Functional Characterization of an Amino-terminal Region of HDAC4 That Possesses MEF2 Binding and Transcriptional Repressive Activity*

Like the full-length histone deacetylase (HDAC) 4, its amino terminus (amino acids 1–208) without the carboxyl deacetylase domain is also known to effectively bind and repress myocyte enhancer factor 2 (MEF2). Within this repressive amino terminus, we further show that a stretch of 90 amino acids (119–208) displays MEF2 binding and repressive activity. The same region is also found to associate specifically with HDAC1 which is responsible for the repressive effect. The amino terminus of HDAC4 can associate with the DNA-bound MEF2 in vitro, suggesting that it does not repress MEF2 simply by disrupting the ability of MEF2 to bind DNA. In vivo, MEF2 induces nuclear translocation of both the full-length HDAC4 and HDAC4-(1–208), whereas the nuclear HDAC4 as well as HDAC4-(1–208) in turn specifically sequesters MEF2 to distinct nuclear bodies. In addition, we show that MyoD and HDAC4 functionally antagonize each other to regulate MEF2 activity. Combined with data from others, our data suggest that the full-length HDAC4 can repress MEF2 through multiple independent repressive domains.

Two families of transcription factors, namely the myogenic regulatory factors and the myocyte enhancer factor 2s (MEF2s), 1 play key and decisive roles during myogenesis (1)(2)(3). Myogenic regulatory factors and MEF2s can physically interact with each other to activate synergistically muscle-specific genes and are involved in regulating each other's expression (1). Several myogenic differentiation model systems based on cell cultures (e.g. C2C12 and L6) are available, which recapitulate some of the key steps during myogenesis and are useful in dissecting the signaling pathways controlling myogenesis. In proliferating C2C12 myoblasts, it is known that MyoD and several MEF2 members are already present (4,5). However, they are functionally repressed to ensure that differentiation does not start precociously (6 -8). The underlying repression mechanisms have just begun to be unraveled, which represent one of the major breakthroughs in the study of myogenesis in recent years. Both myogenic regulatory factors and MEF2s are bound and repressed by members of the histone deacetylase family (HDACs) (9,10). HDACs are thought to repress transcription by removing the acetyl group from histone tails resulting in tighter packaging of nucleosomes and reduced access of transcription factors to regulatory regions of genes (11,12). In addition to histones, HDACs are also known to deacetylate other cellular proteins including p53, MyoD, and tubulin, which modulate their biological activities ranging from DNA binding to microtubule dynamics (8,(13)(14)(15). MyoD is bound and repressed by class I HDACs, whereas MEF2s are directly bound and repressed by class II HDACs (9,10). Whereas the MADS/MEF2 motifs in MEF2 proteins directly interact with the class II HDACs, conserved motifs in the amino termini of the class II HDACs bind MEF2s (6, 16 -18). One of the characteristic features of the class II HDACs is that they shuttle between the nucleus and the cytoplasm under different cellular conditions (9,10,12,19). In the nucleus, class II HDACs mainly localize at nuclear bodies, the distinct punctate/or speckle-like structures with unknown identity (20,21). For HDAC5 and -7, they are preferentially nuclearly localized in proliferating myoblasts (22,23). At the start of differentiation in response to the Ca 2ϩ /calmodulin-activated protein kinasemediated signaling pathway, HDAC5/7 dissociate from MEF2s and translocate to the cytoplasm where they are retained by associating with 14-3-3 (22)(23)(24)(25)(26)(27). Although HDAC4 is equally effective in repressing MEF2 activities in reporter assays and in inhibiting differentiation of C2C12 cells when stably expressed (6,16), HDAC4 differs from HDAC5 and -7 in that it is mainly cytoplasmic in proliferating C2C12 and NIH3T3 cells (28 -30). At the start of differentiation, HDAC4 translocates to the nucleus (28,29). The underlying mechanism and its biological significance remain unclear.
