Pivotal Role of the Cyclin-dependent Kinase Inhibitor p21WAF1/CIP1 in Apoptosis and Autophagy*

Programmed cell death (PCD) is involved in a variety of biologic events. Based on the morphologic appearance of the cells, there are two types of PCD as follows: apoptotic (type I) and autophagic (type II). However, the molecular machinery that determines the type of PCD is poorly defined. The purpose of this study was to show whether the presence of the cyclin-dependent kinase (CDK) inhibitor p21WAF1/CIP1, a modulator of apoptosis, determines which type of PCD the cell undergoes. Treatment with C2-ceramide was associated with both the cleavage of caspase-3 and poly(ADP-ribose) polymerase and the degradation of autophagy-related Beclin 1 and Atg5 proteins, without a change in the cyclin-CDK activity, which culminated in apoptosis in p21+/+ mouse embryonic fibroblasts (MEFs). On the other hand, C2-ceramide did not cleave caspase-3 or poly(ADP-ribose) polymerase and kept Beclin 1 and Atg5 proteins stable in p21-/- MEFs, events that this time culminated in autophagy. When expression of the p21 protein was inhibited by small interfering RNA or when the overexpression of Beclin 1 or Atg5 was induced, autophagy rather than apoptosis was initiated in the p21+/+ MEFs treated with C2-ceramide. In contrast, the exogenous expression of p21 or the silencing of Beclin 1 and Atg5 with small interfering RNA increased the number of apoptotic cells and decreased the number of autophagic cells among C2-ceramide-treated p21-/- MEFs. γ-Irradiation, which endogenously generates ceramide, induced a similar tendency in these MEFs. These results suggest that p21 plays an essential role in determining the type of cell death, positively for apoptosis and negatively for autophagy.

Programmed cell death (PCD) 3 is a major terminal pathway for normal development, but various diseases such as cancer can eventuate when PCD becomes dysregulated (1,2). Beyond this, the precise details of the molecular events that occur in the various PCD pathways are not completely known. Currently, morphologic and biochemical criteria are the primary means used to distinguish between the types of PCD and their molecular signaling pathways. For example, apoptosis exhibits a particular morphologic appearance, including chromatin condensation and DNA ladder formation. Because of its involvement in disease processes, especially tumorigenesis, apoptosis has attracted increasing scientific attention in the last decade, with the result that extensive progress has now been made in elucidating the molecular mechanisms at work in apoptotic pathways. For example, we now know that initiator caspases, such as caspase-8, and executioner caspases, including caspase-3 and caspase-6, play a central role in the transduction of the apoptotic signal and execution of apoptosis in mammalian cells (3). In contrast, autophagy is associated with the presence of autophagosomes and autolysosomes (2) and is generally a caspase-independent process (4). The recent discovery of the expression of autophagy-related (Atg) genes under conditions of nutrient depletion or in the absence of growth factors has represented a key piece of information about the autophagic process (5,6). Like apoptosis, autophagy is also detected under both physiologic and pathologic conditions, with the latter including neurodegenerative diseases and cancers (7)(8)(9). There is now increasing evidence of cross-talk between the two PCD pathways (10 -12). However, there are still many unanswered questions about the nature of this cross-talk (13), including the identity of the molecules that regulate the cell fate.
One molecule of potential interest is p21 WAF1/CIP1 , which was originally identified as a universal inhibitor of cyclin-dependent kinases (CDKs) that belongs to the CIP/KIP family of CDK inhibitors (14 -17). Recently, accumulating evidence has indicated that p21 has functions in addition to CDK inhibition (18,19). In particular, p21 positively or negatively regulates the apoptotic signaling pathway that ultimately determines cell death (20). However, the association between p21 and autophagy is not understood. Therefore, in this study, we assessed the role of p21 in inducing apoptosis and autophagy using p21 ϩ/ϩ and p21 Ϫ/Ϫ mouse embryonic fibroblasts (MEFs). C 2 -ceramide, a cell-permeable ceramide analogue that induces apoptosis or autophagy in some types of cells (21)(22)(23)(24)(25), was used to induce cell death.
