Class II Histone Deacetylases Downregulate GLUT4 Transcription in Response to Increased cAMP Signaling in Cultured Adipocytes and Fasting Mice

3T3-L1 cells were maintained and transfected via electroporation as previously described (12). Cells were treated with a final concentration of 25 μmol/L forskolin (Calbiochem), 25 μmol/L isoproterenol hydrochloride (Sigma), or 0.1 μmol/L TO-901317 (Sigma).

C57BL/6 mice were used for all experiments. In some experiments, transgenic mice were used that carried a human GLUT4 promoter/chloramphenicol acetyltransferase (CAT) reporter construct that is fully functional (895-hG4-CAT), and another line carrying a similar reporter with a loss of function mutation in the GLUT4 LXRE, as previously described (10). All mice were kept on a 12-h light/dark cycle in a temperature- and humidity-controlled room with free access to water and standard chow. Food was removed at ∼5 p.m. the night before when mice were challenged with fasting, but they had free access to water. The fasted animals were killed the following morning and tissues were isolated, flash-frozen, and stored for further analysis.

Obesity was induced by ad libitum feeding with a 60% lard diet (Research Diets) for 8 weeks. For these studies, nonobese control animals were maintained by ad libitum feeding with a 10% fat diet (Research Diets) for the same period. All procedures using animals were approved by the Institutional Animal Care and Use Committee at the University of Oklahoma Health Sciences Center.

Cells in experiments using small interfering (si) RNA were transfected as previously described (12).

Chromatin immunoprecipitation (ChIP) reactions were performed as previously described (10) for cultured adipocytes with one modification: the α-LXR ChIP was performed with an additional crosslinking step using 2 mmol/L disuccinimidyl glutarate in 1× PBS and 1 mmol/L MgCl₂ for 45 min before crosslinking with 1% formaldehyde. For adipose tissues from fed or fasted mice, the following modifications were made: ∼200 mg of adipose tissue was placed in Ham’s F12 media containing 1% formaldehyde. Tissue was homogenized briefly with the Tissue Tearor and incubated at room temperature for 15 min, rocking end-over-end. The crosslinking was stopped with 0.125 mol/L glycine for 5 min. The nuclei were pelleted by centrifugation at 13,000 rpm for 15 min. The nuclear pellet was resuspended in high salt lysis buffer (Santa Cruz Biotechnology), sonicated, and treated the same as cultured cells for the rest of the procedure (10). DNA recovered from ChIP reaction was subjected to quantitative PCR (q-PCR) and analyzed as percent of input and normalized to nonimmune rabbit IgG (Cell Signaling). All q-PCR analysis were run using a CFX96 real-time (RT)-PCR detection system thermal cycler (Bio-Rad).

Samples were treated as previously described (12). Denatured samples were fractionated by SDS-PAGE using 10% polyacrylamide gels for cAMP-responsive element–binding (CREB) and phospho-CREB blots and 7.5% polyacrylamide gel for HDAC blots and transferred to polyvinylidene fluoride membranes. Membranes were probed with rabbit α-CREB (Cell Signaling), rabbit α-phospho–CREB (Cell Signaling), rabbit polyclonal α-HDAC5 antibodies (Millipore), or rabbit polyclonal α-HDAC4 (Cell Signaling) antibodies overnight at 4°C. After incubation with labeled secondary antibodies, membranes were quantified using the Odyssey Infrared Imaging System (Li-Cor Biosciences).

RNA was extracted and prepared as previously described (12). Primer sequences for GLUT4 and CAT mRNA were as described (10). All q-RT-PCRs were run using a CFX96 RT-PCR detection system thermal cycler (Bio-Rad).

3T3-L1 adipocytes were differentiated and electroporated as previously described (12). Luciferase assays were performed 24 h after transfection using the Dual-Glo luciferase kit (Promega).

Mean and standard error of the mean is reported. Differences were analyzed by a one-way ANCOVA or Student t test.

