The Arf/p53 Protein Module, Which Induces Apoptosis, Down-regulates Histone H2AX to Allow Normal Cells to Survive in the Presence of Anti-cancer Drugs*

Background: It is unclear how DNA-damaging agents target cancer cells over normal somatic cells. Results: Arf/p53-dependent down-regulation of H2AX enables normal cells to survive after DNA damage. Conclusion: Transformed cells, which harbor mutations in either Arf or p53, are more sensitive to DNA-damaging agents. Significance: Cellular transformation renders cells more susceptible to some DNA-damaging agents. Anti-cancer drugs generally target cancer cells rather than normal somatic cells. However, the factors that determine this differential sensitivity are poorly understood. Here we show that Arf/p53-dependent down-regulation of H2AX induced the selective survival of normal cells after drug treatment, resulting in the preferential targeting of cancer cells. Treatment with camptothecin, a topoisomerase I inhibitor, caused normal cells to down-regulate H2AX and become quiescent, a process mediated by both Arf and p53. In contrast, transformed cells that harbor mutations in either Arf or p53 do not down-regulate H2AX and are more sensitive to drugs unless they have developed drug resistance. Such transformation-associated changes in H2AX expression rendered cancer cells more susceptible to drug-induced damage (by two orders of magnitude). Thus, the expression of H2AX and γH2AX (phosphorylated form of H2AX at Ser-139) is a critical factor that determines drug sensitivity and should be considered when administering chemotherapy.

Anti-cancer drugs generally target cancer cells rather than normal somatic cells. However, the factors that determine this differential sensitivity are poorly understood. Here we show that Arf/p53-dependent down-regulation of H2AX induced the selective survival of normal cells after drug treatment, resulting in the preferential targeting of cancer cells. Treatment with camptothecin, a topoisomerase I inhibitor, caused normal cells to down-regulate H2AX and become quiescent, a process mediated by both Arf and p53. In contrast, transformed cells that harbor mutations in either Arf or p53 do not down-regulate H2AX and are more sensitive to drugs unless they have developed drug resistance. Such transformation-associated changes in H2AX expression rendered cancer cells more susceptible to drug-induced damage (by two orders of magnitude). Thus, the expression of H2AX and ␥H2AX (phosphorylated form of H2AX at Ser-139) is a critical factor that determines drug sensitivity and should be considered when administering chemotherapy.
Cancer chemotherapy drugs usually act by inducing cancer cell death. Although some drugs were specifically developed for targeted cancer chemotherapy (1,2), most anti-cancer drugs used today are DNA-damaging agents that induce cancer cell death by interfering with checkpoint responses (3)(4)(5)(6)(7)(8)(9). Because these anti-cancer drugs are administered without a specialized delivery system, they often cause side effects. However, cancer cells are still preferentially targeted over normal somatic cells. This raises the fundamental question of why cancer cells are more sensitive to drug-induced damage than normal cells. Although it is generally acknowledged that the higher growth rate of cancer cells contributes to their drug sensitivity, it might also be reasonable to consider other, as yet unidentified, mechanisms that may underlie preferential cancer cell killing.
p53 is involved in apoptosis induction. Therefore, cancer cells harboring intact p53 are usually more sensitive to anticancer drugs than cells harboring mutated p53 (10 -13). However, anti-cancer drugs preferentially kill cancer cells harboring mutations in the Arf/p53 protein module rather than normal somatic cells that possess intact Arf and p53. This is paradoxical and poses many questions, such as how can normal cells survive drug treatment without undergoing p53-mediated cell death?
A recent study showed that the regulation of histone H2AX, which is responsible for efficient DNA damage checkpoint responses, is altered after cellular transformation. This is because down-regulation of H2AX is dependent on regulation by the Arf/p53 protein module, which is widely mutated in transformed cells (14). This suggests a potential mechanism underlying the transformation-coupled alterations in the damage checkpoint response and the resulting sensitivity of cancer cells to anti-cancer drugs. It also poses the following hypothesis. Normal cells survive treatment with anti-cancer drugs because H2AX is down-regulated in an Arf/p53-dependent manner, resulting in impaired DNA checkpoint responses, whereas transformed cells are killed. If this hypothesis is correct, the critical question is whether the sensitivity to DNA damaging drugs changes after cellular transformation involving muta-tions in the Arf/p53 protein module and the subsequent loss of H2AX regulation.
