Constitutive Activation of DNA Damage Checkpoint Signaling Contributes to Mutant p53 Accumulation via Modulation of p53 Ubiquitination

Abstract

Many mutant p53 proteins exhibit an abnormally long half-life and overall increased abundance compared to wild-type p53 in tumors, contributing to mutant p53’s gain-of-function oncogenic properties. Here a novel mechanism is revealed for the maintenance of mutant p53 abundance in cancer that is dependent on DNA damage checkpoint activation. High level mutant p53 expression in lung cancer cells was associated with preferential p53 mono vs. polyubiquitination, suggesting a role for the ubiquitin/proteasome system in regulation of mutant p53 abundance in cancer cells. Interestingly, mutant p53 ubiquitination status was regulated by ATM activation and downstream phosphorylation of mutant p53 (serine 15), both in resting and genotoxin-treated lung cancer cells. Specifically, either inhibition of ATM with caffeine or mutation of p53 (serine 15 to alanine) restored MDM2-dependent polyubiquitination of otherwise monoubiquitinated mutant p53. Caffeine treatment rescued MDM2-dependent proteasome degradation of mutant p53 in cells exhibiting active DNA damage signaling, and ATM knockdown phenocopied the caffeine effect. Importantly, in cells analyzed individually by flow cytometry, p53 levels were highest in cells exhibiting the greatest levels of DNA damage response, and interference with DNA damage signaling preferentially decreased the relative percentage of cells in a population with the highest levels of mutant p53. These data demonstrate that active DNA damage signaling contributes to high levels of mutant p53 via modulation of ubiquitin/proteasome activity toward p53.

Implication

The ability of DNA damage checkpoint signaling to mediate accumulation of mutant p53 suggests that targeting this signaling pathway may provide therapeutic gain.

Introduction

Mutant p53 confers pro-growth/tumorigenic properties to cells, including increased proliferation, chemoresistance, metastatic potential, and genomic instability (1). The gain-of-function (GOF) phenotype of mutant p53 relies partially on its abnormally long half-life, which drives massive accumulation in cancer cells to levels 100-fold or more higher than that observed for wild-type p53 under unstressed conditions in non-transformed cells (2, 3). The increased abundance of mutant p53 in cancer cells is the consequence of a combination of acquired functional properties resulting from the intrinsic effects of the mutation as well as deregulation of the cellular environment in which the mutant protein is expressed. Understanding the mechanisms that promote mutant p53 abundance in cancer cells will prove critical to the design of treatments which might decrease levels of oncogenic mutant p53. It is conceivable that such treatments could partially reverse the neoplastic phenotype of cells that might be dependent on GOF activities of mutant p53 ((4); reviewed in (5)).

Both loss of function and any emergent GOF properties of mutant p53 arise first and foremost from the physical/chemical effects of p53 DNA binding domain mutations, which disrupt sequence specific DNA binding activity and for certain mutants, further destabilize this already thermodynamically unstable domain (6–9). This further disruption of thermodynamic stability by many tumor associated mutations leads to misfolding and aggregation of p53, which itself contributes to mutant p53 oncogenic GOF properties (10). Other p53 DNA binding domain mutants, referred to as contact mutants, exhibit impaired DNA binding in the absence of any observable conformational instability (10). Despite disparate physical/chemical properties, both types of p53 DNA binding domain mutants are aberrantly stable and accumulate in cancer cells (11).

A number of other intrinsic functional properties of mutant p53 that contribute to its abundance have been characterized. The inability of mutant p53 to transactivate the p53 target gene MDM2 has been suggested to contribute to mutant p53 abundance due to a level of expression of MDM2 insufficient to drive p53 degradation (5). Aberrant conformations of mutant p53 have been shown to cause mutant p53 to bind to upregulated Hsp90, trapping it and MDM2 in a complex, preventing proper MDM2-dependent ubiquitination of p53 (11). Many mutant p53 alleles, such as L194F (T47D cells) (12), R175H (skBr3) (12), R280K (MDA MB-231) and R273H (MDA468 and others) (11, 13) interact with Hsp90. The p53 mutants R273H and R273C, which are studied in this report, are classified as tumor-associated contact mutants (6) that exhibit minimal intrinsic thermodynamic instability, but retain Hsp90 binding.

Several cellular mechanisms, whose disruption contribute to the increased stability of mutant p53, have also been described. Mutant p53 is stabilized in many cancer cells, but not in non-transformed cells, suggesting a unique role of the cancer cell environment in the aberrant accumulation of mutant p53 (14). For example, oncogene activation, DNA damage, ionizing radiation, and reactive oxygen species have all been shown to stabilize mutant p53 (15, 16). Other work has shown that in the absence of MDM2 or p16INK4a, mutant p53 becomes stabilized, though the mechanistic link between p16INK4a and mutant p53 stability is unclear (14). Additionally, an absence of ubiquitinated species of mutant p53 has been noted in certain cancer cell lines (11). It is possible that this paucity of ubiquitinated p53 species is responsible, at least in part, for the observed increase in mutant p53 stability in those cell lines. Further defects in the degradation of mutant p53 have been suggested to occur, possibly through impaired delivery of mutant p53 to the proteasome, which may be facilitated in part by MDM2 (17).

Though a number of hypotheses have been proposed to explain the aberrant accumulation of mutant p53 in cancer cells, the contribution of p53 post-translational modifications has not been extensively investigated. We show here that DNA damage leads to p53 serine 15 phosphorylation of mutant p53 by ataxia-telangectasia mutated (ATM) kinase, which facilitates the monoubiquitination, and inhibits the polyubiquitination, of mutant p53 by MDM2, resulting in increased abundance of mutant p53. Inhibition of DNA damage checkpoint kinases, and ATM specifically, resulted in restoration of polyubiquitination of mutant p53 and degradation of mutant p53 in a subpopulation of lung cancer cells with the most active DNA damage checkpoint signaling activity. These data raise the possibility that basal activation of the DNA-damage checkpoint in lung cancer cell populations may lead to inhibition of MDM2-dependent polyubiquitination of mutant p53, impairing degradation or other ubiquitin/proteasome related functions that contribute to the dysregulation of mutant p53 abundance in lung cancer cells.

Materials and Methods

Cell lines

H1048 (R273C) and H1299 (p53 null) lung carcinoma cell lines were purchased from ATCC and maintained in RPMI (Gibco) plus 10% FBS (Hyclone). WI38 (WT p53) normal lung fibroblasts were purchased from ATCC and maintained in Eagle’s MEM (ATCC) plus 10% FBS. ABC1 cells (P278S) (adenocarcinoma) expressing shGFP or shMDM2 (18) were maintained in MEM plus 10% FBS and (1 ug/ml) puromycin (Gibco). H1793, H1975 (both R273H; NSCLC), MDA MB 231 (R280K; breast), U2OS (WT p53; osteosarcoma), and 293T cells (WT p53; human embryonic kidney) were purchased from ATCC and maintained in DMEM plus 10% FBS.