Cell Viability Assay-The effect of C 2 -ceramide on the cell viability of MEFs was determined by using a WST-1 reagent (Roche Applied Science), as described previously (26). MEFs (2 ϫ 10 4 cells/well) were seeded in 96-well flat-bottomed plates. After 24 h, cells were fed with culture medium containing 0.1% fetal bovine serum and treated with C 2 -ceramide at concentrations ranging from 0 to 100 M for 24 h, as described previously (23). The cells were then exposed to 10 l of WST-1 reagent for 1 h at 37°C. Absorbance at 450 nm was measured in a microplate reader. The viability of cells treated with vehicle alone was considered to be 100%.
Cell Cycle Analysis-For the cell cycle analysis, MEFs treated with or without C 2 -ceramide as described above were trypsinized, fixed with 70% ethanol, and stained with propidium iodide by using the cellular DNA flow cytometric analysis reagent set (Roche Applied Science), as described previously (23). Samples were analyzed for DNA content with the FACScan using CellQuest software (BD Biosciences).
Detection of Apoptosis-Nuclei were stained with Hoechst 33258 to detect the chromatin condensation or nuclear fragmentation characteristic of apoptosis, as described previously (23). MEFs treated with or without C 2 -ceramide, as described above, were fixed with 4% paraformaldehyde and stained with Hoechst 33258 (1 g/ml) for 15 min. Cells were counted and scored for the incidence of apoptotic chromatin changes under a fluorescence microscope.
Electron Microscopy-MEFs, grown on gelatinized plastic coverslips, were treated with or without C 2 -ceramide, as described above, fixed for 2 h with 2.5% glutaraldehyde (EM Science, Hatfield, PA) in 0.1 M cacodylate buffer, pH 7.4, postfixed in 1% OsO 4 in the same buffer, and then subjected to electron microscopic analysis as described previously (23). Representative areas were chosen for ultra-thin sectioning and viewed with a Hitachi 7600 electron microscope.
Quantification of Acidic Vesicular Organelles (AVOs) with Acridine Orange Staining-Autophagy is the process of sequestrating cytoplasmic proteins into the lytic component and is characterized by the development of AVOs. To detect and quantify AVOs in C 2 -ceramide-treated MEFs or MEFs at 2 h after amino acid deprivation with Earle's balanced salt solution (Invitrogen) as described previously (27), we performed the vital staining with acridine orange, followed by FACScan analysis, as described previously (23).
Assessment of the Involvement of Microtubule-associated Prote1 Light Chain 3 (LC3)-LC3, a mammalian homologue of yeast Atg8, is recruited to the autophagosome membrane during autophagy (28), which makes LC3 a marker for autophagy. In an earlier study, cells transfected with the expression vector of the green fluorescent protein-linked LC3 (GFP-LC3) showed a diffuse distribution of GFP-LC3 under control conditions, whereas a punctate pattern of GFP-LC3 expression (GFP-LC3 dots) was induced (29). Therefore, we transiently transfected MEFs with the GFP-LC3 expression vector using the FuGENE 6 transfection reagent (Roche Applied Science) for 24 h, treated them with or without C 2 -ceramide, fixed the cells with 4% paraformaldehyde, and determined the proportion of cells expressing GFP-LC3 dots (Ն10 dots/cell) under a fluorescence microscope, as described previously (23). The GFP-LC3 vector was kindly supplied by Dr. Noboru Mizushima (Tokyo Medical and Dental University, Tokyo, Japan).
Reverse Transcription (RT)-PCR-Two micrograms of total RNA was isolated from cells using the RNeasy mini kit (Qiagen) and subjected to reverse transcription using the Thermo-Script TM RT-PCR system (Invitrogen) according to the manufacturer's instructions. Random hexamer was used in the RT reactions. The primer kit for Beclin 1 was purchased from SuperArray (Frederick, MD). The primers for GAPDH were 5Ј-CCTGGAGAAACCTGCCAAGTAT-3Ј and 5Ј-AGAGTGG-GAGTTGCTGTTGAAG-3Ј (Sigma). The amplification program for Beclin 1 was 30 s at 95°C, 30 s at 55°C, and 30 s at 72°C for 40 cycles after denaturing at 95°C for 4 min, and for GAPDH it was 30 s at 95°C, 30 s at 55°C and 30 s at 72°C for 30 cycles after denaturing at 95°C for 4 min, as described previously (30). These cycles were followed by a 10-min elongation step at 72°C. The PCR products were analyzed by 2% agarose gel electrophoresis. The level of Beclin 1 mRNA (normalized to GAPDH) was quantified by using Scion imaging software (Scion Corp., Fredrick, MD) and averaged over three experiments.
