Calcium/Calmodulin-dependent Protein Kinase Activates Serum ...

THE JOURNAL OF BIOLOGICAL CHEMISTRY © 2003 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 278, No. 22, Issue of May 30, pp. 20047–20058, 2003 Printed in U.S.A.

Calcium/Calmodulin-dependent Protein Kinase Activates Serum Response Factor Transcription Activity by Its Dissociation from Histone Deacetylase, HDAC4 IMPLICATIONS IN CARDIAC MUSCLE GENE REGULATION DURING HYPERTROPHY* Received for publication, September 30, 2002, and in revised form, March 14, 2003 Published, JBC Papers in Press, March 26, 2003, DOI 10.1074/jbc.M209998200

Francesca J. Davis‡, Madhu Gupta§‡‡, Blanca Camoretti-Mercado¶, Robert J. Schwartz储, and Mahesh P. Gupta‡** From the Departments of ‡Surgery and ¶Medicine, University of Chicago, Chicago, Illinois 60637, the §Hope Children’s Hospital, Oak Lawn, Illinois 60453, the ‡‡Department of Physiology and Biophysics, University of Illinois, Chicago, Illinois 60612, and the 储Department of Cell Biology, Baylor College of Medicine, Houston, Texas 77030

Serum response factor (SRF) plays a pivotal role in cardiac myocyte development, muscle gene transcription, and hypertrophy. Previously, elevation of intracellular levels of Ca2ⴙ was shown to activate SRF function without involving the Ets family of tertiary complex factors through an unknown regulatory mechanism. Here, we tested the hypothesis that the chromatin remodeling enzymes of class II histone deacetylases (HDAC4) regulate SRF activity in a Ca2ⴙ-sensitive manner. Expression of HDAC4 profoundly repressed SRFmediated transcription in both muscle and nonmuscle cells. Protein interaction studies demonstrated physical association of HDAC4 with SRF in living cells. The SRF/ HDAC4 co-association was disrupted by treatment of cells with hypertrophic agonists such as angiotensin-II and a Ca2ⴙ ionophore, ionomycin. Furthermore, activation of Ca2ⴙ/calmodulin-dependent protein kinase (CaMK)-IV prevented SRF/HDAC4 interaction and derepressed SRF-dependent transcription activity. The SRF䡠HDAC4 complex was localized to the cell nucleus, and the activated CaMK-IV disrupted HDAC4/SRF association, leading to export of HDAC4 from the nucleus and stimulation of SRF transcription activity. Thus, these results identify SRF as a functional interacting target of HDAC4 and define a novel tertiary complex factor-independent mechanism for SRF activation by Ca2ⴙ/CaMK-mediated signaling.

Serum response factor (SRF)1 is a key regulator of several extracellular stimuli-regulated genes important for cell growth, apoptosis, and differentiation. SRF was first identified by its ability to confer the serum-activated expression of the

* This work was supported by NHLBI, National Institutes of Health, Grant RO-1 HL-68083 and the American Heart Association National Center Grant-in-aid 150108N (to M. P. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ** To whom correspondence should be addressed: Dept. of Surgery, MC-5040, University of Chicago, 5841 S. Maryland Ave., Chicago, IL 60637. Tel.: 773-834-7811; Fax: 773-702-4187; E-mail: mgupta@ surgery.bsd.uchicago.edu. 1 The abbreviations used are: SRF, serum response factor; SRE, serum response element; DAPI, 4⬘,6-diamidino-2-phenylindole; HAT, histone acetyltransferase(s); HDAC, histone deacetylase; CREB, cAMPresponse element-binding protein; PBS, phosphate-buffered saline; CaMK, Ca2⫹/calmodulin-dependent protein kinase; HA, hemagglutinin; TRITC, tetramethylrhodamine isothiocyanate; TSA, trichostatin A; FITC, fluorescein isothiocyanate; SRF-FL, full-length SRF; ANF, atrial naturatic factor; MHC, myosin heavy chain. This paper is available on line at http://www.jbc.org

c-fos gene in replicating cells (1). Paradoxically, SRF was also found to regulate expression of several muscle genes, which are expressed specifically in postmitotic myocytes (2). SRF acts through binding as a homodimer to the DNA consensus sequence CC(A/T)6GG, the serum response element (SRE), also referred as the CArG box, which is found essential for tissuespecific expression of numerous striated as well as smooth muscle-specific genes (2, 3). SRF plays a central role in the induction and maintenance of cardiac myogenic program, as exemplified by disruption of the SRF gene, which prevented mesoderm differentiation and cardiac development (4). SRF is differentially expressed in embryonic and adult cardiac myocytes, being at least 2 orders of magnitude greater than those detected in cells of endodermal origin (5). SRF has the ability to physically interact and synergistically cooperate with many other known cardiac-myogenic factors, such as GATA-4, Nkx2.5, TEF-1, myocardin, and CRP1/2 (6 –10). Functionally relevant CArG boxes in the promoter regions of several cardiac-restricted, contractile, Ca2⫹transporting, and metabolic protein genes have been identified, indicating a direct role of SRF in their transcriptional regulation (6, 11, 12). A large body of evidence also indicates that SRF-containing complexes are the end point targets in pathways associated with conversion of hypertrophic stimuli to cardiac cellular response (e.g. induction of “fetal gene program”) where re-expression of genes abundant in the embryonic heart occur (including activation of ANF, skeletal ␣-actin, and ␤-MHC genes, and repression of SRCa2⫹ATPase and ␣-MHC genes) (13–17). However, the underlying mechanism that enables SRF to transduce intracellular signals to the hypertrophic response of cardiac myocytes is not yet known. Several mechanisms have been shown to regulate the SRF transcription activity, including physical interaction of SRF with a number of positive and negative cofactors (6 –10), phosphorylation-dependent change in the DNA and/or protein binding ability of SRF (18), regulated nuclear translocation of SRF (19), and alternative splicing of SRF mRNA primary transcript (20, 21). Recently, we have shown that an alternative spliced isoform of SRF is highly expressed in the failing hearts of both humans and animals, which act as a dominant negative isoform to repress SRF-dependent cardiac muscle gene expression (22). While studying the mechanism of SRF in the hypertrophic growth of cardiac myocytes, several recent reports have evoked our interest to examine the role of chromatin-remodeling enzymes, histone acetyltransferases (HAT) and deacetylases (HDAC), in SRF-mediated cardiac muscle gene activation. By changing the acetylation state of histones, histone acetylases/

20047

20048

HDAC4 Regulates the Transcription Activity of SRF

deacetylases modify the chromatin structure that alters the accessibility of DNA to transcription factors. These enzymes are often recruited as cofactor complexes on the promoter of different genes, where they associate with transcription factors involved in expression of the target gene (23). During differentiation of P19 cells to smooth muscle cells, SRF was shown to be recruited to target genes along with hyperacetylation of histones at smooth muscle-specific regulatory regions of chromatin (24). Similarly, SRF and the HAT-containing co-activator, cAMP-response element binding protein (CREB)-binding protein, was shown recruited to the CArG box of the SM22 promoter during gene activation (25). SRF also physically interacts with the HAT-containing activators, CREB-binding protein/p300, during CArG box-mediated activation of the c-fos gene expression (26). Moreover, Rho-A signaling mediated activation of SRF function on the c-fos promoter was observed only when local chromatin was hyperacetylated, indicating that histone modulation is required for the conversion of intracellular signals to cellular response via a CArG box-dependent mechanism (27). Recently, several myogenic factors, including MyoD, twist and another MADS-containing factor, MEF2, have been shown to associate directly with cofactors having HAT (CREB-binding protein, p300, p300/CBP-associated factor) or HDAC (HDAC4/5) activities, leading to a change in their muscle gene activation potential (28 –30). Based on these reports, a pertinent question may be raised of whether SRF-mediated cardiac muscle gene transcription could also be regulated by histone acetylases/deactetylases and whether this regulation is altered by hypertrophic signals that up-regulate cardiac muscle gene expression? We were particularly interested in class II HDACs, because, as opposed to class I HDACs, the members of class II, which include HDAC4, -5, -7, and -9, have been shown to exhibit tissue-specific expression. HDAC4 and HDAC5 are expressed at high levels in the heart, skeletal muscle, and brain, where they have been suggested to be involved in cell differentiation and development (31–33). In this study, we demonstrate that SRF directly associates with HDAC4 in cardiac myocytes via its MADS domain, resulting in repression of SRF gene activation potential. Both proteins, upon interaction, co-localize to the nucleus of cardiac myocytes. We also show that the SRF/HDAC4 association is a target for upstream signaling events that increase intracellular Ca2⫹ levels and activate CaMK signaling during hypertrophic growth of myocytes. The increased activity of Ca2⫹/CaMK signaling results in dissociation of SRF from HDAC4, leading to export of HDAC4 from the nucleus and restoration of the SRF transcriptional activity. These results demonstrate that under basal conditions the transcription activity of SRF is negatively controlled by HDAC4 and that this is reversed upon activation of Ca2⫹/CaMK signaling. The unmasking of SRF transcription activity in this manner through dissociation from HDACs could be a key regulatory mechanism involved in transducing multiple and diverse intracellular signals that activate SRF-dependent cardiac muscle gene transcription. MATERIALS AND METHODS

