Cyclin-dependent kinase inhibitors uncouple cell cycle progression from mitochondrial apoptotic functions in DNA-damaged cancer cells.

DNA damage results in transcriptional induction of p53 target genes, including the cyclin-dependent kinase (CDK) inhibitor p21(Cip1) (CDKN1A) and the proapoptotic Bcl-2 family member p53 up-regulated modulator of apoptosis (PUMA). Depending on the cellular context, p21(Cip1) and PUMA mediate cell cycle arrest and apoptosis, respectively. By imposing cell cycle arrest at the expense of apoptosis, p21(Cip1) can sharply reduce the effectiveness of DNA-damaging anticancer agents in colorectal cancer cells. We investigated the link between cell cycle progression and the onset of apoptosis in DNA-damaged cells by analyzing the activation of the apoptotic cascade in p21(Cip1)-deficient HCT116 colorectal cancer cells. DNA damage induced a similar level of p53 activation and PUMA induction in p21(Cip1)-deficient cells compared with wild-type isogenic counterparts. p21(Cip1) did not act as a direct blocker of PUMA. However, only p21(Cip1)-deficient cells showed extensive cytochrome c release, mitochondrial membrane depolarization, and caspase activation. An increase in caspase activation occurred as these cells reached M-phase and incurred polyploidy. When ectopically expressed in p21(Cip1)-deficient HCT116 cells, p21(Cip1), its family member p27(Kip1), and the structurally unrelated CDK inhibitor p16(Ink4a) were similarly effective at causing cell cycle arrest and inhibiting DNA damage-induced apoptotic events such as cytochrome c release, mitochondrial membrane depolarization, and activation of the caspase cascade. These observations suggest that by blocking dysregulated cell cycle progression, CDK inhibitors can influence the sensitivity of the mitochondria to proapoptotic signals in DNA damage-induced cancer cells.

Many tumor-specific treatments are based on inducing DNA damage by radiation or chemotherapeutic agents. In response to DNA damage, the tumor suppressor protein p53 accumulates in the nucleus and regulates transcription of target genes, including PUMA 1 (p53 up-regulated modulator of apoptosis) (1)(2)(3) and CDKN1A (4). PUMA encodes a BH3-only pro-tein member of the Bcl-2 family and is a key mediator in the p53-dependent apoptotic response (5)(6)(7). The CDKN1A gene encodes the cyclin-dependent kinase (CDK) inhibitor p21 Cip1/WAF1 , which is important in executing the p53-dependent cytostatic program (8 -11). Depending on the cellular context, the outcome of a p53 response results in either cell cycle arrest or apoptosis.
The heterogeneity of cancer arises from the sequential acquisition of genetic alterations during oncogenic progression. This creates a major challenge in predicting how a specific tumor will respond to therapy, because each tumor has a different genetic signature, cellular context, and pathology. Malignant cells generally show a greater tendency to die in response to elevated levels of p53 than nontransformed cells, in part because of the additional cellular stresses that accompany malignant progression such as uncontrolled proliferation, hypoxia, and growth factor depletion (12). However, the ability of some tumor cells to undergo cell cycle arrest and evade apoptosis in response to DNA-damaging drugs is a contributing cause of drug resistance. These observations have drawn attention to the genetic context of a tumor cell as a critical determinant of the apoptotic or cell cycle arrest response to DNA damage.