Immunoprecipitation-MEFs were treated with or without 25 M C 2 -ceramide for 1 h. Proteins were extracted using lysis buffer (10 mM Tris-HCl, pH 7.8, 1% Nonidet P-40, 150 mM NaCl, 1 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 0.035 mg/ml aprotinin). Equal amounts of protein (200 g) were immunoprecipitated with 1 g of anti-Beclin 1 antibody or anti-p21 antibody at 4°C overnight, and the immunoprecipitated protein was pulled down with protein A-agarose beads (Roche Applied Science) and separated by 15% SDS-PAGE. Membranes were probed with anti-p21 antibody (Santa Cruz Biotechnology) or anti-Beclin 1 antibody. The same samples were also used for immunoblots against Beclin 1 or p21. For a loading control, the membranes were reprobed with ␤-actin antibody (Sigma). To determine the efficacy of immunoprecipitation, after immunoprecipitation with anti-Beclin 1 antibody, we used the supernatant for Western blotting with the same antibody. Because Beclin 1 was undetectable, we speculated that immunoprecipitation with anti-Beclin 1 antibody worked very well.
Pulse-Chase Labeling-MEFs were cultured to 70% confluence. For each chase time point, 5 ϫ 10 6 cells were washed once in ice-cold phosphate-buffered saline and incubated in DMEM (without cysteine/methionine; Sigma) for 1 h at 37°C with 5% CO 2 . Cells were then pulsed for 30 min by using [ 35 S]methionine Promix (10 Ci/10 6 cells; Amersham Biosciences). Cells were then chased for the indicated time points in DMEM in the presence or absence of 25 M C 2 -ceramide. Total cell lysates were then prepared and immunoprecipitated with 1 g of anti-Beclin-1 antibody (Santa Cruz Biotechnology) and protein A-Sepharose beads (Roche Applied Science) for overnight at 4°C. Immunocomplexes were separated by 15% SDS-PAGE. This gel was dried and developed by autoradiography.
Cyclin-CDK Activity-MEFs treated with or without C 2 -ceramide, as described above, were lysed. Equal amounts of proteins (200 g) were immunoprecipitated with 1 g of anti-p21 antibody at 4°C overnight, and the immunoprecipitated protein was then pulled down with protein A-agarose beads (Roche Applied Science) and separated by 15% SDS-PAGE. Membranes were probed with anti-CDK2 and -CDK4 antibodies.
Plasmid Vectors-The p21 expression vector (14) was kindly provided by Dr. Bert Vogelstein (The Johns Hopkins Oncology Center, Baltimore). The Beclin 1 expression vector (31) was kindly provided by Dr. Beth Levine (University of Texas Southwestern Medical Center, Dallas). The Atg5 expression vector was used as described previously (32). As a control, a ␤-galactosidase expression vector (Invitrogen) was used. The FuGENE 6 transfection reagent was used to transfect cells with these plasmids 24 or 48 h before C 2 -ceramide treatment.
Ionizing Radiation-Radiation treatment was administered using a cesium 137 irradiator (model E-0103; U. S. Nuclear Corp., Burbank, CA) at a dose rate of 3.216 Gy/min, as described previously (33).
Statistical Analysis-All the experiments were repeated at least three times. Values are expressed as means Ϯ S.D. Statistical analysis was performed by using Student's t test (twotailed). The criterion for statistical significance was p Ͻ 0.05.

