Cytokine Response Modifier A (CrmA) Inhibits Ceramide Formation in Response to Tumor Necrosis Factor (TNF)-α: CrmA and Bcl-2 Target Distinct Components in the Apoptotic Pathway

The previously described cell lines expressing CrmA or mutant CrmA and those overexpressing Bcl-2 were derived from a TNF-α–sensitive MCF-7 parental line (4, 6, 41) (gift of Muneesh Tewari and Dr. Vishva Dixit, University of Michigan, Ann Arbor). Cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) and 0.2% sodium bicarbonate. 500 μg/ml G418 was added to the CrmA cell line and its vector while 150 μg/ml hygromycin was added to the Bcl-2 cell line and its vector. Experiments were done in the absence of G418 or hygromycin. Cell viability was determined by the ability to exclude trypan blue.

Cells were seeded at 2 × 10⁵ cells/ml in a 10 ml volume, rested overnight, and then treated with 1.2 nM TNF-α. Lipids were collected according to the method of Bligh and Dyer (42). In brief, cells were pelleted, washed once with PBS, and then extracted with 3 ml chloroform/methanol (1:2, vol/vol) in 13 × 100 mm screw-top glass tubes. The monophase was mixed, 0.7 ml water was added, and the samples were rested for 10 min. The organic and aqueous phases were subsequently separated by addition of 1 ml chloroform and 1 ml water followed by vigorous shaking and centrifugation at 1,000 rpm. The organic phase was carefully removed and transferred to a new tube, and the samples were dried under N₂. Lipids were then resuspended in 1 ml of chloroform. Ceramide levels were measured using a modified diacylglycerol kinase assay (43, 44) using external standards. In brief, 80% of the lipid sample was dried under N₂. The dried lipid was solubilized in 20 μl of an octyl-β-d-glucoside/dioleoyl phophatidylglycerol micellar solution (7.5% octylβ-d-glucoside, 25 mM dioleoyl phosphatidylglycerol) by 2 cycles of sonication in a bath sonicator for 60 s followed by resting at room temperature for 15–20 min. The reaction buffer was prepared as a 2 × solution containing 100 mM imidazole HCL, pH 6.6, 100 mM LiCl, 25 mM MgCl₂, 2 mM EGTA. To the lipid micelles, 50 μl of the 2 × reaction buffer were added, 0.2 μl of 1 M dithiothreitol, 5 μg of diglycerol kinase membranes, and dilution buffer (10 mM imidazole, pH 6.6, 1 mM diethylenetriaminepentaacetic acid, pH 7) to a final volume of 90 μl. The reaction was started by adding 10 μl 2.5 mM [γ-³²P]ATP solution (specific activity of 75,000– 200,000 cpm/nmol). The reaction was allowed to proceed at 25°C for 30 min. Lipids were extracted as described above and a 1.5-ml aliquot of the organic phase was dried under N₂. Lipids were then resuspended in a volume of 100 μl methanol/chloroform (1:20, vol/vol) and 20 μl was spotted on a 20-cm silica gel thin layer chromatography plate. Plates were developed with chloroform/acetone/methanol/acetic acid/H₂O (50:20:15:10:5), air dried, and then subjected to autoradiography. The radioactive spots corresponding to phosphatidic acid and ceramide phosphate, and the phosphorylated products of diacylglycerol, and ceramide, respectively, were scraped into a scintillation vial containing 4 ml scintillation fluid and counted on a scintillation counter. Linear curves of phosphorylation were produced over a concentration range of 0–640 pM of external standards (dioleoyl glycerol and CIII ceramide; Sigma Chemical Co., St. Louis, MO). Diacyglycerol and ceramide levels were always normalized to lipid phosphate which was measured according to the method of Rouser et al. (45). In brief, 20% of the lipid sample was dried down under N₂ and oxidized with 150 μl of 70% perchloric acid on a heating block at 160°C for 45 min. The tubes were allowed to cool, and then 830 μl of H₂0 was added, followed by 170 μl of 2.5% ammonium molybdate, and 170 μl of 10% ascorbic acid with vortexing after each addition. The tubes were then incubated at 50°C for 15 min, allowed to cool, and absorbance read at 820 nm and compared to standard.

