Distinction between Pore Assembly by Staphylococcal α-Toxin versus Leukotoxins

The staphylococcal bipartite leukotoxins and the homoheptameric α-toxin belong to the same family of β-barrel pore-forming toxins despite slight differences. In the α-toxin pore, the N-terminal extremity of each protomer interacts as a deployed latch with two consecutive protomers in the vicinity of the pore lumen. N-terminal extremities of leukotoxins as seen in their three-dimensional structures are heterogeneous in length and take part in the β-sandwich core of soluble monomers. Hence, the interaction of these N-terminal extremities within structures of adjacent monomers is questionable. We show here that modifications of their N-termini by two different processes, using fusion with glutathione S-transferase (GST) and bridging of the N-terminal extremity to the adjacent β-sheet via disulphide bridges, are not deleterious for biological activity. Therefore, bipartite leukotoxins do not need a large extension of their N-terminal extremities to form functional pores, thus illustrating a microheterogeneity of the structural organizations between bipartite leukotoxins and α-toxin.


INTRODUCTION
Staphylococcal bipartite leukotoxins and α-toxin belong to a single family of structurally related β-stranded pore-forming toxins ( Figure 1). Leukotoxins are constituted by a class S protein (32 kd) that binds to the surface of target cells prior to a class F protein (34 kd) distinct in sequence [1]. Then, these proteins oligomerise to octamers or hexamers [2][3][4][5], induce Ca 2+ -activation [6], and form monovalent cationselective β-barrel transmembrane pores [7]. Each protomer contributes to the pore by transforming a central β-sheet domain into a β-hairpin [8,9]. Five loci encoding leukotoxins are characterized [9]. Several of these loci, encoding the Panton-Valentine leucocidin (PVL), gamma-hemolysin, [10] and LukE-LukD [11], may be present together and are expressed in a single isolate. The S and F components can then combine to give a specific leukotoxin. However, they do not combine with α-toxin, also present in almost all isolates, for an action onto natural target cells [12], that is, human polymorphonuclear cells or erythrocytes. Panton-Valentine leucocidin, which is composed of LukS-PV and LukF-PV, is associated with furuncles [13] and pneumonia [14]. Bipartite leukotoxins show a complementary spectrum for lytic functions towards human blood cells including lymphocytes and erythrocytes, accounting for the bacterial virulence. Furthermore, leukotoxins and α-toxin differ from each other by the respective cationic and anionic selectivities of their pores [6,15,16].
X-ray diffraction and other techniques have been used to study the heptameric pore of α-toxin [17][18][19][20][21]. The crystal structure of the assembled α-toxin [20] revealed that the transmembrane β-barrel that forms the pore corresponds to the stem domain of a mushroom-shaped structure. It also revealed that the N-terminal extremity, also called the amino latch, of each protomer interacts with two adjacent protomers in the mouth of the pore lumen. The 3D structures of the soluble form of two F monomers, LukF-PV from the Panton-Valentine leucocidin (Figure 1(a)) [8] and HlgB from gamma-hemolysin [22], have also been determined. They showed that the central stem domain is prefolded as three β-strands stacked to the core of the soluble proteins. This core is formed of a β-sandwich to which the N-terminal extremity of the soluble F proteins contributes to one strand. Despite the high similarity between the F and S structures, the N-terminal extremity of LukS-PV appeared shorter [23]. The F and S proteins must undergo a number of conformational modifications during oligomerization and β-barrel formation to evolve to a functional transmembrane pore [24,25].
Since the possible unfolding of the N-terminal extremity to interact with adjacent subunits is unclear, and that modifying the N-termini for bioengineering purposes could considerably influence their function, we report that some biological activity is retained by recombinant proteins after glutathione S-transferase (GST) fusion and expression in Escherichia coli. We also constrained this portion of the proteins to the β-sandwich by engineering disulfide bonds, and investigated the residual functions of new leukotoxins in order to verify if N-terminal extremities of leukotoxins must explore a large conformational space for their pore-forming activity.