Cell Culture and Transfection Primary cultures of cardiac myocytes were prepared from 2-day-old neonatal rats as previously described (34). After differential plating to eliminate fibroblasts, myocytes were further purified using a Percoll density gradient (Amersham Biosciences) and plated at a density of 4 ⫻ 106 in 100-mm plates, precoated with 2% gelatin. Cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and 100 units/ml penicillin and 100 mg/ml streptomycin. COS and C2C12 cells were typically plated in Falcon six-well tissue culture dishes at a density of ⬃3–5 ⫻ 105 cells/well and maintained in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum and penicillin/streptomycin combination. Cells were transfected 48 h after

plating, using LipofectAMINE reagent (Invitrogen) according to the manufacturer’s protocol. The pCMV-␤ gal was used as a reference plasmid in all transfections. After 48 h, transfected cells were harvested, and cell lysates were prepared and assayed for luciferase (Luciferase Assay System, Promega, Madison, WI), ␤-galactosidase activities, and protein content (Bio-Rad protein reagent).

Plasmid Constructs pcDNA3Gal4-HDAC4 (designated as Gal-HDAC4), pcDNA3Gal4HDAC4-D840N (designated as Gal-HDACD840N), pcDNA3Myc-HDAC4 (having a Myc tag at the COOH terminus of HDAC4), and pcDNA3MycHDAC4-D840N were kindly provided by Dr. T. Kouzarides (Wellcome/CRC Institute, University of Cambridge, Cambridge, UK) and have been described (35). Luciferase reporter constructs for skeletal ␣-actin and expression vectors pCGN and pCGNSRF-FL have been previously described (7). pBS-SRF constructs having deletions of different regions of SRF were kindly provided by Dr. R. Prywes (Columbia University, New York, NY). The reporter construct with five SREs was purchased from Stratagene. pcDNA-SRF-M (lacking exon 5) was obtained from Dr. P. Kemp (Cambridge University, Cambridge, UK) (20). pcDNA-SRF-S (lacking exons 4 and 5) has been described elsewhere (22). pcDNA3-CaMK-IVdCT (active) and pcDNA3-CaMK-IVdCTK75E (kinase-dead) plasmids were provided by Dr. D. L. Black (Howard Hughes Medical Institute, UCLA, Los Angeles, CA) (36). The plasmids SR-␣-CaMK-II active (␣-CAMK-T286D) and SR-␣CaMK-II inactive (␣-CaMK-K42M) were provided by Dr. M. R. Rosner (The Ben May Institute, University of Chicago) (37). The plasmid pCMV-p300 was obtained from Dr. D. M. Livingston (Dana-Faber Institute, Harvard University, Boston, MA) (38).

Western Analysis and Co-immunoprecipitation of Proteins Unless otherwise specified, all common salts and reagents were obtained from Sigma. Whole cell lysates from frozen tissues or from cultured cells were prepared according to the method described before (22). For co-immunoprecipitation studies, whole cell lysate (⬃500 –700 ␮g of protein) was first incubated with 0.5 ␮g of normal rabbit IgG and 20 ␮l of protein A/G-agarose beads at 4 °C for 30 min. The precleared lysate was obtained by separation of beads by centrifugation at 1,000 ⫻ g for 5 min at 4 °C and incubated with 20 ␮g of anti-SRF antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) at 4 °C for 60 min. The absence of primary antibody in parallel reaction mixes served as negative control. Protein A/G-agarose (20 ␮l) was then added to each sample and incubated at 4 °C overnight on a rotating device. Pellets were obtained by centrifugation, washed four times in phosphate-buffered saline (PBS; 9.1 mM dibasic sodium phosphate, 1.7 mM monobasic sodium phosphate, and 150 mM sodium chloride, pH 7.4) and reconstituted in 50 ␮l of PBS. The SRF immunoprecipitate (25 ␮l) or the negative control was subjected to SDS-PAGE on a 10% gel. Proteins were transferred to polyvinylidene difluoride membrane in a tank transfer system with a buffer (25 mM Tris-HCl, pH 8.3, 0.192 M glycine, 20% methanol), at 4 °C overnight. Membranes were blocked with 10% nonfat milk in PBST (PBS, Tween 20, 0.5%). Anti-HDAC4 (H-92) or anti-Gal4 (RK5C1) polyclonal antibody (1:500) (Santa Cruz Biotechnology) was used as the primary antibody. The immunoblot analysis was carried out with the appropriate secondary antibody (1:2000) coupled to horseradish peroxidase. An enhanced chemiluminescence kit (Amersham Biosciences) was used for the signal detection. The blots were routinely stripped and reprobed with the SRF antibody to ensure expression of the SRF protein.

In Vitro Protein-Protein Interaction Assays GST Pull-down Assay—GST fusion proteins were expressed in bacteria and purified as described previously (9). Five micrograms of GST or GST-SRF-FL bound to glutathione-agarose beads was incubated with [35S]Myc-HDAC4 in a protein interaction buffer of the following composition: 20 mM HEPES, pH 7.5, 75 mM KCl, 1 mM EDTA, 2 mM MgCl2, 2 mM dithiothreitol, and 0.1% Nonidet P-40. (Note that to assay the role of Ca2⫹ on the binding of SRF and HDAC4, the basal protein interaction buffer was altered by the addition of increasing concentrations of Ca2⫹ (0 – 4 mM) or EGTA (10 mM).) The binding reaction was conducted for 2 h at room temperature on a rocking platform. The beads were then pelleted at 7500 rpm for 3 min, washed three times with 1⫻ protein interaction buffer, and suspended in Laemmli buffer (Bio-Rad), and proteins bound to beads were resolved on SDS-PAGE and detected by autoradiography. In Vitro Co-immunoprecipitation (Myc Pull-down Assay)—Equal amounts (10-␮l aliquots) of [35S]SRF-FL and [35S]Myc-HDAC4 were incubated in a co-immunoprecipitation buffer (50 mM Tris, pH 7.4, 75


