AMP-Activated Protein Kinase Regulates GLUT4 Transcription by Phosphorylating Histone Deacetylase 5

Muscle samples were obtained from the vastus lateralis of five male subjects (23 ± 2 years of age, 75 ± 5 kg, 21.9 ± 0.3 kg/cm²) using the percutaneous needle biopsy technique with suction, after completing a medical questionnaire and giving informed written consent. All experimental procedures were approved by the Deakin University Human Research Ethics Committee.

Muscle samples were washed and trypsin digested before being filtered through a 100-μm cell filter. Fibroblasts were removed by seeding the supernatant in an uncoated flask and incubating for 25 min at 37°C. Myoblasts were cultured in extracellular matrix–coated flasks in growth media (α-MEM supplemented with 10% fetal bovine serum, 0.5% antibiotic, and 0.5% antifungal) at 37°C and 5% CO₂. Myoblasts were grown to 80% confluence before being differentiated to myotubes via serum withdrawal, where media fetal bovine serum was replaced with 2% horse serum.

All experimental protocols were performed 4 days postdifferentiation following overnight incubation in serum free media. Cells were exposed to either 1 mmol/l AICAR, 10 μmol/l 6-(1,3-dioxo-¹H,³H-benzo[de]isoquinolin-2-yl)-hexanoic acid hydroxyamide (Scriptaid), or an equivalent volume of vehicle (DMSO) as a control for 60 min at 37°C. Cells designated for protein extraction were harvested immediately after treatment, while cells designated for RNA extraction were harvested 6 h after treatment. For glucose uptake studies, cells were washed and incubated in fresh PBS containing 0.1% BSA/1 μmol/l insulin for 1 h at 37°C before incubation in 10 μmol/l 2-deoxy-d-[1-³H]-glucose (200 μCi/mmol) for 15 min. Cells were lysed in 0.3 M NaOH and scintillation counted (Perkin Elmer, Wellesley, MA).

Total and nuclear proteins were extracted, and concentrations were determined and immunoblotted as previously described (16). Membranes were exposed to anti-GLUT4 (Biogenesis, Poole, U.K.), MEF2, myosin heavy chain, myogenin, MyoD, 14-3-3 (Santa Cruz Biotechnology, Santa Cruz, CA), HDAC5, pT172 AMPKα, AMPKα (Cell Signaling Technology, Beverly, MA), histone 3 (H3), and acetylated H3K9 and 14 (acetyl-H3; Upstate Biotechnology, Lake Placid, NY), pyruvate dehydrogenase (PDH; Molecular Probes, Eugene, OR), and α-tubulin and β-actin (Sigma, St. Louis, MO) antibodies. HDAC5 phospho-Ser259 (raised against LRKTApSEPNLKC) and phospho-Ser498 (raised against LSRTQpSSPLPQ) antibodies were produced by immunizing rabbits with the aforementioned peptides coupled with keyhole limpet hemocyanin. Antisera were harvested and passed through dephosphorylated HDAC5 affinity purification columns followed by the relevant phospho-peptide affinity purification column. Antibodies were then eluted from the columns and dialyzed.

Total RNA was extracted using the RNeasy mini kit (Qiagen, Valencia) and was reverse transcribed to cDNA. Real-time RT-PCR was conducted with GLUT4 forward (CTT CAT CAT TGG CAT GGG TTT) and reverse (AGG ACC GCA AAT AGA AGG AAG A) primers (5′ to 3′) using SYBR green chemistry (Applied Biosystems, Scoresby, Victoria, Australia). Changes in GLUT4 gene expression were normalized to the housekeeping gene cyclophilin.

All cloning was performed using the Gateway cloning system (Invitrogen, Carlsbad, CA). HDAC5 was subcloned into the pDEST26 expression vector. Wild-type AMPK α₁, β₁, and γ₁ subunits were subcloned into the pDEST27 expression vector. Dominant-negative (K45R) and constitutively active (T172D) mutations in pDEST27 AMPK α₁ and S259A and S498A mutations in pDEST26 HDAC5 were performed using the Quickchange kit (Stratagene, La Jolla, CA).

