Nuclear Receptor Corepressor Is a Novel Regulator of Phosphatidylinositol 3-Kinase Signaling

Abstract

The nuclear receptor corepressor (NCoR) regulates the activities of DNA-binding transcription factors. Recent observations of its distribution in the extranuclear compartment raised the possibility that it could have other cellular functions in addition to transcription repression. We previously showed that phosphatidylinositol 3-kinase (PI3K) signaling is aberrantly activated by a mutant thyroid hormone β receptor (TRβPV, hereafter referred to as PV) via physical interaction with p85α, thus contributing to thyroid carcinogenesis in a mouse model of follicular thyroid carcinoma (TRβ^PV/PV mouse). Since NCoR is known to modulate the actions of TRβ mutants in vivo and in vitro, we asked whether NCoR regulates PV-activated PI3K signaling. Remarkably, we found that NCoR physically interacted with and competed with PV for binding to the C-terminal SH2 (Src homology 2) domain of p85α, the regulatory subunit of PI3K. Confocal fluorescence microscopy showed that both NCoR and p85α were localized in the nuclear as well as in the cytoplasmic compartments. Overexpression of NCoR in thyroid tumor cells of TRβ^PV/PV mouse reduced PI3K signaling, as indicated by the decrease in the phosphorylation of its immediate downstream effector, p-AKT. Conversely, lowering cellular NCoR by siRNA knockdown in tumor cells led to overactivated p-AKT and increased cell proliferation and motility. Furthermore, NCoR protein levels were significantly lower in thyroid tumor cells than in wild-type thyrocytes, allowing more effective binding of PV to p85α to activate PI3K signaling and thus contributing to tumor progression. Taken together, these results indicate that NCoR, via protein-protein interaction, is a novel regulator of PI3K signaling and could serve to modulate thyroid tumor progression.

Thyroid hormone nuclear receptors (TRs) are ligand-dependent transcription factors that mediate the biological activities of thyroid hormone (T3) in growth, development, differentiation, and maintenance of metabolic homeostasis. There are two TR genes, TRα and TRβ, located on chromosomes 17 and 3, respectively, that encode four major T3-binding TR isoforms (α1, β1, β2, and β3). The TRs are ligand-dependent transcription factors, consisting of modular functional structures with the N-terminal A/B, central DNA-binding, and C-terminal ligand-binding domains. In the presence of T3, TRs associate with coactivators to regulate target gene transcription. In the absence of T3, TRs assume a different conformation that recruits corepressors to mediate gene silencing. This ligand-dependent switch in recruitment of coactivators or corepressors caused by TRs alters chromatin structures to signal changes in transcription programs.

In the past decades, strides have been made in understanding the role of corepressors in the biology of TRs. These advances have been mainly focused on the actions of corepressors in nucleus-initiated transcription of TR. However, recent studies show that the nuclear receptor corepressor (NCoR) is localized not only in the nucleus but also in the cytoplasm (6, 16). The redistribution of nuclear NCoR to the cytoplasm provides a mechanism for controlling differentiation of neural stem cells into astrocytes (6). Moreover, Sardi et al. showed that cytoplasmic NCoR forms complexes with a cleaved product of ErbB4 (a member of the epidermal growth factor receptor family) and the signaling protein TAB2 and translocates into the nucleus to regulate astrogenesis in the developing brain (16). Still, it is not clear whether, in addition to its presence in neural cells, NCoR is also localized in the cytoplasm of other cell types to mediate cellular functions that are independent of transcription regulation.

