AMP-activated Protein Kinase Regulates β-Catenin Transcription via Histone Deacetylase 5*

AMP-activated protein kinase (AMPK) is a key regulator of energy metabolism; it is inhibited under obese conditions and is activated by exercise and by many anti-diabetic drugs. Emerging evidence also suggests that AMPK regulates cell differentiation, but the underlying mechanisms are unclear. We hypothesized that AMPK regulates cell differentiation via altering β-catenin expression, which involves phosphorylation of class IIa histone deacetylase 5 (HDAC5). In both C3H10T1/2 cells and mouse embryonic fibroblasts (MEFs), AMPK activity was positively correlated with β-catenin content. Chemical inhibition of HDAC5 increased β-catenin mRNA expression. HDAC5 overexpression reduced and HDAC5 knockdown increased H3K9 acetylation and cellular β-catenin content. HDAC5 formed a complex with myocyte enhancer factor-2 to down-regulate β-catenin mRNA expression. AMPK phosphorylated HDAC5, which promoted HDAC5 exportation from the nucleus; mutation of two phosphorylation sites in HDAC5, Ser-259 and -498, abolished the regulatory role of AMPK on β-catenin expression. In conclusion, AMPK promotes β-catenin expression through phosphorylation of HDAC5, which reduces HDAC5 interaction with the β-catenin promoter via myocyte enhancer factor-2. Thus, the data indicate that AMPK regulates cell differentiation and development via cross-talk with the wingless and Int (Wnt)/β-catenin signaling pathway.

␤-Catenin is a key mediator of Wingless and Int (Wnt)/␤catenin signaling pathway, which is required for early embryonic development (14,15), cell proliferation, and differentiation (16 -18). We postulate that AMPK regulates cell differentiation through cross-talk with the Wnt/␤-catenin signaling pathway. We previously observed that AMPK phosphorylates ␤-catenin, which enhances ␤-catenin stability (19). In this study we further observed that the mRNA expression of ␤-catenin was promoted by AMPK and suggested that histone deacetylase 5 (HDAC5) has an essential role in linking AMPK with ␤-catenin expression.
Epigenetic modifications including histone acetylation and methylation and DNA methylation regulate gene transcription (20,21). Histone acetylation is regulated by histone acetyltransferase and HDAC (22). HDAC5 belongs to the class IIa HDAC family and acts as a conserved transcriptional repressor. HDAC5 interacts with myocyte enhancer factor-2 (MEF2) to target specific gene promoters (23). HDAC5 is phosphorylated by several kinases, including calmodulin-dependent protein kinases (24), protein kinase D (25), salt-inducible kinase (26,27), and protein kinase A (28). Recently, it was also reported that HDAC5 was phosphorylated by AMPK on Ser-259 and Ser-498 (29). We identified a MEF2 binding site on the ␤-catenin promoter. These data prompted us to hypothesize that AMPK regulates ␤-catenin expression through phosphorylation of HDAC5. Here, we present data showing that AMPK phosphorylates HDAC5, which promotes its nuclear export, leading to the acetylation of histones that are bound to the ␤-catenin promoter and enhanced ␤-catenin expression.
Cell Culture-Mouse C3H10T1/2 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% antibiotics at 37°C with 5% CO 2 . Mouse embryonic fibroblasts were isolated as previously described with minor modifications (30). Briefly, mouse embryos at E13 of wild-type and AMPK␣2 knock-out mice (31) were minced in PBS after removing head and visceral tissues. The minced tissue mixture was treated with 0.1 mM trypsin for 15 min, 37°C. After trypsinization, 5 ml DMEM plus 10% FBS was added to the mixture, and the cell suspension was incubated at 37°C with 5% CO 2 .
Immunoblotting Analyses-Immunoblotting analysis was performed as previously described using an Odyssey Infrared Imaging System (LI-COR Biosciences) (32). Band density was normalized to ␤-tubulin content.
Isolation of Nuclei and Cytosol-Nuclei and cytosol were isolated using a nuclear extract kit according to the manufacturer's instructions (Active Motif).