We previously established that multiple class II HDACs are capable of regulating the GLUT4 promoter (12). Although we demonstrated that HDAC4, -5, and -9 each inhibited both GLUT4 enhancer factor (GEF)– and myocyte-specific enhancer factor 2A (MEF2A)–dependent gene transcription in vitro, simultaneous siRNA-mediated knockdown of all three HDACs was required to mediate changes in GLUT4 gene expression in vivo. These data suggest a possible redundant or compensatory function between the HDAC isoforms (12). To test this possibility, we over-expressed siRNA-resistant forms of HDAC4 and HDAC5 to rescue the inhibitory function in 3T3-L1 preadipocytes in which HDAC4, HDAC5, and HDAC9 were knocked down (Fig. 1A). As previously described, cells were transiently transfected with a combination of HDAC4-, HDAC5-, or HDAC9-specific siRNA (Ambion) or a nonspecific siRNA 72 h before the rescue (12). Cells were transfected again with either empty vector or transactivating factors (MEF2A and GEF) to activate the endogenous promoter, and plasmids encoding either human HDAC4 or HDAC5 as indicated (Fig. 1A). Endogenous GLUT4 mRNA was inhibited by overexpressing each HDAC4 or HDAC5 individually, indicating that each of these class II HDACs can inhibit GLUT4 expression in preadipocytes.

Next, we used an in vitro transcription assay to quantify promoter activity after rescue of the class II HDAC knockdown (12). Preadipocytes were transiently transfected with scrambled or HDAC-specific siRNA for 72 h, as described above. After 72 h, cells were transfected a second time with the minimal functional human GLUT4 promoter driving the expression of a firefly luciferase gene (-895-hGLUT4-luc), as previously characterized (11). Activity of the human GLUT4 promoter in transient transfection assays is low compared with the activity of the promoterless reporter. To boost the signal, the reporter construct was cotransfected with MEF2A and GEF (11). To recover HDAC function, cells were cotransfected with human HDAC4, -5, or -9 as indicated (Fig. 1B). Transfection efficiency was accounted for by normalization to the control plasmid Renilla luciferase (pRLTK-luc), for transfection efficiency. As previously described, knockdown of the class II HDACs in preadipocytes, GLUT4 expression increases significantly (Fig. 1B) (12). Rescue of the knockdown was achieved by overexpression of human HDAC4, -5, or- 9, suggesting redundant function. To begin determining which HDAC isoform(s) are physiologically relevant, we used a ChIP to determine if HDAC4 and/or HDAC5 (antibodies for HDAC9 are not available for ChIP) were bound to the GLUT4 promoter in preadipocytes in vivo. When cells were treated with scrambled siRNA, we observed that HDAC5 bind the GLUT4 promoter, whereas HDAC4 binding was no different than nonspecific IgG (Fig. 1C). This indicated that HDAC5 preferentially binds to the GLUT4 promoter in a preadipocyte. Only when cells were treated with siRNA specific for HDAC5 and HDAC9 was there a significant increase in HDAC4 binding to the GLUT4 promoter. Taken together, our data strongly support a model in which HDACs can be functionally redundant but do not directly compete for binding to the GLUT4 promoter in preadipocytes.

A role for HDAC regulation of GLUT4 expression in mature adipocytes has not been studied. We have previously shown that HDAC4 and HDAC5 are expressed in 3T3-L1 adipocytes but are largely excluded from the nuclei as a function of differentiation (11,12). This raised the possibility that HDAC4 or HDAC5 can be redistributed to the nucleus in 3T3-L1 adipocytes in response to specific intracellular signals. For example, cAMP has been implicated in HDAC5 redistribution into the nucleus from the cytosol in neurons (13). To test if this occurs in adipocytes, we elevated cAMP levels with 25 μmol/L forskolin. Forskolin-dependent cAMP signaling was confirmed by Western blotting analysis of phospho-CREB. Forskolin treatment increased phospho-CREB levels by 30 min, and they remained elevated for at least 6 h (Fig. 2A). We next tested the effect of forskolin treatment on HDAC localization. 3T3-L1 adipocytes were treated with 25 μmol/L forskolin for the indicated times. Nuclear extracts were prepared and subjected to Western blotting analysis for HDAC4 and HDAC5 (Fig. 2B). HDAC4 and HDAC5 nuclear content both increased as a result of forskolin treatment. To determine whether nuclear accumulation resulted in increased binding to the GLUT4 promoter, ChIP was performed using antibodies against HDAC4 or HDAC5 (Fig. 2C). The ChIP experiments revealed that HDAC4 binding to the GLUT4 promoter was preferentially increased as a result of forskolin treatment.

