The Adenoviral E4orf6 Protein Induces Atypical Apoptosis in Response to DNA Damage*

Adenoviral proteins interact with host-cell proteins to either exploit or inhibit cellular functions for the purpose of viral propagation. E4orf6, the 34-kDa gene product of the E4 gene, interacts with the double-strand break repair (DSBR) protein DNA-dependent protein kinase and cooperates with binding partner E1B-55K to degrade MRE11, preventing viral DNA concatemer formation. We previously demonstrated that E4orf6 radiosensitizes human tumor cells through the inhibition of DSBR, notably in the absence of E1B-55K. Here, we report that E4orf6 prolongs the signaling of DNA damage by inhibiting the activity of protein phosphatase 2A (PP2A), the phosphatase responsible for dephosphorylating γH2AX. The inhibition of PP2A occurs without significant disruption of the DNA re-ligation rate. Prolonged signaling of DNA damage in the presence of E4orf6 initiates caspase-dependent and independent cell death. This is accompanied by poly(ADP-ribose) polymerase (PARP) hyperactivation and the translocation of apoptosis-inducing factor (AIF) from the mitochondria to the nucleus. Knockdown of AIF by shRNA rescues the radiosensitization induced by E4orf6. Taken together, these data suggest that E4orf6 disrupts cellular DSBR signaling by inhibiting PP2A, leading to prolonged H2AX phosphorylation, hyperactivation of PARP, and AIF translocation to the nucleus. The function of E4orf6 as an inhibitor of PP2A and activator of PARP in the absence of other adenoviral gene products is of importance in delineating the adenovirus-host cell interplay.

replication imposed by cellular DNA repair mechanisms (3), and V(D)J recombination, a cellular mechanism employed for the generation of antigenic diversity within the immunoglobulin genes. Both concatemer formation and V(D)J recombination depend on the non-homologous end-joining pathway (4 -8) of DNA double-strand break repair (DSBR), 2 and require proteins such as DNA-dependent protein-kinase (DNA-PK), DNA ligase IV, and MRE11. These same DSBR proteins are also required for the repair of DNA double-strand breaks (DSBs) following exposure to radiation. The primary mechanism of cytotoxicity from ionizing radiation (IR) is the induction of DNA damage and DSBs, which are known to be the most lethal of DNA lesions (9 -12). In addition, an increased capacity of DSBR has been identified as contributing significantly to the radioresistance of gliomas (13). Due to these observations, the inhibition of DNA DSBR is an attractive method for radiosensitization; therefore, we recognized the ability of E4orf6 to interfere with cellular DSBR as a great opportunity to exploit E4orf6 for the purpose of radiosensitizing radioresistant tumor cells.
We have previously demonstrated that E4orf6, independent of E1B-55K and therefore independent of p53 and MRE11 degradation, is capable of significantly radiosensitizing tumor cells by inhibiting cellular DSBR in response to radiation-induced DSBs (14). In E4orf6-expressing cells, we found prolonged levels of H2AX phosphorylation at Ser-139 (␥H2AX) and DNA-PK autophosphorylation at Thr-2609 at 360 min postirradiation, a time when DNA repair should be complete and the repair proteins should be dephosphorylated. However, it was unclear as to whether E4orf6 inhibited the physical re-ligation of the dsDNA breaks or inhibited a step downstream of re-ligation and how this DNA damage signal was translated into increased radiosensitivity. Therefore, we sought to determine the reason for prolonged signaling of damage, as well as the cell death pathway responsible for the radiosensitization. Here, we report our findings on the atypical mechanism of cell death induced by E4orf6 in the presence of DNA damage. In irradiated E4orf6-expressing cells, poly(ADP-ribose) polymerase (PARP) becomes hyperactivated and apoptosis-inducing factor (AIF) translocates from the mitochondria to the nucleus, inducing cell death. The ability to radiosensitize tumor cells through expression of a single gene makes E4orf6 a promising genetic tool to be used in conjunction with standard radiation therapy in the treatment of radioresistant tumors, such as glioblastoma multiforme (GBM).

