Lethal toxin from Clostridium sordellii induces apoptotic cell death by disruption of mitochondrial homeostasis in HL-60 cells

Authors

  • Patrice Petit,

    1. Institut Cochin, Inserm U567, Department of Developmental Genetic and Molecular PathologyICGM, 24 rue du Faubourg Saint-Jacques, 75014 Paris, France.
    Search for more papers by this author
  • Jacqueline Bréard,

    1. Inserm U461, Faculté de Pharmacie Paris XI, 92296 Chatenay-Malabry, France.
    Search for more papers by this author
  • Valérie Montalescot,

    1. Institut Cochin, Inserm U567, Department of Cell Biology, 22 rue Méchain, 75014 Paris, France.
    Search for more papers by this author
  • Noomen Ben El Hadj,

    1. Institut Cochin, Inserm U567, Department of Cell Biology, 22 rue Méchain, 75014 Paris, France.
    Search for more papers by this author
  • Thierry Levade,

    1. Inserm U466, CHU Rangueil, 31403 Toulouse Cedex 04, France.
    Search for more papers by this author
  • Michel Popoff,

    1. Institut Pasteur, Unité des Bactéries Anaérobies et Toxines, 75724 Paris Cedex 15, France.
    Search for more papers by this author
  • Blandine Geny

    Corresponding author
    1. Institut Cochin, Inserm U567, Department of Cell Biology, 22 rue Méchain, 75014 Paris, France.
    2. Institut Pasteur, Unité des Bactéries Anaérobies et Toxines, 75724 Paris Cedex 15, France.
      For correspondence at Institut Pasteur, Unité des Bactéries Anaérobies et Toxines, 75724 Paris Cedex 15, France. E-mail geny@pasteur.fr ; Tel. (+33) (1) 44 38 95 87; Fax (+33) (1) 40 61 31 23.
    Search for more papers by this author

For correspondence at Institut Pasteur, Unité des Bactéries Anaérobies et Toxines, 75724 Paris Cedex 15, France. E-mail geny@pasteur.fr ; Tel. (+33) (1) 44 38 95 87; Fax (+33) (1) 40 61 31 23.

Summary

Lethal toxin (LT) from Clostridium sordellii (strain IP82) inactivates in glucosylating the small GTPases Ras, Rap, Ral and Rac. In the present study we show that LT-IP82 induces cell death via an intrinsic apoptotic pathway in the myeloid cell-line HL-60. LT-IP82 was found to disrupt mitochondrial homeostasis as characterized by a decrease in mitochondrial transmembrane potential and cardiolipin alterations, associated with the release of cytochrome c in the cytosol. Time-course studies of caspase activation revealed that caspase-9 and caspase-3 were activated before caspase-8. Moreover, although LT-IP82-induced cell death was abrogated by caspase-inhibitors, these inhibitors did not suppress mitochondrial alterations, indicating that caspase activation occurs downstream of mitochondria. Protection of mitochondria by Bcl-2 overexpression prevented mitochondrial changes as well as apoptosis induction. Furthermore, evidence is provided that LT-IP82-induced apoptosis is not a consequence of cortical actin disorganization, suggesting that Rac inactivation does not initiate the apoptotic process. Cell exposure to LT-IP82 leads to a co-localization of the toxin with a mitochondrial marker within 2 h. Therefore, we suggest that LT-IP82 could act at the mitochondrion level independently of its enzymatic effect on small GTPases.

Introduction

Different species of Clostridium, bacteria involved in gastrointestinal diseases and gangrene, produce large molecular weight cytotoxins (250–300 kDa). Large clostridial toxins are exotoxins which induce actin cytoskeleton disruption and rounding up of cultured adherent cells. This group of toxins includes toxins A and B from Clostridium difficile, α-toxin from C. novyi and lethal toxin (LT) from C sordellii, involved in gas gangrene and haemorrhagic diarrhoea in humans and animals. Intestinal cells, which are exposed to lethal and haemorrhagic toxins, are particularly sensitive to these bacteria. Cells from the innate immunity, mainly neutrophils, rapidly colonize infectious sites to eliminate bacteria by phagocytosis and to mediate their destruction via superoxide ion production and secretion of immobilizing and lytic products (Lew, 1998). However, these cells could also be targets for C. sordellii toxins and their destruction could impair the defence against bacterial invasion.

Lethal toxin is the major pathogenic factor extracted from C. sordellii and has been shown to be closely related to toxin B from C. difficile as their amino acid sequences are 88% identical and both toxins cross-react immunologically (Martinez and Wilkins, 1992). Large clostridial toxins possess a glucosyltransferase activity allowing glucosylation at Threonin 35/37 of small G-proteins of the Ras superfamily by direct inhibition of effector binding (Popoff, 1987). Small GTPases include, among others, members of the Ras family (Ras, Rap and Ral) involved in signal transduction and members of the Rho family (Rho, Rac, Cdc42) implicated in actin cytoskeleton remodelling but also in a variety of cell functions such as cell cycle progression and gene transcription (Van Aelst and D’Souza-Schorey, 1997; Symons, 2000). The targets of toxins A and B from C. difficile are Rho, Rac and Cdc42, whereas LT, isolated from C. sordellii, strain IP-82, has been reported to inactivate Ras, Ral, Rap and Rac (Popoff et al., 1996; Richard et al., 1999). Modification of Rho GTPases by toxin A or toxin B can interfere with cell survival and cultured intestinal cells have been reported to undergo apoptosis after toxin exposure (Fiorentini et al., 1997; Brito et al., 2002).

