Hypoxia Activates Tumor Suppressor p53 by Inducing ATR-Chk1 Kinase Cascade-mediated Phosphorylation and Consequent 14-3-3γ Inactivation of MDMX Protein*

Background: It remains unclear how the hypoxia-ATR-Chk1 pathway induces p53. Results: Hypoxia suppresses MDMX activity via ATR/Chk1 phosphorylation of Ser-367 and subsequent MDMX-14-3-3 interaction, leading to p53 activation. Conclusion: Induction of MDMX-14-3-3 binding accounts for p53 activation by hypoxia. Significance: Our study elucidates a molecular mechanism underlying hypoxia-induced p53 activation. It has been known that p53 can be induced and activated by hypoxia, an abnormal condition that often occurs in rapidly growing solid tumors or when normal tissues undergo ischemia. Although the ATR-Chk1 kinase cascade was associated with hypoxia-induced p53 activation, molecules that directly link this hypoxia-ATR-Chk1 pathway to p53 activation have been elusive. Here, we showed that hypoxia could induce phosphorylation of MDMX at Ser-367 and enhance the binding of this phosphorylated MDMX to 14-3-3γ, consequently leading to p53 activation. A Chk1 inhibitor or knockdown of ATR and Chk1 inhibited the phosphorylation of MDMX at Ser-367 and impaired the binding of MDMX to 14-3-3γ in addition to p53 activation in response to hypoxia. In primary mouse embryonic fibroblast cells that harbor a mutant MDMX, including the S367A mutation, hypoxia also failed to induce the binding of this mutant MDMX to 14-3-3γ and to activate p53 and its direct targets. These results demonstrate that hypoxia can activate p53 through inactivation of MDMX by the ATR-Chk1-MDMX-14-3-3γ pathway.

It has been known that p53 can be induced and activated by hypoxia, an abnormal condition that often occurs in rapidly growing solid tumors or when normal tissues undergo ischemia. Although the ATR-Chk1 kinase cascade was associated with hypoxia-induced p53 activation, molecules that directly link this hypoxia-ATR-Chk1 pathway to p53 activation have been elusive. Here, we showed that hypoxia could induce phosphorylation of MDMX at Ser-367 and enhance the binding of this phosphorylated MDMX to 14-3-3␥, consequently leading to p53 activation. A Chk1 inhibitor or knockdown of ATR and Chk1 inhibited the phosphorylation of MDMX at Ser-367 and impaired the binding of MDMX to 14-3-3␥ in addition to p53 activation in response to hypoxia. In primary mouse embryonic fibroblast cells that harbor a mutant MDMX, including the S367A mutation, hypoxia also failed to induce the binding of this mutant MDMX to 14-3-3␥ and to activate p53 and its direct targets. These results demonstrate that hypoxia can activate p53 through inactivation of MDMX by the ATR-Chk1-MDMX-14-3-3␥ pathway.
Low oxygen tension or hypoxia is an inevitable pathophysiological status during solid tumor progression. It occurs due to the lack of blood supply in rapidly growing solid tumors, which possess anoxic (0%) or hypoxic (Ͻ8% of oxygen) regions with poor prognosis (1,2). In response to these anoxic or hypoxic conditions, cancerous cells produce several important proteins that either favor or prevent their survival and advanced progression. Among those, although HIF1␣ (hypoxia-inducible factor 1␣) is induced to promote angiogenesis and anaerobic metabolism as a tumor-promoting factor (2,3), the tumor suppressor p53 is often activated as well to stop tumor growth and progression in wild type p53 containing cancer cells (4) by inducing apoptosis and/or cell cycle arrest and repressing angiogenesis (5,6).