Subconfluent Cos7 cells were transiently transfected with HDAC5 wild-type and mutant plasmids using FuGENE 6 (Roche Applied Science, Indianapolis, IN), according to manufacturer’s specifications. HDAC5 was immunopurified with 2 μg antibody and 30 μl protein A agarose beads before being incubated with 50 mU purified AMPK (Upstate Biotechnology) in the presence of [γ-³²P]ATP (∼5,000 cpm/pmol) for 60 min as previously described (22). Gels were dried and exposed to a phosphor imager (Fujifilm) before being reconstituted, transferred to a nitrocellulose membrane, and immunoblotted for HDAC5.

Coimmunoprecipitation assays were performed as previously described (15) from transiently transfected subconfluent Cos7 cells.

Gene reporter assays were performed as previously described (23) with minor modifications. Cos7 cells were transiently transfected with combinations of plasmids encoding the human GLUT4 promoter tagged to the firefly luciferase gene (hG4-Luc; 250 ng), MEF2A plasmid (250 ng), or various HDAC5 plasmids (250 ng). To control for transfection efficiency, cells were cotransfected with Renilla luciferase (25 ng) and total transfected DNA made up to 1 μg with non-expressing plasmid vector (pcDNA). Cells were AICAR or vehicle (DMSO) treated 24 h after transfection and harvested 48 h after transfection. Dual luciferase assays were performed using a kit (Promega, Madison, WI) according to the manufacturer’s instructions and were measured using a manual luminometer (Hidex, Turku, Finland).

Chromatin immunoprecipitation (ChIP) assays were performed on vehicle (DMSO)- and AICAR-treated primary myotubes using a kit (Upstate Biotechnology) according to the manufacturer’s instructions. Primary myotubes were double cross-linked in 2 mmol/l disuccinimidyl glutarate for 60 min at room temperature and in 1% formaldehyde for 12 min before being quenched with 125 mmol/l glycine. Cells were lysed, input DNA stored, and HDAC5 immunoprecipitated using an HDAC5 antibody (Active Motiff) or unrelated antibody (mouse IgG) for controls. DNA was extracted and real-time RT-PCR was conducted using SYBR green chemistry (Applied Biosystems) with primers designed to amplify a 75-bp fragment of the GLUT4 promoter surrounding the MEF2 binding region (forward: 5′-CCT GAC ATT TGG AGG CTC-3′; reverse: 5′-GGA GCA ATG CCC CAA AG-3′). To control for differences in initial DNA concentration, all ChIP cycle threshold (CT) values were normalized to the input CT value for that sample.

All data are presented as means ± SE with control samples assigned the arbitrary value of 1.0 and treatment samples expressed relative to control. For data from semiquantitive assays, values are designated as arbitrary units. Means were compared using a t test or one-way ANOVA with a significance level of P < 0.05.

We have previously found that HDAC5 dissociation from MEF2 and subsequent HDAC5 nuclear export is associated with increased GLUT4 gene expression in human skeletal muscle (16). From these data, it is not possible to determine whether HDACs, or specifically HDAC5, directly regulate GLUT4 transcription. To test this, human skeletal muscle biopsies of the vastus lateralis were grown in culture and differentiated into human primary myotubes. These cells differentiated into elongated myotubes (supplementary Fig. 1 available in an online appendix at http://dx.doi.org/10.2337/db07-0843) concomitant with increased muscle-specific myogenic marker expression, including myosin heavy chain, MyoD, and myogenin (Fig. 1A). GLUT4 and MEF2, a transcription factor required for GLUT4 expression, are expressed in these terminally differentiated myotubes, although GLUT4 was expressed at lower levels than in extracts from human skeletal muscle (Fig. 1A). Furthermore, these myotubes displayed insulin-dependent glucose uptake (Fig. 1B).