20049

FIG. 1. HDAC4 represses SRF-mediated gene transcription. A schematic representation of promoter/reporter constructs, skeletal-␣-actinluciferase, and 5⫻ SRE-luciferase is shown above the bar diagram. Subconfluent COS cells were grown in a six-well plate and transfected with 600 ng of the reporter plasmid together with 200 ng of expression plasmids for full-length SRF (pCGNSRF-FL), HDAC4 (pCDN3HDAC4-myc), or empty vector, either alone or in combination as shown below each bar diagram. In each case, a ␤-galactosidase expression plasmid was included to normalize for transfection efficiency. For each reporter plasmid, luciferase activity (mean ⫾ S.E.) is derived from 4 –7 different transfection experiments. mM NaCl, 5 mM EDTA, and 0.1% Nonidet P-40) for 1 h at room temperature. A negative control included a binding reaction with [35S]MycHDAC4 and either unprogrammed TNT lysate or 35S-labeled luciferase. One microgram of c-Myc antibody (mouse monoclonal; Santa Cruz Biotechnology) was added to each binding reaction, and the reaction was further allowed to incubate for 1 h at room temperature. Proteinagarose beads (10 ␮l) were then added, and reactions were continued at room temperature with continuous rocking for an additional 1 h. The beads were pelleted and washed three times with 1⫻ co-immunoprecipitation buffer by repeated centrifugation at 7,500 rpm for 3 min. Beads were then resuspended in 20 ␮l of Laemmli sample buffer, denatured by boiling for 3 min, and run on a 10% SDS-PAGE gel. Full-length protein synthesis was confirmed on a separate SDS-PAGE gel. For mapping the domain of SRF required for its interaction with HDAC4, reactions with the various 35S-labeled fragments of SRF and [35S]Myc-HDAC4 were subjected to Myc pull-down assays as described above and visualized by autoradiography. In Vitro CaMK Phosphorylation of GST-SRF or Myc-HDAC4 —GSTSRF, Myc-HDAC4, or GST alone (1 ␮g) was incubated in a kinase reaction buffer (50 mM HEPES, pH 7.4, 10 mM MgCl2, 100 ␮M ATP, 5 mM CaCl2, 30 ␮g/ml calmodulin, and 0.6 ng/␮l CaMK-II) in a 50-␮l reaction volume. [␥-32P]ATP was added to specified reaction tubes at 0.1 ␮Ci/reaction. Reactions were allowed to proceed at room temperature for 1 h. A 10-␮l aliquot of the kinase reaction was subjected to SDSPAGE to ensure phosphorylation of the proteins. A positive control included a known peptide substrate, autocamtide-2 (Biomol, Plymoth Meeting, PA). Subsequent GST pull-down assays were conducted with stated combinations of phosphorylated/unphosphorylated GST-SRF and phosphorylated/unphosphorylated Myc-HDAC4 as described above. Beads were pelleted, washed thoroughly, and resuspended in 50 ␮l of Laemmli’s sample buffer. Protein bound to beads were resolved on 10% SDS-PAGE and subsequently analyzed by autoradiography.

Electromobility Gel Shift Assay Double-stranded oligonucleotides were 5⬘-end-labeled with T4 polynucleotide kinase (Invitrogen) and [␥-32P]ATP. The binding reaction was carried out in a total volume of 25 ␮l containing ⬃10,000 cpm (0.1– 0.5 ng) of the labeled DNA, 2–5 ␮g of the specified nuclear extract, and 1 ␮g of poly(dI-dC). The binding buffer consisted of 10 mM Tris-HCl (pH 7.4), 100 mM NaCl, 0.1 mM EGTA, 0.5 mM dithiothreitol, 0.3 mM MgCl2, 8% glycerol, and 0.5 mM phenylmethylsulfonyl fluoride. After incubation at room temperature for 20 min, the reaction mixtures were loaded on 5% native polyacrylamide gels (44:1, acrylamide/bisacrylam-

ide), and electrophoresis was carried out at 150 V in a 0.5⫻ TBE buffer in a cold room. For competition and antibody experiments, unlabeled competitor DNA or the antibody was preincubated with nuclear extracts at room temperature for 15–20 min in the reaction buffers prior to the addition of the labeled DNA probe. Probe was human ␣-cardiac actin CArG (sense): 5⬘-AAG GGG ACC AAA TAA GGC AAG GTG G-3⬘.

Immunostaining and Confocal Microscopy Myocytes grown on 2% gelatin-coated coverslips and COS cells grown on two-well chamber slides (Nalge Nunc International) were transfected with specified DNA and subjected to the following treatments. Cells were washed with PBS, fixed with ice-cold methanol for 5 min, rehydrated with 5% Triton-X, and permeabilized with 0.02% Nonidet P-40. Cells, shielded from light, were blocked with a 0.3% bovine serum albumin solution and incubated with specified primary antibody at room temperature for 2 h. For localization of overexpressed MycHDAC4, we utilized the fluorescein-conjugated c-Myc (9E10) FITC antibody (Santa Cruz Biotechnology). For localization of overexpressed full-length HA-SRF, we utilized the rhodamine-conjugated HA (F-7) antibody (HA (F-7) TRITC; Santa Cruz Biotechnology). After three rapid washes, cells were mounted using the Slow-Fade Antifade kit with DAPI (Molecular Probes, Inc., Eugene, OR). Cells were observed using a ⫻63 Planapo objective and photographed using a Zeiss Axioplan microscope equipped with a PXL cooled charged coupled device camera (Roper, Tucson, AZ). Cellular detail in each field was obtained by Nomarski (differential interference contrast) imaging. For color florescence in each field, nuclear DAPI staining was visualized using 330 – 380-nm excitation and 460 – 470-nm emission filters; FITC staining was visualized by 493–509-nm excitation and 510 –550-nm emission filters; and rhodamine staining was visualized using a 556 –580-nm excitation filter and 600 – 660-nm emission filters. Single images were processed by no neighbor deconvolution, using Openlab 3 (Improvision, Coventry, UK) with digital confocal processing. All imaging was carried out in the Cancer Center Digital Light Microscopy Laboratory at the University of Chicago. RESULTS

HDAC4 Repressed the Transcription Activity of SRF—To test the hypothesis that a functional relationship exists between SRF and HDAC4, cells were co-transfected with various combinations of expression plasmids encoding either full-length SRF or HDAC4 and the skeletal-␣-actin reporter plasmid, con-

20050


FIG. 2. HDAC4 represses SRF through its enzymatic activity as well as by its ability to antagonize the activity of p300. A, C2C12 cells were transfected with 600 ng of the 5⫻ SRE-luciferase reporter plasmid and 200 ng of expression plasmids encoding SRF-FL, HDAC4, or mutant HDAC4 (HDAC-D840N) either alone or in combination as shown below each bar diagram. B, 40 h after transfection, a selected group of cells was also treated with 330 nM TSA. C, cell lysates obtained from different transfectants were analyzed by Western analysis using an anti-HDAC4 antibody. D, cells were transfected with different amounts of pCMV-p300 expression vector either alone or with increasing amounts (50 and 100 ng) of HDAC4 or HDAC-D840N expression vectors. In each experiment, a ␤-galactosidase expression plasmid was included to normalize for transfection efficiency. Bars indicate normalized reporter plasmid activity (mean ⫾ S.E.) derived from 5– 8 different transfection experiments.

taining ⫺394/⫹24 bp promoter fragment linked to the luciferase gene. The promoter region of skeletal ␣-actin gene used here contained two CArG boxes and several other adjacent cis-regulatory elements (Fig. 1). Overexpression of HDAC4 reduced (⬎80%) the basal activity of the skeletal ␣-actin gene in COS cells as well as in differentiated C2C12 cells, whereas SRF overexpression activated (4-fold) expression of the reporter gene in both cell types. When SRF was co-expressed with HDAC4, the trans-activation effect of SRF was markedly suppressed (⬎75%) (Fig. 1). We then used an artificial promoter/ reporter construct, which has five repeated and adjacent sequences of the CArG box motifs to determine whether this HDAC4-dependent repression was mediated through SRF binding to the CArG box sequence. The basal activity of this construct was also repressed (⬎80%) by overexpression of HDAC4 (Fig. 1), whereas HDAC4 abrogated SRF-mediated activation of the reporter gene activity. However, HDAC4 did not inhibit reporter activity of the plasmid in which CArG sequences were either mutated or deleted (not shown), thus indicating that HDAC4 repressed SRF-dependent transcription mediated through intact CArG box sequences. All members of the histone deacetylase superfamily contain a conserved Asp840 residue. Mutation of this residue (D840N)

and of the analogous residue in members of class I HDACs (e.g. Asp176 in HDAC1) has been shown to eliminate the catalytic activity of the enzyme (35, 39). To determine whether the deacetylase activity of HDAC4 was responsible for the repression of SRF function, we examined the effect of a catalytic defective mutant of HDAC4 (D840N) on SRF-induced gene expression. As shown in Fig. 2A, HDAC4 with mutation of the Asp840 residue (D840N) also reduced (50%) SRF-mediated gene transcription, albeit to an extent lesser than what resulted from the wild-type HDAC4. This finding indicated that additional mechanism(s) apart from the catalytic activity of HDAC4 might be involved in repressing the SRF transcription activity. We then examined the effect of a specific histone deacetylase inhibitor, trichostatin A (TSA), which caused a near maximal in vitro effect at 330 nM concentrations as used in the present study (40). Cells overexpressing SRF-FL with or without expression of HDAC4 were treated with TSA (330 nM) for 48 h. In these cells, overexpression of HDAC4 was verified by Western analysis, and results showed that the level of HDAC4 was not changed by the TSA treatment (Fig. 2C), whereas TSA partially reduced the ability of HDAC4 to repress SRF-mediated gene transcription (40% as opposed to 75% in the absence of TSA). In addition, basal SRF-dependent gene transcription was


