Regulation of MEF2 by Histone Deacetylase 4- and SIRT1 Deacetylase-Mediated Lysine Modifications

DNA constructs and antibodies.

HDAC4 and HDAC5 expression plasmids were generated by subcloning the coding sequences from PBJ-HDAC4 and -HDAC5 (kindly provided by S. Schreiber) into the pcDNA3 vector. The HDAC4 coiled-coil domain (amino acids 67 to 257) was cloned into the pGBKT7 vector (Clontech), and a yeast two-hybrid assay was performed according to the company-provided protocol. Ubc9 cDNA was amplified from a human brain cDNA library and cloned into the pcDNA3 vector. The SUMO expression construct was kindly provided by Michael Nevels (Princeton University). The MCK-luciferase and 3XMEF2-Luc reporters were kindly provided by Rhonda Bassel-Duby and Eric Olson (University of Texas, Southwestern Medical Center). To generate the SUMO-MEF2 fusion protein, SUMO-1 (amino acids 1 to 96) was amplified by PCR and ligated with the MEF2D N terminus into the pcDNA3 vector. Expression plasmids for HDAC4 K559R, MEF2D K424R, and MEF2D I423A mutants were created by a Quikchange site-directed mutagenesis kit (Stratagene). Anti-SUMO antibody and anti-MEF2D antibody were purchased from Zymed Laboratories and BD Biosciences, respectively. Anti-acetylated lysine antibody was kindly provided by Minoru Yoshida (RIKEN Institute, Japan). Anti-MHC (MF20) was used as described previously (6).

Tissue culture, transfections, and Western blots.

293T and 10T1/2 cells were maintained in Dulbecco's modified Eagle medium (DMEM) with 10% fetal bovine serum at 37°C in a humidified atmosphere with 5% CO₂. Calcium phosphate-mediated transfections were performed as described previously to transfect 293T cells (41). Thirty-six hours after transfection, cells were harvested and lysed with RIPA buffer containing 100 mM N-ethylmaleimide. Five-hundred-microgram aliquots of protein were immunoprecipitated with anti-MEF2D antibody overnight, followed by immunoblotting using anti-SUMO or anti-acetylated lysine antibody. Endogenous MEF2 sumoylation was detected as described previously (17). Briefly, C2C12 cells were lysed in RIPA buffer containing 1% SDS. After a quick sonication, cell lysates were diluted 10-fold for immunoprecipitation with an anti-MEF2 antibody. For the endogenous MEF2 acetylation assay, C2C12 cells were induced to differentiate by serum withdrawal for different time periods, as indicated. The cells were treated with nicotinamide for 4 h before being harvested. The cells were lysed in 170 mM NETN (0.5% NP-40, 1 mM EDTA, 20 mM Tris-HCl, 170 mM NaCl), and MEF2 was immunoprecipitated and analyzed by Western blotting as stated above.

Coimmunoprecipitation assays.

For the interaction of HDAC4 and Ubc9, HDAC4 and Ubc9 were coexpressed in U2OS cells, which were then lysed in RIPA buffer. After a brief sonication, cell lysates were precleared with protein G beads for 1 h. The supernatants were then incubated with anti-Myc (9E10) antibody and protein G beads overnight. For the interaction of HDAC4 and SIRT1, 293T cells coexpressing SIRT1 and HDAC4 were lysed in 170 mM NETN and incubated with 9E10 antibody and protein G beads. In both cases, the immunoprecipitation products were analyzed by Western blotting using an anti-FLAG antibody to detect immunoprecipitated HDAC4.

Myogenic conversion assay.

Polyamine-based transfection (Genejammer; Stratagene) was carried out to transiently transfect 10T1/2 cells using company-provided instructions. Twenty-four hours after transfection, the growth medium (GM) was switched to differentiation medium (DM) (DMEM with 2% horse serum). Four days later, the cells were harvested and lysed in cracking buffer containing 4 M urea (27). The muscle cell differentiation marker myosin heavy chain (MHC) was detected in 30-μg aliquots of protein by a Western blot using anti-MHC antibody (MF20).