To determine whether the GLUT4 gene is regulated by HDACs, primary myotubes were exposed to a single 60-min treatment with the HDAC inhibitor Scriptaid. Scriptaid possesses relatively low toxicity and is structurally similar to the class of hydroxamic acid-containing HDAC inhibitors (24) that inhibit HDAC activity by interacting with a zinc binding site found in a small polar pocket within the COOH-terminal deacetylase domain and do not appear to dissociated HDACs from their DNA binding partner or result in HDAC relocalization (25). Following Scriptaid treatment, GLUT4 mRNA was elevated sevenfold (P < 0.05) when compared with control (Fig. 1C). Expression of the housekeeping gene cyclophilin did not change (2^−CT values: 1.18 × 10⁻⁹ ± 1.01 × 10⁻¹⁰ vs. 1.07 × 10⁻⁹ ± 1.27 × 10⁻¹⁰; P = 0.72). Scriptaid treatment also increased H3 acetylation approximately ninefold (P < 0.05) when compared with control (supplementary Fig. 2), consistent with the increased GLUT4 transcription being a result of chromatin remodeling. There was no change in nuclear HDAC5 abundance following Scriptaid treatment (supplementary Fig. 3), confirming that Scriptaid does not alter HDAC localization.

To determine whether the GLUT4 gene is directly regulated by HDAC5, a gene reporter assay using the human GLUT4 promoter coupled to the firefly luciferase gene (hG4-Luc) was performed. This promoter contains all the regulatory elements located within −895 bp of the transcription start site that are required for normal GLUT4 expression in vivo (23). To control for transfection efficiency, cells were also transfected with a plasmid expressing Renilla luciferase. Cotransfection with MEF2A plasmid increased GLUT4 reporter gene expression ∼4.5-fold (P < 0.05), which was returned to control levels with the transfection of wild-type HDAC5 (Fig. 1D). Expression of the MEF2A and HDAC5 plasmids was verified by immunoblotting. Together, these data show that the GLUT4 gene is regulated by HDAC activity and identify HDAC5 as a repressor of GLUT4 gene expression.

Since increases in AMPK activity are associated with increases in GLUT4 expression, we sought to determine whether AMPK might regulate HDAC5. Specifically, phosphorylation of HDAC5 serines 259 and 498 causes HDAC5 dissociation from MEF2. To test whether AMPK could act as an HDAC5 kinase, HDAC5 plasmids expressing either wild-type, a serine 259 to alanine mutant (S259A), a serine 498 to alanine mutant (S498A), or a dual serine 259 and 498 to alanine mutant (S259/498A) were transiently transfected into subconfluent Cos7 cells. HDAC5 was immunoprecipitated from lysates of these cells and incubated with purified AMPK in the presence of [γ-³²P]ATP. Incubation of wild-type HDAC5 with AMPK resulted in HDAC5 phosphorylation that was not observed when AMPK was absent (Fig. 2A).

Mutation of serine 259 was sufficient to block AMPK phosphorylation of HDAC5 (Fig. 2A), supporting previous observations that this site is a priming site required for the subsequent phosphorylation of other sites (26). This site is targeted by a number of kinases, including members of the family of AMPK-related, LKB1-regulated kinases (26,27). Mutation of serine 498 showed that AMPK only weakly phosphorylates serine 259 in vitro (Fig. 2A). Differences in HDAC5 expression could not account for these results (Fig. 2A). Although residual phosphorylation of HDAC5 is present in the HDAC5 mutants and when AMPK is not added to the reaction, this is likely due to endogenous HDAC5 and AMPK being pulled down during the immunoprecipitation. Both phosphorylation sites share features with other AMPK sites, namely a hydrophobic residue at −5 (Leu/Leu), a basic residue within −1 to −3, and a hydrophobic residue at +4 (Leu/Pro, LRKTAS²⁵⁹EPNLKV, and LSRTQS⁴⁹⁸SPLPQ) (28).