Although HDAC4 contains a canonical deacetylase domain at its carboxyl terminus, it is enzymatically inactive on its own and exerts its repressive effect through recruitment of a corepressor complex consisting of HDAC3, SMRT, and N-CoR (31). Interestingly, the amino termini of HDAC4 (aa 1-208), -7 (aa 1-372), and -9 (aa 1-590, also called MEF2-interacting transcription repressor or MITR), in the absence of their carboxylterminal deacetylase domains, are also capable of repressing MEF2s (17,23,32). In the case of MITR, the underlying re-pression mechanisms are thought to be due to recruitment of either the carboxyl-terminal binding protein (CtBP), a transcription corepressor, or HDAC1 (32,33). Although a conserved CtBP-binding motif has been found in both MITR and HDAC4, mutations that disrupt CtBP binding to MITR only partially relieve the repressive effect of MITR, suggesting the existence of a second repressive region in the amino termini of MITR and HDAC4 (33). The identity of this second repressive region in either HDAC4 or MITR remains uncharacterized.
In this report, we looked further into the mechanism by which the amino terminus of HDAC4 (i.e. aa 1-208) represses MEF2.
Transfection, Cell Lysis, and Luciferase Reporter Assays-Cells were first transfected with various plasmids using LipofectAMINE TM Plus reagents (Invitrogen) following the manufacturer's instruction and cultured in either DMEM with 10% fetal bovine serum for 24 -36 h (for HeLa and COS-7) or growth medium for 36 h followed by culturing in differentiation medium for another 30 h (for C2C12) prior to harvest. Cells were then lysed in the lysis buffer (50 mM HEPES, pH 7.6, 1% (v/v) Triton X-100, 150 mM NaCl, 1 mM EGTA, 1.5 mM MgCl 2 , 100 mM NaF, 20 mM p-nitrophenyl phosphate, 20 mM ␤-glycerophosphate, 50 M sodium vanadate, 2 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 0.5 g/ml leupeptin, 0.7 g/ml pepstatin), followed by removal of insoluble debris with a bench-top centrifuge at 14,000 rpm for 2 min to obtain whole cell extracts (WCEs). Luciferase activity was determined with a luciferase assay kit (Roche Applied Science) and a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego). Luciferase units were normalized against total protein amount present in each sample determined by protein assay reagent from Bio-Rad.
Fluorescence Imaging of Live Cells-Fluorescent images of live HeLa or C2C12 myoblasts containing GFP fusion proteins were acquired by an Olympus IX70 fluorescent microscope linked to a charge-coupled device digital camera (Spot RT, Diagnostic Instruments Inc.).
Electrophoretic Mobility Shift Assays-A double-stranded oligonucleotide probe containing a consensus MEF2 site, 5Ј-GATCGCTCTA-AAAATAACCCTGTCG-3Ј (sense) (sc-2521, Santa Cruz Biotechnology), was end-labeled with T4 polynucleotide kinase and [␥-32 P]ATP. Unincorporated ATP was removed by QIAquick spin column (Qiagen). For each binding reaction, 10 g of HeLa WCE was added to a 20-l reaction mixture containing ϳ22,000 cpm of probe, 1 g of poly(dI-dC), and 20 g of bovine serum albumin in the binding buffer (15 mM HEPES, pH 7.6, 40 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 5% glycerol) with or without the recombinant GST-HDAC4-(1-208), GST-HDAC4-(119 -208), or GST. All the components above except the probe were first pre-mixed on ice for 30 min. Upon addition of the probe, the binding reactions were carried out for another 20 min at room temperature. The samples were loaded onto a 3.75% non-denaturing polyacrylamide gel, and electrophoresis was performed at 200 V for 2 h at 4°C. The gel was subsequently dried and subjected to autoradiography.
