Nuclear receptor corepressor complexes in cancer: mechanism, function and regulation

Abstract

Nuclear receptor corepressor (NCoR) and silencing mediator for retinoid and thyroid hormone receptors (SMRT) function as corepressors for diverse transcription factors including nuclear receptors such as estrogen receptors and androgen receptors. Deregulated functions of NCoR and SMRT have been observed in many types of cancers and leukemias. NCoR and SMRT directly bind to transcription factors and nucleate the formation of stable complexes that include histone deacetylase 3, transducin b-like protein 1/TBL1-related protein 1, and G-protein pathway suppressor 2. These NCoR/SMRT-interacting proteins also show deregulated functions in cancers. In this review, we summarize the literature on the mechanism, regulation, and function of the core components of NCoR/SMRT complexes in the context of their involvement in cancers and leukemias. While the current studies support the view that the corepressors are promising targets for cancer treatment, elucidation of the mechanisms of corepressors involved in individual types of cancers is likely required for effective therapy.

Introduction: discovery of NCoR and SMRT as promiscuous nuclear receptor corepressors

The ‘on’-and-‘off’ control of gene expression involves a cascade of transcription factors and is assisted by two classes of cofactors. These cofactors can activate (the coactivators) or repress (the corepressors) gene transcription via modifying specific residues of histones, thereby regulating the accessibility of chromatin to the basal transcription machinery [1]. Two of the first identified corepressors are nuclear receptor corepressors (CoRs), which include NCoR (nuclear receptor corepressor) [2] and SMRT (silencing mediator for retinoid and thyroid hormone receptors) [3]. Prior to the cloning of NCoR and SMRT, several studies have reported that thyroid hormone receptors (TR) and retinoic acid receptors (RAR) have an intrinsic ability to mediate ligand-independent repression, in a manner that is dependent on soluble and titratable factors [4-6]. The search for these factors led to the cloning of NCoR [2] and its homologous protein SMRT [3] (a truncated version) as well as full-length SMRT [7,8]. Binding of TR and RAR to their respective ligands, thyroid hormone (T3) and retinoic acid (RA), disrupts their interactions with NCoR or SMRT, allowing for coactivator recruitment and subsequent activation.

Subsequent studies reveal that NCoR and SMRT also mediate ligand-independent interaction with various other nuclear receptors (NRs), including vitamin D receptors (VDR) [9], peroxisome proliferator-activated receptors (PPAR) [10], liver X receptors (LXR) [11], and with orphan receptors including Rev-Erb [12,13], chicken ovalbumin upstream promoter transcription factor (COUP-TF) [14], and dose-sensitive sex reversal-AHC critical region on the X chromosome, gene 1 (DAX-1) [15]. NCoR and SMRT also have the ability to interact with steroid hormone receptors, namely, estrogen receptor (ER) [16], androgen receptor (AR) [17,18], and progesterone receptor (PR) [19]. Although these interactions are generally weaker than the interactions with non-steroid receptors, the steroid receptor interactions with NCoR/SMRT can, nevertheless, be stabilized by binding to the corresponding antagonists. One of the best-studied examples is tamoxifen, an ERα antagonist used to treat ERα-positive breast cancers. Tamoxifen binding to ERα stabilizes the interaction between ERα and NCoR, which is considered a contributing factor for tamoxifen inhibition of ERα target gene expression [20].

Mechanism of action of NCoR/SMRT

NCoR and SMRT are 270 kDa proteins that share highly homologous domains. The C-terminal region of NCoR and SMRT contains three (NCoR) or two (SMRT) nuclear receptor interaction domains (IDs) that mediate direct interactions with the ligand-binding domain (LBD) of NRs [2] (Figure 1). The functional motifs of IDs contain the so-called CoRNR boxes, which have the consensus sequence L-X-X-I/H-I-X-X-X-L/I (L: leucine, I: isoleucine, X: any amino acids) [21-25]. The binding affinity is predominantly regulated by the conformational change of NR LBD domains that takes place in response to ligands (agonists or antagonists). Binding of agonists or antagonists to NRs induces the movement of the helix 12 region of NRs whose position dictates the binding ability of NCoR/SMRT. In general, agonists destabilize the corepressor interaction, resulting in the dismissal of CoRs and subsequent recruitment of coactivators to drive agonist-dependent transcriptional activation. Antagonists do the opposite by stabilizing NCoR/SMRT interactions. In addition, the specificity and fine tuning of the binding between NRs and N-CoR/SMRT are also shown by the preferential use of different CoRNR boxes within a single CoR molecule and the preferential use of NCoR or SMRT for the binding [26,27].

