Inactivation of the Myocyte Enhancer Factor-2 Repressor Histone Deacetylase-5 by Endogenous Ca2//Calmodulin-dependent Kinase II Promotes Depolarization-mediated Cerebellar Granule Neuron Survival*

Cerebellar granule neuron (CGN) survival depends on activity of the myocyte enhancer factor-2 (MEF2) transcription factors. Neuronal MEF2 activity is regulated by depolarization via a mechanism that is presently unclear. Here, we show that depolarization-mediated MEF2 activity and CGN survival are compromised by overexpression of the MEF2 repressor histone deacetylase-5 (HDAC5). Furthermore, removal of depolarization induced rapid cytoplasm-to-nuclear translocation of endogenous HDAC5. This effect was mimicked by addition of the calcium/calmodulin-dependent kinase (CaMK) inhibitor KN93 to depolarizing medium. Removal of depolarization or KN93 addition resulted in dephosphorylation of HDAC5 and its co-precipitation with MEF2D. HDAC5 nuclear translocation triggered by KN93 induced a marked loss of MEF2 activity and subsequent apoptosis. To selectively decrease CaMKII, CGNs were incubated with an antisense oligonucleotide to CaMKIIα. This antisense decreased CaMKIIα expression and induced nuclear shuttling of HDAC5 in CGNs maintained in depolarizing medium. Selectivity of the CaMKIIα antisense was demonstrated by its lack of effect on CaMKIV-mediated CREB phosphorylation. Finally, antisense to CaMKIIα induced caspase-3 activation and apoptosis, whereas a missense control oligonucleotide had no effect on CGN survival. These results indicate that depolarization-mediated calcium influx acts through CaMKII to inhibit HDAC5, thereby sustaining high MEF2 activity in CGNs maintained under depolarizing conditions.

In early postnatal cerebellum, transcripts for two mammalian MEF2 isoforms, MEF2A and MEF2D, are markedly increased in parallel with enhanced expression of the GABA A receptor ␣6 subunit, a marker for differentiation of mature cerebellar granule neurons (CGNs) (9). Similarly, primary cultures of CGNs, that require medium containing depolarizing extracellular potassium for their survival (13), demonstrate high levels of endogenous MEF2A and MEF2D proteins and MEF2 transcriptional activity. We have previously reported that MEF2A and MEF2D are phosphorylated and subsequently cleaved by caspases in CGNs that are deprived of depolarizing potassium (14). The resulting loss of MEF2 transcriptional activity contributes to CGN apoptosis under these conditions. Likewise, Gaudilliere et al. (15) recently showed that down-regulation of MEF2A using RNA interference significantly compromises depolarization-mediated survival of CGNs. Thus, MEF2 proteins transduce activity-dependent calcium signals into an essential pro-survival pathway in CGNs.
The precise mechanism(s) by which calcium influx regulates MEF2 activity in CGNs is presently unclear. Previous work in muscle indicates that activity of the calcium/calmodulin-dependent, serine/threonine phosphatase, calcineurin, is required for MEF2 function (16). Similarly, calcineurin positively regulates MEF2 DNA binding and transcriptional activity in CGNs (17). In muscle, calcineurin modulation of MEF2 activity is integrated with regulation by a pathway involving calcium/ calmodulin-dependent protein kinase (CaMK) activity (5). Muscle MEF2 activity is repressed by interaction with class II histone deacetylases (HDACs) (18 -20). Overexpression of constitutively-active CaMK isoforms I and IV in muscle and nonmuscle cells (e.g. COS or NIH3T3 fibroblasts) promotes serine phosphorylation of HDACs. Phosphorylation of HDACs induces their nuclear export and promotes their association with cytosolic scaffolding proteins of the 14-3-3 family, ultimately resulting in derepression of MEF2 activity (21). Although these overexpression studies suggest a role for CaMK regulation of HDAC function in muscle, the endogenous CaMK isoform involved has yet to be identified. Moreover, a role for endogenous CaMK activity in modulating MEF2 function in neurons has not yet been investigated.
Here, we show that depolarization-mediated MEF2 activity and CGN survival are compromised by overexpression of the class II HDAC, HDAC5. Moreover, removal of depolarizing potassium or addition of a CaMK inhibitor induces nuclear translocation of endogenous HDAC5, loss of MEF2 activity, and CGN apoptosis. Finally, antisense-mediated down-regulation of CaMKII␣ expression is sufficient to drive nuclear translocation of HDAC5 and induce CGN apoptosis under depolarizing conditions. These results demonstrate that CaMK-mediated inhibition of HDAC function plays a significant role in the calcium regulation of MEF2 activity and survival in CGNs. In addition, these data are the first to identify an endogenous CaMK isoform (CaMKII) that regulates HDAC function.

