Minireview: The Versatile Roles of Lysine Deacetylases in Steroid Receptor Signaling

Abstract

Lysine deacetylases have been known to regulate nuclear receptor function for many years. In the unliganded state, nuclear receptors that form heterodimers with retinoid X receptors, such as the retinoic acid and thyroid hormone receptors, associate with deacetylases to repress target genes. In the case of steroid receptors, binding of an antagonist ligand was initially reported to induce association of deacetylases to prevent activation of target genes. Since then, deacetylases have been shown to have diverse functions in steroid receptor signaling, from regulating interactions with molecular chaperones to facilitating their ability to activate transcription. The purpose of this review is to summarize recent studies on the role of deacetylases in steroid receptor signaling, which show deacetylases to be highly versatile regulators of steroid receptor function.

The first histone acetyltransferases (HATs) and histone deacetylases (HDACs) were identified more than 15 years ago (1–5). Since then, there has been an intense research effort directed at understanding their biology and identifying their substrates. Because histones had been known to be acetylated since the early 1960s (6), initial focus was on the function of these enzymes in transcription through their regulation of histone acetylation. HATs were found to be recruited to genes through interaction with transcription factors that activated transcription. In contrast, HDACs were found to associate with complexes known to have repressive functions in regulating transcription. Coupled with evidence showing that transcriptionally active regions of the genome were enriched with acetylated histones and transcriptionally inert regions were depleted of acetylated histones, the first concrete models of the functions of HATs and HDACs in regulating transcription were formulated (7, 8). These models posited that HATs recruited to genes acetylated promoter histones, contributing to the relaxing of chromatin structure, which facilitated the association of the machinery necessary for transcription. On the other hand, HDACs were recruited to genes being inactivated where they deacetylated histones to facilitate chromatin folding repressive to transcription. In addition, HDACs were thought to actively maintain the deacetylated state of histones in transcriptionally inactive regions of the genome. This general model that casts HATs as transcriptional coactivators and HDACs as corepressors has guided many in their understanding of the role of acetylation in transcription.

Mammalian HDACs comprise a large family of proteins that is divided into 4 classes (reviewed in Reference 9). The class I HDACs, HDACs 1–3, and 8, are nuclear localized. Class II HDACs are divided into 2 subgroups. The IIa HDACs, HDACs 4, 5, 7, and 9, have very low levels of deacetylase activity in vitro due to an amino acid substitution in their catalytic domains and have never been demonstrated to be deacetylases in vivo. HDACs 6 and 10 are the class IIb deacetylases. HDAC6 is cytoplasmic and is known to deacetylate α-tubulin and heat shock protein (Hsp)90. Little is known about the function of HDAC10 and HDAC11, the only class IV deacetylase. Class III is made up of the nicotinamide adenine dinucleotide (NAD+)-dependent sirtuins. The identification and development of small molecule inhibitors of class I and IIb HDACs as anticancer therapeutics established these enzymes as clinically relevant drug targets and expanded knowledge of their function in cell biology, development, and disease.

Several findings have directly challenged the original idea that HDACs function exclusively in transcriptional repression. First, the general model described above predicts that HDAC inhibition would activate gene expression due to increased histone acetylation. However, expression profiling studies as well as study of individual genes have shown that HDAC inhibitors (HDACi) can rapidly repress gene expression through transcriptional mechanisms (reviewed in Reference 10). In addition, depletion of HDAC expression has been shown to inhibit gene activation in response to a biological stimulus, eg, the negative impact of HDAC depletion or HDACi treatment on signaling through Stat transcription factors (11–14). Second, the above model predicts that HDACs would be associated primarily with silenced genes. However, several genome-wide studies have shown that HDACs are enriched in the regions of active or potentially active genes rather than silent genes (15–17). In fact, they colocalize with HATs and acetylated histones at the promoters of transcribed genes.

Finally, the long-standing model for the role of acetylation in transcription is histone centric, describing the functions of HDACs and HATS only as they relate to histones. Since the discovery that p53 is regulated by acetylation (18), many other transcription factors have been shown to be acetylated (19). In addition, recent proteomic studies identified thousands of acetylated proteins including many transcriptional coregulators, such as components of ATP-dependent chromatin remodeling complexes and various histone-modifying complexes (20, 21). The impact of acetylation on the function of most of these proteins is largely unknown. However, acetylation of some transcription factors, such as nuclear factor-κB (NF-κB) (22–25), has been studied in greater detail. It is now generally understood that, like phosphorylation, acetylation can stimulate or inhibit protein function. The original model of HAT and HDAC function in transcription does not include the association of HDACs with active gene regions or the acetylation of nonhistone transcriptional regulatory proteins and thus, cannot account for rapid and potent transcriptional repression of some genes by HDACi (26–30), which strongly suggests that HDACs can facilitate gene expression, probably through deacetylation of nonhistone transcriptional regulatory proteins.

The rapidly growing number of acetylated nonhistone proteins has prompted some to use alternative nomenclature for these enzymes, ie, lysine acetyltransferases (KATs) and lysine deacetylases (KDACs), because these terms more accurately reflect the activity of these enzymes in vivo. This is particularly relevant to the steroid receptor field because there is evidence that steroid receptors and many of the coregulators with which they interact are acetylated (discussed in more detail below). Therefore we will adopt these terms for the duration of this review, referring to deacetylases and acetyltransferases in general as KDACs and KATs, respectively. For individual KDACs, we will also use alternative terminology, eg, KDAC1 instead of HDAC1, KDAC6 instead of HDAC6, etc.

