Apoptogenic Ganglioside GD3 Directly Induces the Mitochondrial Permeability Transition*

Early events in apoptotic cascades initiated by ceramides or by activation of the surface receptor CD95 (Fas/APO-1) include the formation of ganglioside GD3. GD3 appears to be both necessary and sufficient to propagate this lipid-mediated apoptotic pathway. Later events common to many apoptotic pathways include induction of the mitochondrial permeability transition (PT) and cytochrome c release, which in turn triggers downstream caspases and cell death. The links between GD3 formation and downstream stages of apoptosis are unknown. We report that ganglioside GD3 directly induces the PT in isolated rat liver mitochondria at 30–100 μm in the presence of exogenous substrate (succinate) and at ∼3 μm in the absence of exogenous substrate. In contrast, other gangliosides tested (e.g. GM1) have only weak stimulatory effects in the presence of succinate and protect against PT induction in the absence of respiratory substrates. GD3-mediated induction of PT was antagonized by known PT inhibitors, namely cyclosporin A, ADP, trifluoperazine, and Mg2+. GD3 induced PT even in the presence of submicromolar Ca2+; GD3 is therefore the first biological PT inducer identified that does not require elevated Ca2+. Exposure to GD3 also led to mitochondrial cytochrome c release. In contrast, C2-ceramide, which can initiate the lipid-mediated apoptotic cascade in susceptible cells, failed to either induce PT or release cytochrome c. These observations suggest that GD3 propagates apoptosis by inducing the PT and cytochrome crelease. This model provides a mechanistic link between the earlier and later stages of CD95-induced/ceramide-mediated apoptosis.

Cross-linking of CD95 activates an apoptotic cascade in many cell types, including myeloid and lymphoid cells as well as primary cells from the liver, heart, and lung (1,2). One branch of this apoptotic cascade is proteolytic (3,4), whereas the other involves lipid mediators (5)(6)(7). Signaling along the lipid pathway is initiated within 5-15 min after CD95 activation by the sequential activation of phosphatidylcholine-specific phospholipase C and acidic sphingomyelinase (5)(6)(7). Acidic sphingomyelinase-mediated cleavage of sphingolipids produces ceramides, which are themselves sufficient to induce apoptosis in hematopoetic cells (8 -11). Cells genetically unable to acti-vate acidic sphingomyelinase (Niemann-Pick lymphoblasts) do not undergo normal CD95-mediated apoptosis; this defect can be bypassed by the addition of exogenous ceramides (12). The lipid-mediated pathway that includes ceramide production may be the most efficient of the multiple pathways that can mediate CD95-induced apoptosis (12)(13)(14)(15). TNF 1 -mediated cytotoxicity also involves ceramide and shares aspects of the upstream apoptotic pathways with CD95 (16 -20).
Ceramides appear to require the presence of mitochondria and a "cytosolic factor" to trigger downstream apoptotic events (21). Ceramide elevation during the apoptotic cascade rapidly (5 min) induces the activation of ganglioside GD3 synthase (␣-2,8-sialyltransferase) within the Golgi (22,23). This portion of the lipid-mediated pathway appears essential for ceramideinduced apoptosis (12). DeMaria et al. (13) have shown that: 1) activation of GD3 synthase is sufficient to induce apoptosis in cultured lymphoid cells (HuT78), 2) the addition of exogenous GD3 is sufficient to induce apoptosis in HuT78 and myeloid (U937) cells, and 3) inhibition of GD3 synthase is sufficient to prevent CD95-induced apoptosis. These results indicate that GD3 is both necessary and sufficient for CD95-and ceramidemediated apoptosis.
The immediate downstream targets of GD3 have not been identified. We hypothesize that mitochondria are the immediate downstream target of GD3. Consistent with this is the known transport of gangliosides from the Golgi to the mitochondria (24). This hypothesis is also supported by the observation (13) that GD3-mediated apoptosis includes the loss of mitochondrial membrane potential (⌬⌿). Loss of ⌬⌿ is a classic hallmark of the mitochondrial permeability transition (PT) (25)(26)(27)(28). Thus, the data suggest that the presence of GD3 may initiate an event or series of events that culminate in the induction of the PT and subsequent activation of the downstream stages of apoptosis.
PT induction has been proposed to contribute to both cellular apoptosis and necrosis in a variety of situations, including ischemia reperfusion, excitotoxicity, and T-lymphocyte cell death (29 -34). The PT involves the opening of a pore in the inner mitochondrial membrane that allows free diffusion of solutes with mass under 1500 daltons (25)(26)(27)(35)(36)(37). PT induction abolishes oxidative phosphorylation, leads to loss of the mitochondrial proton gradient, and allows efflux of mitochondrially sequestered calcium into the cytoplasm (25)(26)(27)(35)(36)(37). These consequences on energy metabolism and calcium homeostasis, as well as the biochemical identity of the first PT inducers studied (high calcium, oxidants, inorganic phosphate), led to a recognition of the probable role of PT in ischemia reperfusion injury (29).