Figure 1.

Diagram of NCoR/SMRT corepressors showing their regions involved in various interactions with nuclear receptors, other transcription factors as well as interactions with other components of the core complex.

Biochemical purification of NCoR and SMRT has shown that both NCoR and SMRT are present in large protein complexes (1.6-2 MDa size) that also contain histone deacetylase 3 (HDAC3), transducin β-like protein 1 (TBL1)/TBL1-related protein 1 (TBLR1), and G-protein pathway suppressor 2 (GPS2) [28-30]. It is likely that the core complex contains four subunits, two of which can be either NCoR or SMRT, and either TBL1 or TBLR1. A recent work has shown that TBL1 may exist as a tetramer [31]. The molecular weight of the core complex is about 412 KDa. Multiplication of this by 4 gives rise to ~1.6 MDa, which is in the range of the observed size (1.6-2 MDa). Clearly, the NCoR/SMRT complex should also contain other sub-stoichiometric components, such as Sin3A/HDAC1 [32-37], Class II HDACs [38,39], the subunits of chromatin-remodeling complexes [40], KAP-1 [40], and histone acetyl-transferases such as CBP [1,41-43]. The functional interactions of these sub-stoichiometric components with the core NCoR/SMRT complex may offer additional layers of regulation of target gene transcription.

NCoR/SMRT nucleates the formation of the core complex through interactions provided by the N-terminal portion of the proteins. Historically, the N-terminus of NCoR and SMRT has been characterized to contain three independent repression domains (RD1, RD2 and RD3) [2]. While the function of RD2 and RD3 has yet to be clarified, RD1 has been shown to mediate NCoR/SMRT interactions with both GPS2 and TBL1/TBLR1 [28,44]. Interestingly, the region of RD1 that interacts with TBL1/TBLR1 appears to overlap with the recently-reported interaction with the DNA-binding domain of ERα [45] (Figure 1).

None of the three RD domains interacts with HDAC3. A previously-noted motif in NCoR and SMRT is the SW13/ADA2/NCoR/TFIIB (SANT) domain, two copies of which are located between RD1 and RD2 [46]. The first SANT domain along with a short upstream region is referred to as the deacetylase activation domain (DAD), which directly binds to HDAC3. The binding involves a conformational change of DAD and is critical for the activation of the enzymatic activity of HDAC3 [28,47,48]. Interestingly, a recent work [49] by Adikesavan et al. has shown that the DAD of SMRT can directly bind to p53 (Figure 1). This binding blocks HDAC3 interaction with DAD leading to a net increase in histone acetyltransferase (HAT) activities, which contributes to the activation of p53 target gene in response to DNA damage. SMRT (but not NCoR) has also been shown to function as a coactivator for ERα in MCF-7 breast cancer cells [50], underscoring a specificity between SMRT and NCoR in regulating transcription.

The second SANT domain in CoRs has been shown to function as a histone interaction domain (HID), in line with the reported role of SANT domains in recognizing histones in the context of other proteins [51]. By occupying the nonacetylated histone tails, HID inhibits HAT activities, suggesting a feed-forward mechanism by which the two SANT motifs cooperatively promote histone deacetylation and repression of target genes [52,53]. Similar modes of cooperation have also been reported between HDAC3 and TBL1, which can also directly bind to histones [30,54,55].

Although NCoR and SMRT were originally identified as transcriptional corepressors for NRs, they have emerged as promiscuous corepressors for many other sequence-specific transcription factors that function in different cellular processes [42]. Thus, various studies have shown that SMRT and NCoR interact with Bcl-6/LAZ3, a transcription factor recently shown to play an important role in repressing inflammation [56-59], MyoD [60], HES-related repressor proteins [61]. NCoR/SMRT also interact with the evolutionarily related POU homeodomain factors Pit-1 [62,63], Oct-1 [64], the Notch-activated adapter protein Su(H)/RBP-Jκ/CBF1 [65-67], Pbx [68,69], serum response factor (SRF) [70], NF-κB [71-74], AP-1 [70], and signal transducers and activators of transcription 5 (STAT5). Other factors that are associated with NCoR and SMRT include PLZF [75,76], ETO family proteins involved in acute myeloid leukemias [77-79], SMADs [80], c-Myb [81], aryl hydrocarbon receptor [82], Sharp [83], and Kaiso [84].