RESULTS
Inhibitory Effect of C 2 -ceramide on the Cell Viability of p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs-p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs were treated with C 2 -ceramide at concentrations of 0, 5, 10, 25, 50, and 100 M in DMEM with 0.1% fetal bovine serum for 24 h, and then cell viability was determined. Treatment with C 2 -ceramide reduced the viability of p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs in a dose-dependent manner (Fig. 1A). We found that p21 Ϫ/Ϫ MEFs were more resistant to C 2 -ceramide at concentrations of 10, 25, and 50 M than were p21 ϩ/ϩ MEFs (p Ͻ 0.02). Because treatment with 25 M C 2 -ceramide for 24 h reduced the viability of both types of MEF to 40 -60% of the control cells treated with vehicle alone for 24 h, we used this concentration of C 2 -ceramide for the following experiments.
Induction of Apoptosis by C 2 -ceramide in p21 ϩ/ϩ MEFs but Not in p21 Ϫ/Ϫ MEFs-To determine the type of cell death caused by C 2 -ceramide, we first performed apoptosis detection assays using cell cycle analysis and Hoechst 33258 DNA staining. The DNA flow cytometric analysis showed that treatment with C 2 -ceramide for 24 h increased the proportion of the p21 ϩ/ϩ MEFs in the sub-G 1 phase from 5.3-23.3%, which is characteristic of apoptosis; in contrast, there was no significant increase in the percentage of p21 Ϫ/Ϫ MEFs (Fig. 1B). Nuclei of MEFs treated with or without C 2 -ceramide were then stained with Hoechst 33258 to determine the percentage of cells with apoptotic morphology, such as chromatin condensation or nuclear fragmentation. The percentages of p21 ϩ/ϩ MEFs showing apoptosis significantly increased 24 h after exposure to C 2 -ceramide (p Ͻ 0.0005), whereas there was no significant increase in the percentages of p21 Ϫ/Ϫ MEFs showing apoptosis in response to C 2 -ceramide (Fig. 1C). Both caspase-3 and PARP were cleaved in p21 ϩ/ϩ MEFs 6 and 24 h after C 2 -ceramide treatment, but these cleavages were undetectable in treated p21 Ϫ/Ϫ MEFs (Fig. 1D). Interestingly, C 2 -ceramide inhibited the expression of PARP to an undetectable level in p21 Ϫ/Ϫ MEFs 6 h after the treatment but did not affect the expression level of caspase-3. The viability of p21 ϩ/ϩ MEFs inhibited by C 2 -ceramide treatment was restored in a dose-dependent manner by Z-VAD-FMK, a caspase inhibitor with broad specificity (Fig. 1E), confirming the role of caspases in C 2 -ceramide-induced apoptotic cell death in p21 ϩ/ϩ MEFs. Protection of Z-VAD-FMK was incomplete in cells treated with 25 M C 2 -ceramide, suggesting that Z-VAD-FMK leads to death by autoph-agy under these conditions. However, Z-VAD-FMK did not induce autophagy in p21 ϩ/ϩ MEFs (Fig. 1F), which therefore excluded this possibility. These results indicate that C 2 -ceramide cleaves caspase-3 and then PARP, thereby inducing apoptosis in p21 ϩ/ϩ MEFs, whereas no significant apoptosis was detectable in p21 Ϫ/Ϫ MEFs treated with C 2 -ceramide.
Induction of Autophagy by C 2 -ceramide in p21 Ϫ/Ϫ MEFs but Not in p21 ϩ/ϩ MEFs-Recent investigations have demonstrated that C 2 -ceramide triggers nonapoptotic cell death with autophagic features in some types of cells (22)(23)(24)(25). Therefore, we analyzed the ultrastructure of p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs treated with C 2 -ceramide with electron microscopy. In p21 ϩ/ϩ MEFs, treatment with 25 M C 2 -ceramide for 24 h caused nuclear fragmentation, which is a feature of apoptosis ( Fig. 2A). On the other hand, in p21 Ϫ/Ϫ MEFs, some autophagic vacuoles containing cytoplasmic contents were observed under control conditions, but remarkably, C 2 -ceramide increased the number of autophagic vacuoles in the cytoplasma in cells with an intact nucleus, whereas early autophagosomes that were filled with normal cytoplasm were not observed. Microscopically, a substantial number of p21 ϩ/ϩ MEFs became smaller 24 h after C 2 -ceramide treatment, whereas the majority of p21 Ϫ/Ϫ MEFs were larger and had many vacuoles even 6 h after exposure to C 2 -ceramide (Fig. 2B). Because autophagic cells undergo progressive atrophy following growth factor withdrawal (12), it is possible that p21 Ϫ/Ϫ MEFs treated with C 2 -ceramide do not live long enough to become small.
