Main

The p53 tumor suppressor is a promising target for cancer therapy; several compounds targeting p53 are currently being tested in clinical setting.1 In vivo studies support the idea of pharmacological restoration of p53 to combat cancer.2, 3, 4 Activation of p53 can lead to growth arrest, senescence or cell death, but elucidation of the molecular mechanisms driving the life/death decision by p53 remains one of the grand challenges in p53 biology.5 As the p53-mediated senescence or growth arrest can prevent cancer cell killing by chemotherapy thus leading to poor clinical outcome,6 it is imperative to understand the mechanism of p53-mediated cell fate decisions for the efficient clinical application of drugs activating p53.

We have previously shown that in spite of different transcriptional programs induced by p53 in breast cancer cells upon administration of different p53-activating compounds, p53 binds the same set of genes, irrespective of the type of treatment.7 This finding supports the view that the heterogeneous response and selective regulation of p53 target genes is likely to be influenced by other signal transduction pathways. A wealth of studies have looked into the p53 interactions with its partners and the type of p53 posttranslational modifications, but it still remains elusive, when, how and which factors direct p53 to a certain transcriptional program.5 A number of p53-modifying enzymes have been identified, including checkpoint kinases ATM/ATR, Chk2,5 as well as mitogen-activated protein kinases (MAPK) p38 and c-Jun N-terminal kinase JNK8 induced by oxidative stress.

Cancer cells frequently have increased burden of oxidative stress9 and therefore are likely to be more sensitive to the damage promoted by further reactive oxygen species (ROS) insults. Recent studies have revealed the dependency of cancers on redox-regulating mechanisms, such as the glutaredoxin and the thioredoxin systems, to be the cancer-specific vulnerability thereby offering a target for treatment of malignancies.9, 10 The NADPH-dependent selenoprotein thioredoxin reductase (TrxR), often overexpressed in cancer, is one of the promising anti-cancer drug targets, which is inhibited by several anti-cancer drugs in clinical use.11, 12

In the present study, we identified ROS-activated JNK as a crucial p53 co-regulator, revealing a strategy to switch the p53 transcriptional response from growth arrest to apoptosis upon its pharmacological activation.

Results

Transient versus sustained changes in gene expression upon p53-mediated growth arrest and apoptosis

To address the mechanisms of the differential biological outcome upon p53 activation, we used as molecular probes p53-reactivating molecules RITA and nutlin (Nut), which inhibit p53/MDM2 interaction.13 As a model, we applied a pair of cell lines, breast carcinoma MCF7 and colon carcinoma HCT116, in which activation of p53 by 10 μM Nut and 0.1 μM RITA leads to growth inhibition, whereas 1 μM RITA induces apoptosis.14, 15 As high doses of Nut induce p53-independent cell death (Supplementary Figure S1C), we used 10 μM Nut, inducing p53-dependent response.

We compared the kinetics of gene expression changes in MCF7 cells upon treatment with 10 μM Nut, 0.1 or 1 μM RITA at 10 time points using microarray analysis. Systematic clustering analysis showed that the genes involved in cell cycle regulation, metabolic and biosynthetic processes were continuously repressed upon 1 μM RITA; in contrast, these genes were only transiently repressed after Nut and 0.1 μM RITA treatment (clusters 0001 and 0002 in Figure 1a and Supplementary Table S1).

Figure 1
figure 1

Different kinetics of gene expression upon pharmacologically activated p53. (a) Systematic clustering analysis of gene expression profiles revealed distinct kinetics of transcription of several gene clusters upon treatment with different doses of RITA or Nut for 0, 2, 4, 6, 8, 10, 12, 16, 20, 24 h in MCF7 cells. The plots show mean profile (bold) and RMS deviation (gray) for genes, found to have clustered profiles upon 1 μM RITA. (b) Heatmaps of representative p53 target genes differentially regulated by 0.1 and 1 μM RITA in MCF7 cells. Values are normalized to untreated control

Full size image

Another gene cluster, comprising the stress response genes, was continuously induced by 1 μM RITA, but only transiently upregulated by Nut and 0.1 μM RITA (cluster 0004 in Figure 1a).

Next, we analyzed whether known p53 target genes are regulated in a similar differential manner. We found that 1 μM RITA led to the sustained induction of ENC1, GADD45A, PMAIP1, LIF and SESN1, and inhibition of pro-survival genes IGF-1R, MCL1, MYC, BCL2, PIK3CA and PIK3CB; however, changes in the expression of these genes were only transient upon 0.1 μM RITA (Figure 1b).

In conclusion, the induction of apoptosis was associated with the sustained p53-mediated transcriptional response. This prompted us to investigate the factors underlying this phenomenon.