Choice, construction, and purification of modified proteins
We considered alignment of N-terminal LukS-PV and LukF-PV sequence alignment extracted from the alignment of entire sequences (Figure 1(a)) [23]. We chose two pairs of LukF-PV amino acids, T5 and T21, and S8 and K20 for cysteine substitutions due to their respective close locations in the structure (Figure 1(a)) [8]. The distance between Cβ atoms of T5 and T21 (5.7Å) and angles between thiol groups may allow the formation of disulfide bonds that would bridge 3 the first two β-strands of the protein (SSBond software, [26]). The other pair of amino acids seemed even more favorable for disulphide bridge formation, with a distance between Cβ atoms of 4.1Å. The situation was less obvious for LukS-PV, which has no counterpart to LukF-PV T5 (Figure 1(a)). Furthermore, the first three amino acids of the shorter LukS-PV N-terminal extremity could not be traced in the crystal structures of either wild type [23] or recombinant protein (unpublished results). Therefore, this N-terminal extremity may be poorly structured. To allow bridging with R16, which can be aligned with T21 of LukF-PV, we finally chose to substitute either LukS-PV N2 or D1 by a cysteine, and to introduce another cysteine residue upstream from D1 (called LukS-PV-1C) (Figure 1(a)).
GST∼LukS-PV and GST∼LukF-PV fusion proteins were purified for functional analysis by chromatography on glutathione-sepharose 4B followed by hydrophobic interaction chromatography (HIC, alkyl-superose, i.e., resource ISO-Ge Healthcare, USA). GST∼LukS-PV and GST∼LukF-PV eluted at 0.73 M and 0.78 M (NH 4 ) 2 SO 4 , respectively. All double cysteine-mutated proteins were purified by affinity chromatography on glutathione-sepharose 4B followed by cation-exchange FPLC chromatography (CEC) using an NaCl gradient from 0.05 M to 0.7 M [24]. The GST-tag was removed thanks to the PreScission protease. Nevertheless, it remains an octapeptide at the N-terminus which does not hamper the biological activities of the toxins [8,24] (Amersham Biosciences). Hydrophobic interaction chromatography (HIC) was further applied using an (NH 4 ) 2 SO 4 gradient from 1.3 M to 0.5 M to improve purifications. LukS-PV mutants eluted at 0.15 M NaCl and 1.02 M (NH 4 ) 2 SO 4 , whereas those of LukF-PV eluted at 0.1 M NaCl and 0.95 M (NH 4 ) 2 SO 4 . To avoid any disulfide links formation between free sulfhydryl groups of the cysteine residues of the fusion proteins and the GSH, all buffers used for the CEC and the HIC contain 1 mM DTT (except for GST fusion proteins). Controls for homogeneity were performed using SDS-PAGE, and the proteins were then stored at −80 • C. We labelled two fully functional mutants, LukF-PV S27C and LukS-PV G10C, with fluorescein 5-maleimide (Molecular Probes, Leiden, The Netherlands) [24].

Oxidation and accessible thiol-titration
The cysteine mutants were first reduced with 10 mM DTT, before treating with CuSO 4 , 1,10-phenanthroline (Sigma, USA). Briefly, proteins at a concentration of 20 μM were dialysed in 50 mM Hepes, 0.5 M NaCl, 10 mM DTT pH 7.5, and further equilibrated against 50 mM Hepes, 0.5 M NaCl, pH 7.5 to undergo oxidation. Proteins were then adjusted to 2 mL (20 μM) of the same buffer complemented with 1.5 mM CuSO 4 , 5 mM 1,10-phenanthroline and incubated for 2 hours at 4 • C. Protein solutions were then equilibrated against 50 mM Hepes, 0.5 M NaCl, 1 mM EDTA-Na 2 , pH 7.5, and could then be frozen without loss of disulfide bonds and biological activity.
For the titration of free thiols, proteins were precipitated in 5% (w/v) trichloroacetic acid and left for 5 minutes at 0 • C, pelleted by centrifugation and washed three times with the same solution. The precipitate was dissolved in 400 μL of N 2 -saturated 0.2 M Hepes, 0.2 M NaCl, 1 mM EDTA-Na 2 , 2% (w/v) SDS, pH 8.0. The remaining precipitated material (< 10% of total proteins) was removed by centrifugation and 300 μL of the supernatant was added to 30 μL of 10 mM 5-5 -di-thio-bis (2-nitro) benzoic acid (DTNB) (ε = 13, 600 M −1 ·cm −1 ). After 10 minutes of reaction at room temperature, the amount of titrated thiols was estimated by OD at 412 nm and the molarity was compared to the protein concentration determined by the Lowry method.

