The p14 alternate reading frame (ARF) tumor suppressor plays a central role in cancer by binding to mdm2 (Hdm2 in humans) and enhancing p53-mediated apoptosis following DNA damage and oncogene activation. It is unclear, however, how ARF initiates its involvement in the p53/mdm2 pathway, as p53 and mdm2 are located in the nucleoplasm, whereas ARF is largely nucleolar in tumor cells. We have used immunofluorescence and coimmunoprecipitation to examine how the subnuclear distribution and protein-protein interactions of ARF change immediately after DNA damage and over the time course of the DNA damage response in human tumor cells. We find that DNA damage disrupts the interaction of ARF with the nucleolar protein B23(nucleophosmin) and promotes a transient p53-independent translocation of ARF to the nucleoplasm, resulting in a masking of the ARF NH2 terminus that correlates with the appearance of ARF-Hdm2 complexes. The translocation also results in an unmasking of the ARF COOH terminus, suggesting that redistribution disrupts a nucleolar interaction of ARF involving this region. By 24 hours after irradiation, DNA repair has ceased and the pretreatment immunofluorescence patterns and complexes of ARF have been restored. Although the redistribution of ARF is independent of p53 and likely to be regulated by interactions other than Hdm2, ARF does not promote UV sensitization unless p53 is expressed. The results implicate the nucleolus and nucleolar interactions of the ARF, including potentially novel interactions involving its COOH terminus as sites for early DNA damage and stress-mediated cellular events.
- DNA damage
- UV radiation
The p14 alternate reading frame (ARF) tumor suppressor plays a central role in cancer by stabilizing p53, a mediator of stress-induced cell cycle arrest and apoptosis. ARF inhibits the interaction of p53 with its key negative regulator, mdm2 (Hdm2 in humans), a ubiquitin ligase that targets p53 for degradation through the proteasome ( 1, 2). Because mdm2 expression is induced by p53, a feedback mechanism is established that prevents the accumulation of either protein under normal conditions ( 3). In response to oncogenic hyperproliferative signals, ARF is induced ( 4– 6). Through an interaction involving the NH2-terminal domain of ARF and the COOH-terminal domain of mdm2, ARF blocks the ability of mdm2 to target p53 for degradation ( 7, 8). This results in increased accumulation of p53 protein followed by enhanced p53-mediated transcriptional transactivation and p53-mediated cell cycle arrest or apoptosis ( 1, 2, 9). This mechanism provides an important barrier to oncogene-induced hyperproliferation, and cancer cells frequently escape this barrier through loss or deregulation of p53 or ARF.
Through an interaction with mdm2/Hdm2 in normal mouse and human fibroblasts, respectively, ARF has also been implicated in the p53-mediated cell cycle arrest and apoptosis pathway triggered in response to DNA damage ( 10, 11). However, ARF accumulates in the nucleolus in most human tumor cell lines, whereas p53 and Hdm2 remain largely extranucleolar ( 1, 12– 14); thus, the interaction of these proteins to regulate the DNA damage response would need to involve changes in their subnuclear distribution and would likely require a shift in ARF-binding partners. ARF is known to undergo a variety of interactions within the nucleolus, a subnuclear structure organized around ribosomal gene clusters and involved in ribosome assembly, and there is evidence for a p53-independent role for ARF as an inhibitor of rRNA processing ( 15). Within the nucleolus ARF interacts with 5.8S rRNA ( 15) and with B23(nucleophosmin) ( 16, 17), an abundant nucleolar protein that also binds 5.8S rRNA and regulates ribosome biogenesis and assembly. In the murine system, some 28 additional nucleolar proteins have been identified that interact with ARF ( 17). Although the functions of these nucleolar interactions of ARF are not well understood, they could play a role in determining the subnuclear compartmentalization of ARF and thereby regulate the ability of ARF to stabilize p53.
Recently, it has been observed that nucleolar disruption is associated with treatments that induce p53, including exposure to UV radiation, hypoxia, cisplatin, and other drugs ( 18). This has led to the proposal that the nucleolus could be a central early sensor of a variety of stress signals, such as DNA damage ( 18, 19), and could transmit these signals to the system that regulates p53 levels. A nucleolar protein, such as ARF, which has been observed to relocate to the nucleoplasm following nucleolar disruption ( 20), could serve as a link between the nucleolar stress response and downstream events that activate the p53 pathway.