ATM-independent induction of p53-dependent DDR upon 1 μM RITA

In line with the previous findings,16 we observed an increased phosphorylation of H2AX on Ser139 (γH2AX), a hallmark of DNA damage response (DDR), and phosphorylation of p53 on Ser15 (p-S15-p53) upon 1 μM RITA in a time-dependent manner in MCF7, HCT116 (Figure 2a) and U2OS cells (Supplementary Figure S1A). In contrast, low dose of RITA only barely affected the DDR (Figures 2a, 4a and 5b and Supplementary Figure S1A), as did Nut,17 which correlates with the inefficient apoptosis induction, as evidenced by cleaved PARP levels (Figure 4a and Enge et al.14 and Grinkevich et al.15).

Figure 2
figure 2

p53- and dose-dependent induction of DDR by RITA in cancer cells, but not in non-tumorigenic cells. (a) MCF7 and HCT116 cells were treated with 0.1 and 1 μM RITA for indicated periods and levels of p53, γH2AX and p-S15-p53 were detected by immunoblotting. Actin was used as a loading control. (b) RITA did not induce γH2AX in non-tumorigenic MCF10A and 184A1 cells as assessed by immunoblotting. (c) Wild-type p53-expressing cells HCT116 and p53-null HCT116 p53−/−, H1299, Saos2 cells were treated with 1 μM RITA for indicated periods (upper and lower panels); HCT116 cells were pretreated with pifithrin-α for 2 h and treated with 1 μM RITA for 12 h (middle panel). Proteins were analyzed by western blotting. (d) HCT116 cells were treated with 1 μM RITA or doxorubicin for 12 h and DNA strand breaks were assessed by comet assay. Right panel, quantification of cells containing strand breaks (mean±S.E.M., n=3)

Full size image

Importantly, the induction of γH2AX by RITA was observed only in the presence of p53, that is, in the p53-positive HCT116 cells, but not in the p53-null HCT116 p53−/−, osteosarcoma Saos2 and lung adenocarcinoma H1299 cells (Figure 2c, left and right panel). The p53 dependence of γH2AX induction was further supported by the ablation of γH2AX by the p53 inhibitor pifithrin-α (PFTα)18 and upon p53 depletion with siRNA (Figure 2c, middle panel and Supplementary Figure S1B, respectively). We ruled out the possibility that DDR was induced upon DNA fragmentation during apoptosis, as the pretreatment with a pan-caspase inhibitor Z-VAD-fmk did not prevent γH2AX, whereas it did prevent PARP cleavage (Supplementary Figure S1D).

Alkaline comet assay revealed that a ‘comet tail’, indicating DNA strand breaks, was barely detectable upon 1 μM RITA, whereas positive control doxorubicin produced a clear pattern (Figure 2d). We did not detect DNA strand breaks using pulse-field electrophoresis assay either (Supplementary Figure S1E). Thus, in line with the previously published data,16, 19 RITA produces a low number of strand breaks, if any. Importantly, RITA did not induce the DDR and apoptosis in the non-tumorigenic cells MCF10A and 184A1 (Figure 2b and Supplementary Figure S2F).

To find out the mechanism of DDR induction, we tested the involvement of checkpoint kinases. Depletion of ATM by siRNA did not prevent γH2AX and p53 accumulation, neither did the pretreatment with the ATM inhibitor KU55933 or with the major DDR kinase inhibitor caffeine (Supplementary Figures S1F–H), ruling out the involvement of these kinases. However, the kinetics and the extent of p53 accumulation were partially affected by caffeine (Supplementary Figure S1H), suggesting that DDR contributes to the faster and robust induction of p53, perhaps via amplification of the signaling to p53.

In conclusion, the induction of DDR was p53-, but not ATM/ATR-dependent and correlated with the induction of apoptosis.

Generation of ROS leads to DDR and confers synthetic lethality upon p53 reactivation

Since ROS can cause DDR,20 we reasoned that the induction of γH2AX could be due to the p53-dependent induction of ROS resulting from the inhibition of TrxR1 by 1 μM RITA, reported previously by us.21 More detailed analysis of the effect of RITA on TrxR1 in in vitro enzymatic assay revealed that while RITA inhibited the reducing activity of TrxR1 on two different substrates, it barely affected its NADPH oxidase function (Figure 3a), which endows the enzyme with pro-oxidant activity.22, 23 Thus, both the inhibited reductase and the sustained oxidase activities of TrxR1 upon RITA should contribute to ROS accumulation. Indeed, 1 μM RITA induced intracellular ROS in MCF7 and HCT116 cells, whereas a low dose of RITA or Nut did not trigger ROS (Figure 3b, Supplementary Figures S2A, S3A and B). In line with our previously published results,21 ROS were not induced in non-transformed MCF10A cells (Supplementary Figures S2E and S3C), correlating with the absence of DDR. Thus, DDR was associated with the induction of ROS.