Apoptotic cell death, a process essential for normal development, tissue homeostasis as well as removal of mutated or infected cells, is characterized by a sequence of morphological and biochemical events including cytoplasmic shrinkage, membrane blebbing, chromatin condensation and DNA fragmentation leading to cell death while sparing plasma membrane integrity (Kroemer et al., 1995; Vaux and Korsmeyer, 1999). Studies on the apoptotic molecular mechanisms have demonstrated the essential role of a family of cysteine proteases, caspases, which cleave cellular substrates at a P1 aspartate residue (Cohen, 1997; Los et al., 1999). They are expressed as zymogens and activated by cleavage at aspartate sites to generate the mature enzyme. Caspase activation occurs via two main pathways. In the extrinsic pathway, engagement of surface receptors of the Fas/TNF-R family by their cognate ligands leads to the recruitment of procaspase-8, an initiator caspase, which then undergoes autoactivation (Schulze-Osthoff et al., 1998). Active caspase-8 released from this complex cleaves and activates downstream effector caspases, such as caspase-3, that themselves cleave various cellular substrates leading to cellular dysfunction and cell death. In certain cell types, caspase-8 also cleaves the pro-apoptotic protein Bid that couples this death signalling pathway to the mitochondria. The intrinsic pathway is initiated by mitochondrial outer membrane alterations and release of cytochrome c as well as several other proteins that reside in the intermembrane space. The presence of cytosolic cytochrome c, due to either extrinsic or intrinsic signalling, initiates the formation of a complex, the apoptosome, which recruits procaspase-9 (Liu et al., 1996; Li et al., 1997). Autoactivated caspase-9 then activates effector caspases to precipitate apoptotic death. The antiapoptotic proteins Bcl-2 or Bcl-xL act a the level of mitochondria and prevent cytochrome c release (Cosulich et al., 1999).

In the present work, we demonstrate that LT-IP82 induces apoptotic cell death by the intrinsic pathway in myeloid HL-60 cells. Moreover, we attempted to determine whether modification of small G-proteins by LT was required for cell orientation towards apoptosis.

Results

Lethal toxin induces apoptotic cell death in a dose-dependent manner

As assessed by several methods, LT (IP82) was found to induce cell death with features of apoptosis in a dose-dependent manner in HL-60 cells. After LT-IP82 treatment (100 ng ml−1, 16 h), cell nuclei were stained with Hoechst 33258 and examined under UV light. About half of the cells presented a nucleus with chromatin condensation and nucleus fragmentation, compared to untreated cells which mostly (more than 95%) possessed a regular nucleus (Fig. 1A). The effect of LT-IP82 was also studied on DNA fragmentation another typical feature of apoptotic cell death. As reported in Fig. 1B, a nucleosomal DNA ladder, resulting from DNA digestion by specific nucleases, was found to be induced and to increase with the concentration of LT-IP82. Moreover, as reported in Fig. 1C, LT-IP82 also induced a dose-related increase in the percentage of cells labelled with Annexin-V FITC, which binds to phophatidylserine residues exposed on the outer leaflet of the plasma membrane in apoptotic cells. After a 16 h treatment with LT-IP82 (100 ng ml−1), 60% of the cells were labelled with Annexin-V FITC, amongst which 25% only were positive to PI, an indicator of further cell degeneration, ‘secondary necrosis’. To determine the time necessary for LT-IP82 to lead to apoptotic cell death, a time-course study of the toxin effect was also performed. As reported in Fig. 1D, the percentage of cells with exposed phosphatidylserine started to increase after 8 h of treatment with LT-IP82 (100 ng ml−1), whereas the percentage of necrotic cells labelled with PI increased moderately after 8 and 10 h and more rapidly thereafter.

Figure 1.

LT-induction of cell death by apoptosis in HL-60 cells.
A. Nuclear fragmentation analysis: nuclei from control HL-60 cells and cells treated for 16 h with LT-IP82 (100 ng ml−1, for 16 h) were labelled with Hoechst 33258 dye and photographed under a UV light microscope.
B. DNA fragmentation analysis: DNA from HL-60 cells treated or not with LT-IP82 was analysed by electrophoresis in a TBE agarose gel and visualized by ethidium bromide staining.
C. LT dose-reponse on PS exposure and cell death: HL-60 cells treated or not as in A and labelled with Annexin-V-FITC (filled squares) and PI (empty triangles). LT-IP82-treated cells labelled with Annexin-V represent the total dead cells amongst which cells labelled with PI were in secondary necrosis. Data expressed in percentage ± SD of positive cells, are from duplicate independent measurements made in three different experiments.
D. LT-IP82 time-course effect on PS exposure and cell death. HL-60 cells were treated LT-IP82 (100 ng ml−1) for various times after which they were labelled with Annexin-V-FITC and PI. Results are represented as in C.