Among the types of cancer that are affected by this problem is colorectal cancer, which is the third most common malignancy worldwide (13). A major effort has been made to understand the genetic bases of colorectal cancer progression (14) and its response to a variety of anticancer agents such as the antimetabolite 5-fluorouracil, alkylating agent cisplatin, and topoisomerase II inhibitor adriamycin. Studies done with the HCT116 human colorectal cancer cell line have shed light on how the genetics of the tumor cell is crucial in dictating the response to anticancer agents. For example, the effectiveness of 5-fluorouracil is attributed to the wild-type p53 status of HCT116 cells, whereas these same cells are resistant to adriamycin treatment (15). HCT116 cells that are rendered deficient in p21 Cip1 by deletion of the CDKN1A locus become hypersensitive to apoptosis induced by DNA damage (16). p21 Cip1 can act as an inhibitor of apoptosis in a variety of systems (17)(18)(19)(20), dramatically limiting the effectiveness of anticancer agents (21). Xenograft tumors from p21 Cip1 -deficient HCT116 cells (22) are hypersensitive to elimination by ␥-irradiation. Signals that abrogate p21 Cip1 expression via c-Myc favor the ability of p53 to activate apoptosis in response to radiation and topoisomerase II inhibitors (23). Thus, we wanted to further elucidate how p21 Cip1 and cell cycle arrest can interfere with the events that trigger cell death in DNAdamaged cancer cells.
To discern how cell cycle progression is coupled to the onset of apoptosis in DNA-damaged cancer cells, we utilized the wild-type and p21 Cip1 -deficient HCT116 cells. The p21 Cip1 -de-ficient HCT116 cells undergo aberrant cell cycle progression in the presence of DNA damage leading to cell death. We have shown that reintroducing p21 Cip1 or other CDK inhibitors such as p27 Kip1 and p16 Ink4a can prevent p21 Cip1 -deficient HCT116 cells from proceeding through S-phase and accumulating a 4N and higher DNA content in the presence of drug-induced DNA damage. In so doing, these CDK inhibitors prevent apoptosisinitiating events including cytochrome c release from the mitochondria, mitochondrial membrane depolarization, and caspase-3 activation.
Immunoblotting-The appropriate cells were lysed in TNE buffer (10 mM Tris, pH 7.8, 1% Nonidet P-40, 1 mM EDTA) supplemented with protease inhibitors (Roche Applied Science). Protein concentration was measured using the BCA kit (Pierce) according to the manufacturer's protocol. 20 -40 g of protein lysate was separated by 12% SDS-PAGE and transferred to a polyvinylidene difluoride membrane (Immobilon-P; Millipore). The membrane was probed with antibodies specific for PUMA (a kind gift from Drs. Vogelstein, Kinzler, and Yu), p53 (sc-98; Santa Cruz), Bcl-x L (BD Biosciences), or p21 Cip1 (ab-11; Neomarkers), developed with the ECL system (Amersham Biosciences), and stripped with Restore TM Western blot stripping buffer (Pierce) shaking at 37°C for 20 min.
Apoptosis Assays-Annexin V staining was performed using the fluorescein isothiocyanate (FITC) Annexin V kit (BD Biosciences) according to the manufacturer's specification. To measure DNA content, floating and adherent cells were collected, washed with PBS, and resuspended in 0.6 ml of PBS. 1.5 ml of 100% ethanol was added dropwise to the cell suspension, and the cells were left at Ϫ20°C for a minimum of 30 min. The fixed cells were washed and resuspended in 0.3-0.5 ml of DNA staining buffer (3.8 mM sodium citrate, 20 g/ml propidium iodide, 0.125 g/ml RNase A). Cells with a DNA content less than 2N were scored as having a sub-G 1 DNA content. To stain cells with FITC-VAD-fmk (CaspACE™; Promega) and propidium iodide, cells were cultured for the indicated times with 5 M FITC-VAD-fmk and 70 ng/ml adriamycin. Both floating and adherent cells were pooled, washed, and fixed with BD Cytofix/Cytoperm™ (BD Biosciences) solutions according to the manufacturer's specifications. Samples were resuspended in FACS buffer (1% BSA and 0.1% sodium azide in PBS) with 10 g/ml propidium iodide and 0.125 g/ml RNase A. All samples were analyzed by flow cytometry on the FACSCalibur flow cytometer (Becton-Dickinson). Data were formatted using Flowjo software.