During the past decade, a number of studies have been reported to elucidate how hypoxia might induce p53. The first candidate molecule that was proposed to mediate p53 induction by hypoxia is HIF1␣ (7,8). HIF1␣ was shown to directly associate with p53 and to stabilize it in response to hypoxia. However, this notion was challenged by a kinetic study showing that HIF1␣ and p53 were induced at distinct hypoxia stages (9). p53 was stabilized in more stringent hypoxic conditions than HIF1␣ and inversely proportional to HIF1␣ levels with repression of HIF1␣ activity in response to hypoxia (9 -12). It seems that HIF1␣ is insufficient to up-regulate p53 (6) and less likely responsible for extremely low oxygen-triggered p53 induction through direct interaction. Instead, ATR (ATM and Rad3-like serine/threonine kinase) and/or Chk1, a DNA damage checkpoint kinase as a downstream target of ATR, have been proposed as the responsible molecules in response to hypoxia (9,(13)(14)(15)(16). ATR was found to phosphorylate p53 at serine 15 in response to hypoxia, whereas ATM could not (9,14). It was shown previously that serine 15 phosphorylation by UV-activated ATR or ␥ irradiation-activated ATM and/or DNA-dependent protein kinase (DNAPK) can partially stimulate p53 transcriptional activity (17)(18)(19)(20). However, serine 15 phosphorylation could not account for the full scale of p53 induction by hypoxia, as mutation of this residue could only partially affect p53 activity in response to ionizing irradiation in animals and does not seem to affect p53 stability (21,22). Chk1 also was thought to have a role in hypoxia-responsive p53 induction (9,15,16). It is unclear whether Chk1 directly phosphorylates p53 or other components such as MDM2 or MDMX that in turn affects p53 stability. Hence, the exact mechanism of how ATR and Chk1 induce p53 activation in response to hypoxia still remains elusive. The tight regulation of p53 has been mainly attributed to the MDM2-MDMX-p53 loop, whose direct or indirect interferences lead to p53 activation by various cellular stresses, including DNA damage stress (23). Although MDM2 has been described to be a prime negative feedback regulator of p53 (24 -27), another crucial regulator of p53, MDMX, also monitors the intracellular level and activity of p53 together with MDM2 (28,29). Of note, MDMX can also function independently of MDM2 (30). Like MDM2, MDMX is indispensable for p53 inactivation, as knocking out the TP53 gene also rescues the lethality of mdmx null mice (31,32). Recently, in vivo studies by either deleting the C-terminal RING domain of MDMX (MDM4) or generating the C462A mutation, both of which disrupt MDM2-MDMX binding, in mice demonstrate that the MDM2-MDMX binding is essential for negating p53 function in the animals (33,34) These results are consistent with the previous reports showing that MDM2 and MDMX can form a heterocomplex in vitro and in cells (35)(36)(37). Nonetheless, surprisingly, no direct connection has been unambiguously built between this loop and hypoxia-caused stress.
Several members of the 14-3-3 family, including 14-3-3␥, have been reported to bind primarily to phosphorylated MDMX at serine 367 residue in response to stress signals such as DNA damage stimuli (38 -40). Previously, we also demonstrated that UV-mediated Chk1 activation induces phosphorylation of MDMX at Ser-367 and enhances the interaction between MDMX and 14-3-3␥, leading to p53 activation (41). This phenomenon was also demonstrated in a mouse study by knocking-in triple mutations (S341A, S367A, and S402A) at MDMX phosphorylation sites (42). Thus, we speculated that hypoxia might activate p53 by blocking the activity of MDMX via the ATR-ChK1 kinase cascade, which leads to MDMX phosphorylation at Ser-367 and promotes MDMX interaction with 14-3-3, suppressing MDMX activity. Here, in this study, we tested this hypothesis and verified this hypoxia-ATR-ChK1-MDMX-14-3-3-p53 pathway.

EXPERIMENTAL PROCEDURES
Cell Culture and Hypoxia Treatment-Human osteosarcoma U2OS cells and human embryonic kidney (HEK) epithelial 293 cells were cultured in DMEM supplemented with 10% FBS, 10 units/ml penicillin, and 0.1 mg/ml streptomycin at 37°C in a 5% CO 2 humidified atmosphere. Mouse embryonic fibroblast (MEF) 5 wild type (WT) or MDMX triple mutant (3SA) MEF cells (37) were cultured in DMEM supplemented with 15% FBS, 10 units/ml penicillin, 0.1 mg/ml streptomycin, 1ϫ nonessential amino acids (Cellgro), and 1ϫ ␤-mercaptoethanol (Millipore). Prior to exposure to hypoxic condition, cells were cultured and incubated at 5% CO 2 /37°C for 24 -48 h with ϳ50% confluence. Hypoxia was established by incubating cells in an anaerobic system (Thermo Forma), which reduces O 2 levels less than ϳ0.1% close to anoxic condition. The O 2 levels were also monitored by a methylene blue indicator strip during the incubation of cells inside the hypoxia chamber with the time periods indicated in figure legends. Corresponding normoxia control cells were incubated under a normal culture condition.
