FMRP-dependent Mdm2 dephosphorylation is required for MEF2-induced synapse elimination

Abstract

The Myocyte Enhancer Factor 2 (MEF2) transcription factors suppress an excitatory synapse number by promoting degradation of the synaptic scaffold protein, postsynaptic density protein 95 (PSD-95), a process that is deficient in the mouse model of Fragile X Syndrome, Fmr1 KO. How MEF2 activation results in PSD-95 degradation and why this is defective in Fmr1 KO neurons is unknown. Here we report that MEF2 induces a Protein phosphatase 2A (PP2A)-mediated dephosphorylation of murine double minute-2 (Mdm2), the ubiquitin E3 ligase for PSD-95, which results in nuclear export and synaptic accumulation of Mdm2 as well as PSD-95 degradation and synapse elimination. In Fmr1 KO neurons, Mdm2 is hyperphosphorylated, nuclear localized basally, and unaffected by MEF2 activation, which our data suggest due to an enhanced interaction with Eukaryotic Elongation Factor 1α (EF1α), whose protein levels are elevated in Fmr1 KO. Expression of a dephosphomimetic of Mdm2 rescues PSD-95 ubiquitination, degradation and synapse elimination in Fmr1 KO neurons. This work reveals detailed mechanisms of synapse elimination in health and a developmental brain disorder.

Introduction

Maturation of functional neural circuits requires the orchestrated formation and elimination of synapses. Synapse elimination or pruning contributes to the development and experience-dependent refinement of cortical circuits (1–3). The myocyte enhancer factor 2 (MEF2) family of activity-dependent transcription factors promotes elimination of functional and structural excitatory synapses (4–8). Reduced synaptic PSD-95 is predictive of dendritic spine elimination in vivo and suggests that regulation of PSD-95 levels determines synapse number (9). Consistent with this observation, MEF2 activation causes ubiquitination and degradation of PSD-95 by promoting synaptic accumulation of Murine Double Minute-2 (Mdm2), an ubiquitin E3 ligase for PSD-95 (7). The mechanisms that regulate synaptic accumulation of Mdm2 are unknown.

Importantly, MEF2-induced synapse elimination and PSD-95 degradation are deficient in the mouse model of Fragile X Syndrome (FXS), a leading genetic cause of intellectual disability and autism. Such a deficit may contribute to the excess of dendritic spines and circuit hyperexcitability associated with FXS (Bagni and Greenough, 2005). FXS is caused by transcriptional silencing of Fmr1 which encodes an RNA binding protein Fragile X Mental Retardation Protein (FMRP). Our recent work uncovered a deficit in MEF2-triggered PSD-95 ubiquitination and synaptic accumulation of Mdm2 in Fmr1 KO neurons, suggesting that dysregulation of Mdm2 in Fmr1 KO neurons underlies the deficit in synapse elimination. Altered synaptic accumulation of Mdm2 in Fmr1 KO neurons is likely due to the enhanced interaction and increased protein levels of a known Mdm2 interactor, Eukaryotic Elongation Factor 1α (EF1α), whose translation is regulated by FMRP. Knockdown of EF1α in Fmr1 KO neurons restores MEF2-induced Mdm2 accumulation at synapses, PSD-95 degradation and synapse elimination (7). The molecular mechanisms by which EF1α inhibits Mdm2 function in Fmr1 KO neurons is unknown. Understanding this process may provide novel therapeutic targets to recover proper synapse refinement in FXS.

In this study, we investigated the posttranslational modification and intracellular trafficking of Mdm2 during MEF2-induced synapse elimination. MEF2 activation in wildtype (WT) neurons results in dephosphorylation of Mdm2 by Protein phosphatase 2A (PP2A) and subsequent nuclear export and synaptic accumulation in WT neurons. Replacement of endogenous Mdm2 in WT neurons with phosphomimetic mutants prevents both MEF2-induced PSD-95 degradation and synapse elimination. In Fmr1 KO neurons, we present evidence that EF1α prevents the PP2A interaction with and dephosphorylation of Mdm2 upon MEF2 activation, rendering Mdm2 hyperphosphorylated and trapped in the nucleus. Replacement with dephosphomimetic Mdm2 mutants in Fmr1 KO neurons restores MEF2-induced PSD-95 degradation and synapse elimination. This work identifies a novel mechanism by which MEF2 activation regulates synapse number through modulation of Mdm2 trafficking and provides a detailed mechanism of this deficiency in a neurodevelopmental disorder.

Results

MEF2 induces PP2A-dependent Mdm2 dephosphorylation in WT but not in Fmr1 KO neurons

How MEF2 leads to synaptic accumulation of Mdm2 and how this is dysregulated by EF1 α in Fmr1 KO neurons is unknown. Mdm2 is not a known target of MEF2 nor induced upon MEF2 activation (7,10). Therefore, we suspected that MEF2 regulates posttranslational modifications of Mdm2 and this may be altered in Fmr1 KO neurons. Phosphorylation (P) of Serine-163 (S163) of Mdm2 regulates its substrate specificity and subcellular localization (11,12). To determine whether S163 phosphorylation of Mdm2 is basally different between WT and Fmr1 KO brain, we performed western blots of WT and Fmr1 KO total brain lysates using a specific anti-P-Mdm2 (S163) antibody and detected a higher S163 phosphorylation of Mdm2 in Fmr1 KO (Fig. 1A). To evaluate whether and how MEF2 affects Mdm2 phosphorylation, we lentivirally expressed a tamoxifen-inducible, constitutively active MEF2 (MEF2-VP16ERtm) into dissociated cortical neuron cultures. Because MEF2 deletion or knockdown leads to cell death (13,14), use of MEF2-VP16-ERtm is the most specific and robust means to acutely detect MEF2-dependent effects (5,7,8). When MEF2 was activated, by application of 4-Hydroxytamoxifen (4-OHT; 6 h) to cultures expressing MEF2-VP16ERtm we observed less phosphorylation of Mdm2 in WT neuron cultures. In contrast, Mdm2 phosphorylation was unaffected by MEF2 activation in Fmr1 KO neuron cultures (Fig. 1B), revealing a novel deficit in Mdm2 regulation in Fmr1 KO.