To further understand the role of ARF in the DNA damage response of human tumor cells, we have used immunofluorescence microscopy to examine the subnuclear distribution of ARF following DNA damage. We provide evidence for a transient p53-independent redistribution of ARF from the nucleolus to the nucleoplasm following DNA damage that involves time-dependent changes in ARF NH2- and COOH-terminal epitope accessibilities. Whereas the redistribution of ARF is not driven by activation of the p53 pathway, the UV sensitization effect of ectopic ARF is dependent on coexpression of wild-type p53. The results have implications for the possible regulatory role of the nucleolus and nucleolar interactions of ARF in the cellular DNA damage response.
Materials and Methods
Cell lines. DU145 prostate cancer cells and H358 lung cancer cells were obtained from the American Type Culture Collection (Rockville, MD). An estrogen receptor–negative clone of MCF-7 breast cancer cells was provided by Dr. Joseph Lustgarten (Sidney Kimmel Cancer Center). DU145 cells coexpress two endogenous mutant p53s and cannot activate p53-responsive genes ( 21). H358 cells lack endogenous expression of p53. Both cell lines express endogenous ARF ( 22). The MCF-7 subclone expresses endogenous wild-type p53 and lacks endogenous ARF ( 22). Cell lines were maintained as described previously ( 23).
Cell viability assays. Cell viability assays were carried out in 96-well plates as described previously ( 22, 23). Briefly, cells grown in medium lacking phenol red were either treated with UV-C radiation using a Stratalinker (Stratagene, La Jolla, CA) and replated in 96-well plates at 10,000 cells per well or treated with adenoviral vectors for 3 hours ( 22) followed 1 day later by treatment with UV-C radiation and replating. Adp14 has been described previously ( 23). Adp53 (INGN201, AdCMVp53) and AdLuc (luciferase, control) were provided by Introgen Therapeutics, Inc. (Houston, TX). Seventy-two hours after replating, viability was determined by the bioconversion of a tetrazolium compound [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt (MTS); Promega Corp., Madison, WI] to a colored formazan product, measured by absorbance at 490 nm ( 23), which is linearly proportional to cell mass over the range of cell densities examined. Viability was expressed as a percentage of initial cell mass.
Apoptosis assay. Cytoplasmic fractions of 4 × 105 cells were collected 48 hours after vector treatment. Apoptosis was measured using a quantitative Nucleosome ELISA kit (Calbiochem, San Diego, CA) according to the instructions provided in the kit.
Immunofluorescence. Primary antibodies (used at 1:100 in PBS + 3% bovine serum albumin) were as follows: monoclonal antibody to the ARF NH2 terminus (clone 14P01, Labvision Neomarker, Fremont, CA), rabbit polyclonal antibody to B23(nucleophosmin) (Santa Cruz Biotechnology, Santa Cruz, CA), and rabbit polyclonal IgG prepared to ARF COOH terminus using a synthetic ARF COOH-terminal peptide to immunize (sequence: AGRARCLGPSARGPG). Secondary antibodies (used at 1:1,000 in PBS + 3% bovine serum albumin) were chicken anti-mouse IgG Alexa Fluor 594 and goat anti-rabbit IgG Alexa Fluor 486 (Molecular Probes, Inc., Eugene, OR). Hoescht 33342 for staining nuclei (used at 2.5 μg/mL) was obtained from Sigma-Aldrich (St. Louis, MO), and ProLong Gold Antifade Reagent was obtained from Molecular Probes. Tissue culture-treated glass coverslips (12 mm, round, Fisher Scientific, Pittsburgh, PA) were sterilized in ethanol, air-dried, and placed one per well in 12-well tissue culture plates. Cells were plated at ∼2.5 × 105 per well (∼60% confluency), allowed to attach overnight, placed in 300 μL DMEM without phenol red (Mediatech, Inc., Herndon, VA), and treated with UV-C using a Stratalinker at various doses or with 100 μg/mL cisplatin (Platinol, Bristol-Myers Squibb, New York, NY) for 1 hour. In some experiments, cells were exposed for 3 hours to adenoviral vectors encoding p53 or ARF as described above, and drug and radiation treatments were carried out 1 day later. At specific times after drug or radiation treatment, coverslips were washed with PBS, fixed with 3.7% formaldehyde in PBS for 10 minutes (room temperature), washed again with PBS, and permeabilized with 0.5% NP40 in PBS for 20 minutes (room temperature). Primary antibody (40 μL) dilution was placed on the coverslip followed by incubation in a humidified chamber for 1 hour (room temperature). Coverslips were washed in PBS and incubated in a humidified chamber for 1 hour with 40 μL of the appropriate secondary antibody. Coverslips were again washed with PBS and excess liquid was blotted onto a Kimwipe. Coverslips were mounted onto slides using ProLong Gold Antifade Reagent. Slides were stored at 4°C and examined the next day using a Deltavision Restoration Deconvolution Microscope (Applied Precision LLC, Issaquah, WA) mounted on an Olympus IX-70 base. A 1.4 numerical aperture, ×40 objective was used to capture all images. Images were deconvolved with the SoftWorx version 2.5 iterative method. For quantitation purposes, nuclear fluorescence intensity was measured for each wavelength, and the ratio of the wavelengths was used to determine epitope availability changes over time.