Two studies have added new insight into the function and mechanism of CoR-mediated repression. Ebert et al. showed that NCoR can interact with methyl-CpG-binding protein 2 (MeCP2) and phosphorylation of T308 blocks the interaction of the MeCP2 repression domain with NCoR and suppresses the ability of MeCP2 to repress transcription [85], thus linking NCoR to DNA methylation-dependent gene silencing. Another work shows that the NCoR complex binds and deacetylases P-TEFb, an elongation factor important for RNA Pol II-mediated transcription [86], thus revealing a role for corepressors in deacetylating general transcription factors to regulate transcription.

Deregulation of NCoR/SMRT corepressors and other complex subunits in cancer

ER and AR are steroid receptors that play important roles in the initiation and progression of breast and prostate cancers. NCoR/SMRT interact with both ER and AR and serve as potential drug targets in the treatment of these cancers. Recent genome-wide studies have shown that NCoR/SMRT and HDAC3 bind to thousands of genes [56,87], consistent with their use by other transcription factors such as c-Jun and NFκb that impact on fundamental cellular activities. In this section, we summarize the literature on the involvement of NCoR/SMRT and other complex subunits in cancers and leukemias. NCoR and SMRT have been largely studied in the context of ER, AR and leukemia fusion proteins. While previous studies have focused on NCoR, SMRT and HDAC3, some new development has been reported for TBL1/TBLR1 and GPS2, suggesting their important roles in cancer development.

NCoR and SMRT subunits

Breast cancer

ERα is expressed in approximately 75% of breast cancers where ERα - dependent activation plays an important role in the growth of these cancers. NCoR/SMRT are among the best-characterized ERα corepressors [88]. Since coactivators and corepressors compete for ERα binding, the ratio between coactivators and corepressors is expected to modulate ERα activity in benign and malignant cells. Results from independent studies have supported this model by showing that diminished expression of NCoR/SMRT drives breast cancer initiation and progression. For example, an unbiased pathway analysis has revealed multiple alterations linked to loss of NCoR and SMRT corepressor complexes in luminal A breast tumors [89]. In another study that examines ERα corepressor levels in breast cancer, NCoR levels were found to be downregulated in invasive ductal carcinomas [90].

Downregulation of NCoR/SMRT has also been implicated in endocrine resistance. Selective ER modulators (SERMs), such as tamoxifen, an ERα antagonist, are beneficial in the initial treatment of ERα-positive breast cancer, because these SERMs can inhibit cancer growth by inhibiting the transcriptional activity of ERα. The activity of tamoxifen is in part attributed to its ability to stabilize the binding of ERα to corepressors [91]. Reduced expression of NCoR/SMRT corepressors has been shown to contribute to tamoxifen resistance in breast cancers [92-95]. Two studies have provided mechanistic insights into signal-dependent downregulation of NCoR and SMRT in breast cancer cells at the level of protein degradation, which contributes to cancer progression and acquisition of endocrine resistance. Frasor et al. reported that NCoR level was reduced due to estrogen-dependent upregulation of Siah2, which functions as an E3 ligase for NCoR [96,97]. Interestingly, two recent studies have shown that downregulation of Siah2 is correlated with acquisition of breast cancer tamoxifen resistance [98,99]. In another work, Stanya et al. reported that Pin1 can promote proteasomal degradation of phosphorylated SMRT in response to HER-2 signaling [100].

Several studies have reported alternative mechanisms by which NCoR regulates ERα-dependent transcription in breast cancers. First, NCoR has been shown to regulate ERα expression. More than half of the ERα-positive breast cancers also express PR. Elevated expression levels of PR-B isoform increase the interaction of the receptor with NCoR on the half-PRE site of the ERα promoter, an event incompatible with PR-coactivator interactions. Silencing of NCoR was able to reverse the down-regulation of ERα expression induced by PR-B overexpression [101]. In another work, by using chromatin immunoprecipitation (ChIP) assays, Konduri et al. [102] have demonstrated that ERα represses p53-mediated transcriptional activation in human breast cancer cells by recruiting NCoR/SMRT and HDAC1. Finally, reduction in the ratio of ERβ to ERα expression appears to be correlated with breast tumorigenesis. Bartella et al. [103] showed that ERβ recruits NCoR, leading to hypoacetylation of histones and displacement of RNA-polymerase II at the ERα promoter, providing new mechanistic insight into the antagonism between ERβ and ERα in breast cancer cell growth.