Intracellular Staining for Flow Cytometry-For intracellular staining of the active form of caspase-3 and MPM-2, cells were fixed and permeabilized using the Cytofix/Cytoperm TM solutions and labeled with FITC-labeled active caspase-3 antibody (BD Biosciences) according to the manufacturer's protocol or fluorescein-conjugated mouse monoclonal anti-MPM-2 antibody (Upstate Biotechnology) for 1 h at 4°C. Bromodeoxyuridine (BrdUrd) incorporation in the cells was measured using the FITC BrdUrd kit (BD Biosciences) according to the manufacturer's protocol. Intracellular staining of cytochrome c in digitoninpermeabilized cells was carried out essentially as described (24).
Rhodamine 123 (Rh123) Uptake Assay-Floating and adherent cells were pooled and incubated for 30 min with 2 M Rh123 (Molecular Probes) in PBS at 37°C. Cells were then washed in PBS, resuspended in FACS buffer, and Rh123 fluorescence was measured by flow cytometry.

p21 Cip1 Prevents Caspase Activation Downstream of PUMA-
To determine the effects of p21 Cip1 on cells after p53 is induced by DNA damage, we characterized the cell cycle and apoptotic machinery in the HCT116 cell line and its isogenic derivative lacking p21 Cip1 (p21 Ϫ/Ϫ cells). 24 h after adriamycin treatment, cell cycle analysis of HCT116 wild-type cells by BrdUrd incorporation and DNA staining with 7-amino-actinomycin D demonstrated that most of the cells were arrested in G 1 (2N DNA content) and G 2 /M (4N DNA content) (16) with few cells in S-phase (Fig. 1A). In contrast, HCT116 p21 Ϫ/Ϫ cells contained a higher percentage of cells in S-phase (Fig. 1A) and a considerably lower percentage of cells in G 1 and G 2 /M phases. This observation is an indication that the cells lacking p21 Cip1 were unable to maintain tight G 1 and G 2 /M checkpoint controls after DNA damage. Additionally, intracellular staining for the Mphase marker MPM-2 (25) revealed that p21 Ϫ/Ϫ cells entered mitosis more readily than the wild-type cells (Fig. 1B). The abundance of cells in S-phase and mitosis after adriamycin treatment shows that the absence of p21 Cip1 allows these cells to progress through the cell cycle unimpeded after sustaining DNA damage. Many of the p21 Ϫ/Ϫ cells treated with adriamycin also contained lobulated nuclei or displayed abnormal nuclear morphology (Fig. 1C), which is a morphological indicator of apoptosis associated with an aberrant mitosis (26).
PUMA induction is essential for p53-dependent apoptosis in HCT116 cells (6). PUMA localizes to the mitochondria and binds to Bcl-x L and Bcl-2 to induce the release of cytochrome c, an event that subsequently triggers caspase activation and apoptosis (1,2). Despite the differences in the sensitivity of the parental and p21 Ϫ/Ϫ cells to adriamycin treatment, a comparable increase in PUMA protein levels in response to adriamycin was observed between the two populations ( Fig. 2A). This confirmed earlier observations that the expression of p21 Cip1 and the ensuing block in cell cycle progression do not affect the ability of p53 to activate the expression of PUMA and several other proapoptotic genes (23). Closer analysis showed that the wild-type cells appear to contain slightly higher levels of PUMA and p53 than the p21 Ϫ/Ϫ cells ( Fig. 2A), suggesting that the loss of p21 Cip1 has altered the sensitivity of these cells to p53 and PUMA levels. Thus, in DNA-damaged HCT116 cells, the p53-dependent induction of PUMA is not sufficient to trigger substantial cell death. However, under conditions in which PUMA induction is sufficient for apoptosis, as in DLD1 colorectal cancer cells conditionally expressing ectopic PUMA (DLD1-PUMA) (2), overexpression of p21 Cip1 (Fig. 2C) did not protect against PUMA-induced apoptosis (Fig. 2B). These results indicate that p21 Cip1 does not directly block PUMA itself or the downstream events that trigger apoptosis under conditions in which PUMA expression alone is sufficient for apoptosis. In contrast to apoptosis in DLD1-PUMA cells, apoptosis in DNA-damaged HCT116 p21 Ϫ/Ϫ cells must require other events in addition to PUMA induction and these events must be sensitive to inhibition by p21 Cip1 .