Western analyses. Cells were lysed in SDS-free lysis buffer [10 mmol/L sodium phosphate (pH 7), 0.15 mol/L NaCl, 1% NP40, 0.5% sodium deoxycholate, 1 mmol/L EDTA, 0.5 mmol/L phenylmethylsulfonyl fluoride, and one tablet per 10 mL complete protease inhibitors (Roche, Nutley, NJ)]. The lysates were incubated for 15 minutes on ice, cleared by centrifugation, electrophoresed on 12.5% SDS-PAGE gels (40 μg for actin analyses or 150 μg for all others), and blotted onto polyvinylidene difluoride membranes (Osmonics, Minnetonka, MN). Western analysis was carried out as described previously ( 22) using the following primary antibodies (at 1:500 dilution): rabbit polyclonal anti-ARF, full-length protein (Zymed Laboratories, Inc., South San Francisco, CA), mouse monoclonal anti–mdm2/Hdm2 SMP14 (Santa Cruz Biotechnology), and rabbit polyclonal anti-B23(nucleophosmin) (Santa Cruz Biotechnology). As a control, the same lysates were evaluated for actin levels (mouse monoclonal anti-actin, 1:500, Santa Cruz Biotechnology). Bands were quantitated using a Kodak digital camera and analysis software.
Coimmunoprecipitation/Western. Cells (5 × 106, two 10-cm culture dishes) were lysed in 500 μL lysis buffer as described above for Western analysis. Equivalent amounts of each lysate (∼2 mg) were incubated either with 40 μL agarose-conjugated anti–mdm2/Hdm2 (epitope corresponding to amino acids 154-167 of the human protein) or 40 μL anti-B23 antibody (Santa Cruz Biotechnology) followed by incubation for 4 hours at 4°C with rotation. B23 immunoprecipitates were incubated an additional hour in the presence of 80 μL protein G agarose (Santa Cruz Biotechnology). Immunoprecipitates were washed twice with buffer containing 0.14 mol/L NaCl, 0.008 mol/L sodium phosphate, 0.002 mol/L potassium phosphate, and 0.01 mol/L KCl (pH 7.4; binding/wash buffer, BupH Modified Dulbecco's PBS, Pierce Biotechnology, Rockford, IL). Hdm2 immunoprecipitates were eluted from the immobilized antibody with 75 μL glycine buffer (pH 2.8; Immunopure IgG Elution Buffer, Pierce Biotechnology) and neutralized by the addition of 2 μL of 1 mol/L Tris (pH 9) followed Western analysis as described above. B23 immunoprecipitates were resuspended directly in electrophoresis sample buffer, boiled, and subjected to Western analysis as described above and in ref. 22. Control immunoprecipitations were carried out similarly using normal rabbit serum or normal mouse IgG conjugated to agarose.
DNA damage/repair assay. A slot blot immunoassay was used to follow DNA damage and repair. DNA was prepared from cells at fixed times after treatment with 40 J/m2 UV-C radiation using DNA purification columns from Qiagen, Inc. (Valencia, CA), adjusted to 500 ng/mL in 10 mmol/L Tris-HCl, 1 mmol/L EDTA, and denatured by boiling for 3 minutes followed by immediate cooling on ice. A Hybri-Slot Manifold was set up using a pre-wet (H2O) nitrocellulose membrane and gentle vacuum according to the manufacturer's instructions. Duplicate samples of DNA (200 μL) for each time point were then collected on the membrane. Membranes were air-dried and microwaved for 2 minutes followed by treatment as for Western blotting. An antibody specific for cyclobutane pyrimidine dimers, the major photoproduct produced by UV-C treatment, was purchased from Kamiya Biomedical Co. (Seattle, WA; anti-thymidine dimer, clone KTM53) and used at 1:40 dilution.