Current Model for the Roles of KATs and KDACs in Steroid Receptor Action

Steroid receptors have been established to use KATs and KDACs to directly regulate transcription. Upon ligand binding, steroid receptors interact with various KATs, such as CREB-binding protein/p300 and p300/CBP-associated factor (PCAF), and recruit them to target genes. Unlike other nuclear receptors such as thyroid hormone receptor, steroid receptors do not interact with complexes containing KDACs to repress transcription at target genes prior to ligand binding. However, steroid receptors bound to antagonists have been shown to interact with KDACs (31–33). These findings led to models of steroid receptor action that show agonist- and receptor-dependent recruitment of KATs to target genes which then acetylate histones to facilitate transcription. Conversely these models portray antagonist-dependent recruitment of KDAC-containing complexes to target genes that maintain nucleosomal histones in a hypoacetylated state to prevent gene activation (34–36). Thus, in accordance with general models of the role of KATs and KDACs in transcription, KATs are generally considered to facilitate steroid receptor-activated gene expression whereas KDACs repress or prevent it. However, very early studies of the HDACi, sodium butyrate, showed that it impaired steroid-mediated activation of several genes (37–40), first raising the possibility that KDACs play a more complicated and versatile role in steroid receptor signaling. More recent studies have supported these original findings and provided some insight into the mechanism by which this may occur. The purpose of this review is to summarize the current state of knowledge on the contribution of KDACs to gene expression regulated through steroid receptor signaling.

Hsp90 Acetylation and Steroid Receptor Processing

The interaction of steroid receptors with the molecular chaperone, heat shock protein 90 (Hsp90), is critical for their function. Inhibitors of Hsp90, such as geldanamycin, efficiently impair steroid receptor signaling (41–43). In addition, acetylation of Hsp90 inhibits its interaction with steroid receptors, resulting in misfolding, impaired ligand binding, and receptor degradation (44, 45). Hsp90 is deacetylated in large part by the class IIb KDAC6 (46). Accordingly, KDAC6-knockout mice have increased levels of acetylated Hsp90 and decreased function of glucocorticoid receptor (GR) (47). KDAC6 depletion in cell-based systems has also been shown to inhibit androgen receptor (AR) function through increased Hsp90 acetylation (48, 49). In addition, pan-HDACi such as trichostatin A (TSA) and vorinostat, which inhibit both class I and IIb HDACs, cause increased acetylation of Hsp90 and inhibition of estrogen receptor (ER) activity (50). Thus, KDAC6 facilitates steroid receptor signaling through deacetylation of Hsp90.

Whereas acetylation of Hsp90 is detrimental to steroid receptor signaling, treatment with pan-HDACi does not always result in inhibition of steroid receptor action. TSA treatment did not inhibit transcription of the mouse mammary tumor virus (MMTV) promoter induced by the progesterone receptor (PR) in a T47D-derived cell line (51). Also, TSA treatment was shown to stabilize levels of ERα in breast cancer-derived cell lines (52). In both of these studies the time of exposure to TSA was relatively short (<3 hours). Inhibition of Hsp90 by acetylation is dependent not only on KDAC6 inhibition but also the activity of KATs that catalyze the acetylation and their proximity to Hsp90. It is possible that Hsp90 acetylation accumulates at different rates in different cell types. In fact, we observe differential effects on Hsp90-GR interactions in 2 cell lines (53). In one cell line, TSA treatment disrupted this interaction within 1 hour of treatment whereas in another there was no effect on GR-Hsp90 interaction for up to 5 hours, even though histone acetylation increased dramatically over this time period. Thus, the time of exposure to pan-HDACi may be critical to general impairment of steroid receptor activity through Hsp90 acetylation. Short exposures are clinically relevant because the Food and Drug Administration-approved HDACi vorinistat and romidepsin have average serum half-lives in the range of 1.5–8 hours in humans (54–56).

Another important factor to consider in interpretation of experiments involving pan-HDACi and their effects on steroid receptor chaperoning is the concentration used. Two studies clearly document concentration-dependent effects of HDACi. At low concentrations TSA stimulates GR-dependent activation of the MMTV promoter (57) and AR-induced expression of the prostate-specific antigen (PSA) and FKBP5 genes (58) but represses them at higher concentrations in the same cell lines. This might be due to off-target effects of TSA. HDACi are structurally diverse, belonging to several chemical classes. The issue of off-target effects can be addressed by comparing chemically distinct HDACi. In the case of the PSA gene, the same concentration-dependent effects were observed with HDACi belonging to distinct chemical classes (TSA, butyrate, and Romidepsin) (59), strongly indicating that off-target effects were not involved. Selective inhibition of KDAC complexes may also explain concentration-dependent effects of HDACi. A recent chemoproteomic profiling study of HDACi clearly showed selective effects on KDAC complexes, each complex having different sensitivities to various HDACi (60). In fact, even among the pan-HDACi of the hydroxamate class, there are different potencies for inhibiting KDAC6; TSA and pabinostat were less potent KDAC6 inhibitors relative to the class I KDACs than vorinostat and Scriptaid. Thus, it cannot be assumed that inhibitory effects of pan-HDACi on steroid receptor signaling are mediated through inhibition of KDAC6 and resultant inactivation of Hsp90 through unopposed acetylation. In fact, deacetylation of Hsp90 is likely to be one of several ways in which KDACs facilitate steroid receptor signaling because class I-selective HDACi, which do not inhibit KDAC6, have also been shown to impair steroid-induced gene expression.

The Role of KDACs in GR-Regulated Transcription

Class I KDACs facilitate GR transactivation

The activation of several genes by glucocorticoids was shown in the early 1980s to be impaired by cotreatment with the class I-selective HDACi, sodium butyrate (37, 38). However, because KDACs and KATs had not yet been identified, the mechanism remained undefined. More recent studies on the GR-activated MMTV promoter and cellular GR target genes have built a strong case for gene-selective cooperation between GR and class I KDACs to facilitate glucocorticoid-activated transcription.

Agonist-bound GR initiates chromatin remodeling that derepresses the MMTV promoter and provides access to other transcription factors (61). HDACi such as sodium butyrate and TSA impair transactivation of MMTV by glucocorticoids (40, 57, 62) and rapidly repress basal transcription from the promoter (30). Short-term HDACi treatment does not inhibit GR-induced chromatin remodeling at the promoter, indicating that the defect in activation is downstream of GR association and chromatin remodeling (30). Accordingly, TSA-mediated repression of MMTV basal transcription was accompanied by rapid loss of hypophosphorylated RNA polymerase II (RNA pol II) from the promoter, consistent with its dependence on the MMTV core promoter (30, 63). In vitro transcription experiments showed that KDAC activity was dispensable for de novo transcriptional initiation but required for efficient reinitiation of transcription from the MMTV core promoter (63). Altogether these results suggest that, in the absence of KDAC activity, GR associates with the promoter and is able to induce chromatin remodeling and formation of the de novo initiation complex but is unable to maintain efficient reinitiation, thereby limiting its ability to fully activate transcription.