Figure 1

MEF2 triggers PP2A-dependent de-phosphorylation of Mdm2 in WT but not Fmr1 KO neurons. (A) Western blots of Mdm2 and phosphorylated (S163; p)-Mdm2 from WT or Fmr1 KO brain lysates (n = 4). (B) Western blots of Mdm2 and p-Mdm2 from WT or Fmr1 KO cortical neuron cultures transfected with MEF2VP16ERtm and treated with vehicle (V) or tamoxifen (T) for 6 h (n = 3). (C) Western blots of Mdm2 and p-Mdm2 from WT cortical neuron cultures transfected with MEF2VP16ERtm and treated with vehicle or tamoxifen and with vehicle or okadaic acid as indicated for 6 h (n = 3). (D) PP2A phosphatase activity from WT or Fmr1 KO cortical neuron cultures transfected with MEF2VP16ERtm and treated with vehicle or tamoxifen for 6 h (n = 4). (E) Western blots of Mdm2 and PP2A after co-immunoprecipitation (IP) with Mdm2 from WT or Fmr1 KO cortical neuron cultures transfected with MEF2VP16ERtm and treated with vehicle or tamoxifen for 6 h (n = 3). *P < 0.05.

Given the dramatic reduction of Mdm2 phosphorylation within 6 h after MEF2 activation, we suspected that MEF2 induced a phosphatase-mediated dephosphorylation of Mdm2. The dephosphorylation of Mdm2 at S163 (corresponding to S166 in Hdm2 in human) is mediated by Protein phosphatase PP2A (15,16). Preincubation of cultures in a PP2A inhibitor, Okadaic acid (2 nM) (17), blocked MEF2-induced decreases in Mdm2 phosphorylation (Fig. 1C) demonstrating that PP2A activity is required for MEF2-induced Mdm2 dephosphorylation and suggesting that this process may be deficient in Fmr1 KO neurons. To test whether MEF2 activation or tamoxifen treatment on its own regulates PP2A activity, we immunoprecipitated the PP2A complex with an antibody against the more ubiquitous, catalytic subunit of PP2A, the PP2A-C, from cultured WT or Fmr1 KO neurons and performed an in vitro phosphatase assay. We validated the specificity of the antibody to pull down PP2A-C and enrich for phosphatase activity (see Fig. 4C4, 4C5). MEF2 activation reduced PP2A activity similarly in both WT and Fmr1 KO neurons. Tamoxifen application alone had no effect on PP2A activity. (Fig. 1D). These data suggest that MEF2-induced Mdm2 dephosphorylation is unlikely through a modulation of PP2A activity. Therefore, we hypothesized that MEF2 stimulates a complex between PP2A and Mdm2 and this may be deficient in Fmr1 KO neurons. To test this hypothesis, we performed co-immunoprecipitation with the antibody against Mdm2. As predicted, MEF2 activation induces association of Mdm2 and PP2A in WT neurons, but this was not observed in Fmr1 KO neurons (Fig. 1E). These results suggest that the deficit in MEF2-stimulated Mdm2 dephosphorylation in Fmr1 KO neurons is due to a deficit in complex formation of Mdm2 with PP2A.

MEF2 triggers Mdm2 nuclear export in WT but not Fmr1 KO neurons

Mdm2 phosphorylated at S163 and nearby S183 is more localized to the nucleus of fibroblasts and other cell types (11,12). As higher phosphorylation of S163 Mdm2 is observed in Fmr1 KO (Fig. 1A), we confirmed an elevated nuclear distribution of Mdm2 in Fmr1 KO hippocampi (Fig. 2A). Because MEF2 activation induces Mdm2 dephosphorylation in WT neurons, we suspected that MEF2 activation would alter the subcellular distribution of Mdm2. As predicted, MEF2 activation reduced the nuclear Mdm2 and increased the cytoplasmic Mdm2 in WT neurons (Fig. 2B). A slight difference regarding the ratio of nuclear vs cytoplasmic Mdm2 between Fig. 2A and B is likely due to the variability between tissues and culture lysates. However, consistent with the inability to be dephosphorylated, the nuclear or cytoplasmic distribution of Mdm2 in Fmr1 KO neurons is unaffected by MEF2 activation (Fig. 2B). To determine if the redistribution of Mdm2 by MEF2 activation results from enhanced nuclear export, we applied a nuclear export inhibitor Leptomycin-B (LMB) (18). As shown (Fig. 2C), LMB partially, but significantly, inhibited MEF2-induced cytoplasmic accumulation of Mdm2 in WT neurons, suggesting that MEF2-induced dephosphorylation of Mdm2 stimulates its nuclear export. To test whether nuclear export is required for synaptic accumulation of Mdm2 in WT neurons upon MEF2 activation, we again applied LMB and found LMB reduced, but did not completely block, MEF2-induced increases in Mdm2 in synaptoneurosomes of WT cultures (Fig. 2D). These results suggest that MEF2 triggers dephosphorylation and nuclear export of Mdm2 and this is necessary for synaptic localization of Mdm2.