Figure 3
figure 3

Induction of ROS leads to DDR and p53-dependent apoptosis. (a) Dose-dependent inhibition of the ability of TrxR1 to reduce 5,5′-dithiobis (2-nitrobenzoic acid) (DTNB) and thioredoxin (Trx), but not its NADPH oxidase activity by RITA, as measured in in vitro assay using purified TrxR1. (b) ROS were measured in MCF7 cells treated for 6 h with 0.1 and 1 μM RITA (upper left panel), Nut or 400 μM H2O2 (upper right panel) and 5 μM TrxR inhibitor Aura as a positive control (lower panel). (c) Antioxidants NAC and resveratrol prevent the induction of ROS by 1 μM RITA. (d) Antioxidants NAC (upper panel) and resveratrol (lower panel) inhibited the induction of γH2AX and PARP cleavage by RITA as analyzed by western blotting. (e) MCF7 and HCT116 cells were treated with 2 μM Aura or 5 μM Nut, or their combination (with or without NAC pretreatment) for 48 h; cells were photographed before proceeding to crystal violet staining. (f) HCT116 cells were treated with 2 μM Aura, 10 μM Nut or 0.1 μM RITA, or their combination (with or without NAC pretreatment) for 48 h; then cells were harvested and proceeded to double staining with Annexin V and propidium iodide (PI) followed by FACS analysis (mean±S.E.M., n=3). PI only: necrotic and/or late apoptotic cells; PI+Annexin V: apoptotic cells; Annexin V only: early apoptotic cells; *expected additive effect. (g) U2OS cells were treated with 1 μM Aura or 0.05 μM RITA, or their combination (with or without NAC or 10 μM SP600125 pretreatment) for 48 h and assessed as in (f) (mean±S.E.M., n=3); *expected additive effect

Full size image

Next, we addressed the question whether generation of ROS is the cause of DDR and whether it contributes to p53-mediated cell death. We found that ROS scavenger N-Acetyl-L-cysteine (NAC), as well as a low dose (1 μM) of antioxidant resveratrol24 inhibited ROS induced by RITA (Figure 3c, Supplementary Figures S2B, S3A and B). Notably, both antioxidants prevented the induction of γH2AX (Figure 3d and Supplementary Figure S2C), supporting our idea that the induction of DDR is triggered by ROS.

However, NAC or resveratrol did not prevent the accumulation of p53 by RITA, indicating that p53 induction is not due to ROS.

NAC and resveratrol partially reversed the apoptosis triggered by RITA, as evidenced by the decreased PARP cleavage and the rescue of cell viability (Figure 3d and Supplementary Figures S2C and S4A). These data corroborated the involvement of ROS in p53-mediated apoptosis. Accordingly, fluorescence-activated cell sorting (FACS) analysis of propidium iodide (PI) stained cells confirmed that resveratrol, as well as another potent antioxidant nordihydroguaiaretic acid (NDGA)25 significantly inhibited apoptosis upon RITA (Supplementary Figure S2D).

To reinforce the role of ROS as a possible denominator of the apoptotic response upon p53 reactivation, we examined whether blocking TrxR by auranofin (Aura), a well-known inhibitor of TrxR26 and inducer of ROS (Figure 3b and Supplementary Figure S2A), could convert the growth arrest/senescence induced by Nut into apoptosis. As evident from Figure 3e, a low dose of Aura was synthetic lethal with p53 activation by low dose Nut. This is manifested by the robust induction of apoptosis upon Aura/Nut combination, while both agents barely induced cell death, when taken alone; importantly, NAC reverted the synthetic lethality induced by the combination of Nut and Aura (Figures 3e and f, and Supplementary Figure S4C). We also observed a synergistic effect upon the combination of low dose of RITA and Aura; similarly, NAC reverted the synthetic lethality upon low dose of RITA and Aura as well (Figures 3f and g and Supplementary Figure S4C).

To elucidate whether ROS contribute to the apoptosis induction upon other types of pharmacological activation of p53, we used cisplatin (CDDP), which has been shown to activate p53 and inhibit the TrxR enzyme activity.27, 28 We found that pretreatment with NAC, resveratrol or NDGA prevented the elimination of tumor cells by CDDP (Supplementary Figure S4B).

Taken together, our data suggest that disabling the oxidative defense mechanisms in cancer cells, for instance, via TrxR inhibition, is synthetic lethal when combined with the pharmacological restoration of p53.

Activation of JNK triggered by ROS mediates the synthetic lethality upon p53 activation and inhibition of TrxR

Next, we investigated which factor mediates the synthetic lethal effect of p53 activation combined with TrxR inhibition. The activation of MAP kinase JNK upon inhibition of TrxR29 and its ability to modulate p53 make JNK an attractive candidate mediator of ROS signaling to p53. We found that RITA induced JNK phosphorylation (p-JNK) in a p53- and dose-dependent manner (Figures 4a and b). Notably, the induction of p-JNK was conferred by the increased ROS levels, as NAC and resveratrol inhibited JNK activation (Figure 4c and Supplementary Figure S5A). Importantly, JNK serves as a critical mediator of the p53-induced apoptosis, as evidenced by the rescue of PARP cleavage, growth suppression and subG1 fraction by JNK inhibitor SP600125 and siRNA-mediated depletion of JNK (Figures 4d and h and Supplementary Figures S5B and C).