Until recently it was unclear whether the virally derived MMTV promoter was a special case among GR target genes. However, we carried out expression profiling to determine the impact of the class I-selective HDACi, valproic acid (VPA), on the GR transcriptome in a hepatoma-derived cell line (53). We found that transactivation of about 50% of cellular genes up-regulated by short-term dexamethasone (Dex) treatment was impaired by cotreatment with VPA. A structurally distinct, class I-selective HDACi, apicidin, had remarkably similar effects on GR transactivation, strongly indicating that the impaired transactivation was mediated through the KDAC-inhibiting activity of these drugs rather than off-target effects. Because the class I-selective HDACi do not inhibit KDAC6 and thus, do not cause disrupted GR-Hsp90 interaction, class I KDACs are likely to facilitate transactivation of these genes beyond the point of GR processing. In fact, deacetylase activity is present at GR binding regions in most genes tested as measured by a significant increase in histone acetylation upon exposure to VPA. This finding strongly indicates that KDACs are present and active in regions where they could come in contact with GR-assembled transcription complexes.

Several laboratories have independently shown that KDACs 1 and 2 are involved in GR transactivation. Class I KDACs 1–3 have been found to be associated with the MMTV promoter in both the presence and absence of glucocorticoids or progestins (51, 64). Depletion of KDACs 1 or 2 through small interfering RNA (siRNA) resulted in attenuated transactivation of the MMTV promoter by GR; this attenuation was more severe when both were depleted (64, 65). KDAC1 was found to associate with GR by coimmunoprecipitation assays and by fluorescence resonance energy transfer using fluorescent protein-tagged GR and KDAC1 localized to an MMTV array (64, 66). Qiu et al (66) discovered that KDAC1 is acetylated at several lysines in its C-terminal region, which results in inactivation of its deacetylase function. This group also provided evidence that, upon GR activation, KDAC1 at the MMTV promoter becomes deacetylated concomitant with an increase in its mobility at an MMTV array (64). The activation of KDAC1 at the MMTV promoter upon its transactivation by GR is consistent with a report showing that histones in the proximal MMTV promoter region are acetylated prior to GR activation and become deacetylated as transcription increases after Dex treatment (67). Whether acetylated KDAC1 is deacetylated at the promoter or is replaced by deacetylated KDAC1 from the nucleoplasm upon GR binding to the promoter has not been definitively addressed.

KDAC2 is also required for MMTV activation. Unlike KDAC1, KDAC2 is not acetylated but can be inactivated through heterodimerization with acetylated KDAC1 (65). KDACs 1 and 2 are known to form either homodimers or heterodimers in HDAC-containing complexes (68) but the identity of the complex associated with the MMTV promoter is unknown. Our study of the role of KDACs in GR transactivation of cellular genes confirms an important role for KDAC1 (53). Depletion of KDAC1 through siRNA was sufficient to fully mimic the effects of HDACi on most genes examined at which GR transactivation was significantly impaired in the presence of HDACi. Depletion of both KDAC1 and 2 resulted in full to partial impairment of GR transactivation at several additional genes. Interestingly KDAC2 depletion alone did not affect GR transactivation at any gene tested. It is possible that KDAC1 homodimers predominate in complexes because its expression may be significantly higher than that of KDAC2 in the cell line tested. Alternatively, KDACs 1 and 2 may exist in distinct complexes. There is growing evidence that the 2 KDACs are not functionally redundant (17, 60, 69, 70).

KDACs are required for GR-mediated transcriptional repression

Transcriptional repression by GR has been well studied due to its role in mediating the antiinflammatory effects of glucocorticoids. GR can repress gene expression either by binding DNA directly or by tethering to another transcription factor bound to DNA and inhibiting its ability to activate transcription (reviewed in Reference 71). Known examples of tethering transrepression include GR interactions with NF-κB, AP-1, and Nur77. Not all GR-mediated repression is sensitive to the presence of HDACi, indicating that KDACs are required in a gene-selective fashion (72, 73). In the cases in which HDACi treatment abrogates GR-mediated repression, association of KDACs with the target promoter has been observed. KDAC1, but not KDACs 2 or 3, was recruited to the Hes1 promoter in response to Dex treatment, accompanied by loss of histone acetylation and p300 (74). However, in most cases in which KDAC involvement was tested, KDAC2 was found to be associated with GR-mediated repression. At several genes repressed by GR in lipopolysaccharide-induced macrophages, KDACs 2 and 3 were recruited to enhancers in response to Dex treatment concomitant with a reduction in both histone H3 acetylation and expression of associated genes (75). In addition, KDAC2 was recruited along with GR and the corepressor silencing mediator of retinoid and thyroid hormone receptor (SMRT) to genes containing negative glucocorticoid response elements (GREs) that differ from classical GREs but can mediate glucocorticoid-induced transcriptional repression (76). The MUC5AC gene was found to be repressed by Dex through 2 classical GREs upstream from the transcription start site in lung epithelia-derived cells (77). Chromatin immunoprecipitation (ChIP) assays showed that GR and KDAC2 associated with these sequences within 1 hour of Dex treatment. In addition, depletion of KDAC2 expression with siRNA abrogated the Dex-induced repression of MUC5AC expression.

The proopiomelanocortin (POMC) gene is also down-regulated by glucocorticoids. Bilodeau et al (78) reported that one mechanism by which this occurs involves GR transrepression of the nuclear receptor Nur77, which binds the promoter and is necessary for activation of POMC transcription by CRH. Upon binding to agonist, the GR is recruited to the promoter and interacts with Nur77 through protein-protein contacts facilitated by the presence of Brg-1 (Figure 1A). The HDACi, VPA, TSA, or butyrate, prevented GR transrepression of POMC, and ChIP assays showed that KDAC2, but not KDACs 1 and 3, were recruited to the POMC promoter upon Dex treatment, independent of CRH exposure. The association of GR and KDAC2 at the POMC promoter was dependent on the presence of Brg-1 and correlated with a loss of histone acetylation and inhibition of RNA pol II function, including impaired recruitment and promoter clearance.

