The Nuclear Death Domain Protein p84N5 Activates a G 2 /M Cell Cycle Checkpoint Prior to the Onset of Apoptosis*

In contrast to extracellular signals, the mechanisms utilized to transduce nuclear apoptotic signals are not well understood. Characterizing these mechanisms is important for predicting how tumors will respond to genotoxic radiation or chemotherapy. The retinoblastoma (Rb) tumor suppressor protein can regulate apoptosis triggered by DNA damage through an unknown mechanism. The nuclear death domain-containing protein p84N5 can induce apoptosis that is inhibited by association with Rb. The pattern of caspase and NF- k B activation during p84N5-induced apoptosis is similar to p53-independent cellular responses to DNA damage. One hallmark of this response is the activation of a G 2 /M cell cycle checkpoint. In this report, we characterize the effects of p84N5 on the cell cycle. Expression of p84N5 induces changes in cell cycle distribution and kinetics that are consistent with the activation of a G 2 /M cell cycle checkpoint. Like the radiation-induced checkpoint, caffeine blocks p84N5-induced G 2 /M arrest but not subsequent apoptotic cell death. The p84N5-induced checkpoint is functional in ataxia telangiectasia-mu-tated kinase-deficient cells. We conclude that p84N5 induces an ataxia telangiectasia-mutated kinase (ATM)-independent, caffeine-sensitive G 2 /M cell cycle arrest prior to the onset of apoptosis. This conclusion is consistent with the hypotheses

Programmed cell death is essential for normal development, tissue homeostasis, and host defense mechanisms. Programmed cell death is recognized by a collection of distinctive morphological and biochemical characteristics termed apoptosis (1). The pathways leading to apoptosis in response to extracellular stimuli, such as tumor necrosis factor, or in response to mitochondrial apoptotic signals, such as cytochrome c release, are well characterized (2,3). Initiator caspases are typically recruited to protein complexes composed of death receptors and/or adapter molecules. These proteins contain signature protein interaction motifs like the death domain, the death effector domain, or the CARD domain. The locally high concentration of recruited caspase proenzyme triggers its activation by proteolytic processing thereby initiating a caspase proteolytic cascade. Apoptotic signals can also originate from within the nucleus. For example, DNA damage caused by radiation triggers a stress response that can result in apoptotic cell death (4). The mechanisms utilized by nuclear signals to initiate apoptosis are not well understood.
A number of nuclear transcription factors can induce apoptosis, including p53 (5,6). Although p53 has a well documented role in the response of the cell to DNA damage, the mechanism used by p53 to initiate apoptosis is controversial. Although it may trigger apoptosis from within the nucleus by altering the expression of genes directly involved in the execution of apoptosis (7), non-nuclear mechanisms unrelated to transcriptional regulation have also been proposed (8). Few proteins other than transcription factors are known to require nuclear localization to initiate apoptosis. Nuclear localization of expanded polyglutamine repeat proteins is required for their ability to induce apoptosis that causes progressive neurodegenerative diseases like Huntington's disease or spinocerebellar ataxia (9,10). Activation of caspase-8 is a required step in this process (11). Activation of caspase-8 apparently occurs by a novel mechanism involving recruitment of the proenzyme to characteristic protein aggregates that are associated with these diseases.
Recently, we demonstrated that the nuclear protein encoded by the N5 gene (p84N5) 1 was capable of initiating p53-independent apoptosis (12). Since p84N5 contains a death domain that is required for its ability to induce apoptosis, it may participate in a nuclear apoptotic pathway. Consistent with this hypothesis, p84N5-induced apoptosis has a pattern of caspase and NF-B activation that is similar to radiationinduced apoptosis (13). The N5 gene was originally isolated based on the ability of p84N5 to bind an amino-terminal domain of the retinoblastoma tumor suppressor protein (Rb) (14). Rb can regulate both p53-dependent and p53-independent apoptotic responses to DNA damage (15). Association with pRb inhibits p84N5-induced apoptosis, identifying p84N5 as a potential mediator of the inhibitory effects of pRb on p53-independent apoptosis. One characteristic feature of DNA damageinduced apoptosis, especially in the absence of wild-type p53, is the activation of a G 2 /M cell cycle checkpoint prior to cell death (16). In the current study, we test whether p84N5 activates a G 2 /M cell cycle checkpoint.