RNA Interference-siRNA duplex specific to human ATR or Chk1 were synthesized by Dharmacon. The target sequences, which were also utilized in the previous studies (46,47) are as follows: ATR, 5Ј-AAGCCAAGACAAATTCTGTGT-3Ј and Chk1, 5Ј-AAGCGTGCCGTAGACTGTCCA-3Ј. U2OS cells were transfected with 20 nM (final concentration) ATR or Chk1 siRNA duplex or scrambled RNA duplex using siLentFect lipid reagent (Bio-Rad), and the following day, the cells were transfected once again. Approximately 24 h after the second transfection, cells were subjected to hypoxic conditions and incubated for indicated time periods in figure legends followed by WB and/or co-IP analyses. In some cases for ATR knockdown, U2OS cells were transfected with shRNA or infected by lentivirus encoding shRNA and incubated for ϳ48 h prior to hypoxia treatment.
Cell Cycle Analysis-U2OS cells were double-transfected with scramble, ATR, or Chk1 siRNAs, and ϳ24 h after the second transfection, cells were subjected to normoxic or hypoxic conditions. Wild type or 3SA MEF cells were split the day before hypoxia treatment. Then, cells were harvested at indicated time points and fixed with 70% EtOH (final concentra-tion) at least for 30 min on ice. After washing with 1ϫ PBS, cells were stained with phosphatidylinositol staining buffer in PBS (50 g/ml propidium iodide (PI), 0.1 mg/ml RNase A, and 0.05% Triton X-100) and incubated for 40 min at 37°C followed by fluorescence-activated cell sorter (FACS) analysis.
Immunofluorescence Staining and Fluorescence Microscopic Analysis-U2OS cells were transfected with 1 mM random siRNA or Chk1-specific siRNA for 72 h or treated by 1 mM hydroxyurine for 12 h. Cells were fixed for immunofluorescence staining with rabbit anti-␥-H2AX antibody and secondary goat anti-rabbit antibody conjugated Alexa Fluor 546 (red), respectively. DNA was stained with DAPI (blue). Stained cells were analyzed under the Zeiss Axiovert 200 M fluorescent microscope (Zeiss).

RESULTS AND DISCUSSION
To verify and work out the condition for the effect of hypoxia on p53 activation, U2OS cells were incubated under very lowoxygen condition (Ͻ 0.1% O 2 ) up to 16 h. The result showed that, indeed, p53 and its targets p21 and Puma were induced in a time-dependent manner under this condition (Fig. 1, A and  B). Hypoxia-mediated p53 activation could be also extended by 18 -24 h in various cancer cells, including U2OS cells (supplemental Fig. S1A) (9,14). In addition, the treatment of CoCl 2 , a hypoxia mimetic compound, which can induce biochemical and molecular responses similar to those elicited by hypoxia (8), induced p53 in MEF cells at 16 h and U2OS cells in a time-dependent manner up to 24 h under hypoxia as well (supplemental Fig. S1B). These results confirm that hypoxia indeed induces p53 independently of cell types. Interestingly, hypoxia led to the phosphorylation of serine 367 residue of MDMX (Ser-367-MDMX), which is known to be the major phosphorylation site of MDMX for the binding of 14-3-3 protein(s) in response to stress signals, such as DNA damage stimuli (38 -41). As expected, this phosphorylation enhanced the binding of 14-3-3␥ to MDMX, leading to p53 activation (Fig. 1C). These results suggest that hypoxia can induce phosphorylation of MDMX at Ser-367 and subsequent induction of MDMX binding to 14-3-3␥, leading to suppression of MDMX activity toward p53 and consequent p53 activation.