Figure 1.

KDAC2 facilitates GR transrepression at the POMC and CSF2 genes. A, GR tethers to the POMC promoter through interaction with Nur77. KDAC2 interacts with GR in a Brg1-dependent manner to disrupt the interaction of RNA Pol II with the POMC promoter as well as its ability to clear the promoter and initiate transcription. B, Deacetylated GR interacts with both KDAC2 and NF-κB and inhibits the ability of DNA-bound NF-κB to activate the CSF2 promoter.

GR is well established to transrepress some NF-κB-activated promoters, such as that of the CSF2 (GM-CSF) gene, through direct interaction with NF-κB. Agonist-bound GR impairs IL-1β-mediated CSF2 promoter activation by a TSA-sensitive mechanism. Ito et al (79, 80) determined that, in the presence of Dex, both GR and KDAC2 coimmunoprecipitate with the p65 subunit of NF-κB. However, under conditions in which KDAC2 expression is depleted, GR is lost from the NF-κB complex. Neither in vitro binding of GR to a GRE nor glucocorticoid-mediated activation of a transfected reporter or endogenous target gene was impaired by HDAC2 depletion; therefore, the authors concluded that KDAC2 is specifically required for transrepression (Figure 1B). Interestingly they found that KDAC2 levels are low in aveolar macrophages from patients with chronic obstructive pulmonary disease and, when treated with Dex, these cells show little transrepression of CSF2. However, if KDAC2, but not KDAC1, is overexpressed, transrepression is regained (80).

Altogether these studies provide strong evidence that KDAC2 plays an important role in repression of gene expression by GR that is often independent of KDAC1, thereby providing more evidence that the two KDACs are not functionally redundant in their impact on GR-regulated gene expression. Moreover, it is clear that GR cooperates with specific KDACs to activate or repress transcription in a gene-specific fashion, indicating that context may necessitate this cooperation and also determine its transcriptional effect.

Functional effect of GR acetylation

Steroid receptors have been found to be acetylated both in vitro and in cultured cells (81, 82). The most well-studied receptor acetylation is found in the hinge region, which has a concentration of basic amino acids that serve as a nuclear localization signal (NLS). Two groups examined the role of GR acetylation in regulating transcription. Ito et al (80) probed the role of GR acetylation in transrepression of the CSF2 promoter. They found that, overall, GR acetylation increased in the presence of Dex. However, GR that coimmunoprecipitates with p65 had a relatively low level of acetylation. Mutation of the last 2 lysines (K494 and K495 in the human GR) in a KTKK hinge region motif to either asparagine or alanine reduced overall GR acetylation, facilitated its interaction with p65, and caused efficient transrepression of the CSF2 promoter even in the presence of TSA. Thus, acetylation of GR may hinder its ability to interact with NF-κB and interfere with its action at the CSF2 promoter. Their findings imply that KDAC2 facilitates transrepression by maintaining GR in a deacetylated state (Figure 1B).

Nader et al (83) found that GR can be acetylated by Circadian Locomotor Output Cycles Kaput (CLOCK), a KAT integrally involved in maintaining circadian rhythms along with its heterodimerization partner Aryl hydrocarbon receptor nuclear translocator-like (ARNTL or BMAL) (Figure 2A). CLOCK/BMAL overexpression resulted in impaired activation of the MMTV, TSC22D3 (Gilz), and G6PC (glucose 6-phosphatase) promoters by GR, whereas depletion of CLOCK and BMAL enhanced activation upon Dex treatment. They found that CLOCK interacts with the ligand-binding domain of GR and that a CLOCK mutant without KAT activity failed to impair glucocorticoid-activated transcription of the MMTV promoter in transfection assays. CLOCK overexpression increased GR acetylation, and mutation of 4 lysines in the GR hinge region to alanine (K480/492/494/495A) abolished it. Accordingly, this GR mutant was not impaired in its ability to transactivate the MMTV promoter in the presence of CLOCK/BMAL. Because the target lysines are localized to the GR NLS, nuclear translocation was examined and found to be unaffected by CLOCK/BMAL overexpression. ChIP assays showed decreased association of GR with GREs in the endogenous TSC22D3 (Gilz) and G6PC (glucose 6-phosphatase) genes upon Dex treatment in the presence of overexpressed CLOCK/BMAL.

Figure 2.

Acetylation of steroid receptors impacts their ability to associate with target genes in a receptor-dependent fashion. A, Hinge region-acetylation of the GR by CLOCK/BMAL impairs its ability to associate with target genes in a gene-selective fashion. It should be noted that CLOCK may not be the only KAT that can acetylate the GR. The KDAC(s) involved in GR deacetylation have not been identified. B, Acetylation of ERα in the DBD region by p300 increases its ability to bind DNA in vitro and activate a transfected reporter containing the pS2 promoter. TSA-sensitive KDACs or the class III KDAC, Sirtuin 1 (Sirt1), are reported to deacetylate the ERα DBD (104). C, Hinge-region acetylation of MR impairs its ability to associate with the SGK1 and TSC22D3 (GILZ) genes. MR has been found to associate with nuclear co-repressor (NCoR) and KDAC3, and deacetylation of the MR has been shown to be dependent on KDAC3. The KAT that acetylates MR has not been identified.

In a follow-up study, this group examined the kinetic relationship between CLOCK expression and GR acetylation in peripheral blood mononuclear cells (PBMCs) from human subjects (84). This study provides the strongest evidence to date that GR is acetylated in vivo. They found that acetylation of endogenous GR in PBMCs mirrored endogenous CLOCK levels; both are high at 8:00 am and low at 8:00 pm, consistent with high and low levels of serum cortisol. When they examined a number of GR target genes, they found 2 patterns that were independent of whether the genes were repressed or activated by GR. For one group of genes, responsiveness to hydrocortisone in cultured PBMCs did not change over a 24-hour period despite changes in GR acetylation. However, over the same time period, another group of genes responded most robustly to hydrocortisone when the levels of GR acetylation were low, indicating sensitivity to CLOCK-mediated GR acetylation. Together these studies indicate that GR acetylation may negatively affect GR function in a gene-selective fashion. Thus, one possible function of KDACs is to deacetylate GR for efficient modulation of transcription in some gene contexts (Figure 2A).