Our recent discovery that TRβ or its mutant TRβPV (PV) forms complexes with p85α, the regulatory subunit of phosphatidylinositol 3-kinase (PI3K), and activates PI3K signaling (5) provides an opportunity to address the functional role of cytoplasmic NCoR in the context of TR biology. PV was identified in a patient with resistance to thyroid hormone (21). PV has a C insertion at codon 448 that produces a frame shift in the carboxyl-terminal 14 amino acids of TRβ1 (13), resulting in the complete loss of T3 binding activity and transcriptional capacity (11). A knock-in mutant mouse harboring the PV mutation that recapitulates human resistance to thyroid hormone (TRβPV mouse) was created (8). Moreover, as a homozygous TRβ^PV/PV mouse, it spontaneously develops a follicular thyroid carcinoma similar to human thyroid cancer through pathological progression of capsular invasion, vascular invasion, anaplasia, and metastasis (18, 24). Using this mouse model, we found that PV physically associates with p85α to constitutively activate the PI3K activity and, via downstream effectors, to increase cell proliferation and motility to promote thyroid carcinogenesis in TRβ^PV/PV mice (5). That PV is also associated with NCoR in thyroid tumors of TRβ^PV/PV mice (1) raises the possibility that NCoR could play a role in thyroid carcinogenesis by modulating the interaction of PV with PI3K. In the present study, we tested this hypothesis and found that NCoR competed with TRβ or PV for binding to p85α in the nucleus as well as in the cytoplasm. An alteration of cellular NCoR protein levels by overexpression led to reduced PI3K/protein kinase B (AKT) signaling. Conversely, knocking down cellular NCoR by small interfering RNA (siRNA) approaches increased PI3K activity, phosphorylation of AKT, and cell motility. In thyroid tumors, cellular NCoR protein abundance was significantly lower than it was in wild-type thyroids, thereby facilitating the interaction of PV with p85α to activate PI3K signaling. Thus, the present study identified a novel function of NCoR that regulated PI3K signaling via protein-protein interaction to alter the activity of TRβ or PV independent of nucleus-initiated transcription.

MATERIALS AND METHODS

Mouse strain.

All aspects of animal care and experimentation were approved by the National Cancer Institute Animal Care and Use Committee. The mice harboring the TRβPV gene (TRβ^PV/PV mice) were prepared via homologous recombination as previously described (8). TRβ^PV/PV mice used in the present study were offspring of many generations of intersibling mating over 8 years (more than 40 generations). Littermates were used in phenotypic characterization in all studies. Genotyping was carried out by PCRs as previously described (8).

Primary thyroid cultured cells.

Primary thyroid cells from wild-type and TRβ^PV/PV mice were prepared and cultured in a manner similar to that described by Furuya et al. (5).

GST-binding assay.

Binding of [³⁵S]methionine-labeled TRβ1, PV, NCoR, and its truncated domains to glutathione S-transferase (GST)-p85α or binding of the [³⁵S]methionine-labeled receptor interacting domain (RID) of NCoR to GST-p85α or GST fused to truncated p85α was carried out as described by Furuya et al. (5). The plasmids for full-length and truncated GST-p85α proteins were kindly provided by J. K. Liao (Brigham and Women's Hospital and Harvard Medical School, Cambridge, MA). NCoR expression plasmids were generously provided by J. Wong (Baylor College of Medicine, Houston, TX). Around 1 μg of GST or GST-fused protein was used in each binding reaction. Their concentrations were determined by Coomassie blue-stained band intensities using bovine serum albumin standards after migration in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel. For each experiment, the SDS-PAGE gel was stained with Coomassie blue, dried, and autoradiographed. In vitro-translated [³⁵S]methionine-labeled proteins were synthesized using the TNT T7 quick coupled transcription/translation system according to the manufacturer's procedure (Promega, Inc., Madison, WI). For relative binding affinity studies, identical amounts of in vitro-translated TRβ1, PV, or NCoR RID were used. To that purpose, various volumes of [³⁵S]methionine-labeled TRβ1, PV, and NCoR RID lysates (1, 2, and 4 μl) were first analyzed by electrophoresis in an SDS-PAGE gel and autoradiographed, and the band intensities for each volume of protein lysates were quantified with NIH IMAGE software (ImageJ 1.34s; Wayne Rashband, NIH) (http://rsb.info.nih.gov/ij). NCoR RID, TRβ1, and PV contain different numbers of methionines (8 methionines for the RID protein and 13 methionines each for the TRβ1 and PV proteins), and therefore, the volume of protein lysates used for each reaction was adjusted accordingly.

Western blot analysis.

Fractionation of thyroid nuclear and cytosolic fractions of wild-type and TRβ^PV/PV mice was carried out in a manner similar to that described previously (5). Determination of the protein abundance of NCoR and the key regulators in PI3K-AKT pathways in the cytosolic and nuclear factions of thyroid extracts by Western blot analysis was carried out as described by Ying et al. (22). In coimmunoprecipitation experiments, the immunoprecipitation step and the subsequent Western blot analysis were carried out as described by Furumoto et al. (4).