To study the differences in how DNA-damaged HCT116 wildtype and p21 Ϫ/Ϫ cells process similar proapoptotic signals, we sought to identify the step in the apoptosis cascade that is inter-rupted by the presence of p21 Cip1 in these cells. First, we investigated the activation of caspases in DNA-damaged cells. The sensitivity of HCT116 p21 Ϫ/Ϫ cells to p53-induced apoptosis was inhibited when adriamycin-treated cells were concurrently exposed to carbobenzoxy-valyl-alanyl-aspartyl-(O-methyl)-fluoromethyl ketone Z-VAD-fmk, a pan-caspase inhibitor (Fig. 2D).
There was a small percentage of cell death in the p21 Ϫ/Ϫ cell population that was not inhibited by Z-VAD-fmk at 96 h. This observation raises the possibility that a caspase-independent mechanism of cell death is also triggered in DNA-damaged cells, but the majority of cell death was prevented when caspase activity was inhibited. To confirm the presence of activated caspases, intracellular staining of fixed cells with an antibody that specifically recognizes the active form of caspase-3 was carried out, and a higher level of activated caspase-3 was observed in p21 Ϫ/Ϫ cells than in wild-type cells (Fig. 2E). Cells that have been concurrently treated with adriamycin and Z-VAD-fmk to inhibit caspase activity did not display strong intracellular staining for the active form of caspase-3, confirming the specificity of the antibody. These results indicate that cell death induced by DNA damage is predominantly executed by caspases and p21 Cip1 impedes the progression of the apoptotic pathway at a point prior to caspase activation.
p21 Cip1 Prevents Amplification of the Cytochrome c Release Process-The intrinsic apoptosis pathway initiates caspase activation through the release of cytochrome c from the mitochondria (27) to induce formation of the apoptosome, consisting of cytochrome c, ATP, and Apaf-1. Procaspase-9 is recruited to the complex and activated (28), triggering the subsequent activation of effector caspases (caspase-3, -6, and -7). Among the many substrates of effector caspases, ICAD (29) and Acinus (30) are responsible for mediating DNA fragmentation and nuclear condensation, while cleavage of NDUFS1, the p75 kDa subunit of respiratory complex I, results in mitochondrial dysfunction such as disruption of electron transport and loss in transmembrane potential (31).
We investigated whether wild-type and p21 Ϫ/Ϫ cells display differences in the ability to release cytochrome c from the mitochondria in response to DNA damage. First we verified that adriamycin treatment of HCT116 p21 Ϫ/Ϫ cells induced cytochrome c release, as characterized by a disperse pattern of cytochrome c immunofluorescence staining that accompanied nuclear condensation in these cells (Fig. 3A). Addition of Z-VAD-fmk to block caspase activity inhibited apoptosis, as determined by the absence of nuclear condensation, but it did not prevent cytochrome c release (Fig. 3, A and C). This suggests that the release of cytochrome c from the mitochondria is likely to be a direct effect of an increase in the expression levels of PUMA and other proapoptotic factors. To quantitatively characterize this process under various conditions, we used flow cytometry to analyze fixed cells that were intracellularly stained for cytochrome c. Prior to fixation, the cell membrane was permeabilized with digitonin, allowing cytosolic cytochrome c released from the mitochondria to diffuse out of the cell (24). In both wild-type and p21 Ϫ/Ϫ cells, treatment with adriamycin caused an increase in the proportion of cells showing a loss of cytochrome c signal, indicating cytochrome c release from the mitochondria (Fig. 3B). However, the proportion of cells showing a loss of cytochrome c signal, as well as the extent of that loss, was more pronounced in p21 Ϫ/Ϫ cells than in wild-type cells (Fig. 3, B and C).