Because DNA damage stimuli such as UV irradiation can induce p53 activation through phosphorylation of MDMX at Ser-367 by Chk1 (41) and ATR-Chk1 kinase cascade plays an important role in hypoxic stress (9,13,14), we deduced that hypoxia-mediated MDMX phosphorylation might employ the same ATR-Chk1 signaling pathway to activate p53. To test our speculation, we utilized a Chk1 inhibitor, UCN-01 (48) to determine whether it affects the phosphorylation of endogenous MDMX at Ser-367 in response to hypoxia. When U2OS cells were subjected to hypoxia, as expected, MDMX phosphorylation was induced at Ser-367; however, when treated with UCN-01 under the same hypoxic condition, MDMX phosphorylation at this site was reduced dramatically, although the total amount of MDMX was comparable ( Fig. 2A). Subsequently, the hypoxia-enhanced interaction between exogenous Myc-MDMX and FLAG-14-3-3␥ was considerably reduced by UCN-01 treatment (Fig. 2B). Consistently, UCN-01 treatment also reduced hypoxia-induced endogenous MDMX phosphorylation at Ser-367 and weakened the interaction between endogenous MDMX and 14-3-3␥, correlated well with a significant decrease in p53 activation (Fig. 2C). Thus, Chk1 plays an important role in hypoxic response as reported previously (9, 13, 14) but phosphorylates MDMX in this experiment.  To confirm this observation and to demonstrate the involvement of the ATR-Chk1 kinase cascade in MDMX phosphorylation and MDMX-14-3-3 binding, we introduced si-or shR-NAs for Chk1 and ATR into U2OS cells. First, knockdown of Chk1 under hypoxic condition abolished MDMX phosphorylation at Ser-367 as well as the interaction between endogenous MDMX and 14-3-3␥ (Fig. 3A). To exclude the possibility that hypoxia might activate Chk1 by inducing DNA damage, we checked whether knockdown of Chk1 would impair DNA damage by assessing ␥-H2AX focus formation. Surprisingly, knockdown of Chk1, similar to hydroxyurea treatment, also caused DNA damage (supplemental Fig. S2). This result indicates that hypoxia-induced MDMX Ser-367 phosphorylation is catalyzed by Chk1 that is activated in response to hypoxia rather than DNA damage signals. If not, knockdown of Chk1 would still induce Ser-367 phosphorylation, as the depletion of this kinase also caused DNA damage (supplemental Fig. S2). However, this was not the case (Fig. 3). Next, we also performed a similar set of experiments by knocking down ATR, the upstream kinase of Chk1. The result demonstrated that the depletion of ATR also weakens the endogenous MDMX phosphorylation at Ser-367 as well as MDMX interaction with 14-3-3␥ (Fig. 3B).
If MDMX phosphorylation necessary for 14-3-3␥ binding is essential for hypoxia-induced p53 activation, mutations in MDMX phosphorylation sites where 14-3-3 can bind should be expected to abolish p53 activation in response to hypoxia. Although Ser-367 is a key phosphorylation site of MDMX for the binding of 14-3-3 proteins, including 14-3-3␥, phosphorylation at Ser-342 of MDMX was also reported to involve the interaction between 14-3-3 and MDMX (40,49). To test this idea, we first employed MDMX 3SA MEF cells, which harbor a knock-in MDMX mutant with S341A (S342 in human), S367A, and S402A (S403 in human) mutations, for our experiments under hypoxic stress. When wild type MEF cells were subjected to hypoxia for 18 h, as expected, p53 was activated with the induction of its direct downstream target protein p21 (Fig. 4A). Increased endogenous 14-3-3␥ binding to MDMX was also detected in these cells in response to hypoxia (Fig. 4A). However, 3SA MEF cells failed to show significant induction of p53 and its target protein in response to hypoxia (Fig. 4A), similar to the previous study in which p53 activity also was considerably hampered in 3SA MEF cells compared with wild type MEF cells in response to DNA damage (42). In addition, 3SA MEF cells were unable to show enhanced interaction between MDMX and 14-3-3␥ after hypoxia treatment (Fig. 4A). This result could rule out the