Figure 2

MEF2-induced nuclear export and subsequent synaptic accumulation of Mdm2 in WT but not Fmr1 KO neurons. (A) Western blots of Mdm2 from WT or Fmr1 KO hippocampi after a cytoplasmic/nuclear fractionation. Tubulin and Lamin A/C serve as fractionation markers (n = 3). (B) Western blots of Mdm2, tubulin and Lamin A/C after a cytoplasmic/nuclear fractionation of from WT or Fmr1 KO cortical neuron cultures transfected with MEF2VP16ERtm and treated with vehicle or tamoxifen (n = 3). (C) Western blots of Mdm2, tubulin and Lamin A/C after a cytoplasmic/nuclear fractionation of WT or Fmr1 KO cortical neuron cultures transfected with MEF2VP16ERtm and treated with vehicle, tamoxifen and/or Leptomycin-B (LMB; n = 3). (D) Western blots of Mdm2, GluR1 and Tubulin after synaptoneurosome preparation of dissociated WT or Fmr1 KO cortical neurons transfected with MEF2-VP16ERtm and treated with vehicle, tamoxifen and/or Leptomycin-B as indicated (n = 3). *P < 0.05.

EF1α inhibits PP2A-dependent dephosphorylation of Mdm2 in Fmr1 KO neurons

In Fmr1 KO neurons, Mdm2 is hyperphosphorylated, localized in the nucleus and unaffected by MEF2 activation (Figs 1 and 2). We previously reported that an interacting protein of Mdm2, the eukaryotic elongation factor 1α (EF1α), inhibited MEF2-induced accumulation of Mdm2 at synapses in Fmr1 KO neurons (7), likely due to de-repression of EF1α translation (19,20). EF1α may inhibit MEF2-induced redistribution of Mdm2 by preventing Mdm2 dephosphorylation in Fmr1 KO neurons. To test this idea, we first determined the distribution of critical molecules in this pathway, EF1α, PP2A and phosphorylated Mdm2, in nuclear versus cytoplasmic fractions of the hippocampus. As shown (Fig. 3A), EF1α exhibits elevated distribution in both nucleus and cytoplasm of Fmr1 KO hippocampi while PP2A exhibits similarly distribution between WT and Fmr1 KO hippocampi. As predicted, phospho-Mdm2 is particularly elevated in the nucleus of Fmr1 KO hippocampi. These data support the hypothesis that elevated EF1α leads to phosphorylated Mdm2 in Fmr1 KO. To further validate this observation, we lentivirally expressed EF1α in WT cortical neuron cultures with the intention to mimic the altered signaling in Fmr1 KO. As shown (Fig. 3B), in comparison to vector control, overexpression of EF1α prevents MEF2-induced association between PP2A and Mdm2, and Mdm2 dephosphorylation. These findings suggest that reduction of EF1α in Fmr1 KO will correct the deficit of MEF2-induced Mdm2 dephosphorylation in Fmr1 KO.

Figure 3

The level of EF1α correlates with Mdm2 phosphorylation. (A) Western blots of EF1α, PP2A, Mdm2 and phospho-Mdm2 from WT or Fmr1 KO hippocampi after a cytoplasmic/nuclear fractionation. Tubulin and Lamin A/C serve as fractionation markers (n = 3). (B) Western blots of Mdm2 and PP2A after co-IP with PP2A from WT cortical neuron cultures co-transfected with control vector or pCMV6-EF1α along with MEF2VP16ERtm and treated with vehicle or tamoxifen for 6 h (n = 3). *P < 0.05.

To test this idea, we used lentiviral-mediated transfection to express a validated shRNA against EF1α (7) in Fmr1 KO neurons. In comparison to a previously validated control shRNA (7), knockdown of EF1α reduces basal Mdm2 phosphorylation by 36% ± 10% (P < 0.05), implicating EF1α in the basal hyperphosphorylation of Mdm2 in Fmr1 KO neurons. More importantly, EF1α knockdown also restored MEF2-induced dephosphorylation of Mdm2 as compared to control shRNA expressing Fmr1 KO neurons (Fig. 4A). To understand the mechanism by which EF1α inhibits Mdm2 dephosphorylation, we determined if Mdm2 phosphorylation affects its interaction with EF1α. With in vitro prepared WT or phosphomimetic Mdm2 (S163D-S183D-Mdm2), we found the phosphomimetic Mdm2 has a stronger interaction with EF1α (GST- EF1α), in comparison to WT-Mdm2 (Fig. 4B). Because knockdown of EF1α promotes Mdm2 dephosphorylation, we suspected that EF1α inhibits the interaction of Mdm2 and PP2A. To test this hypothesis, we mixed in vitro translated, ³⁵S-labelled, Mdm2 and EF1α, with immunoprecipitated PP2A complex from the WT mouse brain lysate to perform a protein interaction competition experiment. Increasing amounts of ³⁵S-EF1α were pre-incubated with a fixed amount of ³⁵S-Mdm2 and resulted in less ³⁵S-Mdm2 pull down with the PP2A complex (Fig. 4C). The results suggest that EF1α competes with PP2A for binding Mdm2. To determine if EF1α inhibits PP2A-mediated dephosphorylation of Mdm2 we performed an in vitro kinase and phosphatase assay. We first phosphorylated His-tagged Mdm2 by mixing with immunoprecipitated Akt from the WT mouse brain lysate (Fig. 4D, lane 2). The phosphorylated His-Mdm2 was subsequently used for the phosphatase reaction with immunoprecipitated PP2A again from the WT mouse brain lysate. As shown (Fig. 4D, lane 3 to 5), EF1α inhibits PP2A-mediated in vitro dephosphorylation of Mdm2 in agreement with our findings that EF1α inhibits the interaction of PP2A with Mdm2. As a control, the PP2A inhibitor, okadaic acid (2 nM), blocked the dephosphorylation, confirming a PP2A-dependent mechanism. These results suggest that elevated levels of EF1α in Fmr1 KO neurons inhibit MEF2-induced dephosphorylation and synaptic accumulation of Mdm2 by competing for interactions with PP2A.