Transfection and plasmids

Cells were plated to 70–80% confluency. For expression, 2ug of pCMV, pCMV-WTp53, pCMV-p53-273H Neo, pCMVMdm2, pCDNA3.1-His-Ub, pcDNA3.1 –WTp53, pcDNA3.1 WTp53Ser15A, pcDNA3.1-p53-273H, pcDNA3.1-273HSer15, pcDNA3.1-p5316KR, pcDNA3.1-p53-273H/16KR were transfected in 60 mm plates using Lipofectamine 2000 (Invitrogen) following the manufacturer’s recommended protocol. Cells were incubated at 37 degrees in 5% CO2 for 40–48 hrs prior to treatment or lysis.

Cell treatments

Lysis: Cells were washed 2X in PBS and lysed in RIPA buffer (150 mM NaCl (Fisher), 50 mM Tris-HCl (Fisher), pH 8.0, 1% NP-40 (Spectrum), 0.1% SDS, 0.5% deoxycholic acid sodium salt, 5 mM EDTA) and Complete Mini protease inhibitors (Roche). N-ethylmaleimide (NEM) (Sigma) was added to the lysis buffer at a working concentration of 5 μM immediately prior to lysis. MG132 was used at a working concentration of 20 μM and was added to cells for 5 hrs. Caffeine (Sigma) was used at a concentration of 2 mM and was added to culture media 20 min prior to treatment with doxorubicin and was maintained in the media throughout treatment.

Nickel-NTA pulldown

An equal number of cells was lysed in 1 ml of guanidinium lysis buffer (6M guanidinium-HCl, 0.1M Na₂HPO4 (Sigma)/NaH₂PO4 (Fisher) pH 8.0, 0.01 M Tris pH 8.0, 5mM imidazole, 10mM β-mercaptoethanol (Fisher)). Lysates were incubated with 45 μl of Ni2+-NTA-agarose beads (Thermo Scientific) for 4 hrs at room temperature by end-over-end rotation. The beads were washed with the following buffers: guanidinium buffer (6M guanidinium-HCl, 0.1 M Na₂HPO4/NaH₂PO4 pH 8.0, 0.01 M Tris pH 8.0, 10 mM β-mercaptoethanol); urea pH 8 buffer (8M urea (Fisher), 0.1 M Na₂HPO4/NaH₂PO4 pH 8.0, 0.01 M Tris pH 8.0, 10 mM β-mercaptoethanol); buffer A (8M urea, 0.1 M Na₂HPO4/NaH₂PO4 pH 6.3, 0.01 M Tris pH 6.3, 10 mM β-mercaptoethanol) plus 0.2% Triton-X 100 (Acros); buffer A; buffer A plus 0.1% Triton X 100. Proteins were eluted with 200 mM imidazole in 5% SDS (Fisher), 0.15 M Tris pH 6.7, 30% glycerol (Fisher), 0.72M β-mercaptoethanol. Samples were then subjected to SDS-PAGE and Western blotting.

Western blotting

Protein concentrations were determined using BCA Protein Assay Kit (Pierce), and 5–20 ug protein extract was added to LDS Sample Buffer (Life Technologies) plus dithiothreitol (Sigma) and Western blotting was carried out as suggested by the manufacturer (Invitrogen). Blots were visualized using an Odyssey scanner (Licor).

Antibodies

For immunoblotting: p53 (1:500 DO-1 Santa Cruz). Phospho p53 serine 15 rabbit (1:1000 Cell Signaling). Phospho p53 ser15 mouse (1:1000 Cell Signalling) MDM2 (1:500 N-20 Santa Cruz). Actin (1:2000 A-2066 Sigma). Hsp70 (1:500 K-20 Santa Cruz). Ubiquitin (1:500 P4D1 Santa Cruz). Anti-ATM (1:500 Ab-3 Calbiochem) and anti-phospho-ATM (1:500 10H11.E12 Santa Cruz). Goat anti-mouse IgG Alexa Fluor 680 and Goat anti-rabbit IgG Alexa Fluor 680 (1:2500 Life Technologies). For flow cytometry and immunofluorescence: Anti-p53 (1:200 FL393 (goat) Santa Cruz). Bovine-anti-goat IgG CFL 594 (1:200 Santa Cruz). Anti-p53 (1:20 DO-1 Santa Cruz). Anti-p53 (D0-1) Alexa Fluor 488 (1:20 Santa Cruz). Goat anti-mouse IgG CFL594 (1:40 or 1:400 Santa Cruz). Anti-phospho-ser139-γH2AXAlexa Fluor 488 (1:100 Millipore). Anti-ATM (PE conjugate) (1:100 Novus Biologicals). Anti-phospho-ATM (1:50 10H11.E12 Santa Cruz).

Immunofluorescence

Cells were fixed and stained for immunofluorescence following an established protocol (19). Briefly, cells were fixed in 4% paraformaldehyde (Thermo Scientific) and were permeabilized in ice-cold PBS (American BioAnalytical) with 0.2% Triton X and were blocked 30 min in PBS plus 0.5% BSA (Fisher). Cells were stained overnight with an anti-p53 antibody p53 FL393 (goat) (1:200) followed by incubation in bovine anti-goat IgG conjugated to CFL 594 (1:200) overnight at 4 degrees. Alternatively cells were stained with a p53 antibody (DO1) followed by staining with a secondary antibody goat anti-mouse CFL 594. Cells were air dried and mounted on coverslips with anti-fade with DAPI (Invitrogen).

Flow cytometry

Cells were fixed and stained following a previously established protocol (20). Briefly, cells were fixed in 70% ethanol (Koptec) added dropwise with vortexing. More than 24 hrs later, cells were stained with primary antibody (anti-p53 D01, anti-phospho-γH2AX-Alexa Fluor 488, anti-ATM or anti-p-ATM, (see “antibodies”) for 2 hrs rotating at 4 degrees, followed by 2 hr incubation in goat anti mouse CFL 594 (1:20 Santa Cruz). Alternatively, cells were rotated 2 hrs at 4 degrees with anti-p53 (DO-1) Alexa Fluor 488 coupled antibody. Flow cytometric analysis was carried out on a BD FACS Aria II flow cytometer using BD FACS Diva software. For flow sorting of H1048 cells after p53 staining, cells with intensities in the highest or lowest decile of intensity were lysed and immunoblotted with anti-p53 antibody.

Results

Ubiquitination of mutant p53 in lung cancer cells

To evaluate the impact of ubiquitination status of mutant p53 in lung cancer on its abundance, we first analyzed p53 ubiquitination by immunoblot in the lung cancer cell lines H1048 (small cell; R273C) (21), H1793 (non-small cell; 273H/209*) (22) and H1975 (non-small cell; R273H). We also included MDA-MB231 breast cancer cells (R280K) to compare to earlier studies (11). Monoubiquitination of p53 was significantly more prevalent than polyubiquitination (expected to migrate primarily above 100 kDa; (23)) in these cell lines compared to control 293T cells that express WT p53 and demonstrate robust polyubiquitination (Fig. 1a). To determine if inhibition of proteasome-mediated degradation of p53 would increase detection of low-abundance polyubiquitinated mutant p53 species in H1048, H1793 or H1975 cells, we treated the cell lines with the proteasome inhibitor MG132. Although we observed an increase in the intensity of monoubiquitinated species of mutant p53 in all cell lines (shown by the arrow; Supp. Fig. 1), treatment with MG132 did not enhance detection of polyubiquitinated mutant p53 (Fig. 1a), suggesting that polyubiquitination may be blocked in these cell lines (as has been observed in MDA-MB231 (11)), resulting in accumulation of monoubiquitinated mutant p53.