If forskolin-dependent changes in HDAC4 binding regulate GLUT4 transcription, then forskolin treatment should decrease GLUT4 mRNA and GLUT4 transcriptional activity. To test this prediction, we measured GLUT4 mRNA and found that forskolin treatment overnight resulted in a 60% reduction of GLUT4 mRNA (Fig. 2D). Next, we assayed for transcriptional activity by transiently transfecting adipocytes using the -895-hGLUT4-luc reporter assay (11,12). The -895-hGLUT4-luc reporter was activated by cotransfection with MEF2A and GEF and treated without or with 25 μmol/L forskolin for the duration of the transfection period to ensure that cAMP was elevated while transcription of the reporter construct was occurring. As expected, cotransfection of -895-GLUT with MEF2A and GEF increased promoter activity sevenfold. Forskolin treatment resulted in a decrease (about fourfold) in transcriptional activity despite the presence of activating transcription factors (Fig. 2E).

To activate cAMP using a more physiologic mechanism, we used isoproterenol, a β-adrenergic receptor agonist, to initiate cAMP signaling. We performed a time-course to measure phospho-CREB. Similar to forskolin, isoproterenol increased phospho-CREB by 30 min of treatment, and it remained elevated for at least 3 h (Fig. 3A). Isoproterenol treatment increased nuclear accumulation of both HDAC4 and HDAC5 (Fig. 3B). We next performed ChIP assays to determine if HDAC4 and HDAC5 binding to the GLUT4 promoter also increased as a result of isoproterenol treatment. The ChIP assay revealed that the association with the GLUT4 promoter increased in both HDAC4 and HDAC5, with HDAC4 increasing more than HDAC5 (Fig. 3C).

Similar to treatment with forskolin, overnight treatment with isoproterenol reduced GLUT4 by ∼50% (Fig. 3D). To confirm that this is a transcriptional effect, we tested isoproterenol to regulate expression of the -895-hGLUT4-luc transcriptional report. Overnight treatment with isoproterenol inhibited GEF- and MEF2A-dependent activation of -895-hGLUT4-luc (Fig. 3E). Cotransfection with plasmids encoding HDAC4 or HDAC5 inhibited transcriptional activation of -895-hGLUT4-luc, with no further potentiation of the effect when isoproterenol was added (Fig. 3E).

To determine if isoproterenol-dependent changes in HDAC association with the GLUT4 reporter are responsible for the changes in GLUT4 transcription, we used siRNA knockdown of HDAC to determine if HDAC protein expression was required for this effect. To test this, preadipocytes were transiently transfected with scrambled siRNA or HDAC-specific siRNA 72 h before the assay under conditions (i.e., knockdown of HDAC4/5/9) previously shown to increase GLUT4 mRNA in preadipocytes (12). After 72 h, cells were transfected a second time with the -895-hGLUT4-luc plasmid to assay for GLUT4 transcriptional activity. The -895-hGLUT4-luc was transactivated with plasmids encoding MEF2A and GEF and treated without or with 25 μmol/L forskolin or 25 μmol/L isoproterenol as indicated (Fig. 4). We observed that forskolin- and isoproterenol-dependent inhibition was abolished when HDACs were knocked down. Our data indicate that HDAC proteins must be present for either compound to inhibit GLUT4 transcription. Taken together, these data support our hypothesis that forskolin and isoproterenol treatment alters GLUT4 expression through HDAC association with the GLUT4 promoter, likely through a cAMP-dependent signaling pathway.

