Gβγ Binds Histone Deacetylase 5 (HDAC5) and Inhibits Its Transcriptional Co-repression Activity*

In a yeast two-hybrid screen designed to identify novel effectors of the Gβγ subunit of heterotrimeric G proteins, we found that Gβγ binds to histone deacetylase 5 (HDAC5), an enzyme involved in a pathway not previously recognized to be directly impacted by G proteins. Formation of the Gβ1γ2-HDAC5 complex in mammalian cells can be blocked by overexpression of Gαo, and this inhibition is relieved by activation of α2A-adrenergic receptor, suggesting that the interaction occurs in a signal-dependent manner. The C-terminal domain of HDAC5 binds directly to Gβγ through multiple motifs, and overexpression of this domain mimics the C terminus of G protein-coupled receptor kinase 2, a known Gβγ scavenger, in its ability to inhibit the Gβγ/HDAC5 interaction. The C terminus of HDAC4 shares significant similarity with that of HDAC5, and accordingly, HDAC4 is also able to form complexes with Gβ1γ2 in cultured cells, suggesting that the C-terminal domain of class II HDACs is a general Gβγ binding motif. Activation of a Gi/o-coupled receptor results in a time-dependent activation of MEF2C, an HDAC5-regulated transcription factor, whereas inhibition of the interaction with a Gβγ scavenger inhibits MEF2C activity, suggesting a reduced potency of HDAC5-mediated inhibition. Taken together, these data imply that HDAC5 and possibly other class II HDACs can be added to the growing list of Gβγ effectors.

Signaling through G protein-coupled receptors (GPCRs) 3 is a ubiquitous mechanism mediating cellular responses to such diverse stimuli as photons of light, odorants, and hormones. Information from such stimuli is initially transduced through heterotrimeric G proteins; activated receptors catalyze the dissociation of the G␣ and G␤␥ subunits and, thus, modulation of the subunits' downstream effectors (1). In addition to regulating the activation state of a partner G␣ subunit, the G␤␥ functional monomer interacts with an increasingly apparent number of signaling proteins (2), including well known effectors such as phosphoinositide 3-kinases and phospholipases C and more recently discovered interacting partners such as the ubiquitin-related protein PLIC-1 (3) and the glucocorticoid receptor (4).
In a recent effort to expand our understanding of G␤␥ function, we conducted a yeast two-hybrid screen seeking to identify novel G␤␥-interacting proteins. Results from this screen led to the identification of an isoform of the receptor for activated C kinase (RACK1) and several other WD40-repeat containing proteins as G␤␥ binding partners (5). These results led to the hypothesis that interactions between WD40 domains may be a general phenomenon and that the WD40 domain may serve in part to increase the speed and specificity of signaling events by acting as a versatile scaffold in diverse signaling pathways (5,6). Importantly, RACK1 was identified as an effector-specific modulator of G␤␥ signaling in vivo (7).
An emerging paradigm in the field of GPCR function that is typified by the RACK1/G␤␥ interaction is the role of cross-talk with other signaling paradigms. For example, in its ability to interact with molecules downstream of receptor tyrosine kinases and transforming growth factor receptors, RACK1 may serve as a nexus mediating the cross-talk between numerous types of pathways (8). Furthermore, both G␤␥ and receptor tyrosine kinases modulate the activity of the ␤ isoform of phosphoinositide 3-kinase (for review, see Ref. 9), and GPCR activation can result in transactivation of receptor tyrosine kinases in the modulation of mitogen-activated protein kinase (MAPK) cascades (for review see 10).
One potential signaling partner for G protein pathways that has been largely unexplored is protein acetylation, a signaling mechanism of increasingly apparent importance (11). The most well studied acetylation events govern access to the information encoded in the genome. At appropriate times, the acetylation states of histones associated with specific genes are controlled by a plethora of histone acetyltransferases (HATs) and histone deacetylases (HDACs). By altering local chromatin structure and providing binding sites for numerous transcription factors, HATs typically act to increase transcription at a specific site, whereas HDACs often act as transcriptional co-repressors by opposing the actions of HAT enzymes (12).
Two types of enzyme catalyze histone deacetylation, namely the NAD ϩ -dependent sirtuins and the Zn 2ϩ -dependent HDACs (13,14). The latter family is subdivided into at least two classes. Class I HDACs (HDACs 1, 2, 3, and 8) are most similar to the yeast regulator of potassium dependence 3 (Rpd3p). These enzymes are ubiquitously expressed in mammalian tissues and cell lines. Homologous to the class I HDACs are class II HDACs, which are most similar to a yeast HDAC known as Hda1p. Class II HDACs (HDACs 4, 5, 6, 7, 9, 10, and 11) generally have more restricted distributions than those of class I; HDAC5, for example, is most highly expressed in cardiac smooth muscle, skeletal muscle, and brain (15,16).