Primary antibodies for phosphorylated S473 AKT (no. 9271) and total AKT (no. 9272) were purchased from Cell Signaling Technology, Inc., and used at a 1:1,000 dilution. Anti-matrix metalloproteinase 2 (anti-MMP2) antibodies were purchased from Santa Cruz Biotechnology, Inc. (SC0729; at a 1:500 dilution). Anti-p85α antibodies were purchased from Upstate (no. 06-195; 1:500 dilution) and from Santa Cruz (sc-423, sc-1637). Anti-NCoR antibody from Affinity BioReagent (no. PA1-844A; 1 μg/ml) was used. In some experiments, the polyclonal affinity-purified anti-antibody PHQQ (1.5 μg/ml), generously provided by T. Hollenberg (Harvard Medical School, Boston, MA), was used in Western blot analysis. Affinity-purified polyclonal anti-p110α antibodies were generous gifts from Jon Backer (1 to 2 μg/ml; Albert Einstein College of Medicine, Bronx, NY). Anti-histone deacetylase-3 (anti-HDAC3) antibody was obtained from NOVUS Biologicals (catalog no. NB 500-126; 1:1,000 dilution). For the negative control in coimmunoprecipitation, a mouse antibody, MOPC, was used (MOPC141; Sigma, Inc.). MOPC contains immunoglobulin G2bκ derived from ascites originated from mineral oil-induced plasmacytoma. For control of protein loading, the blots were stripped and rereacted with antibodies against α-tubulin (T6199; 1:500 dilution; Sigma) or poly(ADP-ribose) polymerase (PARP) (SC7150; 1:500 dilution; Santa Cruz).

Determination of PI3K activity.

Primary thyrocyte cultured cells or tumor cells of TRβ^PV/PV mice (5 × 10⁵ cells) were plated on 6-cm dishes. Forty nanomolars of control siRNA (siCONTROL nontargeting siRNA no. 2, D-001210-02-05) and siRNA against mouse NCoR (siGENOME SMART pool reagent no. L-058556-00; DHARMACON, Lafayette, CO) were transfected into cells with lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After 48 h, the cells were harvested and were homogenized with 400 μl of lysis buffer (50 mM HEPES [pH 7.5], 150 mM NaCl, 2 mM EDTA, 10 mM Na₄P2O₇, 2 mM Na₃VO₄, 10 mM NaF, 1% Igepal [Sigma], and 10% glycerol). After centrifugations at 4°C for 1 h at 14,000 × g, the protein concentrations of the supernatants were measured. The 300 μl of protein lysate was incubated with 5 μg of anti-p85α monoclonal antibody (SC1637; Santa Cruz) for 20 h. Forty microliters of 1:1 protein G-agarose (Roche, Inc.) was added and rocked gently for 2 h at 4°C. The subsequent processing of the protein G-agarose-anti-p85α monoclonal antibody-p85α immunocomplex and the PI3K activity was carried out using a PI3K enzyme-linked immunosorbent assay kit according to the manufacturer's instructions (Echelon Biosciences, Inc.) (see also reference 3).

Cell motility and proliferation assays.

Primary thyroid cultured cells or cultured tumor cells of TRβ^PV/PV mice were transfected with control siRNA or NCoR siRNA as described above. Forty hours after transfection, a modified Boyden chamber assay using serum as a chemoattractant was performed as previously described (10). Two independent experiments were performed, each in triplicate. Increases (n-fold) in motility were calculated as percentages of control values. For cell proliferation assays, 6 h after transfection of siRNA, the medium was changed to regular growth medium. After culture for an additional 24 h, cells were harvested and replated on new 3.5-cm dishes (4 × 10⁵ cells/dish). The numbers of cells were counted every 24 h, using a Coulter cell counter.

Determination of NCoR mRNA expression by real-time RT-PCR.

Primary thyroid cells isolated from wild-type mice and tumor cells from TRβ^PV/PV mice were preincubated with T3-depleted medium for 24 h and then added with T3 (100 nM) in the presence or absence of cycloheximide (100 μg/ml) for 24 h. Total RNA was isolated from six independent samples. To determine the effect of T3 on the expression of NCoR mRNA in wild-type primary thyroid cells and thyroid tumor cells from TRβ^PV/PV mice, real-time reverse transcription (RT)-PCR was carried out as described by Ying et al. (24). The following intron-flanking primer sequences were used for determination of NCoR mRNA: forward (positions 7387 to 7408), CCCTCTTCAACAGGTTCTACTC; reverse (positions 7594 to 7572), CACAGCTCAGTCGTCACTATCA. All PCR products were analyzed by agarose gel electrophoresis (2% agarose), followed by ethidium bromide staining to ensure amplification of the appropriately sized product.