LT toxin does not activate a sphingomyelinase

In cell-lines from myeloid origin, such as U937 and HL-60, apoptosis induced by various agents such as TNFα, Fas ligand, daunorubicin and Vitamin D3 have been reported to involve the activation of a sphingomyelinase leading to generation of ceramide during the first steps of the cell death process (Kolesnick and Kronke, 1998; Levade and Jaffrezou, 1999). Therefore, sphingomyelinase activity was measured over a period of 150 min after LT-IP82 addition (100 ng ml−1) to HL-60 cells. During this period, no obvious change in sphingomyelin cell content and in ceramide production was observed indicating that the toxin did not induce cell death by activating a sphingomyelinase (Fig. 2).

Figure 2.

Sphingomyelinase activity and ceramide production in LT-treated cells. 3 H-choline labelled cells were treated with LT (100 ng ml −1 ) for various length of time up to 150 min and lipids were extracted. Sphingomyelinase activity (circles) was estimated by quantification of radioactive sphingomyelin isolated as detailed in the experimental procedure section. Simultaneously, ceramide products (triangles) was measured in 3 H-palmitic acid-labelled cells after lipid extraction and thin layer chromatography. This is a representative of three different experiments.

Lethal toxin induces disruption of mitochondrial homeostasis in a dose- and time-dependent manner

Several parameters of mitochondria status were then investigated and results are illustrated in Fig. 2. A gradual drop in two different mitochondrial functional markers was observed in cells treated with increasing concentrations of LT-IP82. Labelling by nonyl acridine orange (NAO) was decreased in a dose-dependent manner, denoting cardiolipin peroxidation induced by the production of superoxide anions at the mitochondrial level (Fig. 2A). In addition, LT induced a decrease in mitochondrial membrane potential (Δψm) as measured with the DiOC6(3) fluorescent probe that accumulates in intact and not in depolarized mitochondria (Fig. 3A). Caspase-3 activation was also measured and, as shown in Fig. 3A, caspase-3 activity was induced by LT and detected even at the lowest LT-IP82 concentration (10 ng ml−1), a dose sufficient to reduce mitochondrial functions. As shown in Fig. 3B, LT-IP82 also induced, within the same dose range, the release of cytochrome c in the cytosol, a crucial step for caspase activation downstream of mitochondrial perturbations.

Figure 3.

Dose-response and time-course LT effects on mitochondrial membrane potential, cardiolipin staining, caspase 3-like activity and cytoplasmic release of cytochrome c.
A. After a 16 h treatment of HL-60 cells with LT-IP82 (10, 30 or 100 ng ml−1), percentage of cells with high mitochondrial membrane potential using DiOC6(3) ΔΨm high) (filled circles), high cardiolipin staining using nonyl acridine orange (NAO) (filled squares) and caspase 3-like activity using Phiphilux (filled triangles) was measured by flow cytometry analysis.
B. Cytosolic release of cytochrome c from mitochondria in response to various doses of LT-IP82 visualized by Western blot analysis. Upper gel, cytochrome c present in the cytosol and lower gel, cytochrome c present in the pellet fractions.
C. After various times (2–16 h) of HL-60 cell treatment with LT-IP82 (100 ng ml−1), mitochondrial membrane potential (ΔΨm high) (filled circles), cardiolipin staining (NAO) (filled squares) and caspase 3-like activity (filled triangles) measured as in A.
D. Cytochrome c release in the cytosol visualized as in 2B. Data in A and C are from three different experiments with two individual measurements.

Time-course studies performed at the highest LT-IP82 dose of 100 ng ml−1 revealed a time-dependent decrease in NAO fluorescence, starting after a 4 h treatment, that strictly paralleled the drop of mitochondrial membrane potential (Fig. 3C). A time-dependent increase of caspase-3 activity was also observed with time (Fig. 3C). Monitoring of cytochrome c release revealed that two hours of LT treatment was sufficient to induce some cytochrome c release in the cytosol (Fig. 3D) and the amount of cytosolic cytochrome c increased substantially thereafter.

Z-VAD partly prevents and Bcl-2 fully protects HL-60 cells from LT-induced apoptosis

To further clarify the sequence of events between mitochondrial depolarization, caspase activation and apoptosis, a permeable pan-caspase irreversible inhibitor, Z-VAD-fmk was used. Variations in mitochondrial ΔΨm and phosphatidylserine externalization were monitored in the presence of various LT-IP82 concentrations. As shown in Fig. 4A, caspase inhibition blocked apoptosis, as visualized by diminished annexin-V binding at 30 ng ml−1. At this dose, reduction of ΔΨm occurred with a similar magnitude, whether or not Z-VAD-fmk was present (Fig. 4B). Thus, mitochondrial perturbations occurred independently of caspase activation, a feature typical of apoptosis induced by the intrinsic pathway. At high LT dose (100 ng ml−1), Z-VAD-fmk inhibitory activity on cell death was only moderate as assessed by annexin-V labelling. Such an observation has been reported with various apoptosis inducers used at high doses. It is thought that, when major mitochondrial dysfunctions are induced, reactive oxygen species accumulate in cells where caspases are inhibited, leading to cell death by a process more akin to necrosis (Green and Reed, 1998).