In addition to distinct tissue distributions, a subset of class II HDACs are characterized by N-terminal extensions with well studied roles in protein-protein interactions (17,18) and subcellular distribution. In addition to their long (ϳ400 -600 amino acid) N-terminal domains, some class II HDACs also have a relatively short (ϳ150 amino acid) and well conserved domain at their extreme C termini. In addition to a signal-responsive nuclear export signal (19), other functions of this domain are emerging (20).
The current report investigates a role for the interaction of G␤␥ with class II histone deacetylases, including HDAC5, which is the second novel G␤␥ binding partner discovered in the previous yeast two-hybrid screen. Interestingly, the C terminus of a related class II HDAC, HDAC7, has been shown to interact with a GPCR, endothelin receptor A, and possibly to modulate an endothelin-1-stimulated MAPK pathway (20). Here, we show that G␤␥ interacts with the C-terminal domain of HDAC5 and provide evidence that signaling through G␤␥ has an inhibitory effect on HDAC5 function. Thus, HDAC5 may be added to the long and growing list of effectors of the signaling capacity of G␤␥.

EXPERIMENTAL PROCEDURES
Materials-Rabbit anti-pan-G␤ (T20) and a control irrelevant antibody (rabbit anti-His probe (H15)) were from Santa Cruz Biotechnology. Mouse anti-FLAG (M2) and anti-HA antibodies and antibody-Sepharose conjugates were from Sigma. HEK293A cells (Invitrogen) were passaged at 37°C in a 5% CO 2 atmosphere in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Invitrogen) and penicillin/streptomycin. Spodoptera frugiperda (Sf9) (Invitrogen) was grown in Sf-900 II SFM medium (Invitrogen) at 28°C with gentle shaking. Trichostatin A (TSA) and other chemicals were of the highest grade available and were typically from Sigma.
Endogenous Immunoprecipitations and Deacetylase Assays-Heart tissue from two normal adult Sprague-Dawley rats was lysed with 15 strokes of a tight-fitting Dounce homogenizer in phosphate-buffered saline with 1% deoxycholate, 1% Igepal, 0.5% SDS, 5 mM dithiothreitol, and protease inhibitors (lysis buffer). The soluble portion of this lysate was pre-cleared by incubation with protein G-Sepharose beads. Four micrograms of purified rabbit polyclonal antibodies against G␤ or the His probe were added, and the mixtures were incubated on ice overnight. A 50-l bed volume of protein G-Sepharose was added, and the tubes were incubated with gentle mixing at 4°C for 1 h. Immunoprecipitates were collected via centrifugation and washed twice with 1 ml of lysis buffer and once with phosphate-buffered saline (PBS). Ten percent of the resin was removed for Western blot analysis. The remaining resin was assayed for deacetylase activity using a commercially available assay kit essentially as described (Upstate).
Plasmid Construction-Plasmids directing the expression of bovine G␥ 2 and N-terminally HA-tagged bovine G␤ 1 in mammalian cells under the control of the human cytomegalovirus promoter were created using conventional molecular biological techniques. Plasmids directing the expression of C-terminally FLAG-tagged human HDACs 4, 5, and 6 were generously provided by Dr. Ed Seto, University of South Florida. A plasmid directing the expression of G␣ oA with an internal EE epitope tag was obtained from the Guthrie cDNA Resource Center. The ␣ 2A -adrenergic receptor expression vector was obtained from Dr. Lee Limbird, Vanderbilt University.
Viruses were created and amplified according to the Bac-to-Bac system (Invitrogen).
Protein Purification-Sf9 cells (1 ϫ 10 6 ) were infected at an approximate multiplicity of infection of one with baculovirus GST-HDAC5ct. Following a 60-h incubation at 28°C, cells were collected by centrifugation. The cells were swollen for 20 min on ice in hypotonic buffer (20 mM Tris pH 8.0, 5 mM MgCl 2 , and 1 mM phenylmethylsulfonyl fluoride). The cells were lysed using a hand-held rotary homogenizer. NaCl was added to a final concentration of 150 mM, and the lysates were cleared by centrifugation. Glutathione-Sepharose purification proceeded as recommended (Amersham Biosciences), except that proteins were eluted overnight at 4°C with 50 mM Tris, 20 mM glutathione, 250 mM NaCl, 5 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride. The proteins were concentrated, and glutathione was removed by buffer exchange with PBS using centrifugal concentrators (Millipore). G␤ 1 ␥ 1 was purified from bovine rod outer segments as described (21).
