Histone Deacetylase 3 Coordinates Deacetylase-independent Epigenetic Silencing of Transforming Growth Factor-β1 (TGF-β1) to Orchestrate Second Heart Field Development*

Background: Histone-modifying genes play critical roles in the pathogenesis of human congenital heart disease. Results: HDAC3 recruits PRC2 complex to mediate epigenetic silencing of TGF-β1 in a deacetylase-independent manner within second heart field progenitor cells. Conclusion: HDAC3-mediated epigenetic silencing of TGF-β1 is required for normal heart development. Significance: This is the first report of HDAC-mediated epigenetic regulation of second heart field development. About two-thirds of human congenital heart disease involves second heart field-derived structures. Histone-modifying enzymes, histone deacetylases (HDACs), regulate the epigenome; however, their functions within the second heart field remain elusive. Here we demonstrate that histone deacetylase 3 (HDAC3) orchestrates epigenetic silencing of Tgf-β1, a causative factor in congenital heart disease pathogenesis, in a deacetylase-independent manner to regulate development of second heart field-derived structures. In murine embryos lacking HDAC3 in the second heart field, increased TGF-β1 bioavailability is associated with ascending aortic dilatation, outflow tract malrotation, overriding aorta, double outlet right ventricle, aberrant semilunar valve development, bicuspid aortic valve, ventricular septal defects, and embryonic lethality. Activation of TGF-β signaling causes aberrant endothelial-to-mesenchymal transition and altered extracellular matrix homeostasis in HDAC3-null outflow tracts and semilunar valves, and pharmacological inhibition of TGF-β rescues these defects. HDAC3 recruits components of the PRC2 complex, methyltransferase EZH2, EED, and SUZ12, to the NCOR complex to enrich trimethylation of Lys-27 on histone H3 at the Tgf-β1 regulatory region and thereby maintains epigenetic silencing of Tgf-β1 specifically within the second heart field-derived mesenchyme. Wild-type HDAC3 or catalytically inactive HDAC3 expression rescues aberrant endothelial-to-mesenchymal transition and epigenetic silencing of Tgf-β1 in HDAC3-null outflow tracts and semilunar valves. These findings reveal that epigenetic dysregulation within the second heart field is a predisposing factor for congenital heart disease.

(E7.0), 2 when a subset of cells derived from the antero-lateral mesoderm forms the cardiac crescent (2). This crescent contains the first and second heart field cardiac progenitor cells (3). The first heart field progenitor cells are the principal contributor to the primary heart tube, the left ventricle, atrioventricular canal, and atria (4). The second heart field progenitor cells extensively contribute to the outflow tract, semilunar valves, atria, right ventricle, primary atrial septum, and ventricular septum (5)(6)(7)(8)(9). Indeed, ablation or genetic manipulation of second heart field progenitor cells and their derivatives leads to outflow tract malrotation, overriding aorta, double outlet right ventricle, aberrant semilunar valve development, and ventricular septal defects (10 -12). In humans, about two-thirds of congenital cardiac defects involve the outflow tract, semilunar valves, or ventricular septum (13).
genital cardiovascular anomalies, such as aortic dilatation, outflow tract defects, bicuspid aortic valve, semilunar valve stenosis, and ventricular septal defects (30 -32). How the second heart field progenitor cells interpret elevated TGF-␤ at the chromatin level in these connective tissue disorders remains to be elucidated. Extracellular matrix tightly regulates the bioavailability of active TGF-␤ (33). For instance, extracellular matrix proteoglycans, such as decorin (DCN), biglycan (BGN), and fibromodulin (FMOD), sequester TGF-␤ to limit its activity (34). Mice lacking DCN, BGN, or FMOD display elevated TGF-␤ activity and phenotypes observed in Marfan syndrome or Ehlers-Danlos syndrome (35)(36)(37). Indeed, proteoglycan deficiencies have been demonstrated in patients with these syndromes (38). Proteoglycans expressed in the developing outflow tract and semilunar valves also play a key role in the assembly of collagen fibers in the extracellular matrix (37,39). Patients with collagen deficiency, such as COL3A1 mutations in Ehlers-Danlos syndrome, show elevated TGF-␤ levels (29). How extracellular matrix homeostasis is regulated remains an area of intense research.
The active form of TGF-␤ binds to TGFBR2 and TGFBR1, which in turn phosphorylate SMAD2/3 (R-SMADs) to promote assembly of heteromeric complex with SMAD4. Activated SMAD complex accumulates in the nucleus, where it recruits transcription co-factors and chromatin modifiers to regulate the expression of target genes (33). Several factors, in addition to TGF-␤, play critical roles to trigger activation and amplification of the intracellular TGF-␤ signaling pathway at multiple levels (33). For instance, the transcription factor SNAI1 interacts with SMAD3/4 to repress endothelial gene expression and thereby augments TGF-␤-mediated mesenchymal transition (40). EndMT, a specialized form of EMT, is the complex biological process in which endothelial cells trans-differentiate into mesenchymal cell types, including smooth muscle-like and fibroblast-like cells. EndMT has been implicated in several pathological processes, including fibrotic disorders and cardiac valvular diseases. A complex orchestration of several signaling pathways, including TGF-␤ signaling, initiates and promotes EndMT. However, the molecular and epigenetic mechanisms regulating termination of EndMT remain elusive.
Histone deacetylases (HDACs) are chromatin-modifying enzymes that regulate the epigenome (41). The mammalian HDACs are classified into five subfamilies based on their phylogenetic analysis and sequence homology. Class I HDACs (HDAC1, HDAC2, HDAC3, and HDAC8) play critical roles at various stages of development (42). For instance, global loss of HDAC1 or HDAC3 results in early embryonic lethality around E9.5 (42). Similarly, mice lacking HDAC2 or HDAC8 display lethality at birth (43)(44)(45). Our group and others, using gene inactivation studies in mice, have demonstrated vital functions for class I HDACs in cardiomyocyte proliferation, differentiation, and hypertrophy (43, 46 -49). However, functions of HDACs in second heart field development remain undefined.