Figure 4.

Effects of Z-VAD on phosphatidylserine exposure and mitochondria membrane potential in LT-treated HL-60 cells. Cells pretreated (stripped bars) or not (filled bars) with Z-VAD-fmk (20 µm) before LT-IP82 addition at 30 and 100 ng ml −1 . After 16 h, percentage annexin V-FITC positive-cells (A) and of cells with high mitochondrial membrane potential (B) was measured by flow cytometry. Experiments were repeated in three occasions and data are expressed as in Fig. 1.

Bcl-2 is known to protect cells from apoptosis by acting at the mitochondrial level and preventing cytochrome c release. Thus, we studied the effect of LT-IP82 in a HL-60 clone stably overexpressing Bcl-2. This clone was found to be unsentitive to LT-induced apoptosis (data not shown). Mitochondrial ΔΨm was evaluated in these cells treated with increasing amount of LT-IP82. As shown in Fig. 5A and in contrast to wild-type HL-60 cells, the presence of Bcl-2 was able to maintain a high ΔΨm at all LT-IP82 concentrations. Measurement of caspase-3 activity in Bcl-2 expressing cells exposed to LT-IP82 showed that Bcl-2 abrogated activation of caspase-3 (Fig. 5B), consistent with the notion that caspase-3 activation lies downstream of mitochondrial perturbations.

Figure 5.

Protective effects of Bcl-2 on LT-induced apoptosis.
A. Measurement by flow cytometry of mitochondrial membrane potential (ΔΨm) in wild-type HL-60 cells (empty circles) and in bcl-2-transfected HL-60 cells (filled circles) treated for 16 h with various concentrations of LT-IP82.
B. Capase 3-like activity measured using Phiphilux in the same cells. Experiments were repeated on three occasions.

Caspase-9 activation occurs prior to caspase-8 activation

To further delineate the sequence of caspase activation during LT-induced apoptosis, caspase-9, -3 and -8 cleavage was studied in kinetics. Incubation of HL-60 cells with 100 ng ml−1 of LT-IP82 led to the activation of all three caspases as denoted by the labelling with the corresponding specific antibodies of smaller fragments corresponding to the mature enzyme (Fig. 6). The use of an antibody reacting specifically with the 37/35 kDa fragments of active caspase-9 showed that they can be detected as soon as four hours after LT addition and increased with time. In contrast, no such fragments are found in HL-60/Bcl-2 cells at this time, consistent with the known protective effect of Bcl-2 on cytochrome c release and caspase-9 activation. A faint 37 kDa band was visualized in these cells after 16 h of incubation, indicating partial rupture of Bcl-2-mediated protection. Caspase-3 cleavage followed a similar kinetics with generation of the active enzyme fragments of 19 and 17 kDa after four hours of incubation. Caspase-3 activation was also inhibited by Bcl-2 and the apparition of a faint 19 kDa cleavage fragment at 16 h in the HL-60/Bcl-2 clone correlated with the minimal caspase-9 activation seen at this time. In contrast, procaspase-8 maturation was delayed, as the fragments of 43/41 kDa, derived from the 55/53 kDa pro-caspase isoforms, were detected only after 6 h of incubation. Moreover, caspase-8 activation was never detected in the Bcl-2 expressing clone, even after prolonged incubation times. The absence of caspase-8 activation in Bcl-2-expressing cells demonstrate that this caspase did not initiate the apoptotic signalling but was activated in a feed-back mechanism, downstream of mitochondrial perturbations. Indeed, the respective kinetics of caspase-9, -3 and -8 cleavage are coherent with an induction of apoptosis by LT in HL-60 cells via the intrinsic pathway.

Figure 6.

Time-course study of LT effects on caspase-9, -3 and -8 activation in HL-60 cells. Western blot analysis was performed with the indicated antibodies in HL-60 (HL-60 wt) and bcl-2-transfected HL-60 (Bcl-2) cells treated in kinetics with LT-IP82 (100 ng ml −1 ). The anti-caspase-9 antibody reacts with the cleaved fragments of the active caspase. The anti-capase-3 antibody recognizes the pro-enzyme (32 kDa) and the 19/17 kDa fragments of the active enzyme. The anti-caspase-8 antibody reacts with both 55 and 53 kDa isoforms of the pro-caspase and the 43 and 41 kDa corresponding cleavage fragments. The nature of the additional cross-reactive band of intermediate molecular weight detected with this antibody is presently unknown.