The Role of KDACs in AR-Regulated Transcription

Class I KDACs facilitate AR transactivation

As in the case of GR, there is strong evidence that KDACs positively impact AR-regulated transcription in a gene-selective fashion. The HDACi, TSA, has been shown by several groups to repress dihydroxytestosterone (DHT)-activated gene expression when used at higher doses (>100 nM) (58, 59, 85). Welsbie et al (85) recently investigated the mechanism behind their observation that TSA repressed AR transactivation of a large number of target genes in LNCaP cells and selected AR target genes in a variety of prostate cancer-derived cell lines. The concentrations of TSA (>100 nM) or other pan-HDACi that were required for suppressed transactivation correlated with increased acetylation of histones and α-tubulin. In addition, they and others showed that class I-selective HDACi (butyrate and depsipeptide), also repress AR transactivation of selected genes (59, 85). The fact that class I-selective HDACi can cause this repression suggests that it is not simply due to inhibition of KDAC6 and consequent acetylation of Hsp90 and degradation of AR. Accordingly, although the AR-Hsp90 interaction was not directly measured, they did show that the rate of AR degradation was not changed in the presence of pan-HDACi.

Welsbie et al (85) went on to show that siRNA-mediated depletion of HDACs 1 or 3 had the greatest impact on AR transactivation in LNCaP cells, but the effect was gene selective rather than universal. Because transactivation of the PSA gene was suppressed by TSA, ChIP assays were performed to monitor complex assembly at the PSA enhancer region in response to androgen. TSA exposure did not affect the association of AR with the enhancer, but greatly diminished the recruitment of RNA pol II, steroid receptor coactivator 1 (SRC1), and p300. A kinetic analysis of AR and RNA pol II association with the PSA enhancer revealed that cycling of AR on and off the enhancer was altered, with the time between peaks of AR association being lengthened in the presence of TSA. In addition, corecruitment of RNA pol II was dramatically stunted during each cycle. These results indicate KDAC activity is dispensable for AR association with the PSA enhancer per se, but it is required for the ability of the AR to assemble dynamic transcription complexes that result in the efficient recruitment of RNA pol II. The authors propose that KDACs are needed to reset AR-responsive enhancers/promoters after each round of transcription to allow for the efficient recruitment of RNA pol II.

Class I KDACs repress AR transactivation

In addition to their repressive effects on AR action, HDACi have also been shown to augment AR transactivation of target genes, indicating a repressive transcriptional role of KDACs. TSA was originally shown to potentiate DHT activation of the MMTV promoter, a finding that was suggested to be due to a significant increase in AR levels (86). Recently, Chng and colleagues (58) documented a stimulatory effect of TSA on AR transactivation of the PSA and FKBP5 genes. At low doses (<100 nM) TSA significantly stimulated DHT-activated expression, consistent with the above-described study of the MMTV promoter, which used TSA at less than 50 nM (86). ChIP sequencing was used to identify AR and ERG binding sites in the prostate cancer-derived cell line, VCaP, which expresses high levels of ERG, a member of the ETS family of transcription factors that is often overexpressed in prostate cancer tumors (87). The results led them to classify AR-regulated genes into 2 groups, those with AR-binding sites only, and those with colocalized AR- and ERG-binding sites, which included the PSA and FKBP5 genes. Depletion of ERG expression resulted in enhanced AR transactivation at about a third of AR-activated target genes. Through ChIP sequencing analysis in the presence and absence of DHT, they found that KDACs 2 and 3 are present at many AR-binding sites. However, the presence of these KDACs was enhanced at the sites at which AR and ERG were colocalized. Coimmunoprecipitation experiments showed that ERG, AR, and KDACs 1 and 2 form a complex that also contains the lysine methyltransferase, EZH2. In LNCaP cells, which do not express ERG, the stimulatory effect of TSA at low concentration on the PSA and FKBP5 genes was not observed. The authors conclude that KDACs recruited to genes with colocalized AR- and ERG-binding sites attenuate AR transactivation in the presence of agonist. These findings indicate that the role of KDACs in AR-regulated transcription may be determined, in part, by the expression level and nearby binding of other transcription factors.

The functional effect of AR acetylation

Like GR, AR can be acetylated at lysine residues in its hinge region NLS in a ligand-dependent fashion in vitro and in cultured cells (88). The potential function of AR hinge region acetylation at a KLKK motif has been studied in some detail mostly with the use of AR mutants in which lysine residues are substituted with amino acids that mimic acetylated lysine (Q or T) or with amino acids that either mimic unacetylated lysine (R) or cannot be acetylated (A). In spite of a slight delay, AR acetylation-mimic mutants were able to achieve efficient ligand-dependent nuclear accumulation (89, 90), whereas AR mutants containing lysine-to-arginine substitutions were defective in nuclear translocation (91). Double (K632A/K633A) and triple (K630A/K632A/K633A) lysine-to-alanine AR mutants have significantly slower nuclear accumulation than wild-type (wt) AR upon ligand exposure (92, 93). Altogether these studies indicate that AR acetylation in the hinge region allows nuclear translocation and/or nuclear accumulation of AR (Figure 3). Accordingly, siRNA-mediated depletion of the KAT, Tip60, which can acetylate AR in its hinge region (94), impaired nuclear accumulation of the wt AR (91).

Figure 3.

KDACs differentially regulate nuclear accumulation of PR and AR through hinge region acetylation. Acetylation of PR in the hinge region inhibits nuclear translocation/accumulation whereas hinge-region acetylation of AR facilitates it. Thus, KDACs promote PR signaling and impair AR signaling through deacetylation of their hinge-region lysines.