Cells were seeded at 2.5 × 10⁵/well of a 6-well plate in a 2-ml volume of RPMI with 2% FBS and treated with ¹⁴C₆ceramide at 20 μM. Cells were harvested by scraping at the indicated time points, washed twice with PBS, and the radioactivity retained in the pellet was counted in a scintillation counter.

For experiments analyzing PARP proteolysis, cells were seeded and treated as described for the ceramide measurement experiments. At the indicated time points, cells were harvested by scraping in media followed by centrifugation at 4°C and washed once with ice-cold PBS. The cell pellet was then resuspended in 50 μl PBS and lysed with sample buffer (30 mM Tris-HCL [pH 6.8], 10% glycerol, 6% β-mercaptoethanol, 4% SDS) and boiled for 10 min. Protein concentrations were determined using the Bio-Rad assay.

Equal amounts of protein, usually 100 μg, were resolved by SDS-PAGE and transferred to a nitrocellulose membrane. PARP and its cleaved fragment were detected using a rabbit polyclonal antiserum at a dilution of 1:2,000, and a goat anti–rabbit secondary antibody at a dilution of 1:5,000. The signal was visualized by enhanced chemiluminescence (Amersham Intl., Buckinghamshire, UK).

Nuclear extracts were prepared as described previously (46, 47). In brief, 2 × 10⁶ cells were washed once in PBS. The cell pellet was rapidly frozen in dry ice and isopropanol, and then thawed by adding 100 μl of ice-cold buffer A (10 mM Hepes, pH 7.9, 10 mM KCl, 1.5 mM MgCl₂, 1 mM dithiothreitol) resulting in 100% cell lysis. The nuclei were pelleted by microcentrifugation at 3,500 rpm for 10 min at 4°C. The supernatant was discarded and the nuclei suspended in 15 μl of buffer C (20 mM Hepes, pH 7.9, 0.4 M NaCl, 1.5 mM MgCl₂, 25% vol/vol glycerol, 0.2 mM EDTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride). The suspension was mixed gently for 20 min at 4°C, and then microcentrifuged at 14,000 rpm for 20 min at 4°C. The supernatant was diluted with 50–70 μl of buffer D (20 mM Hepes, pH 7.9, 50 mM KCl, 20% vol/vol glycerol, 0.2 mM EDTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride) and aliquots stored at −80°C. Protein concentrations were determined using the Bio-Rad assay. Electrophoretic mobility-shift assay reactions were performed in 20 μl volume, using 8–10 μg of nuclear extract in the presence of 1 μg of poly(d[I-C]) 1 μg of pd(N)₆, and 10 μg of BSA. Incubations were in the presence of HDKE buffer with the following final concentrations: 20 mM Hepes, pH 7.9, 50 mM KCl, 1 mM EDTA, and 5 mM dithiothreitol. Reactions were started by the addition of 10,000–50,000 cpm radiolabeled nuclear factor (NF)- κB oligonucleotide probe (Promega Corp., Madison, WI). After incubation for 10 min, the reactions were terminated by adding 6 μl of 15% ficoll solution containing indicator dyes. Equal amounts of the reactions were loaded on 5% nondenaturing polyacrylamide gels in 1 × Tris/Borate/EDTA and run at 200 volts. Specificity of the NF-κB bands was determined by competition with excess unlabeled probe and lack of competition with 100fold excess of a probe containing a one basepair mutation in the NF-κB binding sequence.