Human polymorphonuclear cells (PMNs) and flow cytometry measurements
Human PMNs from healthy donors were purified from buffy coats as previously reported [27], and suspended at 5 × 10 5 cells/mL in 10 mM Hepes, 140 mM NaCl, 5 mM KCl, 10 mM glucose, 0.1 mM EGTA pH 7.3. Flow cytometry measurements from 3000 PMNs were carried out using a Fac-Sort flow cytometer (Becton-Dickinson, Le Pont de Claix, France) equipped with an argon laser tuned to 488 nm [28]. We evaluated the intracellular calcium using flow cytometry of cells previously loaded with 5 μM Fluo-3 (molecular probes) in the presence of 1.1 mM extracellular Ca 2+ by measuring the increase of Fluo-3 fluorescence. Pore formation and monovalent cation influx were revealed by the penetration of ethidium through the pores; cells were incubated 30 minutes with 4 μM ethidium prior to the addition of toxins in the absence of extracellular Ca 2+ [24,28]. Fluo-3 and ethidium fluorescence were measured using Cell Quest Pro software (Becton, Dickinson and Company). The results from at least four different donors were averaged and expressed as percentages of a control of human PMNs treated with the wild-type (WT) PVL. Base level values were obtained for each series of data from a control without addition of toxin. These were systematically subtracted from the other assays. Standard deviations values never exceeded 10% of the obtained values; they were removed from figures for clarity. The dissociation constant (k D[S] ) of LukS-PV for the PMN membrane and that of LukF-PV for the PMN membrane-bound LukS-PV (k D[F] ) were previously reported to be 0.07 nM and 2.5 nM, respectively [29]. LukS-PV and mutants were applied at 1 nM while LukF-PV and derivatives were applied at 10 nM as an excess of toxins to their ligand(s). The binding properties of the modified proteins onto PMNs were estimated through competition experiments carried out in the absence of extracellular Ca 2+ . Fluorescein-labelled LukS-PV G10C (0.1 nM) and LukF-PV S27C (2 nM), and increasing concentrations of the respective mutants (from 1 to 1000 nM) were added 15 minutes before measuring the fluorescence retained at the surface of PMNs. For LukF-PV competitions, the PMNs were initially incubated for 10 minutes with 1 nM of LukS-PV. The IC 50 value corresponds to the concentration of nonfluorescent competitor needed for 50% cell fluorescence inhibition, and is determined from the best fit of independent triplicates of the residual cell fluorescence [24]. The apparent inhibition constant, k Iapp , was calculated from the following equation:

Determination of pore radius
The pore formation induces a disruption in the cell membrane. It results in an increase of the cell size due to the difference of osmolarity between the cytoplasm and the medium. Indeed, the osmolarity is weaker in the medium than in the cytoplasm. The relative variations in PMN size were assessed by measuring of the forward light scatter (FSC) of cells (5×10 5 cells/ml) treated with WT or mutant 20 nM LukS-PV, WT or mutant 100 nM LukF-PV, and 30 mM PEG polymers (1000, 1500, 2000, 3000 Da) of different hydrodynamic radii (0.94, 1.12, 1.22, and 1.44 nm, resp.) [25]. FSC values were collected at 15 minutes after toxin application. If the sizes of the PEG molecules are similar or greater than the diameter of lumen, they cannot pass through the pores, and then cannot balance the osmolarity between both compartments. In that case, the FSC variations are weak. If the sizes of the PEG molecules are smaller than the diameter of lumen, they can pass through the pore with balancing the osmolarity between the two compartments. It results in a rise of the FSC variations.