EXPERIMENTAL PROCEDURES
Cell Culture-SAOS-2, 293, and C-33A cell lines were obtained from American Type Culture Collection and maintained in Dulbecco's modified Eagle's medium with 10% heat-inactivated fetal bovine serum and antibiotics (100 units/ml penicillin, 100 g/ml streptomycin) in a 5% CO 2 incubator at 37°C. The AT22IJE-T cell line and the ATM-expressing derivative were cultured under the same conditions except for the addition of hygromycin as described (17). Viable cells were counted after trypan blue staining using a hemocytometer. Caffeine was added to the culture media to a final concentration of 2 mM.
Plasmids and Adenovirus-The full-length p84N5 cDNA was subcloned into the pCEP4 (Invitrogen, Carlsbad, CA) expression vector as described previously (12) to create the expression vector pCMVN5. This plasmid was used to express p84N5 upon calcium phosphate-mediated transfection. The recombinant p84N5-expressing adenovirus (AdN5) was made as described previously (13,18) by cloning the p84N5 cDNA into the pAdCMV (AS)-BGHpa vector. The green fluorescent proteinexpressing adenovirus (AdGFP) was made similarly and was provided by Dr. T. J. Liu (M. D. Anderson Cancer Center). The Ad/E1Ϫ adenovirus is made from the pXCJL and pJM17 plasmids and lacks adenoviral E1 but contains no foreign gene. Recombinant adenovirus was purified by CsCl equilibrium density gradient centrifugation and viral particle numbers estimated by A 260 in the presence of SDS as described (19). Infectious titer was determined by an end point dilution of the viral stock on 293 cells. Viral infections were typically performed by adding an appropriate number of infectious units to cells at a multiplicity of infection of 10 and incubating under normal growth conditions overnight.
Transfections and Western Blotting-For transfections, 293 cells were seeded on 100-mm dishes and transfected the following day by calcium phosphate precipitation (20) using 10 g of total DNA. Transfections typically included 1 g of the pEGFP-C1 plasmid to measure the transfection efficiency the following day under fluorescence microscopy. Transfected cells were extracted in a buffer containing 50 mM Tris, pH 7.4, 250 mM NaCl, 5 mM EDTA, 0.1% Nonidet P-40, 50 mM NaF, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin on ice for 10 min. The total protein concentration of the soluble extract was determined by Bradford assay according to manufacturer's instructions (Bio-Rad). 25 g of total protein for each sample, normalized for transfection efficiency, was resolved by 10% SDS-polyacrylamide gel electrophoresis, blotted, and stained as described previously (13). Antibody directed against p84N5 (14) was described previously. All other antibodies were used as directed by the manufacturer (Santa Cruz Biotechnology, Santa Cruz, CA). Primary antibodies were detected using a peroxidase-conjugated secondary antibody and Enhanced Chemiluminescence according to manufacturer's recommendations (Amersham Pharmacia Biotech).
Cell Cycle and Apoptosis Analysis-Routine analysis of cell cycle distribution was determined by propidium iodide (PI) staining and flow cytometry. Infected cells were harvested by trypsinization, resuspended in growth media, washed once in phosphate-buffered saline (PBS), resuspended in 0.5 ml PBS, and fixed by addition of ice-cold 95% ethanol to 60% while vortexing. Cells were stored in ethanol overnight at Ϫ20°C prior to staining. Cells were then washed in PBS and resuspended in PBS containing 10 g/ml propidium iodide and 250 g/ml RNase A. Cells were incubated at 37°C for 15 min prior to flow cytometric analysis using a Coulter EPICS Profile instrument (Beckman Coulter Inc., Fullerton, CA). Cell cycle distributions were determined from histograms using Multicycle (Phoenix Flow Systems, San Diego, CA).