20051

FIG. 3. Characterization of in vivo HDAC4/SRF interaction. A and B, 500 ␮g of nuclear extract obtained from cultured cardiac myocytes (A, lane 1), neonatal rat heart (A, lane 2), or COS cells (B) was precleared with 0.5 ␮g of normal IgG and then subjected to immunoprecipitation (IP) with 20 ␮g of anti-SRF antibody conjugated with agarose beads. Beads were separated from supernatant (supnt.) by centrifugation, washed, and subsequently analyzed by the Western blot analysis using anti-HDAC4 antibody. In B, a band below the position of the HDAC4 band is nonspecific. C, COS cells were co-transfected with plasmids encoding full-length SRF and Gal4-tagged HDAC4 or HDAC4 with a mutation (D840N) in the catalytic domain (as shown in the right panel). The next day following transfection, cell extract was prepared and subjected to immunoprecipitation (IP) with an anti-SRF antibody. The resultant beads were analyzed by Western analysis using an anti-Gal4 antibody. No-IP, the extract not subjected to immunoprecipitation.

significantly enhanced (10-fold in COS cells and 30-fold in C2C12 cells) by TSA treatment of cells, consistent with the hypothesis that SRF activity is negatively controlled by histone deacetylases. These results indicate that deacetylase activity of HDAC4 alone may not be the sole mechanism contributing to repression of the SRF transcription activity. HDACs may control transcription activity by competing with activators exhibiting HAT activities, such as p300/CREB-binding protein (29). To test whether the negative effect of HDAC4 on SRF function was, in part, resulting from interference with the activity of HAT activator p300, we examined the effect of HDAC4 on p300-mediated gene expression. COS cells were co-transfected with increasing concentrations of expression plasmids encoding HDAC4 and/or p300 together with the reporter plasmid having multiple SREs. As shown in Fig. 2D, overexpression of p300 activated the reporter gene activity by almost 7-fold. Co-expression of HDAC4 (wild type as well as the mutant lacking the deacetylase activity) resulted in a dose-dependent repression of the reporter gene activity by p300. Thus, HDAC4 repressed the SRF transcription activity in part by antagonizing the activity of the HAT activator, p300. SRF Binds to HDAC4 —p300 has been shown to bind to SRF in living cells (26). To determine whether HDAC4 also binds to SRF, cardiac and COS cell extracts were subjected to immunoprecipitation with an anti-SRF antibody, conjugated with agarose beads. Extract treated with IgG was utilized as negative control. The resultant beads were repeatedly washed and subsequently analyzed by the Western analysis using an antiHDAC4 antibody. For cardiac samples, we used two different extract preparations, one from the cultured cardiac myocytes and the other from neonatal rat heart. As shown in Fig. 3A, HDAC4 was pulled down by the anti-SRF antibody from both cardiac extract preparations but not by the control IgG. Similarly, HDAC4 was also co-immunoprecipitated with SRF from COS cell extracts (Fig. 3B). These results indicated that SRF

and HDAC4 in their native forms bind to each other in vivo. To further strengthen this observation, we overexpressed SRF in COS cells together with Gal-tagged HDAC4 or HDAC4 having a mutation in the catalytic domain (Gal-HDAC-D840N). Cell lysates were prepared and subjected to immunoprecipitation with the anti-SRF antibody. The resultant beads were analyzed by the Western analysis using an anti-Gal4 antibody. As shown in Fig. 3C, both wild type and mutant Gal-tagged HDAC4 were co-immunoprecipitated with SRF and again not with IgG alone. Thus, these data indicated that both native and recombinant HDAC4 associated with SRF in living cells and that this interaction was independent of the catalytic activity of HDAC4. Since SRF and HDAC4 could associate in vivo by either direct or indirect means, we asked if these two proteins physically co-associate. First, we synthesized in vitro [35S]MycHDAC4, [35S]SRF-FL, and [35S]luciferase in a TNT-coupled rabbit reticulocyte lysate system. The synthesis of full-length proteins was verified by SDS-PAGE, and only samples having ⬎90% full-length proteins were utilized for further analysis. [35S]Myc-HDAC4 was incubated with either [35S]SRF-FL or [35S]luciferase in protein-binding buffer for 2 h at room temperature, followed by immunoprecipitation with a monoclonal anti-c-Myc antibody conjugated with agarose beads. Beads were pelleted and washed, and proteins bound to beads were resolved on SDS-PAGE. In this assay, both the antibody target protein (Myc-HDAC4) and the partner protein (SRF) could be visualized on the same gel. By this approach, we found that Myc-HDAC4 was capable of pulling down SRF but not luciferase, which served here as a negative control (Fig. 4B). Second, we examined the ability of GST-SRF to pull down radiolabeled [35S]Myc-HDAC4. The glutathione beads prebound with GST or GST-SRF-FL were incubated with [35S]Myc-HDAC4 for 2 h at room temperature. By this experiment, we also found that [35S]Myc-HDAC4 directly interacted with GST-SRF-FL but not with the GST control (Fig. 4A). Together, these results demon-

20052


FIG. 4. Characterization of direct physical interaction between SRF and HDAC4 and mapping the region of SRF necessary for HDAC4 binding. A, in the rabbit reticulocyte lysate (R.R.) Myc-HDAC4 constructs was transcribed/translated with [35S]methionine, and 20% input of labeled proteins was resolved by SDS-PAGE. The remaining portion was divided into two equal halves and incubated with either GST-SRF or GST-linked glutathione-agarose beads. Proteins bound to beads were analyzed by SDS-PAGE. B, equal amounts of 35S-labeled SRF fragments were incubated with [35S]Myc-HDAC4 in protein binding buffer for 1 h at room temperature. In a separate binding reaction [35S]luciferase was included as a negative control. Following protein binding, the reaction mix was subjected to immunoprecipitation with a monoclonal anti-c-Myc antibody conjugated with protein-agarose beads. The beads were pelleted, washed, and analyzed by SDS-PAGE. C, schematic representation of different SRF constructs analyzed and summary of their interaction ability with Myc-HDAC4. The position of the MADS box containing the DNA-binding and dimerization domain of SRF is depicted by different shades.

strated direct physical interaction between SRF and HDAC4, both under in vitro and in vivo assay conditions. Mapping the Domain of SRF That Mediates Its Interaction with HDAC4 —To define the region of SRF bound to HDAC4, in vitro synthesized truncated 35S-labeled SRF fragments were incubated with equal amounts of [35S]Myc-HDAC4 and examined by the Myc co-immunoprecipitation assay as described above. The amino-terminal fragment of SRF (1–336 amino acids) that includes its MADS box (DNA binding/dimerization region) bound with Myc-HDAC4 to the same degree as fulllength SRF, whereas the carboxyl terminal fragment of SRF (337–508 amino acids) failed to bind, thus suggesting that the interacting domain of SRF with HDAC4 is within the first 336 amino acids of the protein. A deletion mutant of SRF encompassing 141–244 amino acids, which includes its MADS box region, was retained with Myc-HDAC4 to the same degree as full-length SRF, indicating that this region of SRF is sufficient for physical interaction with HDAC4. In order to determine whether the entire MADS box is involved in this interaction, we examined an additional SRF (245–508 amino acids) fragment, which includes the COOH-terminal region of the MADS box together with the entire trans-activation domain of SRF. As shown in Fig. 4B, this fragment also interacted weakly with

HDAC4 as compared with the full-length SRF and the SRF amino acid 141–242 fragment (as measured by scanning densitometry and expressed as a percentage of input). In addition, the SRF 1–204 fragment that contains the amino terminus region of the MADS domain failed completely to bind to HDAC4. From these results, we conclude that amino acids within 205–336 positions of SRF are capable of binding to HDAC4. These combined results also demonstrate that whereas the MADS domain of SRF is sufficient for interaction, it is the COOH-terminal part of the MADS domain that is obligatory for its direct physical association with HDAC4. HDAC4 Is Part of SRF Complex on C(Ar)G Box Sequences— To begin to explore the mechanism of HDAC4-mediated repression of SRF transcription, we asked whether HDAC4 could be recruited as part of the SRF complex on the CArG box of muscle promoters. To test this possibility, COS cells were transfected with three different isoforms of SRF: SRF-FL, SRF-M (deletion of exon 5), and SRF-S (deletion of exons 4 and 5) (22) either alone or together with HDAC4. Cell lysate was prepared and analyzed by mobility gel shift analysis using human cardiac-␣-actin CArG box sequences as a labeled probe. Unlabeled probes as well as antibodies against SRF and HDAC4 were used to demonstrate the presence and specificity of complexes. As shown in Fig. 5, cell lysate overexpress-


20053

FIG. 5. HDAC4 facilitates SRF binding to C(Ar)G box sequences. COS cells were transfected with expression plasmids encoding different isoforms of SRF (SRF-FL, SRF-M, and SRF-S) and HDAC4. Cell nuclear extract was prepared and analyzed by the mobility gel shift analysis using human cardiac ␣-actin C(Ar)G box sequences as a probe. In A (lane 12) and B (last lane), nuclear extract was preincubated with 5 ␮l of antiHDAC4 and anti-SRF antibodies, respectively. SS, supershifted complex. A 100fold excess of unlabeled oligonucleotide (same as probe) was used as a competitor.