EXPERIMENTAL PROCEDURES
Cell Lines, Irradiations, Infections, and Reagents-U251 cells (American Type Culture Collection, Manassas, VA) were incubated at 37°C with 5% CO 2 and maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, L-glutamine, and penicillin/streptomycin. All irradiations were carried out with a 137 Cs irradiator at a dose rate of 4.0 Gy (gray)/min. All chemicals were purchased from Sigma unless otherwise noted. Infections were performed as previously described (14).
Pulsed-field Gel Electrophoresis-PFGE was performed using the CHEF-DR II system (Bio-Rad). U251 cells, infected with either null or E4orf6 virus, were irradiated at 40 Gy and harvested by trypsinization. Cells (1.5 ϫ 10 6 ) were resuspended in phosphate-buffered saline, mixed with an equal volume of 2.0% agarose, and embedded in plugs. Solidified plugs were immersed in lysis buffer (100 mM EDTA, pH 8.0, 10 mM Tris-Cl, pH 8.0, 1% w/v N-lauroylsarcosine sodium salt) with freshly added 100 g/ml proteinase K and incubated at 55°C overnight. Lysed plugs were rinsed with storage buffer (10 mM Tris-Cl, pH 8.0, 10 mM EDTA, pH 8.0) and stored in storage buffer at 4°C. For electrophoresis, plugs were cut into thirds and loaded into the wells of a 1.0% agarose gel (in 0.5ϫ TBE buffer (44.5 mM Tris/boric acid, 1 mM EDTA)). Electrophoresis was carried out at 12°C in 0.5X TBE with parameters of 66 V (2.0 V/cm), linearly increasing (50 -5000 s) pulse times, and a duration of 48 h. After electrophoresis, the gels were incubated for 20 min in ethidium bromide (1.0 g/ml in 0.5ϫ TBE), destained for 20 min in 0.5ϫ TBE, and imaged and quantified on a Typhoon 9210 variable mode imager (GE Healthcare). All values were normalized to their respective samples at 40 Gy without repair time (total induced damage) and shown as fold decrease in damage over time.
Protein Phosphatase 2A (PP2A) Activity Assay-PP2A activity was measured according to manufacturer's protocol (catalog 17-313, Upstate Biotechnology, Lake Placid, NY). Approximately 3 ϫ 10 6 cells were directly lysed by sonication in imidazole buffer containing complete mini mixture (Roche Applied Science), pepstatin, aprotinin, leupeptin and phenylmethylsulfonyl fluoride. Each sample (500 g) was subjected to immunoprecipitation with PP2A antibody (catalog 05-421). As outlined in the protocol, beads containing precipitated PP2A were added to a phosphatase reaction with threonine phosphopeptide in a shaking incubator. Samples were then aliquoted into three wells of a 96-well plate, into which malachite green detection solution was added. Plates were incubated for 15 min at room temperature and then read at 650 nm on an automated plate reader (Molecular Devices, Sunnyvale, CA).
AIF Knockdown-Non-selectable pKD-NegCon-v1 (negative control shRNA) and pKD-AIF-v3 (AIF shRNA) plasmids were purchased from Upstate Biotechnology. The plasmids were digested with NotI and EcoRI restriction enzymes (Promega, Madison, WI) for the shRNA to be subcloned into pSilencer 4.1-CMV puro vector (Ambion, Austin, TX) to obtain the puromycin resistance gene while maintaining the original H1 RNAP III promoter from the Upstate vectors. The reengineered plasmids were used in transfections with U251 cells and Lipofectamine PLUS reagent (Invitrogen). At 48 h post-transfection the U251 cells were treated with 3 g/ml puromycin for 48 h, a time point when mock transfected cells were dead. The surviving NegCon and AIF cells were then plated at limited dilution for harvesting of individual clones. Selected cells were maintained in 0.5 g/ml puromycin, and the addition of 100 M non-essential amino acids and 55 M 2-mercaptoethanol aided in their growth.
