Mdm2 Induces Mono-Ubiquitination of FOXO4

Background The Forkhead box O (FOXO) class of transcription factors are involved in the regulation of several cellular responses including cell cycle progression and apoptosis. Furthermore, in model organisms FOXOs act as tumor suppressors and affect aging. Previously, we noted that FOXOs and p53 are remarkably similar within their spectrum of regulatory proteins [1]. For example, the de-ubiquitinating enzyme USP7 removes ubiquitin from both FOXO and p53. However, Skp2 has been identified as E3 ligase for FOXO1, whereas Mdm2 is the prime E3 ligase for p53. Principal Findings/Methodology Here we provide evidence that Mdm2 acts as an E3 ligase for FOXO as well. In vitro incubation of Mdm2 and FOXO results in ATP-dependent (multi)mono-ubiquitination of FOXO similar to p53. Furthermore, in vivo co-expression of Mdm2 and FOXO induces FOXO mono-ubiquitination and consistent with this result, siRNA-mediated depletion of Mdm2 inhibits mono-ubiquitination of FOXO induced by hydrogen peroxide. Regulation of FOXO ubiquitination by Mdm2 is likely to be direct since Mdm2 and FOXO co-immunoprecipitate. In addition, Mdm2-mediated ubiquitination regulates FOXO transcriptional activity. Conclusions/Significance These data identify Mdm2 as a novel E3 ligase for FOXOs and extend the analogous mode of regulation between FOXO and p53.


Introduction
Forkhead box O (FOXO) transcription factors have recently gained considerable attention because of their potentially critical role in aging [1,2]. The paradigm in this respect is the C. elegans FOXO ortholog DAF-16. Lifespan extension through a number of genetic and non-genetic interventions in these nematodes requires at least in part DAF-16 [2]. Especially, the effects of lowered insulin signaling critically depend on DAF-16 and DAF-16 acts downstream of the insulin signaling pathway consisting of the lipid kinase phosphoinositide-3 kinase (PI-3K) and the serine/threonine protein kinase B (PKB/AKT). PKB directly phosphorylates DAF-16/FOXO and this results in nuclear exclusion and therefore reduced DAF-16/FOXO transcriptional activity [3,4].
Aging may also result from the accumulating damage caused by reactive oxygen species (ROS) [5]. In this respect regulation of cellular anti-oxidant capacity by DAF-16/FOXO provided rationale for its effect on lifespan. Interestingly, FOXO itself is also regulated by ROS and treatment of cells with hydrogen peroxide, which increases cellular oxidative stress, results in nuclear translocation of FOXO [6,7]. FOXOs are regulated through a multitude of post-translational modifications (PTMs) including phosphorylation, acetylation and ubiquitination (re-viewed in [1]). Whereas PKB-mediated phosphorylation results in exclusion of FOXO from the nucleus, the mechanism and/or PTMs responsible for relocalization to the nucleus after increased cellular oxidative stress, remain poorly understood. However, the enzymes responsible for adding these modifications are remarkably similar between p53 and FOXO (for a discussion see [1]). With respect to the regulation of ubiquitination we previously identified USP7 as a de-ubiquitinating enzyme for FOXO4 [8] and USP7 is also a de-ubiquitinating enzyme for p53 [9]. FOXOs are relatively stable proteins with a half-life of approximately 8-10 hrs in untransformed cells [8]. In transformed/oncogenic cells, especially cells transformed through activation of PI-3K signaling, FOXO protein half-life is shortened [10][11][12]. This is likely due to PI-3K/PKB mediated upregulation of Skp2 in these cells, as Skp2 has been identified as an ubiquitin E3 ligase responsible for FOXO poly-ubiquitination and degradation [10]. Consistent with Skp2 regulation by PKB and FOXO being degraded in a Skp2dependent manner, several other PKB targets have been reported to be degraded in a Skp2-dependent manner as well [11,13]. Previously, we demonstrated that the signaling function of FOXO4 is regulated by mono-ubiquitination especially after increased cellular oxidative stress [8]. Mono-ubiquitination correlates with increased nuclear localization of FOXO and hence increased transcriptional activity. Consequently, USP7 expression inhibited FOXO4 transcriptional activity due to deubiquitination of FOXO4 and re-localization to the cytosol. To further understand the regulation of mono-ubiquitination of FOXOs we searched for ubiquitin E3 ligases that would ligate ubiquitin onto FOXO4. Here, we report the identification of Mdm2 as an E3 ligase for FOXO4 that mediates monoubiquitination of FOXO4 after increased cellular oxidative stress.