Prostate cancer

AR is the driving force for prostate cancer development and progression. Many patients will benefit from the androgen deprivation therapy in combination with the treatment of AR antagonists (flutamide or bicalutamide). The underlying principle for the use of AR antagonists is that these compounds will compete with agonists for binding to AR, while stabilizing the binding between AR and NCoR/SMRT corepressors, thus preventing coactivator recruitment and AR-dependent activation of target genes [104,105].

The involvement of AR corepressors in prostate cancer has been well documented [106]. Earlier studies have suggested that a reduction of NCoR/SMRT levels may contribute to androgen-independent AR activation and prostate cancer progression [107-109]. Nevertheless, several studies have shown that the corepressor regulation of AR may be more complex than previously thought. It has been shown that overexpression of NCoR/SMRT does not repress AR-dependent gene expression in prostate cancer cell lines, but rather activates it [110]. A caveat is that overexpressed proteins may not faithfully recapitulate the regulation observed under physiological conditions. Another work has shown that the ability of NCoR to enhance antagonist-mediated AR repression is in fact modulated by post-translational modifications of NCoR. Protein kinase A (PKA) is able to directly bind and phosphorylate the Ser-70 residue in the RD1 domain of NCoR. The phosphorylation enhances nuclear localization of NCoR and potentiates the antagonist activity on AR-dependent transcription in prostate cells [111].

Several recent studies have re-affirmed the importance of AR-NCoR/SMRT axis in regulating AR function and prostate cancer progression. It has been shown that reduced recruitment and loss of corepressor SMRT/NCoR alter the ligand response and AR functions in a manner that contributes to prostate cancer progression [112]. In another study, Yoo et al. studied the role of AR corepressors during androgen-independent prostate cancer progression [113]. It was found that casein kinase 2 (CK2)-mediated phosphorylation of NCoR strongly correlates with androgen-independent growth and invasion of prostate cancer cells, and with poor prognosis of prostate cancer patients, suggesting that CK2-NCoR axis is a potential therapeutic target in prostate cancer. Perhaps a more exciting evidence for a direct role of NCoR in antagonizing androgen-independent prostate cancer growth is provided by the study showing that the ubiquitin ligase Siah2 switches AR from repressed to the activated state by selectively targeting NCoR bound, transcriptionally-inactive AR for ubiquitin-dependent degradation. Thus, Siah2 promotes activation of AR target genes, leading to the growth of androgen-independent prostate cancer cells [114]. Taken together, the consensus of the current studies is in line with the model that the efficacy of NCoR/SMRT corepressors in repressing AR activity is directly correlated with the recruitment of these corepressors on AR target genes. Reduction of such recruitment, which would predispose genes for activation, may occur through various mechanisms, such as reduced gene expression, proteasomal degradation or altered modification of the corepressors.

Leukemia

Leukemia is a prototypic type of cancer in which a pro-oncogenic function of NCoR and SMRT was first documented. One form of acute promyelocytic leukemia (APL) is caused by abbrant expression of PML-RARα (promyelocytic leukemia-retinoic acid receptor α) or PLZF-RARα (promyelocytic leukemia zinc finger-retinoic acid receptor α) fusion proteins. As a result of the fusion, the binding affinity between RARα and NCoR/SMRT is increased such that the corepressors cannot be released by the physiological dose of RA [75,115-118]. Accordingly, these fusion proteins promote leukemogenesis due to reduced sensitivity to retinoic acid-dependent transcriptional activation of target genes involved in cell differentiation. The involvement of NCoR/HDAC in the pathogenesis of leukemias is further confirmed by the studies showing that expression of an NCoR fragment that disrupts corepressor interaction restores RA sensitivity in resistant cells [119], and that the NCoR/HDAC complex is a key regulator of the transcriptional repression mediated by PML-RARα in vivo [120]. APL patients that have PML-RARα is sensitive to high doses of RA treatment, whereas APL patients harboring PLZF-RARα translocations do not. This difference is due to the fact that in PLZF-RARα, the PLZF moiety can also bind to NCoR/SMRT corepressors, which cannot be released by high doses of RA, thus explaining the phenotypic difference between PML-RARα and PLZF-RARα. The presence of trichostatin A, a specific inhibitor of histone deacetylases, increases histone acetylation and leads to transcription activation [115].