Using forskolin and isoproterenol treatment, we have shown that HDAC4 and HDAC5 associate with the GLUT4 promoter to reduce the rate of transcription. We next determined if HDAC4 and HDAC5 associate with the GLUT4 promoter in vivo when GLUT4 mRNA is downregulated. To test this, we subjected mice to an overnight fast, a condition known to decrease GLUT4 gene expression (2,4,6,14–16). Mice were allowed to have free access to food (fed) or were fasted overnight and killed. White adipose tissue and total proteins were extracted for Western blotting analysis of phospho-CREB (Fig. 5A). Mice fasted overnight demonstrated a significant increase in phospho-CREB levels compared with fed control mice, while not displaying any significant changes in total CREB, consistent with elevated cAMP signaling. Overnight fasting reduced GLUT4 mRNA in perigonadal white adipose tissue by 50% (Fig. 5B), which is consistent with previous literature (8,16–18).

Next, we determined if the changes in GLUT4 mRNA expression under the fasting protocol correlated with changes in class II HDACs association with the GLUT4 promoter. Nuclei were prepared from adipose tissue of mice that were fasted overnight or were allowed free access to food. We did not observe a difference between nonspecific IgG and HDAC4 or HDAC5 binding to the GLUT4 promoter in animals that were allowed access to food overnight (Fig. 5C). However, when animals were challenged with overnight fasting, we observed a large increase of both HDAC4 and HDAC5 binding to the GLUT4 promoter. Similar to results observed in 3T3-L1 adipocytes challenged with either forskolin or isoproterenol, HDAC4 binding was higher than HDAC5.

Previous studies have demonstrated that the nuclear orphan receptor LXR activation is linked to HDAC-mediated transcriptional repression (19). Work from our laboratory has demonstrated a role for LXR in GLUT4 transcriptional regulation (10,20). Therefore, we tested the role of LXR signaling on isoproterenol-mediated inhibition of the GLUT4 promoter. Toward this end, we transiently transfected adipocytes with the -895-hGLUT4-luc or Δ-LXR-hGLUT4-luc construct, as indicated (Fig. 6A). The Δ-LXR-hGLUT4-luc construct has a loss of function mutation in the LXRE sites on the GLUT4 promoter, as previously characterized (10). The luciferase constructs were activated with transcription factors (MEF2A and GEF) and treated with isoproterenol alone or with TO901317 together (Fig. 6A). We observed a significant increase in activity when both constructs were activated with transcription factors. Cells treated with isoproterenol alone displayed a statistically significant decrease in GLUT4 transcriptional activity of the -895-hGLUT4-luc. However, the Δ-LXR-hGLUT4-luc was not affected by isoproterenol treatment, indicating that LXR plays a key role in cAMP-dependent inhibition of the GLUT4 promoter. To confirm the role of LXR, we treated the cells with the LXR agonist TO901317 and with isoproterenol. Interestingly, GLUT4 transcriptional activity was restored, despite the presence of isoproterenol. The LXR agonist, TO901317, was able to overcome the inhibitory effect of isoproterenol-dependent regulation of GLUT4 mRNA.

Because TO901317 relieved the inhibition of GLUT4 transcription mediated by isoproterenol, we next determined if TO901317 changed HDAC binding to the GLUT4 promoter. We treated the cells with TO901317 or isoproterenol, or both, and performed ChIP analysis. We observed a large increase in HDAC4 and HDAC5 binding to the GLUT4 promoter in response to isoproterenol (Fig. 6B). The addition of TO901317 prevented the isoproterenol-dependent increase in HDAC binding to the GLUT4 promoter (Fig. 6B). In contrast to HDACs, LXR binding to the LXRE was reduced by isoproterenol treatment and returned to normal when treated with TO901317 (Fig. 6B). TO901317 did not affect the accumulation of HDAC4 or HDAC5 in the nucleus (Fig. 6C). TO901317 treatment had no effect on the induction of phospho-CREB (data not shown).