HDACs lack intrinsic DNA-binding domains but are recruited to the chromatin via their interaction with transcription factors, co-factors, and large multiprotein transcriptional complexes (41). For instance, HDAC3 is an integral part of NCOR (nuclear receptor corepressor) or its homologue SMRT (silencing mediator of retinoic and thyroid receptors). Interaction of HDAC3 with the deacetylase-activating domain of NCOR/SMRT is required for its enzyme activity. Interestingly, recent evidence suggests that enzymatic activity of HDAC3 is dispensable, but its interaction with NCOR/SMRT is essential for transcriptional repression. However, molecular mechanisms that mediate the deacetylase-independent function of HDAC3 remain to be defined. A recent study shows that HDAC4 controls histone methylation in response to elevated cardiac overload, suggesting that HDACs may recruit histone methyltransferases to the chromatin.
In this study, we demonstrate a deacetylase-independent role for HDAC3 as an epigenetic silencer of Tgf-␤1 within the second heart field with direct implications for human TGF-␤ pathway-associated cardiovascular anomalies. As a part of the NCOR complex, HDAC3 recruits EZH2 (enhancer of zeste homologue 2), the major histone methyltransferase of PRC2 (polycomb repressor complex 2), to mediate epigenetic silencing of Tgf-␤1 specifically within the second heart field-derived mesenchymal cells and thereby promotes termination of EndMT. Genetic deletion of Hdac3 in the murine second heart field results in increased TGF-␤ bioavailability within mesenchymal cells, perpetual activation of mesenchymal cells, aberrant EndMT, and altered extracellular matrix homeostasis, observed in patients with semilunar valve pathologies. Together, these results uncover that epigenetic silencing mediated by HDAC3 in a deacetylase-independent manner orchestrates second heart field development, which may be a molecular target in human cardiovascular anomalies.
(horse anti-mouse/rabbit/goat IgG) (Vector Laboratories). Recombinant TGF-␤ was purchased from R&D Systems. Alcian blue, alkaline alcohol, orcein, alcoholic hematoxylin, ferric chloride, Lugol's iodine, woodstain scarlet acid fuchsin, phosphotungstic acid, saffron, Bouin's fixative, Weigert's iron hematoxylin A, Weigert's iron hematoxylin B, phosphomolybdic acid-phosphotungstic acid, aniline blue, and Van Gieson's solution were purchased from Electron Microscopy Sciences. Harris modified hematoxylin, eosin Y, ethanol, xylenes, glacial acetic acid, paraformaldehyde, paraffin, potassium ferricyanide, potassium ferrocyanide, and deoxycholic acid were purchased from Fisher. Polyethylenimine, linear, was purchased from Polysciences. X-gal was purchased from 5 Prime. Vectashield mounting medium, the Vectastain Elite ABC kit, and the DAB Peroxidase Substrate kit were purchased from Vector Laboratories. The RNeasy minikit and GST bead slurry were purchased from Qiagen. Power SYBR Green PCR Master Mix, Superscript first strand synthesis kit, TOPO-TA cloning kit, DMEM high glucose with sodium pyruvate, penicillin/ streptomycin, and horse serum were purchased from Invitrogen. The CellsDirect TM one-step quantitative RT-PCR kit, insulin-transferrin-selenium, Epoxy M-450 Dynabeads, and TRIzol were purchased from Life Technologies, Inc. Rat tail collagen type I was purchased from BD Biosciences. iScript reverse transcription supermix was purchased from Bio-Rad. The sandwich ELISA assay kit for TGF-␤1 was purchased from R&D Systems. The sandwich ELISA assay kit for phospho-SMAD2/3 was purchased from Cell Signaling. The Quik-Change II XL site-directed mutagenesis kit was purchased from Stratagene. Passive lysis buffer and the Dual-Luciferase reporter assay kit were purchased from Promega. Fetal bovine serum, donkey serum, gelatin, and magnetic anti-FLAG beads were purchased from Sigma. Agarose-IgG and IgA bead slurry were purchased from Santa Cruz Biotechnology and Life Technologies. The EZ-ChIP assay kit and HDAC assay kit were purchased from Millipore. The TaKaRa DNA ligation kit was purchased from Clontech.
Hematoxylin and Eosin Staining-Hematoxylin and eosin staining was performed by deparaffinizing sections in xylenes, rehydrating through an ethanol gradient, 30-s or 2-min stain with 30% or 100% Harris modified hematoxylin, and a 30-s counterstain with eosin Y. Slides were rinsed and dehydrated with ethanol, cleared with xylenes, and mounted with Vectashield mounting medium.
Movat's Pentachrome Staining-Movat's pentachrome staining was conducted by deparaffinizing and rehydrating slides, followed by a 20-min stain in Alcian blue, a 1-h differentiation in alkaline alcohol, a 20-min stain in Orcein-Verhoeff solution (Orcein, alcoholic hematoxylin, ferric chloride, and Lugol's iodine), a 2-min stain with woodstain scarlet acid fuchsin, a rinse in acetic acid, and a 10-min differentiation in 5% phosphotungstic acid, followed by a 15-min stain in saffron. Sections were dehydrated in ethanol, cleared in xylenes, and mounted with Vectashield mounting medium.
Masson's Trichrome Staining-Masson's trichrome staining was performed by deparaffinizing and rehydrating sections through an ethanol gradient followed by a 1-h mordant in Bouin's fixative at 56°C. Samples were then washed and stained for 5 min in a solution of Weigert's iron hematoxylin A and Weigert's iron hematoxylin B. Following washing, samples were differentiated in phosphomolybdic acid-phosphotungstic acid for 15 min, stained in aniline blue solution for 20 min, and differentiated in 1% acetic acid for 3 min. Samples were then dehydrated in ethanol, cleared in xylenes, and mounted with Vectashield mounting medium.
LacZ Staining-Tissue samples were dissected in PBS and then fixed in 2% paraformaldehyde for 30 min at 4°C. After washing in PBS at room temperature, samples were stained overnight in LacZ staining solution (5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 2 mM MgCl 2 , 0.01% deoxycholic acid, 0.04% Nonidet P-40, 0.1% X-gal, in 1ϫ PBS) at 37°C in the dark. Samples were then washed in PBS and fixed overnight in 4% paraformaldehyde.