FIG. 6. Binding of HDAC4 to SRF is regulated by calcium-dependent signaling. A, primary cultures of cardiac myocytes were treated with ionomycin (10 ␮M), angiotensin-II (100 ng/ml), or vehicle alone for 4 days. Cells were harvested 4 days following treatment when hypertrophy of myocytes became apparent. Cell lysate was prepared and subjected to immunoprecipitation (IP) with anti-SRF antibody or control rabbit IgG. Beads (B) and supernatants (S) were separated and analyzed by Western analysis using an anti-HDAC4 antibody. B, cells were harvested at different days following ionomycin treatment, and SRF/HDAC4 interaction was analyzed by the same immunoprecipitation method as in A. The HDAC4 activity was determined by densitometry scanning of bands. C, [35S]HDAC4 was in vitro synthesized and incubated with GST-SRF for 1 h in binding buffer containing different composition of CaCl2 (2– 4 mM) in the absence or presence of EGTA (10 mM) as shown above the gel. Binding of HDAC4 to SRF was analyzed by the GST pull-down assay.

ing SRF-FL produced two specific complexes, C-1 and C-2, of different mobilities. Overexpression of HDAC4 together with SRF increased the formation of slower migrating complex (C-1 complex), perhaps by forming multimeric complexes as demonstrated by reduction of the faster migrating C-2 complex. As expected, cell lysate with overexpressed SRF-M and SRF-S isoforms produced different mobility C-1 complexes. Again, co-expression of HADC4 increased the intensity of the respective C-1 complexes with no further change in their gel mobility but decreased the intensity of the C-2 complex. Furthermore, inclusion of antibodies against HDAC4 and SRF abolished and/or supershifted the C-1 complex (Fig. 5, A (lane 12) and B (last lane)); however, no change was observed with preimmune serums (not shown). These results indicated that

HDAC4 may actually facilitate formation of the SRF complex to the cardiac-␣-actin promoter, consistent with a previous report showing HDAC4 being part of the MEF2 complex on the muscle-specific promoters (42). Intracellular Ca2⫹ Levels and CaMK-dependent Phosphorylation Regulate SRF-HDAC4 Interaction—Since our results, thus far, indicated that the SRF-HDAC4 association is functionally relevant in vivo, we were interested further in identifying intracellular signals that might regulate interaction of these two proteins. To demonstrate whether hypertrophic stimuli, whose major second messenger is Ca2⫹ and/or Ca2⫹-mediated signaling events, could affect the SRF/HDAC4 interaction, we treated primary cultures of rat cardiac myocytes with a

20054


FIG. 7. Ca2ⴙ/CaMK-dependent phosphorylation of HDAC4 disrupts the SRF/HDAC4 co-association. Cold HDAC4 and 35S-labeled HDAC4 (lane 1) were synthesized in vitro. Cold HDAC4 and GST-SRF were subjected to in vitro phosphorylation by incubation in a Ca2⫹/ calmodulin kinase reaction buffer containing CaMK-IV and [␥-32P]ATP. The phosphorylation status of GST-SRF (lane 5) and HDAC4 (lane 8) was determined by SDS-PAGE. 35S-Labeled (but unphosphorylated) HDAC4 (lanes 2– 4) or 32P-phosphorylated HDAC4 (lanes 6 and 7) was incubated with either GST-SRF or GST-[32P]SRF in protein binding buffer for 1 h, and the interaction ability of proteins was determined by the GST pull-down assay. Note that phosphorylation of HDAC4 inhibits its binding to SRF regardless of the phosphorylation status of SRF.

FIG. 8. CaMK-IV abrogates HDAC4mediated repression of SRF transcription activity. COS cells grown in a six-well plate were transfected with 600 ng of 5⫻ SRE-luciferase reporter plasmid and various combinations of expression plasmids as shown below each bar diagram. full-length SRF and HDAC4 plasmids were used at 200 ng/well. CaMK-IV expression plasmid was used at three different amounts, 200, 400, and 800 ng. CaMK-IV(dead), CaMK-II, and CaMKII(dead) plasmids were used in 400-ng amount. In each transfection, the ␤-galactosidase expression plasmid was included to normalize for transfection efficiency. Bars indicate normalized luciferase activity (mean ⫾ S.E.) derived from 5– 8 different transfection experiments and presented as percentage of the activity obtained from SRF alone.

Ca2⫹ ionophore (ionomycin) or angiotensin-II. Cells were harvested at different time points of treatments, and cell lysate was prepared and subjected to co-immunoprecipitation analysis to detect SRF-HDAC4 association. In most of the cultures, hypertrophy of myocytes became apparent after 3 days of treatment. The results from co-immunoprecipitation assays showed that the HDAC4/SRF association was appreciably reduced after 24 h of treatment with ionomycin and completely lost by the 3rd and 4th day of treatment when hypertrophy of myocytes became obvious. Similar results were obtained by angiotensin-II treatment of cells; however, no change was observed in vehicle-treated controls (Fig. 6, A and B). These results indicated that hypertrophic agonist-mediated increase in intracellular calcium and/or downstream signaling events causing hypertrophy disrupted the SRF-HDAC4 interaction. To check whether Ca2⫹ directly interfered with SRF-HDAC4 interaction, we examined direct effect of Ca2⫹ on SRF/HDAC4 association in an in vitro protein-binding assay. Equal amounts of [35S]HDAC4 and GST-SRF-FL were incubated in a binding buffer with increasing concentrations of Ca2⫹, ranging from 0

to 4 mM, followed by analysis of protein binding by the GST pull-down assay. The results showed that the physical interaction of the two proteins was progressively lost as the concentration of calcium gradually increased, with no detectable binding at 4 mM Ca2⫹. Furthermore, the addition of EGTA (10 mM) in the binding buffer with high levels of Ca2⫹ partially restored the association of SRF with HDAC4 (Fig. 6C). A possible explanation for this partial restoration of the SRF/HDAC4 association is that the concentration of EGTA used in the assay caused additional chelation of other divalent ions needed for protein binding. These results demonstrated that Ca2⫹ directly interferes with SRF/HDAC4 association in vitro, consistent with an earlier report showing loss of HDAC4-MEF2 binding in the presence of high levels of Ca2⫹ (43). Since such a high level of Ca2⫹ (4 mM) may not exist in a living cell, we asked whether Ca2⫹–activated downstream signals could be involved in regulation of SRF-HDAC4 association. One of the most significant downstream events of increased intracellular Ca2⫹ levels is the activation of CaMKs and phospho-


20055

FIG. 9. Subcellular localization of exogenously expressed HDAC4 in COS cells. COS cells maintained in 10% fetal bovine serum were transfected with Myc-HDAC4 alone (i), Myc-HDAC4 plus HA-SRF (ii), or Myc-HDAC4 and HA-SRF plus active CaMK-IV (iii). Forty-eight hours post-transfection, cells were fixed and stained with FITC-Myc antibody (green) (a) and counterstained with DAPI (blue) for nuclear staining (b). The same field was also visualized by “differential interference contrast” imaging (d). The overlays of images from each field were created with Open-lab3 (Improvision) (c). Whereas Myc-HDAC4 appeared to be pancellular under basal conditions (i-a), it translocated to the nucleus in cells overexpressing HA-SRF (ii-a). The inset in (ii-a) depicts the same field stained with TRITC-HA antibody (red) representing nuclear localization of exogenous HA-SRF. Note that in cells co-transfected with active CaMK-IV (iii), Myc-HDAC4 appeared to relocate to the cytoplasm.