Figure 4

EF1α inhibits PP2A-mediated dephosphorylation of Mdm2 in Fmr1 KO neurons. (A) Western blots of Mdm2 and p-Mdm2 from Fmr1 KO cortical neuron cultures transfected with MEF2VP16ERtm plus control or EF1α shRNA and treated with vehicle or tamoxifen for 6 h (n = 3). (B) PhosphoImager results of 35S-labelled WT-Mdm2 or S163D-183D-Mdm2 pulled down by GST-EF1α (n = 3). The purity of GST-EF1α is shown (B₃). (C) PhosphoImager results of preincubated, 35S-labelled Mdm2 and EF1α followed by a pull-down with immunoprecipitated PP2A. The arrowheads indicate the full-length Mdm2 and EF1α proteins. The quantification of pull-down (C₂), IgG control (C₃ and C₄) and a confirmation of PP2A activity after pull-down (C₅) are shown (n = 4). (D) Western blots of p-Mdm2 after a kinase reaction with immunoprecipitated Akt followed by a phosphatase reaction with immunoprecipitated PP2A. The purity of recombinant His-Mdm2 and GST-EF1α are shown (D₂). *P<0.05, **P<0.01

Mdm2 dephosphorylation is required for MEF2-induced PSD-95 degradation and functional synapse elimination

Our results suggest a model in where MEF2 induces dephosphorylation of Mdm2 and nuclear export which is required for its synaptic localization, PSD-95 degradation and synapse elimination. Furthermore, we hypothesize that the deficit in MEF2-induced dephosphorylation of Mdm2 in Fmr1 KO neurons may underlie the deficit in MEF2-triggered synaptic localization of Mdm2, PSD-95 degradation and synapse elimination. To test these hypotheses, we assessed the role of Mdm2 phosphorylation and dephosphorylation on MEF2-induced decreases in PSD-95 levels and functional synapse elimination in WT and Fmr1 KO neurons.

To determine if dephosphorylation of Mdm2 is required for MEF2-induced decreases in PSD-95 in WT neurons we used a lentiviral-mediated molecular replacement approach in dissociated neuron cultures to knockdown endogenous Mdm2, using a shRNA, and replace it with a myc-tagged wildtype Mdm2 or phosphomimetic of Mdm2 (S163D-S183D). Neurons were cotransfected with MEF2VP16ERtm. Five days post-transfection, dissociated neuron cultures were treated with tamoxifen to activate MEF2, or vehicle, for 6 h. As we have shown (7), this molecular replacement strategy maintains wildtype Mdm2 levels and minimizes the potential effect resulted from overexpressing Mdm2. Consistent with our model, MEF2 activation increases PSD-95 ubiquitination (Fig. 5A) and decreases PSD-95 levels (Fig. 5B) in cultures expressing WT-Mdm2, but not S163D-S183D-Mdm2, indicating that Mdm2 dephosphorylation is necessary for this process. S163D-S183D-Mdm2 did not affect PSD-95 levels in vehicle treated cultures (93% ± 12% in comparison to WT-Mdm2 expressing WT cultures).

Figure 5

MEF2-induced Mdm2 dephosphorylation is required for synapse elimination. (A,B) Western blots of Ubiquitin after IP with PSD-95 (A) or western blots of PSD-95 and Tubulin (B) from dissociated WT cortical neurons transfected with MEF2-VP16ERtm, Mdm2 shRNA, and a shRNA insensitive, myc-tagged WT Mdm2 or S163D-S183D-Mdm2. Cultures were treated with vehicle or tamoxifen for 6 h (n = 3). # indicates the heavy chain of IgG. Ubiquitinated PSD-95 has been previously shown as multiple high molecular weight bands instead of smears (7). (C) Simultaneous voltage-clamp recordings from biolistically transfected (with tamoxifen-inducible MEF2VP16ERtm and WT-Mdm2 or S163D-S183D-Mdm2; T) and neighbouring, untransfected (U) CA1 pyramidal neurons in wildtype organotypic hippocampal slice cultures treated overnight with vehicle or tamoxifen (10 μM). Representative evoked EPSCs are shown as insets (scale bars = 50pA/10ms) and representative mEPSC traces are shown to the right (scale bars = 10pA/500ms); n = 19–26 cell pairs. n = 17–29 cell pairs. *P < 0.05; **P < 0.01.