Co-immunoprecipitation Assays-COS-7 cells were co-transfected with various plasmids. 36 h after transfection, the cells were crosslinked with 20 g/ml dithiobis(succinimidyl propionate) (Pierce) for 5 min followed by lysis in RIPA buffer (25 mM HEPES, pH 7.4, 1% Nonidet P-40, 0.1% SDS, 0.5% sodium deoxycholate, 0.5 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 0.5 g/ml leupeptin, and 0.7 g/ml pepstatin). Protein A-Sepharose beads were first preincubated with 50 mg/ml of bovine serum albumin for 30 min to reduce nonspecific binding. 200 g of extracts were then added along with 2 g of anti-Xpress antibody for 2 h at 4°C. After extensive washing with the RIPA buffer, bound proteins were eluted out by boiling and subjected to SDS-PAGE and immunoblotting.
Pull-down Assays Using Biotinylated Oligonucleotide Probes Containing Either a Wild Type or a Mutant MEF2 Site-The sense strand of a wild type MEF2 oligonucleotide (5Ј-AAAAAAGGCTCTAAAAATA-ACCCCC-3Ј) (6) was synthesized with a covalently linked 5Ј biotin and annealed to its antisense strand to form a double-stranded probe. The mutant biotinylated MEF2 oligonucleotide is identical to the wild type one except for two base changes (in lowercase letters) in the MEF2 site: 5Ј-AAAAAAGGCTgTAAAcATAACCCCC-3Ј. COS-7 WCE containing FLAG-MEF2-(1-105) was preincubated with that containing either xp-HDAC4-(1-208) or FLAG-HDAC4FL at a 1.5 to 1 (v/v) ratio in the presence of 1 g/ml herring sperm DNA to minimize any nonspecific binding of cellular proteins to the MEF2 oligonucleotides in the following step. The combined extracts were then added to streptavidin-Sepharose beads (Amersham Biosciences) with or without the pre-bound biotinylated MEF2 oligonucleotides and incubated for 2 h at 4°C. Following extensive washing (1ϫ phosphate-buffered saline, 5% (v/v) glycerol), proteins retained were eluted out by boiling in 1ϫ SDS sample buffers.
Western Blot Analysis-Eluted proteins from pull-down or immunoprecipitation assays, together with their input controls (20 -30 g of WCEs), were resolved by SDS-PAGE, transferred to a polyvinylidene difluoride membrane (Immuno-Blot TM PVDF; Bio-Rad), and probed with various antibodies. Protein bands were visualized using the enhanced chemiluminescence kit (ECL, Amersham Biosciences).

RESULTS
The Amino-terminal Region (aa 119 -208) of HDAC4 Represses MEF2-dependent Transcription-It was shown previously that an amino-terminal fragment of HDAC4 (i.e. aa 1-208) binds MEF2 and represses MEF2-dependent transcription (17). A smaller fragment (aa 119 -208) of HDAC4 was also shown to bind MEF2 (17). However, it is not known whether it still processes the MEF2 repressive activity. To confirm the previous result and to further define the region that contains the repressive activity, we performed a reporter assay in HeLa cells using a MEF2-dependent luciferase reporter (3xMEF2luc). In agreement with previous results, the full-length HDAC4 and HDAC4-(1-208) significantly repressed MEF2-dependent transcription (Fig. 1A) (17). Interestingly, HDAC4-(119 -208) also repressed MEF2 in a dose-dependent manner (Fig. 1A). Although HDAC4-(119 -208) was less repressive compared with the full-length and HDAC4-(1-208), at a higher dose it could still repress the activity of the MEF2 reporter below the basal level. To test whether HDAC4-(119 -208) also repressed the endogenous MEF2, we co-transfected myogenic C2C12 cells with 3xMEF2-luc along with either the full-length HDAC4 or its two amino-terminal derivatives. As shown in Fig.  1B, all three HDAC4 constructs repressed the activity of the endogenous MEF2s to different extent with the full-length HDAC4 being the most repressive (ϳ95%) and HDAC4-(119 -208) being the least repressive (ϳ50%). Therefore, our results here demonstrated that HDAC4-(119 -208) can repress MEF2 in vivo.