Effects of IGF-1 on NCoR-mediated PI3K activity.

Kidney HEK 293 cells stably expressing TRβ1 (ZLTRβ) or TRβPV (ZLTRβPV) prepared in a manner similar to that described by Ying et al. (23) were grown in Dulbecco's modified Eagle's medium with 10% (vol/vol) fetal bovine serum supplemented with G418 (250 μg/ml). ZLTRβ or ZLTRβPV cells were transfected with pCMX NCoR (4 μg) by using lipofectamine 2000, following the manufacturer's protocols. Sixteen hours later, transfected cells were treated with or without insulin-like growth factor 1 (IGF-1; 100 nM) for 2 h. To determine the IGF-1 effects on PI3K activities caused by NCoR, levels of phosphorylated and total AKT were analyzed using Western blot analysis.

Fluorescence confocal microscopy.

Subcellular localization of endogenous NCoR and transfected Flag-tagged p85α (Flag-p85α) in primary thyrocytes of wild-type mice and thyroid tumor cells derived from TRβ^PV/PV mice was evaluated by using fluorescence confocal microscopy as described previously (5). The primary antibodies were anti-Flag M2 antibodies (F3165; 0.5 μg/ml; Sigma) for Flag-p85α and anti-NCoR antibody for NCoR (SC-8994; 1:200 dilution; Santa Cruz). Nuclei were also stained with DAPI (4′,6′-diamidino-2-phenylindole; Vector Laboratories). Laser confocal scanning images were captured by using an Ultraview (Perkin Elmer) confocal head on a Zeiss TV200 inverted microscope.

Statistical analysis.

All statistical analyses were carried out using StatView 5.0 (SAS Institute, Inc.) as described previously (9). Statistical analysis was performed with the use of analysis of variance (ANOVA), and P values of <0.05 were considered significant. All data are expressed as means ± standard deviations (SD).

RESULTS

NCoR physically interacts with PI3K.

To evaluate whether NCoR could regulate the association of p85α with TRβ1 or PV, we first used coimmunoprecipitation assays to examine whether NCoR could interact directly with p85α in cells. Lane 3 in Fig. 1Aa shows that endogenous NCoR was associated with transfected Flag-tagged p85α (Flag-p85α) when anti-Flag antibody was used in the immunoprecipitation followed by Western blot analysis using anti-NCoR antibody. In cells without transfected p85α, no NCoR was detected (Fig. 1Aa, lane 2). Lane 1 represents the negative control in which an irrelevant monoclonal antibody, MOPC, was used in the immunoprecipitation step. Direct Western blot analysis (Fig. 1Aa, lanes 4 and 5) showed that the NCoR protein abundance remained the same with or without the transfected p85α in CV1 cells. The association of NCoR with p85α was further confirmed by immunoprecipitation with anti-Flag antibody for detection of transfected Flag-tagged NCoR, followed by Western blot analysis for detection of associated endogenous p85α (Fig. 1Ab, lane 3). Lanes 1 and 2 of Fig. 1Ab show the negative controls.

FIG. 1.