Identification of oligomers
Denaturing polyacrylamide gel electrophoresis (SDS-PAGE) was carried out on oxidized leucotoxin-treated human PMNs. Preparations at 5 × 10 7 cells/mL in 10 mM Hepes, 140 mM NaCl, 5 mM KCl, 10 mM glucose, 0.1 mM EGTA, pH 7.3, were incubated with 100 nM of LukS-PV and LukF-PV derivatives in the presence of 10 μL/mL of a mammalian cell-tissue antiprotease cocktail (Sigma, USA). After a 45minute incubation at 22 • C, the biological activity was assessed by optical microscopy, the cells were washed twice and then resuspended in 1 mL of the same buffer containing an antiprotease cocktail (1 μL/mL) as above. The cells were ground in a FastPrep apparatus (QBiogene, Bio101, Illkirch, France) using FastPrep Blue tubes for an orbital centrifugation (10 seconds, 3600 rpm, room temperature). The membranes were harvested by ultracentrifugation for 20 minutes at 20000 × g at 4 • C. The membrane pellets were resuspended in 100 μL of the same buffer complemented with 2 μL of antiprotease cocktail containing 1% (w/v) saponin (Sigma), incubated for 30 minutes at room temperature and then centrifuged for 30 minutes at 22000 × g. The supernatants were adjusted to 1 mM glutaraldehyde (in the above buffer) and incubated for 10 minutes at 50 • C. One third volume of loading buffer (0.5 M Tris-HCl pH 8.5, 2% (w/v) SDS, 0.04% (w/v) bromophenol blue, 30% (v/v) glycerol) containing 100 mM ethanolamine to block the cross-linking reaction was added and assays were heated to 100 • C for 5 minutes. Finally, 10 μL of the solution was loaded onto Trisacetate, pH 8.1 polyacrylamide 3-8% (w/v) gels (Invitrogen, Calif, USA). Proteins were subjected to electrophoresis for 75 minutes at 150 V at room temperature in 50 mM Tris, 50 mM Tricine, pH 8.2, 0.1% (w/v) SDS, and then electroblotted onto nitrocellulose membranes for 1 hour at 30 V in 25 mM Tris, 192 mM glycine, pH 9.3, 20% methanol using a transfer Xcell II blot module (Invitrogen). The leukotoxins oligomers or components were characterized by immunoblotting using affinity-purified rabbit polyclonal antibodies and a peroxidase-labeled goat antirabbit antibody using ECL detection (Amersham Biosciences, Saclay, France) as previously described [24]. The apparent molecular masses were estimated from protein migration according to Precision Plus Protein Standards (Bio-Rad, USA).

GST-fusion proteins remain biologically active
GST∼LukS-PV and GST∼LukF-PV were purified to homogeneity and appeared as homogeneous proteins with apparent molecular masses of 57 and 60 kd, because of the fusion with the 26 kd GST (Figure 2(a)). The apparent binding of leukotoxin derivatives with cell membranes was assessed in competition with an increased amount of the functional fluorescein-labelled ( * ) LukS-PV G10C * or LukF-PV S27C * . Apparent k I of 0.039 nM was found for the WT LukS-PV binding to membranes and a value of 3.6 nM was found for the WT LukF-PV binding to LukS-PV-membrane complexes (data not shown). In these conditions, values of k Iapp determined for GST∼LukS-PV and GST∼LukF-PV (0.2 and 3.4 nM, resp.) remained close to the values obtained for the WT proteins. However, when GST∼LukF-PV was applied to the bound GST∼LukS-PV, a k Iapp of 17.6 nM was recorded, this marked difference may be due to some steric hindrance caused by GST molecules.
Combinations of LukS-PV with GST∼LukF-PV or of GST∼LukS-PV with LukF-PV led to Ca 2+ induction in treated cells comparable to the WT proteins (Figure 2(b)). However, the combination of both fusion proteins required a longer time, that is, 4 minutes instead of 2 minutes, to reach a fluorescence maximum. Subsequent decrease of fluorescence is mainly due to the release of the fluorescent probe by pores and disrupted cell membranes.
The permeability to monovalent cations mediated by the pores was measured via the entry of ethidium and its combination with nucleic acids. Despite ethidium fluorescence being less sensitive than Ca 2+ assay, these two influxes were demonstrated to be independent [7,24]. Fusion proteins tested alone did not generate entry of Ca 2+ (not shown) and ethidium (Figure 2(c)). Ethidium entry showed a higher variation than what was observed for calcium, but the fusion proteins retained significant biological activity (Figure 2(c)). The activity of GST∼LukS-PV + LukF-PV lies between those obtained for PVL and for PVL with a 1 : 10 dilution of LukS-PV, whereas the activity of LukS-PV + GST∼LukF-PV was similar to that of PVL with the 1 : 10 dilution of LukS-PV (Figure 2(c)). The combination of both fusion proteins showed a considerably reduced ethidium influx, even if compared with the 1 : 10 dilution of LukS-PV (Figure 2(c)). This difference can again be explained by the steric hindrance caused by GST molecules. After osmotic protection with PEG molecules, all couples involving at least one fusion protein showed decays of the forward light scatter values comparable to those of the WT toxin (Figure 2(d)). The inflexion point calculated for each curve clearly indicated permeability to PEG molecules with a hydrodynamic radius cutoff between 1.12 and 1.2 nm. Thus, the diameter of the leukotoxin pores was not affected by GST fusion proteins [25]. We aimed at identifying oligomers formed by the fusion proteins (67 and 60 kd, resp.) within the human PMNs membranes to confirm the pore formation. When retrieved from cell membranes, we noticed that oligomers formed by LukS-PV and LukF-PV were very sensitive to detergents (saponin 1% and SDS 0.04%), as we only detected mixtures of monomers and dimers (Figure 3, lane 3). Therefore, we used a cross-linking agent, glutaraldehyde, applied consecutively after the saponin treatment, to help to stabilize the oligomers removed from the membranes in the presence of an excess of cell membrane proteins. Cross-linking proved efficient for both combination of fusion proteins (Figure 3, lanes 4-6), but essentially showed tetramers for all, in spite of the fact that hexamers might be suspected for the LukS-PV + GST∼LukF-PV combination (Figure 3, lane 4, see arrow). The use of anti-LukS-PV and anti-LukF-PV antibodies, alone or in combination, allowed the detection of similar high molecular mass oligomers containing GST fusion leukotoxins (data not shown). The materials were specific for the leukotoxins, since no immunoreactive product was observed when analyzing human PMNs alone (Figure 3, lane 2). The intensity of these bands rapidly decreased while the  number of subunits increased. The increase of the steric hindrance due to the fusion GST proteins probably contributes to the intrinsic instability of leukotoxins pores when extracted from membranes and may impact the resolution of the complete oligomers ( Figure 3, lanes 4-6). Therefore, we decided to characterize the oligomers for only WT proteins and double cysteine mutants (see below).