Cell death induced by LT is not related to the toxin effect on actin organization but could involve one of the small GTPases modified by the toxin

LT-IP82 has been shown to induce actin cytoskeleton disorganization, likely through its inactivating effect on Rac (Popoff et al., 1996). Therefore, we investigated whether apoptosis induction could be linked to cytoskeletal modifications. HL-60 cells were treated for 16 h with various agents known to disorganize the actin cytoskeleton. The effects of LT-IP82 were compared to those of a variant extracted from C. sordellii strain 9048 (LT-9048). LT-9048 has slightly different targets and, like LT-IP82, inactivates Ras, Rap, Rac but, unlike it, also inactivates Cdc42 and does not modify Ral. Iota toxin (Ia-Ib) from C. perfringens type E and cytochalasin D, that also disorganize the actin cytoskeleton, were studied in parallel. The former is known to have a direct effect on actin monomers (Richard et al., 1999) and the latter blocks actin microfilaments formation. As shown in Fig. 7A, at the concentrations used, all these toxins markedly decreased actin polymerization after an overnight treatment as shown by immunofluorescence microscopy analysis. LT-IP82 also induced the formation of long filopodia in HL-60 cells, speculated to be due to activation of Cdc42 subsequent to Rac inactivation (Popoff et al., 1996). Aside from LT-IP82, these toxins were unable to induce apoptosis in HL-60 cells and this was observed for LT-9048 used at doses as high as 1 µg ml−1. Cytochalasin D could induce apoptosis in a limited number of cells (less than 10%) after an overnight treatment.

Figure 7.

Effects of actin depolymerizing agents on cortical actin organization and cell death.
A. Cortical actin cytoskeleton examined by fluorescent microscopy after cell labelling with FITC-phalloidin in untreated HL-60 cells (a, control cells) and in cells treated for 16 h with LT-IP82 (100 ng ml−1) (b), LT-9048 (100 ng ml−1) (c), Iota toxin: ia-ib (1200 ng ml−1) (d), and cytochalasin D (1500 ng ml−1) (e).
B. Cortical actin organization studied as in A in HL-60 cells overexpressing Bcl-2 untreated (a) or treated for 16 h with LT-IP82 (100 ng ml−1) (b).Localization of LT in HL-60 cells upon LT treatment. Control HL-60 cells and HL-60 cells, treated either for 2 h or for 16 h with LT-IP82 (100 ng ml−1) as indicated, were examined by confocal immunofluorescence microscopy after a double staining with rabbit polyclonal LT antibodies revealed with a secondary antibody labelled with FITC (green) and with mouse monoclonal antibodies specific of endosomes (anti-EEA1), lysosomes (anti-Lamp-1) or mitochondria (mAB 1273) as indicated and revealed with a secondary antibody labelled with TRITC (red). Lane 1 (red) shows the intracellular localization of the different membrane structures, lane 2 anti-LT background and intracellular localization of this toxin, and lane 3, merged images with yellow fluorescence indicating co-localization of the markers.

Thus, no relationship was observed between toxic effects on actin polymerization and cell death. Interestingly, a cortical actin disorganization induced by LT-IP82 was also observed in bcl-2-transfected cells (Fig. 7B) which were protected from toxin-induced cell death.

LT localizes in mitochondria

Large clostridial toxins appear to have a tripartite organization similar to that of diphtheria toxin. They are known to be endocytosed and can also reach a subcellular acidic compartment in which they can be degraded. To study whether LT-IP82 could act at the mitochondrion level, we investigated its intracellular sublocalization in different organelles including endosomes, lysosomes and mitochondria. Confocal analysis using antibodies against EEA1 and Lamp1 to study endosomal and lysosomal localization, respectively, revealed that, after a 2 h treatment, LT-IP82 was not found in the endosomal compartment (Fig. 8). Some LT-IP82 was seen to co-localize with Lamp1 antibodies, indicating that at that time, part of the toxin has reached the lysosomal compartment (Fig. 8). At the same time, a co-localization between LT-IP82 and a mitochondrial protein was clearly observed and concerned most of these cellular organelles. After a long-term treatment (16 h), LT-IP82 and mitochondria were still partly co-localized; however, intracellular mitochondrial structures were no longer well defined.

Figure 8.

Localization of LT in HL-60 cell upon LT treatment. Control HL-60 cells and HL-60 cells, treated either for 2 h or for 16 h with LT-IP82 (100 ng/ml) as indicated, were examined by confocal immunofluorescence microscopy after a double staining with rabbit polyclonal LT antibodies revealed with a secondary antibody labelled with FITC (green) and with mouse monoclonal antibodies specific of endosomes (anti EEA1), lysosomes (anti Lamp-1) or mitochondria (mAB 1273) as indicated, and revealed with a secondary antibody labelled with TRITC (red). Lane 1 (red) shows the intracellular localization of the different membrane structures, lane 2 anti LT background and intracellular localization of this toxin, and lane 3 merged images with yellow fluorescence indicating co-localization of the markers.

Discussion

In the present work, we investigated the effects of lethal toxin from C. sordellii on the human promyelocytic cell line, HL-60. Our results clearly show that LT-IP82 induces cell death with several features of apoptosis including changes in nucleus morphology, DNA fragmentation, exposure of phosphatidylserine and activation of several caspases.