To assess the development of AVOs in C 2 -ceramide-treated MEFs, we performed vital staining with acridine orange and quantified the results with flow cytometry. Treatment with 25 M C 2 -ceramide for 24 h increased the proportion of p21 Ϫ/Ϫ MEFs showing prominent red fluorescence from 4.4 to 44.3%, whereas the treatment did not significantly increase the development of AVOs in p21 ϩ/ϩ MEFs (from 5.1 to 3.2%) (Fig. 2C). Moreover, when we investigated the proportion of cells exhibiting GFP-LC3 dots in response to C 2 -ceramide, indicating autophagy, p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs treated without C 2 -ceramide showed diffuse distribution of GFP-LC3, as shown in Fig.  2D. After a 24-h treatment with C 2 -ceramide, p21 Ϫ/Ϫ MEFs but not p21 ϩ/ϩ MEFs exhibited GFP-LC3 dots. The percentage of autophagic p21 Ϫ/Ϫ MEFs with GFP-LC3 dots was significantly increased from 3.8 to 18.4% by C 2 -ceramide (p Ͻ 0.001), whereas there was no significant increase in the percentage of p21 ϩ/ϩ MEFs with GFP-LC3 dots (from 2.2 to 2.4%) (Fig. 2E).
Interestingly, however, p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs showed a similar potential to stimulate autophagy in response to the classic autophagic inducer, amino acid deprivation (Fig. 2F). These results suggest that C 2 -ceramide induces apoptosis or autophagy in MEFs, depending on the presence or absence of p21, respectively.
Effect of Silencing p21 on C 2 -ceramide-induced Apoptosis in p21 ϩ/ϩ MEFs-On the basis of the above data, we speculated that p21 plays a key role in inducing apoptosis instead of autophagy in MEFs after exposure to C 2 -ceramide. To test our hypothesis, we first determined whether the inhibition of p21 by siRNA affects the induction of apoptosis in C 2 -ceramide-treated p21 ϩ/ϩ MEFs. Expression of p21 protein was remarkably suppressed by transfection with 100 nM siRNA directed against p21 for 72 h, but not by transfection with 100 nM nontargeting (control) siRNA (Fig.  3A). After siRNA transfection, p21 ϩ/ϩ MEFs were exposed to 25 M C 2 -ceramide for an additional 24 h. As shown in Fig. 3, B and C, compared with control siRNA transfection, p21 knockdown significantly decreased the percentage of apoptotic cells (p Ͻ 0.02) and increased the percentage of autophagic cells expressing GFP-LC3 dots (p Ͻ 0.02) after a 24-h treatment with 25 M C 2 -ceramide. These results indicate that inhibition of the p21 protein suppresses the ability of p21 ϩ/ϩ MEFs to undergo apoptosis and increases their sensitivity to autophagy caused by C 2 -ceramide.
Effect of Exogenous Expression of p21 on C 2 -ceramide-induced Autophagy in p21 Ϫ/Ϫ MEFs-As a next step in testing our hypothesis about the role of p21 in PCD, we attempted to induce p21 protein expression in p21 Ϫ/Ϫ MEFs. Immunoblotting revealed the exogenous expression of p21 in p21 Ϫ/Ϫ MEFs transfected with the p21 expression vector, but not with the ␤-galactosidase construct (Fig. 4A). To determine whether p21 expression affects the induction of apoptosis or autophagy in p21 Ϫ/Ϫ MEFs treated with C 2 -ceramide, the cells were co-transfected with GFP-LC3 and p21 or ␤-galactosidase expression vectors, followed by exposure to 25 M C 2 -ceramide for 24 h and Hoechst 33258 staining. As shown in Fig. 4B, apoptotic cells with nuclear condensation but not GFP-LC3 dots were detected in p21-transfected cells, whereas autophagic cells exhibiting GFP-LC3 dots but not an apoptotic morphology were observed in ␤-galactosidase-transfected cells. The exogenous expression of p21 significantly increased the percentage of apoptotic cells (p Ͻ 0.004) (Fig. 4C) and decreased the extent of autophagy in p21 Ϫ/Ϫ MEFs treated with C 2 -cera- mide (p Ͻ 0.007) (Fig. 4D). These results indicate that introduction of the p21 protein conferred on p21 Ϫ/Ϫ MEFs the ability to undergo apoptosis and suppressed their potential to undergo autophagy.
