MDM2-Mediated Degradation of p14ARF: A Novel Mechanism to Control ARF Levels in Cancer Cells

We here show a new relationship between the human p14ARF oncosuppressor and the MDM2 oncoprotein. MDM2 overexpression in various cancer cell lines causes p14ARF reduction inducing its degradation through the proteasome. The effect does not require the ubiquitin ligase activity of MDM2 and preferentially occurs in the cytoplasm. Interestingly, treatment with inhibitors of the PKC (Protein Kinase C) pathway and use of p14ARF phosphorylation mutants indicate that ARF phosphorylation could play a role in MDM2 mediated ARF degradation reinforcing our previous observations that ARF phosphorylation influences its stability and biological activity. Our study uncovers a new potentially important mechanism through which ARF and MDM2 can counterbalance each other during the tumorigenic process.


Introduction
Knowledge of the mechanisms governing the unbalance between tumor suppressors and oncogenes, is critical to understanding the pathogenesis and evolution of cancer.
Among the most important tumor suppressors, p14ARF (mouse p19ARF) (Alternative Reading Frame) appears to play a major role, being its direct contribution to cancer largely demonstrated over the last years [1,2].
ARF is involved in oncogenic checkpoint by sensitizing incipient cancer cells to undergo growth arrest or apoptosis. There is growing evidence that ARF signaling is complex, and involves p53-dependent or independent pathways aiming mainly at restraining abnormal cell growth and at maintaining genomic stability. The discovery of a plethora of ARF interactors and the observation that also viral, genotoxic, hypoxic and oxidative stresses activate an ARF response, suggest that ARF has a wider role to protect the cell [3]. Primary cells in normal conditions maintain ARF at low levels; however, when cells are stimulated by oncogenic insults, its concentration in the cell drastically increases. This phenomenon is generally accompanied by a parallel disruption of the inhibitory interaction between Mdm2 and p53, resulting in the accumulation of transcriptionally active p53 that stimulates either apoptosis or cell cycle arrest [4,5]. The strong ARF effect on cell proliferation requires that cells should have developed mechanisms that can promptly reduce ARF intracellular levels when its activity is no more required. However, the mechanisms that regulate ARF turnover are only recently going to be elucidated. ARF degradation depends, at least in part, on the proteasome and, although ARF lacks lysine residues in its sequence, it can undergo N-terminal ubiquitination independently of Mdm2 and p53 [6]. Quite recently, a specific ubiquitin ligase for ARF called ULF was identified [7]. On the other hand, there are evidences that ARF can be degraded in vitro by the 20S proteasome in the absence of ubiquitination and that this process can be counteracted by TBP-1 (Tat Binding Protein 1), a multifunctional component of the regulatory subunit of the proteasome [8]. Furthermore, the REG γ proteasome has been implicated in the ubiquitin-independent regulation of p19ARF turnover, supporting the notion that ubiquitination could be not necessarily implicated in ARF turnover [9].
Interestingly, we and others recently demonstrated that, following PKC (Protein Kinase C alpha) activation, ARF levels increase [10,11]. Furthermore, a point mutation that mimicks phosphorylation in the conserved Thr8 induces ARF accumulation mainly in the cytoplasm and inhibits its biological activity [11].
So far, ARF subcellular localization appears to play an important role in its stability and biological functions, although in not unequivocal manner. It appears that nucleolar localization of ARF may serve for its storage or stabilization [12,13]. In the nucleolus, ARF assumes a stable structure thanks to its association to B23/NPM, while in the nucleoplasm it is subjected to a more rapid turnover. In some cases, the increase in ARF levels causes Mdm2 to relocate to the nucleolus [14,15] and this has been linked to p53 stabilization. Others reported that, although nucleolar localization of ARF causes its stabilization, this is not essential to regulate ARF's activity towards p53 [16,17]. Interestingly, it has been reported that non-nucleolar forms of ARF are subjected to rapid degradation by the proteasome, with MDM2 playing a role in the modulation of this phenomenon although with mechanisms far from being fully elucidated [18]. MDM2 has multifaceted roles in protein degradation. In fact, aside its well-described role as E3-ubiquitin ligase, MDM2 is able to shuttle p63 to the cytoplasm mediating its interaction with proteins specifically involved in its turnover [19]. Moreover, MDM2 has been shown to mediate proteasome-dependent but ubiquitin-independent degradation of p21Waf1/Cip1 [20] and of Retinoblastoma proteins [21]. More recently it has been reported that MDM2 can interacts with components of the 19S proteasome [22] claiming a wider view of its mechanism of action. We here investigate on a new interrelationship between ARF and Mdm2 in which Mdm2 appears to be implicated in the regulation of ARF turnover mediating its degradation through the proteasome.