EXPERIMENTAL PROCEDURES
Materials-A truncated mutant of MEF2A lacking the C-terminal transcriptional activation domain, pcDNA3.1-MEF2A131, was kindly provided by Dr. Ron Prywes (Columbia University, New York). Empty pcDNA3.1 vector was obtained from Invitrogen (Grand Island, NY). Adenoviral green fluorescent protein (Ad-GFP) was a gift from Dr. Jerry Schaack (University of Colorado Health Sciences Center, Denver, CO). The pGL2-MEF2-luciferase reporter plasmid and HA-tagged HDAC5 and HDAC4 were provided by Dr. Saadi Khochbin (INSERM, France). Hoecsht dye number 33258, 4Ј,6-diamidino-2-phenylindole (DAPI), monoclonal antibody to the FLAG epitope, and monoclonal antibody to ␤-tubulin were from Sigma. Monoclonal antibody to the HA epitope and rabbit polyclonal antibodies to phospho-CREB (Ser-133), HDAC4, and HDAC5 for immunocytochemistry were obtained from Cell Signaling Technologies (Beverly, MA). Rabbit polyclonal antibody to active (cleaved) caspase-3 was from Promega (Madison, WI). Monoclonal antibody to MEF2D for immunocytochemistry and Western blotting was purchased from Transduction Laboratories (Lexington, KY). Monoclonal antibody to CaMKII␣ for immunocytochemistry was from Chemicon (Temecula, CA). According to the manufacturer, this antibody may show some cross-reactivity with CaMKII␦ or ␥ subunits. FITC-and Cy3-conjugated secondary antibodies for immunofluorescence were from Jackson Immunoresearch Laboratories (West Grove, PA). Rabbit polyclonal antibody to MEF2 for immunoprecipitation, rabbit polyclonal antibody to HDAC5 for Western blotting, and protein A/G-agarose were obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Horseradish peroxidase-linked secondary antibodies and reagents for enhanced chemiluminescence detection were purchased from Amersham Biosciences (Piscataway, NJ). KN93, KN62, phosphorothioate 18-mer antisense oligonucleotide complementary to CaMKII␣ subunit mRNA nucleotides 33-50 (unlabeled and FITC-labeled), and an unlabeled, scrambled, missense control oligonucleotide were from Calbiochem (La Jolla, CA). According to the manufacturer, this antisense construct is highly specific for the ␣ subunit of CaMKII and does not affect CaMKII␤ expression.
Cell Culture-Rat CGNs were isolated from 7-day-old Sprague-Dawley rat pups (15-19 g) as described previously (22). Briefly, neurons were plated on 35-mm diameter plastic dishes coated with poly-L-lysine at a density of 2.0 ϫ 10 6 cells/ml in basal modified Eagle's medium containing 10% fetal bovine serum, 25 mM KCl, 2 mM L-glutamine, and penicillin (100 units/ml)/streptomycin (100 g/ml) (Invitrogen). Cytosine arabinoside (10 M) was added to the culture medium 24 h after plating to limit the growth of non-neuronal cells. Using this protocol, the cultures were ϳ95% pure for granule neurons. In general, experiments were performed after 7 days in culture.
Preparation of Adenoviral Dominant-negative MEF2 and Adenoviral Infection-pcDNA3.1-MEF2A131 was tagged on the C terminus with the FLAG epitope by polymerase chain reaction. The resulting FLAGtagged construct was cloned into shuttle vector, and Ad-Flag-MEF2A131 was prepared using the Ad-Easy adenovirus expression kit according to the manufacturer's instructions (Quantum Biotechnologies Inc.). Recombinant adenoviruses were purified by cesium chloride gradient ultracentrifugation. The viral titer was determined by measuring the absorbance at 260 nm (where 1.0 absorbance units ϭ 1 ϫ 10 12 particles/ml) and infectious particles were verified by plaque assay. Ad-GFP or Ad-FLAG-MEF2A131 was added to CGN cultures on day 4 at a multiplicity of infection (m.o.i.) of 100. On day 7 (72-h postinfec-tion), CGN apoptosis was assessed in GFP-positive or FLAG-immunoreactive cells by DAPI staining.
Quantification of Apoptosis-Apoptosis was assessed by fixing CGNs in 4% paraformaldehyde and staining nuclei with either Hoechst dye (non-permeabilized cells) or DAPI (permeabilized cells for immunocytochemistry). Cells were considered apoptotic if their nuclei were either condensed or fragmented. In general, ϳ500 cells from at least two fields of a 35-mm well were counted. For immunocytochemical studies where cells were plated on glass coverslips, ϳ100 -200 cells from 2-3 coverslips per treatment group were counted. Data are presented as the percentage of cells in a given treatment group which were scored as apoptotic. Experiments were performed on cells isolated from at least three independent preparations.
Immunocytochemistry-CGNs were cultured on polyethyleneiminecoated glass cover slips at a density of ϳ2.5 ϫ 10 5 cells per coverslip. Following incubation as described under "Results," cells were fixed in 4% paraformaldehyde and were then permeabilized and blocked in PBS (pH 7.4) containing 0.2% Triton X-100 and 5% BSA. Cells were then incubated for ϳ16 h at 4°C with primary antibody diluted in PBS containing 0.2% Triton X-100 and 2% BSA. The primary antibody was aspirated and the cells were washed five times with PBS. The cells were then incubated with either Cy3-conjugated or FITC-conjugated secondary antibodies and DAPI for 1 h at room temperature. CGNs were then washed five more times with PBS and coverslips were adhered to glass slides in mounting medium (0.1% p-phenylenediamine in 75% glycerol in PBS). Fluorescent images were captured using a 63ϫ oil immersion objective on a Zeiss Axioplan 2 microscope equipped with a Cooke Sensicam deep-cooled CCD camera and a Slidebook software analysis program for digital deconvolution (Intelligent Imaging Innovations Inc., Denver, CO).
Preparation of CGN Cell Extracts-After incubation for the indicated times and with the reagents specified in Results, the culture medium was aspirated; the cells were washed once with 2 ml of ice-cold PBS, pH 7.4, placed on ice, and scraped into lysis buffer (200 l/35-mm well) containing 20 mM HEPES (pH 7.4), 1% Triton X-100, 50 mM NaCl, 1 mM EGTA, 5 mM ␤-glycerophosphate, 30 mM sodium pyrophosphate, 100 M sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin. Cell debris was removed by centrifugation at 6,000 ϫ g for 3 min and the protein concentration of the supernatant was determined using a commercially available protein assay kit (Pierce). Aliquots (ϳ150 g) of supernatant protein were diluted to a final concentration of 1ϫ SDS-PAGE sample buffer, boiled for 5 min, and electrophoresed through 7.5% polyacrylamide gels. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Millipore, Bedford, MA) and processed for immunoblot analysis.