We treated MCF-7 breast carcinoma cells with TNF-α and measured ceramide levels and cell death concomitantly at several time points (Fig. 1,A). Ceramide levels did not change appreciably in the first 5 h (data not shown) but were significantly increased between 7 and 9 h and continued to increase with time, up to 400% by 20 h. This accumulation was not dependent on new protein synthesis since the addition of cycloheximide to the cells before treatment with TNF-α enhanced the accumulation of ceramide (Fig. 1,B). Although these are delayed and persistent changes in ceramide, it is becoming increasingly apparent that this is the pattern most closely related to the apoptotic responses. Although this raises the concern that ceramide accumulation is a consequence of death, studies with Bcl-2 (see below) negate this possibility. Cell death, as measured by the inability to exclude trypan blue, was not seen until 20 h, occurring several hours after the increase in ceramide levels, indicating that ceramide accumulation occurs long before loss of membrane integrity. To verify that cell death was occurring through induction of apoptosis, we assayed for cleavage of the 116-kD PARP polypeptide to a specific 85-kD apoptotic fragment (9, 10). This proteolytic cleavage has been shown to occur in apoptosis and to be mediated by Yama/CPP32/apopain or related proteases. As shown previously (6), treatment of MCF-7 cells with TNF-α resulted in specific cleavage of PARP to the 85-kD fragment (Fig. 2). Significant PARP cleavage did not occur until 12 h after treatment with TNF-α and was maximal by 25 h. These results indicate that ceramide accumulation precedes one of the early signs of apoptosis, i.e., PARP cleavage, by at least 3–4 h. Moreover, loss of cell membrane integrity, as determined by trypan blue uptake, is unlikely to contribute to ceramide accumulation since it occurs several hours after the dramatic increase in endogenous ceramide levels.

The ability of CrmA to inhibit, with variable efficiency, the ICE family of cysteine proteases (6, 7, 22, 26) was used to examine the relationship of these proteases to ceramide along the death pathway. MCF-7 cells stably expressing CrmA were treated with 2 nM TNF-α or increasing concentrations of C₆-ceramide, a cell-permeable ceramide analogue, and compared with control (vector) cells. As shown previously (4), CrmA offered almost complete protection from TNF-α–induced cytotoxicity (Fig. 3,A). However, CrmA offered no protection from the cytotoxic effects of ceramide so that cells died equally in the presence or absence of CrmA (Fig. 3 A). Comparison of another pair of vector and CrmA-expressing MCF-7 clones described previously (4) yielded similar results (data not shown). Therefore, CrmA appeared not to interfere with the downstream effects of ceramide.

Since endogenous ceramide elevation may drive the cell to die, we next examined whether CrmA interferes with ceramide generation. We measured endogenous ceramide levels in response to TNF-α in control and in CrmAexpressing MCF-7 cells. Ceramide levels increased dramatically in vector cells at 18 and 24 h (Fig. 3,B). However, CrmA-expressing cells treated similarly with TNF-α showed no increase in ceramide levels. To determine whether the inhibition of ceramide accumulation was dependent on the ability of CrmA to inhibit proteases, we used two approaches. First, we used a point-mutant of CrmA that has no antiproteolytic activity due to the substitution of Arg for Thr at amino acid 291 in the reactive site loop (6). In MCF-7 cells expressing this mutant, there was no significant inhibition of ceramide accumulation (Fig. 3,C) and no protection from TNF-α–induced apoptosis (6). Second, we used the synthetic peptide Ac-YVAD-CHO that, like CrmA, is a potent competitive inhibitor of ICE (7). Pretreatment of cells with this peptide, resulted in significant inhibition of ceramide accumulation after TNF-α treatment (Fig. 3 C). In contrast, other protease inhibitors not known to inhibit the ICE family of proteases, including N-tosyl-l-phenylalanine chloromethyl ketone and leupeptin, were unable to inhibit ceramide accumulation after TNF-α treatment (data not shown). Therefore, CrmA inhibits TNF-α–induced apoptosis by acting at a point upstream of the generation of ceramide, and this inhibition is dependent on its antiproteolytic activity.