Figure 4
figure 4

ROS-mediated activation of JNK contributes to the p53-mediated apoptosis, DDR and transcriptional repression of oncogenes. (a) Dose-dependent induction of p-JNK, p-Ser33-p53, p-Ser15-p53, γH2AX, PARP cleavage and inhibition of Wip1, Mcl1 and MdmX by RITA as assessed by western blotting. (b) HCT116 and HCT116 p53−/− cells were treated with 1 μM RITA; analyzed as in (a). (c) Pretreatment with NAC for 6 h prevented the induction of p-JNK, γH2AX, PARP cleavage and inhibition of Wip1, Mcl1 and MdmX by RITA as analyzed by immunoblotting. (d) JNK inhibitor SP600125 prevented the induction of p-JNK, p-Ser33-p53, γH2AX, PARP cleavage and inhibition of Wip1, Mcl1 and MdmX by RITA, as assessed by western blotting. (e) SP600125 blocked the induction of p-JNK, p-Ser33-p53, γH2AX, PARP cleavage and inhibition of MdmX and Wip1 by combination treatment with 0.05 μM RITA and 1 μM Aura for 24 h, as assessed by western blotting.(f, g) Depletion of JNK by siRNA prevented the induction of γH2AX (f), p-Ser33-p53 and inhibition of Wip1 and MdmX (g), analyzed by immunoblotting. (h) Inhibition of JNK by siRNA prevented apoptosis induction by RITA, as measured by FACS of PI-stained cells. (i) SP600125 blocked the repression of MCL1, PPM1D, PIK3CA, PIK3CB, EIF4E and MDM4 (MdmX) mRNA upon RITA treatment as assessed by qPCR (mean±S.E.M., n=3) in HCT116 (12 h treatment) and MCF7 (8 h treatment) cells

Full size image

Furthermore, we found that the synthetic lethal effect of the low dose of RITA/Aura combination is mediated by p-JNK. Aura combined with the low dose of RITA led to the robust induction of p-JNK. Notably, the induction of apoptosis upon this combination treatment, manifested as PARP cleavage, was prevented by JNK inhibitor (Figure 4e).

Thus, we concluded that JNK is a crucial player downstream of ROS in the molecular pathway leading to the synthetic lethality upon p53 activation combined with TrxR inhibition.

JNK activated by ROS contributes to the induction of DDR and converts p53 into an efficient inhibitor of oncogenes

Next, we focused on elucidating the mechanisms by which JNK enhances p53-induced apoptosis. As JNK has been shown to mediate the UV-induced γH2AX,30 we assessed the role of JNK in γH2AX accumulation. JNK inhibitor SP600125, as well as JNK depletion by siRNA, markedly reduced γH2AX (Figures 4d and f and Supplementary Figure S5B), implying that JNK is the critical kinase mediating DDR. In addition, JNK mediated p53 phosphorylation at Ser33 (Figures 4d, e and g and Supplementary Figure S5D).

Strikingly, the p53-dependent downregulation of several oncogenes, anti-apoptotic factor Mcl1 and p53 inhibitors Wip1 and MdmX (Figure 4b, Supplementary Figure S1B), was rescued by NAC, resveratrol and JNK inhibitor (Figures 4c and e and Supplementary Figures S5A and B). Further evidence of the crucial role of ROS and JNK in oncogene downregulation by p53 was provided by the experiments in which we used the combination of the low dose of RITA and Aura. Although Wip1 expression was increased upon the low dose of RITA, its induction was partially prevented by Aura (Figure 4e). Moreover, the combination of low dose of RITA and Aura led to a dramatic downregulation of MdmX. Notably, Aura-mediated downregulation of Wip1 and MdmX was rescued by JNK inhibitor (Figure 4e).

The JNK-dependence of Wip1 and MdmX downregulation was further supported by the JNK depletion experiments (Figure 4g and Supplementary Figure S5D).

Interestingly, the p53 level was largely unaffected by the JNK inhibitor, as well as the induction of its target genes Puma and Noxa (Figures 4c and d and Supplementary Figures S5A and B).

Next, we assessed whether the observed inhibition of oncogenes occurs on mRNA level. In line with our previous results,15 RT-qPCR experiments demonstrated that pharmacological activation of p53 led to a decreased mRNA level of a set of oncogenes, including MCL1, Wip1-encoding PPM1D, MDM4 (MdmX), as well as PIK3CA and PIK3CB, encoding catalytic subunits of PI3 kinase, and translation factor EIF4E (Figure 4i). The rescue of oncogene inhibition by JNK inhibitor corroborated the key role of JNK (Figure 4i). In addition, we have previously shown that MCL1, PIK3CA and PIK3CB are not downregulated by the low dose of RITA or nultin; however, their inhibition converts the p53-mediated growth arrest into apoptosis.15, 31

Therefore, we concluded that the induction of ROS upon RITA leads to the activation of JNK that mediates the phosphorylation of H2AX at Ser139, phosphorylation of p53 at Ser33 and the inhibition of the expression of a set of pro-survival oncogenes by p53, conferring apoptosis induction.