Involvement of Beclin 1 and Atg5 in C 2 -ceramide-induced Cell Death-To determine whether the autophagy-related proteins Beclin 1 and Atg5 are involved in C 2 -ceramide-induced PCD, we examined their expressions in p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs after exposure to C 2 -ceramide. In p21 Ϫ/Ϫ MEFs, the expression of Beclin 1 and Atg5 proteins remained unchanged before and 3 h after treatment with C 2 -ceramide (Fig. 5A). In contrast, the expression of Beclin 1 and Atg5 proteins was suppressed to very low levels in p21 ϩ/ϩ MEFs 3 h after exposure to C 2 -ceramide. In addition, the pulse-chase assay showed that the level of Beclin 1 was less in p21 ϩ/ϩ MEFs treated with C 2 -ceramide for 1 h than in cells not treated with C 2 -ceramide, indicating that the stability of Beclin 1 was affected (Fig. 5A). Furthermore, we determined whether the effects of p21 on Beclin 1 levels were because of its effect on transcription. As shown in Fig. 5A, C 2 -ceramide did not affect the expression level of Beclin 1 mRNA in p21 ϩ/ϩ and p21 Ϫ/Ϫ MEFs (quantification showed no significant change).
To determine the interaction between Beclin 1 and p21, we examined their binding in p21 ϩ/ϩ MEFs using immunoprecipitation. As shown in Fig. 5B, in p21 ϩ/ϩ MEFs, the expression of Beclin 1 protein was suppressed even 1 h after exposure to 25 M C 2 -ceramide. At that time point, p21 expression remained stable. Immunoprecipitation with anti-Beclin 1 antibody followed by immunoblotting with anti-p21 antibody revealed that Beclin 1 was not directly associated with p21, when Beclin 1 was gradually degraded by the C 2 -ceramide. This finding was also supported by another immunoprecipitation with anti-p21 antibody and then Western blot with anti-Beclin 1 antibody (Fig.  5B). Furthermore, we determined whether change in cyclin-CDK activity was required. As shown in Fig. 5B, in p21 ϩ/ϩ MEFs after treatment with or without 25 M C 2 -ceramide for 3 h, the expressions of CDK2 and CDK4 proteins immunoprecipitated with anti-p21 antibody were not changed. These results suggest that the induction of apoptosis in C 2 -ceramidetreated p21 ϩ/ϩ MEFs is because of inhibition of the autophagic pathway through the destabilization of autophagy-related proteins but not because of a change in the cyclin-CDK activity.
We next knocked down the expression of Beclin 1 or Atg5 in p21 Ϫ/Ϫ MEFs using siRNA (Fig. 5C) to determine whether inhibition of their expressions affected C 2 -ceramide-induced autophagy in p21 Ϫ/Ϫ MEFs. After siRNA transfection, we performed Hoechst 33258 staining and the GFP-LC3 dot assay. Compared with control siRNA transfection, Beclin 1 or Atg5 knockdown significantly increased the percentage of cells showing apoptosis (p Ͻ 0.002) (Fig. 5D) and suppressed the induction of autophagy in p21 Ϫ/Ϫ MEFs treated with C 2 -ceramide (p Ͻ 0.002) (Fig. 5E). In contrast, the expression of Beclin 1 and Atg5 proteins was down-regulated in p21 ϩ/ϩ MEFs during C 2 -ceramide treatment. Therefore, we concluded that the inhibition of Beclin 1 or Atg5 protein in p21 ϩ/ϩ MEFs might increase the frequency of apoptosis more than might control siRNA.