Clonogenic Survival Assay-To analyze clonogenicity in infected and irradiated cells, a modified clonogenic assay was used, due to the sensitivity of the infected cells to the combination of irradiation and replating. U251 clones transfected with negative control or AIF-targeting shRNA were plated in 35-mm plates in triplicate at equal density, infected the next day at an MOI predetermined to yield maximal infectivity, and irradiated 48 h later with a 137 Cs source (dose rate of 4.0 Gy/min) at a range of IR doses. When unirradiated control plates were near confluence (at or near day 7 post-IR), the cells were fixed and stained with crystal violet as previously described (15). Crystal violet was solubilized in 33% acetic acid and the absorbance at 540 nm was measured in triplicate for each well as described by Bernardi et al. (16).

E4orf6 Does Not Inhibit the Re-ligation of DSBs-
We previously demonstrated that E4orf6 radiosensitizes human tumor cells by inhibiting DSBR as measured by prolonged ␥H2AX and Thr-2609-phosphorylated DNA-PK levels and by sublethal damage repair assay in U251 and RKO tumor cells, respectively (14). Because detection of DSBs (H2AX phosphorylation) and DSB repair complex formation (DNA-PK autophosphorylation) were not inhibited by E4orf6, we hypothesized that E4orf6 was interfering with the late stages of repair resulting in the prolonged signaling of damage at times when complete repair would result in dephosphorylation of H2AX and DNA-PK. However, it remained possible that the physical re-ligation of DSBs is inhibited by E4orf6, perhaps through interference with DNA LigaseIV, the enzyme responsible for ligating breaks during NHEJ. To determine whether E4orf6 inhibits re-ligation of the DSBs, we employed the PFGE method to quantify the DNA DSBs over time based on the ability of damaged DNA to migrate through an agarose gel. U251 cells were infected with null (control) or E4orf6-expressing adenoviral vectors (previously optimized (14)). The infected cells were irradiated at 48 h post-infection and harvested at varying times post-irradiation. As measured by PFGE, U251 cells expressing E4orf6 repaired DSBs at a rate similar to those infected with control virus (Fig.  1). There was a trend of E4orf6 attenuating repair at early times post-irradiation; however, this trend was not significant and levels of damage for E4orf6 virus-and null virus-infected U251 cells were overlapping at 360 min post-irradiation (data not shown). In contrast, treatment with the phosphatidylinositol 3-kinase inhibitor wortmannin, which effectively blocks DSBR by inhibiting DNA-PK prior to ligase activation, resulted in the persistence of significant levels of damaged DNA at 120 min post-irradiation ( Fig. 1) and beyond (data not shown).
E4orf6 Inhibits PP2A-Based on the similar rates of re-ligation in null virus-and E4orf6 virus-infected U251 cells, we hypothesized that E4orf6 must inhibit a post-ligation step involving ␥H2AX dephosphorylation. It was recently demonstrated that PP2A is responsible for the dephosphorylation of ␥H2AX following repair of radiation-induced DSBs (17). Therefore, we sought to determine whether the activity of PP2A was altered by E4orf6, resulting in the prolonged phosphorylation of H2AX, by performing in vitro phosphatase reactions with lysates from infected U251 cells. Whereas radiation treatment led to an increase in PP2A activity in null virus-infected U251 cells, PP2A activity in response to IR was significantly inhibited in cells expressing E4orf6 (Fig. 2). The well known inhibitor of PP2A okadaic acid completely inhibited PP2A activity, and no PP2A activity was recovered by a negative control IgG antibody. These data are consistent with the pro-longed H2AX phosphorylation in the presence of E4orf6 (14) and suggest that the inhibition of PP2A activity contributed to the increased sensitivity of E4orf6-expressing cells to IR.