Figure 1. Ubiquitination of mutant p53 in lung cancer cell lines.

(a) Lung cancer cell lines containing endogenous mutant p53 (H1975, H1793, H1048) or MDA-MB231 breast cancer cells were treated with MG132 for 7 hrs, and the ubiquitination status of p53 was compared to untreated samples by Western blotting (DO-1). MDA-MB231 breast cancer cells were included to compare our data to previously published results (12). Additionally, 293T cells containing WT p53 were included as a positive control for polyubiquitination. Actin was immunoblotted as a loading control. (b) H1299 cells were transfected with vectors expressing p53-273H, MDM2 or His-tagged ubiquitin and, 40–48 hours later, mock-treated, or treated with MG132 for 7 hours, and ubiquitinated mutant p53 was pulled down from lysates with Ni-NTA beads and subjected to Western blotting (DO-1) (left panel). Total cell lysates were also directly immunoblotted for p53 (right panel). Hsp-70 was immunoblotted as a loading control. Non-ubiquitinated p53 is indicated by (*) and mono, di and tri-ubiquitinated species of p53 are indicated by the arrows.

To confirm the putative higher molecular weight p53 bands as ubiquitinated species, we transfected p53-null H1299 lung cancer cells with expression plasmids for the p53 contact mutant R273H (p53-273H), MDM2 (due to low levels of endogenous MDM2 in H1299 cells) and His-tagged ubiquitin. Following pull-down of His-ubiquitin using Ni-NTA beads and immunoblotting for p53, mono and multiple-monoubiquitinated species of p53-273H were apparent (Fig 1b, shown by arrows), with barely detected polyubiquitinated species, even after concentration on Ni-NTA beads. To compare with ubiquitination patterns seen in cells expressing endogenous mutant p53, total lysates of p53-R273H-expressing H1299 cells, either mock or MG132-treated, were analyzed by p53 immunoblotting for native, mono, and polyubiquitinated species as in Fig. 1a. Indeed, no significant increase in polyubiquitinated p53-273H species was seen in the MG132-treated cells, similar to the findings with endogenous mutant p53 in H1048, H1975, H1793, and MDA-MB231 cells (Fig. 1a).

Monoubiquitination of mutant p53 increases following DNA damage in lung cancer cells and is dependent on the DNA damage checkpoint

To further interrogate the regulation of mutant p53 ubiquitination, as a means of understanding regulation of mutant p53 abundance, we investigated how a treatment known to further increase mutant p53 stability would impact mutant p53 ubiquitination. We thus treated H1048 cells with the chemotherapeutic agent and topoisomerase II inhibitor doxorubicin (Dox), which induces DNA double-strand breaks and further stabilizes mutant p53 alleles (24, 25). We found that after a 2 hour Dox pulse, native mutant p53 levels increased substantially, paralleled by increased levels of monoubiquitinated mutant p53, and the induction of both species was sustained over a 5 hour time course after the 2 hour Dox pulse (Fig. 2a). To show the relative non-modified vs. monoubiquitated band intensities, we also show a higher exposure in Supp. Fig. 2.

Figure 2. Regulation of mutant p53 monoubiquitination following DNA damage.

(a) H1048 cells were mock treated or treated with 2 μm Dox for 2 hours and harvested 0 1, 3 or 5 hrs post-removal of Dox, followed by Immunoblotting for native and monoubiquitinated mutant p53 (DO-1), and p53-phospho S15. Actin was immunoblotted as a loading control. (b) H1048 cells were treated with 2 mM caffeine for 1 hour prior to Dox treatment or 1 μM KU60019 (ATM inhibitor) 8 hours prior to Dox treatment. Cells were then treated with Dox for 2 hours and harvested at the indicated time points after Dox for immunoblotting with anti-phospho-ATM, anti-p53 and anti-phospho-S15 p53 antibodies. Actin was immunoblotted as a loading control. (c) H1299 cells were transfected with p53-273H and MDM2 and, 40–48 hours later, plates were treated with 2 μM Dox for 2 hours in the presence or absence of 2 mM caffeine. Caffeine-treated samples were pre-treated with 2 mM caffeine for 20 minutes prior to addition of Dox, and cells were harvested 0, 3 or 5 hours following removal of Dox, followed by immunoblotting for total p53 (DO-1), phospho-p53 S15 (separate gel with same lysates), and MDM2. Actin was immunoblotted as a loading control.

To investigate the role of the DNA damage signaling cascade in Dox-induced accumulation of mutant p53 in lung cancer cells, we examined the phosphorylation status of S15 (an ATM target) of mutant p53 after Dox treatment to determine if mutant p53, like WT p53, was impacted by the DNA damage signaling cascade. We found that S15 phosphorylation increased significantly after Dox treatment, along with induction of native p53 (Fig. 2a). Notably, S15 phosphorylation was observed even before Dox treatment, suggesting ongoing DNA damage signaling in the absence of Dox (Fig. 2a). To confirm that S15 phosphorylation was indeed the result of bona fide DNA damage signaling after Dox, H1048 cells treated with Dox were analyzed for p53 S15 phosphorylation in the presence or absence of the DNA damage signaling inhibitors caffeine (ATM/ATR inhibitor) and Ku60019 (ATM inhibitor) (Fig. 2b). Both inhibitors substantially blocked the induction of S15 phosphorylation by Dox (Fig. 2b). As a control to confirm both inhibitors blocked ATM activity, the activating autophosphorylation of ATM was analyzed by immunoblotting, and the increase in S1981 phosphorylation after Dox was indeed blocked by both inhibitors (Fig. 2b).

To understand the specific role of DNA damage signaling in regulating p53 ubiquitination after DNA damage in both normal and cancer cells, p53-273H was expressed in either H1299 lung cancer cells or non-transformed WI38 human lung fibroblasts, and cells were harvested after a 2 hour doxorubicin pulse (0h) at intervals up to 5 hours following removal of the drug (Fig. 2c and Supp. Fig. 3a–b, respectively) in the presence or absence of caffeine. We found that, as in H1048 cells, Dox treatment led to an increase in monoubiquitination and abundance of exogenous p53-273H in H1299 and WI38 cells (Fig. 2c, Supp. Fig. 3a–b (26)). Densitometry of the native and monoubiquitinated p53 bands is shown in Supp. Fig. 4. This effect was accompanied by the expected increase in S15 phosphorylation of p53-273H in both H1299 and WI38 cells. Caffeine treatment abrogated both the induction of S15 phosphorylation and increase in monoubiquitination after Dox (Fig. 2c, Supp. Fig. 3a–b and densitometry in Supp. Fig 4), suggesting that PI3K-family checkpoint kinase activity regulates S15 phosphorylation and ubiquitination status after Dox treatment. These data suggest that monoubiquitination of mutant p53 occurs in response to DNA damage in a DNA damage checkpoint kinase-dependent but cellular context independent manner, and suggests that monoubiquitination of mutant p53 may contribute either directly to regulation of mutant p53 abundance in lung cancer cells, or alternatively, serve as an accumulated intermediate due to inhibition of p53 polyubiquitination after DNA damage.