Two recent studies have provided new insight into the regulation of RARα fusion proteins by NCoR/SMRT. One study shows that PML-RARα and PLZF-RARα have gained the ability to recognize specific spliced variants of NCoR and SMRT variants that are poorly recognized by RARα [121]. Another work showed that failure to dissociate corepressors, or failure to recruit co-activators results in RA-resistant variants of the PML-RARα oncoprotein in APL [122].

The t(8;21) chromosomal translocation is involved in nearly 15% of total acute myeloid leukemia (AML) cases, which generates the AML1-ETO fusion protein between the DNA-binding domain of AML1 and a nearly full-length ETO. The role of NCoR/SMRT corepressors in AML was first revealed by their interaction with ETO, the fusion partner in the t(8;21) AML and NCoR/SMRT [77,123,124]. ETO is a member of a small family of corepressors that can form oligomers and bind to NCoR/SMRT and HDACs [125,126]. Its homologous protein ETO-2 is also involved in AML by fusing to AML1 [127]. The current literature is consistent with the model that the AML1-ETO/ETO-2 promote leukemogenesis in part by recruiting the NCoR/SMRT/HDAC to repress transcription of target genes [77,128-130]. Interestingly, a recent work has shown that another leukemia fusion protein E2A-Pbx1 drives leukemogenesis by escaping ETO-2-mediated repression in order to activate Hox target genes [131]. Thus, the role of corepressors in different types of leukemias is likely context-dependent, emphasizing the need to dissect the individual mechanisms for therapeutic targeting of the corepressors.

Other subunits of the NCoR/SMRT complex

HDAC3

HDACs deacetylate both histones and proteins, and play important roles in the regulation of chromatin biology and gene transcription [77,132-134]. The 18 human HDACs can be divided into 4 classes [28]. Class I, II and IV belong to Zinc-dependent, NAD-independent histone deacetylases, whereas class III belongs to NAD-dependent histone deacetylases. Class I HDACs include HDAC1, 2, 3 and 8 isoforms, which are encoded by separate genes. The Class I HDACs generally show ubiquitous expression, and HDAC1-3 have been well studied. HDAC1 and HDAC2 exist as a dimer in several HDAC1/2-shared complexes, including the NURD chromatin-remodeling complex [135,136], the Sin3A-containing corepressor complex and the CoREST/LSD1-containing histone deacetylase/histone demethylase complex [137,138]. Whereas HDAC1 and HDAC2 generally show constitutive nuclear localization, HDAC3 shows both cytosolic and nuclear localization [139]. Consistent with this, HDAC3 contains both nuclear import and export domains [140]. In the nucleus, HDAC3 specifically associates with NCoR/SMRT corepressor complexes [28].

Several studies have implicated a role for HDAC3 in cancer. Upregulation of HDAC3 has been observed in several types of cancers including colon, lung, prostate and breast cancers, in a manner that correlates with poor survival and prognosis of the patients [141-148]. HDAC3 has served as a potential drug target in the treatment of PML-RARα, PLZF-RARα and AML1-ETO leukemias [77,126,149,150]. Independent studies using cancer and leukemia-derived cell lines have confirmed pro-survival roles for HDAC3. The pro-survival activity of HDAC3 revealed by these studies may be mediated by its roles in transcriptional repression of growth-arrest genes such as p21 [55,151,152], cell cycle progression [153-155], genomic stability [156], or its involvement in apoptosis/cell death/survival pathways [141,142,150,157-159]. In addition to regulating cell proliferation, HDAC3 has also been shown to play an important role in metastasis of cancer cells through its ability to enhance epithelial-mesenchymal transition (EMT) [160]. HIF-1alpha-induced HDAC3 was found to be crucial for hypoxia-induced EMT and metastatic phenotypes. HDAC3 crosstalks with WDR5 to activate the expression of mesenchymal genes expression, while repressing epithelial gene expression. While these studies have supported an anti-tumorigenic function of HDAC3, Bhaskara et al. have shown that mice depleted of HDAC3 in liver develop hepatocellular carcinoma [156,161], suggesting that HDAC3 may have context-dependent function in different cancers.

An active area in HDAC study is the development of isoform-specific HDAC inhibitors [162-164]. While pan-HDAC inhibitors have promising effects in the treatment of cancers, the clinical use of these inhibitors is complicated by their off-target effects [165-168]. It has been proposed that HDAC3 may be responsible for most of the anti-tumor effects of HDAC inhibitors [154,169]. A few studies reported the development and use of HDAC3-specific inhibitors that target the active form of HDAC3 [169,170]. Inhibition of HDAC3 using such a selective inhibitor RGFP966 promotes apoptosis and diminishes cell growth in cutaneous T cell lymphoma cell lines due to DNA damage and compromised S phase progression [169].