Western Blot Analysis-Nonspecific binding sites were blocked in PBS (pH 7.4) containing 0.1% Tween 20 (PBS-T) and 1% BSA for 1 h at room temperature. Primary antibodies were diluted in blocking solution and incubated with the membranes for 1 h. Excess primary antibody was removed by washing the membranes three times in PBS-T. The blots were then incubated with the appropriate horseradish peroxidaseconjugated secondary antibody diluted in PBS-T for 1 h and were subsequently washed three times in PBS-T. Immunoreactive proteins were detected by enhanced chemiluminescence. Autoluminograms shown are representative of at least three independent experiments.
Immunoprecipitation of MEF2D-CGN lysates were prepared as described above, except in lysis buffer containing 0.1% Triton X-100. 4 g of polyclonal antibody against MEF2D was added to 500 l of lysate (ϳ1 g/l CGN protein concentration), and samples were mixed for 16 h at 4°C by continuous inversion. Agarose-conjugated protein A/G (40 l) was added and samples were mixed for a further 4 h. Immune complexes were pelleted, washed three times with lysis buffer containing 0.1% Triton X-100, and boiled in 1ϫ SDS-PAGE sample buffer. Immune complexes were resolved on 7.5% polyacrylamide gels, transferred to polyvinylidene difluoride, and immunoblotted for MEF2D and HDAC5.
CGN Transfection and Reporter Gene Expression-CGNs were transiently transfected using a calcium phosphate co-precipitation method described previously (14). Cells were co-transfected with 1 g of MEF2luciferase expression plasmid (pGL2-MEF2-luc), 1 g of pCMV-␤-gal as an internal control for transfection efficiency, and 2 g of either pcDNA3.1, MEF2A131, or HA-tagged HDAC5. The total amount of DNA for each transfection was kept constant. Following transfection, neurons were placed in conditioned medium and the medium was subsequently changed to serum-free containing either 25 mM KCl or 5 mM KCl. After incubation, cell extracts were prepared using reporter lysis buffer and the activities of luciferase and ␤-galactosidase were measured with the respective enzyme assay system kits (Promega).
Incubation with Antisense Oligonucleotides to CaMKII␣-Phospho-rothioate, FITC-labeled or unlabeled, antisense oligonucleotides to CaMKII␣, or a scrambled missense control oligonucleotide, were added directly to the culture medium (each at a final concentration of 1 M), as was previously described for primary hippocampal neuronal cultures (23). Following incubation for 8 -24 h, cells were processed for immunocytochemical analysis and/or quantification of apoptosis, as described under "Results." Data Analysis-The results shown represent the means Ϯ S.E. for the number (n) of independent experiments that were performed. Statistical differences between the means of unpaired sets of data were evaluated by one-way analysis of variance, followed by a post-hoc Dunnett's test. A p value of Ͻ0.01 was considered statistically significant.

Adenoviral Expression of a Dominant-negative MEF2 Mutant Blocks Depolarization-mediated Survival of CGNs-In
CGNs co-transfected with a MEF2-luciferase reporter (24) and empty pcDNA3.1 vector, removal of the depolarization stimulus resulted in a dramatic loss of MEF2 activity (14 Ϯ 2% of control, p Ͻ 0.01, Fig. 1A). This effect was mimicked under depolarizing conditions by co-transfection with a truncated, dominant-negative form of MEF2 (MEF2A131) that lacks the transcriptional activation domain. Expression of MEF2A131 decreased endogenous MEF2 activity in the presence of depolarization to 39 Ϯ 6% (p Ͻ 0.01) of that observed in cells transfected with empty vector (Fig. 1A). Thus, dominant-negative MEF2 competes with endogenous MEF2 transcription factors for DNA binding, resulting in a loss of endogenous MEF2 activity.
Previous studies showing that MEF2 transcription factors are critical for CGN survival utilized either transient transfection protocols of MEF2 mutants (12,14) or, more recently, RNA interference methodology to knock down expression of MEF2 proteins (15). One limitation of each of these approaches is that only a small percentage (usually Ͻ5%) of the CGNs are transfected. As a means of interfering with MEF2 activity in a larger, and possibly more representative, percentage of the CGN culture, we prepared a FLAG epitope-tagged MEF2A131 and expressed this construct in adenovirus. CGNs maintained in depolarizing medium were then infected with either adenovirus expressing green fluorescent protein (Ad-GFP, to serve as a control for adenoviral infection) or expressing the FLAGtagged dominant-negative MEF2 (Ad-FLAG-MEF2A131). Each adenovirus was infected into the cells at an m.o.i. of 100 (equivalent to 10,000 infectious particles per cell). At this titer, the relative infection efficiency observed over three independent experiments was ϳ35% for Ad-FLAG-MEF2A131. Following infection, cells were incubated for 72 h and MEF2A131-expressing cells were identified by immunocytochemistry with a monoclonal anti-FLAG and a Cy3-conjugated secondary antibody. GFP-positive cells were detected by fluorescence under a FITC filter. Apoptosis was quantified by staining nuclei with DAPI and counting the percentage of GFP-positive or FLAG-MEF2A131-positive cells containing condensed and/or fragmented chromatin. As shown in Fig. 1B, infection with Ad-GFP did not result in significant CGN apoptosis (2 Ϯ 1%), whereas cells expressing the FLAG-MEF2A131 demonstrated massive apoptosis (89 Ϯ 7%), even in the presence of depolarization. In agreement with previous work (12,14,15), our results using an adenoviral dominant-negative MEF2 mutant demonstrate that MEF2 transcriptional activity is a critical component of depolarization-mediated survival signaling in CGNs.
