Report

A Calcium-Regulated MEF2 Sumoylation Switch Controls Postsynaptic Differentiation

See allHide authors and affiliations

Science  17 Feb 2006:
Vol. 311, Issue 5763, pp. 1012-1017
DOI: 10.1126/science.1122513

Abstract

Postsynaptic differentiation of dendrites is an essential step in synapse formation. We report here a requirement for the transcription factor myocyte enhancer factor 2A (MEF2A) in the morphogenesis of postsynaptic granule neuron dendritic claws in the cerebellar cortex. A transcriptional repressor form of MEF2A that is sumoylated at lysine-403 promoted dendritic claw differentiation. Activity-dependent calcium signaling induced a calcineurin-mediated dephosphorylation of MEF2A at serine-408 and, thereby, promoted a switch from sumoylation to acetylation at lysine-403, which led to inhibition of dendritic claw differentiation. Our findings define a mechanism underlying postsynaptic differentiation that may modulate activity-dependent synapse development and plasticity in the brain.

The MEF2 family of transcription factors is highly expressed in the brain when neurons undergo dendritic maturation and synapse formation (1). MEF2A is especially abundant in granule neurons of the cerebellar cortex throughout the period of synaptogenesis (1) (fig. S1). In view of reported functions for transcription factors in distinct aspects of dendritic morphogenesis (24), we investigated a potential role of MEF2A in synaptic dendritic development in the cerebellar cortex.

During cerebellar development, granule neuron dendritic morphogenesis culminates in the differentiation of dendritic claws on which mossy fiber terminals and Golgi neuron axons form synapses (58). To visualize granule neurons undergoing postsynaptic differentiation, we transfected organotypic cerebellar slices prepared from postnatal day 9 (P9) rat pups with an expression plasmid encoding green fluorescent protein (GFP) (9). Transfected granule neurons in the internal granule layer had the typical small cell body with associated parallel axonal fibers and few dendrites (Fig. 1, A and B). Many dendrites harbored structures with the appearance of dendritic claws that were identified on the basis of classic descriptions as (i) located at the end of a dendrite, (ii) having cuplike or sicklelike appearance, and (iii) displaying undulating or serrated inner surfaces (58) (Fig. 1, C and D). Dendritic claws showed punctate expression of the postsynaptic protein PSD95 (Fig. 1D). PSD95 puncta density was greater in the claw region than in the shaft of dendrites (Fig. 1D). Thus, granule neuron dendritic claws in cerebellar slices represent sites of postsynaptic differentiation.

Fig. 1.

MEF2A is essential in dendritic claw morphogenesis in cerebellar slices. (A) Schematic of developing rat cerebellar cortex: external granule (EGL), molecular (ML), Purkinje cell (PL), and internal granule (IGL) layers. Granule neurons (GN) elaborate dendritic claws that contact mossy fiber (MF) axons. (B to D) Representative images of GFP-positive granule neurons within transfected cerebellar slices (9). Asterisk and arrowheads indicate the cell body and axons, respectively (B). Numbers in (C) indicate dendritic claws shown at higher magnification at right; arrows indicate dendritic claws. Neurons are immunostained for GFP and PSD95 in (D). The claw and the shaft of the dendrite are indicated, and arrowheads indicate PSD95-positive puncta within the claw (9). PSD95 puncta density is significantly higher in the claws than in the shaft of dendrites (P < 0.001, t test; total neurons measured, n = 22) (9). (E) (Left) Cerebellar slices transfected with the control U6 or U6/mef2a plasmid, together with expression plasmids encoding MEF2A-WT or MEF2A-Res and GFP, and analyzed for dendritic claws (9). (Upper right) Lysates of 293T cells transfected with control U6 or U6/mef2a plasmid together with expression plasmids for MEF2A-WT or MEF2A-Res were immunoblotted for MEF2A (9). (Lower right) The number of dendritic claws was significantly reduced by MEF2A knockdown and in MEF2A-WT–expressing neurons but not in MEF2A-Res–expressing neurons in the presence of MEF2A knockdown, when compared with U6-transfected controls (P < 0.001, ANOVA followed by Bonferroni-Dunn post hoc test; total neurons measured, n = 262). Scale bars: 5 μm (C); 3 μm (D and E).