Results
In our attempt to identify potential E3 ligases for FOXOs we co-expressed several candidate E3 ligases with FOXO and noticed that Mdm2 co-expression resulted in an apparent reduction in FOXO4 expression (Fig. 1a). In transient expression experiments reduced protein expression can occur through various mechanisms including promoter squelching. However, as Mdm2 induces ubiquitin-mediated breakdown of target proteins we first analyzed whether Mdm2 could catalyze ubiquitin addition to FOXO4 in vitro. To this end we reconstituted a functional E1-E2-Mdm2 ubiquitin ligase system using purified proteins and added GST-FOXO4 as a substrate. Only in the presence of rATP this resulted in the addition of multiple ubiquitin moieties to GST-FOXO causing a laddering indicative for poly-ubiquitination or multiple mono-ubiquitination (Fig. 1b). Importantly, this in vitro reconstituted system displayed similar activity towards GST-p53, but not to GST alone, suggesting that in vitro FOXO is as good a substrate for Mdm2 as is p53. Mdm2 uses a C-terminal RING finger domain, critical to its function as an E3-Ligase. A Mdm2 mutant lacking this domain was tested and found unable to ubiquitinate FOXO4 (Fig. 1b), highlighting the specificity of Mdm2 and the dependency on its E3-ligase activity to mono-ubiquitinate FOXO4 in vitro. To further address whether the in vitro observed laddering represents poly-ubiquitination or multiple mono-ubiquitination, a number of ubiquitin mutants instead of wild-type ubiquitin were analyzed in the in vitro ubiquitination assay. Using ubiquitin-K48A, defective in K48-mediated ubiquitin branching which targets proteins for the proteasome, and ubiquitin-K7R and methylubiquitin both defective in mediating poly-ubiquitination all resulted in same pattern of mdm2-mediated GST-FOXO laddering (Fig 1c). Taken together these results clearly show that in an in vitro reconstituted system Mdm2 can act as an E3-ligase for FOXO4 and that in contrast to what has been reported for p53, Mdm2 catalyzes multiple mono-ubiquitination of GST-FOXO4 rather than poly-ubiquitination.
Next, we tested whether Mdm2 also can ubiquitinate FOXO4 in vivo. Co-expression of flag-FOXO4 and myc-Mdm2 induced mono-ubiquitination of FOXO4 (Fig. 1d). We did not observe substantial poly-ubiquitination, also not in the presence of the proteasome inhibitor MG132 (data not shown). Also, the delta-RING domain Mdm2 mutant did not induce FOXO4 monoubiquitination (Fig 1e). Albeit consistent with our in vitro data this questions the mechanism underlying the reduced detection of FOXO4 protein by immunoblotting after overexpression of Mdm2. Reduced detection of FOXO4 concomitant with Mdm2 overexpression would normally be taken to indicate proteasomal degradation of FOXO4 and this should be reversed by MG132 treatment. However, we did not observe substantial rescue of FOXO4 protein expression after MG132 treatment, despite observing accumulation of auto-poly-ubiquitinated Mdm2 species, which indicates that the MG132 treatment did work (Fig. 1f). Again this is consistent with the observed lack of poly-ubiquitination and suggests that either FOXO4 is degraded through another pathway for example caspase-mediated breakdown, or alternatively, that (multiple) mono-ubiquitinated FOXO4 is targeted to a cellular compartment for example PML bodies from which it is not efficiently extracted. To confirm our MG132 experiments, we performed FOXO4 half-life studies. Transfection of Mdm2 did not affect the half-life of co-transfected FOXO4 ( Supplementary  Fig. S1). Taken together these results indicate that Mdm2 expression does not lead to FOXO4 degradation by means of regulating its protein stability through the proteasome. This finding is consistent with our MG132 experiments. Thus in all approaches we come to the conclusion that Mdm2 does not substantially affect FOXO4 protein half-life.