Overexpression of the MEF2 Repressor HDAC5 Induces CGN Apoptosis in Depolarizing Medium-The activity of MEF2 transcription factors is repressed in skeletal, cardiac, and smooth muscle by members of the class II family of HDACs including HDAC4, 5, 7, and 9 (reviewed in Ref. 6). Consistent with a potential role for HDAC regulation of MEF2 activity in CGNs, overexpression of HA-tagged HDAC5 significantly de-creased MEF2-driven luciferase activity (56 Ϯ 6% of the empty vector control, p Ͻ 0.01) in cells cultured in the presence of depolarizing potassium ( Fig. 2A). As a result of the loss in endogenous MEF2 activity, overexpression of HDAC5 induced significant CGN apoptosis (42 Ϯ 1% compared with 17 Ϯ 1% in cells transfected with empty vector, p Ͻ 0.01, Fig. 2B). Similar results were obtained following transfection with another class II family member, HDAC4 (data not shown). Thus, depolarization-mediated MEF2 activity and CGN survival are compromised by overexpression of HDAC transcriptional repressors.
Removal of Depolarizing Potassium or CaMK Inhibition Induces Cytoplasm-to-Nuclear Translocation of Endogenous HDAC5-We next examined the subcellular localization of endogenous HDAC5 and MEF2D in CGNs maintained in either depolarizing or non-depolarizing medium. Cells cultured in the presence of depolarizing extracellular potassium demonstrated immunocytochemical localization of HDAC5 in the cytoplasm with little to no immunoreactivity observed in nuclei. In contrast, MEF2D was localized exclusively to nuclei, and therefore, there was essentially no overlapping staining for MEF2D and HDAC5 in control CGNs (Fig. 3, left panels). Removal of depolarizing potassium for 4 h induced a dramatic cytoplasmto-nuclear translocation of endogenous HDAC5, resulting in significant overlap of MEF2D and HDAC5 staining (Fig. 3, middle panels).
Previous work in skeletal and cardiac muscle has shown that the subcellular localization of class II HDACs during muscle differentiation can be regulated by CaMK activity (19,20). Overexpression of either constitutively active CaMKI or CaMKIV induced phosphorylation of two serine residues (Ser-259 and Ser-498 in HDAC5) resulting in docking of HDACs to cytoplasmic scaffolding proteins of the 14-3-3 family (21). In this manner, CaMK-mediated phosphorylation of HDACs results in their nuclear exclusion, allowing for derepression of MEF2 activity that is required for muscle differentiation. To were incubated for 4 h in serum-free medium containing either 25 mM KCl (25K), 5 mM KCl (5K), or 25KϩKN93 (5 M). Following incubation, cells were fixed in 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked in 5% BSA. HDAC5 and MEF2D were immunocytochemically localized by incubating the cells with a polyclonal antibody to HDAC5 and a monoclonal antibody to MEF2D followed by Cy3-conjugated and FITC-conjugated secondary antibodies, respectively. Nuclei were identified by DAPI staining. Digitally deconvolved images were captured using a 63ϫ oil objective. The images shown are representative of results obtained in three independent experiments. The bottom row shows the merged images of the DAPI, FITC, and Cy3 channels. Note that MEF2D was exclusively nuclear under all conditions, whereas the localization of HDAC5 changed from cytoplasmic in 25K to nuclear in the 5K and 25K ϩ KN93 conditions. The nuclear colocalization of HDAC5 with MEF2D and DAPI under the 5K and 25KϩKN93 conditions is evident by the resultant white overlapping staining in the merged images. Scale bar, 10 m. examine if depolarization maintains HDAC5 in the cytoplasm of CGNs via endogenous CaMK activity, cells were incubated in depolarizing medium containing the CaMK inhibitor KN93. Inclusion of KN93 mimicked the effect of removing the depolarization stimulus, resulting in nuclear translocation of HDAC5 and its co-localization with MEF2D (Fig. 3, right panels). In a similar manner, either removal of the depolarization stimulus or addition of KN93 also induced the nuclear translocation of HDAC4 in CGNs (data not shown). These data suggest that depolarization promotes the nuclear exclusion of HDAC5 and 4 via the activity of an endogenous CaMK.
CaMK Inhibition Results in Dephosphorylation of HDAC5 and Its Association with MEF2D-To evaluate the phosphorylation status of endogenous HDAC5 in CGNs, we examined the electrophoretic mobility of HDAC5 on SDS-polyacrylamide gels. In cell lysates obtained from CGNs maintained in depolarizing medium, HDAC5 appeared as a broad band or doublet on Western blots (Fig. 4A, first lane). Removal of depolarizing potassium (Fig. 4A, second lane) or addition of the CaMK inhibitor, KN93 (Fig. 4A, third lane), increased the electrophoretic mobility of HDAC5 and resulted in the disappearance of the upper band of the doublet, indicative of a loss of the phosphorylated HDAC5 species. These results suggest that a CaMK activated by depolarization in CGNs regulates the phosphorylation status of HDAC5.