An alternative strategy to inhibit HDAC3 function is to prevent its activation by blocking the formation of HDAC3 complexes with corepressors, based on findings that the enzymatic activity of HDAC3 is strictly dependent on its binding to NCoR/SMRT [28,171]. The critical importance of corepressor interaction for HDAC3 activity has been recently demonstrated in vivo [172], which shows that mice harboring loss-of-function mutations in the DAD domain of both NCoR and SMRT have essentially no active HDAC3 in all tested tissues, which correlates with globally-increased histone acetylation. These mice, unlike the embryonic lethality of the total HDAC3 knockout mice [173], are vital, suggesting that free HDAC3 is not non-functional. A follow-up study by the Lazar group [174] recently showed that an HDAC3 mutant that is enzymatically inactive but retains the ability to bind to corepressors can rescue the HDAC3-null defect in liver, indicating that HDAC activity is not required for all HDAC3 functions, but its interaction with corepressors such as NCoR is critical. This would challenge the rationale of using traditionally-developed HDAC inhibitors that target the active site of HDACs to inhibit the function of HDAC3 (and possibly other HDACs) in cancers. Nevertheless, the above-mentioned strategy of using inhibitors to prevent the formation of stable HDAC3-corepressor complexes would continue to be warranted in targeting HDAC3 function in cancers.

It has been shown that the inactive form of HDAC3 is in complexes with Hsc70 and TRiC chaperones [139]. These chaperones are released from HDAC3 in its mature form after corepressor association. A recent work has shown that knockdown of NCoR and SMRT in cultured cells reduces the steady-state protein expression of HDAC3 in all tested cell types including colon, prostate and breast cancer cells [175], indicating that the E3 ligase-driven HDAC3 degradation pathway is still intact in cancer cells. Elucidating the nature of HDAC3-specific E3 ligases may offer a new approach to target HDAC3 in cancer cells. It is noteworthy that the steady-state level of liver HDAC3 in the above-mentioned DAD mutant knock-in mice does not differ significantly from the control mice. The difference between living animals and cultured cell may be caused by the use of different assay conditions. Alternatively, the mice may have been forced to compensate for the loss of HDAC3 given that HDAC3 is required for the early embryonic development.

Previous studies have suggested that the C-terminus of HDAC3 is required for its interaction with CoR [140,171]. Surprisingly, it was shown that if proteasomal activity is blocked, a C-terminus-truncated HDAC3 can still form a complex with CoR in cells [175]. Proteasome-dependent rapid turnover thus may explain earlier failures to detect an interaction between CoR and the C-terminus-truncated HDAC3. Despite the binding to CoR, the C-terminus-truncated HDAC3 remains associated with chaperones and cannot be activated. These data are consistent with a novel function for HDAC3 C-terminus in coupling CoR binding, chaperone dismissal, and subsequent activation and stabilization of HDAC3, suggesting that the C-terminus acts at a step after the binding of corepressors to facilitate the release of chaperones from HDAC3 (Figure 2). A role for C-terminus in regulating HDAC3 activity is also supported by the finding that phosphorylation of the conserved C-terminal residue S424 increases the activity of HDAC3 [176]. A recent work reported the 3D structure of a complex containing a C-terminus-truncated HDAC3 and SMRT-DAD [177]. It is worth noting that the C-terminus is cleaved after the formation of the complex, consistent with its role involved in the complex formation and HDAC3 activation.

Figure 2.

A hypothetical model of HDAC3 activation by corepressors. In the inactive state, HDAC3 is complexed with Hsc70 and TRiC chaperones. Corepressors bind to HDAC3 N-terminus through the DAD domain. The DAD-triggered HDAC3 activation occurs in a stepwise fashion and involves the formation of an intermediate DAD-HDAC3-chaperone complex, whose subsequent conversion into the mature, active state requires a C-terminus dependent conformational change to dissociate chaperones. “***” denotes exposed HDAC3 active site.