In Vitro Binding Assays-GST or GST-HDAC5ct were incubated at noted concentrations in a final volume of 100 l with G␤ 1 ␥ 1 . Binding buffer consisted of PBS with 5 mM dithiothreitol, 5 mM ethylenediamine tetraacetic acid, and 0.2 mg/ml bovine serum albumin. After a 1-h incubation at 4°C, 30 l of a 50:50 slurry of glutathione-Sepharose resin was added to each reaction. The complexes were allowed to bind to the resin for 30 min at 4°C with constant mixing. The resin was collected by brief centrifugation and then washed three times with 1 ml of PBS plus 0.2% Tween 20 and 5 mM dithiothreitol. The complexes were eluted from the resin with SDS-PAGE running buffer and separated by electrophoresis.
Cell Culture and Transfections-Cells were transfected using Lipofectamine 2000 (Invitrogen). Typically, for a 6-well plate at 80 -90% confluence, 1 g of total plasmid DNA was complexed with 5 l of lipid reagent in 500 l of Opti-MEM I (Invitrogen). After a 20-min incubation at room temperature, the complexes were added to the cells with 2 ml of Dulbecco's modified Eagle's medium and 10% serum but no antibiotics. For some transfections, empty vector (pcDNA3.1) was added to keep total DNA at 1 g.
Immunoprecipitations-Forty-eight hours after transient transfections, cells were washed with PBS, and 400 l of PBS with 0.5% Igepal, 0.1% SDS, protease inhibitors, and 1 mM dithiothreitol (lysis buffer) was added. Following a 5-min incubation at 4°C, the cells were recovered with a cell lifter. The cells were lysed further by 10 passages through a 23-gauge needle with a 1-cc syringe. Cellular debris was removed by centrifugation and a pre-clearing step using protein G-Sepharose (Amersham Biosciences). Complexes were immunoprecipitated for 1 h at 4°C using either anti-FLAG (M2)-agarose or anti-HA agarose (Sigma). The pellets were washed three times with 1 ml of lysis buffer. Proteins were eluted with SDS-PAGE loading dye prior to electrophoresis. Typically, one-half of the supernatant was analyzed by electrophoresis and Western blot.
Western Blots-Samples separated with SDS-PAGE were transferred to polyvinylidene fluoride membranes (Millipore). Primary antibodies were detected with specific secondary antibodies covalently linked to Alexa Fluor 680 (Invitrogen) or IRDye800 (Rockland, Inc.). Fluorescent signal was imaged with an Odyssey infrared imaging system (LI-COR Inc.).
Assays for Reporter Gene Transfection-Plasmids encoding MEF2C and luciferase downstream of three MEF2C binding sites (3ϫ MEF2luciferase) were kindly provided by Dr. Eric Olson (22). LacZ under control of a cytomegalovirus promoter (Invitrogen) was included as a transfection control. HEK293A cells on 12-well plates were transfected with 0.1 g of MEF2C, 0.2 g of 3ϫ MEF2-luciferase, 0.0125 g of CMV-LacZ, and other plasmids as noted using Lipofectamine 2000 transfection reagent as described above. Total DNA was adjusted to 0.5 g with empty vector (pcDNA3.1). Two days after transfection, luciferase and ␤-galactosidase were harvested with 250 l of passive lysis buffer as recommended (Promega). Luciferase and ␤-galactosidase activities were determined with reagents obtained from Promega. Luciferase output was measured with a Wallac Victor 2 V 1420 multi-label HTS counter equipped with a 700-nm infrared cutoff filter. Measurements were linear with respect to time and dose of extract.

RESULTS
Yeast Two-hybrid Screen-Bovine G␤ 1 was used as bait to screen a mouse brain library in a previously described yeast two-hybrid assay (5). Two clones whose expression resulted in reporter gene transcription were found to encode the C-terminal ϳ160 amino acids of HDAC5 (data not shown). This result was confirmed by retransformation of the plasmids and co-immunoprecipitation from yeast lysates as described (5) (data not shown). In addition, the interaction was found to be independent of the co-expression of mammalian G␥ (data not shown), suggesting that G␤ 1 formed a functional dimer with a yeast G␥-like protein in situ and that binding was mediated by G␤ rather than G␥.
Investigating an in Vivo Interaction-Because of an apparent insensitivity of several anti-HDAC5 antibodies in immunoblotting and immunoprecipitation experiments, the HDAC content of immunoprecipitated G␤ complexes was assessed with in vitro deacetylase assays to confirm the association of G␤ with HDACs in situ. Lysates of heart tissue from normal adult rats were subjected to immunoprecipitation with a rabbit polyclonal anti-G␤ antibody or a control irrelevant rabbit polyclonal antibody. The anti-G␤ antibody specifically precipitated G␤, as shown in Fig. 1A. As displayed in Fig. 1B, the anti-G␤ antibody precipitates displayed significantly more deacetylase activity than did the control precipitates.