Next, we addressed the question whether and how the repression of PPM1D, downstream of JNK, contributes to the synthetic lethal effect.

Inhibition of Wip1 by p53 promotes the induction of γH2AX

p53 activation is opposed by Wip1, an oncogene that removes inactivating phosphorylation marks from Mdm2 and activating phosphorylation marks from p53, p14/p19ARF and checkpoint kinases,32 and dephosphorylates γH2AX,33, 34 thereby attenuating DDR. Moreover, Wip1 is a p53 target gene that serves as one of the critical determinants of the biological outcome.34, 35 These data and our present results showing that Wip1 activation was induced by Nut and low dose of RITA (Supplementary Figures S6D and E and Figure 4e, respectively), but was repressed by 1 μM RITA in the JNK-dependent manner, prompted us to test how inhibition of Wip1 contributes to p53 activity.

We found that the decline of Wip1 mRNA and protein levels upon RITA strongly correlated with γH2AX induction (Figures 5a and b).

Figure 5
figure 5

Inhibition of Wip1 promotes the induction of γH2AX upon RITA treatment. (a) Wip1 mRNA was repressed after 8 h treatment with 1 μM RITA, but not 0.1 μM RITA, as assessed by qPCR (mean±S.E.M., n=3). (b) Downregulation of Wip1 protein level correlated with the induction of γH2AX upon RITA treatment as analyzed by immunoblotting. (c) MCF7 and U2OS cells stably transfected with empty vector shRNA or Wip1 shRNA were treated with 0.1 and 1 μM RITA for indicated periods and γH2AX was assessed as in (b). (d) HCT116 and U2OS cells transfected with either empty vector or FLAG-Wip1 were treated with 1 μM RITA for indicated times. Proteins were detected by western blotting

Full size image

Inhibition of Wip1 has a role in DDR induction upon RITA, as its depletion by shRNA in combination with the low dose of RITA significantly increased the level of γH2AX, comparable with the level induced by 1 μM RITA (Figure 5c).

The negative contribution of Wip1 to DDR was supported by a significantly reduced induction of γH2AX by RITA upon Wip1 overexpression, as assessed by western blotting (Figure 5d).

Wip1 depletion enhances the transactivation function of p53 and contributes to the synthetic lethality

We reasoned that the enhancement of DDR resulting from Wip1 downregulation could facilitate the p53 transcriptional activity leading to a more robust response. To assess this notion, we analyzed the p53 transcriptional response upon Wip1 silencing in MCF7 cells treated with the low dose (0.1 μM) RITA for 4 and 16 h using microarray analysis.

Wip1 depletion per se only weakly affected the expression of p53 target genes (Figure 6a). Low dose of RITA induced several p53 targets including GPRC5A,7 TNFRSF10B (KILLER/DR5), FAS, RPRM and ENC1 (PIG10);36 however, their induction was not observed at a late time point. Unsupervised clustering analysis indicated similarity of the profiles obtained at 16 h of treatment with control samples (Figure 6a).

Figure 6
figure 6

Depletion of Wip1 confers a sustained transcriptional activation of p53 target genes, but does not facilitate transrepression. (a and b) Microarray analysis of MCF7 cells with (indicated in violet) or without (indicated in gray) Wip1 depletion by shRNA, treated with 0.1 μM RITA or DMSO for indicated time points revealed that p53-mediated transactivation was enhanced by Wip1 silencing. (c) Wip1 downregulation led to the increased induction of p53-activated genes (upper panel) but did not augment the repression of pro-survival genes by p53 (lower panel) upon low dose of RITA as analyzed by qPCR (mean±S.E.M., n=3). Insert demonstrates the efficiency of Wip1 depletion, as assessed by immunoblotting. (d) MCF7 cells transfected with either empty vector or FLAG-Wip1 were treated with 1 μM RITA or DMSO for 8 h and mRNA levels of PIK3CA, PIK3CB and IGF-1R were assessed by qPCR (mean±S.E.M., n=3). (e) MCF7 cells stably transfected with shWip1 or control shVector were treated with 0.1 and 1 μM RITA or DMSO for 24 h, and cells were stained with Annexin V followed by FACS analysis (mean±S.E.M., n=3, *P<0.05, **P<0.01, by two-tailed t-test). (f) Model of the synthetic lethality upon activation of p53 and inhibition of TrxR. Inhibition of TrxR lead to the accumulation of ROS and activation of JNK, facilitating p53 function upon its release from Mdm2. In turn, activated p53 induces pro-oxidant genes, which increases the level of ROS, further activating JNK, and thus, p53. Activated JNK converts p53 to an inhibitor of Wip1 and MdmX, therefore amplifying p53 activity. Transcriptional repression of Mcl1, eIF4E and PI3K abolishes survival signaling, contributing to apoptosis induction. Thus, the dual targeting of p53 and TrxR (i.e., by RITA) leads to the robust apoptosis