TBL1 and TBLR1

TBL1 is a protein initially found to play a role in the X-linked late-onset deafness in human [178]. Its counterpart in fly, ebi, has been shown to play a role in fly eye development by regulating the EGF receptor-mediated signaling pathways [179]. Interestingly, Lazar and Wong groups later discovered that TBL1 is a stoichiometric subunit of NCoR and SMRT nuclear receptor corepressor complexes [30,44]. TBL1 directly interacts with histone tails and facilitates the function of HDAC3 [30,44]. By purifying FLAG-NCoR and FLAG-HDAC3 interacting proteins, Zhang et al. discovered a TBL1-like protein, which they named TBLR1 (TBL1-like protein 1) [28]. TBLR1/TBL1, like NCoR and SMRT, are the alternative subunits of CoR/HDAC3 corepressor complexes. Interestingly, the expression of TBLR1 is reduced in mature hematopoietic cells compared to progenitor cells [180], suggesting a role for TBLR1 in regulating cell differentiation.

Both TBL1 and TBLR1 have been shown to have “coactivator” functions. Although the classical Gal4-fusion assay failed to support an “activation domain” in TBL1/TBLR1, it confirmed that TBL1/TBLR1 harbor transferable repression domains [30,44]. This shows that TBL1/TBLR1 are not classical coactivators. The “coactivator” activity of TBL1/TBLR1 appears to be attributed to their ability to drive proteasomal degradation of NCoR/SMRT and CtBP corepressors in a signal-dependent fashion [181]. Perissi et al. showed that TBL1/TBLR1 promote the clearance of NCoR/SMRT and CtBP corepressors by recruiting 19S proteasome particles to drive the ubiquitination and degradation of NCoR/SMRT and CtBP, thus allowing the recruitment of coactivators. This takes place not only on NRs but also on c-Jun and NF-κB binding sites of affected genes [181,182]. Phosphorylation of TBL1 and TBLR1 on specific residues is required to activate their E3 ligase activities. TBL1 and TBLR1 have imperfect F-boxes. Therefore, an interesting model is that these proteins normally reside in the NCoR/SMRT complex in the inactive state. Phosphorylation and/or possibly other post-translational modifications activate their latent E3 ligase activities, leading to degradation of the corepressors.

Dysregulation of Wnt-β-catenin signaling pathway has been associated with tumorigenesis. Another form of post-translational modifications on TBL1/TBLR1, SUMOylation, has been shown to dissociate TBL1/TBLR1 from the NCoR complex, and subsequently allow them to complex with β-catenin to dock on and activate the promoter of Wnt target genes, which is important for the tumorigenicity of SW480 colon cancer cells [183]. These results suggest that the nature of post-translational modification of TBL1 and TBLR1 determines whether they may function as tumor suppressors or as oncogenes to facilitate repression or activation of context-dependent target genes.

A recent development has implicated TBLR1 involvement in AR regulation and prostate cancer progression. TBLR1 is found to be primarily localized in the nucleus of benign prostate cells and the level is reduced in prostate cancer cells. TBLR1 binds to AR and potentiates AR-dependent transcription of target genes in a phosphorylation- and 19S proteasome-dependent manner. Interestingly, TBLR1 selectively activates genes important for growth suppression and differentiation but not pro-proliferative AR target genes, suggesting a tumor suppressor function of TBLR1 in prostate cancer [184].

GPS2

In search for proteins that can rescue the lethal phenotype of yeast harboring a dominant active yeast Gβγ mutant, Spain et al. have discovered GPS2 as a G-protein pathway suppressor and shown that GPS2 can suppress RAS/MAPK-dependent JNK activation [185]. Another group independently found that GPS2 interacts with Tax, a human T-cell lymphotrophic virus transcription factor, and suppresses TNFalpha-dependent activation of the JNK pathway [186]. Biochemical purification of novel NCoR/HDAC3-interacting proteins identified GPS2 as an integral component of the NCoR/SMRT/HDAC3 complex [28]. A coiled-coil region of GPS2 (amino acids 1-105) directly interacts with the N-terminal regions of both NCoR/SMRT and TBL1/TBLR1 to form a heterotrimeric complex. A recent structural study has confirmed these domain interactions and also shown that TBL1 is organized into a tetrameric structure [31].