rylation of their substrates. Both SRF and HDAC4 have been previously shown to be substrates for CaMK-dependent phosphorylation (44, 45). It has also been documented that CaMK signaling induces hypertrophy in cultured cardiac myocytes as well as in transgenic mice overexpressing a constitutively active form of CaMK-IV (46). To test whether CaMK-dependent phosphorylation of SRF and/or HDAC4 affects their binding abilities, we subjected both proteins to in vitro phosphorylation. Myc-HDAC4, GST-SRF, and GST were incubated separately in a Ca2⫹/ calmodulin kinase reaction buffer in the presence or absence of active CaMK-IV and [␥-32P]ATP. The phosphorylation status of each protein was determined by SDS-PAGE. Both GST-SRF and HDAC4 were highly phosphorylated, but no phosphorylation of GST was detected, thus serving as negative control. The effect of protein phosphorylation on their interaction ability was determined by the GST pull-down assay. As shown in Fig. 7, unphosphorylated HDAC4 interacted strongly with GST-SRF (lanes 3 and 4), but phosphorylated HDAC4 failed to bind to either phosphorylated or unphosphorylated GST-SRF (Fig. 7, lanes 6 and 7). These results showed that CaMK-dependent phosphorylation of HDAC4 abrogated its ability to bind to SRF, irrespective of SRF phosphorylation status. CaMK-IV Signaling Reversed HDAC-mediated Repression of SRF Transcription Activity—The disruption of SRF-HDAC4 complex upon CaMK-dependent phosphorylation may activate

the SRF transcription activity by removal of HDAC4-mediated repression. We tested this hypothesis by doing transient transfection studies, in which cells were co-transfected with various combinations of expression plasmids encoding CaMK-II, CaMK-IV, HDAC4, and/or SRF together with a reporter plasmid having multiple SRF binding sites. As shown in Fig. 8, overexpression of active CaMK-IV removed the HDAC4-mediated repression of the SRF transcriptional activity in a concentration-dependent manner, with complete restoration of SRF function at an 800-ng amount of the plasmid; however, the inactivated CaMK-IV had no effect. Furthermore and surprisingly, the active form of CaMK-II had no effect on HDAC4mediated repression of the SRF activity, thus suggesting a specific effect of CaMK-IV on the removal of HDAC4-mediated repression. How did CaMK-IV mediate reversal of HDAC4-mediated repression? To address this question, we examined whether activated CaMK-IV influences the intracellular localization of HDAC4. As shown in Fig. 9, cells overexpressing HDAC4 revealed a pancellular distribution of HDAC4, which appeared to be micropunctated due to association with as yet unidentified subcellular organelles, consistent with earlier reports of distribution of HDAC4 in mitotic cells (32). Immunostaining of COS cells overexpressing both SRF and HDAC4 indicated a substantial increase in nuclear localization of HDAC4 (Fig. 9).

20056


FIG. 10. Subcellular localization of exogenously expressed HDAC4 in primary neonatal rat cardiac myocytes. Myocytes were isolated as described under “Materials and Methods” and grown on gelatin-coated coverslips. Cells were co-transfected with Myc-HDAC4 and HA-SRF with (ii) or without (i) active CaMK-IV. Forty-eight hours post-transfection, cells were fixed and stained with FITC-Myc antibody (green) (a) or TRITC-HA antibody (red) (i-b) and counterstained with DAPI (blue) for nuclear staining. The same field was also visualized by differential interference contrast imaging (d), and overlays of images from each field were created with Openlab 3 (Improvision) as presented in ii-c. Myc-HDAC4 and HA-SRF were strictly nuclear in cardiac myocytes maintained under basal conditions (i-a, i-b); however, overexpression of CaMK-IV caused cytoplasmic relocation of HDAC4 in these cells (ii-a and ii-c).

Interestingly, although SRF itself appears to be predominantly nuclear in localization, a large population of proliferating COS cells showed cytoplasmic expression of SRF as well (data not shown). In contrast, immunostaining of cardiac myocytes also expressing ectopic SRF as well as HDAC4 revealed strictly nuclear localization of both proteins (Fig. 10). However, when cells overexpressing SRF and HDAC4 were further co-transfected with constitutively active CaMKs, both in COS cells and in cardiac myocytes, CaMK-IV, but not CaMK-II, resulted in relocation of HDAC4 to the cytoplasm (Figs. 9 and 10). This observation is consistent with our data from transcription assays, in which CaMK-IV, but not CaMK-II, was able to relieve HDAC4-mediated repression of the SRF transcription activity. Collectively, these results indicate that phosphorylation of HDAC4 by CaMK-IV disrupts the HDAC4䡠SRF complex, leading to the export of HDAC4 from the nucleus and derepression of SRF transcription activity. DISCUSSION

Our studies described here identify SRF as a specific and functional target of a class II histone deacetylase, HDAC4. We demonstrate that SRF physically interacts with HDAC4 in cardiac myocytes as well as in nonmuscle cells, leading to the localization of HDAC4 to the cell nucleus. The functional consequence of HDAC4/SRF interaction is profound repression of

SRF-mediated gene transcription. We also show that this protein interaction is sensitive to enhanced Ca2⫹ levels and to activation of Ca2⫹/CaMK signaling by hypertrophic agonist stimulation. Activation of Ca2⫹/CaMK-dependent signaling disrupts SRF-HDAC4 interaction, leading to redistribution of HDAC4 to cytoplasm and unmasking the SRF gene activation function. This type of restoration of SRF function by hypertrophic stimuli could explain why certain SRF-regulated genes, such as reversal of the “fetal gene program,” become reactivated during cardiac hypertrophy. HDAC4 Represses SRF Transcriptional Potential—Histone deacetylases could repress gene transcription by either deacetylation of the NH2-terminal tail of core histones, which results in chromatin condensation, or as part of co-repressor complexes on the promoters of genes by competing with other transcription factors and/or interfering with basal transcription machinery (35, 47). Several lines of evidence obtained from both in vitro and in vivo experiments demonstrate a direct role of HDAC4 in repression of the SRF transcription activity. First, HDAC4 directly associates with SRF independently of SRF binding to DNA; second, CArG box sequences are sufficient to mediate a profound repression effect of HDAC4; and third, HDAC4 represses SRF-mediated gene expression regardless of cell type (i.e. muscle or nonmuscle cell background). HDAC4,

HDAC4 Regulates the Transcription Activity of SRF like other members of the class II family of HDACs, contains two distinct functional domains at each terminus of the protein. The carboxyl terminus domain is homologous to the yeast HDA-1 and has catalytic activity (33), whereas the amino terminus domain has been implicated in mediating protein-protein interaction with different transcription factors, including MyoD, the MEF2 family of factors, and the chaperone protein 14-3-3 (47– 49). Since HDAC4 does not bind to DNA directly, it is these two characteristics of the protein (deacetylase and protein binding) to which its repression function is probably attributable. Our results, obtained by using a catalytically inactive mutant of HDAC4, showed that it was slightly less active than the wild type HDAC4. Moreover, HDAC4 was able to repress SRF function to a substantial degree in the presence of a specific histone deacetylase inhibitor, TSA, indicating that the deacetylase activity, although playing a role, was not the sole mechanism by which HDAC4 inhibited SRF-dependent gene transcription. In this context, however, it should be noted that the enzyme-dead mutant of HDAC4 retains the ability to bind to other endogenous histone deacetylases of class I and II and that their involvement in repressing SRF function cannot be disregarded. Several other independent studies have also demonstrated that additional mechanisms aside from the inherent deacetylase activity of HDAC4 are operative in repressing MEF2-dependent transcription as well as in regulation of skeletal muscle cell differentiation (42, 49). Another mechanism by which HDAC4 could repress SRF function is by precluding binding of co-activators to SRF. Experiments conducted to map the HDAC4-interacting domain of SRF demonstrated that the COOH-terminal portion of MADS box of SRF is required for its binding to HDAC4. The HAT activator, p300, has also been previously shown to bind to the MADS box and to activate the transcription activity of MEF2 as well as SRF (26, 29). In our experiments, both the wild type and the enzyme-dead mutant of HDAC4 canceled the gene activation induced by p300, and it thus seems likely that the negative effect of HDAC4 is partly due to competition between HDAC4 and p300 to bind to SRF. Previously, p300 has been shown to compete with HDACs in binding to MADS domain of MEF2 (43). Another possible explanation for repressed SRF function is that the association of SRF with HDAC4 in the absence of DNA results in a conformational change in SRF that increases its DNA binding activity (as shown in Fig. 5) but consequently decreases its transcription potency. Although these possibilities need to be formally tested, there are examples where the altered DNA binding activity of SRF has been accounted for by its change in transcriptional potential (17). Through immunostaining of proteins, we have demonstrated intranuclear co-localization of HDAC4 and SRF in cultured primary cardiac myocytes. We are not aware of any previous reports of HDAC4 localization in cardiac myocytes, although its nuclear localization in differentiated skeletal muscle cells (C2C12 cells) has been previously demonstrated, where it serves to maintain the cells in differentiated state (42, 49). Likewise, we believe that nuclear localization of HDAC4 plays an important role in maintaining the differentiated state of cardiac myocytes. In mitotic COS cells, however, we found that ectopic HDAC4 has pancellular distribution. Cells maintained under similar conditions, but overexpressing SRF together with HDAC4, revealed a predominantly nuclear localization of HDAC4, indicating that SRF promotes import of HDAC4 to the cell nucleus. This observation is similar to a previous report of MEF2 capable of mediating the nuclear localization of HDAC4 (48, 50). Thus, it appears that SRF, similar to MEF2, is capable of importing one of its own repressors into the cell nucleus. In the case of cardiac myocytes, it is possible that high endoge-