We determined the effect of MEF2A knockdown induced by RNA interference (RNAi) on granule neuron dendritic morphogenesis. We transfected cerebellar slices with the U6/mef2a plasmid that encodes MEF2A hairpin RNAs (MEF2AhpRNA) or the control U6 plasmid, together with a GFP expression plasmid (9). The MEF2AhpRNA-expressing granule neurons had 60% fewer dendritic claws than control U6-transfected neurons, and their dendrites displayed tapered or bulbous tips instead of claws (Fig. 1E). In these dendrites, PSD95 puncta density was low in the tip region and no greater than in the shaft (fig. S2). The MEF2AhpRNA-induced dendritic claw phenotype was not due to a reduction in dendritic growth (fig. S3). Together, these results suggest that MEF2A plays a key role in the morphogenesis of dendritic claws in the cerebellar cortex.

To exclude the possibility that the dendritic phenotype induced by MEF2A knockdown is the result of off-target effects of RNAi, we performed a rescue experiment. MEF2A RNAi induced the effective knockdown of MEF2A protein encoded by wild-type (WT) MEF2A cDNA but failed to reduce the expression of MEF2A encoded by an RNAi-resistant cDNA (MEF2A-Res) (Fig. 1E). In cerebellar slices, MEF2A-Res, but not MEF2A-WT, reversed the MEF2AhpRNA-induced dendritic claw phenotype (Fig. 1E). Expression of MEF2A-Res induced dendritic claws of similar number, morphological appearance, and PSD95 density as those in control U6-transfected neurons (Fig. 1, D and E; fig. S2). These experiments indicate that the MEF2AhpRNA-induced dendritic claw phenotype is the result of the specific knockdown of MEF2A.

To establish MEF2A function in dendritic claw development in vivo, we induced MEF2A knockdown in the postnatal cerebellum using electroporation-mediated gene transfer (9, 10). We injected a control U6 or U6/mef2a plasmid that also encoded GFP into the cerebellar cortex of P3 rat pups and assessed dendritic claws in the cerebellum of these animals at P12. Granule neurons in control-transfected cerebella had PSD95-positive postsynaptic dendritic claws at the tips of their dendrites (Fig. 2, A and B). In addition, expression of the presynaptic protein synaptophysin was found juxtaposed to the surface of ∼80% of dendritic claws (Fig. 2C). As in cerebellar slices, MEF2A knockdown reduced the number of dendritic claws in the cerebellum in vivo (Fig. 2D). Together, our findings suggest a physiological, cell-autonomous function for MEF2A in the morphogenesis of dendritic claws in the developing cerebellar cortex.

Fig. 2.

MEF2A is required for postsynaptic dendritic differentiation in vivo. (A) The cerebellum of P3 rat pups was injected and electroporated with the U6-cmvGFP plasmid (9). Representative electroporated granule neurons with cell bodies in the IGL. Scale bars: 50 μm (left), 10 μm (middle), 5 μm (right). Asterisk, arrowheads, and arrows, respectively, indicate the cell body, parallel fibers, and dendritic claws of granule neurons. (B and C) Sections of U6-cmvGFP–electroporated cerebellum were immunostained with antibodies to GFP and PSD95 (B) or synaptophysin (C). An image of a dendritic claw analyzed as in Fig. 1D is shown, and PSD95 density is quantified below (B). PSD95 puncta density is significantly higher in the claw than in the shaft of dendrites (P < 0.001, t test; total neurons measured, n = 13). (D) Representative granule neuron dendrites from cerebella transfected with the control U6-cmvGFP or U6/mef2a-cmvGFP plasmid. Quantification indicates that MEF2A knockdown significantly reduces dendritic claw number in vivo (P < 0.001, t test; total neurons measured, n = 124). Scale bar, 5 μm (B to D).