Physical interaction of NCoR with p85α in cells. (A) Coimmunoprecipitation of NCoR and p85α in cells. (Aa) CV-1 cells were transfected with the expression plasmid of Flag-p85α (8 μg) (lanes 3 and 5), and cellular lysates were prepared for coimmunoprecipitation. Cell lysates were immunoprecipitated (IP) with monoclonal anti-Flag antibody, M2 (5 μg) (lanes 2 and 3), or an irrelevant monoclonal antibody (MOPC) as a control (lane 1), followed by Western blot analysis with anti-NCoR antibody as described in Material and Methods. Inputs (5%) are shown in lanes 4 and 5. (Ab) Association of endogenous p85α with transfected NCoR. CV1 cells were transfected with the expression plasmid of Flag-NCoR (pCMX-NCoR; 8 μg) (lanes 3 and 5). For coimmunoprecipitation, cell lysates were immunoprecipitated with monoclonal anti-Flag antibody (5 μg of M2) (lanes 2 and 3) or an irrelevant monoclonal antibody (MOPC) as a control (lane 1), followed by Western blot analysis with anti-p85α antibody. Inputs are shown in lanes 4 and 5. (Ac) Association of endogenous p85α with endogenous NCoR. Cellular lysates (1.2 mg) were immunoprecipitated with anti-p85α antibodies (lanes 3, 4, 8, and 9) or with a control antibody, MOPC (lanes 5 and 10), followed by Western blot analysis using anti-NCoR antibody (HPQQ; 2 μg/ml). Lanes 1, 2, 6, and 7 represent positive controls for direct Western blot analysis. CV1 cells (lanes 1 through 5) and HeLa cells (lanes 6 through 10) were treated with or without T3 (100 nM) (lanes are marked) as described in Materials and Methods. (Ad) Association of endogenous p110α with endogenous p85α but not with endogenous NCoR. Cellular lysates (1.2 mg) were immunoprecipitated with anti-p110α antibody (5 μg) (lanes 1 and 4), anti-p85α (5 μg) (lanes 2 and 5), or control rabbit immunoglobulin G (IgG; 5 μg) (lanes 3 and 6), followed by Western blot analysis using anti-NCoR antibody (HPQQ; 2 μg/ml) (top), anti-p110α antibody (middle), or anti-p85α (lower). Lanes 7 and 8 represent positive controls for direct Western blot analysis. Cells were treated with or without T3 (100 nM) as marked. (Ae) HDAC3 is not associated with endogenous p85α/NCoR complexes. Cellular lysates (1.2 mg) were immunoprecipitated with anti-p85α antibody (lanes 3, 4, 8, and 9) or a control antibody, MOPC, followed by Western blot analysis using anti-HDAC3 antibody (1:1,000 dilution). Lanes 1, 2, 6, and 7 represent positive controls for direct Western blot analysis. CV1 cells (lanes 1 through 5) and HeLa cells (lanes 6 through 10) were treated with or without T3 (100 nM) (lanes are marked) as described in Materials and Methods. (B) Binding of p85α to NCoR determined by GST pulldown assays. (Ba) Equal amounts of GST-p85α fusion proteins (full-length and truncated domains as marked) were each incubated with 10 μl of ³⁵S-labeled TRβ1 (lanes 1 to 5) synthesized by in vitro transcription/translation as described in Materials and Methods (lane 6 shows the 5% ³⁵S-labeled TRβ1 input). Coomassie blue staining of the SDS-PAGE gel shows that similar amounts of GST-fused proteins (lanes 1 through 4) and GST proteins (lane 5) were used (lower). After drying, the gel was autoradiographed (upper). (Bb) GST-p85α fusion proteins were incubated with 10 μl of ³⁵S-labeled domains of NCoR proteins as indicated (lanes 1 to 5) and prepared by in vitro transcription/translation. Lanes 6 to 10 represent the 5% input of the protein lysates used in the GST pulldown assay. (Upper) autoradiography; (lower) Coomassie blue staining. (Bc) Schematic representation of NCoR and its domains. (Bd) GST-p85α fusion proteins (full-length and truncated domains) and GST proteins were each incubated with 10 μl of ³⁵S-labeled NCoR RID (lanes 1 to 5) synthesized by in vitro transcription/translation (lane 6 shows the 5% ³⁵S-labeled NCoR RID input). (Upper) autoradiography; (lower) Coomassie blue staining. (Be) Schematic representation of p85α and its domains. (C) Localization of endogenous NCoR with transfected Flag-p85α in primary thyroid cultured cells of wild-type mice (Ca to Cd) and thyroid tumor cells of TRβ^PV/PV mice (Ce to Ch) were visualized by confocal fluorescence microscopy. Flag-p85α (Ca and Ce) (green) and NCoR (Cb and Cf) (red) were stained with anti-Flag (M2) and anti-NCoR (SC-8994; 1:200 dilution; Santa Cruz) antibodies followed by secondary antibodies for visualization. Panel Ca shows that p85α was localized in the nucleus (arrow) as well as in the cytoplasmic (arrowhead) compartments. Panel Cb shows that NCoR was mainly localized in the nucleus (arrow), but cytoplasmic distribution was also observed (arrowhead). Panels Cc and Cg show the merged images of p85α and NCoR for wild-type and tumor cells, respectively. Panels Cd and Ch show nuclear staining with DAPI.