The majority of cells with cytochrome c released from the mitochondria in the wild-type and p21 Ϫ/Ϫ populations was not affected by Z-VAD-fmk (Fig. 3C), but Z-VAD-fmk had a strong effect on inhibition of cell death as determined by a sub-G 1 DNA content (Fig. 3D) in the same experiment. This is consistent with the immunofluorescence staining data in Fig. 3A showing that cytochrome c can be released from the mitochondria in the absence of caspase activity. In addition, loss of mitochondrial membrane potential was more pronounced in the p21 Ϫ/Ϫ cells than in the wild-type HCT116 cells (Fig. 3E). Collectively, these results suggest that the amplification of the mitochondria pathway is inhibited in DNA-damaged cells expressing p21 Cip1 .
p21 Cip1 Uncouples the Cell Cycle from Activation Caspases-To search for clues that can explain the competence of p21 Ϫ/Ϫ cells, but not wild-type cells, to trigger cytochrome c release from the mitochondria and caspase activation, we investigated the stage in the cell cycle at which caspase activation occurs upon adriamycin treatment. For this purpose, we measured the accumulation of activated caspases as adriamycin-treated cells progressed through cell cycle over time. We used a fluorescent analog of Z-VAD-fmk in which the FITC group is substituted for the carbobenzoxy (z) N-terminal blocking group. This reagent irreversibly binds to activated caspases, inhibiting their activity to initiate downstream events such as DNA fragmentation while fluorescently marking cells that have achieved activation of caspases (32). Similar increases in the percentage of cells with active caspases after adriamycin treatment were observed in both wild-type and p21 Ϫ/Ϫ cells with a 2N DNA content, but there was a clear difference between the wild-type and p21 Ϫ/Ϫ cells with a DNA content of 4N or higher. The wild-type population showed small changes in the percentage of cells with activated caspases after 72 h of adriamycin treatment, whereas the corresponding p21 Ϫ/Ϫ cells displayed a more prominent change in the percentage of cells with activated caspases (Fig. 4, A and B). Thus, the majority of the gain in caspase activity primarily occurred in p21 Ϫ/Ϫ cells that had progressed beyond G 2 /M phase. These results suggest that the cell cycle is a contributing factor in enhancing the sensitivity of DNA-damaged p21 Ϫ/Ϫ cells to trigger caspase activation in response to PUMA induction.
Cell Cycle Uncoupling from Caspase Activation Is a General Property of CDK Inhibitors-To further investigate whether the sensitivity of HCT116 p21 Ϫ/Ϫ cells to adriamycin is a result of deregulated cell cycle progression, we determined whether other CDK inhibitors can compensate for the absence of p21 Cip1 . HCT116 p21 Ϫ/Ϫ cells were infected with adenovirus encoding ␤-galactosidase (␤-gal), p21 Cip1 , p27 Kip1 , or p16 Ink4a . Both p21 Cip1 and p27 Kip1 preferentially target Cdk2, but p27 Kip1 is structurally and functionally divergent from p21 Cip1 outside the CDK inhibitory domain (33)(34)(35). Furthermore, p16 Ink4a is structurally unrelated to p21 Cip1 or p27 Kip1 and directly targets Cdk4/6 but not Cdk2 (36). Despite these differences, the various CDK inhibitors prevented DNA fragmentation (sub-G 1 DNA content), an indication of cell death in p21 Ϫ/Ϫ cells that were exposed to adriamycin (Fig. 5, A and B) or ␥-radiation (Fig. 5A). Cells expressing p21 Cip1 , p27 Kip1 , or p16 Ink4a underwent G 1 arrest when left untreated and G 1 and G 2 arrest when treated with adriamycin (Fig. 5C). Together, these results show that preventing cell cycle progression after DNA damage can hinder the ability of these cells to trigger apoptosis.
To characterize the DNA damage response of p21 Ϫ/Ϫ cells expressing the different CDK inhibitors, we analyzed the release of cytochrome c from the mitochondria (Fig. 6A), loss of mitochondrial membrane potential (Fig. 6B), and active caspase-3 (Fig. 6C) in these cells. p21 Cip1 , p27 Kip1 , and p16 Ink4a inhibited each of these processes. Thus, p27 Kip1 and p16 Ink4a are as effective as p21 Cip1 in inhibiting the amplification of the mitochondrial-dependent caspase activation pathway and thereby reducing the sensitivity of cancer cells to DNA damageinduced apoptosis.