For kinetic analysis of the cell cycle, bromodeoxyuridine (BrdUrd, 1 M) was added to the culture media at the indicated times after infection. After 20 min BrdUrd media were removed, and the cells were washed three times with pre-warmed and gassed media, and then either fixed with ethanol as above (time 0) or incubation was continued for 6 h before fixation. After fixation, the cells were prepared for kinetic flow cytometric analysis. Fixed cells were digested in 0.04% pepsin (EM Science, Cherry Hill, NJ) in 0.1 N HCl for 20 min while rocking at room temperature. After incubation with 2 N HCl for 20 min at 37°C, 0.1 M sodium borate (twice the HCl volume) was added. The nuclei were then centrifuged and washed with PBS containing 0.5% Tween 20 and 0.5% bovine serum albumin (PBTB). After centrifugation the nuclei were resuspended in anti-BrdUrd monoclonal antibody IU-4 (1:100 v/v, Caltag, South San Francisco, CA) in PBS plus 0.5% Tween 20 (PBT) and then incubated at room temperature for 1 h in the dark. Another washing with PBTB followed, and then the nuclei were incubated for 1 h in the dark at room temperature with fluorescein isothiocyanate-conjugated goat anti-mouse immunoglobulin G (IgG; 1:100 v/v; Sigma) in PBT and 1% normal goat serum. After a final wash in PBTB, the nuclei were resuspended in 10 g/ml PI (Sigma) at a concentration of 10 6 nuclei/ml in PBTB.
Bivariate distributions of BrdUrd content (fluorescein isothiocyanate) versus DNA content (PI) were measured using an Epics 752 flow cytometer (Coulter Corp., Hialeah, FL) equipped with narrow beam (5-m) excitation optics, a low velocity quartz flow cell and Cicero data acquisition and display electronics (Cytomation, Fort Collins, CO). Excitation was at 488 nm using a 5-watt argon ion laser operating at 200 milliwatts. After blocking incident laser light, BrdUrd was measured using a logarithmic amplifier with a 530-nm short pass filter and DNA content collected after a 610-nm long pass filter. There was no spectral overlap of the emitted fluorescence using this optical configuration. Doublets and clumps were excluded from the analysis by gating on a bivariate distribution of the red peak versus integral signal. Data from 30,000 events were collected in the final gated histograms. Bivariate DNA versus BrdUrd histograms were analyzed using the "Summit" software (Cytomation), and one-dimensional DNA histograms were fitted using Modfit LT (Verity Software House, Topsham, ME).
The analytical methodology for calculation of kinetic parameters has been described in detail elsewhere (21)(22)(23)(24). The analysis is based on our established approach that we introduced to compute cell kinetic parameters using the extra information inherent in a bivariate DNA versus incorporated BrdUrd flow cytometry histogram together with our more recent method for the simultaneous estimation of the durations of G 2 and Mitosis (T G 2 ϩM), and S phase (T S ), and the potential doubling time (T pot ) using single sample dynamic data from bivariate DNA-thymidine analogue cytometry (25). In this study labeling index (LI) was measured directly using data taken immediately (20 min) after pulse labeling and computed for the 6-h time points from LI ϭ e c(T G2 ϩ M ) (e v Ϫ 1) where c is the growth rate of the population, and is a dimensionless quantity based on the division status of labeled cells, which may be measured from these labeled populations (22,23,25).
Apoptotic cells were identified by staining with Annexin-V-Alexa TM 568 according to the manufacturer's recommendations (Roche Molecular Biochemicals). At least 150 cells for each infection were counted under fluorescence microscopy using a Texas Red filter, and the percentage of cells staining with Annexin-V-Alexa TM 568 was determined.

Adenoviral-mediated Expression of p84N5 Induces a G 2 /M
Phase Cell Cycle Arrest-The full-length, wild-type N5 cDNA was used to generate a recombinant, E1-deleted, replicationdefective adenovirus (AdN5) that expressed wild-type p84N5 under control of the cytomegalovirus early promoter and the bovine growth hormone polyadenylation signal (13,18). This adenovirus was used to drive expression of p84N5 in infected cells. The cell cycle distribution of cells at different times after infection was determined by propidium iodide staining and flow cytometry. Two days after infection, the percentage of cells containing 2 N DNA content indicative of the G 2 /M phase increased substantially in AdN5-infected SAOS-2 and C33-A cells relative to cells infected with an adenovirus designed to express the green fluorescent protein (AdGFP) (Fig. 1, A and  B). AdGFP-infected SAOS-2 cells had a mean of 15.8% of cells in the G 2 /M phase of the cell cycle 2 days after infection, whereas AdN5-infected cells had a mean of 29.0% of cells in the G 2 /M phase of the cell cycle at the same time point. The mean percentage of AdGFP-infected C33-A cells with G 2 /M DNA content 2 days after infection was 13.6% compared with 32.8% for AdN5-infected cells. To ensure that the changes in cell cycle distribution observed were caused by expression of p84N5, similarly treated SAOS-2 cells were extracted, and the protein extracts were analyzed for p84N5 by Western blotting. AdN5infected cells show a 3-5-fold increase in the accumulation of p84N5 compared with AdGFP-infected cells (Fig. 1C).