To determine if Mdm2 dephosphorylation is necessary for MEF2-induced functional elimination of excitatory synapses, we used the same molecular replacement strategy in wildtype organotypic hippocampal slice cultures to knockdown endogenous Mdm2, using a shRNA, and replace it with either WT Mdm2 or the phosphomimetic S163D-S183D-Mdm2. Neurons were co-transfected, using biolistics, with a tamoxifen-inducible, GFP-tagged MEF2VP16ERtm and treated, 1 day post-transfection, with either vehicle or 4OHT for 24 hrs. Our previous work demonstrated that activation of MEF2 for 24 h results in a functional and structural synapse elimination (4,5,7). Simultaneous whole cell patch clamp recordings from transfected and neighbouring untransfected CA1 neurons were used to assess the effect of MEF2 activation and Mdm2 on synaptic function. MEF2 activation (Tamoxifen treatment) decreased evoked EPSC amplitude and miniature (m) mEPSC frequency (Fig. 5C1) in comparison to untransfected neurons as previously shown (5,7). Consistent with our observation of PSD-95 levels, S163D-S183D-Mdm2 prevented the MEF2 induced decreases in evoked EPSCs or mEPSC frequency (Fig. 5C2). We also assessed vehicle-treated wildtype cultures transfected with the same constructs to determine the effects of Mdm2 and Mdm2 mutants on synaptic function in the absence of MEF2 activation (baseline synaptic transmission). Expression of WT-Mdm2 or S163D-S183D-Mdm2 did not affect evoked EPSCs or mEPSC frequency (Fig. 5C3, 5C4). Although S163D-S183D-Mdm2 had a trend to decrease evoked EPSC (P > 0.1), this does not confound our interpretation of its effects on MEF2-induced changes in evoked EPSCs (Fig. 5C4, Supplementary Material, Table S1). mEPSC amplitudes and passive electrical properties were unaffected by MEF2 or Mdm2 (Supplementary Material, Table S1). Paired-pulse facilitation (PPF) was unaffected by MEF2 activation, indicating that the MEF2-induced decrease in evoked EPSCs is not due to a decrease in presynaptic function and is consistent with a decrease in functional synapse number as previously demonstrated (5,7). Altogether, our data indicate that dephosphorylation of Mdm2 at S163/183 is necessary for MEF2-induced PSD-95 degradation and functional synapse elimination.

Because Mdm2 is not dephosphorylated in response to MEF2 activation in Fmr1 KO neurons, we hypothesized that replacement of Mdm2 with a dephosphomimetic would rescue MEF2-induced PSD95 degradation and synapse elimination in Fmr1 KO neurons. To test this idea, we used the same molecular replacement strategy, as described above for WT neurons, to knockdown endogenous Mdm2 and replace it with either a WT-Mdm2 or dephosphomimetic Mdm2 (S163A-S183A) in Fmr1 KO neurons. We first tested the ability of lentiviral expressed S163A-S183A-Mdm2 to rescue MEF2-induced PSD95 degradation in dissociated Fmr1 KO neuron cultures. These experiments were performed in parallel with the WT neuron experiments (Fig. 5). In Fmr1 KO cultures expressing WT Mdm2, MEF2 activation by tamoxifen treatment, failed to increase PSD-95 ubiquitination (Fig. 6A) or decrease PSD-95 levels (Fig. 6B), as previously reported (7). In contrast, S163A-S183A-Mdm2 expression increases PSD-95 ubiquitination while decreases PSD-95 level after MEF2 activation (Fig. 6A and B). S163A-S183A-Mdm2 had no effect on PSD-95 levels in vehicle-treated cultures (95% ± 17.5% in comparison to WT-Mdm2 expressing Fmr1 KO cultures). These data validate our hypothesis that dephosphorylation of Mdm2 is crucial to PSD-95 ubiquitination. They also confirmed our previous observation that PSD-95 ubiquitination is not sufficient to cause decreases in PSD-95 level without the activation of MEF2 and the expression of other MEF2 target genes (7).

Figure 6

Dephosphomimetic Mdm2 restores MEF2-induced synapse elimination in Fmr1 KO neurons. (A and B) Western blots of Ubiquitin after IP with PSD-95 (A) or western blots of PSD-95 and Tubulin (B) from dissociated Fmr1 KO cortical neurons transfected with MEF2-VP16ERtm, Mdm2 shRNA, and a shRNA insensitive WT Mdm2 or S163A-S183A-Mdm2. Cultures were treated with vehicle or Tamoxifen for 6 h (n = 3 for A and 4 for B). # indicates the heavy chain of IgG. (B) Simultaneous voltage-clamp recordings from biolistically transfected (with tamoxifen-inducible MEF2VP16ERtm and WT-Mdm2 or S163A-S183A-Mdm2; T) and neighbouring, untransfected (U) CA1 pyramidal neurons in Fmr1 KO organotypic hippocampal slice cultures treated overnight with vehicle or tamoxifen (10 μM). Representative evoked EPSCs are shown as insets (scale bars = 50pA/10ms) and representative mEPSC traces are shown to the right (scale bars = 10pA/500ms); n = 19–26 cell pairs. *P < 0.05; **P < 0.01.