Construction of Expression Vector-The 3.8-kb ␤-catenin promoter fragment containing the MEF2 binding site was amplified from mouse DNA using PCR with following primers: forward (5Ј-GGAGTGGCTTCCCTACAT-3Ј) and reverse (5Ј-CAGGCATCTGAATCACAAAC-3Ј). PCR products were subcloned between Xhol and Nhel sites of the pGL4promoter luciferase plasmid (Promega). The construct was verified by digestion with restriction enzymes and sequencing.
Transfection-Plasmid transfection was performed using Lipofectamine (Invitrogen) according to the manufacturer's instructions. Briefly, cells were changed to no antibiotic medium 12 h before transfection. Transfections were carried out when cells reached 95% confluence using a 1:3 ratio of DNA (g):Lipofectamine (l). For shRNA interference, HDAC5 shRNA and control shRNA (Santa Cruz Biotechnology, Santa Cruz, CA) were delivered into cells, and transfected cells were selected using puromycin (2 g/ml).
Immunoprecipitation-After washing with cold PBS, cells were lysed with ice-cold lysis buffer containing 20 mM Tris (pH 7.5), 1% Triton X-100, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM PMSF, 2.5 mM sodium pyrophosphate, 1 g/ml leupeptin, 1 mM Na 3 VO 4 , and 100 mM NaF. Lysates were collected and centrifuged at 14,000 ϫ g for 10 min at 4°C, and 200 l supernatant was precleared with 20 l of Protein A-Sepharose bead slurry (50%) (Rockland, Inc., Gilbertsville, PA) for 2 h with shaking. The supernatant (150 l) was transferred to a new tube, mixed with anti-HDAC5 antibody (1:50 dilution), and incubated with rocking at 4°C overnight. Protein A-Sepharose bead slurry (50%, 20 l) was added, and incubation was continued with rocking for 2 h. Immunoprecipitates were collected and washed with 100 l of lysis buffer three times. Then an equal volume of 2ϫ SDS loading buffer was added, and samples were boiled for 5 min. Immunoblotting was conducted using an antibody against 14-3-3 (Cell Signaling).
DNA Mutagenesis-WT HDAC5 was obtained from Addgene. Ser-259 and Ser-498 of HDAC5 were mutated to Ala using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The MEF2 binding site on ␤-catenin promoter was deleted using the same kit. All mutations or deletion were sequence-verified.
Luciferase Reporter Activity Assay-To measure transcriptional activity of the ␤-catenin promoter, C3H10T1/2 cells were transfected with the ␤-catenin promoter reporter vector, the HDAC5 vector, or vectors with mutated HDAC5. Renilla luciferase vector (Promega) was transfected as an internal control. Each transfection was performed 6 times. Beginning 24 h after transfection, cells were treated with AICAR or vehicle for 12 h. Cells were then harvested, and luciferase activity was measured using a dual luciferase assay kit (Promega) according to the manufacturer's instructions. Data from each experiment were normalized to the Renilla luciferase activity.
Immunocytochemistry-Cells grown on coverslips were fixed in 4% paraformaldehyde and incubated with anti-tag or mouse IgG (1:100) at 4°C overnight. A fluorescent secondary antibody (1:500) was added, and fluorescence was examined as previously described (33).
Chromatin Immunoprecipitation Assay-Chromatin immunoprecipitation assay was performed using a kit (Millipore, Billerica, MA) according to the manufacturer's instructions. Briefly, C3H10T1/2 cells that had been treated with AICAR (0.25 mM) or vehicle were cross-linked with 1% formaldehyde for 10 min at room temperature and quenched with 125 mM glycine. Cells were lysed, and the lysate was sonicated to break up the DNA to 200 -1000-bp fragments. DNA-protein complex was precipitated with HDAC5 antibody (Active Motif) or a mouse IgG antibody (negative control). DNA was purified, and real-time PCR was conducted using a SYBR Green kit from Bio-Rad to amplify a 233-bp fragment surrounding the MEF2 binding site of catenin promoter. Primer sequences were as follows: forward (5Ј-TATGTTTACGTTAGGAAGGGTT-3Ј) and reverse (5Ј-TCTGCTTCTGCCTGGTTG-3Ј). Relative DNA content was normalized to input DNA of each sample.