Calcium signaling strongly influences the activity of MEF2s (11, 12). Calcium entry through voltage-sensitive calcium channels (VSCCs) triggers MEF2 phosphorylation at distinct sites and calcineurin-mediated dephosphorylation at undetermined sites, which lead to enhanced MEF2-dependent transcription (1114). Calcineurin has emerged as a critical regulator of dendritic spine morphology in hippocampal neurons (15). We reasoned that calcineurin might, therefore, control postsynaptic dendritic differentiation via a MEF2-regulated transcriptional mechanism.

We first determined the site of calcineurin-mediated dephosphorylation of MEF2A. Because calcineurin stimulates MEF2-dependent transcription, we reasoned that calcineurin might induce the dephosphorylation of MEF2A at Ser408, whose phosphorylation inhibits MEF2-dependent transcription (16, 17). Using antibodies that recognize MEF2A when phosphorylated on Ser408 (fig. S4) (9), we found that endogenous MEF2A was phosphorylated on Ser408 in neurons (Fig. 3A). When neurons were depolarized, MEF2A underwent rapid and robust dephosphorylation at Ser408, an effect that was blocked in neurons treated with nimodipine, an inhibitor of L-type VSCCs, or cyclosporin A (CsA), an inhibitor of calcineurin (Fig. 3A; fig. S5). In human 293T embryonic kidney cells, MEF2A was constitutively phosphorylated at Ser408, and coexpression of activated calcineurin induced dephosphorylation of MEF2A at this site (Fig. 3B). These results suggest that calcineurin mediates activity-induced dephosphorylation of MEF2A at Ser408 in neurons.

Fig. 3.

MEF2A Ser408 dephosphorylation promotes a sumoylation-to-acetylation switch at MEF2A Lys403. (A) Activity-dependent dephosphorylation of MEF2A Ser408 in neurons. Lysates of granule neurons depolarized with 25 mM KCl were immunoblotted for MEF2A, phospho-Ser408 MEF2A, or ERK1/2. Nimodipine (Nim), 20 μM; CsA, 4 μM. (B) Calcineurin dephosphorylates MEF2A Ser408. Lysates of 293T cells transfected with MEF2A and hemagglutinin (HA)-tagged constitutively active calcineurin subunit A (HA-CnA*) or control vector were immunoblotted for MEF2A, MEF2ApS408, or HA. NS, nonspecific band. (C) MEF2A Lys403 is both sumoylated and acetylated. Lysates and GAL4-immunoprecipitates of 293T cells transfected with wild type, K403R or E405D mutant G4-MEF2A, and HA-SUMO1 were immunoblotted for HA, acetyl-lysine (AcK), or MEF2A. Both K403R and Glu405 replaced by Asp (E405D) mutations block MEF2A sumoylation, which indicates that Lys403 is a bona fide SUMO acceptor site (24). (D) Calcineurin inhibits sumoylation and promotes acetylation of MEF2A. Cells transfected with G4-MEF2A and HA-CnA* were analyzed as in (C). (E) Ser408 is required for MEF2A sumoylation. Cells transfected with wild type, K403R or S408A mutant G4-MEF2A, and HA-SUMO1 were analyzed as in (C). (F and G) Endogenous MEF2A is sumoylated in neurons. (F) Granule neurons in nondepolarizing concentrations of KCl (5 mM) were lysed in the presence or absence of the isopeptidase inhibitor NEM and immunoblotted for MEF2A. Asterisk indicates a form of MEF2A of appropriate size for sumoylated MEF2A (fig. S8). (G) Granule neurons in media containing nondepolarizing (5 mM) or depolarizing (25 mM) concentrations of KCl in the presence of Nim or its control vehicle dimethyl sulfoxide (DMSO) were lysed in the presence of NEM and immunoblotted as in (F). MEF2A is also acetylated in neurons in a VSCC- and calcineurin-dependent manner (fig. S9).