Immunohistochemistry-For immunohistochemistry, sections were deparaffinized in xylenes and pretreated using heat antigen retrieval in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6). Immunohistochemistry was conducted using the Vectastain Elite ABC kit and DAB Peroxidase Substrate kit according to the manufacturers' guidelines. Sec-tions were incubated with cleaved caspase-3 antibody (1:100) for 1 h at room temperature or HDAC3 antibody (1:250) or phospho-SMAD2/3 antibody (1:50) overnight at 4°C. Biotinylated universal pan-specific antibody (horse anti-mouse/rabbit/goat IgG) was used for phospho-SMAD2/3 immunostaining in place of the Vectastain Elite ABC kit secondary antibody according to the manufacturer's guidelines. For counterstaining, slides were rinsed and then incubated with 30% hematoxylin for 30 s after 3,3Ј-diaminobenzidine developing. All slides were ethanol-dehydrated, cleared with xylenes, and mounted with Vectashield mounting medium. For immunofluorescent staining, sections were deparaffinized and rehydrated through xylenes and an ethanol gradient. Slides were rinsed in PBS, and antigen retrieval was performed in sodium citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0) for 10 min at 95°C. After rinsing, sections were blocked in 10% donkey serum, 0.3% Triton X-100 in PBS for 1 h at room temperature. Sections were then washed in PBS and incubated with smooth muscle actin antibody (1:100) in 10% donkey serum and PBS for 1 h at room temperature or with MF-20 (1:50) or troponin (1:25) antibodies in 10% donkey serum in PBS overnight at 4°C. Finally, slides were washed in PBS, incubated in secondary antibody (donkey anti-mouse 546, 1:500, with Hoechst, 1:1000, in 10% donkey serum, PBS) for 1 h at room temperature, rinsed in PBS, and mounted with Vectashield mounting medium.
Biometric Analysis-Aortic diameter was measured from ϫ2 images of dissected hearts at three levels (tubular aortic trunk, proximal aortic arch, and intermediate aortic arch) using NIS-Elements analysis software (Nikon). Ventricular wall thickness and valve area measurements were made from ϫ2 and ϫ10 images, respectively, of position-matched hematoxylin and eosin-stained sections using NIS-Elements analysis software (Nikon). Nuclei numbers per section in the valves were counted manually in ϫ10 hematoxylin and eosin-stained sections using ImageJ counting tools. Differences between groups were compared using Student's t tests.
Real-time Quantitative PCR-Total RNA was extracted and reverse transcribed using iScript reverse transcription supermix (Bio-Rad) or the CellAmp whole transcriptome amplification kit (Takara) according to the manufacturer's guidelines. Transcript expression was measured by quantitative RT-PCR using SYBR Green PCR Master Mix. Signals were normalized to corresponding GAPDH controls and represented as relative expression ratios of experimental samples relative to Hdac3 F/F controls. Primer sequences are available upon request.
Apoptosis Assay-Sections of aortic and pulmonic valves from five Hdac3 Isl1KO and five control E13.5 hearts were immunostained with cleaved caspase-3 antibody and counterstained with hematoxylin. Images of the aortic and pulmonic valves were taken at ϫ20 magnification. The total number of nuclei and the number of cleaved caspase-3 positive nuclei within the valve cusp were manually counted for each sample. The aortic and pulmonic valves were compared separately. The number of cleaved caspase-3-positive nuclei per 1000 nuclei served as an apoptosis index.
Microarray Analysis-Heart tissue was dissected from E9.5 mouse embryos and snap-frozen in liquid nitrogen. RNA was extracted from pools of Hdac3 Isl1KO or Hdac3 F/F control hearts using an RNeasy minikit. Microarray analysis was performed in triplicate from pooled samples by the University of Massachusetts Genomics Core Facility using Affymetrix Mouse Gene 2.0 ST arrays (Affymetrix). Raw microarray data were annotated using the Bioconductor and Oligo packages in R (53)(54)(55). Significance of expression differences between Hdac3 Isl1KO and Hdac3 F/F samples was determined using Student's t test. Heat maps of microarray data were generated using the pheatmap package in R (56). Relative expression of each transcript is reported as log-transformed expression value for each sample, normalized to the median expression value of the transcript across all six samples. MIAME-compliant full microarray data sets can be accessed at the Gene Expression Omnibus (GEO, GSE73666).
Ingenuity Pathway Analysis-Microarray data from E9.5 Hdac3 Isl1KO and control hearts were analyzed using Ingenuity Pathway Analysis (IPA). The IPA Diseases and Functions utility was employed to investigate phenotypes associated with the molecular changes. Hierarchical heat maps were generated using both the Diseases and Biological Functions categories and Tox Functions categories of the Diseases and Functions utility. Plots of phenotypically relevant categories were constructed based on subcategory p values. The Upstream Regulators utility was employed to determine potential regulators of differentially expressed genes. Significant regulators were sorted by significance and by number of associated genes. Differentially expressed genes associated with the upstream regulator TGF-␤1 were exported, and a clustered heat map was generated using R software.
Lentiviral Infection-Lentiviral medium was generated by transfecting 100-mm plates of subconfluent 293T cells with 5 g of lentiviral cDNA, 5 g of pCMV-dR8.2, and 2.5 g of pCMV-VSVG, in 10 ml of 2% FBS medium. The medium was changed to a fresh 10 ml of 2% FBS medium 24 h after transfection. Viral medium was collected 24 h later and filtered through a 40-m cell strainer. Isolated mouse embryonic heart tissue was infected with filtered viral medium supplemented with 10 g/ml Polybrene reagent. GFP viral medium was used to maintain a constant viral medium volume. Infected cells were harvested for analysis 24 or 48 h after infection.
Cell Culture, Transient Transfection, and Luciferase Assay-HEK293T cells and murine endothelial cells were maintained in DMEM with 10% FBS, 100 mg/ml penicillin, and 100 mM/ml streptomycin in a 37°C incubator with 5% CO 2 . HEK293T cells were transfected in subconfluent 100-mm plates with 2.5 g of DNA and 5 l of polyethylenimine, linear, in 10 ml of 10% FBS medium. Luciferase assays were conducted by transfecting subconfluent murine endothelial cells in 6-well plates with 1 g of DNA and 2 l of polyethylenimine, linear, in 2 ml of 10% FBS medium. The DNA amount was maintained constant using pcDNA3.1(Ϫ) or pLJM1-EGFP DNA. Cells were lysed with passive lysis buffer 16 h after transfection, and lysates were analyzed using the Dual-Luciferase reporter assay kit according to the manufacturer's guidelines. Luciferase activity was measured using a Berthold microplate reader according to the manufacturer's guidelines.