As ceramide generation occurred in cells destined to die due to treatment with TNF-α and not in the protected, CrmA-expressing cells, it became important to verify that this increase in ceramide levels, as well as the suppression of this increase by CrmA, were specific events correlating with cell death or survival, respectively. We used Bcl-2, another antiapoptotic molecule that has recently been shown to protect from ceramide-induced apoptosis (48, 49). MCF-7 cells expressing Bcl-2 or vector controls were treated with TNF-α or increasing concentrations of C₆-ceramide. Control cells died in response to both treatments, but Bcl-2–expressing cells displayed resistance to TNFα–induced cell death as seen in cells expressing CrmA (Fig. 4,A). However, unlike CrmA-expressing cells, Bcl-2–expressing cells were resistant to ceramide-induced cell death. Additionally, generation of endogenous ceramide was nearly equal in both vector and Bcl-2 cells in response to TNF-α, indicating that Bcl-2 does not interfere with ceramide generation (Fig. 4 B). Therefore, Bcl-2 functions at a point downstream of ceramide to inhibit apoptosis. More importantly, this delayed accumulation of ceramide is not a consequence of cell death since it is still observed in the viable Bcl-2–expressing cells. Thus, these data demonstrate that the elevation in ceramide is proximal to the biochemical and morphological changes of cell death.

To determine the relationship between TNF-α, ceramide, and subsequent PARP cleavage and apoptosis, it became imperative to study the effects and kinetics of ceramide on PARP cleavage. Consistent with results in Molt 4 cells (50), treatment of MCF-7 cells with ceramide resulted in cleavage of PARP (Fig. 5,A), and PARP cleavage after 4 h of ceramide treatment was equivalent to that caused by 12–16 h of TNF-α treatment (compare with Fig. 2). By using ¹⁴C-labeled ceramide and evaluating the kinetics of its uptake by MCF-7 cells at several time points, we found that ceramide was taken up slowly by these cells with maximal uptake reached at only 4 h (Fig. 5 B). Therefore, the delay in PARP cleavage after ceramide treatment can be attributed to the delay in uptake of exogenous ceramide. Hence, these experiments support the hypothesis that ceramide accumulation following TNF-α treatment may represent a trigger for PARP cleavage and apoptosis.

Next, it became important to determine the effect of CrmA or Bcl-2 expression on ceramide-induced PARP cleavage in MCF-7 cells. Although both Bcl-2 and CrmA inhibited TNF-α–induced PARP cleavage, only Bcl-2 expression provided protection from ceramide-induced cleavage. PARP was cleaved after exogenous ceramide treatment despite high levels of expression of CrmA, indicating that ceramide bypasses the CrmA target and activates a downstream protease capable of cleaving PARP (i.e., prICE) (Fig. 5 C). Densitometric quantitation of PARP cleavage showed 17.7 and 20.4% cleavage in the ceramide– or TNF-α–treated vector cells, respectively, whereas CrmAexpressing cells treated with ceramide or TNF-α had 16.6 and 0% cleavage, respectively. Therefore, in MCF-7 cells, CrmA expression does not significantly interfere with activation of PARP cleavage by ceramide.