Intriguingly, Ser408 lies near a conserved SUMO (small ubiquitin-like modifier) acceptor site centered at Lys403 within a domain of MEF2A that represses transcription (17, 18). Sumoylation of transcription factors typically induces transcriptional repression (19, 20). MEF2 proteins can function as activators or repressors of transcription in a signal-dependent manner (11, 12, 21). We asked whether Ser408 dephosphorylation might regulate MEF2A sumoylation at Lys403 and MEF2's transcriptional repression function. First, we demonstrated that MEF2A Lys403 is modified by sumoylation in vitro and in cells (Fig. 3C; fig. S6). Interestingly, MEF2A was also acetylated in cells in a Lys403-dependent manner (Fig. 3C).

To determine how dephosphorylation of Ser408 might regulate the Lys403 modifications, we expressed the MEF2A transactivation domain fused to the DNA binding domain of GAL4 (G4-MEF2A) together with a constitutively active form of calcineurin. Activated calcineurin inhibited sumoylation and enhanced acetylation of MEF2A in cells (Fig. 3D). A G4-MEF2A mutant in which Ser408 was replaced with alanine (G4-MEF2AS408A) had reduced sumoylation and enhanced acetylation as compared with G4-MEF2A (Fig. 3E). Expression of the SUMO E2 ligase Ubc9 in cells increased sumoylation and inhibited the acetylation of G4-MEF2A, but not of G4-MEF2AS408A (fig. S7). Together, these results suggest that the calcineurin-induced dephosphorylation of MEF2A at Ser408 promotes a sumoylation to acetylation switch at Lys403.

In granule neurons, endogenous sumoylated MEF2A was detected as an N-ethylmaleimide (NEM)–sensitive MEF2 immunoreactive band of appropriate molecular size by immunoblotting with antibodies to MEF2A (Fig. 3F; fig. S8). Membrane depolarization of neurons led to an almost complete reduction of sumoylated MEF2A, an effect that tightly correlated with Ser408 dephosphorylation (Fig. 3G). Endogenous MEF2A was acetylated in depolarized neurons (fig. S9). Incubation of depolarized neurons with the VSCC inhibitor nimodipine or the calcineurin inhibitor CsA increased sumoylation and decreased acetylation of endogenous MEF2A (Fig. 3G; fig. S9).

We assessed the consequences of Ser408 dephosphorylation-induced Lys403 modifications of MEF2A on transcription. Replacement of Ser408 with alanine (S408A) or Lys403 with arginine (K403R) in G4-MEF2A similarly enhanced transcription in neurons or 293T cells (Fig. 4A; fig. S10). The S408A and K403R mutants of MEF2A are both deficient in sumoylation, yet the S408A mutant enhances MEF2A acetylation (Fig. 3E). In view of these results, the phenocopy of the S408A and K403R mutants in the reporter assay supports the conclusion that sumoylation is the critical modification of Lys403, leading to repression of transcription. Acetylation of Lys403 may thus serve to prevent sumoylation of MEF2A.

Fig. 4.