Finally, to further substantiate a role for endogenous Mdm2 in regulating the ubiquitin status of FOXO4 in vivo, we used siRNA against the human ortholog of Mdm2 ( Supplementary Fig. S2). As reported previously, increasing cellular oxidative stress by treating cells with hydrogen peroxide induced mono-ubiquitination of both FOXO4 and FOXO3a [8] and Supplementary Fig. S3. Importantly, siRNA-mediated knockdown of Mdm2 significantly reduced hydrogen peroxide-induced mono-ubiquitination of FOXO4 (Fig. 1h). Together, these data provide compelling evidence that Mdm2 can mediate FOXO4 mono-ubiquitination in vitro and in vivo.
To test the possibility that Mdm2 could directly regulate FOXO we analyzed binding between Mdm2 and FOXO4. Upon coexpression of Flag-Mdm2 and HA-FOXO4, FOXO4 was coimmunoprecipitated with Mdm2, and vice-versa (Fig. 2a,b). Consistent with our in vitro data, this result suggests that Mdm2 directly binds and regulates FOXO4, rather than through regulating the activity of de-ubiquitinating enzymes such as USP7. Next, we tested binding between endogenous FOXO4 and Mdm2 proteins and observed reciprocal co-immunoprecip- itation between endogenous FOXO4 and Mdm2 in HEK293T cells (Fig 2c). Mono-ubiquitination of FOXO4 results in increased transcriptional activity of FOXO4 [8] and this is reversed by USP7mediated de-ubiquitination. To see whether Mdm2 would also regulate FOXO4 transcriptional activity we performed FOXO reporter assays. Expression of increasing amounts of Mdm2 resulted in a bell-shaped regulation of FOXO4 transcriptional activity. Low amounts of Mdm2 transfected induced a reproducible increase in FOXO4 transcriptional activity on two different FOXO responsive reporters (Fig. 3a (6xDBE-luciferase) and Fig. 3b (p27-luciferase)), consistent with the notion that Mdm2 can induce mono-ubiquitination of FOXO4. In contrast, higher amounts of Mdm2 transfected resulted in reduced FOXO4 transcriptional activity. This suggests that the multiple mono-ubiquitination of FOXO4 induced by Mdm2 in vitro and possibly in vivo by high levels of Mdm2 represents inactivation of FOXO4. In effect this would be similar to polyubiquitination and degradation, but in contrast this leaves FOXO4 to be re-activated by de-ubiquitination by USP7 [8]. To assess the structural requirements of Mdm2 to regulate FOXO4 transcriptional activity we compared the effect of wild-type Mdm2 on FOXO4 transcriptional activity with that of Mdm2 mutated in its RING domain, and with that of Mdm2 mutated in its p53 interaction domain (Fig. 3c). The RING domain mutant of Mdm2 did not affect FOXO4 transcriptional activity indicating that Mdm2 regulated FOXO4 activity requires a functional E3 ligase domain. In contrast, p53 binding to Mdm2 appears not involved as Mdm2 defective in binding to p53 regulated FOXO4 in a manner identical to wild-type Mdm2. Ectopic expression of FOXO4 in cells induces cell cycle arrest and induction of apoptosis [1] and this can be monitored by a reduction in colony formation ( [14] and Fig 3d). Similar to decreased transcriptional activity at higher levels of Mdm2 co-expression Mdm2 represses the ability of FOXO4 to inhibit colony formation.