To assess rescue of MEF2-induced functional synapse elimination in Fmr1 KO neurons, we used biolistic transfection of Fmr1 KO slice cultures, as described above for WT slice cultures, to knockdown Mdm2 and replace with either WT-Mdm2 or S163A-S183A-Mdm2 with or without MEF2 activation. Tamoxifen-induced MEF2 activation in WT-Mdm2 expressing Fmr1 KO neurons failed to depress evoked EPSCs or mEPSC frequency (Fig. 6C), consistent with previous observations (5,7). In contrast, MEF2 activation decreased evoked EPSCs and mEPSC frequency in S163A-S183A-Mdm2 expressing Fmr1 KO neurons (Fig. 6C2), effectively restoring functional MEF2-induced changes in synaptic transmission. In vehicle treated cultures, (in the absence of MEF2 activation), WT-Mdm2 or S163A-S183A-Mdm2 did not affect synaptic transmission (Fig. 6C3 and C4), passive membrane properties or PPF (Supplementary Material, Table S1). Altogether, our data suggest that Mdm2 dephosphorylation is required for MEF2-induced synapse elimination in both WT and Fmr1 KO neurons.

Discussion

Nuclear export of Mdm2 is required for its synaptic accumulation upon MEF2 activation

In previous work, we demonstrated that MEF2 induced synaptic localization of Mdm2 which resulted in ubiquitination and degradation of the Mdm2 substrate, PSD-95, and ultimately synapse elimination (7). However, the mechanisms by which MEF2 regulated Mdm2 were unknown. Here we demonstrate that MEF2 activation triggers a PP2A-mediated dephosphorylation of Mdm2 and subsequent nuclear export of Mdm2 (Fig. 7). By replacing Mdm2 with a phosphomimetic that cannot be dephosphorylated, we show here that dephosphorylation of Mdm2 is necessary for MEF2-induced degradation of PSD-95 and synapse elimination (Fig. 7). Proteasomal degradation of PSD-95 has been known to play a role in synaptic plasticity (21,22), but how Mdm2 is regulated in neurons or synapses has not been studied. In other cell types, the phosphorylation state of Mdm2 can determine its nuclear and cytoplasmic distribution (11). Similarly, we find in neurons that MEF2 activation results in a redistribution of Mdm2 from the nucleus to the cytoplasm and nuclear export is required for synaptic accumulation of Mdm2. Together, these findings suggest that MEF2 activation promotes synaptic accumulation of Mdm2 by initiating dephosphorylation and nuclear export of Mdm2. MEF2 likely also regulates other mechanisms to target Mdm2 to synapses and/or PSD-95 once in the cytoplasm.

Figure 7

Working model of Mdm2 dephosphorylation in MEF2-induced synapse elimination in WT neurons and the molecular basis of the deficit in Fmr1 KO neurons. (A) In WT neurons, MEF2 activation stimulates association of PP2A with Mdm2 which dephosphorylates Mdm2 and promotes its nuclear export. Movement of Mdm2 to the cytoplasm may be sufficient to confer synaptic accumulation or another mechanism may be required. Synaptic accumulation of Mdm2 results in ubiquitination and proteasomal degradation of PSD-95 and synapse elimination. (B) In Fmr1 KO neurons, basal levels of Mdm2 phosphorylation are elevated and MEF2 activation fails to cause dephosphorylation of Mdm2. Elevate levels of EF1α are bound to Mdm2 and compete for binding of Mdm2 with PP2A.

Interestingly, expression of the dephosphomimetic Mdm2 alone in Fmr1 KO neurons, which promote PSD-95 ubiquitination, was insufficient to cause PSD-95 degradation or functional synapse elimination; MEF2 activation was also required. Our previous work demonstrated that MEF2-induction of PCDH10, a MEF2-target gene that traffics ubiquitinated PSD-95 to the proteosome, is required for both PSD-95 degradation and synapse elimination. Thus, without MEF2 activation, degradation of PSD-95 would not be expected to occur. We also find that the MEF2-induced transcript, Arc, is necessary for synapse elimination, although its necessity for PSD-95 degradation is unknown (8). Other MEF2-induced transcripts may also be required for PSD-95 degradation and synapse elimination.

MEF2 activation causes dephosphorylation of Mdm2

MEF2 activation results in Mdm2 dephosphorylation which is blocked by PP2A inhibition. However, MEF2 activation does not increase overall PP2A activity, but stimulates an association of the PP2A complex with Mdm2; a process that is deficient in Fmr1 KO neurons. The PP2A complex consists of a scaffold subunit A, a regulatory subunit B, and a catalytic subunit C (23). The B subunit is more heterogeneous and directs the substrate specificity of PP2A. The substrate specificity of PP2A can be also be regulated through the localization of the PP2A complex and interaction with associated proteins (23). MEF2 has at least 300 direct target genes identified through a genome-wide search (10). MEF2-induced transcripts may directly or indirectly modulate the levels of PP2A subunits, PP2A interacting proteins or other regulators of PP2A localization that facilitate an interaction of PP2A and Mdm2. We have also not ruled out the possibility that MEF2 inhibits an Mdm2 kinase (24).