This study showed that Arf/p53-dependent down-regulation of H2AX was abrogated in transformed cells in the presence of anti-cancer drugs, resulting in an increase in the selective killing of transformed cells by two orders of magnitude. These results highlight a novel mechanism underlying the effects of anti-cancer drugs.
Analysis of DNA Damage Induction and Cell Death-DNA damage was induced by camptothecin (CPT), doxorubicin, cisplatin, or hydroxyurea (HU) (Sigma). Survival rates were determined by counting the number of viable cells after 6 days of CPT treatment (experiments shown in Figs. [1][2][3][4][5] or by counting the number of colonies formed after 1-week release from CPT in the presence or absence of PJ34 (ALEXIS) (experiments shown in Fig. 7). The effects of transient H2AX knockdown and overexpression were determined after 2 days of CPT treatment.

RESULTS
Unlike Immortalized Cells, Normal Cells Survive in the Presence of CPT by Down-regulating H2AX-MEFs become immortal because of genomic instability (19) and mutations in the Arf/p53 protein module (either Arf or p53) (21), processes similar to those that occur during cancer development (22). Therefore, to test our hypothesis that cellular transformation results in changes in drug sensitivity, we examined the response of MEFs to CPT both before and after immortalization. CPT inhibits topoisomerase I and causes DNA replication stress in a manner similar to that induced by the anti-cancer drugs topotecan and irinotecan (23,24). Although CPT killed the majority of immortalized MEFs, most primary MEFs survived, showing a flattened and enlarged morphology (Fig. 1, A and B). Sensitivity to CPT increased 100-fold after immortalization (Fig. 1A). Before undergoing cell death, immortalized MEFs arrested at G 2 and showed increased SA-␤-galactosidase activity ( Fig. 1, C and D). In contrast, damaged primary MEFs showed weaker SA-␤-galactosidase activity and did not arrest at a specific point in the cell cycle. Thus, treatment with CPT caused primary MEFs to undergo growth arrest and to adopt a flattened and enlarged morphology. However, the cells did not enter a canonical senescent state.
To examine changes in the DNA damage checkpoint response, we compared the damage responses of primary and immortalized MEFs after CPT treatment. The critical difference between primary and immortalized MEFs appeared to be the responses of checkpoint factors, as illustrated by the increased ␥H2AX signal, p53 accumulation, and the increased levels of phosphorylated p53 (Ser-18 in mice) in immortalized cells (Fig. 1E). Unlike in immortal MEFs, checkpoint factors were rarely activated in primary MEFs. These differences were associated with the levels of H2AX expression. Consistent with the increased checkpoint responses, high levels of H2AX accumulated in immortalized MEFs (Fig. 1E). However, the levels decreased with the appearance of cleaved Parp1 (Fig. 1F). These findings are consistent with apoptosis induction. In contrast, H2AX levels in primary MEFs increased transiently without activating the checkpoint response but, subsequently, decreased again (Fig. 1F), and the cells became growth-arrested and showed a flattened and enlarged morphology (B). This suggests the onset of quiescence in response to the down-regulation of H2AX. Thus, normal cells become quiescent upon down-regulation of H2AX and are able to survive the effects of CPT. This quiescence-associated survival in the presence of CPT is lost upon immortalization.
Cisplatin and Doxorubicin Do Not Preferentially Target Normal Cells, but damaged Cells Accumulate ␥H2AX and Die-Unlike CPT, primary MEFs did not survive in the presence of doxorubicin or cisplatin (Fig. 2, A and B). In fact, they were more sensitive to these drugs than their immortalized counterparts. FACS analysis was not able to clarify the difference between primary and immortalized MEFs after treatment with cisplatin and doxorubicin. In both cell types, doxorubicin induced G 2 arrest ( Fig. 2A) but cisplatin did not (B). Importantly, sensitivity to both of these drugs was associated with H2AX accumulation and increased ␥H2AX expression. Thus, the regulation of cellular H2AX levels is affected by DNA damage, which, importantly, influences cell fate. Cells that downregulate H2AX become quiescent and are able to survive in the presence of CPT, whereas cells that accumulate H2AX and increase their expression of ␥H2AX are killed.