Immunoprecipitation-Tissue samples were homogenized in immunoprecipitation buffer (50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.5% Nonidet P-40, 1 mM EDTA, and 1 mM DTT) containing 1 mM PMSF, phosphate inhibitors (Sigma), and protease inhibitor mixture (Sigma). The homogenized samples were sonicated using a Branson 250 Digital Sonifier with 1-s on and 1-s off pulses at 40% power amplitude for 15 s. Precleared lysates with beads were incubated with primary antibodies for 16 h at 4°C. After incubation for 1 h at 4°C with beads, immune complexes were collected, washed four times with immunoprecipitation buffer, and applied to 4 -12% SDS-polyacrylamide gels for Western blot analysis.
ELISA-Sandwich ELISAs were performed according to the manufacturer's protocol. Briefly, for the phospho-SMAD2/3 or phospho-HDAC3 ELISA, samples were prepared using 100 l of 1ϫ cell lysis buffer (20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM sodium pyrophosphate, 1 mM ␤-glycerophosphate, 1 mM Na 3 VO 4 , 1 g/ml leupeptin). 100 l of sample diluent was added per 100 l of sample (1:1 ratio). Samples were incubated overnight at 4°C. After four washes with 1ϫ wash buffer, samples were incubated with detection antibody for 1 h at 37°C. Samples were visualized using the HRP detection method at 450 nm. For the phospho-HDAC3 ELISA, a microtiter plate was coated with HDAC3 antibody (5 g/ml) in carbonate/bicarbonate buffer (pH 9.6). The sandwich ELISA for TGF-␤1 was performed according to the manufacturer's protocol. Briefly, samples were prepared using assay diluent buffer RD1-21. Samples were incubated for 2 h at room temperature. After four washes with 1ϫ wash buffer, samples were incubated with 100 l of TGF-␤1 conjugate for 2 h at room temperature. After four washes with 1ϫ wash buffer, samples were incubated with 100 l of substrate solution (protected from light) for 30 min at room temperature. After the addition of 100 l of stop solution, reactions were visualized at 450 and 540 nm. Readings at 540 nm were subtracted from readings at 450 nm to correct for optical imperfections in the plate.
Outflow Tract Explant Assay-The outflow tract explant cultures were performed as described previously (57,58). Briefly, a solution of rat tail collagen type I (1.5 mg/ml) containing NaOH to a final concentration of 15 mmol/liter was dispensed into 24-well microculture dishes. Subsequently, gels were placed in a 37°C tissue culture incubator at 5% CO 2 and allowed to polymerize. After 30 min, collagen gels were washed several times with DMEM containing 10% fetal bovine serum, 0.1% insulin-transferrin-selenium, 100 mg/ml penicillin, and 100 mM/ml streptomycin. Outflow tracts were carefully dissected from control and Hdac3 Isl1KO E9.5, E10.5, or E14.5 embryonic hearts. The outflow tracts were placed endocardium face down onto collagen gels and allowed to adhere for 8 h at 37°C in 5% CO 2 . Eight hours after attachment, medium was added, and explants were cultured for up to 24, 48, or 72 h. Radial migration was measured from ϫ4 images of outflow tract explants at 45°intervals using NIS-Elements analysis software (Nikon). Explants were treated with TGF-␤ antibody (1 g/ml) for 24 or 48 h or with recombinant TGF-␤ (10 ng/ml) for 24 or 48 h. Explants were also detached, trypsinized, and incubated with anti-vimentin (mesenchymal marker) antibody-conjugated or PECAM-1 (endothelial marker)-conjugated epoxy M-450 Dynabeads as per the manufacturer's guidelines to isolate cells.
HDAC Activity Assay-HDAC activity was measured by using an HDAC activity assay kit for fluorometric detection as per the manufacturer's protocol. Briefly, 293T cells were transfected with or without plasmid containing either HDAC3-

HDAC3 Orchestrates Second Heart Field Development
FLAG or HDAC3 H134A,H135A -FLAG. Using anti-FLAG antibody, total lysate was immunoprecipitated and then incubated with HDAC assay substrate for 60 min at 37°C. After the addition of diluted activator reagent, samples were incubated at room temperature for 10 min and analyzed by a fluorescent microplate reader at excitation of 360 nm and emission of 450 nm.
Statistical Analysis-Statistical significance between groups was determined using two-tailed Student's t test or 2 test. A p value of Ͻ0.05 was considered significant.
The mature semilunar valves are composed of extracellular matrix, valvular interstitial cells, and endothelial cells. Remodeling of valvular interstitial cells and extracellular matrix, mainly collagens, proteoglycans, and elastin, is essential for maturation of the semilunar valves (61). Hdac3 Isl1KO hearts displayed loss of extracellular matrix trilaminar stratification and valvular interstitial cell compartmentalization in the semilunar valves at various stages of development (Fig. 3, A-F, L, and M). Specifically, loss of HDAC3 resulted in aberrant expression of collagen and proteoglycan in all layers, probably explaining enlarged and thickened semilunar valves (Fig. 3, A-F, L, and M). Elastin fibers appeared disorganized and fragmented in Hdac3 Isl1KO semilunar valves (Fig. 3, A, B, E, F, L, and M). Similarly, Hdac3 Mef2CKO embryos revealed aberrant expression of collagen and proteoglycan in all layers of the semilunar valves (Fig. 3, G and H). Remodeling of the semilunar valves is associated with an increase in apoptosis (18). Hdac3 Isl1KO semilunar valve interstitial cells displayed a 55-65% decrease in apoptosis compared with controls (p Ͻ 0.02) during valve remodeling (Fig. 3, I-K).