Integrity of double-cysteine mutants
Double-cysteine mutants were analyzed in their oxidized forms (Figure 4). Among the LukS-PV and LukF-PV doublecysteine mutants obtained in either reducing or oxidizing conditions, only oxidized LukS-PV N2C-R16C and oxidized LukF-PV T5C-T21C contained very small amounts of homodimers (< 8%). The presence of these dimers did not cause a significant loss of biological activity.

Binding of double-mutated leukotoxins
Using similar conditions as for the fusion proteins, we determined the apparent binding constants of the different mutants to their respective ligands. For LukS-PV mutants, we found k Iapp values, ranging from 0.052 nM to 0.069 nM, similar to those of WT LukS-PV. Values for the binding of LukF-PV T5C-T21C to the LukS-PV mutants-membrane complexes (k Iapp = 2.8 − 5.2 nM) were also close to that of the WT protein (k Iapp = 2.5 nM). In contrast, binding of LukF-PV S8C-K20C was more affected. Indeed, binding of LukF-PV S8C-K20C to LukS-PV D1C-R16C and N2C-R16C gave k Iapp of 18.5 ± 2.5 and 12 ± 2 nM, respectively. Binding became even worse with LukS-PV and LukS-PV(-1C)-R16C with k Iapp = 37 ± 5 and 24 ± 6 nM, respectively. Olivier Joubert et al.

Ca 2+ entry
Staali et al. [7] and Baba Moussa et al. [24] showed that Ca 2+ influx and ethidium entry promoted by leukotoxins can be selectively inhibited. We evaluated the different oxidized LukS-PV and LukF-PV double mutants for their ability to activate PMNs and provoke Ca 2+ -channel opening ( Figure 5).
In our system, we assayed LukS-PV at 0.1 or 1 nM in combination with LukF-PV at 2 or 20 nM, and compared activity of associations of LukS-PV proteins/mutants at 1 nM and LukF-PV proteins/mutants at 20 nM. Table 1 summarizes kinetics of cell-associated Ca 2+ fluorescence for different combinations. When mutants were first combined with the heterologous WT protein, kinetics of Ca 2+ influx comparable with that of the WT toxin were observed for combinations involving LukS-PV(-1C)-R16C or N2C-R16C and LukF-PV at 20 nM ( Figure 5(a)). In contrast, the kinetics for 1 nM LukS-PV D1C-R16C combined with LukF-PV was decreased to a lower value than that obtained for WT combination involving 2 nM of LukF-PV. The same was observed for LukS-PV + LukF-PV S8C-K20C, whereas the combinations involving LukF-PV T5C-T21C were almost as active as the WT toxin ( Figure 5(b), Table 1 Table 1). Altogether, these data indicate that constraining N-terminal extremities might not be detrimental to biological activity.