To functionally characterize the activity present in the complex, in vitro assays were performed in the presence of TSA, which is a selective inhibitor of the Zn 2ϩ -dependent HDAC family of deacetylases. No activity was detected in control or anti-G␤ immunoprecipitates (Fig.  1B), suggesting that HDAC enzymes exist in stable complexes with G␤ in vivo.
Interaction with HDAC5 in Particular-Coimmunoprecipitation of epitope-tagged proteins expressed in cultured cells was employed to confirm that G␤ could interact with HDAC5. First, empty vector or N-terminally HA-tagged G␤ 1 and untagged G␥ 2 were transiently transfected into HEK293A cells with C-terminally FLAG-tagged HDAC5. Two days after transfection, cell lysates were subjected to immunoprecipitation with an anti-HA antibody. As shown in Fig. 2A, HA-G␤ 1 was specifically purified when HDAC5-FLAG was immunoprecipitated, indicating that the proteins can exist in an identical complex in mammalian cells. In the converse experiment, immunoprecipitation of HA-G␤ 1 ␥ 2 also specifically co-purified HDAC5-FLAG, further supporting the ability of HDAC5 and G␤␥ to form a complex in mammalian cells.
Direct G␤/HDAC5 Interaction-The interaction between HDAC5 and G␤␥ has thus been shown in both yeast and mammalian systems, suggesting that the interaction is likely direct. However, to further demonstrate that the interaction is not mediated through other molecules, purified proteins were assayed for binding. The C-terminal domain of HDAC5, HDAC5ct, was expressed and purified as a GST fusion protein from a baculovirus/Sf9 system. This protein was assayed for interaction with G␤ 1 ␥ 1 highly purified from bovine rod outer segments. Equimolar amounts (50 nM) of GST or GST-HDAC5ct were incubated with increasing concentrations of G␤ 1 ␥ 1 followed by purification with glutathione-Sepharose. As displayed in Fig. 3, G␤ 1 specifically co-purified FIGURE 1. G␤␥ exists in a stable complex harboring HDAC activity in vivo. Rat heart lysates were subjected to immunoprecipitation with a polyclonal pan-G␤ antibody or a negative control antibody (rabbit anti-His probe). A, immunoprecipitation of G␤. One percent of the lysates or post-immunoprecipitation flow through (FT) or 10% of the protein G resin was analyzed by immunoblotting with an anti-G␤ antibody. The result is typical of three independent immunoprecipitations. B, G␤ complexes exhibit HDAC activity. The immunoprecipitates (IP) were assayed for in vitro deacetylase activity. Results are displayed as the average from three immunoprecipitations (IP) (* , p Ͻ 0.02 versus control; **, p Ͻ 0.01; two-tailed Student's t test).

JOURNAL OF BIOLOGICAL CHEMISTRY 41771
with GST-HDAC5ct, confirming that the proteins interact directly. In addition, to estimate the HDAC5ct/G␤␥ affinity, an apparent EC 50 was calculated; at 50 nM GST-HDAC5ct, 50% binding is achieved at 330 Ϯ 120 nM G␤␥. Thus, the apparent affinity is similar to that observed for the interaction of G␤␥ with RACK1 (5).
The G␤␥/HDAC5 Interaction Is Controlled by Heterotrimeric G Protein Activation-In regard to the regulation of effector molecules, heterotrimeric G proteins containing G␣-GDP are thought to be inactive, with effector regulation occurring upon the receptor-catalyzed exchange of GTP for GDP on G␣ and resultant dissociation of the heterotrimer. This mechanism allows activation of receptors to be involved in the temporal regulation of downstream pathways. Accordingly, we sought to determine whether coexpression of G␣ could block the ability of overexpressed G␤␥ to form a complex with HDAC5. HEK293A cells transfected with HDAC5-FLAG, HA-G␤ 1 ␥ 2 , EE-tagged G␣ oA , and the G i/o -coupled receptor ␣ 2A -adrenergic receptor were serum-starved overnight. In subsequent anti-FLAG immunoprecipitation assays we observed that complex formation was greatly reduced by G␣ o expression (Fig. 4), and anti-EE reactivity was never observed in the immunoprecipitates (data not shown). Together, these data suggest that HDAC5 binds only to free G␤␥, allowing the possibility that the interaction is controlled by activated receptors.
To test the hypothesis that G␤␥ interacts with HDAC5 upon G protein activation, the ability of stimulation of a G i/o -coupled receptor to relieve inhibition of complex formation by G␣ oA was tested. Following serum starvation as described above, the cells were incubated with a ␤-adrenergic receptor antagonist, 10 M propanolol, for 10 min at 37°C. The cells were then exposed to 10 M (Ϫ)epinephrine to stimulate the ␣ 2A -adrenergic receptor for various times, whereupon complex formation was assayed by anti-FLAG immunoprecipitation. As shown in Fig. 4, receptor stimulation relieved the G␣ o -mediated inhibition of G␤␥/HDAC5 complex formation, suggesting that the interaction is signal-responsive in vivo.