In vitro sumoylation assay.

Recombinant SUMO-1, SAE-1, and SAE-2 were expressed as glutathione S-transferase (GST) fusion proteins using the pGEX series of plasmids. H6-tagged Ubc9 was expressed from plasmid pET15b-Ubc9, which contains the Ubc9 cDNA as an XhoI/BamHI fragment. Recombinant HDAC4 was purified from a baculovirus expression system. In vitro sumoylation reactions were carried out in 20 μl of reaction buffer as described previously (36), supplemented with 2 mM ATP. One hundred nanograms of GST-SAE1, 150 ng of GST-SAE2, 50 ng of H6-Ubc9, 1 to 2 μg of SUMO-1, and 1 to 2 μl of in vitro-translated, ³⁵S-labeled MEF2D were added to the reaction mix. Reactions were carried out for 45 min at 30°C and analyzed by SDS-polyacrylamide gel electrophoresis. Quantitation was performed using a PhosphorImager (Molecular Dynamics). After subtraction of the background, the sum of the signal for modified plus unmodified MEF2 in each lane was set to 100%, and the relative values are given accordingly.

Luciferase reporter assay.

10T1/2 cells were transfected by using the Genejammer transfection reagent (Stratagene) in 24-well plates with plasmids for the MCK-Luc reporter (0.4 μg), MEF2D or the MEF2D K424R (I423A) mutant (0.2 μg), MyoD (10 ng), and pCMX-βGal (0.1 μg). Twenty-four hours after transfection, cells were collected, and the luciferase activity was measured and normalized by the β-galactosidase activity as previously described (45).

HDAC4 interacts with Ubc9 and is subject to sumoylation.

Sequence analysis identified a coiled-coil domain at the N termini of HDAC4 and the closely related HDAC5 (1) (Fig. 1A). This domain is located in a region previously shown to mediate HDAC4 binding to MEF2 and calmodulin (26, 42), suggesting that it might be important for protein-protein interactions. To identify novel HDAC4-interacting proteins, we performed a yeast two-hybrid screen, using this coiled-coil domain (amino acids 67 to 257) as bait. One of the positive clones was found to encode Ubc9, an E2 SUMO-conjugating enzyme (25). As shown in Fig. 1A, Ubc9 specifically interacts with the coiled-coil domains of HDAC4 and HDAC5, but not with the C terminus of HDAC4, in a yeast two-hybrid analysis. A coimmunoprecipitation assay also confirmed the association between Ubc9 and HDAC4 (Fig. 1B). Further supporting the observation that HDAC4 and Ubc9 functionally interact, the coexpression of HDAC4 and Ubc9 in 293T cells leads to specific sumoylation of HDAC4 at lysine 559 (Fig. 1C, lanes 4 and 5). However, the subsequent analysis failed to reveal obvious defects associated with an HDAC4 mutant that cannot be sumoylated (HDAC4 K559R). This sumoylation-deficient HDAC4 mutant appears to be indistinguishable from wild-type HDAC4 in its transcriptional repression activity (Fig. 1D) and deacetylase activity (data not shown). This observation indicates that sumoylation of HDAC4 might not be the key outcome of the HDAC4-Ubc9 interaction. Consistent with this idea, although HDAC5 also interacts with Ubc9, it is not subject to sumoylation (Fig. 1C, lane 6).

HDAC4 promotes MEF2 SUMO modification in vivo and in vitro.

We therefore considered the alternative hypothesis that HDAC4 recruits the Ubc9 machinery to regulate the sumoylation of HDAC4-associated proteins such as MEF2. To this end, we first determined whether endogenous MEF2 is subject to sumoylation. As shown in Fig. 2A, the immunoprecipitation of endogenous MEF2D from C2C12 cells revealed a slower-migrating MEF2D species with the mobility predicted for SUMO-1-conjugated MEF2 (lower panel). Indeed, this MEF2D species can also be recognized by an anti-SUMO-1 antibody (upper panel), strongly indicating that endogenous MEF2 is subject to SUMO modification.