Coimmunoprecipitation from lysates of transiently transfected Cos7 cells identified that both catalytic AMPK α₁ and α₂ subunits associate with HDAC5, consistent with kinase-substrate interactions (Fig. 2B). Alanine mutations of S259 and S498 did not disturb this interaction, demonstrating that the reduction in AMPK-mediated phosphorylation of these mutants was due to the mutation per se and not due to any alterations in protein/protein interactions between AMPK and HDAC5. It also appeared that HDAC5 S/A mutations increased its association with AMPK. While not examined in the present study, this could be due to disrupted phosphorylation-dissociation kinetics caused by the mutations.

From these data, it is clear that S259 and S498 are required for AMPK phosphorylation of HDAC5; however, it is unclear whether these are the sites directly phosphorylated by AMPK. To test this, antibodies were raised recognizing the sequences surrounding S259 and S498 when phosphorylated. Total HDAC5 was immunoprecipitated from lysates of Cos7 cells transiently transfected with wild-type, constitutively active, or dominant-negative AMPK (and regulatory β and γ subunits) and either wild-type or S259/498A HDAC5 and probed with these antibodies. Serine 259 showed high levels of basal phosphorylation (Fig. 2C), consistent with the idea that this site is a hierarchical priming site targeted by a number of kinases and also potentially explaining the weak phosphorylation of this site by AMPK in vitro. Expression of constitutively active AMPK induced phosphorylation of both S259 and S498 above that observed with the expression of wild-type AMPK but increased S498 phosphorylation to a greater extent than the S259 site (Fig. 2C). Furthermore, this was attenuated with expression of dominant-negative AMPK (Fig. 2C). Phosphorylation of HDAC5 at these residues promoted HDAC5 association with 14-3-3 isoforms, as assessed by coimmunoprecipitation, and acetylation at H3K9 and 14. Importantly, these effects were blocked by expression of the S259/498A HDAC5 mutant (Fig. 2C). It is well established that association of HDAC5 with 14-3-3 is required for disruption of the MEF2-HDAC5 complex and HDAC5 nuclear export (29). These results therefore indicate that AMPK phosphorylation of HDAC5 at S259 and S498 is sufficient to induce HDAC5 nuclear export via association with the 14-3-3 chaperone proteins, in turn shifting the chromatin state to one that favored histone acetylation.

To determine whether activation of endogenous AMPK results in HDAC5 phosphorylation, human primary myotubes were treated with AICAR for 60 min. In these samples, AICAR treatment increased AMPK phosphorylation ∼40% (P < 0.05) and increased HDAC5 phosphorylation ∼65% (P < 0.05) and ∼100% (P < 0.05) at S259 and S498, respectively (Fig. 3A). This again showed that AMPK increases phosphorylation at S498 rather than at S259. These results were associated with an ∼55% increase in H3K9 and 14 acetylation (Fig. 3A). To examine the functional relationship between AMPK activation and HDAC5 localization, human primary myotubes were exposed to an acute 60-min treatment of AICAR, and nuclear and whole-cell lysates were analyzed for HDAC5 protein abundance. Following AICAR treatment, a 26% reduction (P < 0.05) in nuclear HDAC5 was observed, without any change in total HDAC5 expression (Fig. 3B), consistent with HDAC5 nuclear export. Immunoblotting for H3 and PDH confirmed nuclear fraction enrichment (Fig. 3B). Taken together, these data suggest that AMPK is an HDAC5 kinase capable of regulating HDAC5 localization and chromatin remodeling by phosphorylating the key regulatory HDAC5 sites, serines 259 and 498.