In terms of AR transactivation, acetylation of the KLKK motif in the AR hinge region is thought to be stimulatory to AR function, although the underlying mechanisms are only vaguely understood. Acetylation-mimic mutants, such as K630Q and K630T, show increased binding to p300, decreased binding to corepressor proteins such as nuclear receptor corepressor, and increased transactivation of MMTV-, TAT- and PSA-reporter genes (95, 96). In contrast, lysine-to-alanine or arginine mutants are resistant to the stimulatory effects of p300 or p300/CBP-associated factor (PCAF) overexpression and show decreased association with coactivators and increased interaction with corepressors (88, 97, 98). Interestingly, a triple K-to-A mutant (K630/632/633A) has promoter-specific effects, stimulating transactivation at the MMTV promoter, inhibiting transactivation of the Pem promoter, and having no effect on the PSA promoter (93), indicating that the effect of AR acetylation may be promoter specific. One caveat to the above-mentioned studies is that all used transfected reporter genes in a few cell lines. Another factor complicating interpretation of experiments with lysine mutants is that AR is known to be methylated in this region. Two studies have provided evidence that hinge-region methylation at K630 or K632 by Set9 is stimulatory to AR function (97, 99). Further study of AR acetylation and methylation and its effects on endogenous target genes is warranted. In addition, acetylation of the AR has not yet been definitively demonstrated in vivo.

The Role of KDACs in ERα-Regulated Transcription

Studies of the role of KDACs in ERα-activated transcription are complicated by the effects of HDACi on ERα expression. These drugs have been reported to repress or increase expression of ERα, depending on the cell line. In ERα-negative breast cancer-derived cell lines, HDACi treatment can activate the ESR1 gene (100) whereas in ERα-positive cell lines the effects are varied, depending on the time of drug exposure. In MCF-7 cells both pan- and class I-selective HDACi caused a decrease in ESR1 transcription and an increase in ERα degradation with treatments of 6–48 hours (101–103). This may involve the action of KDACs 1, 2, and 6 based on siRNA-mediated depletions (103). In contrast, in T47D cells, short-term TSA treatment induces ERα acetylation and increases ERα expression through increased stability (52). Because p300 was shown to directly acetylate ERα, the acetylation and stability of p300 were investigated and both were found to be increased by TSA treatment in T47D cells. TSA enhanced the interaction of ERα with p300 and, in the presence of overexpressed p300, ERα acetylation was increased and its ubiquitination was decreased. Neither the sites of TSA-sensitive ERα acetylation nor the transcriptional activity of the TSA-stabilized ERα was assessed.

ERα is acetylated at multiple lysines, 2 in the DNA binding domain (DBD) and several in the hinge region. Mutation of the DBD lysines (K266, K268 in the human ERα) to glutamine increased the DNA binding activity of ERα in vitro as well as its transactivation function in reporter assays whereas mutation to arginine did not (104) (Figure 2B). In the case of the hinge region lysines (K299, K302, K303 in human ERα), the picture is more complicated. Mutation of these lysines to glutamine, arginine, or alanine increased sensitivity of ERα to estradiol and enhanced its activation of a target reporter in the presence of overexpressed p300 (105, 106). Because all these mutations had the same effect, it is not clear that acetylation, or lack thereof, is the cause of these changes in ERα function. ERα is also methylated and ubiquitinated at K302 (reviewed in Reference 107); therefore it is possible that the mutations prevented these modifications and contributed to the effects on ERα activity. The function of ERα hinge region acetylation in ERα function clearly requires additional study.

There is some evidence that acetylation of ER coactivators may impact ER function. As demonstrated in the study described above, acetylation of p300 may modulate ERα function through increased stability (52). In addition, Chen et al (108) discovered that acetylation of the ER coactivator steroid receptor coactivator 3 (SRC3, also known as ACTR) has a negative impact on ERα transactivation. They found that activation of the pS2 gene by ERα is transient. ChIP assays showed that whereas ERα continued to associate with the promoter, SRC3 and p300 are lost at later time points after estrogen treatment. In vitro experiments demonstrated that p300 acetylated SRC3 (as well as SRC2) at lysines just amino terminal to the first nuclear receptor box in its receptor interaction domain (RID). In support of these findings, a more recent proteomic study of acetylated proteins identified both SRC3 and SRC2 as acetylated (21). Further investigation revealed that acetylation in the RID region of SRC3 inhibited its ability to interact in vitro with agonist-bound ERα and that estrogen treatment induced the acetylation of endogenous SRC3 in MCF-7 cells. Expression of SRC3 mutants showed that mutation of 2 lysines in the RID to either glutamines or arginines caused sustained up-regulation of an ERα-activated reporter gene whereas expression of wt SRC3 caused only transient up-regulation. Based on crystallography studies, the lysines just upstream of the first nuclear receptor box of SRC3 are thought to have electrostatic interactions with acidic amino acids in the ERα ligand-binding domain. Because acetylation neutralizes the charge of the lysine side chains, the electrostatic interactions would likely be disrupted, reducing the affinity of the 2 proteins for each other. However, it is not clear why the glutamine substitutions would not have accomplished this, indicating that the ERα-SRC3 interaction might be more complicated than suggested. Further study of SRC acetylation is required to better understand how it impacts interaction with ERα as well as other steroid receptors.

The Role of KDACs in PR-Regulated Transcription

As has been reported for ERα, long-term exposure (≥24 hours) to pan-HDACi or class I-selective HDACi, results in significant down-regulation of both B and A isoforms of PR in breast cancer cell lines (103). This was attributed, in part, to an HDACi-induced decrease in PR mRNA. Depletion of KDAC2, but not KDACs 1 or 6, also caused down-regulation of PR protein expression independently of any effects on ERα expression. Thus, KDAC2 expression is important for maintenance of PR levels.

There are several reports of the effect of HDACi on the ability of PR to transactivate target genes. In T47D cells with integrated copies of the MMTV promoter, Archer and colleagues (51) reported that short-term (4–5 hours) TSA exposure at concentrations ≥ 100 nM did not impair, but slightly enhanced, PR transactivation of the MMTV promoter. They also showed that depletion of KDAC3 enhanced PR transactivation of the MMTV promoter as well as 3 cellular PR target genes. In the same cell line, longer exposure (≥ 24 hours) to the same concentrations of TSA caused a potent inhibition of PR transactivation of MMTV (109). This coincided with a decrease in PR levels, consistent with the study described above (103). In a different T47D-derived cell line with integrated copies of the MMTV promoter, Narayanan et al (110) reported that TSA treatment for 30 hours at the relatively high concentration of 1 μM, stimulated PR transactivation of the MMTV promoter in the G₁ and G₂/M phases of the cell cycle but had no effect when cells were in S phase. One major difference between the T47D-derived cell lines used by these 2 groups is that the first contains the full MMTV LTR with the mammary-specific enhancer, and the second contains a truncated LTR without the enhancer (111, 112). The additional regulatory elements may influence PR function under conditions of KDAC inhibition, consistent with the idea that promoter context can determine how KDACs contribute to steroid receptor action.