Discussion
Here we provide evidence that Mdm2 is an E3-ligase that can ligate ubiquitin onto FOXO4 both in vitro and in vivo. Monoubiquitination of FOXO4 as induced by hydrogen peroxide treatment of cells requires endogenous Mdm2, because siRNA mediated knockdown of Mdm2 prevents mono-ubiquitination. This shows that Mdm2 is at least functional in regulating monoubiquitination of FOXO4. In vitro, Mdm2 induces a pattern of ubiquitination that normally is considered indicative of polyubiquitination. However, alternative to poly-ubiquitination, extensive mono-ubiquitination on multiple different lysine residues, 19 of which are present in FOXO4, can cause a similar characteristic laddering. This would be consistent with the current view that low mobility species of ubiquitinated p53 represent mono-ubiquitinated p53 ( Fig. 1 and [ 15,16]. To further discriminate between these possibilities we used a number of ubiquitin mutants that are defective in poly-ubiquitination (K48A, K7R and methyl-ubiquitin). When these mutants were used as only ubiquitin donor in the in vitro assay, Mdm2 catalyzed a highly similar, if not identical, pattern of laddering compared to including wild-type ubiquitin in this assay. This strongly suggests that in vitro Mdm2 catalyzes preferentially only (multi-)mono-ubiquitination of GST-FOXO4. Also in vivo overexpression or siRNA-mediated knockdown of Mdm2 resulted in the induction or loss of FOXO4 monoubiqitination respectively. Thus we conclude that Mdm2 regulates mono-ubiquitination of FOXO4. However, we observed an apparent reduction in protein expression of FOXO4, especially after high expression of Mdm2 and this would suggest protein degradation most likely through poly-ubiquitination-mediated proteasomal degradation. This would be at odds with the conclusion that Mdm2 catalyzes mono-ubiquitination. To establish whether proteasomal degradation is causal to the reduced detection of HA-FOXO4 after flag-Mdm2 overexpression we treated cells with MG132. Inhibition of proteasome-mediated degradation did not result in increased detection of HA-FOXO4 despite observed accumulation of Mdm2. Thus, this result does not implicate Mdm2 induced poly-ubiquitination of HA-FOXO4 and subsequent degradation through the proteasome and is consistent with the lack of effect of MG132 treatment on FOXO4 ubiquitination in vivo (Fig 1, and B.M.T.B, unpublished data). Consistent with these observations, protein stability experiments indicate that Mdm2 does not affect FOXO4 protein stability. Consequently, reduced detection of FOXO4 after Mdm2 overexpression is likely due to other mechanism(s). This could be alternative mechanism(s) of degradation such as lysosomal degradation or protease-mediated degradation (e.g. caspase). Alternatively, mono-ubiquitination has been shown in several cases to regulate cellular localization of proteins and thus the apparent reduction in protein expression may equally represent a shift of target protein into a complex or towards a cellular location that results in inefficient extraction of protein and thereby reduced detection after immunoblotting. Indeed, recent developments have provided multiple examples in different signaling pathways that ubiquitination serves other purposes than merely targeting proteins for degradation [17][18][19]. This raises the interesting possibility that the initial function of (mono-)-ubiquitination is to provide a means to regulate protein function similar to for example phosphorylation. However, to terminate the signaling function of (mono-)ubiquitination a cell can choose between either de-ubiquitination or poly-ubiquitination. Depending on the urgency to terminate signaling, poly-ubiquitination and subsequent degradation, may be the preferred mode.
While our study was in progress Yang et al. also reported ubiquitination of FOXO3a by Mdm2 [20]. However, in contrast to our results presented here, their study suggests a role for Mdm2 mainly in the breakdown of FOXO3a. As discussed above, in our experiments only high expression of Mdm2 may result in induced breakdown of FOXO4.
Yang et al. also implicated a role for ERK in the regulation of FOXO3a by Mdm2 and provide evidence that FOXO3a phosphorylation by ERK through an unknown mechanism induces Mdm2 binding to FOXO3a [20]. Importantly, Yang et al. use EGF as stimulus whereas we use peroxide stress. It is of importance to note that the stimuli used by others (EGF, PDGF, insulin) all inhibit FOXO function whereas we and others have shown that oxidative stress (e.g. peroxide) as used here activates FOXOs [6]. This is an essential difference. In addition, for PDGF and insulin stimulation several previous studies have shown that Skp2 is involved in degradation of FOXO induced by these factors [10,11]. In addition, with respect to the issue here, several interesting observations within these studies were made. Firstly, FOXO half-life in 'normal' cells is around 8-10 hrs similar to our previous observations ( [8] and the results in this study). Second, only in cells transformed through PI3K activation (v-Ha-RAS, Active PI3K alleles) FOXO half-life is shortened [21], but again this is in these studies a PKB/AKT and Skp2 mediated process (and not ERK-Mdm2). We are tempted to speculate that Mdm2 induces FOXO mono-ubiquitination; this results in activation of FOXO. Activation can be terminated by USP7 de-ubiquitination or alternatively by Skp2-mediated poly-ubiquitination and degradation. The latter occurs as a result of oncogenic transformation through PI3K/PKB/AKT, but possibly also through ERK signalling. Thus if one considers the possibility of Mdm2 being a 'priming' E3-ligase for FOXO and Skp2 the branching E3-ligase, these different results can be reconciled. Clearly further studies are required to fully appreciate the role of ubiquitination in FOXO regulation in response to various cellular conditions.