20057

FIG. 11. A model for the role of HDAC4 in regulation of SRF function. Under basal conditions, HDAC4 is recruited as part of the SRF complex on CArG box sequences of the target gene, resulting in repression of gene transcription. Stimulation of cells with hypertrophic agonists causes activation of CaMK-dependent signaling, which leads to phosphorylation of HDAC4. Phosphorylated HDAC4 fails to bind to SRF, leading to export of HDAC4 from the nucleus and thus allowing HAT activator p300 to bind to SRF complex, and that results in activation of SRF-dependent gene transcription.

nous levels of SRF are sufficient to mediate nuclear entry of ectopic HDAC4 when expressed alone. Alternatively, HDAC4 is retained in the nucleus of differentiated cardiac myocytes as a result of an as yet unidentified phosphorylation event. Recently, activation of the Ras-MAPK pathway was shown sufficient to cause nuclear localization of HDAC4 in C2C12 myoblast cells (51). Hypertrophic Signaling Dissociates the HDAC4䡠SRF Complex—Ca2⫹/CaMK-signaling is known to play an important role in transcription regulation of a large number of cardiac genes as well as in the induction and progression of hypertrophy (46). Given the role of SRF in the activation of cardiac muscle gene expression during hypertrophy, it was not surprising that association of SRF with a repressor, HDAC4, was regulated by Ca2⫹/CaMK signaling. SRF-dependent transcription of the cfos gene promoter was previously shown to be activated by enhanced intracellular levels of Ca2⫹ that requires CaMK-dependent signaling but does not involve activation of the Ets family of tertiary complex factors (52). Both CaMK-IV and CaMK-II are shown capable of phosphorylating SRF at Ser103, within seconds of rise in intracellular Ca2⫹ levels (44). However, this particular phosphorylation event alone could not explain an increase in the transcriptional activity of SRF in response to enhanced levels of calcium. Evidence from our study suggests that additional events such as phosphorylation of SRF partner proteins are involved in the increased transcriptional potency of SRF. One such partner protein of SRF is HDAC4. We found that phosphorylation of HDAC4 but not SRF, by either CaMK-IV or CaMK-II, was capable of disrupting the HDAC4/SRF association. Also in transfection assays, the constitutively active form of CaMK-IV was capable of overcoming HDAC4-mediated repression of the SRF transcriptional activity. Consistent with these results, we also demonstrated cytoplasmic relocalization of HDAC4 in response to CAMK-IV activity both in cardiac myocytes as well as in COS cells. Interestingly, however, the constitutively active CaMK-II had no effect in vivo on HDAC4-mediated repression or its intracellular localization. A possible explanation for the lack of the effect of CaMK-II in vivo could be its inability to enter the cell nucleus (53), whereas CaMK-IV was found to be predominantly a nuclear isoform (54). It is also likely that the two isoforms of

20058


CaMK phosphorylate different sites on SRF and/or HDAC4, with distinct net effects on SRF-mediated transcription. For instance, it has been demonstrated that CaMK-II and CaMK-IV both phosphorylate CREB at Ser133; however, whereas CaMK-IV stimulates its transcription activity, CaMK-II does not. This was attributed to blocking the effect of Ser133 phosphorylation of CREB through phosphorylation of an additional site, Ser142, which is phosphorylated by CaMK-II but not by CaMK-IV (41). How does dissociation of HDAC4 activate SRF transcription activity? It is conceivable that the release of HDAC4 from SRF in response to CaMK signaling depends on or is accompanied by displacement of another SRF-binding factor that is CaMKsensitive. The transcription co-activator p300, which is activated by Ca/CaMK, might be recruited to CArG box sequences in response to CaMK signaling (43). Since p300 possesses acetyltransferase activity, it may also acetylase SRF, leading to change in its DNA and/or protein binding activity and, hence, activation of its gene transcription potential. Earlier reports have suggested that Ca2⫹/CaMK signaling is a common mechanism to relieve transcriptional repression imposed by interactions of HDACs with transcription factors. For instance, the MEF2/HDAC4/5 interaction was also shown to be disrupted by increased levels of Ca2⫹ and CaMK-dependent phosphorylation, resulting in cytoplasmic relocation of the HDAC䡠14-3-3 complex and derepression of the transcription activity of MEF2 (30, 43). Since multiple Ca2⫹-dependent pathways are known to be involved during hypertrophic response of cardiac myocytes, it remains to be determined which is the rate-limiting pathway contributing to the association/localization of HDAC4䡠SRF complex and subsequent activation of SRF-dependent cardiac muscle gene expression. Identification of such a rate-limiting pathway may open new therapeutic avenues to alter cardiac gene expression in a hypertrophied heart. A simple model of CaMK-dependent activation of SRF function by dissociation from HDAC4 is shown in Fig 11. In summary, we identified a role of class II histone deacetylase HDAC4 in SRF-mediated gene transcription. We believe that during basal conditions, SRF-dependent transcription is under the negative control of histone deacetylases. The SRF䡠HDAC4 complex is a downstream target of Ca2⫹/CaMK signaling for “TCF-independent” regulation of SRF transcription function. Since SRF plays a versatile role in gene regulation, both in proliferating and terminally differentiated cells, the data presented here also provide clues for future studies to address the question whether SRF/HDAC4 association is a mechanistic link between calcium/CaMK signaling and chromatin remodeling. Acknowledgments—We thank Dr. T. Kouzarides for HDAC4 plasmids, Drs. D. L. Black and M. R. Rosner for CaMK plasmids, and Dr. D. M. Livingston for p300 expression vector. We also thank Dr. Rene Arcilla for critically reading the manuscript. REFERENCES 1. Greenberg, M. E., Siegfried, Z., and Ziff, E. B. (1987) Mol. Cell. Biol. 7, 1217–1225 2. Boxer, L. M., Prywes, R., Roeder, R. G., and Kedes, L. (1989) Mol. Cell. Biol. 9, 515–522 3. Treisman, R. (1994) Curr. Opin. Gen. Dev. 4, 96 –101 4. Arsenian, S., Weinhold, B., Oelgeschlager, M., Ruther, U., and Nordheim, A. (1998) EMBO J. 17, 6289 – 6299 5. Belaguli, N. S., Schildmeyer, L. A., and Schwartz, R. J. (1997) J. Biol. Chem. 272, 18222–18231 6. Belaguli, N. S., Sepulveda, J. L., Nigam, V., Charron, F., Nemer, M., and Schwartz, R. J. (2000) Mol. Cell. Biol. 20, 7550 –7558 7. Chen, C. Y., and Schwartz, R. J. (1996) Mol. Cell. Biol. 16, 6372– 6384 8. Gupta, M., Kogut, P., Davis, F. J., Belaguli, N. S., Schwartz, R. J., and Gupta, M. P. (2001) J. Biol. Chem. 276, 10413–10422 9. Wang, D., Chang, P. S., Wang, Z., Sutherland, L., Richardson, J. A., Small, E.,