Normal Cells Survive Treatment with Drugs That Cause DNA Replication Stress-As with CPT, down-regulation of H2AX and the selective survival of normal cells were observed upon treatment with HU, which induces DNA replication stress by depleting the cellular dNTP pool (Fig. 2C). This implies that H2AX is often down-regulated in response to DNA replication stress-associated damage, conferring a "survival" phenotype upon normal cells. In fact, the major damage caused by CPT occurs in association with DNA replication stress (24 -26). Similar to CPT, HU caused G 2 phase arrest in immortalized MEFs but not in primary MEFs. These results indicate that normal cells survive in the presence of drugs that induce DNA replication stress by down-regulating H2AX, a process that is regulated by the Arf/p53 protein module. Therefore, cells become sensitive to these drugs after immortalization.
Cells Harboring Mutated Arf or p53 Are Sensitive to CPT-To address whether changes in sensitivity to CPT are the result of mutations in the Arf/p53 protein module, we examined sensitivity to CPT and changes in H2AX expression in Arf and p53 KO MEFs (both primary and immortal). As expected, neither Arf nor p53 KO MEFs showed changes in drug sensitivity as they became immortal. Indeed, they were as sensitive as immortalized WT MEFs ( Fig. 3A and supplemental Fig. S2). In addition, both primary and immortal Arf and p53 KO MEFs showed increased accumulation of H2AX and increased expression of ␥H2AX and cleaved-Parp1 (Fig. 3, B and C). The cells arrested in G 2 phase (mainly those harboring Arf mutations) and/or showed the appearance of an 8N chromosome peak (mainly those harboring p53 mutations), which often indicates cell death because of mitotic catastrophe (Fig. 3D). These findings indicate that the survival of normal cells in the presence of CPT is dependent on both Arf and p53. Therefore, the survival of normal cells that have down-regulated their expression of H2AX is abrogated by immortalization because of the associated mutations in the Arf/p53 protein module (Fig. 3E).
Because p53 mediates the induction of cell death, Arf KO MEFs with WT p53 are more sensitive to CPT than cells without p53 (Fig. 3A). However, the difference is much smaller than that observed between normal primary MEFs and MEFs without a functional Arf/p53 protein module (Figs. 1A and 3A).
Cancer Cells Are Preferentially Killed unless They Have Acquired Resistance-A key question is whether alterations in H2AX regulation/expression lead to the preferential killing of cancer cells. To investigate this issue, we examined the killing efficiency of CPT in several types of human cancer cells harboring mutations in the Arf/p53 protein module (supplemental Table S1) and compared it with that in NHFs and human mammary epithelial cells (Fig. 4, A and B, and supplemental Fig. S3,  A and B). After exposure to CPT, both NHFs and human mam-  Fig. S3, C and D), and most of the cells died. However, BT474 cells (which are resistant to the drug) were not killed. We also observed the selective killing of cancer cells when studying a mixed culture containing HCT116 cells (a human colon cancer cell line) and primary WT MEFs (Fig. 4D). Thus, cancer cells harboring mutations in the Arf/p53 protein module are sensitive to CPT unless they have acquired resistance.
Although it is still not clear how cancer cells acquire drug resistance, the mechanism underlying the resistance shown by BT474 cells appears to be completely different from that underlying the survival of normal cells. Damaged normal cells lost PCNA and H2AX expression and became quiescent, showing a flattened and enlarged morphology. However, BT474 cells accumulated high levels of H2AX without losing PCNA expression (Fig. 5A) or showing any of the associated morphological changes (B). In addition, the growth of BT474 cells was immediately restored after release from CPT, whereas primary WT MEFs remained quiescent (Fig. 5C). Thus, the mechanism underlying the resistance shown by BT474 is different from that underlying the survival of normal cells. This implies that drug resistance develops in cancer cells as they acquire other mechanisms that permit them to escape cell death.
Drug resistance can be attributed to efflux (the active pumping of a drug out of the cell) as well as to various survival signals. However, the resistance of BT474 cells is probably not associated with drug efflux because damaged BT474 cells accumulated H2AX and showed an increase in ␥H2AX signaling and in the levels of ubiquitinated PCNA and H2AX (Fig. 5, A and D). The cells also showed increased levels of phosphorylated AKT, Bcl-xl, and Bcl-2 (all of which promote cell survival) upon CPT treatment, which was not observed in MCF7 cells (Fig. 5D). This suggests that survival signals, rather than drug efflux, promote the survival of these cells. Unlike drug efflux, which acts for multiple drugs, survival signals are dependent on the response to each type of damage. In fact, BT474 is resistant to HU but not to cisplatin and doxorubicin (supplemental Fig. S4). Nevertheless, resistant cancer cells that express the ␥H2AX signal are more likely associated with survival signal activation.