We then determined whether HDAC4 modulates MEF2 sumoylation. Similar to endogenous MEF2D, ectopically expressed MEF2D also becomes moderately sumoylated when it is coexpressed with SUMO-1 (Fig. 2B, lane 2, upper panel). Remarkably, the coexpression of HDAC4 dramatically increases the level of sumoylated MEF2D (Fig. 2B, lane 3, upper panel). Normalized quantification of the sumoylated MEF2 revealed that HDAC4 can stimulate MEF2D sumoylation more than fourfold. This effect is specific, as an HDAC4 mutant that does not have the coiled-coil domain is inactive in this assay (lane 4). Importantly, the closely related HDAC5, which also contains the coiled-coil domain and interacts with Ubc9, potently promotes MEF2D sumoylation as well (lane 8). In contrast, other HDAC members, such as HDAC1 and HDAC6, have little effect (lanes 5 and 7). Finally, PIAS1 and PIASy, which function as SUMO E3 ligases for p53 and LEF1 (32, 34), do not efficiently stimulate MEF2D sumoylation (Fig. 2B, lane 6, and data not shown). Together, these data demonstrate that MEF2D is a sumoylated protein and that HDAC4 can specifically stimulate MEF2 sumoylation in vivo.

To determine whether HDAC4 has the activity to promote MEF2 sumoylation directly, we examined the effect of recombinant HDAC4 on MEF2 sumoylation in a reconstituted system in vitro. To this end, in vitro-translated MEF2D was incubated with purified recombinant SUMO E1-activating enzymes (SAE1/SAE2), E2 Ubc9, and SUMO-1 in the presence or absence of recombinant HDAC4. As shown in Fig. 2C, a small population of MEF2D becomes sumoylated after incubation with the reconstituted SUMO machinery (upper panel, lanes 1 and 2). Significantly, the addition of purified recombinant HDAC4 to the reaction dramatically increases the production of sumoylated MEF2D, with a concomitant reduction in the unmodified species (Fig. 2C, lanes 3 to 5). Quantification of these experiments revealed that recombinant HDAC4 can stimulate MEF2D sumoylation more than fivefold in a dose-dependent manner (Fig. 2C, lower panel). These results strongly support a direct role for HDAC4 in promoting MEF2 sumoylation.

A sequence analysis of MEF2D revealed the presence of one consensus sumoylation motif surrounding lysine 424. Notably, this motif is highly conserved and can be found in most isoforms of MEF2 from different animal species (Fig. 2D). Supporting the hypothesis that lysine 424 is the key SUMO acceptor, the conversion of lysine 424 to arginine (K424R) completely abrogates HDAC4-dependent MEF2D sumoylation in cells (Fig. 2E) and in the in vitro assay (data not shown). Together, these results demonstrate that HDAC4 can stimulate MEF2 sumoylation on lysine 424 both in vivo and in vitro.

MEF2D lysine 424 is subject to reversible acetylation catalyzed by CBP and SIRT1.

Recently, several lysine-targeting modifications, such as acetylation, sumoylation, and ubiquitination, have been reported as important modulators of transcription (7). Our observation that HDAC4, a protein deacetylase, can stimulate MEF2 sumoylation prompted us to ask whether MEF2 is also subject to lysine acetylation. To test this, we first determined whether MEF2 can be acetylated by the acetyltransferase CBP, which has been shown to promote MEF2 activity and is essential for C2C12 cell differentiation (6, 33). As shown in Fig. 3A, we found that wild-type CBP can potently induce MEF2D acetylation. An enzymatically deficient mutant of CBP fails to do so. Importantly, mutation of the SUMO acceptor lysine 424 dramatically reduces CBP-induced MEF2 acetylation (Fig. 3B), revealing lysine 424 as the major acceptor for both acetylation and sumoylation and suggesting a potential interplay of these two lysine-targeting modifications in regulating MEF2 function.