Mdm2 overexpression causes p14ARF degradation through the proteasome
To analyze the potential involvement of MDM2 in the regulation of p14ARF protein stability we overexpressed a MDM2 expression plasmid in different human and mouse cell lines that present or lack detectable levels of p53 and/or MDM2. As Fig. 1 shows, p14ARF endogenous levels in both H1299 and HeLa cells linearly decrease when increasing amounts of MDM2 expression plasmid were transfected (Fig. 1, A and B). Real-Time PCR analysis on RNA extracted from MDM2 transfected HeLa and H1299 cells shows no significant changes in the relative amount of ARF mRNA (Fig. 1C) indicating that the MDM2 effect is exerted at the post-transcriptional level. Furthermore, MDM2 overexpression results also in reduction of exogenously expressed ARF protein in several human and mouse cell lines (Fig. 1D, E and F) irrespective of the p53 status, indicating that MDM2 affects ARF protein stability in a p53-independent way. Importantly, reduction of endogenous MDM2 intracellular levels by siRNA causes an increase of ARF endogenous levels in both HeLa and H1299 cells, further confirming that endogenous MDM2 can control p14ARF abundance (Fig. 2). We thus investigated ARF protein stability in the presence or absence of MDM2 overexpression following exposure to cycloheximide to block protein synthesis. At the indicated times after exposure to the drug, cells were harvested and the extracts were analyzed by Western Blot. Fig. 3A and B shows that endogenous p14ARF half-life is lowered following MDM2 overexpression.
We next asked whether the proteasome is involved in MDM2-mediated p14ARF degradation. To explore this hypothesis, we transfected U2OS cells with a fixed amount of p14ARF and increasing amounts of MDM2 and treated cells with the proteasome inhibitor MG132.  Western Blot in Fig. 3C consistently shows that treatment with the proteasome inhibitor counteracts the MDM2 effect, indicating the proteasome as the final effector of the MDM2 action on ARF.

ARF degradation by MDM2 requires protein-protein interaction
The physical interaction between p14ARF and MDM2 has been the object of intensive studies and multiple interacting domains have been identified in p14ARF that contribute to the binding to MDM2 [14,17,23,24]. Hence, we decided to analyze which regions of ARF are necessary to observe the MDM2 effect on p14ARF. We thus made use of two different ARF deletion mutants, the ARF 1-65 , encompassing exon 1β of p14ARF and the ARF 65-132 , corresponding to ARF exon 2. Co-immunoprecipitation in U2OS cells showed that the ARF 1-65 mutant is able to interact with MDM2 while that retaining only exon 2 (ARF 65-132 ) is not (Fig. 4A), confirming a major role of exon 1β in binding MDM2 [23,24]. The two ARF deletion mutants were then transfected together with increasing amounts of MDM2. As Fig. 4B and C clearly show, the amino terminal half of p14ARF (ARF 1-65 ) appears degraded by MDM2 overexpression, while the C-terminal half (ARF 65-132 ) is not affected, supporting a role of the binding between the proteins in the MDM2 mediated ARF degradation. Further, we analyzed the capacity of MDM2 to induce degradation of an ARF deletion mutant (ARFΔ 2-14/82-101 ) lacking the nuclear/nucleolar localization signals [14]. Interestingly, this mutant displays a prevalent cytoplasmic localization [14] and retains the capacity to bind (Fig. 4D) and be degraded by MDM2 (Fig. 4E).
Aside the well established role as an ubiquitin ligase, MDM2 can affect protein turnover also through an ubiquitin independent mechanism, inducing a direct association of target proteins with the proteasome [22]. To discriminate between these possibilities, we analyzed the effect of a MDM2 mutant lacking the ubiquitin ligase ring finger domain (MDM2 1-441 ) (Fig. 5A) [25] on exogenously expressed ARF levels. Interestingly, this mutant is able to induce ARF degradation when overexpressed (Fig. 5B), indicating that the MDM2 ubiquitin ligase domain of MDM2 is dispensable for the observed effect on ARF. Next, we analyzed the effect of a MDM2 deletion mutant (MDM2Δ 150-230 ) (Fig. 5A) that, lacking both the import (NLS) and export (NES) nuclear localization signals, shows a prevalent cytoplasmic localization [25]. When overexpressed in U2OS cells, this mutant induces degradation of exogenously expressed ARF (Fig. 5C) indicating that ARF degradation does not require MDM2 localization in the nucleus. Co-immunoprecipitation experiments indicate that both MDM2 mutants are able to bind ARF, reinforcing the notion that the binding between the two proteins could play a role in the observed effect (Fig. 5D).

