also [36,40]), while differences in the end point fluorescence rather depend on the target cells.
Stock solutions of the rHbl components were pre-mixed in appropriate ratios and added as serial dilutions (WST-1 bioassay; start concentrations as indicated in the respective experiments) or in concentrations of 37.5 pmol/mL each (PI) to the cells.

Flow Cytometry Analyses
For flow cytometry analyses, Vero cells were counted and adjusted to 1 × 10 6 cells in 500 µL EC buffer (140 mM NaCl, 15 mM HEPES, 1 mM MgCl 2 , 1 mM CaCl 2 , 10 mM glucose, pH 7.2). rHbl B ± rHbl L 1 or L 2 were added in EC buffer in concentrations of 9.38 pmol/mL each to a total volume of 1 mL. After 1 h incubation at 37 • C with moderate agitation, 2 mL 1% BSA-PBS were added and cells were centrifuged for 5 min at 800 rpm. Cells were washed again in 2 mL 1% BSA-PBS. Samples were incubated with 5 µg/mL mAb 1G8 or 1B8 [30,43] in 1 mL 1% BSA-PBS for 1 h at RT and again washed twice with 2 mL 1% BSA-PBS. Subsequently, samples were incubated with 2 µg/mL Alexa Fluor ® 488 goat anti mouse IgG (life technologies, Carlsbad, CA, USA) for 45 min at 4 • C. Alternatively, mAb 1G8 directly labeled with Alexa Fluor ® 488 was used for detection. After two washing steps with 2 mL 1% BSA-PSB, cells were resuspended in 500 µL 1% BSA-PBS, and transferred to flow cytometry tubes. For the detection of Hbl L 2 , Vero cells were first incubated with rHbl B, L 1 or B+L 1 (4.17 pmol/mL each) and in a second step with a mixture of rHbl L 2 (4.17 pmol/mL) and the neutralizing mAb 1H9 (5 µg/mL) [30]. Fluorescence was measured in a FACS Calibur using the CellQuestPro software (BD Bioscience, San Jose, CA, USA). Cell populations were visualized in the FSC SSC dot-plot. Fluorescence was recorded in fluorescence channel 1 and shown as a histogram.

Determination of Hemolytic Activity
10 µL of toxin containing F837/76 supernatant or rHbl L 2 , L 1 and B (1.5 pmol/µL each) were filled individually or as mixture in different ratios into 3.5 mm diameter stamp holes on sheep blood agar plates. After 3-6 h incubation at 32 • C hemolysis patterns were assessed and documented.