The C Terminus of HDAC5 Mediates G␤␥ Binding-The yeast twohybrid assay identified the C terminus of HDAC5 as sufficient for binding to G␤␥. To demonstrate that this domain is the major binding site for G␤␥, we determined the ability of the C-terminal domain to block the interaction between G␤␥ and full-length HDAC5. A construct encompassing the C-terminal 172 amino acids of HDAC5 (residues 951-1122 of human HDAC5) was created by PCR amplification. HEK293A cells were transfected with plasmids encoding HDAC5-FLAG and HA-G␤ 1 ␥ 2 along with increasing amounts of DNA directing the expression of the C-terminal fragment with an N-terminal His 6 tag. As shown in Fig. 5A, subsequent co-immunoprecipitation assays indicated that the C-terminal fragment was capable of inhibiting the interaction between HDAC5 and G␤␥.
Identification of the C terminus of HDAC5 as a major G␤␥ binding site suggested that it might comprise a novel G␤␥ interaction domain. Scanning protein sequence data bases with this minimal domain did not reveal conservation among multiple protein families, but it did identify the C-terminal 164 amino acids of HDAC4 as containing significant similarity (ϳ68% identity and 80% similarity). The C termini of both HDAC4 and HDAC5 were subcloned into a mammalian expression vector directing the expression of N-terminally GST-tagged proteins. Precipitations with glutathione-Sepharose from lysates of HEK293A cells expressing the GST fusions and HA-tagged G␤ 1 ␥ 2 demonstrated that G␤␥ interacts with both proteins (Fig. 5B).  In addition to the C-terminal domain, full-length HDAC4 was tested for G␤␥ binding by immunoprecipitating C-terminally FLAG-tagged full-length HDAC4. As shown in Fig. 4C, full-length HDAC4 and HDAC5 formed complexes with HA-G␤ 1 ␥ 2 , demonstrating that HDAC4 is another novel G␤␥-binding protein. Furthermore, HDAC6, which has two catalytic domains that are homologous to those of HDAC4 and HDAC5 but lacks the corresponding C-terminal domain, was not observed to form a complex with G␤␥ in a parallel experiment (Fig. 5C), indicating specificity of the interaction of G␤␥ with HDACs 4 and 5.
The G␤␥ Binding Motif Comprises Multiple Non-colinear Sequences-To further distinguish the G␤␥-binding element in the C terminus of HDAC5, the domain was dissected via PCR into several fragments (Fig. 6A). These fragments were expressed with HA-G␤ 1 ␥ 2 in HEK293A cells and were tested for G␤␥ binding by glutathione-Sepharose-mediated precipitation from cell lysates. As displayed in Fig. 6B, fragments consisting of the first 58 (fragment 4) or 88 (fragment 5) amino acids bound G␤␥, as did a fragment consisting of the last 113 amino acids (fragment 6). Because fragments 4 and 6 do not overlap, these results suggest that the domain contains multiple sequences required for full G␤␥ binding. Of additional significance, fragments consisting of the extreme C-terminal 55 (fragment 2) or 28 (fragment 3) amino acids did not pull down a significant amount of G␤␥ under these conditions, suggesting that G␤␥ does not bind to the signalresponsive nuclear export signal located within amino acids 1081 to 1122 (19).
Signaling through G␤␥ Influences Transcription through an HDAC5repressible Factor-One well studied role of HDAC5 and other class II HDACs is co-repression of the transcriptional activity of the muscle differentiation factor MEF2C. When HDAC5 is associated with MEF2C, the deacetylase creates a repressive environment, inhibiting the ability of the transcription factor to activate gene transcription. The repressive effect of HDAC5 is controlled in several ways, for instance by prevention of the interaction with MEF2C and by sequestration from the nucleus, which is enhanced by phosphorylation and by binding to 14-3-3 proteins (19,(23)(24)(25).
To test the hypothesis that G␤␥ impacts HDAC5 function, we employed a quantitative assay of MEF2C transcriptional activity in cultured cells using a procedure originally described by Zhang et al. (22). In this system, ectopic expression of MEF2C activates transcription of a reporter gene (luciferase) that is downstream of multiple MEF2C binding sites. In HEK293A cells, transcriptional activation is repressed in a dose-dependent manner by co-expression of HDACs 4 and 5, as expected (data not shown). In addition, in cells grown in normal serum the activation is potentiated by the cell-permeable HDAC inhibitor TSA (EC 50 ϭ 396 nM) (Fig. 7A), suggesting that the activity of MEF2C is limited by endogenous HDACs in resting cells.