Statistical analyses. Statistical analyses of immunofluorescence results were carried out using the Systat program. A one-way ANOVA followed by a pairwise comparison procedure (Bonferroni t test) was used to evaluate the significance of differences in data sets.
Endogenous and ectopic ARF colocalizes with B23(nucleophosmin) in nucleoli of human tumor cell lines. We carried out immunofluorescence staining on fixed human tumor cells followed by deconvolution microscopy to determine the subcellular distribution of endogenous and ectopic ARF in several human tumor lines before and after DNA damage. Using an antibody specific for the ARF NH2 terminus, we detected a predominantly nucleolar pattern of staining for endogenous ARF in DU145 prostate cancer cells ( Fig. 1A, 1 ), which express mutant p53. Nuclei were stained with the DNA stain Hoescht 33342, and nucleoli were identified by immunostaining for the nucleolar antigen B23(nucleophosmin) ( Fig. 1A, 2), which colocalizes with ARF ( Fig. 1A, 3). Nucleoli were also visualized by differential interference contrast, where they appear as elevated regions within the nucleus ( Fig. 1A, 4). The nucleolar localization of ARF was not affected by ectopic overexpression of wild-type p53 ( Fig. 1A, 5), which results in induction of p53 downstream target genes, p21waf1, Hdm2, and bax ( 22). Similar patterns of nucleolar staining for ARF were observed in p53-null H358 lung cancer cells, although the staining was more heterogeneous from cell to cell ( Fig. 1A, 6) and in MCF-7 breast cancer cells (endogenous ARF-null, endogenous p53 wild-type) that overexpress ectopic ARF ( Fig. 1A, 7). Taken together, these results are consistent with other reports that human and mouse cells lines often display nucleolar localization of ARF ( 1, 7, 13, 24) and with a recent report that a significant proportion of ARF in murine fibroblasts is bound to the abundant nucleolar protein, B23(nucleophosmin) ( 17).
Three hours after irradiation with 80 J/m2 UV radiation, a treatment that cross-links adjacent thymidine nucleotides in DNA, or 24 hours after a 1-hour exposure to 100 μmol/L cisplatin, an agent that forms bifunctional adducts that cross-link adjacent guanidine or guanidine-adenine residues in DNA, the nucleolar pattern of staining disappeared and was replaced by a more diffuse staining pattern over the nuclear and perinuclear region ( Fig. 1B, 1 and 3, respectively). Nucleoli could not be clearly distinguished by differential interference contrast following these treatments ( Fig. 1B, 2 and 4, respectively), suggesting that the loss of nucleolar ARF in this case could be a consequence of nucleolar disruption.
ARF redistributes transiently to the nucleoplasm following DNA damage. We used UV treatment to further explore the time course of ARF redistribution in DU145 cells, where the majority of cells initially display a nucleolar pattern of ARF immunofluorescence. To better visualize the earliest UV-induced effects, before cellular or nucleolar breakdown, we irradiated cells with 40 J/m2 UV, a dose that did reduce cell viability over the 24-hour period we examined. Immunofluorescence of the nucleolar antigen B23(nucleophosmin) was used to identify the nucleolar region. Although treatment with UV can lead to nucleolar breakdown and relocalization of B23(nucleophosmin) to the nucleoplasm ( 18), under the treatment conditions used here, a nucleolar pattern of B23(nucleophosmin) immunofluorescence remained evident over the 24-hour period, indicating that a significant fraction of B23(nucleophosmin) remained nucleolar and nucleoli remained largely intact ( Fig. 2, column 2 ).
Unlike the pattern of B23(nucleophosmin) immunofluorescence, ARF immunofluorescence disappeared from the nucleoli of most cells within the first hours after treatment ( Fig. 2, column 1). Image scaling was set to be identical for all panels, making the diffuse nucleoplasmic ARF visibly undetectable. The fraction of cells with ARF-positive nucleoli reached a minimum at 5 hours post-treatment, after which time the percentage of cells with ARF-positive nucleoli increased and reached pretreatment levels by 24 hours. The disappearance of the nucleolar immunofluorescence pattern of ARF corresponded to the period of the most rapid removal of cyclobutane thymidine dimers from DNA ( Fig. 2B, •). However, ARF did not seem to play a role in the removal of dimers, because their rate of disappearance from genomic DNA was unchanged by overexpression of ectopic ARF ( Fig. 2B, ○), achieved by treatment of cells for 3 hours with 40 plaque-forming units (pfu) per cell of Adp14, 24 hours before UV treatment. Although there were time-dependent changes in the immunofluorescence pattern of ARF, overall levels of ARF protein remained constant over the entire time course as indicated by Western analysis ( Fig. 2C, top). Similarly, overall levels of B23(nucleophosmin) remain largely unchanged over the time course studied ( Fig. 2C, middle), as did actin levels ( Fig. 2C, bottom). These results, together with the results of Fig. 1, suggest that moderate amounts of DNA damage induce a transient redistribution of ARF from the nucleolus to the nucleoplasm in human tumor cells.