To investigate whether AMPK can regulate the GLUT4 gene through HDAC5, further GLUT4 reporter gene assays were performed. Cos7 cells were transfected with hG4-Luc, MEF2A, and wild-type HDAC5 plasmids and treated either with a vehicle control (DMSO) or AICAR. AICAR-treated cells showed ∼40% higher (P < 0.05) GLUT4 reporter gene expression than control cells (Fig. 4A), indicating that activation of endogenous AMPK can reduce HDAC5 repression of the GLUT4 reporter gene due to HDAC5 overexpression. Cells transfected with hG4-Luc, MEF2A, and S259/498A HDAC5 plasmids and treated with AICAR showed no increase in GLUT4 reporter gene expression when compared with control transfected cells (Fig. 4B), indicating that serines 259 and 498 are required for the AICAR/AMPK-mediated increase in GLUT4 reporter gene expression. Exclusive mutation of either site produced similar results (supplementary Fig. 4). Treatment of transfected cells with AICAR in the absence of HDAC5 expression had no influence on GLUT4 reporter gene expression when compared with control (Fig. 4C), further supporting the idea that HDAC5 is required for AMPK induction of GLUT4 transcription and is not due to activation of endogenous transcriptional coactivators. This also supports previous observations that AMPK does not phosphorylate and transactivate MEF2 (30). Immunoblotting confirmed activation of AMPK in response to AICAR treatment (supplementary Fig. 5). Collectively, the results obtained from this series of reporter gene assays imply that HDAC5 phosphorylation on serines 259 and 498 by AMPK leads to an increase in GLUT4 gene expression.

To further examine this paradigm in a nonreconstituted system in vivo, ChIP assays were performed on human primary myotubes in the presence or absence of AICAR using primers to amplify a 75-bp region of the GLUT4 promoter surrounding the MEF2 binding domain (Fig. 5A). Using this method, HDAC5 was found associated with this region of the GLUT4 promoter in control cells, suggesting that HDAC5 regulates the GLUT4 gene in vivo. Furthermore, the amount of HDAC5 associated with the GLUT4 promoter was reduced ∼70% (P < 0.05) following AICAR treatment (Fig. 5A). These results could not be attributed to nonspecific effects of the immunoprecipitating antibody, as an unrelated antibody (mouse IgG) immunoprecipitated vastly lower amounts of DNA (>100-fold), nor to any indirect effects of AICAR, as input DNA was not different with this treatment (control 6.2 × 10⁻⁹ ± 1.6 × 10⁻⁹ vs. AICAR 8.2 × 10⁻⁹ ± 1.6 × 10⁻⁹). A reduction in HDAC5 associated with the GLUT4 promoter region was also associated with an approximate twofold increase in GLUT4 mRNA (P < 0.05) in human primary myotubes treated with AICAR (Fig. 5B). These results are consistent with those obtained from the gene reporter studies and suggest that AMPK-mediated HDAC5 phosphorylation regulates the GLUT4 gene.

Given that overexpression of GLUT4 in skeletal muscle ameliorates many of the symptoms associated with type 2 diabetes (5–7), there has been considerable interest in understanding the regulation of GLUT4 expression. From results obtained in the present study, we propose that phosphorylation of HDAC5 on serines 259 and 498 by AMPK is sufficient to induce GLUT4 transcription. These data provide a molecular mechanism linking our previous observations that AMPK translocates to the nucleus (31) while HDAC5 is dissociated from MEF2 and exported from the nucleus (16) with associated increases in GLUT4 gene expression in human skeletal muscle. These findings also describe a signaling pathway linking cellular energy charge to gene transcription directed at restoring cellular and whole-body energy balance. It is also well established that activation of AMPK is associated with enhanced expression of numerous metabolic genes involved in carbohydrate metabolism, lipid metabolism, and the electron transport chain (32), suggesting that HDAC5 could be a key regulator of metabolic gene expression in skeletal muscle. Indeed, many of these genes possess conserved MEF2 binding domains on their gene promoter regions. In the heart, S259/498A mutation of HDAC5 leads to mitochondrial dysfunction and cardiac sudden death secondary to reduced mitochondrial enzyme expression (33). Preliminary work in cultured skeletal muscle cells suggests that HDAC5 is an important mediator of metabolism, as overexpression of HDAC5 represses numerous metabolic genes (S.L.M., K. Baar, unpublished observations).