The interaction of endogenous p85α with endogenous NCoR was further demonstrated by similar coimmunoprecipitation assays. As shown in Fig. 1Ac, upon immunoprecipitation of cellular lysates of CV1 cells (Fig. 1Ac, lanes 3 and 4) or of HeLa cells (lanes 8 and 9) by anti-p85α antibody, followed by Western blot analysis with anti-NCoR antibodies, endogenous NCoR was found to complex with endogenous p85α. This interaction was not affected by T3 (compare the band intensity of lane 3 to that of lane 4 and that of lane 8 to that of lane 9). Therefore, endogenous p85α interacted with endogenous NCoR and overexpression of either protein is not required.

To determine whether NCoR also bound to the catalytic subunit of PI3K, p110α, we carried out a coimmunoprecipitation assay by using antibody to p110α in the immunoprecipitation step, followed by Western blot analysis using anti-NCoR (Fig. 1Ad, top, lanes 1 and 4) or anti-p85α (bottom, lanes 1 and 4) antibodies. Concurrently, cell lysates were also immunoprecipitated by anti-p85α antibody, followed by Western blot analysis using anti-NCoR (top, lanes 2 and 5) or anti-p110α (middle, lanes 2 and 5) antibodies as positive controls. Although p110α-bound p85α (the interaction is shown in the middle panel) interacted with NCoR (top, lanes 2 and 5), p85α-bound p110α did not interact with NCoR (top, lanes 1 and 4), whether T3 was present or not. Lanes 7 and 8 of Fig. 1Ad represent the positive controls for direct Western blot analysis of cellular lysates without the immunoprecipitation step. These results indicate that p110α did not physically interact with NCoR.

To further evaluate whether p85α/NCoR complexes also recruited other known NCoR-associated proteins, such as HDAC3, lysates of CV1 and HeLa cells were similarly immunoprecipitated with anti-p85α antibodies, followed by Western blot analysis using anti-HDAC3 antibodies. As shown in Fig. 1Ae, whereas direct Western blot analysis showed the presence of HDAC3 (lanes 1, 2, 6, and 7), no HDAC3 was found to associate with p85α/NCoR complexes.

Previously, we showed that p85α binds to TRβ1 or PV via the ligand-binding domain of TRβ1 or PV (5). In this study, using GST pulldown assays, we mapped the interaction region of p85α and TRβ1 to the C-terminal SH2 (Src homology 2) domain (CSH2) (Fig. 1Ba, lane 4; see also panel Be) but not in the SH3 domain (Fig. 1Ba, lane 2) or in the N-terminal SH2 domain (NSH2) (Fig. 1Ba, lane 3). Lane 5 of Fig. 1Ba represents the negative control, and lane 6 shows the input (Fig. 1Ba). The lower panel of Fig. 1Ba shows that identical amounts of GST-fused p85α and its domains were used in the pulldown assays determined by Coomassie blue staining.

The interaction regions in NCoR and p85α were also mapped by GST pulldown assays (Fig. 1Bb). The amino-terminal R1 (Fig. 1Bb, lane 1) and the C-terminal R4 and RID (lanes 4 and 5, respectively) were the regions of NCoR that interacted with p85α. Lanes 6 to 10 of Fig. 1Bb show the input for the respective domains of NCoR (see also Fig. 1Bc). Again, the lower panel of Fig. 1Bb shows that identical amounts of GST-p85α were used in the GST pulldown assays for each reaction.

The region of p85α that interacted with NCoR (RID) is shown (in Fig. 1Bd) for GST-full-length p85α (lane 1), SH3 (lane 2), NSH2 (lane 3), and CSH2 (lane 4) (see also Fig. 1Be). Full-length p85α (Fig. 1Bd, upper, lane 1) and CSH2 were able to bind to RID of NCoR (Fig. 1Bd, upper, lane 4) but not SH3 (Fig. 1Bd, upper, lane 2) or NSH2 (Fig. 1Bd, lane 3). The lower panel of Fig. 1Bd shows the Coomassie blue-stained GST-p85α and the truncated domains, indicating that identical GST proteins were used in the GST pulldown assays for each reaction. Thus, these data indicate that NCoR and TRβ1 interacted with the same region of p85α at the CSH2 domain.