Levels of the p53 target gene product and ARF-binding partner, Hdm2, also remained unchanged over the 24-hour time period following irradiation ( Fig. 3C, a ), consistent with the fact that DU145 lacks transcriptionally active p53 ( 21). In addition, treatment of DU145 cells with a p53 adenoviral vector (40 pfu per cell for 3 hours, 24 hours before UV treatment) to induce Hdm2 did not affect the nucleolar localization of ARF ( Fig. 1A, 5), nor did it change the time course of ARF redistribution after UV treatment (data not shown). Thus, the redistribution of ARF to the nucleoplasm after UV treatment is independent of p53 and is not driven by a process that would require induction of Hdm2.
Reciprocal changes in ARF NH2- and COOH-terminal immunofluorescence intensities following DNA damage. To further characterize the nucleolar/nucleoplasmic redistribution of ARF, we compared immunofluorescence patterns and intensities obtained with the ARF NH2-terminal antibody used in Figs. 1 and 2 to that obtained with a rabbit IgG specific for the COOH-terminal 15–amino acid residues of ARF. Both antibodies produced a nucleolar immunofluorescence pattern in untreated cells ( Fig. 3A) that disappeared during the first 5 hours after irradiation and reappeared by 24 hours (data not shown) in a manner similar to that shown in Fig. 2A. However, as shown graphically in Fig. 3B, the time-dependent variation in total nuclear immunofluorescence intensity following irradiation differed for the two antibodies. Nuclear immunofluorescence intensities were evaluated in random fields of 50 cells, a number sufficient to observe statistically significant differences between data sets. Mean intensities are expressed as a percent of the mean intensity observed in unirradiated cells (0 hour). As shown in Fig. 3B (b), the intensity of ARF NH2-terminal immunofluorescence decreased over time after irradiation to a minimum at the 5-hour time point and then increased to the pretreatment level by 24 hours after irradiation. A one-way ANOVA followed by a pairwise comparison procedure (Bonferroni t test) to analyze the significance of differences between the different data sets showed that the mean NH2-terminal immunofluorescence intensities at 3 and 5 hours after irradiation were significantly lower than the 0-hour (pretreatment) and 24-hour mean intensities (P < 0.001). As shown in Fig. 3B (a), the COOH-terminal immunofluorescence intensity varied in a reciprocal manner, increasing just after irradiation and then decreasing to 0-hour (pretreatment) levels by 24 hours after irradiation. Pairwise comparisons of mean intensities at different time points indicated that the 1- and 3-hour mean intensities were significantly higher than the 0-hour (pretreatment) mean intensity (P < 0.001). The ratio of NH2- and COOH-terminal immunofluorescence intensities, evaluated on a cell-by-cell basis in 50 cells, decreased over the first 5 hours after irradiation and increased to the 0-hour (pretreatment) level at 24 hours after irradiation ( Fig. 3B, c). Pairwise comparisons of the 0- or 24-hour values to the 1-, 3-, and 5-hour values revealed highly significant differences in each case (P < 0.001). Because overall ARF protein levels did not change over this time frame ( Fig. 2C), the results point to a transient masking of the NH2-terminal epitope after DNA damage, accompanied by a transient unmasking of a COOH-terminal epitope, perhaps due to changing interactions involving these regions.