To determine if the dephosphorylation and nuclear translocation of HDAC5 induced by removal of the depolarization stimulus or CaMK inhibition permits its direct interaction with nuclear MEF2 proteins, we performed co-immunoprecipitation experiments. MEF2D immune complexes isolated from CGNs maintained in depolarizing medium did not contain any detectable HDAC5 (Fig. 4B, first lane). This finding is consistent with the immunocytochemical data shown in Fig. 3 that demonstrated little co-localization of MEF2D and HDAC5 under control conditions. In contrast, MEF2D isolated from cells incubated in either non-depolarizing medium or in the presence of KN93 co-precipitated with HDAC5 (Fig. 4B, second and third  lanes). Note that the electrophoretic mobility of MEF2D was significantly retarded in cells cultured in non-depolarizing medium, consistent with its hyperphosphorylation under these conditions as we have previously reported (14,25). Interestingly, CaMK inhibition also resulted in some decreased mobility of MEF2D, suggesting that HDAC5 association with MEF2D may alter the accessibility of MEF2D to phosphatases (e.g. calcineurin). Collectively, these data suggest that inhibition of CaMK activity in CGNs promotes the dephosphorylation and nuclear translocation of HDAC5, ultimately resulting in its co-association with MEF2 transcription factors.
CaMK Inhibition Blocks Depolarization-mediated MEF2 Activity and Survival of CGNs-To determine the functional consequences of HDAC5 nuclear translocation and its interaction with MEF2D, we next examined the effects of CaMK inhibition on MEF2 transcriptional activity and CGN survival. Incubation of CGNs with the CaMK inhibitors, KN93 or KN62, induced a marked decrease in endogenous MEF2 activity measured with a MEF2-driven luciferase reporter gene (Fig. 5A). Moreover, CaMK inhibition induced significant CGN apoptosis in the presence of depolarizing potassium (Fig. 5B). Apoptosis induced by KN93 was characterized morphologically by substantial chromatin fragmentation, microtubule disruption, and caspase-3 activation (Fig. 5C). These results demonstrate that inhibition of endogenous CaMK activity is sufficient to induce nuclear translocation of the MEF2 repressor, HDAC5, resulting in a loss of MEF2 activity and induction of CGN apoptosis.
Antisense to CaMKII␣ Induces Nuclear Translocation of HDAC5-Previous work in skeletal muscle describing the reg-ulation of HDAC localization by CaMK activity has relied on overexpression of various CaMK isoforms (19). To date, identification of the endogenous CaMK isozyme(s) involved in HDAC regulation has not been reported. The CaMK enzyme family is made up of several distinct isoforms, including CaMK I, II, and IV, with four distinct genes (␣, ␤, ␦, ␥) coding for members of the CaMKII subfamily (26). Of the many CaMK isozymes identified, only CaMKII␣ is exclusively expressed in the brain (27). The CaMK inhibitors, KN93 and KN62, lack isozyme specificity and have been shown to inhibit the activities of CaMKI, II and IV (28,29). To selectively evaluate a role for CaMKII in regulating the subcellular localization of HDAC5 in CGNs, cells were incubated with a CaMKII␣-specific antisense oligonucleotide. This antisense construct has previously been reported to reduce CaMKII␣ expression and induce epileptiform . Following incubation, detergent-soluble cell lysates were subjected to SDS-PAGE on 5% polyacrylamide gels and the proteins were subsequently transferred to polyvinylidene difluoride membranes. HDAC5 was detected by immunoblotting (IB) with a polyclonal antibody and a horseradish-peroxidase-conjugated secondary. Immunoreactive proteins were detected by enhanced chemiluminescence. The blot shown is representative of results obtained in three separate experiments. Note that HDAC5 runs as a broad apparent doublet in 25K, whereas under the 5K or 25KϩKN93 conditions the upper band of the doublet disappears, indicative of an enhanced mobility of HDAC5 under these conditions. B, CGNs were incubated exactly as described in A and cell lysates were prepared. MEF2D was immunoprecipitated (IP) from the cell lysates using a polyclonal antibody and immune complexes were resolved on 7.5% gels. The membranes were cut at the 75 kDa standard and the upper portion of the blot was probed for HDAC5 while the lower portion was probed for MEF2D. The blots shown are indicative of data observed in two independent experiments. Note that HDAC5 was only detectable in MEF2D immune complexes obtained from CGNs incubated in either 5K or 25KϩKN93 conditions. HC, immunoglobulin heavy chain from the precipitating antibody. activity in primary cultured hippocampal neurons (23).
Addition of 1 M FITC-labeled, phosphorothioate antisense oligonucleotide to CaMKII␣ to CGN cultures for 8 h resulted in a marked reduction in CaMKII␣ immunoreactivity in cells expressing the FITC-antisense when compared with cells in the same field that did not take up the antisense construct (Fig.  6A). In general, ϳ25% of the CGNs took up the FITC-labeled CaMKII␣ antisense following 8 -24 h of incubation at a concentration of 1 M. Granule neurons expressing the FITC-CaMKII␣ antisense demonstrated a dramatic redistribution of endogenous HDAC5 from the cytosol to the nucleus, even when maintained in depolarizing medium (Fig. 6B). In contrast, cells in the same field that did not express the antisense showed HDAC5 localization that was extensively cytoplasmic. These results indicate that the endogenous CaMKII isozyme regulates depolarization-mediated cytoplasmic localization of HDAC5 in CGNs.
Antisense to CaMKII␣ Does Not Affect CaMKIV-dependent Phosphorylation of CREB-To confirm the selectivity of the CaMKII␣ antisense, we examined its effects on phosphorylation of the CaMKIV substrate, cAMP-response element binding protein (CREB). A recent study demonstrated that CaMKIV activity promotes depolarization-mediated CGN survival via phosphorylation of CREB on the activating Ser-133 (30). As shown in Fig. 7A (upper panels), CREB phosphorylation on Ser-133 was high in CGNs maintained in depolarizing medium. Addition of the CaMK inhibitor, KN93, resulted in a complete loss of CREB phosphorylation (Fig. 7A, lower panels). This latter result is consistent with the relative non-selectivity of KN93 for the various isoforms of CaMK (28,29). In contrast, the FITC-labeled antisense to CaMKII␣ had no significant effect on CREB phosphorylation after 8 h of incubation (Fig.  7B, compare the CREB phosphorylation in FITC-positive cells to that in antisense-negative cells), a time point at which it caused significant nuclear translocation of HDAC5 (see Fig.  6B). These results demonstrate the selectivity of the CaMKII␣ antisense. In addition, they suggest that CaMKII regulation of HDAC localization is an alternative calcium-dependent pathway to CaMKIV/CREB for mediating CGN survival.