Statistics-For all cell culture studies, at least three independent experiments were conducted. All data were expressed as the means Ϯ S.E. Data were analyzed using the General Linear Model (GLM) of SAS (SAS Institute Inc., Cary, NC), and Tukey's Studentized Range test was used to determine significant differences among means. p Ͻ 0.05 was considered to be significant.

RESULTS
AMPK Regulates ␤-Catenin Expression-We first determined whether AMPK regulated ␤-catenin expression. C3H10T1/2 cells were treated with various doses of AICAR, which indicated that 0.25 mM AICAR was sufficient to activate AMPK, and this amount, therefore, was used for further studies (supplemental Fig. 1). In contrast to some previous studies, we did not use higher concentrations of AICAR because doses higher than necessary to activate AMPK might have unwanted side effects.
After treatment of C3H10T1/2 cells with AICAR (0.25 mM) or vehicle (PBS), mRNA expression (6 h) of ␤-catenin was analyzed. We also did a time course analysis of ␤-catenin protein content after AICAR treatment and found that ␤-catenin protein content reaches a maximum after 48 h of treatment (supplemental Fig. 2). AICAR treatment activated AMPK, as indicated by enhanced phosphorylation of acetyl-CoA carboxylase, an exclusive substrate of AMPK (Fig. 1B). At the same time, AICAR treatment increased ␤-catenin content on both mRNA and protein levels (Fig. 1, A and B). We transfected C3H10T1/2 cells with a vector containing wild-type AMPK␣2, kinase-dead AMPK␣2 K45R plasmid, or a GFP control vector to determine whether ectopic AMPK expression affected ␤-catenin expression. Similar to the results obtained from AICAR treatment, cells transfected with AMPK␣2 WT had higher ␤-catenin content than cells transfected with AMPK␣2 K45R, showing that AMPK promoted ␤-catenin expression (Fig. 1, C and D). To further verify these findings, we repeated the above experiments in mouse embryonic fibroblasts. Consistently, mouse embryonic fibroblasts from AMPK␣2 knock-out mice exhibited significantly lower ␤-catenin content compared with mouse embryonic fibroblasts from WT mice (Fig. 1, E and F). To further verify whether AMPK is involved in the Wnt/␤catenin signaling pathway, we analyzed nuclear ␤-catenin content; nuclear ␤-catenin regulates gene expression (37). Data showed that AMPK activation increased nuclear ␤-catenin content (Fig. 1G). All these results showed that AMPK positively regulated ␤-catenin expression.
To further confirm our hypothesis, we used HDAC5 shRNA to knock down endogenous HDAC5. As shown in Fig. 2E, compared with control shRNA treatment, HDAC5 knockdown increased ␤-catenin mRNA levels; H3K9 acetylation levels and ␤-catenin protein content were also increased (Fig. 2F). These data demonstrate that HDAC5 regulates ␤-catenin expression.
The promoter specificity of HDAC5 is mediated by MEF2. To determine whether HDAC5 regulates the transcriptional activity of ␤-catenin promoter directly, we cloned the ␤-catenin promoter ( Fig. 2A) and constructed a ␤-catenin promoter-luciferase vector. We constructed a second vector containing the ␤-catenin promoter with the MEF2 binding site deleted. C3H10T1/2 cells were transfected with the ␤-catenin promoter-luciferase vector or the mutated ␤-catenin promoter-luciferase vector with or without co-transfection of the HDAC5 vector. In cells receiving the ␤-catenin promoter-luciferase vector and HDAC5, luciferase activity was suppressed. When the MEF2 binding site was deleted, this suppression was abolished (Fig. 2G). These data demonstrate that HDAC5 regulates ␤-catenin transcription via the specific MEF2 binding site on the ␤-catenin promoter.