Mono-ubiquitination is observed and studied thus far, for proteins with a relative long half-life, such as PTEN [22], EGF receptor [23] and FOXOs (approximately 10 hrs in untransformed cells [8]). In contrast, mono-ubiquitination has not yet been considered for short lived proteins such as cell cycle regulators (cyclins) and oncogenes (myc, beta-catenin). However, establishing mono-ubiquitination for these short-lived proteins may just be a technical challenge. Indeed, recent results with p53 may provide basis for such a paradigm shift. Whereas, initially p53 served as a classical example of a protein regulated through protein degradation, it is by now clear that mono-ubiquitination of p53 occurs and serves to provide a new signaling function to p53 [15,24]. Instead of acting as a transcription factor, monoubiquitination of p53 serves as a signal to relocate p53 from the nucleus to mitochondrial membrane [24]. Along the same line of reasoning, the role of Mdm2 in ubiquitination of p53 is now being discussed [17,25,26]. Thus, the possibility is being raised that endogenously expressed Mdm2 is actually preferentially involved in mono-ubiquitination of p53, whereas aberrant high expression of Mdm2 may result in poly-ubiquitination and degradation of p53. Essentially, our observations presented here indeed fully support a role for Mdm2, at endogenous level, in regulating monoubiquitination of in this case FOXO4.
In summary, we have identified Mdm2 as an ubiquitin E3 ligase for FOXO4 which functions in oxidative stress-induced FOXO4 mono-ubiquitination. This extends the network of co-regulatory proteins of FOXO and p53 and therefore supports a model of coevolution of stress maintenance mechanisms.

Luciferase Reporter assay
Cells were transfected with the indicated constructs plus 20 ng TK-Renilla per condition as a transfection efficiency control. Lysates were measured after 24 h by the Promega Dual Luciferase reporter assay.

Ubiquitination assays
The in vitro ubiquitination assay was performed essentially as described [29]. Flag-Mdm2 was purified from HEK293T cells with Flag-M2 beads (Sigma). Precipitated protein was washed with RIPA, and eluted off with Flag peptide (Sigma). Eluted protein was dialysed o/n with a buffer containing (25 mM HEPES-pH 8.0, 150 mM NaCl, protease inhibitors). The in vitro ubiquitination assay was initiated with the addition of ATP (f.c.2.5 mM) and quenched after 2 h with Laemmli Sample Buffer. Purified GST and GST-FOXO4 were a kind gift from H. de Ruiter. Purified GST-p53, E1, E2 (UbcH5b) were kind gifts from Dr. K.W. Mulder. Ubiquitin, Ubiquitin-K7R (All lysines mutated to arginine), Ubiquitin-K48A (lysine 48 mutated to Alanine) and methylated Ubiquitin (all lysines methylated) were purchased at Boston Biochem. In vivo ubiquitination assays were essentially performed as described [8]. In brief, HEK293T cells were transfected with the indicated constructs. 48 h. post transfection, cells were left untreated or treated as indicated, and lysed in lysis buffer (8 M Urea, 10 mM Tris-HCl pH 8.0, 100 mM Na 2 HPO 4 / NaH 2 PO 4 , 0.2% TX-100, 5 mM NEM, protease inhibitors).
Ubiquitinated proteins were precipitated using Ni-NTA agarose beads and analysed by WB.

Colony outgrowth assay
Equal amounts of A14 cells were plated in triplicate in 6-well dishes and transfected with 2 mg of the indicated constructs in combination with 0.5 mg pbabe-puro. 24 hours post-transfection cells were placed under selection with 2 mg/ml Puromycin. Every two days the selection medium was refreshed. At 10 days post transfection cells were fixed for 10 minutes with ice-cold methanol and colonies were stained with 0.5% crystal violet, dissolved in 25% methanol. The plates were washed with dH 2 O and dried overnight.