The deficit of Mdm2 dephosphorylation in fragile X syndrome

MEF2-induced synaptic accumulation of Mdm2, ubiquitination and degradation of PSD-95 and synapse elimination are deficient in cortical and hippocampal neurons of Fmr1 KO mice. Our new data find that MEF2 fails to promote an interaction of PP2A with Mdm2 and dephosphorylation of Mdm2 in Fmr1 KO neurons, thus suggesting a mechanism for the deficits in Mdm2 accumulation (Fig. 7B). EF1α is a direct interacting protein of Mdm2 (25) and is elevated in Fmr1 KO neurons due to the loss of translation suppression (7,20). Knockdown of EF1α restores MEF2-triggered synaptic accumulation of Mdm2, PSD-95 degradation and synapse elimination in Fmr1 KO mice. How EF1α inhibits MEF2-Mdm2 movement to the synapse was unknown. We find that EF1α competes with the PP2A complex for interactions with Mdm2 and prevents PP2A-mediated dephosphorylation of Mdm2 in an in vitro assay. Because EF1α has a greater affinity for phosphorylated Mdm2, it can inhibit dephosphorylation of Mdm2 by sequestering phosphorylated Mdm2 and preventing its interaction with PP2A. Therefore, limiting the levels of EF1α protein is likely an important means to allow phosphorylated Mdm2 to interact with PP2A. FMRP suppression of EF1α mRNA translation may be such a mechanism to accomplish this. Mdm2 phosphorylation at serine-163 is primarily mediated by the kinase Akt (24). Because of the known increase in phosphatidylinositol 3'-kinase (PI3K)-Akt signalling in Fmr1 KO neurons (26,27), this pathway may also contribute to higher Mdm2 phosphorylation at S163 in Fmr1 KO neurons. Inhibition of the PI3K-Akt pathway is a candidate therapeutic for FXS (26,27), and it would be of particular interest in the future to investigate the role of the Akt pathway on Mdm2 phosphorylation and/or MEF2-induced synapse elimination, especially in Fmr1 KO neurons.

FXS in humans and mice is associated with an excess of dendritic spines. In Fmr1 KO mice there is good evidence to suggest there are deficits in developmental and experience-dependent synapse elimination (5,28–30). Such deficits likely lead to the known hyperconnectivity and excitability of cortical networks (28,31) and may contribute to cognitive and sensory processing deficits associated with FXS. Our work has identified a cellular model of experience and activity-dependent synapse elimination that relies on the activity-dependent transcription factor MEF2 and Fmr1 (5,7). Here we have further determined the detailed molecular mechanisms by which MEF2 and Fmr1 mediate synapse elimination through dephosphorylation of the E3 ligase for PSD-95, Mdm2. Understanding the normal and deficient molecular pathways will inform the development of targeted therapeutic interventions to correct or aid synapse refinement in FXS and related neurodevelopmental disorders.

Materials and Methods

All experimental protocols involving mice were performed in accordance with the guidelines and regulations set forth by the Institutional Animal Care and Use Committee at The University of Texas Southwestern Medical Center and University of Illinois at Urbana-Champaign.

Reagents, plasmids and transfection

Okadaic acid, Leptomycin B and 4-Hydroxytamoxifen were from Sigma. GST-EF1α and His-Mdm2 protein were from Abcam. Mdm2 shRNA (TRCN0000039484), EF1α (TRCN0000123775) were from Sigma. Control shRNA were from Sigma (Non-target shRNA) and GeneCopoeia (CSHCTR001-HIVU6). pCMV6-EF1α was from Origene. Myc-tagged Mdm2s were obtained from Addgene and sub-cloned into FUGW vector. Transfection in HEK-293 cells is done by Lipofectamine-3000 reagent (Life Technologies). All neuronal transfections in this study were done by either lentivirus (for dissociated cultures) or biolistic gene gun transfection (for slice cultures). Lentiviral particles were generated by transfecting HEK-293 cells with three packaging vectors (VSVG, pMDL and RSV-REV) and lentiviral shRNA plasmid or overexpression plasmid. 6 h after transfection, the culture medium was replaced with Neurobasal A medium plus B-27 supplement (Invitrogen). Virus-containing media were harvested 48–60 h post-transfection and applied onto dissociated cultured neuron directly. Slice cultures were biolistically transfected at 5–6 Days-in-vitro (DIV). The procedure and gold bullet preparation were performed with the Helios Gene Gun system (BioRad) according to the manufacturer’s protocols (32).

Phosphatase activity assay

To perform the PP2A phosphatase assay (reagents from Millipore), the WT or Fmr1 KO cortical neuron cultures lentivirally transfected with MEF2-VP16ERtm were harvested in ice-cold PBS after vehicle or tamoxifen treatment. After washing to PBS two times, cells were lysed by brief sonication. The supernatant containing soluble proteins were added to a 96-well plate pre-coated with PP2A substrate: Threonine Phosphopeptide (K-R-pT-I-R-R). The phosphate released from the dephosphorylation reaction was detected by Malachite Green according to the manufacturer protocol.

In vitro kinase and phosphatase assay

The in vitro kinase reaction was performed as described (33) with immunoprecipitated Akt from WT total brain lysate and recombinant His-Mdm2 (Abcam). The phosphatase assay was performed by incubating immunoprecipitated PP2A also from the WT brain lysate with the samples after kinase reaction for another 30 min. The final reaction mixture was then subjected onto SDS-PAGE and western blotting with anti-phospho-Mdm2 antibody.

In vitro translation, GST pull down and western blotting

In vitro translation was conducted using TNT Cell‐Free Protein Expression (Promega) with 35S-Met/Cys from PerkinElmer. For GST pull-down, GST-EF1α was bound to glutathione beads and incubated with in vitro translated WT Mdm2 or S163D-183D Mdm2 for 1 h at 4 °C. After washing, the samples were subjected to SDS-PAGE and exposed to a PhosphorImager (Amersham Biosciences) for quantification. For Western blotting after SDS-PAGE, the gel was transferred onto a polyvinylidene fluoride (PVDF) membrane. After blocking with 1% BSA in TBS, the membrane was incubated with the appropriate primary antibody for 4 hrs at room temperature, followed by a 30-min wash with TBS. The membrane was then incubated with an HRP-conjugated secondary antibody for 1 h, followed by another 30-min wash. The membrane was then developed with an ECL (enhanced chemiluminescence) solution as described (34). Primary antibodies used in this study were anti-Tubulin (Abcam), anti-PSD-95 (Santa Cruz Biotechnology), anti-PP2A (Millipore), anti-Akt (Cell Signaling), anti-Mdm2 (Santa Cruz Biotechnology), anti-phosphor-Mdm2 (Cell Signaling), anti-GST (Cell Signaling), anti-Actin (Genscript), anti-LaminA/C (Santa Cruz Biotechnology), anti-Myc (Santa Cruz Biotechnology) and anti-EF1α (Abcam).