Moreover, we reintroduced Beclin 1 or Atg5 expression in p21 ϩ/ϩ MEFs to determine whether the restoration of these proteins affected the induction of apoptosis or autophagy triggered by C 2 -ceramide. Overexpression of Beclin 1 or Atg5 significantly reduced the percentage of apoptotic cells (p Ͻ 0.05) and increased the percentage of autophagic cells among the C 2 -ceramide-treated p21 ϩ/ϩ MEFs (p Ͻ 0.005) (Fig. 6, A and  B). These collective results indicate that C 2 -ceramide-induced apoptosis requires a reduction in the expression of Beclin 1 and Atg5 proteins, whereas C 2 -ceramide-induced autophagy requires Beclin 1 or Atg5 protein expression.
It was also important to determine whether our findings are unique to C 2 -ceramide or apply to other apoptosis inducers that endogenously generate ceramide. Therefore, we treated p21 ϩ/ϩ MEFs and p21 Ϫ/Ϫ MEFs with ␥-irradiation, which has been shown to generate the second-messenger ceramide and mediate apoptosis (21). As shown in Fig. 7A, apoptosis was induced in p21 ϩ/ϩ MEFs by ␥-irradiation at 10 Gy more than in p21 Ϫ/Ϫ MEFs. On the other hand, ␥-irradiation caused autophagy in p21 Ϫ/Ϫ MEFs more than in p21 ϩ/ϩ MEFs (Fig. 7B). These results suggest that our findings do apply to other apoptosis inducers.

DISCUSSION
In this study, we showed that stimulation of the cell death signal by C 2 -ceramide degrades the autophagy-related proteins  Beclin 1 and Atg5, which subsequently induces caspasedependent apoptosis in p21 ϩ/ϩ MEFs without a change in the cyclin-CDK activity. On the other hand, in p21 Ϫ/Ϫ MEFs, C 2 -ceramide drastically disrupts the PARP protein while maintaining the expression levels of Beclin 1 and Atg5 proteins, which then triggers autophagy. Furthermore, the inhibition of p21 or induction of Beclin 1 or Atg5 suppresses the apoptotic pathway and turns on the switch that triggers autophagy in p21 ϩ/ϩ MEFs, whereas the introduction of p21 or inhibition of Beclin 1 or Atg5 increases the sensitivity of p21 Ϫ/Ϫ MEFs to apoptosis. These findings suggest that, upon the cell death stimuli, p21 triggers apoptosis by inhibiting the autophagic pathway through the suppression of the stability of autophagyrelated proteins but not by influencing the cyclin-CDK activity, at least in MEFs. Because the assembly of cyclin D-CDK4 complexes is impaired in p21 Ϫ/Ϫ MEFs compared with p21 ϩ/ϩ MEFs (35), the intact presence but not change in cyclin-CDK complexes might be required to direct a cell to an apoptotic rather than an autophagic fate.
p21 forms a complex with cyclin and CDK and inhibits their activity as broad acting CDK inhibitors (14 -17). p21 can also associate with the proliferating cell nuclear antigen to suppress proliferating cell nuclear antigen-dependent cell growth (36). Besides being an inhibitor of cell cycle progression, p21 acts as a mediator of the apoptotic pathway (20). For example, the enforced expression of p21 induces apoptosis or enhances the apoptotic response to chemotherapeutic agents (37)(38)(39)(40)(41). In other situations, however, p21 can function as an anti-apoptotic protein (42). For example, the suppression of p21 renders cells more sensitive to the apoptosis caused by DNA-damaging treatments, including radiation (43)(44)(45). p21 is also cleaved by caspase-3, making growth-arrested cells undergo apoptosis (46). However, the precise mechanisms underlying the different roles of p21 in apoptosis remain controversial.