DISCUSSION
The anti-apoptotic function of p21 Cip1 in the p53 response to DNA damage is not well understood (21). The cellular context and the type of apoptotic stimulus are clearly important factors in determining the outcome. Some evidence suggests that the cytoplasmic localization of p21 Cip1 is important in mediating its anti-apoptotic function via its interaction with procaspase-3 (37) or apoptosis signal-regulating kinase 1 (38). On the other hand, others have linked the anti-apoptotic function of p21 Cip1 to its CDK-inhibiting activity in the nucleus (19,39). Previous work has demonstrated that cells lacking p21 Cip1 continue to undergo rounds of DNA synthesis without completing mitosis, which leads to polyploidy and cell death in response to DNA damage (16). Our work is consistent with these observations and sheds light on the differences in the engagement of the apoptotic machinery in p21 Cip1 -proficient and -deficient cells in response to DNA damage. Furthermore, we provide evidence that the ability of p21 Cip1 to reduce the sensitivity of cancer cells to DNA-damaging agents and to disengage the intrinsic or treated (filled shapes) with adriamycin alone (ϩADR) or adriamycin and 20 M Z-VAD-fmk (ϩADR ϩZ) for the indicated times before they were harvested. Cytochrome c release was detected as described for panel B. D, the graph shows the percentage of cells with a sub-G 1 DNA content over the times indicated from the same experiment described for panel C. To detect the percentage of cells with sub-G 1 DNA content, cells were fixed, stained with propidium iodide, and analyzed by flow cytometry. E, HCT116 WT and p21 Ϫ/Ϫ cells were cultured in the presence (blue) or absence (red) of adriamycin for 72 h. Both floating and adherent cells were harvested and incubated with rhodamine 123 (Rh123) for 30 min at 37°C before analysis by flow cytometry. Mitochondrial membrane depolarization was scored based on loss of Rh123 uptake in the cells. The percentage of cells that fell within the gated population (green box) is shown. apoptosis pathway at a step prior to cytochrome c release from the mitochondria is shared by the CDK inhibitors p27 Kip1 and p16 Ink4a . The data presented here have implications for understanding the relationship between cell cycle progression and apoptosis in tumor cell responses to chemotherapeutic agents.
HCT116 wild-type cells are protected from undergoing cell death in response to DNA damage mainly because of the presence of p21 Cip1 . By characterizing the cell cycle profile of HCT116 wild-type and p21 Ϫ/Ϫ cells after adriamycin treatment, we observed that p21 Ϫ/Ϫ cells entered S-phase and mi- FIG. 4. Increase in caspase activation occurs mainly in HCT116 p21 ؊/؊ cells with a 4N or higher DNA content. A, HCT116 p21 Ϫ/Ϫ cells were cultured with FITC-VAD-fmk in the presence or absence of adriamycin for the indicated times before they were harvested, fixed, stained with propidium iodide, and analyzed by flow cytometry. The positions of cells with a 2N and 4N DNA content are marked with arrows. The red boxes represent the cells with either 2N, 4N, or greater than 4N DNA content that stained positive for active caspases. B, the change in the percentage of HCT116 wildtype and p21 Ϫ/Ϫ cells with a 2N, 4N, or greater than 4N (Ͼ 4N) DNA content that stained positive for active caspases as in panel A after adriamycin treatment for the indicated times was averaged from two experiments and graphed.

FIG. 5. G 1 CDK inhibitors reduce the sensitivity of cells to DNA damage-induced apoptosis.