TGF-␤ Mediates Aberrant EndMT and Extracellular Matrix Remodeling in Hdac3 Isl1KO
Heart-TGF-␤ signaling and its downstream target genes, such as Snai1, are critical factors that induce EndMT in the developing outflow tract cushions (40). EndMT is characterized by loss of endothelial markers, such as TEK, and gain of mesenchymal markers, such as TAGLN (33). Hdac3 Isl1KO hearts exhibited robust down-regulation of Tek and up-regulation of Snai1 and Tagln (Figs. 4F and 5A). Hdac3 Isl1KO outflow tract cushion explants showed ϳ3-fold induction of EndMT (p Ͻ 0.005; Fig. 5, B and C), and these changes were largely abolished by TGF-␤-neutralizing antibody (p Ͻ 0.009; Fig. 5, B-F). Aberrant TGF-␤ signaling activates valvular interstitial cells, which in turn, express smooth muscle genes and produce extracellular matrix scaffold proteins to remodel valvular matrix (24). Hdac3 Isl1KO hearts revealed significant expression changes in smooth muscle and extracellular matrix genes (Fig. 5, A and G-M). Smooth muscle genes, including Tagln, Myh11, and Cnn2, were up-regulated in Hdac3 Isl1KO hearts and semilunar valves (Fig. 5, A, G, and H). Simultaneously, expression of proteoglycans, including Dcn, Bgn, Fmod, and Vcan, and collagens, including Col2a1, Col11a2, Col9a2, Col1a2, and Col3a1, were significantly altered in Hdac3 Isl1KO hearts and semilunar valves (Fig. 5, A and  I-M). We next examined the requirement of TGF-␤ for aberrant expression of extracellular matrix genes in Hdac3 Isl1KO valvular interstitial cells. Loss of HDAC3 resulted in 6 -10-fold expression changes in proteoglycans, collagen, and smooth muscle genes (p Ͻ 0.006), and these changes were largely abol-

HDAC3 Orchestrates Second Heart Field Development
ished by TGF-␤-neutralizing antibody (Fig. 5, H-M). Taken together, these results suggest that aberrant EndMT and extracellular matrix remodeling in Hdac3 Isl1KO hearts are mediated by TGF-␤ signaling.

HDAC3 Epigenetically Silences TGF-␤1 within Mesenchymal Cells by Recruiting EZH2 to the NCOR Complex-During
EndMT, endothelial cells lose their endothelium-specific markers and morphology and acquire a mesenchymal cell-like phenotype (33). In this process, endothelium-derived mesenchymal cells displayed ϳ90% repression of TGF-␤1 expression compared with endothelial cells (p Ͻ 0.003); however, this repression was completely absent in mesenchymal cells lacking HDAC3 (Fig. 8, A and B). We observed ϳ0.6 -1.8% enrichment of HDAC3 at a regulatory region upstream of TGF-␤1 within mesenchymal cells compared with endothelial cells (p Ͻ 0.02; Fig. 8, C and D). We evaluated chromatin occupancy of various histone marks to determine the regulatory mechanism of the HDAC3-enriched region upstream of TGF-␤1 within mesenchymal cells. ChIP-qPCR analysis revealed loss of trimethylation of Lys-27 on histone H3 (H3K27me3) and ϳ8% enrichment of acetylation of Lys-27 on histone H3 (H3K27ac) at the TGF-␤1 regulatory region in Hdac3 Isl1KO semilunar valves compared with control (p Ͻ 0.04, Fig. 8, E and F). Consistent with these findings, Hdac3 Isl1KO semilunar valves displayed ϳ0.9 -3% enrichment of RNA polymerase II and CREBBP (p Ͻ 0.006) and abolished occupancy of EZH2, EED, and SUZ12, the H3K27 methyltransferase components of PRC2 complex, at the TGF-␤1 regulatory region (Fig. 8, G-K). NCOR1 was required for recruitment of HDAC3 to the TGF-␤1 locus (Fig. 8L). Interestingly, NCOR1 occupancy remained unchanged at this locus in Hdac3 Isl1KO semilunar valves, suggesting that HDAC3 is not required for NCOR1 recruitment to the TGF-␤1 regulatory region (Fig. 8M). Co-ChIP-qPCR analysis revealed co-occupancy of HDAC3 with either EZH2, H3K27me3, or NCOR1 at the TGF-␤1 locus in wild-type semilunar valves (Fig. 8N). HDAC3 interacts with EZH2 within mesenchymal cells of semilunar valves (Fig. 8O). Importantly, EZH2 and NCOR1 are required for deposition of the H3K27me3 mark at the TGF-␤1 locus within valvular mesenchymal cells (Fig. 8P). These data suggest that HDAC3 is required to mediate epigenetic silencing of TGF-␤1 within mesenchymal cells of semilunar valves.

Discussion
Previous studies from our group and others have demonstrated that HDACs are critical regulators of various developmental processes, including cardiogenesis (48,49). Among class I HDACs (HDAC1, -2, -3, and -8), global loss of HDAC2 or HDAC8 in mice does not affect morphogenesis of second heart field-derived structures at birth (43,45,46). Interestingly, murine embryos lacking HDAC1 within the second heart field (Hdac1 Isl1KO ) display normal cardiogenesis, including development of the outflow tract and semilunar valves. 3 Similarly, second heart field-derived structures appear normal in mice lacking class II HDACs, such as HDAC4, -5, -6, or -9 at birth (42). Taken together, our data reveal a unique and a specific role for HDAC3 in regulation of second heart field morphogenesis.
Second heart field progenitor cells progressively and restrictively differentiate into various cardiac cell types, including cardiomyocytes, endothelial cells, and mesenchymal or smooth muscle cells (3). Genetic ablation of HDAC3 in the second heart field progenitor cells using either Isl1-Cre or Mef2c-AHF-Cre resulted in ascending aortic dilatation, outflow tract malrotation, overriding aorta, double outlet right ventricle, semilunar valve stenosis, bicuspid aortic valve, and membranous ventric-ular septal defects that closely resembled those seen in patients with congenital cardiovascular diseases. Strikingly, genetic deletion of HDAC3 in mesenchymal or smooth muscle cells (Hdac3 TaglnKO ) recapitulated the cardiovascular anomalies observed in Hdac3 Isl1KO and Hdac3 Mef2CKO embryos; however, genetic deletion of HDAC3 in differentiated cardiomyocytes (Hdac3 Myh6KO ) or endothelial cells (Hdac3 Cdh5KO ) did not recapitulate these defects. These observations define a spatiotemporal function of HDAC3 within mesenchymal or smooth muscle cells of second heart field-derived structures that is required for arterial pole morphogenesis (Fig. 10, A and B).