The interaction between G␤␥ and HDAC5 is enhanced through the activation of a G i -coupled receptor (Fig. 4). To determine whether interaction with G␤␥ correlates with a modulation of HDAC5 activity, we assayed MEF2C activity upon activation of the ␣ 2A -adrenergic receptor.

FIGURE 5. The G␤␥/HDAC5 interaction is mediated through the C-terminal domain.
A, expression of the C terminus inhibits the interaction between HDAC5 and G␤␥. Lysates from HEK293A cells expressing empty vector or HDAC5-FLAG, HA-G␤ 1 , and G␥ 2 and increasing amounts of His 6 -tagged C-terminal domain were subjected to anti-FLAG immunoprecipitation (IPs). The amount of HA-G␤ 1 co-precipitated decreased with increasing expression of the C-terminal domain (middle section). IB, immunoblot. B, the C terminus of HDAC4 also interacts with G␤␥. Lysates from HEK293A cells expressing empty vector (pcDNA), GST, or GST-tagged HDAC4 or HDAC5 C termini (HD4ct and HD5ct) and HA-G␤ 1 ␥ 2 were subjected to precipitation with glutathione-Sepharose. Both C termini precipitated HA-G␤ 1 with similar potency (middle section). IB, immunoblot. C, G␤␥ binding is specific for HDACs 4 and 5. HEK293A cells co-transfected with HDAC 4, 5, or 6 or empty vector (CTRL) and HA-G␤ 1 ␥ 2 were lysed and subjected to anti-FLAG immunoprecipitation (IPs). Both HDAC4 and HDAC5, which have similar C termini, co-immunoprecipitate HA-G␤ 1 , whereas no binding to HDAC6 was detected. Approximately 2.5% of lysates and 50% of immunoprecipitates were analyzed by Western blot. IB, immunoblot. FIGURE 6. Multiple regions of HDAC5ct are required for full G␤␥ binding. A, the C-terminal domain was dissected into several fragments via PCR-based cloning. The portion of HDAC5 corresponding to the yeast two-hybrid hit is labeled Y2H hit. Numbers on the right denote the amino acids of full-length human HDAC5 corresponding to the start and end of the respective fragment. Shading of the bars indicates fragments that pulled down (gray) or did not pull down (black) HA-G␤ 1 ␥ 2 . B, GST pull-downs from intact cells indicate that both halves of the C terminus are required for full binding. Lysates from appropriately transfected cells were subjected to precipitation with glutathione-Sepharose. Half of the pellets and 2.5% of lysates were analyzed by Western blot with the noted antibodies. Mobility of standards is noted on the left (kDa). Similar amounts of fusion proteins were purified (top section), whereas G␤␥ co-purified with fragments 1, 4, 5, and 6 (middle section). G␤␥ was expressed at similar levels in all lanes (bottom section). IB, immunoblot.
Overnight serum starvation of transfected HEK293A cells resulted in an almost complete inhibition of the MEF2C activity (data not shown). However, stimulation of the receptor with 10 M (Ϫ)epinephrine caused a time-dependent activation of MEF2C-mediated transcription (Fig. 7B). Note that because this assay relies on transcription and translation of the reporter gene, the observed time dependence is significantly different from the rate of formation of the protein-protein interaction upon receptor stimulation (Fig. 4). To demonstrate the specificity of this response, the cells were treated overnight with pertussis toxin. In these cells, (Ϫ)epinephrine stimulation failed to induce MEF2C activation (Fig. 7B), consistent with a role of G i/o heterotrimers in the response.
To determine whether modulation of MEF2C output is dependent on G␤␥ or is downstream of G␣ i/o , we utilized a G␤␥ scavenger, the C-terminal 132 amino acids of GPCR kinase 2 (GRK2ct) (26,27). Importantly, we found that expression of GRK2ct inhibits complex formation between G␤␥ and HDAC5. Cells were transfected with HDAC5-FLAG, HA-G␤ 1 ␥ 2 , and His 6 -GRK2ct or empty vector. Subsequent immunoprecipitation with an anti-FLAG antibody showed that less G␤ 1 was co-purified with HDAC5 when GRK2ct was expressed (Fig. 8A).
In addition to disrupting complex formation, we found that expression of GRK2ct inhibited MEF2C-mediated gene transcription (Fig. 8B) in cells grown in normal serum, suggesting that endogenous G␤␥ influences MEF2C activity. To show that the inhibition was due to modulation of HDAC activity, we determined the sensitivity of the inhibition to treatment with a cell-permeable HDAC inhibitor. A reduction of the ability of GRK2ct to inhibit MEF2C activity by an HDAC inhibitor would suggest that the effect is mediated through an alteration of HDAC activity (28). At 30 nM TSA, neither MEF2C-independent nor MEF2C-dependent transcription was altered significantly (Fig. 8B). At this concentration, however, the GRK2ct-mediated inhibition was abated (Fig. 8B), suggesting that scavenging G␤␥ causes an increase in endogenous HDAC activity.