The MDM2-mediated ARF degradation mainly occurs in the cytoplasm
The evidence that a MDM2 mutant, unable to localize in the nucleus, induces ARF degradation, together with the observation that an ARF mutant displaying a predominant cytoplasmic localization (ARFΔ 2-14/82-101 ) is prone to MDM2-mediated degradation, lead to the suggestion that MDM2 induced ARF degradation upon MDM2 overexpression can occur in the cytoplasm. This prompted us to verify the ability of MDM2 to degrade ARF in cells treated with Leptomycin B (LMB), known to block MDM2 shuttling to the cytoplasm [26]. To this purpose, U2OS cells transfected with a fixed amount of ARF were treated or not with Leptomycin B in presence or absence of ectopic MDM2. As shown in Fig. 6A, Leptomycin B treatment not only determines accumulation of ARF but, more importantly, counteracts the MDM2 effect on ARF, strongly indicating that forced nuclear MDM2 localization prevents its ability to affect p14ARF turnover. We recently demonstrated that TPA (a known activator of the signal transduction enzyme PKCα) treatment results in an increase of the ARF cytoplasmic pool [11]. Here we addressed whether PKC activation could influence MDM2-mediated ARF degradation. To this purpose we treated HeLa cells, transfected with increasing amounts of MDM2, with either TPA or Bisindolylmalemide (Bim, an inhibitor of PKC activation). Fig. 6B shows that Bim treatment, as expected, induces a decrease of ARF. Interestingly, MDM2 overexpression further lowers ARF levels compared to controls. Conversely, TPA treatment mostly counteracts MDM2 induced degradation of p14ARF, suggesting that the PKC pathway should not be activated to obtain efficient MDM2-mediated ARF degradation. To further analyse the PKC role in MDM2 mediated ARF degradation, U2OS cells were treated either with PKC α specific siRNA or control siRNA and transfected with increasing amounts of MDM2 (Fig. 6C). Overexpression of MDM2 in si-ctr depleted cells induces ARF degradation as expected. Interestingly, we observe a decrease of ARF levels upon PKC depletion in the absence of MDM2 overexpression (compare lane 5 with lane 2). This is in agreement with the observation that PKC activation induces an increase of ARF protein levels and PKC inactivation causes a decrease of ARF levels (Vivo et al., 2013 and Fig. 6B). We actually observed that also MDM2 levels are negatively regulated by PKC depletion, (compare lane 5 with lane 2). Importantly, in agreement with our previous experiments, MDM2 overexpression in PKC depleted cells can still reduce ARF protein levels (compare lane 5 with lanes 6 and 7).
To go deeper inside the potential role of ARF phosphorylation in MDM2 mediated ARF degradation, we made use of two point mutants, the ARFT8D and ARFT8A in which Threonine 8 has been substituted, in order to mimic, respectively, a phosphorylated and a non phosphorylated form of ARF [11]. Interestingly, the dephosphorylated form of ARF (T8A mutant) is very unstable, with a relatively short half-life while the T8D appears consistently enriched in the cytoplasmic compartment [11].
Both mutants were transfected in U2OS cells together with increasing amounts of MDM2. Interestingly, the phospho-mimetic T8D mutant appears more resistant to MDM2 mediated degradation, while the T8A mutant appears more sensible to the MDM2 effect (Figs. 6D and 1D), confirming that phosphorylation exerts a protective role on ARF.
This result raised the hypothesis that the prevalent nuclear localization of ARF (T8A and WT proteins) could be the result of an efficient clearance of dephosphorylated, and thus unstable, ARF protein from the cytoplasm. To explore this hypothesis, U2OS cells, transfected with wild type or mutant proteins were treated with MG132 and ARF subcellular distribution was analyzed by immunofluorescence. As is shown in Fig. 7A and B we observed, upon MG132 treatment, a consistent increase of ARF cytoplasmic staining in both WT and T8A transfected cells, further suggesting that ARF degradation prevalently occurs in the cytoplasm. Similarly, Western Blot of nuclear and cytoplasmic extracts of U20S cells, transfected and treated as above, shows a consistent increase of both wt ARF and ARFT8A protein levels in the cytoplasmic compartment upon proteasome inhibition (Fig. 7C).