The Amino-terminal Derivatives of HDAC4 Repress MEF2 Without Interfering with MEF2 DNA Binding-In the reporter assays above, the repression of MEF2 by HDAC4 or its aminoterminal derivatives could conceivably be caused by a block to either DNA binding or transactivating function of MEF2. To distinguish between the two possible mechanisms, we fused the genes encoding the full-length MEF2C and the yeast Gal4 DNA binding domain (Gal4DBD, aa 1-147) to generate the Gal4-MEF2 fusion protein, and we tested the repressive effect of the full-length HDAC4 and its amino-terminal derivatives on this fusion protein. As the MEF2 DNA binding domain was not required in this assay system, any repression detected would most likely be due to an effect on the transactivating function of MEF2. As shown in Fig. 2A, both the full-length HDAC4 and its amino-terminal derivatives repressed the transcriptional activity of the fusion protein, suggesting that they repress MEF2 by interfering with the transactivating function of MEF2. To prove directly that the amino-terminal derivatives of HDAC4 repress MEF2 without affecting its binding to DNA, we performed electrophoretic mobility shift assays using HeLa whole cell extracts. Our electrophoretic mobility shift assays resolved at least two MEF2-containing bands (Fig. 2B, lane 1), indicating the existence of different MEF2-containing higher order protein complexes. Importantly, both bands shifted upwards (supershift) in the presence of the bacterially expressed FIG. 1. HDAC4-(119 -208) represses MEF2-dependent transcription. A, HeLa cells were co-transfected with a 3xMEF2-luc reporter along with an empty vector, FLAG-MEF2C, and increasing amounts of the full-length (FL) HDAC4 or its amino-terminal derivatives in various combinations as indicated. The expression level of FLAG-MEF2C, HDAC4, and its amino-terminal fragments in each sample was assayed by immunoblot (IB). B, C2C12 cells were co-transfected with a 3xMEF2-luc reporter along with an empty vector, HDAC4 or its derivatives. Luciferase activity was measured and normalized against the total amount of proteins present in the sample. Fold activation is the ratio of the luciferase activity in cells transfected with various expression constructs over that transfected with the empty vector only. The data in A and B are presented as mean Ϯ S.E. from two separate transfection experiments. GST-HDAC-(1-208) or GST-HDAC4-(119 -208), indicative of binding of GST-HDAC4-(1-208) and GST-HDAC4-(119 -208) to the pre-existing MEF2-DNA complexes (Fig. 2B, lanes 3 and 4). Addition of GST alone had no such supershift effect (Fig. 2B,  lane 5). In HeLa cells transfected with increasing amounts of HDAC4-(1-208) at doses that were sufficient to repress MEF2 activity in the reporter assays, we detected no obvious change in the intensity and migration pattern of the MEF2-DNA complexes compared with the control (Fig. 2B, lanes 7 and 8 versus  6), nor could the transfected HDAC-(119 -208) affect MEF2 binding to DNA (data not shown). Furthermore, in a direct pull-down assay using double-stranded biotinylated oligonucleotides containing either a wild type or a mutant MEF2binding site, we showed that both HDAC4-(1-208) and the full-length HDAC4 could be pulled down by FLAG-MEF2-(1-105) only when the wild type MEF2 oligonucleotide was present (Fig. 2C, lanes 4 and 7). In contrast, the mutant MEF2 oligonucleotide (lanes 5 and 8) and the streptavidin beads alone (lanes 6 and 9) failed to pull down significant amounts of HDAC4-(1-208) and the full-length HDAC4. This suggests that HDAC4 and its amino-terminal derivative can specifically bind to the DNA-bound MEF2. Collectively, our results above demonstrated that the amino-terminal derivatives of HDAC4 repress MEF2 without interfering with MEF2 binding to DNA.