Taken together, we have demonstrated that elevated cAMP levels can downregulate the GLUT4 promoter by inducing HDAC4 binding to the GLUT4 promoter through a mechanism that requires LXRE but can be inhibited by treatment with TO9001317. These data suggest that the regulated expression of GLUT4 during the transition from fasting to feeding is regulated by cAMP signaling through the LXRE. To confirm that LXRE plays a physiologic role in modulation of GLUT4 mRNA, we tested the requirement for the GLUT4 LXRE in two models, fasting and diet-induced obesity, that are accompanied by changes in GLUT4 expression. To estimate GLUT4 transcription in the mouse model, we used transgenic mice engineered to carry a fully functional GLUT4 promoter/reporter construct (-895-HG4-CAT) or a construct with a loss of function mutation in the GLUT4 LXRE (Δ-LXR-CAT) (Fig. 7A) (10). Animals were fed a control diet (10% fat) or a high-fat diet (60% fat) for 8 weeks. The control-diet animals were deprived of food overnight or were fed ad libitum until they were killed (serving as the control for both experimental groups). CAT and endogenous GLUT4 mRNA levels were quantified from perigonadal adipose tissue. As expected, endogenous GLUT4 mRNA was decreased in perigonadal adipose tissue in response to both fasting and diet-induced obesity. Transgenic mRNA levels under the control of the -895-HG4 promoter changed in parallel with endogenous GLUT4 (Fig. 7B). In contrast, the loss of function mutation in the GLUT4 LXRE was unresponsive to fasting or diet-induced obesity, and transgenic mRNA levels were unchanged (Fig. 7B).

GLUT4 gene expression in adipose tissue is sensitive to changes in metabolic state, including the fasting and diabetes (8,18,21). Understanding how the GLUT4 gene promoter responds to changes in metabolism is a key to understanding the molecular events that underlie this important adaptation to changing physiology. Using transgenic mice to elucidate the key regulatory domains present in the GLUT4 promoter, we determined the MEF2 binding domain and domain 1 were both required for GLUT4 gene transcription (9,22).

We demonstrated that the MEF2 and GEF transcription factors work synergistically to transactivate the GLUT4 promoter and that both are necessary but not sufficient to drive GLUT4 transcription (9,22). These elements, when ligand bound, produce a scaffold that allows the binding of coactivators and corepressors to modulate the GLUT4 promoter in response to altered metabolic states. Among the potential corepressors that bind this scaffold are the class II HDACs (11,12).

In this report, we provide several lines of evidence supporting the notion that HDACs can be functionally redundant with regard to GLUT4 transcription but have isoform-specific roles based on the stage of differentiation. We, along with others, have observed that class II HDACs potentially compensate for each other, resulting in the requirement to knockdown multiple HDACs to achieve a phenotype (12,23,24). Therefore, we sought to determine the conditions under which HDACs compensate for each other. We demonstrated that class II HDACs, specifically HDAC4 and 5, are functionally redundant in regards to the GLUT4 promoter in preadipocytes; however, the HDAC5 preferentially binds the GLUT4 promoter in preadipocytes. HDAC4 only binds to the GLUT4 promoter in the preadipocyte stage when HDAC5 is knocked down. Therefore, despite the structural similarities, HDAC4 does not compete with HDAC5 to regulate the GLUT4 promoter in preadipocytes. Rather, HDAC4 can only bind the GLUT4 promoter in preadipocytes when HDAC5 and HDAC9 are depleted. The specific contribution of HDAC9 is not yet clarified because suitable antibodies for ChIP and Western blotting analysis were not available.

The role of HDAC proteins in regulation of GLUT4 in differentiated adipocytes is different from preadipocytes. Early work on GLUT4 transcriptional regulation demonstrated a link between elevated cAMP levels and decreased GLUT4 expression (17,18). Recent work showed that class II HDACs changed their localization pattern in response to elevated cAMP; therefore, we reasoned that class II HDACs regulate the GLUT4 promoter in differentiated adipocytes via activation of cAMP. To this end, we determined that HDAC4 and HDAC5 increased their association with the GLUT4 promoter in response to stimuli (forskolin or isoproterenol) that increased intracellular cAMP. Unlike preadipocytes, HDAC4 and HDAC5 binding to the promoter both increased in circumstances in differentiated adipocytes. Thus, there appears to be a clear, distinct role for both HDAC4 and 5 at different developmental stages.