The anterior part of the second heart field contributes to the arterial pole of the heart, myocardial, and endothelial layers of the outflow tract, right ventricle, ventricular septum, and endocardial cushion mesenchyme of the outflow tract. The posterior part of the second heart field contributes to the venous pole of the heart, atria, dorsal mesenchymal protrusion, and atrial septum (4). Isl1 Cre marks both the anterior and posterior part of the second heart field (3). However, genetic deletion of ubiquitously expressed HDAC3 in the Isl1 Cre expression domain only affected morphogenesis of the anterior heart field-derived structures. Indeed, embryos lacking HDAC3 in the anterior second heart field (Mef2C-AHF-Cre) recapitulated these defects. Thus, HDAC3 mainly functions within the anterior part of the second heart field. Interestingly, embryos lacking HDAC3 in the second heart field display normal ventricular myocardium (data not shown). Previously, we demonstrated that HDAC3 is required in the primary heart field (Nkx2-5 enhancer-Cre) for development of the ventricular myocardium at the early stages of cardiogenesis (48). Taken together, these findings suggest that once the primary heart field progenitors have adopted a cardiac fate to form the nascent heart tube, the second heart field-derived ventricular myocardium can develop independently of HDAC3. This is consistent with prior evidence that the ventricular myocardium remains unaffected after genetic manipulations or ablation of the second heart field progenitor cells (10,11).
Outflow tract malrotation in tandem with ventricular septal defect is one of the most common cyanotic congenital heart defects at birth (13). Proper alignment, orientation, and septation of the cardiac outflow tract into the aorta and the pulmonary artery require intricate coordination and interaction among multiple cell types, including neural crest and second heart field (68). Cardiac neural crest cell-dependent processes, such as outflow tract septation and distal outflow tract morphogenesis, appeared normal in embryos lacking HDAC3 in the second heart field (Fig. 1G). Previously, we demonstrated that neural crest-specific genetic deletion of HDAC3 resulted in anomalous outflow tract septation and distal outflow tract mor-   NOVEMBER 6, 2015 • VOLUME 290 • NUMBER 45 JOURNAL OF BIOLOGICAL CHEMISTRY 27081 phogenesis; however, outflow tract rotation and semilunar valve development were unaffected (69). Instead, second heart field-derived myocardium at the base of the outflow tract governs the normal outflow tract rotation (70). Prior studies also suggest that aberrant EndMT and remodeling of the outflow tract cushions give rise to structural outflow tract defects, particularly outflow tract malrotation (68). Our data establish that loss of HDAC3 triggers EndMT in the outflow tract cushion that can be markedly reduced by inhibiting TGF-␤. Indeed, TGF-␤ and its downstream target genes are critical regulators of EndMT (17). In particular, TGF-␤-activated SMAD proteins directly stimulate the expression of zinc finger transcription factor Snai1 within the outflow tract cushion (40). Studies in chicken and mice have shown that transcriptional repressor Snai1 is both sufficient and required for EndMT (71). Consistent with these findings, embryos lacking HDAC3 in the second heart field or valvular mesenchymal cells display a marked increase in Snai1 expression within the developing outflow tract and semilunar valves. Overall, our data demonstrate that genetic deletion of HDAC3 in the second heart field or in valvular mesenchymal cells augments TGF-␤ bioavailability, which, in turn, promotes aberrant EndMT and remodeling of the outflow tract cushion, and this, in part, may modulate outflow tract rotation. Importantly, our results highlight the underlying fact that outflow tract rotation and septation are independent processes regulated by distinct second heart field and cardiac neural crest lineages.

HDAC3 Orchestrates Second Heart Field Development
Our data reveal that HDAC3 is a critical regulator of extracellular matrix homeostasis within the second heart field-derived structures. Aberrant extracellular matrix is often an underlying pathology in semilunar valve stenosis, bicuspid aortic valve, and aortic dilatation (19,38,72). Patients with connective tissue disorders often exhibit altered composition and distribution of extracellular matrix within thickened semilunar valves and ascending aorta, resembling those seen in Hdac3 Isl1KO and Hdac3 Mef2CKO embryos. We demonstrate that HDAC3 is required to maintain proper composition of both proteoglycans and collagen in the developing semilunar valves and ascending aorta. The small leucine-rich proteoglycans, such as biglycan, decorin, and fibromodulin, sequester TGF-␤ to limit its bioavailability and thereby regulate proper remodeling of the hinge region of the semilunar valves and ascending aorta (37,39). Without this sequestration, an increase of TGF-␤ bioavailability manifests as dilatation of ascending aorta, bicuspid aortic valves, and thickened semilunar valves in connective tissue disorders. Consistent with this model, our data demonstrate that genetic deletion of HDAC3 in the second heart field or valvular mesenchymal cells leads to significant down-regulation of biglycan, decorin, and fibromodulin, probably explaining increased TGF-␤ bioavailability, enlarged hinge region of semilunar valves, bicuspid aortic valves, and dilation of ascending aorta in Hdac3 Isl1KO and Hdac3 Mef2CKO embryos. Interestingly, we observed striking up-regulation of lumican, a small leucine-rich proteoglycan, in the outflow tract and developing semilunar valves of Hdac3 Isl1KO embryos, probably explaining thickened semilunar valves. Patients with bicuspid aortic valve, the most common congenital heart defect, are at increased risk for ascending aortic dilatation (73). Proteoglycans facilitate collagen assembly and synthesis, a critical process for the development of semilunar valves and ascending aorta (37). Indeed, patients with mutations in collagen genes, such as Col1a2 or Col3a1, are often predisposed to bicuspid aortic valves and aortic dilatation. Consistent with these observations, significantly reduced expression of Col1a2 and Col3a1 in Hdac3 Isl1KO embryos might explain bicuspid aortic valves and ascending aortic dilatation.