In our study, C 2 -ceramide induced apoptosis in p21 ϩ/ϩ MEFs, which is associated with the cleavage of caspase-3 and PARP and degradation of Beclin 1 and Atg5. In p21 Ϫ/Ϫ MEFs, however, C 2 -ceramide induced autophagy without activating the caspase cascade. ␥-Irradiation, which generates ceramide, induces similar events in these MEFs. Therefore, as far as ceramide-induced PCD is concerned, p21 functions as an apoptosis inducer. In the greater scheme of things, ceramide has been recognized as an important second messenger molecule involved in the signaling pathways that control cell proliferation, differentiation, and death, especially apoptosis (47,48). Ceramide-mediated apoptosis, in particular, has been extensively studied (21,49,50), and p21 has been found to promote ceramide-induced apoptosis (51). At the same time, ceramide has also been implicated in nonapoptotic autophagic cell death (22)(23)(24)(25). For example, in MCF7 breast cancer cells, C 2 -ceramide stimulated the expression of Beclin 1 and induced autophagy (24). However, whether p21 is involved in ceramide-induced autophagy was not determined in these investigations. Autophagy is a process in which subcellular membranes sequester proteins and organelles to degrade and recycle these materials (52). Autophagy begins with the isolation of double membrane-bound structures inside a cell. These structures are currently considered either to originate from a pre-existing membrane structure called a phagophore or to be formed de novo (29,53). These membrane structures elongate, mature, and form autophagosomes to sequester cytoplasmic proteins and organelles. The autophagosomes then fuse with lysosomes and become either autolysosomes or degradative autophagic vacuoles. After that, the sequestered contents are degraded by lysosomal hydrolases. Recent genetic studies have identified at least 16 Atg genes in yeast that are necessary to form autophagosomes (5,6). For example, Atg5-Atg12 conjugation is required for the formation of autophagosomes (32). To date, more than five mammalian orthologues of Atg genes have been identified. Beclin 1 and LC3 are orthologues of Atg6 and Atg8, which have been implicated in the formation of autophagosomes (28,31). Autophagy has been observed under various cell conditions, including the degradation of proteins in response to nutrient deprivation, differentiation, aging, and cancer therapy (7)(8)(9). However, the molecular machinery that regulates autophagy is not fully understood. In this study, C 2 -ceramide treatment degraded PARP in p21 Ϫ/Ϫ MEFs while maintaining the expression of Beclin 1 and Atg5 proteins, leading to autophagy rather than apoptosis. Unlike C 2 -ceramide-induced autophagy in breast cancer MCF7 cells (24), C 2 -ceramide did not increase the expression of Beclin 1 in p21 Ϫ/Ϫ MEFs. This difference might be because of the fact that MEFs endogenously express enough Beclin 1 or Atg5 to undergo autophagy, whereas endogenous Beclin 1 expression is low in breast cancer cells, including MCF7 cells (31).
In summary, we demonstrated that p21 plays an essential role in deciding the type of PCD, apoptosis or autophagy, MEFs undergo after exposure to C 2 -ceramide. In other words, in cells with p21, cell death stimuli might degrade autophagy-related proteins while keeping the apoptotic pathway intact, leading to apoptosis. Without p21 expression, the cell death signal for apoptosis might be stopped by the degradation of PARP and the expression of autophagy-related proteins might remain unchanged, resulting in the induction of autophagy instead of apoptosis. When apoptosis is induced, the full-length PARP is cleaved by caspases into the p24 and p89 inactive fragments to prevent excessive NAD consumption and ATP loss (54). On the other hand, DNA-damaging agents such as N-methyl-NЈ-nitro-N-nitrosoguanidine were observed to cause the expression of the full-length PARP protein to be lost but did not cleave the protein, then activated PARP and depleted ATP, leading to nonapoptotic cell death (55). In addition, an increase in PARP activation and reactive oxygen species production caused caspase-independent autophagic cell death (34). These findings might explain why autophagy rather than apoptosis was induced when PARP was degraded but not cleaved in C 2 -ceramide-treated p21 Ϫ/Ϫ MEFs. One possible scenario may be that PARP is not cleaved and might be activated, and autophagy is stimulated because of a fall in ATP. Of course, we need to determine whether PARP is actually activated when PARP is degraded in C 2 -ceramide-treated p21 Ϫ/Ϫ MEFs. Further study will add new knowledge about the PCD pathways.