To further dissociate the site of action of CrmA from that of Bcl-2, we employed the alkylating agent mitomycin C. Mitomycin C did not cause any significant change in endogenous ceramide levels (data not shown). We next compared the effects of TNF-α, C₆-ceramide, and mitomycin C on the survival of CrmA– or Bcl-2–expressing MCF-7 cells. Whereas the CrmA-expressing cells were protected from TNF-α–induced apoptosis, they were equally susceptible to mitomycin C (Fig. 6,A) as were vector cells. However, cells expressing Bcl-2 were protected from the cytotoxic effects of mitomycin C as well as TNF-α and ceramide. Biochemically, expression of CrmA prevented PARP cleavage after activation of the ceramide-dependent TNF-α pathway (Fig. 6 B). However, PARP cleavage was almost complete, despite expression of CrmA, after treatment with mitomycin C that activates a ceramide-independent apoptotic pathway. In contrast, Bcl-2 overexpression prevented PARP cleavage induced by both the TNF-α (ceramidedependent) and mitomycin C (ceramide-independent) pathways. These results provide additional evidence that CrmA functions distinctly from Bcl-2. Specifically, CrmA appears to inhibit an event proximal to ceramide accumulation in ceramide-dependent pathways whereas Bcl-2 works by inhibiting a more distal target that is common to both apoptotic pathways.

Activation of the transcription factor NF-κB is one of the major functions of TNF-α. For activation to occur, NF-κB has to dissociate from its cytoplasmic inhibitory protein I-κB so that it can translocate to the nucleus (51). This is accomplished through site-specific serine phosphorylation followed by proteolytic degradation of I-κB (52, 53). The mediators of this activation have not been clarified, although the “death domain”-containing protein TRADD, but not FADD, has been shown to be important for its activation (35, 36). Recently, ceramide generated through the action of an acidic sphingomyelinase was implicated in the activation of NF-κB (54). The antiproteolytic activity of CrmA, as well as its ability to inhibit ceramide accumulation after TNF-α treatment led us to examine the effect of CrmA expression on NF-κB activation. We used electrophoretic mobility shift assays which rely on the ability of activated NF-κB to bind to its specific DNA sequence resulting in retardation of the protein– DNA complex on nondenaturing polyacrylamide gels. Treatment of MCF-7 cells expressing CrmA, Bcl-2, or their corresponding vector with 2 nM TNF-α resulted in equal activation of NF-κB (Fig. 7, A and B). Therefore, CrmA does not interfere with the signaling pathway leading to activation of NF-κB, despite its demonstrated ability to completely inhibit ceramide accumulation. This suggests that a physiologic role for ceramide in NF-κB activation is unlikely.

The results from this study show that ceramide accumulation observed after treatment of MCF-7 cells with TNF-α is completely inhibited by CrmA (Fig. 3,B). Exogenous ceramide can bypass this block and produce apoptosis by activation of downstream proteases (Fig. 3 A and 5 A). Similar results were obtained with BJAB cells after ligation of the Fas receptor (data not shown). These data suggest that CrmA targets a protease acting upstream of ceramide generation. Since CrmA did not inhibit activation of prICE by ceramide or mitomycin C, these results raise the possibility that CrmA may target proteases that act upstream of the PARP protease.

The antiapoptotic activity of CrmA has been thought to be due to its ability to inhibit members of the ICE family of cysteine proteases that are homologues of the C. elegans cell death protease CED-3 (2, 4, 25). However, the potency by which CrmA inhibits the various members of this family varies significantly. Whereas picomolar concentrations of CrmA can potently inhibit ICE, concentrations that are over 1,000-fold higher are required to achieve comparable inhibition of Yama/CPP32/apopain, a putative PARP cleaving protease (7). Similarly, CrmA is a poor inhibitor of ICH-1 (23) and CED-3 (3, 55), indicating that it preferentially targets only a subset of ICE family members. This may explain why CrmA is not a universal inhibitor of apoptosis. Based on our findings, the more physiologically relevant target of CrmA appears to be a protease that is activated by TNF-α and is involved in the activation phase, rather than the execution phase, of apoptosis. The recently described Fas– and TNF-α receptor–associated ICE protease FLICE/MACH, is a likely candidate since it was demonstrated that apoptosis induced by its overexpression can be inhibited by CrmA (16). Other apoptotic stimuli such as mitomycin or sodium azide (data not shown) that use a ceramide-independent pathway to activate the death proteases are not inhibited by CrmA (Fig. 8).