MyoD and HDAC4/HDAC4-(1-208)/HDAC4-(119 -208) Antagonize Each Other to Regulate MEF2-As both MyoD and HDAC4 interact with MEF2 through the MADS/MEF2 motifs (6,38), we asked whether MyoD and HDAC4s antagonize each other in regulating MEF2 activity. As seen in Fig. 4A, the repression of MEF2 by HDAC4 and its amino-terminal derivatives could be effectively relieved by MyoD in a dose-dependent manner. Moreover, MyoD was more effective in relieving MEF2 repressed by the amino-terminal derivatives than by the full-length HDAC4. The result above indicated that overexpression of MyoD could switch MEF2 from a repressed state to a highly activated state. Interestingly, MyoDT4basic, a MyoD mutant defective in MEF2 binding, failed to de-repress MEF2 as effectively as the wild type MyoD did (Fig. 4B) (39). This suggested that direct MEF2 binding is a prerequisite for MyoD to antagonize HDAC4. To demonstrate further that MyoD and HDAC4 compete to regulate MEF2, we studied the effect of HDAC4 and its amino-terminal derivatives on MyoD-mediated synergistic activation of 3xMEF2-luc. In the absence of HDAC4, addition of MyoD significantly activated the reporter gene activity due to interaction of the transfected MyoD with the endogenous MEF2 (Fig. 4C). In the presence of either HDAC4 or its amino-terminal derivatives, the synergistic effect between MyoD and MEF2 was significantly reduced/abolished (Fig. 4C). Therefore, our data above support the hypothesis that MyoD and HDAC4 antagonize each other to regulate MEF2.
The Amino Terminus of HDAC4 Targets MEF2 to Nuclear Bodies-Although HDAC4 is predominantly localized to the cytoplasm (28 -30), MEF2 is mainly a nuclear protein. A question arises as to how cytoplasmic HDAC4 can physically associate with nuclear MEF2 and repress its transcriptional activ- FIG. 3. HDAC4-(119 -208) represses MEF2 by specifically recruiting HDAC1. A, COS-7 cells were separately transfected with FLAG-HDAC4, FLAG-HDAC1, and FLAG-Six1 together with either xp-HDAC4-(119 -208) or an empty vector as indicated. Dithiobis(succinimidyl propionate) (20 g/ml) was added into the culture media for 5 min before cell harvest. 200 g of WCEs were incubated with an anti-Xpress antibody for immunoprecipitation, and the membrane was immunoblotted with an anti-FLAG antibody (bottom panel). 20 g of each input samples were also blotted with an anti-FLAG antibody (top panel). IB, immunoblot. IP, immunoprecipitation. B, C2C12 cells were transfected with a 3xMEF2-luc reporter together with either an empty vector or xp-HDAC4-(119 -208) as indicated. Cells were either left untreated or treated with TSA (50 nM) or Me 2 SO for 24 h in growth medium as indicated and were then shifted to differentiation medium (without TSA or Me 2 SO) for another 36 h before harvest. The luciferase activity and fold activation were determined as in Fig. 1. Data were presented as mean Ϯ S.E. from a duplicate experiment. Me 2 SO, dimethyl sulfoxide.
ity in vivo. To address this question, we studied the cellular distribution pattern of either protein with or without co-expression of the other. HeLa cells were co-transfected with various expression vectors in different combinations, and imaging of live cells containing green fluorescence protein (GFP)-tagged fusion proteins was taken (Fig. 5A, a-l). When expressed alone, GFP-HDAC4 and HDAC4-(1-208)-GFP were predominantly cytoplasmic and Pan-cellular, respectively (a and c), whereas GFP-MEF2C was evenly distributed in the nucleus (Pan-nuclear) (e). Interestingly, in the presence of the co-transfected MEF2 (no fluorescent tag), the localization of GFP-HDAC4 shifted from the cytoplasm to punctate nuclear bodies (b). Similarly, the localization of HDAC4-(1-208)-GFP also shifted from Pan-cellular distribution to punctate nuclear bodies when cotransfected with MEF2 (d). For GFP-MEF2, in the presence of the co-transfected HDAC4 or HDAC4-(1-208), the localization of GFP-MEF2 also shifted from Pan-nuclear distribution to punctate nuclear bodies (f and g). Although HDAC4-(1-208) was less effective than the full-length HDAC4 in clustering MEF2 (f and g), this MEF2-clustering effect was specific, as HDAC4-(119 -208) failed to target MEF2 to nuclear bodies (h). Similar results were also seen in C2C12 cells (i-l). Of note, HDAC4-(1-208) is more effective in targeting MEF2 to nuclear bodies in C2C12 cells than in HeLa cells (compare d/j and g/l). Quantitative analyses of subcellular distribution of HDAC4 and MEF2 based on experiments in Fig. 5A were also presented (Fig. 5, B-D). Our data suggest that MEF2 and HDAC4 associate with each other in vivo and that the amino terminus (i.e. aa 1-208) of HDAC4 specifically targets MEF2 to distinct nuclear bodies.