Ethidium entry induced by pore formation
Pore formation and ethidium entry promoted by the oxidized mutated proteins were more variable than Ca 2+ entry ( Figure 6). Table 2 Table 2) and kinetics for LukS-PV D1C-R16C was dramatically affected. The kinetics of ethidium entry induced by LukS-PV combined with LukF-PV mutants was intermediate between those produced by a same concentration of LukF-PV and its 1 : 10 dilution. Combinations of reduced mutants with heterologous WT proteins or mutants did not show high difference in activity compared to the use of oxidized mutants (data not shown).  (Figure 6(c)), the associations of reduced mutants always induced higher ethidium influxes than those obtained with associations of oxidized mutants (Table 2). From combining oxidized mutants, the pairs involving LukS-PV(-1C)-R16C, N2C-R16C, combined with LukF-PV T5C-T21C showed intermediate activity (Table 2), while only LukS-PV N2C-R16C combined with LukF-PV S8C-K20C could be considered as significant as the last cited ( Figure 6(d)). It has to be noticed that all combinations of oxidized mutants less prone to induce Ca 2+ influx were also less effective in pore formation.

Pore radii
The pore radii of the most active oxidized leukotoxins were evaluated by osmotic protection induced with calibrated PEG molecules (Figure 7). PMNs were protected against osmotic disruption by PEG molecules having a hydrodynamic diameter between 1.12 nm and 1.22 nm. A same value ranging 1.2 nm was obtained for the WT toxin, indicating similar pore radii of these toxins.

Panton-Valentine leucocidin oligomers
Several experiments were carried out, checking for the recovery, the stability, and the occurrence of oligomers formed by PVL. Control of cells without PVL did not give evidence of cross-reacting material ( Figure 8, lane 1). Challenging PVL oligomers with 4 ng of each component in solution was highly dependent on glutaraldehyde concentrations (Figure 8, lanes 3 and 4), while they show a little tendency to spontaneously form dimers (lane 2). Indeed, the use of 0.3 mM glutaraldehyde allowed to obtain various oligomers containing 2 to 10 units, at least, whereas 3 mM applied on these purified proteins dramatically affected signals (lane 4). Application of 30 ng of PVL per assay to PMNs only allowed observing few dimers in the described conditions without any glutaraldehyde treatment (Figure 8, lane 5). It has to be noticed that a nonboiled, but SDS-containing, assay resulted in insufficiently denatured materials, and even if treated by saponin and glutaraldehyde, this did not give interpretable results (Figure 8, lane 6). Saponin treatment of cells largely helped the release of protein-containing oligomers, despite the fact that saponin brought contaminating proteins (50% of the total bulk of proteins). Each assay on cells involve an original volume of 1 mL further concentrated to 100 μL. Cells treated with PVL, saponin, and with a consecutive cross-linking with 3% (v/v) glutaraldehyde allowed to Olivier Joubert et al.  identify oligomers species provided that samples were boiled (Figure 8, lanes 7-11), whether saponin treatment was carried out at 4 • C or 23 • C. In fact, the Schiff reaction promoted by glutaraldehyde is not stable in aqueous solution, but crosslinking certainly occurs in our assays through the Michael addition reaction onto amine groups (R-NH 2 ). Considering the excess of proteins in lysates, the observed oligomers are assumed as preformed oligomers, since the cross-linking reaction cannot be considered to be specific of the PVL proteins. This addition reaction was stopped by ethanolamine Table 2: Times (min) to reach (a) 5%, (b) 50%, and (c) 100% of the entry of ethidium into cells by pores formed by combinations of PVL (taken as reference) and its mutants.