Discussion
The importance of the functional interaction between ARF and MDM2 came out from the initial discovery that ARF has the potential to act as a tumour suppressor by binding to and inhibiting the p53 antagonist MDM2. This knowledge was reinforced and refined over years although with controversies regarding the role of ARF nucleolar localization in this process [16,17,27]. On the other hand, the importance of this interaction is clear from the observation that inactivation/deregulation of the p53-MDM2-ARF axis is crucial in the development of most human cancers. Intriguingly, MDM2 overexpression and p14ARF loss correlate with a more aggressive phenotype in primary human lung tumors underlying an inverse relationship between the two in human tumors [28]. On the other hand, previous observations pointed to a controversial effect of MDM2 action on p14ARF: it was shown that loss of Mdm2 could either suppress [29] or restore [14] the ability of murine p19 ARF to induce p53-independent cell cycle arrest. Thus, depending on the context (cellular types, upstream signals, expression level), Mdm2 could behave either as a mediator or inhibitor of p14ARF function onto cell proliferation [28]. Anyway, the first "upturn" in our way to look at the MDM2/ARF relationship came out from the observation that a short ARF mutant (aa2-29) appeared significantly stabilized after silencing of MDM2 or overexpression of MDMX (a MDM2 related, interfering protein),  leading to a model in which ARF binding to MDM2 causes, in some way, its suicide because brings itself to degradation ("guilt by association model") [18]. Our study well fit in with these observations, demonstrating that MDM2 intracellular increase can, indeed, cause ARF destruction by the proteasome. We demonstrate that the physical interaction between ARF and MDM2 is necessary, although not sufficient in order to observe the effect, as ARFT8D mutant is able to interact with MDM2 [11] but is not degraded. The fact that the region of interaction of MDM2 with MDMX (i.e the RING Finger domain) is dispensable for this effect suggests that, indeed, this protein is not involved. On the other hand, the observation that the RING Finger/Ubiquitin ligase domain of MDM2 is dispensable to affect ARF levels, suggests that, although this cannot be formally excluded, ARF ubiquitination is not required [8,9]. On the other hand, Rodway et al. [18] postulated a role of MDM2 in mediating ARF delivery to the proteasome without any requirement for ubiquitination. This could probably be mediated by MDM2 direct interaction with the proteasome [22]. In particular, Mdm2 has been shown to directly interact with several proteasome subunits through both N-terminal and C-terminal regions while its central domain (aa200-300) appears to exert a negative regulative control on the binding [22]. Our MDM2 mutants (MDM2 1-441 and MDM2Δ 150-230 ) retain at least one of the regions required for the interaction with the proteasome and both of them are able to induce ARF degradation. Intriguingly, we also observe a more efficient ARF reduction with the MDM2Δ 150-230 mutant, that lacks the central domain.
However, as ARF ubiquitination has not been directly addressed in the present study, we cannot exclude the intervention of ULF [7] or any other yet unidentified ubiquitin ligase in the observed effect. However, while ULF mediated ARF degradation appears exerted in the nucleoplasm, different observations in this paper indicate that the MDM2 effect occurs in the cytoplasm. Firstly, mutations in both ARF and MDM2 that force their localization in the cytoplasm, do not affect MDM2 mediated ARF degradation. Furthermore, Leptomycin B treatment, that inhibits MDM2 nuclear export, causes ARF accumulation and counteracts the MDM2 degradation effect on ARF. Finally, immunofluorescence experiments clearly show that MG132 treatment results in an increase of ARF levels in the cytoplasm, strongly suggesting that, indeed, the reason why ARF is not usually detectable in this subcellular compartment resides in a more rapid turnover.
It has to be remarked that, since its initial discovery, ARF was described with a prevalent nucleo/nucleolar localization. Accordingly, it has been reported that a protective role on ARF levels is exerted in the nucleolus by B23/NPM. Moreover, interaction with TBP1, which protects ARF from proteasome degradation, occurs in the nucleus [8].
More recently, ARF has been reported to localize also in the cytoplasm although mainly associated to mitochondria, because of its role in autophagy [30]. Interestingly, cytoplasmic p14ARF staining has been described in some cancers [28].
Finally, we have recently reported that, following activation of PKC, ARF protein is phosphorylated and accumulates in the cytoplasm [11].
Here we show that activation of the PKC pathway by TPA counteracts MDM2 action while inhibition of the pathway by either Bisindolylmalemide treatment or PKCα directed siRNA, causes a decrease of ARF levels and enhances MDM2 mediated ARF degradation. Accordingly, the T8D ARF mutant that mimics the phosphorylated status of the protein, although more present in the cytoplasm, appears to be less sensitive to MG132 treatment and to MDM2 overexpression. These data reinforce our previous suggestion that phosphorylation exerts a protective effect on ARF. It seems that more than one transcription independent mechanisms are involved in the regulation of ARF protein, i.e. proteasome mediated degradation [6][7][8] and phosphorylation.
Overall, our observations lead to the hypothesis that in an oncogenic environment in which MDM2 is overexpressed, ARF levels are kept low due to a rapid turnover caused by MDM2 triggered degradation. Increasing ARF degradation and/or phosphorylation could thus be common strategies that a cell orchestrate to promptly escape ARF growth suppression functions either in physiological conditions or when cells are forced to proliferate during cancer progression. Further experiments are needed to clarify the molecular mechanisms underlining these observations and their relevance for cancer development in vivo.