Monoubiquitination of mutant p53 following DNA damage is MDM2-dependent

Since ubiquitination of p53 can generally occur through the activity of the E3 ubiquitin ligase MDM2, (reviewed in (27, 28)), we examined the requirement for MDM2 in the accumulation of monoubiquitinated species of endogenous mutant p53 in lung cancer cells following DNA damage. Mutant p53-expressing ABC1 non-small cell lung cancer cells expressing shRNA against either MDM2 or a non-target control shGFP, were treated with doxorubicin for 2 hours and harvested at 0, 2, 3 or 5 hours following removal of the drug. MDM2 silencing by MDM2 shRNA was demonstrated by MDM2 immunoblot and densitometry (Fig. 3a, Supp. Fig. 5). We found that monoubiquitination of mutant p53 was induced in GFP shRNA-expressing cells in response to doxorubicin (albeit delayed relative to H1048 cells), but to a much lesser extent in MDM2 shRNA-expressing ABC1 cells (Fig. 3a, Supp. Fig. 5). Surprisingly, the abundance of mutant p53 increased in response to Dox more so in GFP shRNA-expressing cells as opposed to MDM2 shRNA-expressing cells, suggesting a potential paradoxical r ole for MDM2 in promoting accumulation of mutant p53 after DNA damage (Fig. 3a, Supp. Fig. 5).

Figure 3. Monoubiquitination of mutant p53 in lung cancer cells following DNA damage requires MDM2 E3 activity.

(a) ABC1 cells endogenously expressing mutant p53 and either non-target control shGFP or shMDM2 were treated with 2 μM doxorubicin for 2 hours and were harvested 0, 2, 3 or 5 hours following removal of the drug to determine the ubiquitination status of mutant p53 by Western blot. An MDM2 immunoblot demonstrated the level of MDM2 silencing in shGFP and shMDM2 cells (separate immunoblot with same lysates). Actin was immunoblotted as a loading control. (b) H1299 cells were transfected with vectors expressing p53-273H and MDM2 or an MDM2 RING mutant (C464A) defective for ubiquitin ligase activity. 40–48 hrs after transfection, cells were treated with 2 μM doxorubicin and were harvested 0, 3, or 5 hours after removal of the drug, followed by p53 immunoblot to detect native and monoubiquitinated p53 species (DO-1). Actin was immunoblotted as a loading control.

The RING domain of Mdm2 is required for modulation of mutant p53 following DNA damage

The E3 ligase activity of MDM2 can preferentially mono- or polyubiquitinate p53 based on cellular conditions and the MDM2:p53 stoichiometric ratio (29, 30). We thus wanted to determine whether the mechanism for increasing mutant p53 abundance in cancer cells after DNA damage might involve modulating the balance between MDM2-dependent monoubiquitination and polyubiquitination of mutant p53. H1299 cells were transfected with plasmids encoding p53-273H and either MDM2 or an MDM2 RING domain mutant, C464A, which abolishes the ubiquitin ligase function of MDM2 (31). Cells were then treated with Dox for 2 hours and were harvested 0, 3 or 5 hours following removal of the drug (Fig. 3b). In MDM2-expressing cells, unmodified and monoubiquitinated mutant p53 levels increased in response to Dox treatment (Fig. 3b and Supp. Fig. 6). In contrast, Dox treatment of cells expressing the MDM2 RING mutant resulted in no detectable increase of mutant p53 monoubiquitination and, despite increased abundance of unmodified mutant p53 at baseline, no additional increase in p53 levels was observed after Dox treatment (Fig. 3b, Supp. Fig. 6). These data suggest that MDM2 E3 ligase activity specifically impacts mutant p53 abundance (as it is known do to for WT p53 abundance), and that once MDM2 E3 activity is blocked, further mutant p53 accumulation is also inhibited after DNA damage. MDM2, therefore, plays an active role in modulating mutant p53 abundance after genotoxic exposure.

Serine 15 phosphorylation of mutant p53 following DNA damage regulates the extent of mutant p53 mono-ubiquitination

Phosphorylation of S15 of p53 by ATM/ATR checkpoint kinases is required for WT p53 stabilization following DNA damage (reviewed in (29, 32)). Given our observation that S15 of mutant p53-273H is phosphorylated similarly to WT p53 after DNA damage (Fig. 2), and that S15 phosphorylation correlates with mutant p53 monoubiquitination, we wanted to determine whether S15 phosphorylation might also contribute to aberrant mutant p53 stabilization and accumulation by influencing the ratio of monoubiquitinated/polyubiquitinated p53.

To test this hypothesis, we first transfected H1299 or WI38 cells (as a control to see if effects were cancer cell specific) with plasmids expressing p53-273H or p53-273H/S15A, along with an MDM2 expression plasmid (Fig. 4a and Supp. Fig. 3c, respectively). Cells were treated with 2μM Dox for 2 hours and were harvested 0, 3, or 5 hours after removal of Dox. We found an increase in monoubiquitination and abundance of S15-phosphorylated p53 in response to DNA damage in cells expressing p53-273H as was seen previously in Fig. 2a with endogenous mutant p53 in H1048 cells (Fig. 4a, Supp. Fig. 3c and densitometry in Supp. Fig. 7). However, mutation of S15 to alanine abrogated induction of p53 S15 phosphorylation and reduced (but did not eliminate) doxorubicin-induced monoubiquitination after Dox treatment in both H2199 and WI38 cells (Fig. 4a, Supp. Fig. 3c, respectively, and Supp. Fig. 7). This suggests that phosphorylation of S15 of mutant p53 following exposure to DNA damage is required for full monoubiquitination and accumulation of mutant p53 in both normal and cancer cells. The residual increase in monoubiquitination of p53-273H/S15A after Dox compared with the more complete inhibition by caffeine of increased monoubiquitination of p53-273H (Fig. 2c) may be due to compensating modifications at other sites that occur within p53-273H/S15A that are otherwise blocked after caffeine treatment (e.g. S20 phosphorylation) (33).

Figure 4. Phosphorylation of p53-273H on serine 15 following DNA damage is required for monoubiquitination and inhibition of polyubiquitination of mutant p53.

(a) H1299 cells were transfected with vectors expressing p53-273H or p53-273H/S15A and MDM2 and 40–48 hours after transfection, cells were treated with 2 μM Dox and harvested 0, 3, or 5 hours following removal of the drug. The level of p53 or monoubiquitinated p53 was assessed by immunoblot (DO-1). A separate immunoblot (same lysate; using monoclonal anti p-Ser15-p53) was run to detect p-Ser15 p53-273H. Actin was immunoblotted as a loading control. (b) H1299 cells were transfected with vectors expressing either p53-273H or p53-273H/S15A and MDM2 and His-tagged ubiquitin and, 40–48 hours later, ubiquitinated mutant p53 was pulled down using Ni-NTA beads, followed by Western blotting for p53 (DO-1). Panels shown are from the same gel.