Krieg, P. A., and Olson, E. N. (2001) Cell 105, 851– 862 10. Chang, D. F., Belaguli, N. S., Iyer, D., Roberts, W. B., Wu, S. P., Dong, X. R., Marx, J. G., Moore, M. S., Beckerle, M. C., Majesky, M. W., and Schwartz, R. J. (2003) Dev. Cell. 4, 107–118 11. Cheng, G., Hagen, T. P., Dawson, M. L., Barnes, K. V., and Menick, D. R. (1999) J. Biol. Chem. 274, 12819 –12826 12. Moore, M. L., Wang, G. L., Belaguli, N. S., Schwartz, R. J., and McMillin, J. B. (2001) J. Biol. Chem. 276, 1026 –1033 13. Paradis, P., MacLellan, W. R., Belaguli, N. S., Schwartz, R. J., and Schneider, M. D. (1996) J. Biol. Chem. 271, 10827–10833 14. Thuerauf, D. J., Arnold, N. D., Zechner, D., Hanford, D. S., DeMartin, K. M. McDonough, P. M., Prywes, R., and Glembotski, C. C. (1998) J. Biol. Chem. 273, 20636 –20643 15. Bristow, M. R. (1998) Lancet 352, Suppl. 1, 8 –14 16. Zhang, X., Azhar, G., Chai, J., Sheridan, P., Nagano, K., Brown, T., Yang, J., Khrapko, K., Borras, A. M., Lawitts, J., Misra, R. P., and Wei, J. Y. (2001) Am. J. Physiol. 280, H1782–H1792 17. Hines, W. A., Thorburn, J., and Thorburn, A. (1992) Mol. Cell. Biol. 19, 1841–1852 18. Janknecht, R., Hipskind, R. A., Houthaeve, T., Nordheim, A., and Stunnenberg, H. G. (1992) EMBO J. 11, 1045–1054 19. Camoretti-Mercado, B., Liu, H. W., Halayko, A. J., Forsythe, S. M., Kyle, J. W., Li, B., Fu, Y., McConville, J., Kogut, P., Vieira, J. E., Patel, N. M., Hershenson, M. B., Fuchs, E., Sinha, S., Miano, J. M., Parmacek, M. S., Burkhardt, J. K., and Solway, J. (2000) J. Biol. Chem. 275, 30387–30393 20. Kemp, P. R., and Metcalfe, J. C. (2000) Biochem. J. 345, 445– 451 21. Belaguli, N. S., Zhou, W., Trinh, T. T., Majesky, M. W., and Schwartz, R. J. (1999) Mol. Cell. Biol. 19, 4582– 4591 22. Davis, F. J., Gupta, M., Pogwizd, S. M., Bacha, E., Jeevanandam, V., and Gupta, M. P. (2001) Am. J. Physiol. 282, H1521–H1533 23. Imhof, A., Yang, X. J., Ogryzko, V. V., Nakatani, Y., Wolffe, A. P., and Ge, H. (1997) Curr. Opin. Biol. 7, 689 – 692 24. Manabe, I., and Owens, G. K. (2001) Circ. Res. 88, 1127–1134 25. Qiu, P., and Li, L. (2002) Circ. Res. 90, 858 – 865 26. Ramirez, S., Ait-Si-Ali, S., Robin, P., Trouche, D., and Harel-Bellan, A. (1997) J. Biol. Chem. 272, 31016 –31021 27. Alberts, A. S., Geneste, O., and Treisman, R. (1998) Cell 92, 475– 487 28. Hamamori, Y., Sartorelli, V., Ogryzko, V., Puri, P. L., Wu, H. Y., Wang, J. Y., Nakatani, Y., and Kedes, L. (1999) Cell 96, 405– 413 29. Sartorelli, V., Huang, J., Hamamori, Y., and Kedes, L. (1997) Mol. Cell. Biol. 17, 1010 –1026 30. Lu, J., McKinsey, T. A., Nicol, R. L., and Olson, E. N. (2000) Proc. Natl. Acad. Sci. U. S. A. 97, 4070 – 4075 31. Miska, E. A., Langley, E., Wolf, D., Karlsson, C., Pines, J., and Kouzarides, T. (2001) Nucleic Acids Res. 29, 3439 –3447 32. Grozinger, C. M., Hassig, C. A., and Schreiber, S. L. (1999) Proc. Natl. Acad. Sci. U. S. A. 96, 4868 – 4873 33. Wang, A. H., Bertos, N. R., Vezmar, M., Pelletier, N., Crosato, M., Heng, H. H., Th’ng, J., Han, J., and Yang, X. J. (1999) Mol. Cell. Biol. 19, 7816 –7827 34. Nemoto, S., Sheng, Z., and Lin, A. Mol. Cell. Biol. 18, 3518 –3526 35. Miska, E. A., Karlsson, C., Langley, E., Nielsen, S. J., Pines, J., and Kouzarides, T. (1999) EMBO J. 18, 5099 –5107 36. Jiuyong, X., and Black, D. L. (2001) Nature 410, 936 –939 37. Waldmann, R., Hanson, P. I., and Schulman, H. (1990) Biochemistry 29, 1679 –1684 38. Eckner, R., Ewen, M. E., Newsome, D., Gerdes, M., DeCaprio, J. A., Lawrence, J. B., and Livingston, D. M. (1994) Genes Dev. 8, 869 – 884 39. Hassig, C. A., Tong, J. K., Fleischer, T. C., Owa, T., Grable, P. G., Ayer, D. E., and Schreiber, S. L. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 3519 –3524 40. Marks, P. A., Richon, V. M., Breslow, R., Rifkind, R. A. (2001) Curr. Opin. Oncol. 13, 477– 483 41. Sun, P., Enslen, H., Myung, P. S., and Maurer, R. A. (1994) Genes Dev. 8, 2527–2539 42. Lu, J., McKinsey, T. A., Zhang, C. L., and Olson, E. N. (2000) Mol. Cell. 6, 233–244 43. Youn, H. D., Grozinger, C. M., and Liu, J. O. (2000) J. Biol. Chem. 275, 22563–22567 44. Misra, R. P., Bonni, A., Miranti, C. K., Rivera, V. M., Sheng, M., and Greenberg, M. E. (1994) J. Biol. Chem. 269, 25483–25493 45. Zhao, X., Ito, A., Kane, C. D., Liao, T. S., Bolger, T. A., Lemrow, S. M., Means, A. R., and Yao, T. P. (2001) J. Biol. Chem. 276, 35042–35048 46. Passier, R., Zeng, H., Frey, N., Naya, F. J., Nicol, R. L., McKinsey, T. A., Overbeek, P. Richardson, J. A., Grant, S. R., and Olson, E. N. (2000) J. Clin. Invest. 105, 1395–1406 47. Fischle, W., Kiermer, V., Dequiedt, F., and Verdin, E. (2001) Biochem. Cell Biol. 79, 337–348 48. Grozinger, C. M., and Schreiber, S. L. (2000) Proc. Natl. Acad. Sci. U. S. A. 97, 7835–7840 49. Zhang, C. L., McKinsey, T. A., Lu, J. R., and Olson, E. N. (2001) J. Biol. Chem. 276, 35–39 50. Borghi, S., Molinari, S., Razzini, G., Parise, F., Battini, R., and Ferrari, S. (2001) J. Cell Sci. 114, 4477– 4483 51. Zhou, X., Richon, V. M., Wang, A. H., Yang, X. J., Rifkind, R. A., and Marks, P. A. (2000) Proc. Natl. Acad. Sci. U. S. A. 97, 14329 –14333 52. Miranti, C. K., Ginty, D. D., Huang, G., Chatila, T., and Greenberg, M. E. (1995) Mol. Cell. Biol. 15, 3672–3684 53. Srinivasan, M., Edman, C. F., and Schulman, H. (1994) J. Cell Biol. 126, 839 – 852 54. Jensen, K. F., Ohmstede, C., Fisher, R. S., and Sahyoun, N. (1991) Proc. Natl. Acad. Sci. U. S. A. 88, 2850 –2853