Full size image

Notably, Wip1 depletion by shRNA facilitated the induction of several p53 targets by the low dose of RITA and rescued their decline at 16 h (Figure 6a). This indicates that the inhibition of Wip1 boosts transactivation by p53 and confers its sustainability.

qPCR experiments confirmed that upon Wip1 silencing the expression of p53 targets FAS, GDF15 and BTG2 was substantially induced, further elevated upon 0.1 μM RITA and lasted longer (Figure 6c, upper panel), confirming that the inhibition of Wip1 leads to an enhanced and sustained p53-mediated transactivation.

However, Wip1 silencing did not facilitate the downregulation of the pro-survival genes (Figures 6b and c, lower panel), which are inhibited by 1 μM RITA in a p53-dependent manner.15 In line with these data, ectopic expression of Wip1 did not prevent RITA-induced downregulation of Mcl1 (Figure 5d). Further, we found that Wip1 overexpression did not relieve the repression of PIK3CA, PI3KCB and IGF-1R (Figure 6d). These results suggest that in contrast to p53 transactivation function, p53-mediated inhibition of oncogenes is less tightly regulated by Wip1.

Next, we addressed whether inhibition of Wip1 contributes to the biological response triggered by p53. Indeed, Wip1 depletion promoted the apoptosis induced by both 0.1 and 1 μM RITA, as shown in Figure 6e and Supplementary Figures S6A and B.

As shown in Figure 4e and Supplementary Figure S6E, Aura partially prevented the induction of Wip1 by low RITA or Nut in U2OS cells as well as triggered the downregulation of Mdm2 and MdmX and apoptosis (Supplementary Figures S6E and F). We observed a prominent induction of cell death associated with the increase of γH2AX, p53 and decrease of MdmX in the Wip1-silenced cells upon Aura treatment (Supplementary Figures S6C and D). Induction of cell death by Aura in Wip1-depleted cells was comparable with the extent of apoptosis upon Nut/Aura combination, which underscores the crucial role of Wip1.

Taken together, our data demonstrate that the inhibition of Wip1 significantly contributes to the synthetic lethal effects upon p53 activation and TrxR inhibition.

Discussion

Elucidation of the molecular mechanisms governing the cellular responses elicited by p53 is still a challenge in the p53 field, limiting the effective harnessing of p53 activity for cancer treatment.5 Our present study revealed the crucial role of ROS and activated JNK in p53’s proapoptotic function in cancer cells upon pharmacological activation of p53. Further, we elucidated the JNK-mediated mechanisms that play a role in this process and identified a set of key factors downstream of p53 affected by JNK that confer cell death outcome.

Our previous study established that the inhibition of anti-apoptotic factor Mcl1, as well as catalytic subunits of PI3 kinase, translational factor eIF4E and p53 inhibitor MdmX, all well-known oncogenes and a high-priority anti-cancer targets,37, 38, 39, 40 is crucial for the robust apoptosis induction by p53.15 However, it remained unclear which factors control the inhibition of survival genes by p53. Here, we demonstrated that the ablation of this set of oncogenes by p53 is JNK-dependent.

Despite the extensive efforts, we still have much to learn about the molecular mechanisms of p53-mediated transcriptional repression.36 p21 has been implicated in the p53-mediated repression of a number of genes.36 However, p21 is depleted upon p53 activation by RITA, which contributes to the induction of apoptosis,14 ruling out the possibility that the observed transcriptional repression is p21-dependent.

p53 might have a direct role in the transcriptional repression of several oncogenes, including EIF4E, PPM1D and MDM4, as evidenced by p53 binding to its consensus motif within a short distance from the transcriptional starting site of these genes (Supplementary Table S2), identified by us using p53 chromatin immunoprecipitation followed by deep sequencing (ChIP-seq) described in Nikulenkov et al.7 The binding of p53 to MYC and MCL1 promoters we have previously described.15, 31 However, we cannot rule out an indirect effect of p53 that might be attributed to the p53-regulated microRNAs or other p53-upregulated factors.

We have previously found that p53-activated gene PPM1D, whose product Wip1 inhibits p53 signaling,32 is repressed upon RITA.35 However, the mechanism of this intriguing phenomenon has not been identified. Our present study implicates JNK as a crucial factor that can convert p53 from a transactivator to a repressor of Wip1.