H2AX Knockdown in Cancer Cells Induces Tolerance to CPT and a Senescent Morphology-The above results indicate that normal cells resist damage by down-regulating their expression of H2AX, a process regulated by the Arf/p53 protein module. This process is non-functional in cancer cells. However, one question remains. Do reduced H2AX levels directly contribute to resistance? To address this, we knocked down H2AX expression in the colon cancer cell line SW480. SW480 cells treated with a negative control siRNA were killed by CPT. However, SW480 cells treated with siRNA targeting H2AX survived, FIGURE 4. Cancer cells are more sensitive to CPT than normal cells unless they acquire resistance. A-C, cancer cells are generally sensitive to CPT unless they acquire resistance. Cancer cells are more sensitive to CPT than primary NIHs and human mammary epithelial cells (hMEC). However, BT474 breast cancer cells are resistant (A). Sensitivity to damage was determined as outlined in Fig. 1. Survival rates were plotted as in Fig. 1A. B, representative images of NHF and MCF7 cells are shown before and after CPT treatment (other cell lines are shown in supplemental Fig. S3, A and B). C, H2AX and ␥H2AX levels in MCF7 and NHF cells (levels in other cell lines are shown in supplemental Fig. S3, C and D). adopting the flattened and enlarged morphology associated with impairment of DNA damage checkpoint responses (Fig.  6), which was also observed in normal cells. Similar results were obtained for MCF7 and HCT116 cells (supplemental Fig. S5, A  and B). In addition, immortalized MEFs were sensitized to CPT treatment upon H2AX overexpression (supplemental Fig.  S5C). These results suggest that normal cells survive in the presence of CPT by down-regulating H2AX. This lends further support to the concept that the mechanism underlying the survival of normal cells involves the Arf/p53-dependent downregulation of H2AX. Thus, cancer cells that are unable to downregulate H2AX are killed preferentially.

A Parp Inhibitor Sensitizes Cells to CPT, Resulting in Increased H2AX Accumulation and Enhanced Checkpoint
Responses-On the basis of the above results, Arf/p53-dependent down-regulation of H2AX enables cells to survive in the presence of CPT, whereas H2AX accumulation is associated with cell death via the induction of efficient checkpoint responses. This prompted us to investigate the effects of modulating H2AX levels. For this, we used the Parp inhibitor PJ34, which is thought to efficiently induce cancer cell death for the following reasons. Down-regulation of H2AX in normal cells is less efficient during non-homologous end-joining than during homologous recombination (14), suggesting a more efficient checkpoint response during non-homologous end-joining than during homologous recombination, and after damage induced by CPT, non-homologous end-joining becomes the predomi-nant repair mechanism in the absence of Parp1 because of the suppression of homologous recombination (27,28). Therefore, we tested whether PJ34 enhances the killing efficiency of CPT in HCT116 cells via increased expression of H2AX. As expected, treatment with PJ34 significantly increased the level of CPT-induced cell death (3-fold), which was associated with increased H2AX accumulation and damage-related cell signal-  Arf/p53-dependent Cell Survival ing, as indicated by an increase in the levels of ␥H2AX and phosphorylated SMC1 (Fig. 7, A and B). In addition, cells pulsed with CPT for 1 h, either in the presence or absence of PJ34, showed identical levels of damage (during the first hour), as assessed by the number of ␥H2AX-positive cells and the ␥H2AX signal intensity (Fig. 7C, 0 h). This led to a significant increase in ␥H2AX intensity in PJ34-treated cells (see the shifted peak at 1 h). Furthermore, the ␥H2AX signal decayed much more slowly in cells treated with PJ34, which is consistent with a deficiency in Parp-dependent repair (25,29). Thus, Parp inhibition induces H2AX accumulation by modulating the repair system, a process that enhances the checkpoint responses and leads to increased cancer cell death (Fig. 7D).