To investigate the link between S15 phosphorylation and the mono- vs. polyubiquitination status of mutant p53, p53-273H or p53-273H/S15A were ectopically expressed in H1299 cells with MDM2 and His-ubiquitin (Ub), and ubiquitinated p53 was detected by Ni-NTA pulldown followed by p53 immunoblot. We found that p53-273H/S15A exhibited increased polyubiquitination relative to p53-273H when both were co-expressed with MDM2 (Fig. 4b), suggesting that S15-mutated mutant p53 is more efficiently polyubiquitinated in cancer cells than an allele that retains the capacity for phosphorylation by checkpoint kinases. Taken together, these data suggest that phosphorylation of mutant p53 at S15 following DNA damage leads to its accumulation, which is correlated with increased monoubiquitination and decreased polyubiquitination. Moreover, these data suggest that a caffeine-sensitive kinase activity (i.e. ATM, ATR or both) may actively regulate mutant p53 poly- vs. monoubiquitination, and thus abundance, in a manner similar to WT p53, independent of cellular context.

The ATM/ATR inhibitor caffeine rescues polyubiquitination of mutant p53 in cancer cells

Our data thus far indicate that active DNA damage signaling can cause further accumulation of mutant p53 through ATM/ATR signaling, S15 phosphorylation, and inhibition of polyubiquitination that is associated with excess monoubiquitination. To see if active DNA damage signaling in resting untreated cancer cells can also account for mutant p53 accumulation, we assayed phospho-ATM levels in H1299 lung cancer cells vs. WI38 primary lung fibroblasts as a measure of basal DNA damage signaling activity (Fig. 5a). Phospho 1981 (active)-ATM was detected in H1299 cell lysates in the absence of exogenous genotoxin, but was undetectable in WI38 cell lysates, supporting our working hypothesis that active DNA damage signaling occurs in an ongoing fashion in cancer cells, but either not at all, or to a greatly lesser degree, in primary cells (Fig. 5a).

Figure 5. DNA damage signaling regulates mutant p53 ubiquitination.

(a) WI38 and H1299 cell lysates were immunoblotted for phospho ATM. Actin was immunoblotted as a loading control. (b) H1048 cells were treated with 2 mM caffeine for 0–4 hours. Extract was harvested at 1 hour intervals and subjected to Western blotting (DO-1) with higher molecular weight portion of the gel exposed longer than the portion including native p53 to better reveal polyubiquitin conjugates. Actin was immunoblotted as a loading control. (c) H1299 cells were transfected with empty vector (mock) or vectors encoding p53-273H, MDM2 and His-tagged ubiquitin. Forty to forty-eight hours post-transfection, cells were treated with or without 2 mM caffeine for 1 hour followed by addition of 2 μM Dox for 2 hours. Ubiquitinated proteins from cell lysates were pulled down using Ni-NTA beads and immunoblotted using anti-p53 antibody (DO-1). A fraction of the eluent was blotted for his-ubiquitin.

We next wanted to determine if inhibition of DNA damage checkpoint kinases by caffeine treatment would restore polyubiquitination of endogenous mutant p53 in lung cancer cells, phenocopying the effect of the S15A mutation on p53-273H ubiquitination status. H1048 cells were treated with 2 mM caffeine for 0–4 hours, and lysates were analyzed for p53 ubiquitination status by immunoblot. Caffeine treatment rapidly increased the level of polyubiquitinated species of mutant p53 within 1 hour, and levels further increased to 1.6 fold higher at 2 hours (Fig. 5b, Supp. Fig. 8). This suggests that the activity of the ATM/ATR DNA damage checkpoint in H1048 cells modulates the ubiquitination status of mutant p53. Of note, we did not observe a detectable decrease in non-modified mutant p53 levels in total cell extract following caffeine treatment (Fig. 5b and also seen in Fig. 2b) as might be expected with degradation that accompanies an increase in polyubiquitination. This discrepancy could be due to the polyubiquitination only arising in a sub-population of the culture, or not being linked to degradation, and is further addressed in Fig. 6.

Figure 6. Caffeine treatment restores MDM2- and proteasome- dependent degradation of mutant p53 in cancer cells.

(a) H1048 cells were grown on chamber slides and were treated with 2 mM caffeine for 0 or 7 hours, and cells were fixed, followed by staining using an anti-p53 primary antibody (FL393) and bovine anti-goat IgG CFL 594 secondary antibody. (b) WI38 cells were grown on chamber slides and treated for 7 hours with or without caffeine, and were fixed and stained with an anti-p53 antibody (FL393) followed by a bovine anti-goat IgG CFL 594 secondary antibody. (c) H1048 cells fixed and stained with an anti-p53 antibody (DO-1 Alexa Fluor 488) were sorted to collect the cells with the highest 10% and lowest 10% p53 intensity. Equal numbers of cells from each population were then analyzed by anti-p53 (DO-1) and anti-Hsp70 (loading control) immunoblotting, and densitometry revealed that the loading adjusted difference in p53 abundance was 3-fold between high and low decile gate samples. (d) H1048 cells were treated with caffeine with or without MG132 for 7 hours and fixed and analyzed by flow cytometry for p53 abundance. Cells were stained with p53 DO1 Alexa Fluor 488. Values for median intensity of p53 staining and the number of cells in the low and high gates are additionally shown for each sample in Table 1. (e) WI38 cells were treated for 7 hours with or without caffeine, and were fixed and stained with an anti-p53 antibody (DO-1 Alexa Fluor 488) for flow cytometry. The level of endogenous WT p53 was compared in WI38 cells treated with or without caffeine, and the median values of the peaks and the number of cells in the low and high 15% gates are additionally shown in Table 1. (f) Abc1 cells containing mutant p53 and expressing non-target control shGFP (top panels) or shMDM2 (lower panels) were plated on chamber slides and treated with caffeine for 0 or 3 hrs. Cells were fixed and stained with an anti-p53 antibody (DO-1) followed by staining with a goat anti-mouse IgG CFL 594 secondary antibody to determine mutant p53 level by immunofluorescence. (g) ABC1 shGFP and shMDM2 cells were treated with or without caffeine for 3 hours and fixed and analyzed by flow cytometry for p53 abundance. Cells were stained with a primary antibody for p53 (DO1) followed by staining with a secondary antibody (goat anti-mouse IgG-CFL 594). Values for median intensity of p53 staining and the number of cells in the low and high gates are additionally shown for each sample in Table 1.

To further confirm that ATM/ATR signaling controls the extent of mutant p53 polyubiquitination under conditions of active DNA damage signaling, we ectopically expressed p53-273H, MDM2 and His-Ub in H1299 cells, followed by Dox with or without caffeine treatment. Cell lysates were then analyzed for p53-273H ubiquitination by Ni-NTA pulldown and immunoblot. Ni-NTA pulldown of ectopically expressed ubiquitinated mutant p53 showed that incubation with caffeine led to a detectable and significant increase in the level of polyubiquitinated mutant p53 (Fig. 5c and Supp. Fig. 9). Thus, caffeine rescues p53-273H polyubiquitination after DNA damage, consistent with its known effects in inhibiting ATM/ATR, and correlated with the regulation of p53-273H polyubiquitination by p53-273H S15 phosphorylation as documented in Fig. 4b.