possibilities such that ATR-Chk1-mediated direct p53 activation by N-terminal phosphorylation of p53 or hypoxia-mediated MDM2 phosphorylation, which promotes MDM2 autodegradation, as only triple mutations of MDMX are involved in this occasion. Next, wild type and 3SA MEF cells were cultured under a normoxic or hypoxic condition followed by FACS analysis. Although wild type MEF cells showed Ͼ15% increase in G 1 phase cells after hypoxia treatment, 3SA MEF cells were unable to demonstrate any significant increase in G 1 arrest (Fig. 4B). This difference between WT and mutant MDMX-containing MEF cells should be due to the phosphorylation of MDMX at these three sites followed by p53 activation. Because 3SA MDMX in MEF cells possesses three serine mutations, to determine which serine is more critical for hypoxia-induced interaction between MDMX and 14-3-3␥, we dissected and evaluated the outcome derived from individual single mutations of these MDMX phosphorylation sites in the perspective of interaction between two proteins. Co-immunoprecipitation after ectopic expression of FLAG-14-3-3␥ and Myc-tagged MDMX with respective single alanine mutations at Ser-342, Ser-367, or Ser-403 residues showed that the S367A mutation of MDMX completely abolished the interaction between 14-3-3␥ and MDMX regardless of oxygen status (Fig.  4C). Also, although the S342A mutation hampered this interaction, hypoxia treatment was still able to enhance the interaction between S342A-MDMX and 14-3-3␥, relatively to a less degree compared with that for wild type MDMX (Fig.  4C). In contrast, the S403A-MDMX mutation did not hinder the binding of 14-3-3␥ to MDMX, compared with the corresponding wild type MDMX control before and after hypoxia (Fig. 4C). These results, similar to the previous observations in response to DNA damage (38 -42), again firmly support the phosphorylation of MDMX at Ser-367 is the most essen- A, knockdown of Chk1 eliminates hypoxia-induced MDMX phosphorylation and MDMX interaction with 14-3-3␥. U2OS cells were double-transfected with scramble siRNA or siRNA specific to Chk1 (siR-Chk1), and ϳ24 h after the second transfection, cells were subjected to hypoxia. After 6 or 18 h hypoxia, cell lysates were prepared for WB (40 g) or co-IP analyses. 500 g of cell lysates were incubated with anti-14-3-3 antibodies for co-IP and subjected to SDS-PAGE followed by WB analysis. Proteins of interest were detected by antibodies as listed on the left. B, depletion of ATR inhibits hypoxia-induced MDMX phosphorylation and its interaction with 14-3-3␥. Knockdown of ATR was achieved by the introduction of siRNA or shRNA in U2OS cells ϳ24 h after double transfection. Then, cells were subjected to hypoxia for additional 6 or 18 h followed by WB analysis (80 g) and co-IP (500 g) as described above. Corresponding antibodies were listed on the left. tial and responsible event for the interaction between MDMX and 14-3-3␥ in response to hypoxia. Also, nearby Ser-342 appears to possess the potential to facilitate and preserve their intact interaction.
Our study as presented here unambiguously demonstrates that MDMX phosphorylation at Ser-367 is vital to hypoxiainduced p53 activation in addition to DNA damage responses. Moreover, Ser-342 phosphorylation of MDMX seems to function as a favorable factor conducive to Ser-367 phosphorylation and/or 14-3-3 binding to MDMX in the course of hypoxiamediated p53 activation. Although here we have utilized and scrutinized the function of 14-3-3␥ as a binding partner of MDMX in hypoxia-mediated p53 activation in this study, it would not be surprising if other 14-3-3 family members, which have also been shown to interact with MDMX, such as 14-3-3 or 14-3-3⑀, but not 14-3-3 (33-35), may play a role in hypoxia-caused p53 activation as well. It is likely that MDMX phosphorylation-enhanced interaction with 14-3-3 protein(s) might be a common mechanism that accounts for p53 activation via inactivation of MDMX by a number of stress signals.