Effects of PR acetylation on function

The function of PR acetylation was examined by Daniel et al (113). In T47D cells, they showed that PR acetylation increases in response to short-term R5020 treatment in the presence of TSA, indicating that activation of PR by ligand induces an increased rate of acetylation, perhaps due to increased interaction between PR and KAT complexes. PR acetylation, as in other steroid receptors, occurs in the hinge region at a KKFK motif. The study showed that mutation of the hinge-region lysines to noncharged amino acids such as alanine, threonine, or glutamine significantly slowed ligand-induced nuclear accumulation and phosphorylation of PR. In contrast, mutation of these lysines to arginine resulted in a receptor that behaved like wt PR. These results suggest that PR acetylation impairs ligand-induced nuclear accumulation due to neutralization of positive charge in the NLS (Figure 3). With assays performed after PR mutants had fully accumulated in the nucleus, the study showed that PR mutants in which the lysines had been mutated to acetylation mimics, such as threonine and glutamine, impaired the ability of PR to activate the endogenous SGK1 gene but had no effect on transactivation of the F3 (tissue factor) gene. Altogether this study shows that KDACs may facilitate PR signaling through hinge-region deacetylation in 2 ways: first, by allowing efficient nuclear accumulation in the presence of ligand, and second, by facilitating PR-activated gene expression in a gene-selective fashion.

The Role of KDACs in MR-Regulated Transcription

HDACi have been reported to have antihypertensive effects (114). Lee et al (115) recently investigated the effect of HDACi on the mineralocorticoid receptor (MR), which is considered to play a role in the pathology of hypertension. They found that the class I-selective HDACi, VPA and MS-275, significantly attenuate the aldosterone (Aldo)-induced activation of 3 selected MR target genes in cultured HEK-293 cells. VPA pretreatment caused increased acetylation of MR only in the presence of Aldo. However, nuclear translocation of the MR under these conditions was unaffected, indicating that disrupted translocation did not cause impaired activation of MR target genes. At GRE sequences in the TSC22D3 (Gilz) and SGK1 genes, VPA treatment caused attenuated association of both wt MR and RNA pol II in the presence of Aldo. These findings were extended to the kidneys of rats in which hypertension had been induced by exposure to salts; both VPA-induced MR acetylation and impaired Aldo-dependent activation of the same target genes were associated with attenuated binding of MR in VPA-treated animals.

Lee et al went on to show that the acetylation of MR at hinge-region lysines was likely to cause the defects in MR transactivation. Acetylation of MR mutants with hinge-region lysines mutated to arginine or alanine was greatly reduced in the presence of VPA treatment. The ability of these mutants to activate an MR target gene reporter and associate with GRE regions in the TSC22D3 and SGK1 genes in the presence of Aldo was unimpaired by VPA cotreatment. MR mutants with the hinge-region lysines mutated to glutamine or threonine were not tested. MR was found to associate with KDAC3 and nuclear receptor corepressor but not the other class I HDACs. Depletion of KDAC3 resulted in increased acetylation of MR and impaired MR activation of the TSC22D3 and SGK1 genes, thereby mimicking the HDACi. Overall the study results show that KDAC3 may facilitate MR signaling by removing acetylation that impairs efficient association of MR with response elements in target genes (Figure 2C).

Summary

From the research described in this review, it is clear that the role of KDACs in steroid receptor signaling goes beyond their known association with antagonist-bound receptors. The accumulated evidence indicates that KDACs impact steroid receptor signaling at 4 points. First, KDACs are involved in regulation of genes encoding steroid receptors. Several studies show that HDACi or KDAC depletion result in changes in levels of mRNA encoding ERα (100, 101, 116), PR (103), and AR (85). In the case of ERα in breast cancer cells, HDACi cause increased expression in ERα-negative cell lines and decreased expression in ERα-positive cell lines. The mechanisms by which KDACs regulate expression of steroid receptors are only vaguely understood.

Second, KDACs facilitate steroid receptor signaling through deacetylation of the molecular chaperone, Hsp90. It is clear that acetylated Hsp90 does not interact efficiently with steroid receptors, thereby compromising their ability to achieve or maintain a conformation competent for ligand binding. KDAC6 is largely responsible for Hsp90 deacetylation but other KDACs may target it depending on its location in the cell. Depletion of KDAC1 by siRNA or treatment with the class I-selective HDACi, MS-275 results in increased Hsp90 acetylation specifically in the nuclear compartment, suggesting that KDAC1 can also deacetylate Hsp90 (117). Unliganded steroid receptors are known to shuttle between the nuclear and cytoplasmic compartments (118) where they would be exposed to either KDAC6 or KDAC1. However, the 24-hour exposure time to siRNA or MS-275 used in this study leaves open the possibility of secondary or indirect effects on Hsp90 acetylation.

Third, regulation of steroid receptor acetylation by KDACs may influence their nuclear accumulation or ability to bind DNA. Hinge-region acetylation of PR and AR impacts the efficiency of nuclear translocation and accumulation upon exposure to agonist (Figure 3). In the case of AR, acetylation of hinge-region lysines allows nuclear accumulation, whereas acetylation of PR in the hinge region significantly slows nuclear translocation. Acetylation of hinge region lysines in GR and MR does not appear to have an effect on nuclear accumulation, whereas its impact on translocation of ERα has not been directly investigated. In vitro experiments indicate that acetylation of steroid receptors may also modulate their ability to associate with DNA or chromatin (Figure 2). Hinge-region acetylation of GR and PR impairs their association with target genes in a gene-selective fashion. MR hinge-region acetylation inhibits its association with 2 target genes tested. In contrast, ERα acetylation in the DBD region stimulates its ability to bind DNA in vitro.