We carried out coimmunoprecipitation experiments to evaluate changes in ARF-Hdm2 and ARF-B23(nucleophosmin) interactions following irradiation. As shown in Fig. 3C, c, ARF-Hdm2 complexes are low in unirradiated cells where ARF is nucleolar in localization. ARF-Hdm2 complexes increase >4-fold, 1 hour following irradiation as ARF moves to the nucleoplasm, and they decrease to pretreatment levels by 24 hours following irradiation when ARF has relocalized to the nucleolus. A reciprocal pattern was observed for ARF-B23(nucleophosmin) complexes, which are detectable in unirradiated cells, decrease to undetectable levels by 1 hour after irradiation, and are restored to pretreatment levels by 24 hours after irradiation ( Fig. 3C, e). Hdm2 is expressed at low levels in DU145 cells but could be detected in some nuclei at levels above background, where it remained nucleoplasmic throughout this period and did not import to the nucleolus with ARF ( Fig. 3D). In some nuclei, there was evidence for nucleolar exclusion ( Fig. 3D). The results are consistent with a model in which the disruption of the nucleolar ARF-B23(nucleophosmin) complex following DNA damage–induced redistribution of ARF is accompanied by the formation of a nucleoplasmic ARF-Hdm2 complex, resulting in greater masking of the ARF NH2-terminal epitope. Notably, in murine fibroblasts, ARF has been shown to interact with both mdm2 and B23(nucleophosmin) through its NH2-terminal sequences. However, the interaction with mdm2 seems the stronger of the two interactions, as mdm2 competes out B23(nucleophosmin) for ARF binding in vitro ( 25), possibly indicating that mdm2/Hdm2 interacts over a larger surface of the ARF NH2 terminus and achieves greater masking of the ARF NH2-terminal epitope. The fact that the time course of formation of ARF-Hdm2 complexes does not strictly follow the time course of ARF NH2-terminal masking could suggest that an interaction(s) of the ARF NH2 terminus with a protein(s) other than Hdm2 contributes to the masking of the NH2 terminus as well.
We also carried out coimmunoprecipitation experiments to evaluate B23(nucleophosmin) interactions with Hdm2 over time following radiation. Although B23(nucleophosmin) is a major nucleolar antigen, it cycles between the nucleolus, nucleoplasm, and cytoplasm and has been shown to interact with the NH2- and COOH-terminal regions of Hdm2, forming a complex that contributes to the activation of p53 ( 26). As shown in Fig. 3C, d, B23(nucleophosmin)-Hdm2 complexes are nearly undetectable in untreated cells but increase progressively by some 5-fold at 5 hours after irradiation and begin to decrease at 24 hours after irradiation to ∼3-fold untreated levels. The kinetics of accumulation and disappearance of B23(nucleophosmin)-Hdm2 complexes is delayed compared with that of ARF-Hdm2 complexes, indicating that the two types of complexes are regulated differently and may play different roles following irradiation.
p53-mediated response to UV is enhanced by ARF. To determine how ARF affects cell viability after UV radiation in the presence and absence of p53, we carried out the viability assays shown in Fig. 4 . Cells were treated with adenoviral vectors (multiplicity of infection = 100 for each vector) encoding firefly luciferase (AdLuc, control), wild-type p53 (Adp53), ARF (Adp14), or a combination of Adp53 and Adp14 for 3 hours, replated in triplicate at 104 cells per well in 96-well plates as described in Materials and Methods, and scored for viability by the MTS assay 72 hours after vector treatment. Under the replating densities used, cell viability, expressed as a percentage of initial cell mass, remained unchanged in AdLuc (control) cultures ( Fig. 4A). Cell viabilities (72 hours) in Adp14-treated, Adp53-treated, and Adp14 plus Adp53–treated cells were 56 ± 13%, 70 ± 17%, and 12 ± 1% of initial cell mass, respectively ( Fig. 4A), indicating that cell death occurred. This was confirmed by carrying out an ELISA-based apoptosis assay designed to detect the release of histone-bound DNA fragments (oligonucleosomes) from the nucleus into the cytoplasm during the early phase of apoptosis. As shown in Fig. 4B, oligonucleosomal release measured 48 hours after vector treatment and expressed relative to untreated cells correlated inversely with percent initial cell mass remaining at 72 hours ( Fig. 4A). The results add further support to a growing body of evidence for p53-independent tumor suppression by Adp14 ( 27) as well as for the synergistic effect of overexpressed ectopic ARF and p53, which results from increased p53 stability and increased induction of p53 target genes involved in cell growth arrest and apoptosis ( 23).
In a second experiment, vector-treated cells were exposed 1 day later to either 40 or 160 J/m2 before replating, and viability was scored 48 hours after UV irradiation ( Fig. 4C). Cell viability is expressed as a percentage of each vector treatment's no-UV control to visualize the additional reduction in viability attributable to UV treatment. As shown in Fig. 4C, cells treated with Adp14 or AdLuc displayed a similar 30% loss of viability at the higher UV dose relative to their respective no-UV controls. In contrast, cells treated with Adp53 or with a combination of Adp53 and Adp14 displayed an additional 60% loss of viability at the higher dose of UV. Overall cell killing due to the combined effects of ectopic p53 plus ARF plus UV radiation reached >90% ( Fig. 4C, ▿). These results are consistent with a variety of other published studies that have documented a role for p53 in mediating cell cycle arrest or apoptosis in response to UV and other DNA-damaging treatments ( 28– 30). Taken together, the results show that, in the absence of UV irradiation, overexpressed ARF is able to suppress cell viability and induce apoptosis independently of p53, but ARF does not enhance apoptosis in response to UV treatment unless p53 is also expressed.