To further demonstrate the relation between AMPK and HDAC5, we performed immunocytochemical staining after transfecting C3H10T1/2 with an HDAC5 vector containing a tag near the HDAC5 C terminus. As shown in Fig. 3D, in the absence of a stimulus, HDAC5 remained localized inside the nucleus. After AMPK activation, however, a large amount of HDAC5 was translocated to the cytoplasm. These data suggest that AMPK regulates HDAC5 though phosphorylation, which leads to relocation of HDAC5 from nuclei to cytoplasm.
AMPK Regulates ␤-Catenin Expression through Phosphorylation of HDAC5-A previous study showed that AMPK phosphorylates HDAC5 at Ser-259 and -498 (29). To investigate whether AMPK regulates HDAC5 through phosphorylation, we used immunoblotting analyses to assess the presence of phosphorylated HDAC5 after AMPK activation. The results demonstrated that the concentration of phospho-HDAC5 at Ser-258 and -498 correlated with AMPK activity (Fig. 4A). To further test whether AMPK mediates ␤-catenin expression through phosphorylation of HDAC5, we generated two HDAC5 mutants in which either Ser-259 or Ser-498 were substituted with Ala. We then transfected C3H10T1/2 cells with HDAC5 or the mutants; transfected cells were further treated with AICAR. AMPK activation failed to increase the ␤-catenin and ␤-catenin mRNA content was analyzed. C and D, C3H10T1/2 cells were transfected with either a HDAC5 or GFP vector, and ␤-catenin mRNA expression (C) and H3K9 acetylation levels and ␤-catenin protein content (D) were analyzed. E and F, C3H10T1/2 cells were transfected with HDAC5 shRNA or control shRNA and ␤-catenin mRNA expression (E), and H3K9 acetylation level (AC-H3K9) and ␤-catenin content were analyzed (F). G, C3H10T1/2 cells were transfected with a ␤-catenin promoter-luciferase vector or mutated ␤-catenin promoter, which lacks the MEF2 binding site together with or without co-transfection of the HDAC5 vector. Luciferase activity was analyzed. *, p Ͻ 0.05; bars atop with a different letter differ; mean Ϯ S.E.; n ϭ 3; for luciferase assay, n ϭ 6. content in cells transfected with either of the two HDAC5 mutants (Fig. 4B), clearly showing that AMPK induced ␤-catenin expression via phosphorylation of HDAC5 at Ser-259 and -498.
To further analyze the regulation of ␤-catenin expression, we co-transfected C3H10T 1/2 cells with HDAC5 and the ␤-catenin promoter luciferase construct with or without AICAR treatment. AICAR treatment increased luciferase activity (Fig. 4C). We also co-transfected C3H10T1/2 cells with HDAC5 or its mutants together with AICAR treatment. As shown in Fig. 4D, AMPK activation did not increase the luciferase activity in cells transfected with HDAC5 mutants but did increase luciferase activity in the HDAC5 group, demonstrating that HDAC5 phosphorylation by AMPK is necessary for regulation of ␤-catenin expression by AMPK.