Antisense to CaMKII␣ Induces Caspase-3 Activation and Apoptosis of CGNs-To determine the extent of CGN apoptosis induced by antisense treatment, we quantified the number of cells with condensed and/or fragmented nuclei in cultures incubated with either the CaMKII␣ antisense or a missense control oligonucleotide. Because the missense oligonucleotide was not fluorescently labeled, an unlabeled antisense was utilized in these experiments. The overall basal apoptosis of control CGN cultures maintained in depolarizing medium measured 9 Ϯ 1% (n ϭ 3 experiments, performed in triplicate) on day 8 in vitro. On day 7 in vitro, cells were exposed to either the CaMKII␣ antisense or a control missense oligonucleotide for 24 h, each at a final concentration of 1 M. CGN cultures incubated with the antisense to CaMKII␣ demonstrated an overall apoptosis of 33 Ϯ 11% (n ϭ 3), whereas cells treated with the missense construct showed apoptosis comparable to (cleaved) caspase-3 [Casp-3 (a)] were immunostained by incubating the cells with a monoclonal antibody to tubulin and a polyclonal antibody that specifically recognizes the active fragment of caspase-3, followed by FITC-conjugated and Cy3-conjugated secondary antibodies, respectively. Nuclei were identified by DAPI staining. Digitally deconvolved images were captured using a 63ϫ oil objective. The images shown are representative of results obtained in three independent experiments. The bottom panels show the merged images of the DAPI, FITC, and Cy3 channels. Arrows point to cells in the KN93 treated culture that demonstrate significant chromatin fragmentation and immunoreactivity for active caspase-3. Also, note the substantial loss of fine microtubule staining in the KN93-treated culture. Scale bar, 10 m.

FIG. 5. The CaMK inhibitors, KN93 and KN62, block depolarization-mediated MEF2 transcriptional activity and induce apoptosis of CGNs.
A, CGNs (day 7 in vitro) were transiently co-transfected with a MEF2-responsive luciferase reporter plasmid and pCMV-␤-gal. After transfection for 2 h, cells were placed in serum-free medium containing either 25 mM KCl (25K) Ϯ KN93 or KN62 (each at 5 M) or 5 mM KCl (5K). Luciferase and ␤-galactosidase activities were determined 4 h later and luciferase activity was normalized with respect to that of ␤-galactosidase. Data are expressed as a percentage of the activity in control neurons transfected with empty vector. Results represent the mean Ϯ S.E. for three independent experiments each performed in duplicate. *, significantly different from the 25K control (p Ͻ 0.01). B, CGNs (day 6 in vitro) were incubated in either 25K medium Ϯ KN93 or KN62 or 5K medium for 24 h. CGN apoptosis was quantified by Hoechst staining of nuclei. The values shown represent the mean Ϯ S.E. percent apoptosis for three experiments each performed in triplicate. *, significantly different from the 25K control (p Ͻ 0.01). C, CGNs incubated for 24 h in 25K medium alone (left panels) or containing KN93 (right panels) were fixed in 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked in 5% BSA. ␤-tubulin and active the untreated controls (8 Ϯ 1%, n ϭ 3). These results demonstrate that the scrambled missense had no discernible effect on CGN survival when added at the same concentration as the antisense construct.
As mentioned previously, in a given experiment ϳ25% of the entire culture took up the FITC-labeled antisense. Therefore, the overall apoptosis observed in CGN cultures incubated with the unlabeled antisense was consistent with near complete apoptosis of the antisense-positive cells. To verify this result, we quantitated the fraction of FITC-positive cells that were apoptotic following a 24-h incubation with 1 M FITC-labeled CaMKII␣ antisense. Greater than 95% of the FITC-positive (antisense-positive) CGNs were apoptotic under these conditions, as assessed by nuclear morphology. In agreement with these findings, essentially every cell that accumulated the FITC-labeled antisense to CaMKII␣ for 24 h demonstrated substantial immunoreactivity for active (cleaved) caspase-3, whereas surrounding antisense-negative cells in the same field were devoid of active caspase-3 staining (Fig. 8). Thus, antisense-mediated depletion of CaMKII␣ inhibits depolarizationdependent survival signaling and induces CGN apoptosis.

DISCUSSION
MEF2 proteins are members of the MADS (MCM1-agamousdeficiens-serum response factor) family of transcription factors (31)(32)(33). The mammalian MEF2 family consists of four distinct genes (A, B, C, D) that each code for multiple splice variants (31,34). MEF2 proteins play a key role in the differentiation and survival of neurons during CNS development. The expression of MEF2 proteins in the developing CNS is regulated temporally and spatially in an isoform-specific manner, and coincides with neuronal maturation. For example, cerebral cortical neuron development is associated with changes in the expression of MEF2C (35)(36)(37), whereas CGN maturation is coupled to enhanced expression of MEF2A and MEF2D (9). Their differential tissue distribution suggests that MEF2 proteins may be regulated in an isoform-specific manner. Indeed this is the case in CGNs where MEF2A and MEF2D are phosphorylated and ultimately cleaved by caspases following removal of depolarizing potassium, whereas MEF2B and MEF2C are not modified in this manner (14). These findings indicate that signal transduction pathways which regulate MEF2 activity display both tissue-specific and isoform-specific properties.