Our findings define a novel site of CrmA-induced protection from apoptosis that is clearly distinct from that of Bcl-2. Whereas CrmA interferes with ceramide accumulation, Bcl-2 does not modulate this event and appears to work downstream by inhibiting the execution machinery of apoptosis (Fig. 8). This is additionally supported by the ability of both CrmA and Bcl-2 to protect from cell death when induced with TNF-α while only Bcl-2 is capable of protecting cells treated with C₆-ceramide or mitomycin C. These findings implicate Bcl-2 as a more generalized inhibitor of apoptosis acting on a target common to different apoptotic pathways whereas CrmA may specifically inhibit pathways that lead to a sustained ceramide signal such as that stimulated by TNF-α or Fas.

These studies also define distinct pathways that lead to NF-κB activation and apoptosis. The activation of the ceramide pathway has been implicated by some groups to be a necessary step for activation of NF-κB (54, 56, 57). Others have been unable to reproduce these results (47, 58–61). The ability of CrmA to inhibit the generation of ceramide provided a new tool to help resolve these contradictions. Our finding that CrmA does not inhibit NF-κB activation by TNF-α despite a complete shutdown of ceramide generation makes a role for ceramide in NF-κB activation highly questionable. Rather, ceramide accumulation appears to be more intimately related to apoptosis and growth suppression. These results are consistent with published studies showing that CrmA does not interfere with NF-κB activation stimulated by TRADD overexpression, but blocks TRADD-driven apoptosis (36), and that the dominant negative FADD mutant can potently suppress FADD-dependent apoptosis without interfering with NF-κB activation (38). Additionally, in cells undergoing Fas-induced apoptosis which is accompanied by massive accumulation of ceramide, NF-κB activation is strikingly absent (Gamard, C., G. Dbaibo, L. Obeid, and Y. Hannun, manuscript submitted). These studies clearly support the previous findings that ceramide is neither necessary nor sufficient for NF-κB activation.

Taken together, these findings are consistent with an important role for ceramide in TNF-α–induced apoptosis. The sustained elevations in endogenous ceramide levels observed over several hours that are inhibited by CrmA are more closely related to apoptosis than the transient, but more modest, increases in ceramide observed within minutes in other cell lines and implicated in mediating early signaling events (27). Indeed, these sustained levels of ceramide are more compatible with cellular levels achieved with 1–20 μM of exogenous ceramide. Accumulation of ceramide with delayed kinetics has been implicated as a stress response that drives the cell to either undergo apoptosis (29, 34) or cell cycle arrest (29, 62). Indeed, these delayed kinetics of ceramide accumulation are more commensurate with the observed delay in the onset of apoptotic morphology and subsequent cell death after TNF-α. Our findings that this accumulation occurs before PARP cleavage and the ability of Bcl-2 to prevent ceramide-induced death support this emerging role for ceramide. The results obtained using MCF-7 cells suggest that, in the apoptotic response of these cells, ceramide functions much more as a downstream sensor and integrator rather than as an upstream acute signaling switch.

In conclusion, ceramide accumulation is emerging as an important component in a TNF-α–induceable pathway of apoptosis. This pathway can be inhibited by CrmA that blocks ceramide generation or Bcl-2 that interferes with the function of effector molecules downstream of ceramide (Fig. 8). Although both Bcl-2 and CrmA appear to inhibit activation of proteases, these results show that they interfere with the death pathway at two different points. These results will need to be evaluated in other cell lines and further exploration of this pathway will be essential for the understanding of this complex aspect of cell regulation and may identify more specific targets for inhibition of apoptosis.

We thank Muneesh Tewari and Dr. Vishva M. Dixit for providing us with different MCF-7 cell lines and for helpful discussions and advice, Alicja Bielawska for C₆-ceramide and ¹⁴C₆-ceramide synthesis, Milton Campbell and William E. Kraus for help with densitometry, and Marsha Mangum for expert secretarial assistance.