Cell cultures, transfections
U2OS, H1299, Hela cell lines were purchased from the Cell Lines Service (CLS). MEF p53-/-MDM2-/cell line was described in [32]. All cell lines were grown in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum (Euroclone, Life Science) at 37°C in a humidified atmosphere of 5% (v/v) CO2 in air.
For experiments described in Fig. 7C, U2OS cells were transfected by electroporation using the Neon Transfection System (Life Technology) following manufacturer instructions. Briefly, 4x10 6 cells were electroporated with 2.5 μg of either wt ARF or T8D ARF mutant plasmid DNA, and 3.5 μg of ARF T8A expressing plasmid. In order to achieve the same transfection efficiency in untreated and MG132 treated samples, transfected cells were splitted in two aliquots and, 24 hours post transfection, incubated with either DMSO or MG132.

Real Time PCR Analysis
Total cell extracts from HeLa cells were collected for each experimental point using the Pure-Link RNA Mini Kit-Life Technology according to manufacturer instructions. cDNA synthesis was performed using 1ug of total RNA with SuperScript III RT (Life Technology). Real Time PCR was performed using 50 ng of ss cDNA with Syber Green PCR master mix (Applied Biosystems). Relative quantification of p14ARF transcripts was done using 18s RNA for normalization.

Decay rate analysis
Decay rate analysis was already described [33]. Briefly, HeLa cells were either mock transfected or transfected with pCMVMDM2. Twenty-four hours after transfection, cycloheximide was added at a final concentration of 60 μg /ml, cells were harvested at the indicated time points and extracts analyzed by Western Blot with anti-ARF (Clone C-18, Santa Cruz), anti MDM2 (clone 2A10, Calbiochem) and anti actin (Clone I-19, Santa Cruz). Band intensities at the different time points were quantified by Image J Software, normalized to actin and reported in graph as percentage of total protein (protein a t0). Each profile represents the mean of three independent experiments. Standard deviations are also shown.

Silencing experiment
H1299 and HeLa cells were plated in six-well plates at 2,5Ã10 5 cells/well. 24 hours later, cells were transfected with either with 10 pmoles of siRNA targeting MDM2 or siNegative Control (Qiagen) (10nM final siRNA concentration). Cells were harvested 72 hours later and processed as described before. U2OS cells were plated in 24 well plates at 5Ã10 4 cells/well. 24 hours later, cells were transfected with either with 15 pmoles of siRNA targeting PKCα or siNegative Control [33]. SiRNA were purchased from Qiagen (Flexi Tubes). 48 hours later, cells were mock transfected (lane 1) or transfected with 0.3 μg of ARF alone or in combination with increasing amounts of pCMVMDM2 expressing plasmid (0.6 and 0.8 μg). Twenty-four hours later cells were collected and total cell extracts analyzed by Western Blot.