These results are consistent with induction of a G 2 /M cell cycle delay caused by p84N5. However, cell death or an alteration in the duration of other cell cycle phases may cause similar changes in cell cycle distribution. To confirm that p84N5 expression increases the duration of G 2 /M phase, we compared the kinetic parameters of the cell cycle in AdN5-and AdGFP-infected cells. S phase cells were pulse-labeled with BrdUrd at various times after infection. Cell populations at 0 and 6 h after labeling were analyzed by bivariate flow cytometry. The histograms were used to calculate the potential doubling time, the duration of S phase, and the duration of G 2 /M phase (21)(22)(23)(24)(25). Consistent with the hypothesis that AdN5 infection triggers a G 2 /M cell cycle delay, the duration of G 2 /M phase for both SAOS-2 and C33-A AdN5-infected cells increased compared with AdGFP-infected cells (Table I). For both cell lines, the duration of G 2 /M phase increased approximately 2-fold. There was no consistent difference observed in the duration of S phase or the potential doubling time between AdN5and AdGFP-infected cells.
To assess the extent of this G 2 /M delay, we infected cells and subsequently treated them with aphidicolin to block new DNA synthesis. If cells were delayed through G 2 /M, then the G 2 /M peak should disappear over time in the presence of aphidicolin leading to an increase in the fraction of G 1 phase cells. If, however, the cells were more permanently arrested prior to mitosis, the fraction of G 1 and G 2 /M cells would remain constant unless there is loss due to cell death. Cells infected with Ad/E1Ϫ, a nonrecombinant replication-defective adenovirus, did show a decrease in the fraction of G 2 /M phase cells and an increase in G 1 phase cells in the presence of aphidicolin (Fig. 2).
The G 1 fraction increased from 37 Ϯ 1 to 54 Ϯ 2% with aphidicolin, and the G 2 /M fraction decreased from 32 Ϯ 1 to 11 Ϯ 1%. As expected, AdN5-infected cells exhibited an increase in G 2 /M phase cells and cells with sub-G 1 DNA content relative to Ad/E1Ϫ cells. Interestingly, 26 Ϯ 5% of AdN5-infected cells were aneuploid, suggesting that these cells had reinitiated DNA synthesis without completing the intervening mitosis. In the presence of aphidicolin, the fraction of G 2 /M phase cells decreased from 41 Ϯ 3 to 22 Ϯ 3%. However, aphidicolin did not cause a significant increase in the fraction of G 1 phase cells (25 Ϯ 3 versus 32 Ϯ 4%). Furthermore, the apparent fraction of S phase cells increased from 38 Ϯ 1 to 60 Ϯ 2% in cells prevented from synthesizing DNA by aphidicolin treatment. This suggested that the decrease in the fraction of G 2 /M cells upon aphidicolin treatment was not due to completion of mitosis and entry into G 1 phase but rather was due to loss of DNA content in G 2 /M-arrested cells during apoptotic cell death.
We have also compared the expression levels of various regulatory proteins that serve as markers of the cell cycle in pCMVN5 versus pCMV-transfected cells. Consistent with G 2 /M cell cycle arrest, the level of both cyclin A and cyclin B protein increase in pCMVN5-transfected cells relative to pCMV-transfected cells (Fig. 3). Cyclin A and B are normally expressed beginning in S phase, and protein levels peak during G 2 phase. The levels of cyclin D1, cyclin E, cdk4, p16, and p27 are not significantly different in pCMVN5-or pCMV-transfected cells.

N5 Induces a G 2 /M Checkpoint
As expected, the amount of p84N5 increases 3-5-fold upon pCMVN5 transfection relative to pCMV transfection.