We have used immunofluorescence and coimmunoprecipitation analysis to examine the regulation of ARF in human tumor cells following UV radiation–induced and cisplatin-induced DNA damage. Consistent with other published reports ( 1, 12– 14), we observe that ARF colocalizes in the nucleolus of unirradiated tumor cells with the abundant nucleolar protein, B23(nucleophosmin), and we find that ARF-B23(nucleophosmin) complexes can be recovered by coimmunoprecipitation from cell lysates. We observe that DNA damage promotes an immediate redistribution of ARF from a predominantly nucleolar localization to a more diffuse nuclear and perinuclear pattern of immunofluorescence and that this can occur in the absence of generalized nucleolar breakdown. The redistribution correlates with the formation of nucleoplasmic ARF-Hdm2 complexes and with the disruption of nucleolar complexes between ARF and B23(nucleophosmin). Following the peak of the DNA damage response, during which time the majority of thymidine dimers in DNA are repaired, ARF reestablishes its nucleolar localization, and ARF-Hdm2 complexes disappear and are replaced by ARF-B23(nucleophosmin) complexes. The transient redistribution of ARF is independent of p53, as it occurs in DU145 cells that lack expression of endogenous wild-type p53, and it is not driven by changing stoichiometries of the binding partners, ARF, Hdm2, and B23, as intracellular levels of these proteins remain constant throughout the response.
Notably, we observe that Hdm2 remains nucleoplasmic and apparently excluded from nucleoli throughout the DNA damage response possibly due to its low expression levels in DU145 cells. Our observation that ARF can localize in the nucleolus of DU145 cells independently of Hdm2 is consistent with studies in ARF/p53/mdm2–null mouse embryo fibroblasts, where nucleolar localization of ectopic ARF was observed in the absence of mdm2 ( 14). However, the subnuclear distributions of ARF and Hdm2 that we observe before and after DNA damage contrast to distributions observed in normal mouse and human fibroblasts before and after DNA damage ( 11). In the latter case, endogenous ARF and mdm2 (Hdm2) were both found in the nucleoplasm in unstressed cells but were seen to form complexes that accumulate in the nucleolus late in the DNA damage response ( 11). Nucleolar sequestration of the ARF-mdm2 complex has been proposed as a key mechanism whereby ARF inactivates mdm2 ( 14, 31). However, ARF can directly inactivate mdm2 ( 32, 33) and can in some cases suppress growth and stabilize p53 without nucleolar import of mdm2/Hdm2 ( 34, 35). In normal human and mouse fibroblasts, nucleolar accumulation of the complexes after DNA damage was dependent on activation of the p53 pathway or overexpression of ectopic Hdm2 ( 11), suggesting that under conditions of low endogenous ARF or mdm2/Hdm2 expression found in normal cells, high levels of induced or ectopic mdm2/Hdm2 will help drive the complex into the nucleolus. In fact, other studies in murine fibroblasts have shown that, whereas ARF can localize in the nucleolus independently of mdm2, cooperative signals from mdm2 contribute to nucleolar localization of the ARF-mdm2 complex ( 31).
The redistribution of ARF to the nucleoplasm following DNA damage may result from disruptions of nucleolar interactions of ARF that retain ARF in the nucleolus of unstressed cells. Our comparative data on ARF NH2-terminal versus COOH-terminal immunofluorescence show that UV treatment promotes a significant increase in ARF COOH-terminal immunofluorescence as ARF moves out of the nucleolus, possibly reflecting the disruption of a nucleolar complex between the ARF COOH terminus and a putative nucleolar protein “x” that masks this region of ARF in unstressed cells. This raises the possibility that nucleolar interactions, such as the interaction with B23(nucleophosmin) and possibly others involving the ARF COOH terminus, regulate ARF's involvement in the DNA damage response.