DISCUSSION
G␤␥-mediated signaling occurs through interactions with a variety of molecules, including enzymes, ion channels, and small G proteins (2). One major signaling role for G␤␥ is the management of MAPK cascades, which allows communication between G␤␥ and the nucleus (for review, see Ref. 29). Here, we have shown that G␤␥ interacts with another nuclear protein, HDAC5, and that this interaction correlates with inhibition of its nuclear function. Our results therefore suggest that inhibition of the activity of transcriptional co-repressors is another mechanism by which G␤␥ influences gene expression.
Adding to the List of G␤␥ Effectors-The abundance of pathways impacted by G␤␥ is evident by the number of biological roles ascribed to G␤␥ signaling, such as mediation of chemokine-stimulated chemotaxis (30) and alteration of neurotransmitter release from presynaptic vesicles (31). The finding that G␤␥ directly interacts with and affects the function of enzymes involved in protein deacetylation expands the signaling capacity and physiological responsibilities of G proteins.
A New Level of Control of Transcriptional Co-repressors-In addition to expanding the cellular role of G␤␥, the current report also adds to the understanding of HDAC regulation. The regulation of HDAC5 and other class II HDACs is of clear interest. For example, there exists a FIGURE 7. Liberation of G␤␥ corresponds with enhanced MEF2C transcriptional activity. A, MEF2C activity is limited by endogenous HDAC activity. Cells were transfected with MEF2C, a reporter construct with luciferase expression under control of MEF2C, and a CMV-LacZ construct. One day after transfection, increasing doses of TSA or vehicle (0.1% EtOH) were added, and the cells were incubated for ϳ18 h. Upon harvesting, lysates were assayed for luciferase activity, and the results were normalized to ␤-galactosidase activity. Data are presented as the ratio of the luciferase and ␤-galactosidase activities normalized to the ratio from cells without ectopic MEF2C (basal) and without TSA treatment. The data are normalized to the basal/vehicle value. RLU, relative luciferase units. B, activation of a G i -coupled receptor enhances MEF2C-mediated transcription. Cells were transfected with MEF2C and reporter plasmids as well as a plasmid encoding ␣ 2A -adrenergic receptor. After overnight serum starvation in the absence or presence of 0.1 g/ml pertussis toxin (PTx), cells were preincubated for 10 min with 10 M propanolol and then stimulated for various times with 10 M (Ϫ)epinephrine (epi). Reporter gene transcription was measured as described above. Luciferase/␤-galactosidase activity ratios are normalized to the values at t ϭ 0.

FIGURE 8. G␤␥ influences MEF2C activity in an HDAC-dependent manner. A, GRK2ct
inhibits the G␤␥/HDAC5 interaction. Co-immunoprecipitation assays were performed as described previously, except that various amounts of a plasmid encoding rat GRK2ct were transfected. B, GRK2ct inhibits MEF2C activity in an HDAC-dependent manner. Cells were transfected with reporter constructs, without (basal) or with (no inhibitor) a MEF2C plasmid or with MEF2C and His 6 -GRK2ct (␤ARKct). These cells were treated with vehicle or 30 nM TSA for ϳ18 h, after which luciferase and ␤-galactosidase activities were assayed. N.S., values with vehicle and 30 nM TSA are not significantly different (n ϭ 3); *, value with vehicle is significantly different from that with 30 nM TSA (p Ͻ 0.01; two-tailed Student's t test, n ϭ 3). RLU, relative luciferase units. strong link between these enzymes and the physiological differentiation of muscle cells and the pathological development of cardiac hypertrophy. For example, Zhang et al. (32) demonstrated that ectopic expression of a hyperactive HDAC5 mutant inhibited the stimulated expression of genes marking pathologic hypertrophy. In addition, mice lacking a related HDAC, HDAC9, displayed increased susceptibility to hypertrophic stimuli (32). Moreover, histone deacetylase inhibitors have been demonstrated to be effective at limiting pathology in a model of hypertrophy (33).
Accordingly, the regulation of class II HDACs has been well studied (34). Phosphorylation by kinases, including calmodulin-dependent protein kinase I, leads to the dissociation of HDAC5 from MEF2C, relieving its inhibitory effect (35). Phosphorylation induces export of HDAC5 from the nucleus into the cytosol, where it is sequestered by members of the 14-3-3 protein family (19,(23)(24)(25). Direct interaction with G␤␥ provides an additional level through which signaling from cell surface receptors regulates HDAC activity.