The involvement of PT in apoptosis has been recognized more recently. The PT has been proposed to propagate the apoptotic cascade by triggering the mitochondrial release of apoptosis inducing factor, which stimulates nuclear fragmentation, and cytochrome c, which activates the downstream caspases that effect the end stage of apoptosis (38 -41). Several lines of evidence converge to support the view that the PT propagates the downstream stages of apoptosis mediated by the lipid-dependent pathway that includes ceramide and GD3. First, studies in cell-free systems indicate that PT is sufficient to induce the downstream stages of apoptosis (42). Second, loss of ⌬⌿, a hallmark of the PT, can precede and correlate with various markers of apoptotic cell death induced by this family of inducers (CD95, TNF, ceramide, GD3) (13,43). Third, cyclosporin A (CsA), the best established pharmacologic PT inhibitor, delays or prevents the progression of induced apoptosis in cultured cells (34,44). Fourth, treatment with CsA delays the loss of ⌬⌿ during apoptosis mediated by TNF and ceramide (43). Fifth, TNF-induced cytotoxicity in L929 cells, which may occur via apoptosis and/or necrosis (45,46), is mediated by ceramide. Cytotoxicity in this model is delayed by PT inhibitors, suggesting the involvement of PT in cell death (47,48). We tested the hypothesis that GD3 directly mediates PT induction by examining the ability of GD3 and related compounds to induce the PT. Data presented demonstrate that GD3, but not other gangliosides or C 2 -ceramide, accelerated the PT in isolated rat liver mitochondria in either the presence or absence of exogenous substrate. Strikingly, GD3 was capable of inducing the PT in the presence of submicromolar Ca 2ϩ . GD3 is therefore the first physiological inducer clearly shown not to require elevated Ca 2ϩ . Furthermore, GD3, but not C 2 -ceramide, induced the release of cytochrome c from mitochondria. These results are consistent with our hypothesis that GD3 itself directly enlists mitochondria into the apoptotic cascade.

EXPERIMENTAL PROCEDURES
Mitochondrial Isolation-Liver mitochondria were isolated from 4 -6 month old male Fischer 344 rats essentially as described previously (49). Briefly, livers were rapidly homogenized in ice-cold isolation buffer (250 mM mannitol, 75 mM sucrose, 100 M K-EDTA, 10 mM K-HEPES, pH 7.4) supplemented with 500 M K-EGTA (pH 7.4). Homogenates were centrifuged at 1000 ϫ g for 10 min. Supernatants were removed and centrifuged at 10,000 ϫ g for 15 min. Pellets were washed three times in isolation buffer supplemented with 0.5% fatty acid-free bovine serum albumin (Sigma A-6003). The first wash buffer was also supplemented with 500 M EGTA. The final mitochondrial pellet was resuspended in the same buffer without EGTA or bovine serum albumin.
FIG. 1. GD3 promotes but GD1a, GM1, and GT1 inhibit PT induction. Representative traces from a minimum of three experiments on the effects of GD3 (A-C) or GD1a, GM1, and GT1b (D-F) on PT induction in the absence of exogenous succinate. Induction by Ca 2ϩ was studied in liver mitochondria isolated from 4 -6-month-old male Fischer 344 rats (49). As noted under "Experimental Procedures," the basic buffer contains ϳ3 M Ca 2ϩ . The efficacy of 30 M GD3 to promote PT induction varied from rat to rat; the greatest response observed for this dose is shown. In all figures, data for all traces in a single panel were collected at the same time from the same mitochondrial preparation. In all figures, data are representative of studies on 3-10 independent mitochondrial preparations.
PT Induction-PT induction was assessed spectrophotometrically essentially as described previously (49,50) by suspending ϳ0.1-0.2 mg of mitochondrial protein at room temperature in 200 l of 215 mM mannitol, 71 mM sucrose, 5 mM K-HEPES (pH 7.4) in the presence (or, where indicated, absence) of 10 mM sodium succinate. Gangliosides (Alexis, Matreya) were dissolved in water. C 2 -ceramide (Biomol) was dissolved in ethanol, and the appropriate concentrations of ethanol (Յ1.5%) were added to controls run in parallel with ceramide samples. Changes in absorbance at 540 nm (A 540 ) were followed for 2 h using a SpectraMax 250 Plate Reader (Molecular Dynamics).