The G 2 /M Checkpoint Activated by p84N5 Expression Is Sensitive to Caffeine but Does Not Require Functional ATM
Protein-Caffeine has been demonstrated to abrogate cell cycle checkpoint controls that are normally activated in response to DNA damage (26,27). However, the type of DNA damage and the cell cycle phase in which it occurs influence whether caffeine will affect the subsequent cell cycle checkpoint (28). Abrogation of these checkpoints facilitates subsequent cell death by apoptosis. In particular, caffeine blocks G 2 /M cell cycle arrest and increases sensitivity to ionizing irradiation or other genotoxic treatments. The mechanisms utilized by caffeine to block activation of the G 2 /M cell cycle checkpoint are not completely defined. However, caffeine has been demonstrated to inhibit the ataxia telangiectasia-mutated (ATM) kinase (29,30). ATM kinase can phosphorylate and activate the cell cycle regulator Chk2/Cds1. Activation of Chk2/Cds1, in turn, enforces a G 2 /M checkpoint by phosphorylating and inactivating Cdc25C. Cdc25C is normally required to remove an inhibitory phosphate on the mitotic cyclin-dependent kinase Cdk1. Loss of ATM function, therefore, compromises cell cycle checkpoints that are triggered in response to genotoxic stress (31).
We analyzed the effects of caffeine treatment on the p84N5induced G 2 /M cell cycle delay and apoptosis. C33-A cells were infected with AdN5, AdGFP, or Ad/E1Ϫ in the presence or absence of caffeine. The cell cycle distribution of cells was analyzed by propidium iodide staining and flow cytometry at varying times after infection. The percentage of apoptotic cells was determined by staining with annexin-V. As expected, AdN5 infection induced a significant accumulation of cells in the G 2 /M phase of the cell cycle and a 2-3-fold increase in the percentage of apoptotic cell by 2 days after infection (Fig. 4). Treatment with caffeine prevented the accumulation of G 2 /M phase cells normally observed upon AdN5 infection (Fig. 4A). In the experiment shown, the fraction of G 2 /M phase cells 2 days after infection with AdN5 was 36.3% in the absence of caffeine and 18.5% in the presence of caffeine. Similar results were obtained with SAOS-2 cells (data not shown). However, caffeine treatment did not alter the percentage of apoptotic cells observed (Fig. 4B). Given the experimental conditions, it is not possible to determine whether caffeine increases the rate of apoptosis. However, the number of remaining viable cells 48 h after AdN5 infection is typically reduced by 40% in the presence of caffeine (1 ϫ 10 6 viable AdN5-infected cells in the absence of caffeine and 6.1 ϫ 10 5 viable AdN5-infected cells in the presence of caffeine), suggesting that caffeine may sensitize cells to AdN5-induced apoptotic cell death.
Since caffeine abrogated the p84N5-induced G 2 /M cell cycle arrest and caffeine can inhibit the ATM kinase, we tested the hypothesis that ATM may be required for activation of this checkpoint. Immortalized ataxia telangiectasia (AT) fibroblasts that lack wild-type ATM and the same cells reconstituted for ATM function by expression of recombinant ATM (AT/ATM) were infected with AdN5 or AdGFP, and the cell cycle distribution of infected cells was determined. AdGFP infection had little effect on the cell cycle distribution of these cells although a consistent small increase in the fraction of G 2 /M cells 2 days following infection was observed (Fig. 5). As in C33-A and SAOS-2 tumor cell lines, AdN5 caused a large increase in the percentage of cells in the G 2 /M phase of the cell cycle 2 days after infection. On average, the percentage of G 2 /M phase AdN5-infected cells was about 2-fold greater than AdGFPinfected cells for both the AT and AT/ATM cell lines. The presence of reconstituted ATM function did not appear to influence the extent or the rate of the accumulation of G 2 /M phase cells since the cell cycle distributions for AT and AT/ATM cells were very similar. DISCUSSION The results presented here confirm the hypothesis that p84N5 expression activates an authentic G 2 /M cell cycle checkpoint prior to the onset of apoptosis. This conclusion is supported first by the fact that AdN5 causes an accumulation of cells with G 2 /M phase DNA content relative to infection with the control virus. Since similar results have been obtained in two different tumor cell lines and an immortalized human AT fibroblast cell line, the effects observed are unlikely to be cell line-or tumor cell-specific. Neither p53 nor Rb is required for this G 2 /M arrest since SAOS-2 cells are null for both. Furthermore, functional ATM is not required for activation of the G 2 /M checkpoint by p84N5 since AT cells lacking ATM still accumulate in the G 2 /M phase of the cell cycle after infection with AdN5. Second, the accumulation of cells in G 2 /M phase is due to an increase in the duration of this phase of the cell cycle. Based on analysis kinetic data obtained from bivariate flow cytometry of BrdUrd pulse-labeled cells, the calculated duration of G 2 /M doubles upon AdN5 infection relative to AdGFP infection. In contrast, the duration of S phase and the potential doubling time are not consistently dissimilar in AdN5-and AdGFPinfected cells. Third, the relative increase in cyclin A and cyclin B expression observed in pCMVN5-versus pCMV-transfected cells is consistent with an accumulation of cells in the G 2 /M phase of the cell cycle since the expression of these proteins peak during this phase. The accumulation of G 2 /M phase cells upon pCMVN5 transfection also indicates that the effects ob-served are not dependent on adenovirally mediated gene transfer. Finally, treatment of AdN5-infected cells with caffeine, a known inhibitor of the G 2 /M cell cycle checkpoint, prevents accumulation of G 2 /M phase cells.