Based on these results, we suggest a model illustrating how pharmacological release of p53 from the Mdm2 complex, combined with the inhibition of TrxR, results in the excessive accumulation of ROS that leads to further p53 activation by JNK (Figure 6f). In turn, active p53 induces pro-oxidant genes, such as PUMA and PIGs, increasing ROS and further feeding activating signals to JNK and thus, itself. p-JNK converts p53 to an efficient inhibitor of pro-survival oncogenes Mcl1, eIF4E and PI3 kinase, which contributes to apoptosis induction. Further, two p53’s own inhibitors, Wip1 and MdmX, are repressed in the p-JNK-dependent manner, which amplifies p53 activation. Establishment of the JNK-p53 positive feedback loop and the inhibition of p53-Wip1 negative feedback loop result in the enhanced and sustained p53 activation. This produces a robust apoptotic outcome, leading to the effective elimination of cancer cells.

Our study also indicates that the induction of ROS might be the cause of the previously observed DNA–DNA and DNA–protein cross-links upon RITA treatment.19

Molecular pathways by which reconstituted p53 becomes proapoptotic selectively in malignant tumors are a subject of debate.2, 3, 41, 42, 43 Reinstatement of p53 efficiently eliminates advanced lung carcinoma cells whereas leaving early lesions unperturbed. This is due to the amplified stress-activated MAPK signaling in advanced lesions, which engages the MDM2 inhibitor p19ARF (p14ARF in humans), in turn activating p53.2, 3 Notably, another in vivo study has shown that JNK is required for p53 induction upon oncogene activation.43

Our data suggest that elevated ROS in malignant tumors might provide an activating signal to p53 via JNK leading to the enhanced and sustained p53 activity. It is tempting to speculate that this may constitute a basis for the selective elimination of advanced cancers by the reinstated p53, observed in mouse models.2, 3 The relative contribution of the ROS/MAPK pathway in oncogenic signaling and preferential suppression of malignant tumors by p53 is an interesting subject for future studies.

Our results demonstrating that pharmacological activation of p53 in combination with TrxR inhibition and ROS induction confers synthetic lethality could be an important consideration for the clinical application of p53-reactivating drugs. Indeed, we showed that growth arrest/senescence by Nut, which is now being tested in clinic, could be converted to apoptosis upon the low dose of TrxR inhibitor Aura. Furthermore, Aura is an FDA approved drug for the treatment of rheumatoid arthritis; therefore, its repositioning for cancer therapy through the combination strategy as shown in this study will save the time and the cost for developing more effective cancer treatment approaches.

One of the important biochemical differences between normal and cancer cells is a decreased capability of cancer cells to buffer high ROS levels. Our study suggests that dual targeting of p53 and the cellular antioxidant system might allow to maximally exploit the p53-mediated tumor suppression as a therapeutic strategy.

Materials and Methods

Cell culture

Colon carcinoma cell lines HCT116 and HCT116 TP53−/− were kindly provided by B. Vogelstein, John Hopkins University, USA. p14ARF negative HCT116, HCT116 TP53−/−, breast carcinoma MCF7, osteosarcoma U2OS cells were grown under standard conditions; mammary epithelial cell line MCF10A and 184A1 were obtained from Serhiy Souchelnytskyi, Karolinska Institutet. MCF10A cells were kept in medium containing 50% of MEBM (Clonetics, Basel, Switzerland), 50% of Nutrient Mixture F-12 HAM (Sigma, St. Louis, MO, USA) and 5% horse serum supplemented with MEGM SingleQuot (Clonetics); 184A1 cells were kept in MEGM complete medium supplemented with 5% horse serum.

Plasmids, shRNA and siRNA

Plasmids encoding FLAG-Wip1 and shRNA for Wip1 were kindly provided by René H. Medema, Utrecht, Netherlands. ATM and p53 siRNA were from Santa Cruz Biotechnology (Dallas, TX, USA). Custom siRNA targeting both JNK1 and JNK244 was synthesized by Thermo Scientific Dharmacom (Waltham, MA, USA). Plasmid transfection was performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) and siRNA was transfected using HiPerFect (Qiagen, Hilden, Germany) according to the manufacturer’s instructions.

Drug treatments

RITA was obtained from the National Cancer Institute (NCI). Nut and caffeine from Calbiochem (Billerica, MA, USA) were used at 10 μM and 4 mM, respectively. Resveratrol (BIOMOL International, Farmingdale, NY, USA) was used at 1 μM. Pifithrin-α, SP600125, neocarzinostatin, cisplatin (CDDP), H2O2 and NAC (all from Sigma) were used at 10 μM, 40 μM, 200 ng/ml, 50 μM, 400 μM and 5 mM, respectively. Pan-caspase inhibitor Z-VAD-FMK (R&D Systems, Minneapolis, MN, USA) and KU55933 (Tocris Bioscience, Bristol, UK) were used at 10 μM. NDGA, a kind gift from O. Rådmark and Doxorubicin from B. Zhivitovsky, (both from Karolinska Institutet) were used at 10 and 2 μM, respectively.