The distribution of mutant p53 abundance in cell populations is dependent on the DNA damage checkpoint

To understand the consequence of the induction of mutant p53 polyubiquitination by caffeine, we examined levels of mutant p53 (H1048) or WT p53 (WI38) following caffeine treatment at a single cell level by immunofluorescence (IF), noting that overall levels of mutant p53 in bulk cell populations did not change with caffeine treatment (Figs. 2b and 5b). We surprisingly found that inhibition of the checkpoint with caffeine led to clearly decreased levels of mutant p53 in H1048 cancer cells (Fig. 6a) but not WI38 cells (Fig. 6b). This suggests a checkpoint-dependent accumulation of mutant p53 in cancer cells, whereas in the absence of ongoing checkpoint signaling in normal cells, wild type p53 levels do not change in response to caffeine treatment.

To gain a more precise understanding of p53 abundance across cell populations in different cell lines, and with or without checkpoint inhibition, flow cytometry was employed using a fluorescent-labeled p53 antibody. To first validate that flow cytometric analysis accurately quantitated p53 levels, H1048 cells stained for p53 were separated by flow sorting, and cells in the highest and lowest decile of p53 staining intensity were immunoblotted for p53 (Fig. 6c). Consistent with the significant difference in median p53 staining intensity for the two cell populations from the lowest and highest decile gates, the abundance of p53 in the high decile gate as determined by immunoblot and densitometry was approximately 3-fold higher than in the low decile gate (Fig. 6c), suggesting that the low and high staining intensity gated cells express a detectable difference in mutant p53 protein abundance.

We then analyzed a variety of cell lines including H1048 (small cell lung; mutant p53) (Fig. 6d and Supp. Fig. 10a), H1975 (NSCLC; mutant p53) (Supp. Fig. 11a), MDA-MB231 (breast; mutant p53; for comparison to prior report (11) (Supp. Fig. 11b), WI38 (normal fibroblast; WT p53) (Fig. 6e, Supp. Fig. 11c), and U2OS (osteosarcoma; WT p53) (Supp. Fig. 11d), treated with or without caffeine for overall median intensity, as well as the number of cells that fall in the approximately 15% low and 15% high p53 staining intensity gates (as set for the control (no treatment) condition in each experiment) by flow cytometry (Table 1). The latter measure provides a more sensitive indicator for changes in p53 levels in cells at the highest and lowest levels within a population. We observed modest but significant (all p< 0.05) changes in median intensity of the whole cell population in each mutant p53 cancer cell line after caffeine treatment in the same direction as the IF data, where H1048, H1975, and MDA-MB231 median intensities decreased 10–20%, while U2OS and WI38 median intensity did not decrease, but actually increased by ~50% and ~10%, respectively (p< 0.05; Table 1).

Table 1. Flow cytometric analysis of p53 abundance.

Quantitation of p53 level of abundance by flow cytometry. The median level of mutant p53 as measured by flow cytometry using indicated antibodies (see Methods) in each overall cell population, as well as in the low and high p53, γH2AX-p, ATM, or p-ATM abundance gates, is shown in the last 3 columns, along with the percentage of cells in the low and high p53, γH2AX-p, ATM, or p-ATM abundance gates in the 1^st two columns. All differences (increase or decrease) in median overall intensity for pairwise combinations of control and treatment (+/−caffeine, +/−Dox, or sicontrol vs. siATM) for each cell line were p<0.05.

cell line -- gated on p53	%low p53	%high p53	Median low	Median high	Median overall
1975-no caffeine	15.6	16.7	13173	56236	28386
1975- caffeine	22.1	11.4	13878	56523	23752
MDAMB231-no caffeine	15.5	16	24185	99978	48545
MDAMB231-caffeine	22.3	10.5	23622	97283	43054
U20S-no caffeine	16	15.8	4810	29951	17454
U20S- caffeine	3.4	45.6	3180	32152	25935
H1048-no caffeine	15.7	15.9	28848	95285	56699
H1048-caffeine	33.5	6.2	27444	91855	44920
H1048- MG132	19.3	8.5	28822	90108	50980
H1048-MG132 + caffeine	19.4	10.3	28181	90687	52022
WI38-no caffeine	15.7	15.8	14105	59952	39757
WI38-caffeine	8.8	22.9	8998	60334	44033
ABC1 shGFP-0h	16	15.3	2075	6953	4349
ABC1 shGFP-3h caffeine	28.3	7	2214	6877	3637
ABC1-shMdm2-0h	16	16.4	926	2632	1723
ABC1 shMdm2-3h caffeine	15.6	17.1	955	2671	1742
cell line -- gated on H2AX	%low p53	%high p53	Median low	Median high	Median overall
H1048 H2AX non-dox	15.1	14.8	1187	4091	2233
H1048 H2AX dox	8.8	20.7	1242	4132	2506
cell line -- gated on ATM	%low p53	%high p53	Median low	Median high	Median overall
H1048 control si	15.9	15.2	11641	34367	20973
H1048 ATM si	42.5	7.5	16179	42762	20466
cell line -- gated on p-ATM	%low p53	%high p53	Median low	Median high	Median overall
H1048 p-ATM	15.5	14.5	18862	45852	34385

Looking at the number of cells in the lowest and highest 15% gates, we observed about a 1.5–2-fold increase in low gated and 30–50% decrease in high gated cells for mutant p53 expressing H1048, H1975, and MDA-MB231 cells after caffeine (Fig. 6d; Supp. Fig. 11a–b and 10a; Table 1). WT p53 levels in U2OS and WI38 cells followed the pattern of median intensity with 2–5-fold decrease in low gated cells and 1.5–3-fold increase in high gated cells following caffeine treatment (Fig. 6e, Supp. Fig. 11c–d; Table 1). Our data therefore support a checkpoint-dependence for specific maintenance of mutant p53 abundance at the highest levels in a cancer cell population.

Redistribution of mutant p53 abundance in cell populations by caffeine is MDM2 and proteasome dependent

We next wanted to determine whether the effect of caffeine on mutant p53 abundance in lung cancer cells was dependent on MDM2, the key E3 ligase for p53. We therefore analyzed mutant p53 abundance in ABC lung cancer cells expressing control shGFP vs. shMDM2 by IF (Fig. 6f). MDM2-expressing ABC1-shGFP cells treated with caffeine demonstrated lower p53 intensity than mock treatment, whereas knockdown of MDM2 with shRNA abrogated the effect of caffeine on reducing p53 levels (Fig. 6f). These data were verified by flow cytometry, where caffeine led to a near doubling of low gated cells, with a 50% reduction of high gated cells in ABC-shGFP cells (as expected), but no change in high or law gated cells was observed after caffeine in ABC-shMDM2 cells (Fig. 6g; Table 1; Supp. Fig. 10b). Thus, depletion of mutant p53 levels following inactivation of the DNA damage checkpoint by caffeine is dependent on MDM2.

To explore the proteasome dependence of caffeine induced redistribution of mutant p53 abundance, H1048 cells were treated with caffeine in the presence of the proteasome inhibitor MG132. MG132 treatment blocked the ability of caffeine to shift the population abundance of mutant p53 away from higher expression to lower expression, with essentially no change in the percent cells in the low/high gates, consistent with proteasome degradation as the mechanism for the caffeine effect observed in mutant p53 expressing cancer cells (Fig. 6d, Supp. Fig. 10b; Table 1). Thus, caffeine can lower mutant p53 levels in cancer cells by shifting the abundance curve to the left, and it does so in an MDM2- and proteasome-dependent manner.