Several Independent Repression Domains Exist in HDAC4 -
As mentioned in the Introduction, several independent repressive domains are already known to exist in HDAC4 and other class II HDACs. They include the amino-terminal CtBP-bind-ing motif and the carboxyl deacetylase domain that recruits SMRT/N-CoR/HDAC3 (31,33). In a recent report, two motifs in MITR (aa 390 -400 and 400 -490) were shown to bind heterochromatin protein 1 (HP1) which recognizes the methylated lysine in histone tails and represses transcription by recruiting additional HDACs and histone methyltransferases (40). Conserved regions in both HDAC4 and -5 (roughly around aa 400 -540 before the start of the carboxyl deacetylase domain) were also shown to bind HP1 (40). In addition to these known repressive domains, we demonstrate in this report that a small region in the amino terminus of HDAC4 (aa 119 -208), which is physically separate from the CtBP-binding motif (aa 43-57) and without the HP1-binding region and the carboxyl deacetylase domain, is still capable of binding and repressing MEF2. HDAC4-(119 -208) specifically associates with HDAC1 which accounts for its repressive activity. This result is further supported by a previous finding that HDAC1 associates with MITR which also contains a region homologous to HDAC4-(119 -208) (Fig. 6A) (32,33). Thus, it appears that the full-length HDAC4 can repress transcription through at least four independent domains: binding to CtBP through aa 43-57, recruitment of HDAC1 through aa 119 -208, interaction with HP1 through aa 408 -510, and association with HDAC3/SMRT/N-CoR through the carboxyl deacetylase domain (aa 612-1084) (Fig. 6B). Our data also suggest that the four distinct repression domains of HDAC4 cooperate with each other to bring about the most efficient transcription repression, as loss of either two or three repressive domains (in the case of HDAC4-(1-208) and HDAC4-(119 -208), respectively) results in gradual decrease of the repressive activity in the truncated HDAC4s (Figs. 1A, 2A,  and 4C). Furthermore, we think other class II HDACs may also recruit HDAC1 through a similar mechanism as they all contain motifs at the amino termini that are homologous to aa 119 -208 in HDAC4 (Fig. 6A).