The ARF COOH terminus is unusual in that it is encoded by exon 2 sequences that overlap with the p16 tumor suppressor gene but are read in an alternate reading frame ( 12, 36– 38). Although ARF NH2-terminal sequences encoded by ARF's first exon bind mdm2 (Hdm2) and promote the accumulation of p53 when expressed ectopically ( 2), it is now clear that the COOH-terminal region of ARF also contributes to its physical properties ( 39), its nucleolar localization ( 13, 40), and its function. Several mutations in the ARF COOH-terminal region have been identified in melanoma kindred that affect the solubility and nucleolar localization of ARF ( 41), suggesting that interactions involving this region could play a role in the mechanism of melanoma and other cancers. The ARF COOH terminus has been found to interact with and enhance the activity of topoisomerase I ( 42, 43), a nucleolar enzyme that associates with intranucleosomal regions of the 5.8S rRNA genes ( 44). There is therefore growing evidence that the COOH-terminal domain of ARF plays a role within the nucleolus relevant to cancer, and this region may contribute to the regulation of ARF following DNA damage and other forms of cellular stress.
In addition to the interactions with B23(nucleophosmin) and topoisomerase I, ARF undergoes other nucleolar protein-protein interactions whose functions are not well understood ( 17) as well as a specific interaction with 5.8S rRNA ( 15). UV radiation is known to cause a site-specific damage to 28S rRNA that blocks translation ( 45), and it may cause similar damage to 5.8S RNA as well, disrupting the interaction with ARF. In fact, RNase treatment alone can promote the translocation of ARF from the nucleolus to the nucleoplasm ( 46), suggesting that an ARF-RNA interaction may regulate ARF's nucleolar localization. Similarly, cisplatin can damage RNA through the formation of bifunctional adducts similar to those formed on DNA. If an ARF-RNA interaction contributes to the nucleolar localization of ARF, then the accumulation of newly transcribed 5.8S rRNA following repair of UV damage to the 5.8S genes could enable ARF to reestablish its interaction with 5.8S RNA and promote its reentry into the nucleolus.
B23(nucleophosmin) is known to undergo a nucleolar/nucleoplasmic translocation after UV irradiation of cells ( 18) and has recently been reported to stabilize p53 through an interaction with the NH2- and COOH-terminal domains of Hdm2 ( 26). B23(nucleophosmin) may therefore contribute, along with ARF, to the regulation of the p53-mediated response to cellular stress and DNA damage. Importantly, the interaction of B23(nucleophosmin) with Hdm2 is independent of p53 and is reduced in vitro by high levels of p53, suggesting competition between p53 and Hdm2 for binding sites on B23(nucleophosmin) ( 26). Furthermore, whereas B23(nucleophosmin) can bind to p53 and regulate p53 transcriptional activity ( 47, 48), it seems to prefer complex formation with Hdm2 and may be unable to bind to and dissociate preformed p53-Hdm2 complexes ( 26). Taken together, the results suggest a model ( Fig. 5 ) in which DNA damage disrupts a nucleolar interaction involving the ARF NH2 terminus with B23(nucleophosmin) and the ARF COOH terminus with another protein “x.” Disruption of these interactions enables the redistribution of ARF to the nucleoplasm, where it binds Hdm2 and promotes the disruption of preformed p53-Hdm2 complexes during the early phases of the DNA damage response. B23(nucleophosmin) contributes later in the response by binding to the free Hdm2 released from p53 complexes by ARF or to newly synthesized Hdm2 that would accumulate in cells expressing wild-type p53. The early appearance of ARF-Hdm2 complexes that we observe, followed by the delayed appearance and persistence of B23(nucleophosmin)-Hdm2 complexes, is consistent with such a model.
We found that although ectopic overexpression of ARF did not enhance UV-induced apoptosis independently of p53, it did enhance p53-mediated responses to UV, consistent with other results that implicate ARF in the p53-mediated response to DNA damage ( 10, 11). Our results suggest that a disruption of ARF's nucleolar interactions may promote the subnuclear redistribution of ARF needed for efficient cooperation between p53 and ARF to promote tumor cell apoptosis. The results further implicate the nucleolus and nucleolar interactions of ARF as possible p53 upstream regulators of the stress and DNA damage response and suggest that ARF could contribute to the outcome of the p53-mediated responses to a variety of therapies, including cisplatin, which use the p53 pathway, at least in part to achieve their therapeutic effect.
Grant support: California Tobacco-Related Disease Research Program grant 11RT-0074 and National Cancer Institute grant 1R01CA111868-01 (R.A. Gjerset).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Introgen Therapeutics for providing the AdLuc and Adp53 (INGN201, Ad5CMVp53) vectors.
- Received May 25, 2005.
- Revision received August 2, 2005.
- Accepted August 17, 2005.
- ©2005 American Association for Cancer Research.