To regulate gene-specific histone acetylation, HDAC5 needs to interact with MEF2, and as a complex, HDAC5/MEF2 medi-ates gene-specific histone acetylation. To analyze whether activation of AMPK with AICAR decreases the amount of HDAC5/ MEF2 complex binding to the MEF2 binding site on the ␤-catenin promoter, we performed ChIP analysis. AICAR treatment decreased the amount of HDAC5 associated with the ␤-catenin promoter surrounding the MEF2 binding site; no decrease was detected in two HDAC5 mutants (Fig. 4E). These data are consistent with re-location of HDAC5 from nucleus to cytoplasm (Fig. 3D). Finally, we transfected C3H10T 1/2 cells with HDAC5 and its two mutants in combination with AICAR or vehicle treatment, and the subcellular localization of HDAC5 was analyzed. Most of the HDAC5 as well as the mutants were localized inside nuclei before AICAR treatment. After AMPK activation, a portion of HDAC5 relocalized to the cytoplasm, but either mutation partially disrupted relocalization of HDAC5 to the cytoplasm (Fig. 4F). We also analyzed the mRNA expression of a number of Wnt/␤-catenin downstream   Fig. 3D) and its two mutants followed by AICAR (0.25 mM) or vehicle (PBS) for 3 h. Subcellular localization of HDAC5 was analyzed by immunocytochemical staining. *, p Ͻ 0.05; **, p Ͻ 0.01; bars atop with a different letter differ; Mean Ϯ S.E.; n ϭ 3; for luciferase assay, n ϭ 6). genes in mouse embryonic fibroblast cells separated from wildtype or AMPK␣2 knock-out mice, including Pax 3, Pax 7, MyoD, Myf5, and myogenin. All these genes are involved in myogenesis, and their expression is correlated with canonical Wnt/␤-catenin signaling (38). Indeed, their expression was higher in wild type compared with AMPK ␣2 knock-out embryonic fibroblasts, correlated with reduced ␤-catenin content in AMPK ␣2 knock-out embryonic fibroblasts (Fig. 1F). These data showed that AMPK regulates ␤-catenin expression, which had significant effect on myogenesis (supplemental Fig. 3). Taken together, these data showed that AMPK regulates ␤-catenin transcription through phosphorylation of HDAC5, which results in relocation of HDAC5 to the cytoplasm and loss of the interaction of HDAC5 with MEF2 (Fig. 5).
In our previous studies we observed that low AMPK activity correlated with reduced ␤-catenin content in fetal skeletal muscle (55,56). To analyze the association between AMPK and ␤-catenin expression, we first treated C3H101/2 cells with AICAR to activate AMPK. With AMPK activation, both mRNA and protein levels of ␤-catenin were increased. The correlation of AMPK activity with ␤-catenin expression was further confirmed by ectopic expression of AMPK WT or AMPK K45R. Furthermore, higher AMPK activity induced the accumulation of nuclear ␤-catenin, which regulates gene expression (37).
AMPK contains two catalytic subunit isoforms, ␣1 and ␣2. AMPK ␣2 subunit is primarily distributed in the nuclear compartment (57), which suggests that AMPK ␣2 has a major role in the regulation of gene expression. We isolated embryonic fibroblasts with AMPK ␣2 knock-out (Ϫ/Ϫ) or heterozygous (ϩ/Ϫ) from the skeletal muscle of fetal mice originating from the same pregnancy and compared the expression of ␤-catenin between ␣2 (Ϫ/Ϫ) and ␣2 (ϩ/Ϫ) embryonic fibroblasts. We found that ␣2 knock-out embryonic fibroblasts had lower ␤-catenin content compared with heterozygous fibroblasts. In addition, the mRNA expression of Wnt/␤-catenin downstream genes including Pax 3, Pax 7, MyoD, Myf5, and myogenin was reduced, which is evidence of a regulatory role of AMPK on ␤-catenin expression and its downstream signaling. FIGURE 5. Schematic diagram shows the proposed mechanism for AMPK regulation of ␤-catenin transcription. Without AMPK activation, MEF2 recruits HDAC5 to ␤-catenin promoter region, leading to deacetylation and inhibition of ␤-catenin gene expression. On the other hand, activation of AMPK phosphorylates HDAC5 on both Ser-259 and Ser-498 and initiates the interaction between HDAC5 and 14-3-3; this interaction releases HDAC5 from MEF2 and promotes its relocalization from nuclei to cytoplasm, promoting ␤-catenin gene transcription.