An effect of G␤␥ on transcription is not without precedent. For instance, G␤␥ was found to play a role in the induction of intercellular adhesion molecule 1 (ICAM-1) transcription by thrombin stimulation of protease-activated receptor 1 in endothelial cells (36). In addition, Reusch et al. (37) showed that serum stimulation of a smooth musclespecific promoter in vascular smooth muscle cells was up-regulated by G␤␥ activation and inhibited by expression of a G␤␥ scavenger. Accordingly, G␤␥ is thought to play an important role in cell proliferation (38 -40), and a G␤␥ scavenger has been shown recently to inhibit growth in a cancer cell model (41). Several mechanisms are likely responsible for the effect of G␤␥ on gene transcription and cell proliferation, including activation of phosphoinositide 3-kinase (36) and the MAPK cascades (42). We propose that modulation of HDAC activity is yet another mechanism influencing nuclear effects of G protein signaling.
G␤␥ signaling may be responsible for fine tuning the activity of HDAC5 and MEF2C. Whereas activation of G i/o in serum starved-cells resulted in a maximal induction of a MEF2C responsive reporter gene of ϳ2-fold (Fig. 6B), and a G␤␥ scavenger had a significant but modest effect on transcription in cells in normal serum (Fig. 7B), treatment with a small molecule inhibitor of HDAC catalytic activity induced an increase of ϳ10-fold (Fig. 6A). This disparity suggests that G␤␥ influences a relatively small portion of cellular HDAC. Such a modulation may be critical under physiological conditions, however.
Mechanism of HDAC5 Inhibition by G␤␥-We are currently studying the mechanism by which interaction with G␤␥ inhibits HDAC5 cellular activity. G␤␥ might collaborate with other proteins in altering subcellular distribution of HDAC5. Alternatively, interaction with G␤␥ may redeploy a fraction of the cellular HDAC pool to an alternate job.
The impact of the G␤␥/HDAC5 interaction on signaling through MEF2C may be secondary to a yet to be discovered role for the interaction near the plasma membrane. It is abundantly clear that histones are not the only proteins impacted by the activity of HDACs (11). Numerous transcription factors, including STAT3 (signal transducers and activators of transcription 3) (43) and p53 (44), are acetylated; autoacetylation of HATs can impact catalysis and protein-protein interactions (11,45), and the roles of acetylation of proteins such as importin-␣ have yet to be determined (46). Association with HATs and HDACs has been reported to affect the biological roles of extranuclear proteins such as endothelin receptor A (20) and cytosolic phospholipase A 2 (47), but the acetylated substrates mediating these effects have not been identified. In mediating the interaction of HDAC5 with other factors, G␤␥ thus may be relieving MEF2C of a small portion of its complement of class II HDACs, leading to our observations.
On the other hand, it is interesting to note that the G␤␥ dimer may play a direct role in transcriptional regulation. Park et al. (48) showed that G␤␥ directly binds another transcriptional repressor, the adipocyte enhancer-binding protein 1 (AEBP1). Levels of G␥ 5 are regulated during adipogenesis and correlate inversely with AEBP1 transcriptional repression, thus potentially implicating G␤␥ in a direct regulation of transcription (48). It is possible, therefore, that G␤␥ activates MEF2C-dependent transcriptional activity by shunting class II HDACs to other nuclear complexes.
Additional recent data have identified another direct nuclear function for G␤␥. Kino et al. (40) discovered through a yeast two-hybrid screen that G␤␥ is a binding partner of the glucocorticoid receptor. This association correlates with nuclear import of G␤␥ upon glucocorticoid stimulation and plasma membrane association of the glucocorticoid receptor upon activation of the somatostatin GPCR (4). In contrast to our findings, however, where G␤␥ signaling enhanced the activity of a transcription factor, G␤␥ was found by Kino et al. (4) to suppress glucocorticoid-dependent transcriptional activity. The effect of G␤␥ could thus depend on the identity of the transcription factor or the contents of the transcriptionally active complex. On the other hand, it is possible that the impact of G␤␥ on MEF2C activity in vivo is secondary to recruitment of HDACs from MEF2C to nuclear hormone receptors. Interestingly, Kino et al. (4) found no effect of GAL4-fused G␤ expression on basal transcription from a GAL4 promoter, indicating that G␤ was not recruiting co-activators or co-repressors to DNA. The effects of HDACs on transcription are not restricted to modulating the acetylation state of histones, however. The G␤␥-mediated recruitment of class II HDACs could modulate the acetylation state of the transcription factor itself or some other protein; this effect would be predicted to be apparent only in the context of an appropriate transcription factor.
Conclusion-The discovery of a direct link between G proteins and histone deacetylases was a novel and unexpected finding. HDACs are emerging as a valuable drug target in the treatment of cancer and other disorders. Therefore, the newly described association may facilitate drug discovery, because signaling pathways involving G proteins have long been known as a rich source of validated drug targets.