Induction of PARP Activation and PARP-induced Cell Death in Irradiated E4orf6-expressing U251 Cells-Despite the almost complete re-ligation of DSBs, E4orf6 significantly sensitizes cells to die in response to IR. Thus, we hypothesized that the prolonged signaling of damage initiates a cell death program in E4orf6-expressing cells. Time-lapse microscopy suggested that a significant portion of infected cells undergo apoptotic death, observed by the formation of membrane blebs and spikes in the days following irradiation (data not shown). Representative phase microscopy images shown in Fig. 3A demonstrate similar amounts of rounded cells between null virus-and E4orf6 virusinfected U251 populations 48 h following IR. We decided to further characterize apoptosis in these U251 cells. Since caspase-3 is a terminal caspase activated by various apoptotic stimuli, we analyzed caspase-3 cleavage in response to infection and irradiation. The level of cleaved caspase-3 beginning at 48 h post-irradiation were similar to that induced by treatment with staurosporine but did not differ significantly between null virus-infected and E4orf6 virus-infected U251 cells (Fig. 3B); therefore, caspase-dependent cell death is unlikely to be responsible for the significant radiosensitization induced by E4orf6. One well characterized caspase-independent mechanism of DNA damage-induced cell death is via PARP hyperactivation. PARP is activated by nicked DNA and responds by covalently modifying signaling and repair proteins with the addition of poly(ADP-ribose) (PAR) in polymers of varying length. PARP is necessary for cell survival under conditions of mild DNA damage (18,19); however, in the presence of excessive DNA damage (or unrepaired DNA damage) hyperactivation of PARP leads to induction of a caspase-independent cell death program (20). To determine the PARP response in null FIGURE 2. E4orf6 inhibits the activity of PP2A in response to IR. In vitro PP2A activity assays were performed with lysates from U251 cells infected with either null or E4orf6-expressing virus. The experiment was repeated three times, all values were normalized to 0 Gy, and bars representing the standard error are shown (top). Treatment with treatment okadaic acid, a potent and specific inhibitor of PP2A, was used as a positive control for PP2A inhibition. Immunoprecipitation with a IgG antibody was used as a negative control for activity. Immunoblotting was performed to confirm equal immunoprecipitation of PP2A between experimental samples (bottom).  Fig. 4A), a time when the majority of dsDNA breaks have been repaired (Fig. 1). However, in irradiated E4orf6-expressing cells, there was a significant and continual increase in PAR-modified proteins (black asterisks, Fig. 4A). Similar results were obtained in three independent experiments. These results mirror the effects of E4orf6 on prolonged H2AX phosphorylation and DNA-PK autophosphorylation following IR and suggest that this difference in the PARP-dependent cellular response to IR between null virus-infected and E4orf6 virus-infected cells could be responsible for the radiosensitization by E4orf6. To determine whether the PARP activation plays a role in cell death induction in the presence of E4orf6, we decided to analyze AIF, a key effector molecule downstream of PARP hyperactivation.

virus-infected and E4orf6 virus-infected cells, we analyzed the presence of PAR-modified proteins with an antibody recognizing PAR polymers. We observed PAR modified proteins under control conditions and an induction in newly ribosylated proteins within 30 min of radiation treatment. In null virus-infected U251 cells this increase in ribosylation returned to baseline levels within 2 h post-IR (black arrows,
AIF is a mitochondrial resident protein and has been identified as a downstream modulator of PARP activity (21). PARP hyperactivation in the presence of massive DNA damage leads to disruption of the mitochondrial membrane potential, resulting in the release of AIF from the outer mitochondrial membrane (22). Once released from the mitochondria, AIF translocates to the nucleus and aids in events characteristic of apoptosis, including chromatin condensation, high molecular weight DNA fragmentation requiring endonuclease G, and phosphatidylserine exposure (23). We measured AIF translocation to the nucleus by subcellular fractionation of nuclei and immunoblotting. The nuclear levels of AIF were elevated in E4orf6-expressing U251 cells, compared with null virus-infected cells in response to IR (Fig. 4B). Staurosporine (STS), a chemical inducer of apoptosis by both AIF-dependent and caspase-3 dependent means, also induced substantial AIF nuclear translocation. Therefore, the higher levels of AIF released from the mitochondria in response to IR in the E4orf6expressing cells could mediate all, or part of, the observed radiosensitization induced by E4orf6.