A, HCT116 wild-type (WT) and p21 Ϫ/Ϫ cells were infected with adenovirus encoding ␤-galactosidase (Ad␤-gal), p21 Cip1 (Adp21), p27 Kip1 (Adp27), or p16 Ink4a (Adp16). 24 h later, fresh medium with or without adriamycin (left panel) was added or 20 Gy ␥-radiation (right panel) was administered to the cells. At 72 h, floating and adherent cells were pooled, fixed, stained with propidium iodide, and analyzed by flow cytometry. The percentage of cells with a sub-G 1 DNA content was determined, and the average from two experiments is shown. B, cell extracts were prepared from cells in an experiment in panel A after treatment with adriamycin for 24 h and were subjected to immunoblotting with antibodies specific for the indicated proteins. The protein levels of Bcl-x L served as a loading control because its expression did not change. C, 24 h after HCT116 p21 Ϫ/Ϫ cells were infected with Ad␤-galactosidase, Adp21, Adp27, and Adp16, the cells were cultured in the presence (ϩADR) or absence (ϪADR) of adriamycin for an additional 24 h. BrdUrd was added to the culture 1 h before the cells were harvested. Harvested cells were fixed, stained with anti-BrdUrd-FITC and 7-amino-actinomycin D, and analyzed by flow cytometry. The percentage of cells in the gated regions is shown on the dot plot and reported in the bar graph. tosis more readily than wild-type cells, which correlated with a higher frequency of lobulated nuclei and activated caspases. These results are consistent with other works suggesting that aberrant entry into S-phase and mitosis in the presence of DNA damage appears to be an important determinant for chemotherapy outcome (40 -42). DNA damage induced a similar level of p53 activation and PUMA induction in the parental and p21 Cip1 -deficient cells, but only p21 Cip1 -deficient cells showed extensive cytochrome c release, mitochondrial membrane depolarization, and caspase activation. p21 Cip1 did not act as a direct blocker of PUMA, suggesting that p21 Cip1 indirectly inhibits an event that can potentiate the activity of PUMA to trigger apoptosis. Ectopic expression of p21 Cip1 , p27 Kip1 , and p16 Ink4a in p21 Cip1 -deficient HCT116 cells was similarly effective in inhibiting cell cycle progression and reducing the ability of these cells to trigger apoptosis in response to DNA damage, thus demonstrating how CDK inhibitors can block aberrant cell cycle progression to influence the apoptotic response to anticancer therapy.
p21 Cip1 Can Modulate the Sensitivity to PUMA Induction-The failure of p21 Cip1 to block apoptosis in the DLD1-PUMAinducible cells shows that p21 Cip1 cannot directly block PUMA or the downstream apoptotic events that are triggered by PUMA. Therefore, p21 Cip1 would have to block apoptosis at a point either upstream of and/or parallel to PUMA. Our previous work demonstrating that p21 Cip1 does not hinder the induction of PUMA expression by p53 (23) rules out the possibility that p21 Cip1 interferes with the p53-dependent induction of PUMA. This then argues that p21 Cip1 is most likely modulating a parallel signaling regime that calibrates the sensitivity of the cell to PUMA induction. The molecular mechanism underlying this process remains to be identified. However, one possibility is that this process is linked to the Rb-E2F pathway. The ability of Rb to bind to the E2F transcription factor and inhibit the activation of E2F target genes is dependent on the activity of cyclin-CDK complexes (36). Phosphorylation of Rb by cyclin-CDK complexes results in dissociation of Rb from E2F and activation of E2F-responsive genes. Many of the E2F-responsive genes are involved in DNA replication and cell cycle progression, but recent studies have also identified genes involved in apoptosis such as Apaf-1 (43,44), caspase-7 (45), and p73 (46,47). As DNA-damaged cells undergo M-phase and re-enter G 1 -phase, they would be liable to activate E2F-dependent apoptotic gene responses. Therefore, the ability of CDK inhibitors to block CDK activity and inhibit the expression of E2F-responsive proapoptotic genes would reduce the sensitivity of these cells to apoptotic signals.
Additionally, a mitotic checkpoint protein, Mad2, has been identified as a direct E2F target gene. Aberrant expression of Mad2 due to Rb inactivation leads to mitotic defects and aneuploidy (48). Hence, it is conceivable that DNA-damaged cells lacking p21 Cip1 mimic this phenotype leading to hypersensitivity to proapoptotic signals.