A calcium-regulated Lys403-sumoylated transcriptional repressor form of MEF2A promotes dendritic claw differentiation. (A) Granule neurons were transfected with wild-type and K403R or S408A mutant G4-MEF2A, together with the p5G4-E1b-luc and the pRL-TK reporter genes (9). The K403R and S408A mutants of G4-MEF2A had significantly greater transcriptional activity than wild-type G4-MEF2A (P < 0.02, ANOVA followed by Bonferroni-Dunn post hoc test; n = 6). (B) Cells coexpressing MEF2A and increasing amounts of MEF2A-SUMO together with a 3-MRE luciferase reporter gene (pMEF2x3-luc) and pRL-TK. MEF2A-SUMO significantly reduced MRE-dependent transcription at all amounts tested (P < 0.01, ANOVA followed by Bonferroni-Dunn post hoc test; n = 5) (9). Lysates were also immunoblotted for MEF2 or extracellular signal–regulated kinase types 1 and 2 (ERK1/2). (C) Cerebellar slices were transfected with control vector, MEF2A, or MEF2A-SUMO. Dendritic claw number was significantly increased by MEF2A-SUMO when compared with MEF2A-expressing or vector-transfected neurons (P < 0.005, ANOVA followed by Bonferroni-Dunn post hoc test; total neurons measured, n = 53) (9). (D) Cerebellar slices transfected as in Fig. 1C were treated with Nim (20μM), CsA (4μM), or vehicle (DMSO). Dendritic claw number was significantly increased by Nim or CsA as compared with vehicle (P < 0.001, ANOVA followed by Bonferroni-Dunn post hoc test; total neurons measured, n = 162) (9). (E) (Left) Cerebellar slices transfected with the U6/mef2a plasmid together with MEF2A-Res or with K403R or S408A mutants of MEF2A-Res were analyzed as in Fig. 1E. Scale bar, 3 μm. The number of claws was significantly higher in granule neurons that expressed MEF2A-Res but not MEF2A-ResK403R or MEF2A-ResS408A in the presence of MEF2A knockdown when compared with MEF2A knockdown alone (P < 0.001, ANOVA followed by Bonferroni-Dunn post hoc test; total neurons measured, n = 133). (Upper right) Lysates of 293T cells transfected with the control or U6/mef2a plasmid together with MEF2A-WT, MEF2A-Res, K403R or S408A mutant of MEF2A-Res, and FLAG-14-3-3 were immunoblotted with the indicated antibodies. (F) Cerebellar slices transfected with the U6/mef2a plasmid, together with MEF2A-Res or MEF2A-ResS408A-SUMO. The number of claws was significantly higher in neurons expressing MEFA-Res or MEF2A-ResS408A-SUMO in the presence of MEF2A knockdown when compared with MEF2A knockdown alone (P < 0.005, ANOVA followed by Bonferroni-Dunn post hoc test; n = 105).

Fusion of a SUMO moiety to transcription factors mimics the effect of SUMO that is covalently linked to proteins on the native lysine (22). A MEF2A-SUMO fusion protein potently inhibited the ability of coexpressed wild-type MEF2A to induce a MEF2-responsive [MEF2 response element (MRE)–dependent] reporter gene in cells (Fig. 4B). Sumoylation did not appear to alter MEF2A's subnuclear localization or stability (fig. S11). Together, our findings suggest that Lys403-sumoylated MEF2A represses transcription and that Ser408 dephosphorylation inhibits Lys403 sumoylation and, thereby, derepresses MEF2A-induced transcription.

To characterize the role of the calcium-MEF2A signaling pathway in dendritic claw morphogenesis in the cerebellar cortex, we first tested the effect of MEF2A sumoylation on dendritic claw differentiation. Expression of MEF2A-SUMO increased the number of dendritic claws compared with MEF2A-expressing or control-transfected neurons (Fig. 4C), which suggested that a transcriptional repressor form of MEF2A stimulates dendritic claw differentiation. In support of this conclusion, expression of a protein in which the MADS/MEF2 domains were fused to the transcriptional repressor Engrailed (MEF2-EN), which potently repressed MRE-dependent transcription, led to an increase in the number of dendritic claws in cerebellar slices (fig. S12).

We next determined the effect of the endogenous calcium-induced cascade of modifications at Ser408 and Lys403 of MEF2A on dendritic claw differentiation. Incubation of cerebellar slices with nimodipine or CsA increased dendritic claw number (Fig. 4D), which suggested that VSCC or calcineurin activation inhibits dendritic claw development. We then tested the ability of the S408A or K403R mutant of MEF2A-Res to rescue the MEF2AhpRNA-induced dendritic claw phenotype in cerebellar slices. The S408A mutant mimics calcineurin-induced dephosphorylation of MEF2A Ser408, whereas both S408A and K403R mutants of MEF2A are deficient in sumoylation and transcriptional repression (Figs. 3 and 4A). In contrast to MEF2A-Res, neither mutant of MEF2A-Res reversed MEF2AhpRNA-inhibition of dendritic claw morphogenesis (Fig. 4E). Fusion of SUMO with the S408A mutant of MEF2A-Res protein gave this protein the ability to induce dendritic claw differentiation in the presence of MEF2A knockdown (Fig. 4F). Thus, the rescue experiments suggest that an endogenously sumoylated transcriptional repressor form of MEF2A promotes dendritic claw differentiation. Together, our results also suggest that the calcium-induced Ser408 dephosphorylation and consequent inhibition of Lys403 sumoylation of MEF2A suppress dendritic claw morphogenesis (see model in fig. S13).