LT-IP82 has been reported to be endocytosed after binding to cell surface receptors. Indeed, cell incubation with LT-IP82 for two 2 h followed by resuspension in fresh culture medium led to the same level of cell death as the permanent presence of the toxin (data not shown). Lethal toxin does not appear however, to initiate apoptosis by direct or indirect triggering of a membrane receptor with a death domain as neither caspase-8 nor a sphingomyelinase were activated during the initial steps of cell treatment. In contrast, our investigations indicate that LT-IP82 induces cell death via disruption of mitochondrial homeostasis as assessed by an early decrease in mitochondrial membrane potential concomitant with cadiolipin alterations, and the activation of caspase-9 and caspase-3 prior to that of caspase-8. Initiation of LT-IP82-induced apoptosis at the mitochondrion is also supported by experiments showing that: (i) HL-60 cells stably overexpressing Bcl-2 are fully protected against apoptosis induced by LT-IP82; (ii) no caspase is activated in Bcl-2 expressing cells, whereas caspase-8 activation is insensitive to Bcl-2 when apoptosis is induced by the extrinsinc pathway; (iii) the large-spectrum caspase inhibitor, Z-VAD-fmk, is able to protect cells from apoptosis without modifying LT-IP82-induced mitochondrial alterations, showing that mitochondrial disruption is not secondary to caspase activation. Therefore, it appears that the toxin-induced mitochondrial perturbations are the initial events triggering apoptotic signalling.

Apoptosis induction via the intrinsic pathway has already been reported for various bacterial toxins. Indeed, Galmiche et al. (2000) have shown that Helicobacter pylori vacuolating toxin targets mitochondria and induces cytochrome c release via its N-terminal domain. Bantel et al. (2001) have reported that α-toxin from Staphyloccocus aureus, induces cell death and activates caspases via the mitochondrion independently of death receptor signalling. Staphylococcus aureus does not require internalization to be toxic as its α-toxin which induces pore formation is responsible for the bacteria cytotoxicity. Among the bacterial toxins known to inhibit small GTPases, toxins A and B of Clostridium difficile have been shown to induce apoptosis in intestinal cultured cells (Fiorentini et al., 1998; Brito et al., 2002) as well as other cell types (Linseman et al., 2001; Hippenstiel et al., 2002). The level of apoptotic signalling initiation has not been clearly established in these studies but apoptotic cell death was mostly interpreted as a direct consequence of Rho GTPases modification. However, toxin A has also been shown to localize at mitochondria and to cause early mitochondrial damages including a decrease in ATP concentration, reactive oxygen radicals production, and cytoplasmic release of cytochrome c in CHO cells (He et al., 2000). These authors also demonstrated that toxin A is able to directly induce leakage of cytochrome c from isolated mitochondria.

Our data concerning LT-IP82 show that this toxin initiates apoptosis at the level of mitochondria and we attempted to define the underlying triggering events. LT-IP82, which glucosylates Rac, Ras, Ral, induces cortical actin depolimerization, an effect likely to be related to the toxin effect on Rac. However, LT-IP82-induced apoptosis did not appear to be a consequence of actin disruption. Indeed, LT-9048, a LT-IP82 variant, did not induce apoptosis, despite its retained ability to trigger actin depolymerization. Another clostridial toxin, iota toxin from C. perfringens, was observed to dramatically depolymerize actin but had no apoptotic effect. By inactivating Ras, LT-IP82 could block the PI-3 kinase/Akt-mediated cell survival pathway (Marte et al., 1997; Downward, 1998), but LT-9048 also inactivates Ras, and thus the implication of Ras in the death-inducing properties of LT-IP82 appears very unlikely. However, the involvement of Ral inactivation cannot be ruled out as LT-9048 does not affect Ral function. A role for Ral in proliferation has been reported and the Ras/Ral pathway has been suggested to play a role in DNA synthesis and eventually cell transformation (Wolthuis and Bos, 1999). Expression of activated Ral has been shown to be sufficient to induce activation of NF-κB gene expression and cyclin D1 transcription, two key convergent points for mitogenic and survival signalling (Henry et al., 2000). Thus, if LT-IP82-induced apoptosis is linked to its glucosylating effect on a small G-protein, Ral could be candidate and this hypothesis requires further investigation. An important finding in our studies was that LT-IP82 can localize at mitochondria shortly after its uptake by the cell. Therefore, an alternative mechanism leading to apoptotic signalling could be a direct effect of LT on mitochondrial homeostasis. However, at the present time, no protein at the level of this organelle has been identified as a putative LT target. LT-9048 has not yet been cloned. It is likely that differences between the sequences of LT-IP82 and LT-9048 might help in understanding at the molecular level which domain, motif and/or amino acid(s) could be required for apoptosis induction. Extensive comparison between the two strains should allow for a better understanding of the relative implication of mitochondrial localization and/or Ral inhibition in the initiation of the apoptotic process.

Experimental procedures

Toxins and reagents

Lethal toxin (LT) was purified to homogeneity from Clostridium sordellii (strains IP82 and 9048) as extensively reported elsewhere (Popoff, 1987). RPMI-1640, antibiotics and glutamine were purchased from Gibco BRL (Paisley, UK). Fetal bovine serum was from Dutcher (Vilmorin, France). [9,10(n)-3H]palmitic acid (53 Ci mmol−1), [methyl-3H]Choline chloride (82 Ci mmol−1), PGEX-4T-3 vector and PVDF membrane (0.45 µm) were purchased from Amersham Pharmacia Biotech (Orsay. France). [glucose-14C(U)]-Uridine diphosphate glucose (UDP-glucose) (8.36 Ci mmol−1) was from NEN. Complete protease inhibitor cocktail were purchased from Roche (Roche Diagnostics, Meylan, France). Z-Val-Ala-DL-Asp(Omet)-fluoromethylketone (Z-VAD-fmk) was purchased from Bachem (BACHEM Biochimie SARL, Voisins-le-Bretonneux, France). Annexin-V-FITC was obtained from Immunotech (Beckman/Coulter, Marseille, France). Hoechst 33258, 3,3′-dihexyloxacarbocyanine iodide (DiOC6(3)), propidium iodide (PI) and nonyl acridine orange (NAO) were purchased from Molecular Probes (Leiden, NL) and PhiPhilux G1D2 was from OncoImmunin, Kensington, MD (USA). All other molecules and chemical products were from Sigma (Sigma-Aldrich, Saint Quentin Fallavier, France).