Since the discovery of GPS2 as a G-protein pathway suppressor, GPS2 has emerged as a multifunctional protein. GPS2 can act both as a corepressor and a coactivator for various transcription factors including nuclear receptors [187-194], consistent with the presence of distinct repression and activation domains in GPS2 [28,187]. Consistent with its coactivator function, GPS2 has also been shown to facilitate histone demethylation of certain target genes [189,195]. These pro-activation activities of GPS2 may be independent of its interactions with NCoR/SMRT and TBL1/TBLR1. In the context of repression, it has been proposed that GPS2 may, analogous to NCoR/SMRT, tether the corepressor complex to sequence-specific transcription factors. While this idea is still interesting, and in fact is consistent with the reported anti-inflammatory function of GPS2 [196], the nature of transcription factors that use GPS2 to recruit the CoR complex, and the corresponding target genes regulated by the transcription factors, remain to be determined.

Several recent studies have provided evidence that the interaction between GPS2 and CoRs may be a drugable target in cancer. Bi et al. [197] reported that GPS2 is SUMOylated at K45 and K71 within the coiled-coil region that interacts with CoRs. SUMOylation increases the corepressor function of GPS2 for ERα-mediated transcription, initially reported by Cheng et al. [194], in both MCF-7 and T47D breast cancer cells due to enhanced GPS2 incorporation into the TBL1/SMRT complex, which correlates with reduced proliferation of MCF-7 and T47D cells.

Two recent studies reported somatic mutations of GPS2 in a form of aggressive brain tumor, Medulloblastoma [198]. These mutations affect the residues in the coiled-coil region of GPS2 that mediate its interaction with NCoR/SMRT and TBL1/TBLR1. The mutations significantly worsen the prognosis [199]. The impact of these mutations on the GPS2-CoR interaction has yet to be studied. Given the anti-inflammatory function of GPS2, one may speculate that such mutations may reduce the ability of the corepressor complex to suppress cytokine-dependent signaling pathways, which may predispose the affected cells to oncogenesis.

GPS2 turns out to be the first CoR complex subunit that forms fusion proteins with different partners in cancers. Undifferentiated spindle cell sarcoma (UDS) is a type of cancer with poorly defined diagnostic markers. O’Meara et al. identified a novel chromosomal translocation t(17;19)(p13;q13) in a pediatric UDS [200]. Interestingly, this in-frame fusion is between GPS2 and MLL4, which is the active subunit of one of the histone H3K4 methyltransferase complexes [201]. This fusion is oncogenic because it promotes anchorage-independent growth of the cells. It will be interesting to determine whether the fusion protein-containing complexes show multiple histone modification activities conferred by MLL4 (methyltransferase) and GPS2 (deacetylase and/or demethylase), which may be important for deregulation of gene expression in tumors. Besides MLL4, GPS2 has also been reported to fuse to other proteins in a prostate carcinoma cell line [202] and in glioblastoma multiforme [203]. These findings underscore the emerging important role of GPS2 in cancer.

Conclusion and future directions

Since the discovery of NCoR/SMRT nearly 20 years ago, tremendous progress has been made in understanding the mechanisms, regulation and functions of NCoR/SMRT and their interacting proteins in transcription, and their involvement in diverse pathways that regulate metabolism, inflammation and oncogenesis. A major advance in the field is the discovery of the multi-protein complex containing NCoR/SMRT, HDAC3, TBL1/TBLR1 and GPS2. However, to date, we still do not fully understand how these proteins regulate transcription in normal and cancer cells. Future studies may be devoted to understanding (i) the specificity of NCoR, SMRT, and their different isoforms, (ii) how the free form of HDAC3 assembles into the mature NCoR/SMRT-HDAC3 complex, (iii) the mechanism of “coactivator” functions of TBL1/TBLR1 and GPS2 and how these functions may be related to their interactions with NCoR/SMRT, (iv) the regulation of the expression of these proteins in normal and cancer cells, and (v) the genome-wide targets of these proteins. While the literature has supported the concept that deregulated corepressor function facilitates cancer development by disrupting the homeostatic balance between corepressors and coactivators (or other regulatory proteins) in the regulation of chromatin structure, transcription, DNA repair, inflammation and other important biological processes (Figure 3), corepressors may have different mechanisms and play either pro- or anti-tumorigenic roles in different types of cancers and leukemias. It is thus important to fully understand the context-dependent function of corepressors. Given the direct interactions between the various corepressor components and the transcription factors (such as ER, AR and leukemia fusion proteins) involved in cancers and leukemias, these corepressor proteins should serve as important drugable targets for cancer therapy.

Figure 3.

A working model showing that deregulated function of the corepressor complex disrupts homeostatic balance of cellular pathways to promote cancer progression.

Disclosure of conflict of interest

None.