From data obtained in transgenic kinase dead AMPK expressing mice, it appears that AMPK is not the only kinase capable of regulating the GLUT4 gene. Although the AICAR-induced increase in GLUT4 gene expression is abolished in these mice, they show similar increases in GLUT4 gene expression following exercise when compared with control mice (34). This apparent redundancy could be explained by recent studies identifying members of the family of AMPK-related protein kinases (26,27), as well as CaMKs and protein kinase D (35), as kinases capable of phosphorylating serines 259 and 498 on HDAC5 and removing HDAC5-mediated repression of transcriptional activity. HDAC5 can also be regulated in a manner independent of kinase activity. Interaction of Ca²⁺-calmodulin (36) and protein Gβγ (37) with HDAC5 inhibits HDAC5 and increases MEF2 transcriptional activity. A recent study also suggests that HDAC5 activity can be regulated by the ubiquitin proteasome pathway (38). Furthermore, we cannot exclude the possibility that other members of the class IIa HDACs regulate the GLUT4 gene, given that all isoforms are expressed in skeletal muscle to some degree. Together, these data suggest that AMPK activation via perturbations in energy balance is one of a variety of stimuli that can regulate the GLUT4 gene. Further investigation of these mechanisms is currently underway.

In the present study, we propose that AMPK phosphorylation of HDAC5 relieves transcriptional repression of the GLUT4 gene. AMPK is also known to phosphorylate GEF in vitro (29). Although the specific site(s) on GEF have not yet been identified, and it is unknown whether this effect is preserved in vivo, phosphorylation of both HDAC5 and GEF by AMPK could be a mechanism to exert further control over GLUT4 transcription. Supporting this, we have observed increases in DNA binding activity of both MEF2 and GEF during exercise in human skeletal muscle (39). However, further studies examining the relationship between HDAC5 and GEF regulation are required.

HDAC5 belongs to the class IIa family of HDACs, which also includes isoforms 4, 7, and 9. Isoforms 4 and 7 are also regulated by phosphorylation-dependent nuclear-cytoplasmic shuttling (14,40) through interactions with 14-3-3 isoforms. Sequence analysis suggests that the phosphorylation sites involved in this mechanism are largely conserved between these isoforms. This raises the possibility that AMPK could also regulate HDAC4 and HDAC7. While these HDAC isoforms functionally overlap to regulate a number of common cellular processes, there are also examples of specific functional regulation by these individual isoforms (41,42) and suggest that AMPK might also be involved in such processes. There are other AMPK consensus sites (S279 and S661) in addition to S259 and S498 found within HDAC5 using the AMPK consensus motif. It is unknown whether AMPK phosphorylates these sites in vivo. However, it is known that these sites have no effect on HDAC localization (14). The role, if any, that these sites play in HDAC5 regulation remains to be determined. Furthermore, the interaction between the 259 site, thought to be a priming site for subsequent phosphorylation (26), and the kinases that target this and other sites needs to be determined.

In conclusion, the results from the present study demonstrate that phosphorylation of HDAC5 on serines 259 and 498 by AMPK regulates the GLUT4 gene through a region containing the MEF2 binding site on the GLUT4 gene promoter in human skeletal muscle. These data provide further insights into potential therapeutic targets for the treatment and management of insulin resistance and type 2 diabetes.

We thank Dr. Andrew Garnham, A/Prof. David Cameron-Smith, Dr. Ian Jennings, Dr. Belinda Michell, and Amy Larsen for their assistance with the study and A/Prof. Ann-Louise Olson and Frosa Katsis for providing reagents. B.E.K. is an ARC Federation Fellow, and this work was supported in part by grants from the National Health and Medical Research Council (NHMRC), National Heart Foundation, and ARC (to B.E.K.). S.L.M. is an NHMRC Peter Doherty Fellow (400446).