It was recently reported that inhibition of Cdk2 kinase activity by ectopic expression of p27 Kip1 , expression of dominantnegative Cdk2, or antisense oligonucleotides did not prevent cell cycle progression in HCT116 cells (49). Our data are inconsistent with this observation. In our hands, overexpression of various structurally different CDK inhibitors was sufficient to block cell cycle progression in the p21-deficient HCT116 cells, FIG. 6. CDK inhibitors can prevent robust engagement of the mitochondria pathway in HCT116 p21 ؊/؊ cells in response to DNA damage. A, HCT116 p21 Ϫ/Ϫ cells were infected with Ad␤-galactosidase, Adp21, or Adp16. 24 h later, cells were untreated (ϪADR) or treated (ϩADR) with adriamycin for an additional 72 h before they were harvested. Cells were permeabilized with digitonin, fixed, stained for intracellular cytochrome c, and analyzed by flow cytometry. The numbers represent the percentage of cells that fall in the area with low fluorescent intensity. B, HCT116 p21 Ϫ/Ϫ cells were infected with the indicated adenovirus 24 h before they were exposed to 20 Gy of ␥-radiation. After 72 h, the cells were harvested and stained with Rh123 for 30 min before the samples were analyzed by flow cytometry. The green box indicates cells with mitochondrial membrane depolarization as detected by low Rh123 uptake. The percentage of cells in the green box is shown. C, HCT116 p21 Ϫ/Ϫ cells expressing exogenous p21 Cip1 , p27 Kip1 , or p16 Ink4a by means of adenovirus infection were subjected to intracellular staining with an antibody that specifically recognizes the active form of caspase-3 after 72 h of mock (red) or adriamycin (blue) treatment and analyzed by flow cytometry. an isogenic derivative of wild-type HCT116 cells. In addition, the ability of wild-type HCT116 cells to arrest in G 1 in response to adriamycin and ␥-radiation is lost in p21 Cip1 -deficient cells (11,16), supporting the role of endogenous p21 Cip1 in HCT116 cells as an effective inhibitor of cell cycle progression.
A Link between Mitotic Cell Death and Apoptosis-It has been argued that mitotic death is the clinical determinant of cytotoxicity in response to radiation and DNA-damaging agents in tumors of epithelial origins and that apoptosis is clinically relevant only in tumor cells of myeloid and lymphoid origins (50). The distinction between these two forms of cell death is based on the idea that apoptosis is a rapid process triggered by the primary DNA-damaging event, whereas mitotic death, or mitotic catastrophe, results from an aberrant mitosis that does not produce proper chromosome segregation and cell division, leading to formation of large non-viable cells with broken chromosomes (26,51). Thus, apoptosis following mitosis would not be a determinant of treatment sensitivity when it is triggered by a signal subsequent to mitosis in cells that can no longer survive. The present results show that the majority of cell death induced by DNA damage is executed by caspase activity triggered by the mitochondria pathway, and others have demonstrated the requirement for PUMA and Bax to initiate this pathway (6). However, the presence of PUMA and proapoptotic factors induced by p53 in response to DNA damage is not sufficient to initiate apoptosis unless these events are coupled with improper cell cycle progression due to the absence of p21 Cip1 . Therefore, we suggest that mitotic cell death and apoptosis via the intrinsic pathway act in a synergistic manner to trigger cell death in response to DNA-damaging agents.
Many chemotherapeutic drugs exhibit cell cycle phase-specific activity; therefore they depend on cell cycle progression to exert optimal efficiency. Thus, interfering with cell cycle progression can diminish the cytotoxicity of these drugs (52). For example, topoisomerase II inhibitors are most potent during S-phase to create DNA double strand breaks (53). Therefore, it is conceivable that strategies aimed at inhibiting CDK activity in tumor cells might hinder the effectiveness of DNA-damaging agents that are cell cycle phase-specific. Conversely, strategies aimed at suppressing cell cycle checkpoint controls in combination with DNA-damaging agents may be a superior option.