Specific antiserum to cytochrome c was purchased from Pharmingen and anti-caspase-8 and anti-caspase-3 Alexis (Alexis Biochemicals) and anti-capase-9 from Cell Signaling technology. Anti-early endosome A1 (EEA1) and anti-Lamp 1 were from BD Biosciences (Pharmingen and Transduction Laboratories, Le Pont de Claix, France) and anti-mitochondria antibodies mAB 1273 recognizing a 65 kDa protein was purchased from Chemicon (Temecula, CA, USA). Secondary antibodies: donkey anti-rabbit IgG and horse anti-mouse IgG conjugated to horseradish peroxidase were obtained from Amersham Pharmacia Biotech and Cell Signaling Technology (Beverly MA, USA), respectively, and used at the dilution indicated by the manufacturer. Rabbit anti-mouse IgG and anti-goat IgG were from Cappel. Polyclonal antibodies were obtained by immunizing rabbits against the N-term 1–546 of LT-IP82 produced as a fusion protein with a 6-His tag in E. coli. Enhanced chemiluminescence kit (Super signal, West-pico) was purchased from Pierce (Rockford, Il, USA).

Cell culture and cell treatments with agents

HL-60 cells were cultured in suspension in RPMI-1640 medium containing 10% foetal bovine serum, 1% penicillin-streptomycin, 2% l-glutamine at 37°C in a humidified incubator containing 5% CO2. HL-60 cells overexpressing Bcl-2 were obtained from Dr M. Allouche and cultured in the presence of geniticin (0.5 mg ml−1) as reported (Allouche et al., 1997). Cells were treated by direct dilution in the culture medium of different agents as indicated.

Sphingomyelinase activity measurements

(a) Sphingolipid quantification. For sphingomyelin quantification, cells were labelled with 3H-choline, treated with LT-IP82 toxin and lipids were extracted as described above for PLD activity measurement. Radioactive sphingomyelin was isolated as described by Andrieu et al. (1994). Briefly, after evaporation under nitrogen, residue of the organic phase was dissolved in 0.25 ml chloroform and 0.25 ml of 0.5 M methanolic NaOH and subjected to mild alkaline hydrolysis for 4 h at 37°C. After addition of 0.85 ml chloroform, 0.25 ml of 0.5 M methanolic HCl and 0.43 ml water, extracts were vortexed and centrifuged. The lower phase was washed twice with the upper phase of a mixture made of chloroform/methanol/H2O (3:48:47 by volume), dried under vacuum, dissolved in 0.1 ml of chloroform/methanol (2:1) and counted.

(b) Ceramide formation measurement. 

  • Cells were suspended in RPMI 1640 containing 1% FBS and labelled to isotopic equilibrium with 1 mCi ml−1 of [9.10–3H] palmitic acid (53 Ci mmol−1. Amersham, Les Ullis, France). Cells were then washed. Toxin treatment and lipid extraction were performed as above. After dissolution in chloroform/methanol, lipids were separated by thin layer chromatography as detailed elsewhere (Levade et al., 1993).

Apoptosis measurements

(a) Nucleus morphology.  Morphological alterations of chromatin were estimated by cell nuclei analysis using Hoechst 33258 dye. Cells washed twice in PBS containing 1% BSA were made adherent on poly l-lysine covered glass slides and fixed at room temperature for 30 min in a 3% paraformaldehyde solution freshly prepared. After rinsing with PBS, cells stained with 1 µM Hoechst 33258 dye in PBS and, then, photographed under a UV light microscope. The percentage of cells with a nucleus containing condensed chromatin and fragmented nuclei was counted.

(b) DNA fragmentation.  Equivalent number of cells (106) treated or not with LT were pelleted, washed in PBS and then lysed in 1 ml of a buffer made of NaCl (100 mM), Tris-HCl pH 8 (10 mM), EDTA (25 mM) and SDS (0.5%). Cell lysates were incubated with proteinase K for 4 h at 50°C. DNA was extracted first with an equal volume of phenol:chloroform:isoamyl alcohol (25:24:1 v/v/v) and, then, with chloroform:isoamyl alcohol (24:1 v:v). DNA precipitation was achieved by addition of 0.2 ml sodium acetate pH 5.3 (3M) and 2 ml of isopropanol for one hour at − 70°C. Precipitated DNA was collected by centrifugation for 30 min at 14 000 g at 4°C, washed in ice-cold 80% ethanol and air-dried. After resuspension in a buffer made of Tris-HCl, pH 7.5 (10 mM) and EDTA (1 mM), DNA was treated at 37°C with 0.5 mg ml−1 RNAse. DNA samples were then analysed by electrophoresis in a TBE agarose gel at 70 V and visualized by ethidium bromide staining.