Divalent Cation Measurement and Buffer Preparation-The total divalent cation concentration present in the buffer was estimated by a spectrophotometric sequestration assay to be ϳ5 M (60% of which, ϳ3 M, was Ca 2ϩ ) using 2.5 M of the indicators Mag-Fura-2 and Fura Red and the cation chelators EDTA, EGTA, and tetrakis(2-pyridylmethyl)ethylenediamine. Results for Ca 2ϩ concentrations determined by this method are consistent with atomic absorption studies.
Use of this method revealed that preparing buffers from most standard research grade chemicals and deionized water led to levels of divalent cations, particularly Ca 2ϩ , which were unacceptable (as high as ϳ30 M Ca 2ϩ ). We therefore prepared all buffers from deionized water (Ͼ15 megohm) and the highest purity reagents available.
Western Blotting-Mitochondrial supernatants and pellets were collected following a 5-min centrifugation (ϳ14,000 ϫ g). Western blotting was done using minor modifications of established procedures. Proteins were separated using 4 -20% Tris-tricine polyacrylamide gels. Following transfer, membranes were blocked for 1 h in 5% fetal calf sera, 0.2% bovine serum albumin, and 0.5% Tween 20 in phosphate-buffered saline. Incubations with primary antibody were carried out overnight at a dilution of 1:2000. Incubations (30 min) with goat anti-mouse, alkaline phosphatase conjugate (Bio-Rad) were also carried out at a dilution of 1:2000. Nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (Bio-Rad) were used as substrates according to the supplier's instructions. All incubations were carried out at room temperature.

RESULTS
GD3 Directly Induces the PT-PT induction in liver mitochondria isolated from male Fischer 344 rats was monitored spectrophotometrically as loss of absorbance using established methods (49,50). Concentrations of GD3 Ն3 M increased the rate of PT induction in the absence of exogenous substrate (Fig.  1, A and B). This dose of GD3 (3 M) is ϳ40-fold lower than that required to induce apoptotic cell death in 50% of cultured HuT78 cells at 24 h (13). The ability of GD3 to accelerate induction was most notable in the absence of added Ca 2ϩ (i.e. total Ca 2ϩ ϳ3 M, see "Experimental Procedures"). Indeed, Ն12 M exogenous Ca 2ϩ (ϳ15 M total) induced PT in the absence of GD3 with such rapid kinetics that additional effects of GD3 were undetectable (Fig. 1C).
The ability of gangliosides GD1a, GM1, and GT1b to induce PT was examined to control for nonspecific effects of gangliosides, such as the effects on lipid organization or ion permeability (51,52). These gangliosides do not to induce apoptosis in the HuT78 system (13). Unlike GD3, gangliosides (GD1a, GM1, or GT1b) delayed PT induction at 30 M and completely prevented induction at 100 M (Fig. 1D). Some protection persisted even at higher Ca 2ϩ concentrations (Fig. 1, E and F).
Ganglioside specificity was maintained in the presence of exogenous substrate, although the concentration of GD3 required to induce PT increased 10 -30-fold. In the presence of 10 mM succinate, which increases mitochondrial resistance to PT induction, 100 M GD3 rapidly induced PT at all calcium concentrations tested (Fig. 2, A-D); 3-10 M GD3 accelerated PT induction when total Ca 2ϩ was ϳ27 M (i.e. 24 M added Ca 2ϩ , Fig. 2D), a concentration typically used for PT studies. Gangliosides GD1a, GT1b, and GM1 had only minimal effects until total calcium was ϳ27 M (Fig. 2, E-H), again demonstrating a specific response to GD3.
In contrast to GD3, C 2 -ceramide had no consistent effects on PT induction in either the presence or absence of exogenous substrate (Fig. 3), even at concentrations of C 2 -ceramide 4-fold higher than those that induce cellular apoptosis (12). In some cases, such as that shown in Fig. 3A, C 2 -ceramide actually delayed PT induction. Similar results were observed in the presence of increased Ca 2ϩ (not shown).
GD3-Ca 2ϩ Interactions-Our studies show that GD3 acted synergistically with free Ca 2ϩ to induce PT. GD3 induction was enhanced by added Ca 2ϩ in the presence of either endogenous substrate (Fig. 1, A-C, and Fig. 4, traces D-F) or added succinate (Fig. 2, C and D). These data are consistent with previous studies that suggested the necessity of Ca 2ϩ as a co-factor in PT induction (25)(26)(27). The combined effect of GD3 and Ca 2ϩ on PT induction is saturable. For example, when the PT was induced in the presence of total Ca 2ϩ , Ն15 M, additional effects of GD3 were nearly undetectable ( Fig. 1C and Fig. 4A-E, F traces). Similarly, 30 or 100 M GD3 were insensitive to Ca 2ϩ addition (Fig. 4, D and E). This saturation property of GD3 is unlike the interaction of other inducers with Ca 2ϩ .