E4orf6-induced Radiosensitization Is AIF-dependent-To determine whether radiosensitization by E4orf6 is dependent upon AIF translocation, we established U251 clones stably expressing shRNA against AIF or a negative control shRNA that does not target any known proteins. Fig. 5A shows the level of AIF knockdown in a mixed population and an individual clone, as well as the normal levels of AIF in control shRNAexpressing cells. AIF shRNA induced a decrease in AIF protein of ϳ83% in clone AIF.1 compared with the control clone. When the individual clones were analyzed for clonogenic survival following IR, E4orf6 radiosensitized the control shRNA clone to a similar degree as untransfected U251 cells. However, the cells expressing AIF shRNA were significantly more radioresistant (Fig. 5B). This indicates that AIF-dependent cell death significantly contributes to the increase in radiation-induced cell death in the presence of E4orf6. However, AIF knockdown cells  retain some ability to undergo caspase-3 cleavage in response to IR (Fig. 5C). This could be due to the incomplete knockdown of AIF protein and/or the presence of AIF-independent caspase activation in the presence of E4orf6. Interestingly, in the absence of AIF, control infected U251 cells are deficient in IRinduced caspase-3 cleavage but are only modestly more resistant to IR (data not shown) suggesting that caspase-3 cleavage is not a significant contributor to radiation-induced cell death in the absence of E4orf6.

DISCUSSION
Tumor therapy regimens of the most aggressive cancers, such as GBM have become multi-modality by including surgery, chemotherapy, and radiation therapy. Clinically, it is possible to deliver localized and very large doses of IR by stereotactic gamma knife radiosurgery; however, a proportion of GBM tumor cells will survive these high doses of IR and continue to proliferate. There is a need for novel therapeutics, and it is ideal to explore radiosensitizers, as the survival time for patients with GBMs and other radioresistant tumors increases significantly from the addition of therapeutics that potentiate the radiation treatments (24).
We have previously identified E4orf6 as a potent radiosensitizer of glioblastoma cells in vitro (14). Here, we further define the action of E4orf6 in the presence of IR and report an atypical mechanism of cell death by which E4orf6 induces tumor cell radiosensitization. We demonstrate that E4orf6 has the ability to inhibit DSBR by prolonging signaling from DSBs, despite the physical repair of DNA damage. These data suggest that E4orf6 does not significantly inhibit the recruitment or activity of DNA Ligase IV. Because our previous data showed that E4orf6 inhibits the dephosphorylation of DSBR complex proteins H2AX and DNA-PK, we conclude that E4orf6 interferes with the very late stages of repair, after re-ligation of the break but before complex dissociation and dephosphorylation. These late, postligation events of repair have been established to be as important as physical re-ligation in conferring radioresistance. The repair complex proteins must undergo a conformational change and dissociate to signal complete repair (25,26). This finding highlights the sensitive nature with which cells react to prolonged signaling of damage, despite the physical repair of damage. In the case of E4orf6, the DSBs resulting from IR are repaired at a similar rate to control cells, but PP2A activity is compromised by E4orf6 rendering ␥H2AX unable to be dephosphorylated, and therefore, the cells respond as if significant levels of unrepaired damage remain by inducing cell death.
We also report that PARP-induced cell death is a consequence of the prolonged DSB signaling. To our knowledge, this is the first demonstration that an adenoviral protein elicits PARP activation in response to cellular stress, such as DNA damage. As a well defined downstream mediator of PARP activation, AIF translocation contributes a significant amount of cell death in E4orf6-expressing cells exposed to radiation. The rescue of radiosensitization in AIF knockdown cells was incomplete (93%). Although not significant, this trend could be attributed to the incomplete AIF knockdown or the possibility of an AIF-independent pathway contributing to the cell death induced by E4orf6. AIF-independent caspase-3 cleavage is a likely component responsible for this alternate pathway. Caspase-3 cleavage induced by IR is decreased but still present in AIF knockdown cells, suggesting that caspase-3 cleavage is induced in cells expressing E4orf6 but not a major contributor to radiosensitization. Therefore, the caspase-3 cleavage observed in AIF-expressing cells was partially AIF-independent, a possibility that is consistent with published reports of both AIF-dependent and independent caspase activation in response to mitochondrial dysfunction (20,21). In control-infected cells, the contribution of mitotic cell death and cell cycle arrest are likely the major contributors to decreased clonogenicity in response to IR, which is supported by the nearly complete lack of caspase-3 cleavage in the absence of AIF with insignificant effects on radioresistance. In extrapolating from these studies, the possibility exists that PARP-induced AIFdependent cell death is a more general response to the presence of unrepaired DSBs, a possibility that warrants further investigation.