(c) Determination of changes in mitochondrial membrane potential, mitochondrial cardiolipin staining and caspase-3 like activities by flow cytometry.

  • Mitochondrial transmembrane potential (Δψm) was measured as described elsewhere (Vayssière et al., 1994; Petit et al., 1995) with slight modifications (Gendron et al., 2001). Briefly, cells at 5 × 105 ml−1 were incubated with 2.5 nM DiOC6(3) to avoid any cytotoxicity (Rottenberg and Wu, 1998) and with 2 µg ml−1 propidium iodide (PI) for 15 min at 37°C. Cells were analysed using a FACScalibur (Becton Dickinson, San José, CA, USA), gating on the forward and side scatter to exclude debris. The fluorescence was excited at 488 nm and collected in FL-1 [Band Pass 530 ± 30 nm for DiOC6(3)], or FL-3 (Long Pass 670 nm for PI). A minimum of 5 × 103 events were acquired in list mode and analysed with CellquestTM software (Becton Dickinson). For mitochondrial cardiolipin alteration measurement, cells were incubated for 30 min at room temperature with 5 µM nonyl acridine orange (NAO, Molecular Probes, EU, Oregon) and the fluorescence was recorded in FL2 (585 = 10 nm).

The cell-permeable fluorogenic substrate, PhiPhilux G1D2, containing the sequence GDEVDG was used to detect caspase-3-like activity in intact cells. After treatment with LT, cells were harvested and washed twice in PBS. Cells (106) were resuspended in 50 µl substrate solution and incubated for 1 h at 37°C in the dark. After washing, cells were suspended in PBS; the fluorescence emission at 530 ± 30 nm (FL1) was measured as previously described by Gendron et al. (2001).

(d) Phosphatidyl  serine exposure and necrotic cells.  Exposed phosphatidylserine (PS) on the outer plasma membrane and necrotic dead cells death were measured by staining cells with annexin-V-FITC (1 µg ml−1) and PI (2 µg ml−1) for 10 min at 4°C.

(e) Measurement of caspase activation by Western blot analysis.  After treatment with LT, 107 cells were extracted by lysis in 0.1 ml ‘Laemmli buffer’ made of 60 mM Tris buffer pH 6.8, SDS 2% and glycerol 10%. Lysates were sonicated four times on ice during 40 s and, then, ultracentrifuged for 30 min at 100 000 g in a TL100 ultracentrifuge (Beckman). After by SDS-PAGE electrophoresis, proteins were transferred to PVDF membrane (0.45 µm), at 15 mA overnight at 4°C in transfer buffer made of Tris-base (20 mM), glycine (150 mM) and ethanol (20%). Non-specific binding was blocked by incubation in a blocking buffer, for 4 h at room temperature. This was followed by incubation for one and a half hours with specific primary antibody and, after washing the unbound antibody in phosphate-buffered saline (PBS) (pH 7.4) containing 0.3% tween 20 (PBS-Tween), the membranes were incubated with the relevant secondary HRP-conjugated antibody. Following washes in PBS-tween, the blots were developed using enhanced chemiluminescence kit.

(f) Measurement of cytochrome c release.  For the determination of cytochrome c release from mitochondria, control and LT-treated HL-60 cells were washed twice with PBS and incubated for 20 min on ice in a lysis buffer made of 20 mM Hepes, 10 mM KCl, 2 mM MgCl2, 1 mM EGTA, 1 mM EDTA 1 mM dithiothreitol (DTT) and complete protease inhibitor cocktail (1×). Cells were then carefully broken by 25 strokes in a Dounce (pestle B) and centrifuged at 800 g for 10 min. Cytosolic extracts were recovered after centrifugation at 75 000 r.p.m. for 30 min in a TL-100 Beckman ultra centrifuge. For each condition, 100 µg of cytosolic extract proteins were loaded on a 15% SDS-polyacrylamide gel, cytochrome c was visualized by Western blotting using a specific cytochrome c antiserum.

Immunofluorescence microscopy

Cells were fixed with 3% paraformaldehyde for 20 min and incubated in a solution of NH4Cl (50 mM) containing 10 µM CaCl2 and 10 µM MgCl2 for 10 min. After several washes with PBS, cells were permeabilized with Triton X-100 0.3% for 5 min. The primary antibodies, diluted in PBS were then incubated for 1 h with cells and after several washes, the fluorescence-labelled antibodies at the required dilution were incubated for 30 min with cells. After cell washing, samples were mounted in Mowiol and observed with a confocal laser scanning microscope (Bio-Rad MRC-100) attached to a diaphot 300 microscope (Nikon).

Acknowledgements

This work was supported by INSERM and Association de Recherche sur le Cancer (ARC) (grants n°9452,7550 and 4612). N. Ben E. H. was supported by a fellowship from ARC and V. M. is a fellow from Association de Recherche sur la Polyarthrite (ARP).The authors are extremely grateful to Nicole Riché for her dedicated technical assistance.

Ancillary