Recent elegant studies (74) clearly demonstrate that in addition to their complex roles in early development, TGF-␤ signaling pathways are involved in many human diseases as a result of mutations in components of the pathways or aberrant regulation of signaling. Our study identifies HDAC3 as a novel and specific regulator of the TGF-␤ signaling pathway at the extracellular, membrane, cytoplasmic, and nuclear levels. For instance, our data demonstrate that loss of HDAC3 augments expression of extracellular agonist ligands of the TGF-␤ pathway, such as TGF-␤1. Simultaneously, expression of extracellular antagonists of these ligands, such as decorin, biglycan, fibromodulin, and collagen, is strikingly diminished in the absence of HDAC3, probably explaining increased TGF-␤ bioavailability in Hdac3 Isl1KO embryos. Similarly, gain of NRP2, SUMO1, KPNB1, and SMAD4 expression in Hdac3 Isl1KO embryos would probably explain amplification of TGF-␤ signaling at the membrane, cytoplasmic, and nuclear levels. Our findings demonstrate that HDAC3 is required to maintain quiescence of valvular mesenchymal cells, the most prevalent cells of valvular cusps. Pathologic conditions promote activation and differentiation of valvular interstitial cells into myofibroblasts, smooth muscle-like fibroblasts that are observed in patients with semilunar valve disease (25). Myofibroblasts express smooth muscle-specific genes and secrete strikingly higher levels of TGF-␤, proteoglycans, and collagen that alter the composition of the extracellular matrix (19). Hence, semilunar valve pathologies augment TGF-␤ bioavailability, which in turn promotes aberrant EndMT and transcriptional activation of smooth muscle and extracellular matrix genes within the myofibroblasts (24). TGF-␤ activation and aberrant extracellular matrix further promote the activation and differentiation of new valvular mesenchymal cells derived via EndMT (25). Consistent with these observations, our data dem-FIGURE 8. HDAC3 epigenetically silences TGF-␤1 within valvular mesenchymal cells by recruiting PRC2 complex to the NCOR complex. A, relative mRNA levels of TGF-␤1 in isolated cardiac endothelial cells (EC) or cardiac mesenchymal cells (MC) from E10.5 Hdac3 F/F outflow tract cushion explants, infected either with CRE or GFP control lentivirus. B, relative mRNA levels of TGF-␤1 in isolated cardiac endothelial cells or dissected semilunar valves (SL) derived from Hdac3 Isl1KO and Hdac3 F/F E14.5 hearts. C, ChIP-qPCR analysis of HDAC3 occupancy upstream of TGF-␤1 in isolated cardiac endothelial cells or cardiac mesenchymal cells from E10.5 outflow tract cushion explants. D, ChIP-qPCR analysis of HDAC3 occupancy upstream of TGF-␤1 in isolated cardiac endothelial cells or dissected semilunar valves from E14.5 hearts. E-K, ChIP-qPCR analysis of H3K27 trimethylation (E), H3K27 acetylation (F), RNA polymerase II (G), CREBBP (H), EZH2 (I), EED (J), and SUZ12 (K) upstream of TGF-␤1 in Hdac3 Isl1KO and control E14.5 semilunar valves. L, ChIP-qPCR analysis of HDAC3 occupancy upstream of TGF-␤1 in E14.5 valvular mesenchymal cells infected with either control shRNA (sc-shRNA), EZH2 shRNA, or NCOR1 shRNA. M, ChIP-qPCR analysis of NCOR1 upstream of TGF-␤1 in Hdac3 Isl1KO and control E14.5 semilunar valves. N, Co-ChIP for HDAC3 and either EZH2, NCOR1, H3K27me3, H3K27ac, or polymerase II upstream of TGF-␤1 in E14.5 wild-type dissected semilunar valves. O, total lysates from E14.5 wild-type pooled semilunar valves were immunoprecipitated (IP) by EZH2 antibody, and Western blot was performed using HDAC3 antibody. ␣-Tubulin is shown as an input control. HDAC3 was quantified and normalized to total input ␣-tubulin using ImageJ software (mean Ϯ S.E. (error bars), n ϭ 3). P, ChIP-qPCR analysis of H3K27me3 upstream of TGF-␤1 in E14.5 valvular mesenchymal cells infected with either control shRNA, EZH2 shRNA, or NCOR1 shRNA. FIGURE 9. HDAC3 functions in a deacetylase-independent manner to regulate EndMT and epigenetic silencing of TGF-␤1. A, HDAC3-FLAG and HDAC3 H134A,H135A -FLAG expression constructs were transfected in HEK-293T cells. Expression was detected by Western blot from whole cell lysates using FLAG antibody. GAPDH is shown as a loading control. B, HDAC3-FLAG and HDAC3 H134A,H135A -FLAG expression was quantified and normalized to total input GAPDH using ImageJ software (mean Ϯ S.D. (error bars), n ϭ 3). C, HDAC activity of HDAC3-FLAG and HDAC3 H134A,H135A -FLAG expression was quantified against a pseudosubstrate. D, EndMT assay of control-or Cre-infected E10.5 Hdac3 F/F outflow tract cushion explants co-infected with GFP, HDAC3-FLAG, or HDAC3 H134A,H135A -FLAG lentiviruses, imaged 24 h after isolation. E, quantification of average radial migration, measured in eight directions, of control-or Cre-infected E10.5 Hdac3 F/F outflow tract cushion explants co-infected with GFP, HDAC3-FLAG, or HDAC3 H134A,H135A -FLAG lentiviruses, measured 24 h after isolation (mean Ϯ S.E. (error bars), n ϭ 3). F, relative mRNA levels of Tgf-␤1 in control-or Cre-infected E14.5 Hdac3 F/F valvular mesenchymal cells co-infected with control, HDAC3-FLAG, or HDAC3 H134A,H135A -FLAG lentiviruses (mean Ϯ S.E., n ϭ 3). G, a 1309-bp TGF-␤1 promoter luciferase reporter (WT) or a truncated, 1267-bp TGF-␤1 promoter luciferase reporter, lacking an HDAC3-enriched region (mutant) were transfected in murine endothelial cells with and without an HDAC3-FLAG or HDAC3 H134A,H135A -FLAG expression plasmid. Induction is represented as a ratio of firefly and Renilla luciferase activity. H-L, ChIP-qPCR analysis of H3K27 acetylation (H), H3K27 trimethylation (I), EZH2 (J), EED (K), and SUZ12 (L) upstream of TGF-␤1 in control-or Cre-infected E14.5 Hdac3 F/F valvular mesenchymal cells co-infected with control, HDAC3-FLAG, or HDAC3 H134A,H135A -FLAG lentiviruses (mean Ϯ S.E., n ϭ 3). FIGURE 10. Summary of phenotypes and proposed model of HDAC3 function within second heart field progenitor cells and second heart field-derived mesenchymal cells. A, loss of HDAC3 in second heart field progenitor cells leads to outflow tract and semilunar valve pathologies. Strikingly, genetic deletion of HDAC3 in differentiated mesenchymal and smooth muscle cells (Hdac3 TaglnKO ) recapitulates the majority of these phenotypes. However, deletion of HDAC3 in differentiated cardiomyocytes (Hdac3 Myh6KO ) or endothelial cells (Hdac3 Cdh5KO ) did not recapitulate the cardiovascular defects observed in Hdac3 Isl1KO embryos. B, Hdac3 Isl1KO hearts exhibit disorganized collagen and elastin within dilated aortic walls and hyperplastic semilunar valves containing activated myofibroblasts and disorganized extracellular matrix. In both control and Hdac3 Isl1KO cardiac endothelial cells, the upstream regulatory region of TGF-␤1 is occupied by RNA polymerase II and CREBBP and exhibits H3K27 acetylation concomitant with TGF-␤1 expression. In control semilunar valves, endothelial cells undergo EndMT to become mesenchymal cells. In these mesenchymal cells, NCOR1, HDAC3, and PRC2 complex (EZH2, EED, and SUZ12) are recruited to the upstream regulatory region of TGF-␤1, which becomes trimethylated on histone H3 Lys-27, and TGF-␤1 expression is epigenetically silenced. In Hdac3 Isl1KO hearts, EZH2, EED, and SUZ12 are not recruited to the TGF-␤1 regulatory region, RNA polymerase II and CREBBP are present, and histone H3 Lys-27 remains acetylated, favoring aberrant expression of TGF-␤1 in mesenchymal cells. TGF-␤1 activates mesenchymal cells to become myofibroblasts, which perpetuate EndMT and activation of mesenchymal cells through continued induction of TGF-␤1 and aberrant expression of extracellular matrix, including proteoglycans and collagen.
onstrate that loss of HDAC3 leads to strikingly higher expression of TGF-␤1, smooth muscle genes, and aberrant expression of extracellular matrix genes within the valvular mesenchymal cells, and these changes were largely abolished by TGF-␤-neutralizing antibody. Taken together, our findings support a model in which HDAC3-mediated repression of TGF-␤1 within valvular mesenchymal cells is required to prevent perpetual activation of EndMT and to maintain extracellular matrix homeostasis in the semilunar valves.
Catalytically inactive enzymes or "pseudoenzymes" are clearly widespread, occurring in most enzyme families, including HDACs (66). For instance, class II HDACs, such as HDAC4, -5, -7, and -9, lack deacetylase activity due to a His substitution at the key Tyr residue (75). These pseudoenzymes function as scaffolding proteins to recruit various co-factors in a signal-dependent manner (42). Similarly, HDAC1 and -3, class I HDACs, and their deacetylase-dead mutants have the same degree of in vivo effect on cardiomyopathy (76). During endothelial-tomesenchymal transition, epigenetic mechanisms regulating gene activation and silencing remain largely undefined (77). The present study provides the mechanistic basis of a deacetylase-independent function of HDAC3 as a scaffold to recruit methyltransferase components of PRC2 (polycomb repressive complex 2) to epigenetically silence TGF-␤1. Repressor complex-mediated gene silencing is critical to maintain cellular identity of differentiated cells through multiple divisions (78). Recent evidence suggests that PRC2-mediated epigenetic silencing is maintained for many cell generations. EZH2, the major H3K27 methyltransferase of the PRC2 complex, supported by EED and SUZ12, catalyzes methylation of H3K27 to mediate chromatin compaction and thereby regulate differentiation and cell identity (79). Consistent with this model, our data demonstrate that HDAC3 recruits EZH2, EED, and SUZ12 to the NCOR complex in a deacetylase-independent manner and thereby mediates epigenetic silencing of TGF-␤1 within valvular mesenchymal cells. In the absence of HDAC3, the NCOR complex fails to recruit PRC2 complex, resulting in aberrant recruitment of CREBBP, which catalyzes acetylation of H3K27 to mediate activation of TGF-␤1 expression, which in turn promotes perpetual transdifferentiation of valvular mesenchymal cells to myofibroblasts. Subsequently, activated myofibroblasts secrete higher levels of extracellular matrix proteins and promote aberrant EndMT. Future investigations dissecting the modes of interaction among HDAC3, EZH2, EED, and SUZ12 will be important to determine PRC2-independent functions of HDAC3. Similarly, murine models expressing catalytically inactive HDAC3 would further define deacetylase-independent functions of HDAC3 during cardiogenesis.
Elucidation of the role of HDAC3 in the second heart field is directly relevant to human congenital heart disease. Marfan syndrome, Ehlers-Danlos syndrome, and Loeys-Dietz syndrome, caused by mutations and activation of TGF-␤ signaling genes, are associated with congenital cardiovascular anomalies resembling those seen in Hdac3 Isl1KO , Hdac3 Mef2CKO , or Hdac3 TaglnKO embryos. Our data suggest that many of these cardiovascular defects, such as aortic dilatation, outflow tract defects, bicuspid aortic valve, semilunar valve stenosis, and membranous ventricular septal defects, are likely to be caused by defective morphogenesis of the second heart field and aberrant extracellular matrix as a result of anomalous epigenetic silencing of TGF-␤.
Author Contributions-S. L. L., H. P. J., and C. M. T. designed, performed, and analyzed the experiments. S. L. L. and C. M. T. coordinated the study and wrote the paper. C. M. T. conceived the study, acquired funding, and supervised S. L. L. and H. P. J. All authors reviewed the results and approved the final version of the manuscript.