We investigated the mechanism by which sumoylated MEF2A promotes dendritic claw differentiation. Transcription of the gene encoding the transcription factor Nur77 is induced by a calcineurin-MEF2 signaling pathway in immune cells (23). Endogenous MEF2A was found to occupy the endogenous Nur77 promoter in granule neurons (Fig. 5A). Nur77 mRNA abundance and Nur77 promoter–mediated transcription increased in depolarized neurons in a VSSC- and calcineurin-dependent manner (Fig. 5, B and C). Both MEF2A-SUMO and MEF2-EN inhibited depolarization-induced Nur77 transcription (Fig. 5D). In cerebellar slices, expression of a dominant interfering form of Nur77 increased the number of dendritic claws (Fig. 5E). Thus, Nur77 represents a MEF2A target gene whose repression by sumoylated MEF2A contributes to dendritic claw differentiation (fig. S13).

Fig. 5.

Nur77 repression by MEF2A-SUMO contributes to dendritic claw morphogenesis. (A) Endogenous MEF2A is associated with the endogenous Nur77 promoter but not nucleolin (control) in granule neurons as determined by chromatin immunoprecipitation analysis. (B) Depolarization induces Nur77 gene expression in neurons in a VSCC- and calcineurin-dependent manner. RNA of granule neurons treated for 1 hour in the presence or absence of 25 mM KCl and vehicle (DMSO), Nim, or CsA was subjected to reverse transcription polymerase chain reaction (RT-PCR) by using primers specific to Nur77 or GAPDH. (C) Depolarization of granule neurons significantly induced expression of a luciferase reporter gene controlled by a Nur77 promoter containing WT MRE but not of a reporter controlled by a Nur77 promoter containing mutant MRE (MREmut) (P < 0.0001, ANOVA followed by Bonferroni-Dunn post hoc test; n = 6). Treatment with CsA prevented depolarization-induced expression of the WT Nur77–luciferase reporter gene (P < 0.0001, ANOVA followed by Bonferroni-Dunn post hoc test; n = 6), but had no effect on MREmut Nur77–luciferase reporter gene. (D) Nur77–luciferase reporter gene activity was significantly induced in depolarized neurons transfected with the control vector or MEF2A (P < 0.005, ANOVA followed by Bonferroni-Dunn post hoc test; n = 4). Both MEF2A-SUMO and MEF2-EN repressed depolarization-induced Nur77–luciferase reporter activity. (E) Cerebellar slices were transfected with control vector, wild-type Nur77 (Nur77-WT), or dominant-negative Nur77 (Nur77-DN). The number of claws was significantly increased in neurons expressing Nur77-DN compared with control-transfected neurons or neurons expressing Nur77-WT (P < 0.001, ANOVA followed by Bonferroni-Dunn post hoc test; total neurons measured, n = 114).

We have discovered a transcriptional mechanism that may orchestrate postsynaptic dendritic development in the mammalian brain. Our findings indicate that the transcription factor MEF2A plays a key role in the morphogenesis of granule neuron dendritic claws in the cerebellar cortex. The modifications of MEF2A required for postsynaptic differentiation occur within a phosphorylation-regulated sumoylation-acetylation switch (SAS) peptide motif that is conserved in all major MEF2 isoforms except MEF2B, as well as in several other transcription factor families (table S1). Thus, a phosphorylation-dependent switch between sumoylation and acetylation in transcription factors may play a widespread role in signal-regulated transcription and regulate diverse biological processes, including synapse development and plasticity.

Supporting Online Material

www.sciencemag.org/cgi/content/full/311/5763/1012/DC1

Materials and Methods

Figs. S1 to S13

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article