The remaining question was how AMPK regulated ␤-catenin expression. To address this, we analyzed the structure of the ␤-catenin promoter, and one conserved MEF2 binding site (TAAAAATA) (58 -60) was found. Because HDAC5 interacts with MEF2 to regulate gene expression (61,62), we asked whether HDAC5 regulates ␤-catenin expression. We inhibited HDAC5 by scriptaid and observed increased ␤-catenin mRNA content. In addition, HDAC5 shRNA increased both H3K9 acetylation and ␤-catenin expression, whereas HDAC5 overexpression reduced both H3K9 acetylation and ␤-catenin expression. We also tested ␤-catenin promoter activity. HDAC5 transfection reduced ␤-catenin promoter activity, which indicated that HDAC5 inhibits ␤-catenin transcription. When the MEF2 binding site in the ␤-catenin promoter was deleted, the inhibitory effect of HDAC5 on ␤-catenin transcription was abolished, suggesting that the interaction of HDAC with MEF2 was necessary for regulating ␤-catenin transcription.
Phosphorylation of HDAC5 by protein kinases leads to its translocation from nucleus to cytoplasm through interaction with 14-3-3. This is a conserved mechanism by which HDAC class II family members regulate gene transcription (63). We asked if AMPK regulated HDAC5 through phosphorylation and interaction with 14-3-3. With AMPK activation, the concentration of phosphorylated HDAC5 was increased, and more 14-3-3 was co-immunoprecipitated, indicating that AMPK phosphorylates HDAC5, consistent with a previous report showing that AMPK phosphorylates HDAC5 on Ser-259 and Ser-498 (29). After phosphorylation, HDAC5 was exported from nucleus, leading to an increase in H3K9 acetylation. To confirm this, we also transfected cells with an HDAC5 expression vector containing a tag and performed immunocytochemical staining using an anti-tag antibody. Our data showed that HDAC5 localized primarily in the nucleus in normal cells, but large amounts of HADC5 were translocated to the cytoplasm with AMPK activation, further suggesting that AMPK regulated HDAC5 localization.
To further determine whether AMPK regulates ␤-catenin expression through phosphorylation of HDAC5, we mutated Ser-259 and Ser-498 of HDAC5 to Ala, which prevents phosphorylation by AMPK and disrupts the 14-3-3 binding sites (23). Our data showed that AICAR increased ␤-catenin content when cells were transfected with wild-type HDAC5 but not with mutants lacking 14-3-3 binding sites. We also observed that AMPK activation increased ␤-catenin promoter activity when cells were co-transfected with wild-type HDAC5 but not with the mutants. These data suggested that disruption of AMPK phosphorylation sites abolished nuclear exportation. ChIP analysis further confirmed our finding. With AMPK activation, HDAC5 was phosphorylated and translocated to the cytoplasm, reducing the amount of HDAC5 bound to MEF2 binding site. This notion was further confirmed by immunocytochemical staining. The mutation from Ser to Ala of either the AMPK phosphorylation site at Ser-259 or Ser-498 of HDAC5 partially disrupted the translocation of HDAC5.
Our observation that AMPK mediates ␤-catenin transcription through HDAC5 extends the role of AMPK from that of an energy sensor to a key regulator of cell differentiation and tissue development. It has been well established that Wnt/␤-catenin signaling is crucial for cell differentiation and early embryonic development. Consistently, AMPK is also crucial for early embryonic development. A recent report showed that knockout of both AMPK ␣1 and ␣2 in mice leads to embryonic lethality at E10.5 (13), a stage corresponding approximately to the initial myogenesis from MSC.
The obesity epidemic is becoming increasingly serious. It is known that obesity induces low grade inflammation, which inhibits AMPK via a tumor necrosis factor-␣ induced and protein phosphatase 2C-mediated dephosphorylation (64), which has been demonstrated in our previous studies in fetal skeletal muscle (55). In addition, AMPK is also regulated by adipokines such as leptin, adiponectin, and resistin and by cytokines, such as interleukin-6 (65). Furthermore, AMPK is activated during exercise (66,67). Therefore, obesity, inflammation, exercise, and likely other physiological factors that alter AMPK activity affect the differentiation of multi-potent cells through crosstalk between AMPK and Wnt/␤-catenin signaling. Because of the critical role of Wnt/␤-catenin signaling in cell differentiation and tissue development, such cross-talk between AMPK and ␤-catenin signaling likely has important physiological and developmental implications, including exerting long term effects on the properties of skeletal muscle and other tissues.