Interplay among HDAC4, MEF2, and MyoD-It has been shown that the MADS/MEF2 motif of MEF2 is involved in interaction with both MyoD and HDAC4 (6,38). The question arises as to whether MyoD and HDAC4 compete for MEF2 binding in a mutually exclusive manner or whether both proteins can bind MEF2 simultaneously. Although direct biochemical binding data are lacking (mainly due to weak binding between MEF2 and MyoD/E proteins in vitro), we favor the "mutually exclusive model." On one hand, the synergistic activation of a MEF2-dependent reporter by MEF2 and MyoD can be effectively abolished by either the full-length HDAC4 or, to a lesser extent, HDAC4-(1-208) and HDAC4-(119 -208) (Fig.  4C). On the other hand, only the wild type MyoD, but not a MyoD mutant defective in MEF2 binding, can overcome the repressive effect of either the full-length or the amino-terminal derivatives of HDAC4 in a dose-dependent manner (Fig. 4, A  and B). The latter result differs from a previous report (6). The ratio of MyoD over HDAC4 used in the transfection scheme may underlie the difference. In addition, the weaker binding between MEF2 and MyoD/E proteins may partly explain why it is less efficient for MyoD to counteract the repressive effect of HDAC4 on MEF2. Furthermore, the mutually exclusive model was supported by the dynamic nature of class II HDACs during myogenic differentiation. In myoblasts, class II HDACs associate with MEF2 to repress transcription of muscle-specific genes. Although MyoD is present in C2C12 myoblasts, it is in a repressed state itself and is not believed to interact with MEF2 (7). At the beginning of differentiation when calmodulin-activated protein kinase is activated, class II HDACs will be dissociated from MEF2 and shuttled to the cytoplasm (9,10). This would be the time when MyoD (perhaps together with p300/ PCAF) is assembled on MEF2 to synergistically activate the transcription of pro-differentiation genes.
HDAC4 differs from other class II HDACs in that it is mainly cytoplasmic in proliferating C2C12 myoblasts (18,29). We show that the cytoplasmic HDAC4 can be shuttled to the nucleus by co-transfected MEF2 (Fig. 5A, b), in agreement with previous findings (18,30). This suggests that MEF2 may be one of the limiting factors in determining the subcellular localization of HDAC4. Due to limited amount of endogenous MEF2 in myoblasts, a large fraction of HDAC4 that cannot associate with MEF2 ends up in the cytoplasm. As MEF2 protein level increases during differentiation (41), MEF2 would be expected to induce nuclear translocation of HDAC4 in differentiated myotubes, which is in agreement with previous findings (28,29). This could be a unique role played by HDAC4 among class II HDACs to ensure that active MEF2 is repressed again by HDAC4 in myotubes at the end of differentiation process (presumably when calmodulin-activated protein kinase signal is already down-regulated).
From our data in Fig. 5A, it is also interesting to note the intricate interplay between MEF2 and HDAC4 in cells. Whereas MEF2 is the driving force to drag HDAC4 from the cytoplasm into the nucleus (Fig. 5A, b), once in the nucleus, HDAC4 can target MEF2 to distinct clusters of nuclear bodies (Fig. 5A, f). As neither GFP-HDAC4 nor GFP-MEF2 alone localizes at nuclear bodies, our results suggest that another factor(s) may influence the localization of the HDAC4-MEF2 complex in the nucleus. SMRT is one such candidate factor, as it normally localizes at nuclear bodies and also drives cytoplasmic HDAC4 to nuclear bodies when co-expressed with HDAC4 in the cells (21). As HDAC4 was previously found to interact with SMRT through its carboxyl terminus (31), it came as a surprise that aa 1-208 of HDAC4 still retains the ability to target MEF2 to nuclear bodies (Fig. 5A, g and l), suggesting that either SMRT or another unknown factor is responsible for targeting the HDAC4-(1-208)-MEF2 complex to nuclear bodies. It is also noteworthy that HDAC4-(1-208) is more efficient to target MEF2 to nuclear bodies in C2C12 than in HeLa cells (Fig. 5A, b and j, and d and l). The abundance of the factor (e.g. SMRT) in different cell lines may underlie such a difference. Unlike HDAC4-(1-208), aa 119 -208 of HDAC4 fails to efficiently target MEF2 to nuclear bodies (Fig. 5A, h), suggesting that the aa 1-118 of HDAC4 is crucial for targeting MEF2 to nuclear bodies.
It would be interesting to find out next which adaptor molecule(s) bridges HDAC1 and aa (119 -208) of HDAC4 and which molecule(s) (in addition to SMRT) specifically targets the HDAC4-MEF2 complex to nuclear bodies. Other issues worthy of consideration in the near future include whether the endog-enous MEF2 associates with class II HDACs in the free or DNA-bound form and which of the class II HDACs preferentially binds MEF2 in proliferating myoblasts and mature myotubes.