DISCUSSION
Although staphylococcal α-toxin and bipartite leukotoxins fold in a comparable way, differences exist in the sequences and structures of these two subfamilies of toxins that imply differences in function [30]. Another distinction is in stoichiometry. Indeed, α-toxin is known to form heptamers in target cells [20] and hexamers in some different membranes [31]. Recently, octamers were identified in synthetic bilayers and PMNs membrane by using cross-linking between mutated components of gamma-hemolysin [5,32]. Moreover, the presence of strictly alternating S and F proteins was recently proven in the case of leukotoxin pores [33]. Considering the sequence alignment of the S and F components of leukotoxins and of α-toxin [23], differences between these proteins are located at the N-termini, at each side of the central domain and in the last fifty C-terminal residues. The significance of these differences with α-toxin is not understood, so far. It has been shown that deleting the first two residues or labelling the N-terminal extremity of α-toxin dramatically reduce its biological activity [34,35]. It is also noteworthy that leukotoxins oligomers are less stable than those of αtoxin, hampering their crystallization. An investigation using infrared spectroscopy did not give evidence of any significant modification in the β-strands content when passing from the soluble state to the pore formed in planar lipid membranes [4].
In this work, we try to bring out arguments that Nterminal extremities of leukotoxins keeping part of the βsandwich core during pore formation lead to functional toxins, and that a large unfolding of these extremities is not obvious. Despite the fact that the recombinant expression system used in this study adds an N-terminal octapeptide, the proteins that were produced retain a biological activity comparable to native toxins [24,25]. Moreover, GST fusions of LukS-PV or LukF-PV also proved their binding efficacy. LukS-PV∼GST + LukF-PV was only from 3 to 5 folds less efficient than the corresponding WT proteins, and the LukS-PV + LukF-PV∼GST was also a bit less effective than the WT couple. The binding of LukF-PV∼GST was affected when tested on its GST fusion counterpart protein and the resulting biological activity significantly decreased, but remained biologically active. Thus, the decrease in binding might be due to steric hindrance produced by the fusion, especially in the case of the GST couple. Despite a higher sensitivity of the Ca 2+ entry assay, the decrease of Ca 2+ influx induced by the GST couple correlated with its lower pore-forming activity. It can be assumed, therefore, that the decrease in pore-forming activity results from the reduced binding of LukF-PV∼GST (Figure 2(c)). Taking into account all these features, it becomes less realistic that Nterminal extremities of leukotoxins extensively unfold to interact with residues of adjacent monomers located within the lumen of the pore (see Figure 1). Comparatively, the expressed and purified GST∼ α-toxin has no significant activity against rabbit red blood cells, whereas when GST is cleaved, the resulted octapeptide-α-toxin shows a lytic activity diminished by more than 10 folds than that of the native α-toxin purified from S. aureus. This constitutes a functional difference with the Panton-Valentine leucocidin. Moreover, at least LukF-PV was not sensitive to deletions less than 10 amino acids [36,37].  In the second approach used in this study, N-terminal extremities of both PVL components were locked to the protein core by disulfide bonds via site-directed mutagenesis. Different locations for cysteine substitution were chosen for each protein. The very first residues and R16 of LukS-PV appeared as good candidates for substitution by cysteines and subsequent bridging. Indeed, although the first two Nterminal residues are absent in the three-dimensional structure of LukS-PV [23], three positions were chosen by analogy with LukF-PV which produced internal disulfide bonds in majority. Thus, formation of homodimers during assisted oxidation could not be responsible for a great decrease in the biological activity of the mutated toxins. Oxidation with 30 mM H 2 O 2 could alternatively be preferred to Cu 2+ oxidation for some mutants, but in any case, excess of oxidant was removed before any biological assays. The binding of oxidized LukS-PV mutants to human PMNs was not affected, but that of LukF-PV S8C-K20C to LukS-PV mutants was diminished from 6 to 15 folds compared to that of WT proteins. This may suggest some modifications in the overall structures and some adverse compatibilities between new structures.
In fact, when combined with LukS-PV, LukF-PV S8C-K20C showed diminishing both in Ca 2+ induction (> 10 folds) and in pore formation (less than 10 folds), but activities largely decreased for combinations with LukS-PV