To date, the modulation of MEF2 activity by calcium-regulated signals has largely been studied in muscle cells in which MEF2 proteins act as key regulators of myogenic differentiation (5,6). Given that distinct MEF2 isoforms can be regulated differently, even within the same cell (14), it is unknown if signaling pathways that modulate MEF2 proteins in muscle will necessarily translate directly into neurons. Moreover, MEF2 regulation in muscle has been investigated during either myogenic differentiation (38 -40) or hypertrophy (41)(42)(43). In contrast, a principal function of MEF2 proteins in neurons is to promote activity-dependent cell survival (12,15). The signaling pathways that are invoked during muscle differentiation or hypertrophy may be different from those generated in neurons by activity-dependent calcium influx.
Despite these important differences, at least two signaling pathways have been implicated in MEF2 regulation in both muscle and neurons. First, the stress-activated protein kinase, p38 MAP kinase, stimulates MEF2 activity in hypertrophic muscle (41,44) and in CGNs responding to depolarization (12). Second, the phosphatase, calcineurin, positively regulates MEF2 activity in both muscle cells (16,45,46) and CGNs (17). In muscle, p38 MAP kinase and calcineurin cooperate with CaMK to maximally activate MEF2 transcriptional activity (5,47).
CaMK-mediated regulation of MEF2 activity in muscle has been proposed to occur primarily by an indirect mechanism. Class II HDAC proteins act as endogenous repressors of MEF2 transcriptional activity by directly interacting with MEF2 proteins in the nucleus (40). Overexpression of active CaMK I or IV derepresses MEF2 activity in muscle by phosphorylating HDACs and promoting their nuclear export (18,19). CaMKmediated phosphorylation of HDACs on specific serine residues promotes HDAC binding to cytosolic scaffolding proteins of the 14-3-3 family, resulting in sequestration of HDACs in the cytoplasm (21,48). In the present study, we demonstrate that endogenous CaMKII plays a key role in CGNs to exclude HDACs from the nucleus, and in turn, to derepress MEF2 activity and promote cell survival.
We initially confirmed an essential role for MEF2 activity in FIG. 6. An antisense oligonucleotide to CaMKII␣ decreases CaMKII␣ expression and induces nuclear translocation of HDAC5. A, CGNs (day 7 in vitro) were incubated in serum-free medium containing 25 mM KCl (25K) and a phosphorothioate, FITC-labeled antisense oligonucleotide to CaMKII␣ (CaMKII␣-FITC-as, 1 M final concentration). After incubation for 8 h, cells were fixed in 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked in 5% BSA. Expression of endogenous CaMKII␣ was then assessed by immunostaining with an isozyme-specific antibody and a Cy3-conjugated secondary. Nuclei were stained with DAPI and CGNs expressing the antisense construct were identified by positive staining in the FITC channel (indicated by the arrows). Note that cells positive for the antisense exhibited substantially less immunoreactivity for CaMKII␣ than cells in the same field that did not take up the oligonucleotide. In general, ϳ25% of the entire CGN culture took up the FITC-labeled CaMKII␣ antisense. The images shown are indicative of data obtained in two separate experiments. Scale bar, 10 m. B, CGNs incubated exactly as described in A were immunostained for HDAC5 using a polyclonal antibody and a Cy3-conjugated secondary. The arrows point to cells that were positive for the CaMKII␣-FITC-as. Note that CGNs expressing the antisense demonstrated nuclear localization of HDAC5, whereas antisense-negative cells in the same field showed cytoplasmic localization of HDAC5. The lower right panel shows the merged image of the DAPI, FITC, and Cy3 channels; note the white overlapping nuclear staining in cells expressing the antisense and showing nuclear HDAC5. The results shown are representative of three independent experiments. Scale bar, 10 m.
CGN survival by demonstrating that adenoviral infection of a truncated, dominant-interfering mutant of MEF2 (MEF2A131) is sufficient to induce apoptosis of CGNs that are maintained in depolarizing medium. To examine a potential role for HDACs in the regulation of MEF2 activity in CGNs, the effects of overexpressing the class II HDAC, HDAC5, on MEF2 activity and CGN survival were assessed. Transient transfection of CGNs with HDAC5 induced a significant loss of endogenous MEF2 transcriptional activity and substantial apoptosis under depolarizing conditions.
To directly investigate the involvement of endogenous HDAC5 in regulating MEF2 activity, we next analyzed the subcellular localization of MEF2D and HDAC5 in CGNs. As expected, the transcription factor MEF2D was exclusively nuclear under all conditions tested. In contrast, endogenous HDAC5 was extensively cytoplasmic in CGNs cultured in depolarizing medium. Either removal of the depolarization stimulus or addition of the CaMK inhibitor, KN93, induced a marked translocation of HDAC5 into the nucleus where it co-localized with MEF2D. While this article was in preparation, Chawla et al. (49) described a partial inhibition of HDAC5 nuclear export in hippocampal neurons following incubation with a similar CaMK inhibitor, KN62. However, these authors did not examine the effects of CaMK inhibition on MEF2 activity or neuronal survival downstream of HDAC shuttling. In the current study performed in CGNs, the nuclear translocation of HDAC5 was associated with its enhanced mobility on polyacrylamide gels (consistent with its dephosphorylation) and its co-precipitation with MEF2D. Collectively, these data indicate that depolarization-mediated calcium influx stimulates the activity of an endogenous CaMK that, in turn, phosphorylates HDAC5 and inhibits its nuclear localization.