How does prolonged or excessive PARP activation lead to AIF release? Although the precise mechanism remains unclear, data support a model in which the depletion of cellular energy FIGURE 5. E4orf6-induced radiosensitization is AIF-dependent. A, U251 cells transfected with shRNA against AIF, or a negative control shRNA, were analyzed for AIF protein levels in both mixed populations and individual clones. AIF clone 1 shows ϳ83% (normalized for loading with actin) compared with negative (Neg) clone 3. B, shRNA clones were analyzed for radiosensitization by E4orf6 in a modified clonogenic assay. C, immunoblot identifying caspase-3 cleavage in shRNA clones 72 h after 6 Gy IR or following STS (0.5 M, 9 h) treatment with or without the pan-caspase inhibitor, benzyloxycarbonyl-VAD-fluoromethyl ketone (50 M).
from the large scale addition of PAR polymers to cellular proteins (27) leads to a collapse of the mitochondrial membrane potential and AIF release. We attempted to determine whether addition of pyruvate or NAD ϩ , having been shown to restore cellular energy pools (28), could inhibit AIF translocation in E4orf6-expressing and irradiated cells. However, these attempts were not successful. This is likely due to the fact that in the studies in which pyruvate and NAD ϩ were used, the stress was acute, while in our experimental system, E4orf6 causes a prolonged, modest level of sustained PARP hyperactivation.
The rate of repair is important in determining radioresistance or radiosensitivity, as much as the levels of damage remaining at 360 min post-irradiation. With that being said, however, the level of radiosensitization achieved with Wortmannin treatment (10 M) is about one log at 8 Gy (29), similar to the one log of sensitization at 8 Gy achieved by E4orf6 (14); therefore, a discrepancy exists between the DSB re-ligation rate and radiosensitization when E4orf6 is compared with a chemical inhibitor of DSBR. Thus, the high level of radiosensitization by E4orf6 cannot be explained alone by the slight alteration in the rate of early repair. Rather, as suggested by our data, this can be explained by the signaling effects of ␥H2AX and PARP at late times post-irradiation when damage levels are similar between E4orf6 and control-infected cell populations.
Another noteworthy finding it that there is an apparent delay between the prolonged ␥H2AX and PARP signaling at 360 min post-IR and the onset of morphological cell death, beginning at 48 h post-IR as observed with time-lapse microscopy (data not shown). It remains possible that the early inhibition of DSBR, as witnessed by prolonged ␥H2AX levels in E4orf6-expressing cells, is propagated by cellular replication, which was observed prior to cell death in the time-lapse videos. Both null virusinfected and E4orf6 virus-infected cells continued to proliferate following exposure to IR, suggesting defective cell cycle control, likely attributed to the presence of mutated p53 in U251 cells (30); however, it is the cells expressing E4orf6 and prolonged ␥H2AX signaling that propagate the DNA damage response resulting in a significant increase in death, in contrast to control, repair-proficient cells that proliferate without amplifying DNA damage. Another explanation for a delay in the onset of death is the possibility that the E4orf6-expressing cells require more than 360 min for the PARP activation to translate into a death signal, such as from the depletion of cellular energy. This is supported by published data confirming the role of PARP in DNA repair and survival at low doses of IR but inducing cell death at higher levels of damage (18 -20) and our data revealing the translocation of AIF to the nucleus at 48 h post-radiation. This is consistent with reports suggesting the almost immediate onset of death following AIF translocation. We therefore propose a model by which aberrant ␥H2AX signaling is sufficient to induce PARP hyperactivation and eventual AIF translocation to the nucleus, resulting in cell death in the presence of E4orf6 (Fig. 6). Collectively, our results point to an atypical apoptotic response as a significant contributor to E4orf6-induced radiosensitization.