Because the potency of GD3 as a PT inducer became more apparent as less Ca 2ϩ was added, we further probed the interactions between GD3 and Ca 2ϩ . These data show that GD3 also acted independently of free Ca 2ϩ to induce PT. Free Ca 2ϩ contributed by buffer components (ϳ3 M) was reduced by adding 12, 24, or 36 M of the divalent cation chelator EGTA (K d , 10 Ϫ10.86 for Ca 2ϩ ; calculated free Ca 2ϩ , ϳ5, 2, or 1 pM, respectively). The addition of 12 M EGTA, which reduces free Ca 2ϩ to ϳ5 pM, prevented spontaneous PT induction in controls (no GD3) for at least 2 h (Fig. 4A). Treatment with Ն10 M GD3 overcame the EGTA-mediated protection ( Fig. 4A-C, C traces). Higher doses of EGTA progressively delayed, but did not prevent, GD3-mediated PT induction (Figs. 4A, A and B traces, and 5C). GD3 also induced PT with subpicomolar free Ca 2ϩ in the presence of succinate (Fig. 5D).
Inhibitors Confirm PT Induction-We wished to confirm that the optical changes were because of PT induction and not a nonspecific loss of absorbance. We therefore tested the efficacy of four well characterized inhibitors of the PT, CsA, MgCl 2 , ADP, and trifluoperazine (25)(26)(27)53) to prevent GD3-mediated effects (Fig. 5).
CsA-In the absence of succinate, 5 M CsA, but not 0.5 M CsA, prevented PT induction by 100 M GD3 (Fig. 5A, traces  AH, BH, and CH). Both doses of CsA protected against 10 M GD3 (Fig. 5A, traces AL, BL, and CL). In the presence of succinate, both 0.5 and 5 M CsA delayed GD3-induced PT but did not prevent it (Fig. 5B). The higher CsA dose consistently protected, whereas only some preparations were protected by the lower dose.
MgCl 2 -The addition of 3 mM MgCl 2 or 100 M EGTA resulted in essentially equal inhibition in either the presence or absence of succinate (compare traces B and C in Fig. 5, C or D). This finding suggests that MgCl 2 acted by inhibiting a divalent cation-dependent event, consistent with the known role of MgCl 2 as a competitive inhibitor of the calcium uniporter. Note that GD3 induction in the presence of 100 M EGTA is further delayed relative to induction in the presence of 36 M EGTA (Fig. 4A-E, A traces). ADP-In either the presence or absence of succinate, ADP prevented PT induction by 100 M GD3 (Fig. 5, E and F).
Trifluoperazine-The phospholipase A 2 inhibitor trifluoperazine prevented GD3-mediated PT in both the presence and absence of succinate (Fig. 5, G and H). Protection was dose-dependent, and the protective doses corresponded with those that blocked Ca 2ϩ -mediated PT induction (not shown).
The inhibitor data shown (Fig. 5) confirm that the GD3mediated loss of absorbance is because of the induction of the PT.
GD3 Releases Cytochrome c-Further support for the hypothesis that GD3 can directly enlist mitochondria into the apoptotic cascade is that GD3-mediated PT induction released mitochondrial cytochrome c (Fig. 6). In contrast, cytochrome c was not released from mitochondria in either control or C 2ceramide-treated samples. Release was more variable and less complete in the absence of succinate, suggesting that mitochondrial respiration may promote the release of cytochrome c. This possibility is under further investigation.

GD3 Displays Structural Specificity as a PT Inducer-Under
the conditions used in this study, GD3 induced a dose-dependent loss of absorbance (Figs. 1, 2, and 4). Use of a series of inhibitors (CsA, MgCl 2 , ADP, and trifluoperazine; Fig. 5) confirmed that this loss of absorbance resulted from induction of the PT. In contrast, related gangliosides such as GD1a, GM1, or GT1b, which do not induce cell death, failed to induce PT ( Figs. 1 and 2). The structural specificity of the sialogangliosides in PT induction is reminiscent of that observed for hydroxyalkenals, where 4-hydroxyhexenal induces the PT at femtomolar levels, but 4-hydroxynonenal requires micromolar concentrations (49). The protective action of GM1 may contribute to the protective capacity of this compound to prevent cell death in many systems, including models of ischemia (54,55).