Active checkpoint signaling through ATM increases the level of mutant p53 in cancer cells

As ATM is a key target for inhibition by caffeine (Fig. 2b), we explored the direct role of ATM in modulating mutant p53 abundance. Utilizing H1048 cells analyzed by IF or flow cytometry, we analyzed how mutant p53 levels varied with total ATM level, and then how depleting ATM via siRNA affected the distribution of p53 abundance across the cell population (Fig. 7a–c; Table 1). Flow cytometric analysis of control siRNA-treated H1048 cells dually stained for p53 and ATM and then gated for the lowest and highest 15% of ATM intensities, revealed that between the low and high ATM gates, p53 intensity increased by ~3-fold (Fig. 7a; Table 1). Thus, p53 abundance varies in direct proportion to ATM abundance, and possibly in proportion to ATM signaling activity.

Figure 7. Checkpoint kinase signaling upregulates p53-273H abundance.

(a)H1048 cells were transfected with control or ATM siRNA, and the level of mutant p53 in the approximate 15% low and 15% high total ATM gates was assayed by flow cytometry. p53 was stained with p53 DO1 Alexa Fluor 488 and ATM was stained with an ATM (PE conjugate) antibody. (b) The number of cells in the high and low gates gates is shown. (c) H1048 cells were transfected with control or siATM and stained by immunofluorescence to detect the relative levels of p53 using an anti-p53 FL393 antibody and bovine anti-goat IgG CFL 594 secondary antibody. (d) H1048 cells were fixed and stained for flow cytometry to simultaneously determine endogenous levels of p-ATM (S1981) and p53. p-ATM was stained with a p-ATM antibody followed by staining with a secondary antibody (goat anti-mouse CFL-594). Cells were then subsequently stained for p53 using a DO1 p53 Alexa Fluor 488 antibody. The median of p53 levels in the low and high p-ATM gates are shown in Table 1.

By both IF and flow cytometry, ATM depletion, like caffeine, resulted in decreased p53 abundance in H1048 cells, indicating that physiologic levels of ATM are necessary to maintain high levels of mutant p53 in H1048 lung cancer cells (Fig. 7a, c; Table 1; ATM knockdown confirmed in Fig. 7a by flow cytometry and by immunoblot in Supp. Fig. 12). Quantitatively, we show that with ATM knockdown the overall median intensity of mutant p53 decreases slightly (as with caffeine) (Table 1) and that ~3X more cells are seen in the low ATM gate and ~50% fewer in the high gate, consistent with the knockdown (gates set using the lowest/highest 15% intensity within the control si treated population; Fig. 7b). Examination of the distribution of p53 intensities in the low vs. high ATM gate after siATM treatment demonstrates again the increased cells in the low vs. high gate, with the p53 median intensities from each ATM gate still maintaining a ~2.6 fold difference between low and high ATM gates (Fig. 7a–b; Table 1). Thus, knockdown of ATM increases the number of cells in the population with a low p53 abundance while decreasing the number of cells harboring the highest abundance, without actually decreasing the lowest p53 abundances any further.

We next explored the relationship between the activation state of ATM, as measured by S-1981 phosphorylation, with mutant p53 abundance, using flow cytometry analysis of H1048 cells stained simultaneously with p53 and ATM S-1981p antibodies, in the absence of any exogenous DNA damage (Fig. 7d; Table 1). Dually stained cells were gated for p53 analysis from the lowest and highest 15% population intensity for p-ATM abundance (Fig. 7d). Cells with the lowest p-ATM levels exhibited lower p53 staining intensity, whereas cells staining most intensely for p-ATM exhibited higher levels of p53 (Fig. 7d; Table 1), with the median p53 intensity of cells in the high p-ATM gate 2.5-times higher than from the low p-ATM gate (p<0.05), indicating that activated ATM signaling leads to elevated levels of mutant p53 in H1048 cells (Fig. 7d; Table 1). Taken together, these data indicate that checkpoint signaling through ATM leads to higher levels of mutant p53.

The intensity of DNA-damage signaling in mutant p53 lung cancer cells is correlated with p53 abundance

Based on the dependence of mutant p53 accumulation in H1048 lung cancer cells on ATM signaling, we wanted to independently confirm that the intensity of DNA damage signaling in an individual cell correlated with the abundance of mutant p53. We therefore interrogated a population of H1048 cells that were untreated or treated with Dox for 2 hours followed by dual staining with anti-phospho-γH2AX (γH2AX-P_i, a marker reflecting ongoing DNA damage signaling, especially by ATM) and anti-p53 antibodies (Fig. 8; Table 1). Cells were classified into high 15% or low 15% γH2AX-P_i gates using flow cytometry and the staining intensity of mutant p53 was then compared within each gate. We found that cells in the high γH2AX-P_i gate had a higher median level of mutant p53 (~3.5-fold higher) than the median p53 intensity in the low γH2AX-P_i gate in untreated cells (p<0.05; Fig. 8; Table 1). Dox treatment led to the expected increase in the median intensity of γH2AX-P_i, but interestingly no substantial increase in the median intensity of p53 staining in cells isolated from the high γH2AX-P_i gate (Fig. 8; Table 1). The main impact of Dox treatment was to shift many more cells into the high γH2AX-P_i gate (Fig. 8b). Thus, it appears that the intensity of p53 staining, and by inference, abundance of p53, is already at a near maximum in high γH2AX-P_i staining untreated H1048 cells, and that Dox just shifts more cells toward expressing the higher level, rather than further increasing p53 levels in the already high expressing population. These data support the hypothesis that ongoing DNA damage signaling that exists in cancer cells harboring mutant p53 may lead to increased abundance of mutant p53, and that genotoxic treatment increases the total number of cells with high p53 levels (and thus increases total p53 in the cell population as observed by immunoblot), but does not greatly increase p53 levels at the single cell level beyond the maximum already seen in untreated cells that harbor the highest levels of DNA damage signalling.

Figure 8. p53 abundance directly correlates with the intensity of DNA damage signaling in lung cancer cells.

(a) H1048 cells were untreated or treated with a 2 hour Dox pulse and fixed 3 hours following removal of the drug. Cells were stained with a p53 antibody (DO1) followed by staining with a secondary antibody (goat anti-mouse CFL-594) and Alexa Fluor 488 anti-γ-H2AX-Pi-Ser139 antibody. Cells were then gated into high and low levels of γ-H2AX-Pi staining (low gate= < median intensity untreated cells; high gate > median intensity Dox-treated cells), (b) The percent of cells within the low and high gates was analyzed.

Discussion

The increased abundance of mutant p53 in cancer cells contributes to its gain-of-function, so identifying the mechanisms that inhibit its normal turnover could lead to the identification of drugs that promote or restore degradation, which would conceivably disrupt the tumorigenic properties of mutant p53. We have analyzed factors leading to mutant p53 accumulation in lung cancer cells. Our data suggest that aberrant activation of checkpoint signaling in cancer cells may increase levels of mutant p53 by promoting MDM2-mediated monoubiquitination as opposed to polyubiquitination of mutant p53, thus limiting the delivery of mutant p53 to the proteasome, which requires a polyubiquitin chain (n>=4) for proper recognition of substrates for degradation.