Nuclear and cytoplasmic fractionation and synaptoneurosome preparation

Nuclear/cytoplasmic fractionation was conducted by the NE-PER Nuclear and Cytoplasmic Extraction Reagents Kit (Pierce Chemical) as described previously (33) and according to manufacturer’s protocol. Lamin A/C and β-Tubulin were served as nuclear and cytoplasmic markers, respectively. Synaptoneurosome was prepared as previously described (7). In brief, cells were scraped off in a buffer: 118 mM NaCl, 4.7 mM KCl, 1.2 mM MgSO₄, 2.5 mM CaCl₂, 1.53 mM KH₂PO₄, 212.7 mM glucose, 1 mM DTT (pH 7.4), and protease inhibitor cocktail (Calbiochem) followed by cell homogenization with dounce homogenizer. The cell lysate was passed through two 100 µm nylon filters and one 10 µm nylon filter. The synaptoneurosome was then pelleted by centrifugation at 1,000 g for 10 min.

Electrophysiology

Simultaneous whole-cell voltage-clamp recordings were obtained from transfected and neighbouring untransfected neurons (minimum of three independent slice cultures) under visual guidance using IR-DIC and fluorescence to identify transfected neurons as described (35). Recordings from slice cultures were made at 30 °C in a submersion chamber perfused at 3 ml/min with artificial cerebrospinal fluid (aCSF) containing (in mM): 119 NaCl, 2.5 KCl, 26 NaHCO₃, 1 NaH₂PO₄, 11 D-Glucose, 4 CaCl₂, 4 MgCl₂, 0.1 picrotoxin, 0.002 2-chloro-adenosine; pH 7.28, 310 mOsm and saturated with 95% O₂/5%CO₂. aCSF was supplemented with 1 μM TTX for mEPSC measurements. Whole cell recording pipettes (∼4–6 M&Omega;) were filled with intracellular solution containing (in mM): 0.2 EGTA, 130 K-gluconate, 6 KCl, 3 NaCl, 10 HEPES, 10 sucrose, 4 ATP-Mg, 0.4 GTP-Na, 14 phosphocreatine-Tris, 2 QX-314; pH 7.2, 285 mOsm. EPSCs were evoked by a single bipolar electrode placed in stratum radiatum of area CA1 50–100 μm from the recorded neurons with monophasic current pulses (5–120 μA, 100–200 μs). The stimulation current for each slice is adjusted to evoke a 50–100pA monosynaptic EPSC in the untransfected neuron of each cell pair. Therefore, due to the variable stimulation current and placement of the stimulating electrode across slices, comparisons of evoked EPSC amplitudes are only valid between transfected and untransfected neurons within the same slice (as in Supplementary Material, Table S1) and comparisons of evoked EPSC amplitudes across slices (animals or drug treatment conditions) are not valid.

For all recordings, input and series resistances were measured in voltage clamp with a 400-ms, –10 mV step from a –60 mV holding potential (filtered at 30 kHz, sampled at 50 kHz). Cells were only used for analysis if they met the following criteria: series resistance < 25 MΩ and stable throughout the experiment, resting membrane potential < −35mV; input resistance > 75 MΩ; evoked EPSC > 20pA (for the untransfected cell). Waveforms were filtered at 3 kHz, acquired and digitized at 10 kHz on a PC using custom software (LabView; National Instruments). Data analysis was also performed in custom-designed software in LabView (35). mEPSCs were detected off-line using an automatic detection program (MiniAnalysis; Synaptosoft Inc.) with a detection threshold set at a value greater than at least 2 standard deviations of the noise values, followed by a subsequent round of visual confirmation. The detection threshold remained constant for the duration of each experiment. For evoked EPSCs shown in figures the stimulation artefact has been digitally removed for clarity.

Statistical analysis

For electrophysiology assays, a paired t-test or Wilcoxon matched pairs signed rank test was used depending on whether the data was normally distributed (see Supplementary Material, Table S1). For multiple comparisons, a two-way ANOVA and Bonferroni post-hoc test were performed. Independent t-tests were used for Figs 1C and D,3B. One sample t-tests were used when experimental group is normalized to control groups (Figs 1A and B,2A and B, 2D,3A,4A and B). In all figures, error bars represent SEM and ^∗P < 0.05, ^∗∗P < 0.01.

Supplementary Material

Supplementary Material is available at HMG online.

Conflict of Interest statement. None declared.

Funding

This research was supported by the ICR start-up fund from University of Illinois at Urbana-Champaign (N-P.T.), the National Institutes of Health HD052731 (K.M.H.), F32HD062120 (J.R.W.), F32HD069111 (N-P.T.), the Simons Foundations (SFARI# 206919, K.M.H. and SFARI# 336605, N-P.T.) and NARSAD Young Investigator Award (N-P.T.).