Ceramides can induce both apoptotic and necrotic cell death that are associated with mitochondrial dysfunction, including the loss of ⌬⌿ (8 -12, 56). We have demonstrated that biologically active C 2 -ceramide has little or no direct effect on PT induction or cytochrome c release in isolated mitochondria (Figs. 3 and 6). This result differs from a previous report that C 2 -ceramide can induce PT in isolated liver mitochondria (56). The differences may be because of important differences in experimental conditions. The other study used digitonintreated mitochondria that were purified using a Percoll gradient and assayed in the presence of phosphate, rotenone, and oligomycin.
GD3 Acts Both Synergistically with Ca 2ϩ and Independently of Ca 2ϩ to Induce PT-In vitro studies in isolated mitochondria have provided predictive insight into PT induction in cells. However, the PT inducers previously identified by studies in isolated mitochondria, with the exception of the nonphysiologic thiol cross-linking agent phenylarsine oxide, require supraphysiological levels of calcium (e.g. Ն25 M). Therefore, in vivo PT induction, which occurs without such a profound increase in intracellular Ca 2ϩ , must be stimulated by an unidentified component. Our study shows that GD3 may be such a compound.
In addition to acting synergistically with Ca 2ϩ to induce PT (Figs. 1, 2, and 4), GD3 can also act independently of free Ca 2ϩ to induce PT (Fig. 4). This behavior is unlike that previously described for naturally occurring PT inducers (25)(26)(27). The ability of GD3 to induce PT in the presence of submicromolar Ca 2ϩ is important, because intracellular levels of Ca 2ϩ do not rise above 0.4 M during the early stages of CD95-mediated apoptosis (57,58).
Overall, the data presented suggest a biphasic interaction between GD3, divalent cations, and the PT. One phase is apparent when free Ca 2ϩ varies from 0 -15 M (traces C-F in all panels of Fig. 4) and presumably reflects the effects of Ca 2ϩ transported by the mitochondrial uniporter (e.g., Ca 2ϩ accumulation/cycling) (26). The second phase occurs when free Ca 2ϩ varies in the picomolar range, which is well below the concentration at which the mitochondrial Ca 2ϩ uniporter acts (26). GD3-mediated PT induction was progressively delayed by increasing EGTA from 12 to 100 M (Figs. 4, traces A-C, and 5C). A possible explanation for the second phase is that GD3-mediated PT requires a divalent cation that either GD3 or mitochondria bind with an affinity comparable to that of EGTA.
PT Induction by GD3 Can Enlist Mitochondria Into the Ap- optotic Cascade-The data presented for isolated mitochondria link GD3 to PT induction (Figs. 1, 2, 4, and 5) and cytochrome c release in vitro (Fig. 6). These data are analogous to data from cell culture studies that link GD3 to the loss of ⌬⌿ and subsequent apoptotic cell death in vivo (13). Specifically, in cultured cells, elevation of intracellular GD3, whether by synthesis or exogenous addition, is correlated with a loss of ⌬⌿ and (other) downstream events associated with progression through the lipid-mediated apoptotic pathway (13). We demonstrated that exposure to GD3 directly induced the PT, which abolishes ⌬⌿, and released cytochrome c, which can propogate the apoptotic cascade. These data are consistent with reports that PT induction results in cytochrome c release (59,60). These data also confirm that GD3-mediated PT leads to the release of at least one of the apoptogenic factors associated with activation of downstream caspases. Our data thus support the hypothesis that the mechanism by which GD3 propagates the lipid-mediated apoptotic pathway is a direct induction of the PT (Figs. 1 and 2) and release of mitochondrial cytochrome c (Fig. 6).
Conclusions-Evidence of GD3-mediated PT induction and cytochrome c release provide a biochemical mechanism linking the early and late stages of CD95-mediated and ceramideinduced apoptosis. The data presented indicate that PT induction and apoptosis can be propagated by the formation of GD3 independent of elevated free intracellular Ca 2ϩ . This contrasts with previous work that focused on models in which the PT was driven primarily by high Ca 2ϩ . Although Ca 2ϩ appears to be unnecessary in GD3-mediated PT, our data demonstrate that elevated intracellular Ca 2ϩ can synergistically accelerate PT induction both in the presence and in the absence of exogenous respiratory substrates. The observed capacity for synergy between Ca 2ϩ and GD3 in PT induction provides a mechanism by which elevated intracellular Ca 2ϩ may contribute to various forms of cell death involving lipid mediators. These include apoptosis, and probably necrosis, initiated by the activation of CD95 or the TNF receptor as well as cell death subsequent to radiation exposure or HIV infection (8,9). Further investigations are needed to determine the mechanisms by which differ-