This proposed dual role for MDM2 in regulating mutant p53 stability or degradation is somewhat contradictory to the widely held view that its sole function is inactivating and degrading p53 (wild type or mutant). In contrast, recent data indicates that the ability of MDM2 to suppress wild type p53 varies greatly, and is highly dependent on the relative expression levels of the two genes (34). Indeed, MDM2 can even act as a growth suppressor under certain conditions (35). We show for the first time that in the presence of activated DNA damage checkpoint signaling, MDM2 acts to increase mutant p53 abundance, by facilitating monoubiquitination of mutant p53, and inactivation of checkpoint signaling serves as a switch that turns MDM2 from an activator of mutant p53 abundance to facilitating its degradation. Given its dual role, these data suggest MDM2, itself, is not an ideal therapeutic target in mutant p53-expressing tumors, unless only one function (monoubiquitination) can be targeted vs. the other (polyubiquitination).

In support of the hypothesis that mutant p53 stability in at least some contexts could be related to chronic activation of DNA damage signaling, we found that caffeine sensitive S15 phosphorylation of mutant p53 was required for the preferential monoubiquitination of mutant p53 by MDM2, and treatment of lung cancer cells with caffeine both blocked monoubiquitination and promoted polyubiquitination of mutant p53. In addition, a p53-273H mutant defective for phosphorylation at S15 rescued polyubiquitination relative to a p53-273H allele that was wild type at S15, suggesting that S15 phosphorylation inhibits polyubiquitination. Moreover, caffeine treatment led to an MDM2 and proteasome-dependent reduction in mutant p53 levels also seen after siATM, all in the absence of any exogenous genotoxin. Indeed, phospho-ATM was detected in H1299 cells in the absence of exogenous DNA damage, suggesting that the checkpoint in these cells is chronically activated, at least under in vitro cell culture conditions.

Our analysis of cell populations at the single cell level also revealed a robust and direct relationship between DNA damage signaling and mutant p53 abundance. Mutant p53 abundance varied in proportion with the abundance of both activated p-1981-ATM and phospho-γH2AX. Moreover, treatment of cells with Dox increased the number of high expressing p53 cells in a population without changing the maximum intensity of p53 staining. These observations suggest ATM is active and available for DNA damage signaling in mutant p53 expressing cells, though does not speak to the relative activity of ATM compared with p53-null or WT p53 expressing cells.

Whereas ATM activation leads to phosphorylation of S15 and increased abundance of WT p53 (36), we have shown that S15 phosphorylation can also induce accumulation of mutant p53 by modulating ubiquitination. A prior report suggesting that mutant p53 actually blocked MRN activation of ATM and the G2/M checkpoint (37) did not, however, address the effect of ATM on mutant p53 abundance. Instead, we show that checkpoint activation mediated by double strand breaks through ATM leads to accumulation of mutant p53 by modulating ubiquitination, and our immunoblot (Figs. 2, 4, 5) and flow cytometry data (Figs. 6–8; Table 1) clearly shows active ATM as well as phosphorylated downstream targets in the absence or presence of exogenous DNA damage. In addition, and in agreement with our hypothesis, Song et al. speculate that mutant p53 is induced more highly following DNA damage, and that stabilization of mutant p53 is dependent on MDM2 (37).

Taken together, our data provide a model for increased mutant p53 abundance following DNA damage (Fig. 9). We suggest that following an oncogenic challenge, the DNA damage checkpoint becomes chronically activated and modifies p53 by ATM-dependent phosphorylation on serine 15, which then promotes MDM2-mediated monoubiquitination of p53 by a mechanism yet to be determined, arresting conversion to polyubiquitination, and promoting mutant p53 stabilization. Our data therefore suggest that in cancer cells expressing mutant p53, the DNA damage checkpoint could contribute to oncogenesis, as its activation leads to increased levels of mutant p53, potentially resulting in oncogenic GOF activities. Moreover, this mechanism for accumulation of mutant p53 could account for the selection of tumor subclones with enhanced aggressiveness and tumor progression. According to this model, cells with higher levels of genotoxic stress and DNA damage signaling would also have higher GOF p53 levels that could enhance their malignant properties and aid in the dissemination of mutated subclones with greater metastatic potential or chemoresistance.

Figure 9. Model for normal and aberrant regulation of WT and mutant p53 stability and activity after DNA damage.

Normal homeostasis of WT p53 is shown in the circle, and the steps leading to WT p53 stabilization and eventual return to homeostasis after stress, are shown outside the circle. Similar stabilization of mutant p53 following activation of the DNA damage checkpoint is shown on the far right, as well as the proposed mechanism for maintenance of aberrant mutant p53 stability following DNA damage due to lack of resolution of damage arising from the loss of p53 functions that contribute to physiologic repair and resolution of DNA damage.

Prior work investigating mutant p53 stability has demonstrated a strong link to Hsp90-mutant p53 interaction, which has been shown to block MDM2-dependent ubiquitination in various cancer cell lines (11). However, it is unclear if Hsp90 interaction with mutant p53 functions in all cancer cell types and with all mutant p53 alleles. The Hsp90 mechanism also does not account for the known transcriptional activities of GOF p53 (1) as these would presumably be inhibited by stoichiometric interaction with Hsp90. Taken with the Hsp90 data, our data suggest two possible ways that mutant p53 levels can be increased in cancer cells: through DNA-damage or possibly other signaling pathways that promote S15 phosphorylation and monoubiquitination, or by inhibition of ubiquitination via Hsp90 interaction.

Caffeine was only effective at reducing p53 levels in cells with generally higher mutant p53 levels to start with. Since the cells with higher mutant p53 levels are also those with more active DNA damage signaling (Fig. 7–8), inhibition of DNA damage checkpoint signaling may only lead to degradation of mutant p53 in the subpopulation of cells where mutant p53 has already been stabilized by the checkpoint response. Our data therefore complements the Hsp90 model by providing further insight into the detailed mechanism by which mutant p53 degradation is suppressed in cancer cells. Further analysis will be required to understand whether the Hsp90 and checkpoint signaling mechanisms operate simultaneously within the same cell or are mutually exclusive based on cell type, mutation type, or cell subpopulation factors, such as DNA damage signaling.

Highly specific second generation ATM inhibitors are growth inhibitory in cancer cells, but not normal cells (38). In addition, ATM inhibition leads to enhanced tumor xenograft responses in irradiated mice dependent on mutant p53 expression (38), also in potential contradiction to prior reports of mutant p53 inhibiting ATM (37). We now show a potential mechanism for this mutant p53-dependent therapeutic response, by virtue of the ability of ATM inhibitors to promote the degradation of mutant p53 in the highest expressing (and possibly the most malignant cells) cells in the tumor. We suggest that inhibition of the DNA damage checkpoint may be a way to selectively target the growth of cancer cells containing mutant p53, while having minimal negative impact on non-transformed cells.