The ultimate consequence of removing the depolarization stimulus from cerebellar cultures is the apoptotic death of CGNs (13,50). Loss of MEF2 transcriptional activity precedes CGN death under non-depolarizing conditions and apoptosis can be attenuated by expression of a constitutively active FIG. 7. CREB phosphorylation in CGNs is not affected by a CaMKII␣ antisense oligonucleotide. A, CGNs (day 7 in vitro) were incubated for 4 h in serum-free medium containing 25 mM KCl (25K) in the absence or presence of KN93 (5 M). Cells were fixed in paraformaldehyde and CREB phosphorylated on Ser-133 (p-CREB) was identified by staining with a phospho-specific polyclonal antibody and a Cy3-conjugated secondary. Each of the p-CREB images was captured with a 63ϫ oil objective under a Cy3 filter using equal exposure times. Scale bar, 10 m. B, cells were incubated for 8 h in 25K medium containing a FITC-labeled antisense oligonucleotide to CaMKII␣ (CaMKII␣-FITC-as, 1 M final concentration). Following incubation, cells were immunostained for p-CREB. Arrows point to CGNs that were positive for the FITC-labeled antisense. Note that all of the cells demonstrated similar immunoreactivity for p-CREB regardless of whether or not they expressed the CaMKII␣ antisense. The images shown are indicative of three independent experiments. Scale bar, 10 m.

FIG. 8. Antisense to CaMKII␣ induces caspase-3 activation in CGNs maintained in depolarizing medium. CGNs (day 7 in vitro)
were incubated in serum-free medium containing 25 mM KCl (25K) and a phosphorothioate, FITC-labeled antisense oligonucleotide to CaMKII␣ (CaMKII␣-FITC-as, 1 M final concentration). After incubation for 24 h, cells were fixed in 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and blocked in 5% BSA. Activation of caspase-3 was assessed by immunostaining with a polyclonal antibody that specifically recognizes the active (cleaved) form of the caspase (Casp-3(a)), followed by a Cy3-conjugated secondary. Nuclei were stained with DAPI and CGNs expressing the antisense construct were identified by positive staining in the FITC channel (indicated by the arrows). Note that cells positive for the antisense exhibited high immunoreactivity for active caspase-3, whereas cells in the same field that did not take up the oligonucleotide showed no staining for active caspase. The images shown are representative of two independent experiments. Scale bar, 10 m.
MEF2 mutant (12,14). Previous work suggests a role for CaMK activity in depolarization-mediated survival of CGNs (51). In agreement with this, addition of the CaMK inhibitors, KN93 or KN62, to CGNs maintained in depolarizing medium mimicked the loss of MEF2 activity and induction of CGN apoptosis that normally occur under non-depolarizing conditions. The ability of CaMK inhibitors to induce the nuclear translocation of HDAC5 and the consequent loss of MEF2 activity suggests that an endogenous CaMK isozyme(s) actively derepresses MEF2dependent transcription of putative pro-survival genes in CGNs maintained in depolarizing medium.
The regulation of HDAC localization by CaMK in muscle and other non-neuronal cells has been analyzed primarily by overexpression of various CaMK isoforms, including constitutively active mutants of CaMKI and IV (19,48). Thus far, the endogenous CaMK isozyme(s) involved in HDAC regulation has not been identified. CaMKII␣ is a brain-specific CaMK isozyme best known for its role in hippocampal long term potentiation, a cellular model of learning and memory (52). However, CaMKII␣ is also expressed in cerebellum (53), where it has been implicated in CGN neurite outgrowth (51,54). Since the CaMK inhibitors, KN93 and KN62, lack isozyme selectivity (28,29), we utilized an antisense strategy to specifically interfere with CaMKII signaling in CGNs.
Recently, direct addition of a phosphorothioate antisense oligonucleotide to CaMKII␣ to primary hippocampal neurons was shown to substantially decrease CaMKII␣ expression in vitro (23). We found that addition of an identical FITC-labeled CaMKII␣ antisense oligonucleotide to CGN cultures maintained in depolarizing medium resulted in a marked loss of CaMKII␣ immunoreactivity and nuclear translocation of HDAC5. In contrast, this antisense construct had no discernible effect on phosphorylation of the CaMKIV substrate, CREB, demonstrating the specificity of the oligonucleotide for CaMKII. Finally, antisense to CaMKII␣ induced caspase-3 activation and apoptosis, whereas a scrambled missense control oligonucleotide had no effect on CGN survival. Collectively, these results indicate that CaMKII is an important endogenous regulator of HDAC5 localization and MEF2-mediated survival of CGNs.
Our data implicating CaMKII in neuronal survival support previous findings suggesting that reductions in CaMKII activity contribute to neuronal death induced by excitotoxicity or ischemia in vitro and in vivo (55)(56)(57)(58). Since loss of MEF2 activity has recently been implicated in excitotoxic and ischemic neuronal death (59), it will be interesting to determine if nuclear localization of HDAC5 (or other class II family members) is also detected in these conditions. One implication of this would be that HDAC repression may act in a coordinated manner with caspase-mediated degradation of MEF2 proteins to deplete MEF2 activity during neuronal apoptosis (14,59).
In summary, data in the present study establish a principal role for endogenous CaMKII in the modulation of HDAC5 localization in CGNs. By phosphorylating HDAC5, CaMKII excludes this transcriptional repressor from the nucleus and prevents it from interacting with MEF2D. In this manner, CaMKII works in a coordinated fashion with other signaling molecules, such as calcineurin, to maximize MEF2-dependent transcription of putative pro-survival genes. These results add further insight into the complexities of MEF2 regulation in neurons and suggest that multiple calcium-regulated pathways act in concert to modulate the function of these important transcription factors during CNS development.