To ensure that isoproterenol and forskolin were both working through class II HDACs, we used HDAC-specific siRNA knockdown in conjunction with the treatments in our luciferase assay (Fig. 4). This effectiveness of the siRNA-mediated knockdown of HDACs has been previously characterized (12), and the effect of both treatments on cAMP has also been demonstrated (Fig. 2 and 3). By knocking down HDACs and abolishing the inhibitory effect of either treatment, we firmly established the role of HDACs in cAMP-mediated GLUT4 repression.

Although our in vitro data are clear, we needed to establish the physiologic relevance of the finding. Isoproterenol activates a counter-regulatory hormone response, which would be activated in the fasted state, a condition that leads to downregulation of GLUT4 expression. We subjected mice to an overnight fast to test the effects on cAMP activation, GLUT4 mRNA, and HDAC association with the GLUT4 promoter. As expected, fasting increased cAMP signaling and decreased GLUT4 mRNA. More importan, we found that HDAC4 and HDAC5 binding to the GLUT4 promoter increased in vivo, similarly to what we observed in vitro (compare Figs. 2C, 3C, and 5C).

In this report, we have demonstrated that elevated cAMP negatively regulates the GLUT4 promoter via HDACs. How cAMP signals HDAC redistribution in adipocytes to the nucleus is unclear. However, during times of metabolic stress like fasting, the need for glucose storage in adipose tissue is suppressed while enzymes regulating fatty acid release are activated, a function consistent with LXR signaling. The GLUT4 promoter contains an LXRE binding domain, just eight base pairs downstream of the MEF2 binding domain. Our laboratory previously established a role for LXR in GLUT4 regulation in response to inhibiting β-oxidation (10). We have now determined that cAMP signaling, fasting, and adaptation to obesity require the intact LXRE to specifically downregulate GLUT4 gene transcription (Figs. 6 and 7). We observed that treatment of cells with TO901317, an LXR agonist, completely reversed the effects of isoproterenol on GLUT4 promoter activity and HDAC4 and HDAC5 association with the GLUT4 promoter. TO901317 treatment did not reverse the nuclear accumulation of HDAC4 or HDAC5 in isoproterenol-treated cells (Fig. 6C); however, it did reverse the isoproterenol-dependent loss in the LXR association with the GLUT4 promoter (Fig. 6B). This suggests that the cAMP-dependent increases in HDAC association with the GLUT4 promoter are mitigated by the LXRE ligands.

A recent study demonstrated that HDACs can bind and downregulate LXR-dependent gene transcription through a small ubiquitin-like modifier (SUMO)ylation of LXR (19). We have shown that treatment with TO901317 increases the LXR-α association with the GLUT4 promoter (Fig. 6B[10]); therefore, cAMP is not likely to be acting through the SUMOlyation pathway. The LXRE is clearly required for downregulation of GLUT4 mRNA in vitro and in vivo given that the loss of function mutation even though it no longer supports downregulation of the gene. This suggests that when LXR binding to the LXRE is decreased, another ligand may bind and that this ligand may facilitate HDAC association with the GLUT4 promoter. This model will be tested in future studies in our laboratory.

In summary, we have demonstrated that despite the functional redundancy of HDACs, HDAC4 and HDAC5 have physiologically distinct functions at different stages of adipocyte differentiation to regulate the GLUT4 promoter. Furthermore, we have demonstrated that cAMP-dependent downregulation of GLUT4 mRNA expression in vivo is dependent on nuclear localization of HDAC4 and HDAC5 and the presence of a functional LXR binding element in the GLUT4 promoter. This offers a new understanding of how cAMP-signaling pathways alter gene expression in adipose tissue.

This work was supported by a grant from the National Institutes of Health (DK 081545).

No potential conflicts of interest relevant to this article were reported.

J.C.W. researched data and wrote the manuscript. B.A.G. researched data. A.L.O. researched data and reviewed and edited the manuscript. A.L.O. is the guarantor of this work and, as such, had full access to all the data in the study and takes full responsibility for the integrity of the data and the accuracy of the data analysis.