Although solid progress has been made, the study of steroid receptor acetylation is challenging because it is difficult to detect. Acetylation site-specific antibodies that would allow rigorous demonstration of steroid receptor acetylation in vivo are not yet available. Many of the studies done to date have used cell lines with overexpressed receptors and the corresponding acetylation mutants. In addition, few target genes have been examined in functional studies of receptor acetylation; some studies relied solely on transfected reporter genes. Thus, the generality of the functional impact of receptor acetylation of DNA binding and transcriptional regulation needs further attention. Finally, it appears that the hinge-region lysines of steroid receptors are sites of other modifications, including methylation and ubiquitylation. The use of receptor mutants to study the function of receptor acetylation leaves open the possibility that other modifications cause the functional effects measured. Mass spectrometry might address whether different hinge-region modifications are mutually exclusive, occur simultaneously, or are observed in distinct cellular compartments.

The fourth way in which KDACs impact steroid receptor signaling is at the transcriptional level. There is ample evidence that KDACs cooperate with steroid receptors in both repression and activation of transcription at selected target genes. The targets of KDAC action at gene promoters and enhancers have traditionally been considered to be histones. Based on the known functions of histone acetylation, the role of KDACs in transrepression by steroid receptors would be predicted. Indeed, loss of histone acetylation was observed at promoters that required KDAC activity for repression in studies described above (67, 78, 79). It is harder to understand how KDACs might facilitate receptor-activated transcription by deacetylating histones. One possibility involves deacetylation of histones as a promoter-reset mechanism. Steroid receptors and coregulators have been shown to associate with some target genes in a cyclic fashion such that transcription occurs in waves (85, 119, 120). Between rounds of transcription most transcriptional regulatory proteins dissociate from the target gene and histones are deacetylated, presumably by KDACs that associate with the target gene late in a transcription cycle (120). As suggested by Welsbie et al (85) in their study of AR transactivation, hyperacetylated nucleosomes in a promoter region may not allow for promoter resetting that is necessary for a subsequent round of transcription. To our knowledge this hypothesis has not been directly tested, and there are some issues with this idea. First, it is not clear that all steroid-regulated genes undergo waves of transcription. In fact, John et al (121) found the genomic response to glucocorticoids to be kinetically complex. Among genes up-regulated by glucocorticoids, they observed both continuous and transient activation over a 24-hour period. In the latter group, accumulation of nascent transcripts peaks within 1–2 hours and then falls off rapidly without subsequent increases. Second, given that histone acetylation is tightly associated with transcriptional activity, it is unclear how acetylated nucleosomes might block the reassembly of transcription complexes once a transcription cycle has been completed. In fact, Wang et al (15) demonstrated that deacetylase activity associated with promoters of genes primed for activation in resting T cells prevents both histone acetylation and RNA pol II association, indicating that increased histone acetylation facilitates RNA pol II recruitment. This subject clearly requires further study.

Another possible mechanism to explain the positive role of KDACs in transcriptional activation is that they remove acetylation from nonhistone targets. Steroid receptors recruit multiple KATs to target genes, and it is not unreasonable to suggest that they acetylate promoter-associated proteins other than histones, including the receptors themselves. As discussed above, acetylation of GR, PR, and possibly MR inhibits their ability to associate with binding sites in a gene-selective fashion (83, 84, 115). As shown in Figure 4, we propose that KDACs may facilitate transactivation at target genes in 2 other ways, using GR as an example. Many transcriptional coregulators recruited to target genes by GR and other steroid receptors are acetylated, including components of KAT-containing complexes, ATP-dependent chromatin remodelers, histone methyltransferase complexes, the mediator complex, and TFIID (21). Complexes containing class I KDACs may remove inhibitory acetylation from components of coactivator complexes that GR recruits to its binding regions (Figure 4A). These acetylations may inhibit complex function or affect complex stability. In support, Chen et al (108) showed that p300-mediated acetylation of SRC3 causes its dissociation from ERα. Alternatively, these KDACs may remove acetylation from components of the basal transcription machinery that inhibits efficient stimulation of transcriptional initiation (Figure 4B). A case in point, repression of the MMTV promoter by HDACi is mediated through its core promoter; GR binding and associated chromatin remodeling are unimpaired (30, 63).

Figure 4.

Model for the role of KDACs 1 and 2 in GR transactivation. A, Many coactivators of GR transactivation are known to be acetylated. KDACs may facilitate GR action at target genes through removal of inhibitory acetylation on GR-assembled coactivator complexes or those already present in GR-binding regions of target genes. B, KDACs remove acetylation from basal transcription machinery associated with the core promoters of target genes that interfere with efficient interactions with GR-containing transcription complexes at distant GR-binding regions. The KDAC1/2-containing complexes and their mechanism of recruitment to GR target genes have not yet been defined. Ac, acetylation.

The studies described in this review show repeatedly that gene context is very important in determining whether KDACs are required for steroid receptor action and how they impact it. The involvement of transcriptional coregulators as targets of KDACs provides an explanation for the gene-selective effects of KDACi or KDAC depletion on steroid-regulated transcription. Such proteins may be required for steroid receptor action in a context-dependent fashion as was clearly demonstrated for ERα target genes in a recent study (122). The functional impact of acetylation on coregulators needs to be investigated to fully understand the role of KDACs in steroid receptor transactivation.

The cooperation between steroid receptors and KDACs has clinical ramifications since several HDACi have been approved for use in humans. In particular, valproate (VPA) has been used to treat epilepsy and bipolar disorder for many years. For these indications, patients take valproate daily. Many studies have shown that up to 50% of patients on VPA experience metabolic (weight gain, dyslipidemia, hyperinsulinemia) and reproductive abnormalities (reduced fertility, menstrual problems) that may increase their risk for more serious problems, such as cardiovascular disease (123, 124). Because nuclear receptors are essential for regulation of reproduction and metabolism and require the action of KDACs for function, it is possible that the side effects are mediated, in part, by modulation of nuclear receptor action HDACi are now being evaluated in both preclinical and clinical studies for their efficacy against a variety of diseases including HIV and neurologic and inflammatory disorders (9, 125). Thus, human exposure to these drugs may increase in the future. A better understanding of the function of KDACs in signaling and metabolic pathways is needed to fully appreciate the effects these drugs might have on human physiology and guide usage of these drugs to minimize their negative effects.

Funding Statement

This work was supported by National Science Foundation grant MCB-1122088 (to C.L.S.).