Molecular and cell biology assays

Western blotting was performed according to a standard procedure. The following antibodies were used: anti-p53 (DO-1), PARP-1/2, Mcl1, c-Myc, phospho-p53 (Ser15), phospho-p53 (Ser33), phospho-JNK (Cell Signaling, Danvers, MA, USA), phospho-histone H2AX (Ser139) (Millipore, Billerica, MA, USA), ATM (Abcam, Cambridge, UK), Wip1 (Bethyl Laboratories, Montgomery, TX, USA), Noxa, Puma (Calbiochem), β-actin (Sigma). After transfer membranes were cut to detect several proteins on the same membrane; in Figure 4 and Supplementary Figure S5, the proteins were detected in the following order: (1) Wip1, p-JNK (30 min exposure), γH2AX; (2) MdmX, p-S15-p53, actin, Puma; (3) PARP, p-S33-p53, Mcl1, Noxa; (4) p53.

Alkaline comet assay, FACS analysis of propidium iodide (PI)-stained cells and qPCR were performed as described Imreh et al.45 and Enge et al.14 FACS analysis of FITC-Annexin V and PI (from BD Biosciences, San Jose, CA, USA) double-stained cells was performed according to the manufacturer’s protocols. Detection of activated caspases by FAM-FLICA Poly-Caspase Detection Kit from ImmunoChemistry Technologies (Bloomington, MN, USA) was performed using FACS according to the manufacture’s protocol.

Primers are described in Supplementary Table S3.

ROS measurement

ROS were measured as in Hedstrom et al.21 Briefly, after treatment with different compounds, cells were washed once with serum-free medium and incubated with 10 μM 2′,7′-dichlorodihydrofluorescein diacetate (DCF-DA) in serum-free medium for half an hour under standard conditions; then cells were washed twice with PBS, trypsinized, harvested and washed another two times with PBS. The samples were sorted on Becton Dickinson FCAScan using FL1-H channel, and analyzed by CellQuest (San Jose, CA, USA) software 4.0.2.

Microarray analysis

Systematic clustering of gene expression data was performed with CRC clustering method, implemented in geneXplain platform (www.genexplain.com).46 When applied to 2000 10-point profiles with the largest fold change, the method gives about 10 high quality clusters with more than 40 profiles in them, and with the gene lists of the clusters enriched in GO terms at P-value<0.002. To compare gene expression profiles under different treatments, we plot the mean expression profile of the clustered genes in one case and the mean profile of the same (not clustered) genes in other cases. This allowed detecting of conserved and variable features.

p53 ChIP-seq

The detailed description of p53 ChIP-seq experiment is provided in Nikulenkov et al.7 The analysis of p53 binding to inhibited genes was performed as described in Zawacka-Pankau et al.31

In vitro assay of TrxR1 activity

In this inhibition analysis, 50 nM wild-type TrxR1 (24.0 U/mg) was reduced by 150 μM NADPH in TE buffer (pH 7.5) and then RITA at defined concentration ranging from 0.1 to 200 μM was incubated with the reduced enzyme in the dark at room temperature for 60 min. DMSO instead of RITA was used as control. Small aliquots of RITA-inhibited enzyme were taken out at 60-min time point for measuring the DTNB reduction activity and the NADPH-reduced enzyme was used as control. After TrxR1 activity was inhibited, 500 μl RITA-inhibited TrxR1 (50 nM) was applied onto TE-equilibrated NAP-5 column (GE Healthcare Life Sciences, Uppsala, Sweden) and then eluted with 1 ml TE buffer. TrxR1 activities using DTNB (2.5 mM), human Trx1 (20 μM) and juglone (50 μM) as substrates were measured after desalting. DTNB reduction was measured at 412 nm (extinction coefficient of 13 600 M/cm) by adding the desalted enzyme respectively into the DTNB reaction mixture (200 μl) containing 2.5 mM DTNB, 300 μM NADPH and 4.5 nM enzyme in TE buffer (pH 7.5). Trx-coupled insulin assay was performed by measuring the NADPH consumption at 340 nm (extinction coefficient of 6220 M/cm) and reaction system contains 20 μM human Trx1, 160 μM insulin, 300 μM NADPH and 9 nM enzyme in TE buffer (pH 7.5). The NADPH oxidase activity was monitored as the decrease at 340 nm (extinction coefficient of 6220 M/cm) for 60 min. The reaction system (200 μl) contains 50 μM juglone, 200 μM NADPH and 4.5 nM enzyme in TE buffer (pH 7.5). Enzymatic reactions and measurements were performed with 10-sec time interval reads at 25 °C using a VersaMax microplate reader (Molecular Devices, Sunnyvale, CA, USA), with the reaction mixtures without enzyme serving as reference. Activity measurements were performed in triplicate and analyzed with the Prism 5 software (GraphPad, La Jolla, CA, USA).