Immunoblot analyses, Co-immunoprecipitation, subcellular fractionation
Western Blot analysis was performed as previously described [35]. Extracts were blotted onto PVDF Immobilon-P transfer membrane (Millipore cat. NO IPVH00010). Proteins were visualized with an enhanced chemi-luminescence images detection system (Thermo Scientific ECLplus) and images were taken with ChemiDoc XRS System (Bio-Rad Laboratories). Image quantification was performed with Image J analysis software as described above.
CoIPs were performed as described [32]. Briefly, lysates from cells transfected with 1:1 ratio of either pCDNA-ARF and various MDM2 constructs or pCMV MDM2 and various ARF constructs (as described in the text) were incubated with anti-ARF, or with anti-MDM2 antibodies.

Treatments in this study were performed as follows
For the treatment with MG132 proteasome inhibitor: U2OS cells were treated either with DMSO or 10 μM MG132 (Sigma-Aldrich) for 5 hours. Cells were harvested and total cell extracts were prepared for subsequent analysis as described [8].
Treatment with Bisindolylmalemide or TPA was performed as follows: 24 hours after transfection, HeLa cells were treated either with DMSO or Bisindolylmalemide (Calbiochem) at 10 μM final concentration for 16 hours or TPA (Applichem) at 10 μM final concentration for 5 minutes. Cells were harvested and total extracts prepared for subsequent analysis as described.
Treatment with Leptomycin B: 24 hours after transfection, U2OS cells were treated either with Methanol or with Leptomycin B (Sigma-Aldrich) at 10 μM final concentration for 4 hours. Cells were harvested and total extracts prepared for subsequent analysis as described.
Immunofluorescence assay was performed as described [11]. Briefly, cells were fixed with 4% paraformaldehyde (Sigma-Aldrich, Germany) for 10' at RT. Cells were permeabilized with ice-cold 0.5% Triton X-100 for 5 min and then washed with PBS. Cells were then incubated with anti-ARF antibody (14PO2 clone, Neomarkers) for 1hr at RT, followed by incubation with a Cy3-conjugated anti-mouse antibody (ImmunoResearch Laboratory) for 1 hour at RT. DAPI (Sigma-Aldrich, Germany) staining was performed to counterstain nuclei. Coverslip were mounted with Vectashield (VectorLab) and examined under a fluorescence microscope (Nikon). For each transfection point, subcellular localizations were analysed in 100-150 cells and results plotted in graph. Histograms (% of transfected cell ± S.D.) represent the mean of at least three independent transfection experiments.
Nucleo-cytoplasmic extraction of U2OS electroporated cells was performed as previously described in Vivo et al., 2013. Equal amount of either nuclear (6 μg) or cytoplasmic extracts 30 μg) were subjected to SDS-page and Western blot as described.