Since lysine 424 is targeted by both acetylation and sumoylation, we hypothesized that HDAC4 might promote lysine 424 sumoylation by deacetylating this lysine residue. Surprisingly, however, the ectopic expression of HDAC4 does not induce efficient MEF2 deacetylation (Fig. 4A, lane 3). Supporting this conclusion, treatment with a potent HDAC4 deacetylase inhibitor, trichostatin A (TSA), has little effect on the MEF2 acetylation level (Fig. 4B, lane 3), suggesting an HDAC4-independent deacetylation of MEF2. In contrast, a treatment with nicotinamide, an inhibitor of the class III NAD⁺-dependent Sir2 family deacetylases, dramatically increases MEF2 acetylation (Fig. 4B, lanes 4 and 5). Importantly, a differentiation-dependent acetylation of endogenous MEF2 can also be prominently induced by nicotinamide but not TSA treatment (Fig. 4C and data not shown). These results strongly suggest that MEF2 acetylation is dynamically regulated during muscle cell differentiation and that one or more of the Sir2 family proteins, but not HDAC4, function as the MEF2 deacetylase.

Among the Sir2 family members, SIRT1 was shown to repress the muscle cell differentiation program (8). We thus tested whether SIRT1 can function as a MEF2 deacetylase. As shown in Fig. 4D, the cotransfection of SIRT1 potently reduces CBP-mediated MEF2D acetylation, whereas a catalytically inactive SIRT1 mutant (H363Y) and another Sir2 member, SIRT2, do not cause efficient MEF2D deacetylation (Fig. 4D and data not shown). Thus, SIRT1 can function as a MEF2 deacetylase. Interestingly, coimmunoprecipitation assays revealed that SIRT1 interacts with HDAC4 but not with HDAC6 (Fig. 4E). These results reveal that the class III deacetylase SIRT1 can function as a MEF2 deacetylase and can form complexes with the class II deacetylase HDAC4.

SUMO modification negatively regulates MEF2 activity.

Since acetylation and sumoylation both target lysine 424 (Fig. 2E and 3B) and since acetylation positively correlates with MEF2 activity during the muscle cell differentiation process (Fig. 4C), we hypothesized that HDAC4-induced sumoylation negatively regulates MEF2 activity. To examine this, we coexpressed MyoD with wild-type or sumoylation-deficient mutant (MEF2D K424R) MEF2D and assessed their ability to activate a luciferase reporter controlled by the muscle creatine kinase (MCK) enhancer. Consistent with previous reports, wild-type MEF2D and MyoD activate the MCK reporter cooperatively (Fig. 5A, upper panel). Importantly, the sumoylation-deficient mutant MEF2D K424R is consistently more active than wild-type MEF2D in collaborating with MyoD to activate the MCK reporter (Fig. 5A, upper panel), suggesting that sumoylation negatively regulates MEF2 activity. A similar conclusion was made by assessing the transcriptional activity of another sumoylation-deficient mutant, MEF2D I423A (Fig. 5A, upper panel), in which the sumoylation consensus site (IKSE) was destroyed in order to eliminate only MEF2 sumoylation but still keep acetylation (Fig. 5A, lower panel). To further confirm that sumoylation negatively regulates MEF2D activity, we fused SUMO to MEF2D to mimic sumoylated MEF2. Both the SUMO-MEF2D and SUMO-MEF2DK424R fusion proteins fail to synergize with MyoD in activating the MCK-luciferase reporter (Fig. 5B) or a reporter driven by tandem MEF2 binding sites (data not shown). These data support the idea that sumoylated MEF2 is transcriptionally inactive. Lastly, we examined the activity of sumoylation-deficient MEF2 in converting 10T1/2 fibroblasts to myocytes. As shown in Fig. 5C, coexpression of the sumoylation-deficient MEF2D mutant with MyoD induces MHC in 10T1/2 fibroblasts more potently than that of wild-type MEF2. Together, our data indicate that the conjugation of SUMO down-regulates MEF2 activity.