DNA damage induced by ionizing radiation also activates a G 2 /M checkpoint that is sensitive to caffeine. Although caffeine abrogates radiation-induced G 2 /M cell cycle arrest, it sensitizes cells to radiation-induced cell death (32). This indicates that G 2 /M cell cycle arrest is not a prerequisite for subsequent apoptosis. Similarly, caffeine abrogates p84N5-induced G 2 /M checkpoint activation but not apoptosis. The percentage of apoptotic AdN5-infected cells is similar in the presence or absence of caffeine. Although the percentage of apoptotic cells at the 48-h time point is not significantly altered by caffeine, caffeine does reduce the total number of viable AdN5-infected cells. These observations suggest that, like radiation, caffeine sensitizes cells to the effects of AdN5. The G 2 /M arrest induced by DNA damage is persistent. In the absence of normal G 1 and S phase checkpoints, such as in Rb or p53 mutant cells, G 2 /M cells arrested by DNA damage will initiate another round of DNA synthesis without completing the intervening mitosis. Polyploid cells are also observed after infection with AdN5 but not control virus Ad/E1Ϫ. Like DNA damage-induced checkpoint activation, this suggests that AdN5-infected cells arrested at G 2 /M can reinitiate DNA synthesis without completing the intervening mitosis. Blocking DNA synthesis by treatment with aphidicolin does not cause an increase in the percentage of AdN5 cells in G 1 phase as is observed in cells infected with Ad/E1Ϫ, but rather causes an increase in the fraction of cells with S phase DNA content. Since new DNA synthesis is blocked by aphidicolin, the increased fraction of apparent S phase cells is likely due to loss of DNA content during apoptotic cell death of G 2 /M or polyploid cells. Hence, it appears as if cells arrested at G 2 /M by AdN5 either re-initiate DNA synthesis and/or die by apoptotic cell death before completing mitosis and reentering G 1 phase.
Caffeine can inhibit ATM protein kinase activity and subsequent activation of Chk2/Cds1 (29,30). However, AdN5-and radiation-induced (data not shown) G 2 /M cell cycle arrest still occurs in the absence of ATM, indicating that ATM-independent mechanisms must exist to enforce a G 2 /M checkpoint. This finding is consistent with the observations that Chk2 is dispensable for initiation of the G 2 /M phase checkpoint (35) and that ATM-independent mechanisms may exist to regulate Chk2 (28). How p84N5 expression leads to maintenance of Cdk1 in the inactive state required to prevent mitotic entry is unclear. One possibility is that forced p84N5 expression stresses the cell in a manner that induces an ATM-independent response analogous to that observed in response to DNA damage.
DNA damage is one apoptotic signal that unambiguously originates from within the nucleus. The nuclear localized Rb protein negatively regulates DNA damage-induced apoptosis (36,37). How Rb influences DNA damage-induced apoptosis is not completely understood. The observation that an aminoterminal domain of Rb that is dispensable for cell cycle regulation may be required to inhibit some forms of apoptosis 2 (38) suggests that the ability of Rb to regulate apoptosis may be a novel function. The amino-terminal domain of Rb is also required to bind p84N5, and this binding inhibits p84N5-induced apoptosis (12) and G 2 /M arrest. 2 We hypothesize that p84N5 plays a role in a cellular stress response that is similar to that triggered by radiation-induced DNA damage and, therefore, that it is a good candidate for mediating the effects of Rb on this response.