DISCUSSION
Although DNA-damaging drugs are used widely to treat cancer, it is still unclear how they induce death in cancer cells more efficiently than in normal somatic cells. This study identified a novel regulatory mechanism underlying the selective killing of cancer cells that have not yet acquired drug resistance (Fig. 7D). The mechanism can be outlined as follows. Normal cells downregulate H2AX in response to DNA replication stress via a process that is regulated by Arf/p53, resulting in defective checkpoint responses (H2AX is required for efficient checkpoint responses), and unlike benign tumors, malignant cancer cells mostly harbor mutations in the Arf/p53 protein module (either Arf or p53). Thus, they are unable to down-regulate H2AX and are sensitive to drugs that induce DNA replication stress (unless they have acquired drug resistance).
Normal cells generally become growth-arrested after several rounds of proliferation in vivo and in vitro, a process that is associated with H2AX down-regulation and is dependent on both Arf and p53. Although it is still unclear how the downregulation of H2AX is controlled by the Arf/p53 protein mod-FIGURE 7. H2AX accumulation induces cancer cell death. A, treatment with a Parp inhibitor increases CPT-induced cancer cell death. The effects of the Parp inhibitor PJ34 were examined in CPT-treated HCT116 cells using colony formation assays. Survival rates were plotted with the means of three independent experiments and the S.D. Treatment with PJ34 caused a significant increase in the level of cell death. B, Parp inhibitors cause increased H2AX accumulation and heightened the checkpoint response in CPT-treated cells. The effects of PJ34 were examined in HCT116 cells synchronized in S phase using a double thymidine block. Cells were released from double thymidine block for 1 h and then treated with CPT in the presence or absence of PJ34. The level of H2AX accumulation was increased significantly in the presence of PJ34, which was associated with increased activation of the checkpoint response (indicated by expression of ␥H2AX and phosphorylated SMC1). C, Parp inhibition induces increased expression of ␥H2AX in response to CPT. To address the effects of PJ34 on ␥H2AX expression, cells were pulsed with CPT for 1 h in the presence or absence of PJ34 and then released. ␥H2AX levels were then examined by FACS (the top panel shows the experimental scheme). ␥H2AX levels after 1 h of CPT treatment (0 h) were almost identical in the absence and presence of PJ34, as determined by the number of damaged cells and the intensity of the ␥H2AX signal (see the signals denoted by the red bar in each panel). At 1-5 h after release from CPT, an increase in ␥H2AX intensity was observed only in PJ34-treated cells (the peak shifted onto the green line). D, model showing differences in CPT sensitivity in normal cells, cancer cells, and drug-resistant cancer cells. Normal cells survive in the presence of CPT by down-regulating H2AX and becoming quiescent. Transformed cells accumulate H2AX and are killed preferentially unless they develop resistance. MAY 10, 2013 • VOLUME 288 • NUMBER 19 ule (30), the process only occurs in normal cells under the regulation of both Arf and p53 (14), which prevents transformation (19,31). The results of this study show that normal cells down-regulate H2AX and enter a quiescent state in response to CPT, which is identical to the growth arrest observed when cells proliferate normally and clearly differently from CPT-resistant BT474 cancer cells (see supplemental Fig.  S6 and Fig. 5C). This implies that CPT treatment may simply induce premature growth arrest in normal cells, which then enter a quiescent state (unless the damage is too severe). Normal cells only enter such a state after exposure to mild stress and are protected from immortalization. This process is clearly different from the response to acute damage (32,33). Thus, cells generally survive in the presence of CPT or HU unless cells are transformed with mutations in the Arf/p53 protein module. This results in aberrant H2AX regulation and changes in the checkpoint response.

Arf/p53-dependent Cell Survival
p53 is involved in the DNA damage checkpoint response (34 -36), during which it is activated by ATM and ATR and plays a role in inducing senescence and apoptosis (37)(38)(39). Thus, cancer cells harboring mutations in p53 are less sensitive to drugs than cells that do not harbor p53 mutations. However, chemotherapy kills cancer cells harboring Arf/p53 mutations more efficiently than it kills normal somatic cells that possess functional Arf/p53. The results presented in this study show that Arf/p53-dependent down-regulation of H2AX allows normal cells to survive in the presence of cytotoxic drugs that induce DNA replication stress (Figs. 3E and 7D). This is mainly because cells lacking H2AX show defective activation of checkpoint responses, such as ATM and ATR (40 -42), which activate p53-mediated apoptosis. This suggests that the level of H2AX is the critical factor that determines whether cell death is induced. Taken together with the results of the experiments showing that a Parp inhibitor modulates H2AX expression, this strongly suggests that, before embarking upon a course of chemotherapy, H2AX status must be taken into consideration if efficient cell death is to be induced.