Using fluorescence confocal microscopy, we further identified the subcellular sites at which the interaction of NCoR with p85α might occur. Figure 1Ca shows that p85α was localized in the nucleus as well as in the cytoplasmic compartments. Figure 1Cb shows that NCoR was mainly localized in the nucleus, but cytoplasmic distribution was also observed. The images from Fig. 1Ca and Cb are merged in Fig. 1Cc, indicating that p85α and NCoR were localized in the cytoplasmic as well as the nuclear compartments. DAPI staining is shown in Fig. 1Cd. In tumor cells of TRβ^PV/PV mice, similar subcellular patterns were observed for p85α (Fig. 1Ce) and NCoR (Fig. 1Cf). However, the fluorescence intensity of NCoR was apparently reduced (Fig. 1C, compare panel Cb with panel Cf and panel Cc with panel Cg). The merged image shown in Fig. 1Cg indicates that NCoR and PV were localized in the same compartments.

NCoR competes with TRβ1 or PV for binding to p85α.

That NCoR and TRβ1 bound to the same region of p85α via the CSH2 domain (Fig. 1B) prompted us to ascertain whether NCoR competed with TRβ1 or PV for binding to p85α. Figure 2A shows that, indeed, in the presence of increasing concentrations of NCoR (RID), binding of TRβ1 to p85α was decreased in a concentration-dependent manner (Fig. 2A, lanes 2 to 4 compared with lane 1). NCoR also competed with PV for association with p85α in a concentration-dependent manner (Fig. 2B). We further determined relative affinities in the binding of TRβ1, PV, and NCoR to p85α by using a constant amount of GST-p85α (shown in the lower panels of Fig. 2C, D, and E; determined by Coomassie staining) and increasing amounts of ³⁵S-labeled TRβ1, PV, or NCoR (RID). For the comparison of relative affinities in the binding, identical concentrations for TRβ1, PV, and NCoR (RID) at each increment were used. The band intensities were quantified and plotted in Fig. 2F, indicating that the rank order for binding to p85α was PV > TRβ1 > NCoR.

FIG. 2.

NCoR (RID) competes with TRβ or PV for binding to p85α. GST-p85α fusion protein was incubated with 2 μl of ³⁵S-labeled TRβ1 (A) or ³⁵S-labeled PV (B) in the absence (lane 1) or presence (lanes 2 to 4, respectively) of 1, 2, or 4 μl of NCoR (RID). (C to F) GST-p85α fusion proteins were incubated with increasing quantities of ³⁵S-labeled TRβ1 (2.5, 5, 10, and 20 μl) (C), PV (3.2, 6.4, 12.8, and 25.6 μl) (D), or NCoR (RID) (1.7, 3.4, 6.8, and 13.6 μl) (E) (lanes 1 to 4). Similar amounts of ³⁵S-labeled TRβ1, PV, and NCoR (RID) were used after normalization by the number of methionines in each protein as described in Materials and Methods. Band intensities were quantified by using NIH IMAGE software (ImageJ 1.34s; Wayne Rashband, NIH) (http://rsb.info.nih.gov/ij), and the relative binding affinities of TRβ1, PV, and NCoR (RID) to p85α are shown in panel F.

Reduced expression and distribution of NCoR in thyroid tumors of TRβ^PV/PV mice.

We have previously shown that PV binds to p85α more strongly than does TRβ1 and that this interaction leads to the activation of the PI3K signaling cascade in thyroid tumors of TRβ^PV/PV mice (5). The above-described findings that NCoR could compete with TRβ1 or PV for binding to p85α raised the possibility that the protein abundances of NCoR in the thyroids of wild-type and TRβ^PV/PV mice could differ. We therefore compared the protein abundances of NCoR in the thyroids of wild-type and TRβ^PV/PV mice by Western blot analysis. Indeed, NCoR protein levels in thyroid tumors of TRβ^PV/PV mice (Fig. 3Aa, lanes 4 to 8) (n = 5) were 2.6-fold lower than those in wild-type mice (lanes 1 to 3) (n = 3). Similar reductions in NCoR protein abundance were observed in the primary thyroid tumor cultured cells isolated from TRβ^PV/PV mice (Fig. 3Bb, compare lanes 3 and 4 to lanes 1 and 2). These findings suggest that a reduced abundance of cellular NCoR would favor the interaction of p85α with PV.

FIG. 3.

Decreased NCoR protein abundance in thyroid tumors (A) and nuclear and cytosolic distribution of NCoR by Western blot analysis (B). (Aa) Decreased abundance of NCoR proteins in the thyroids of TR^PV/PV mice (lanes 4 to 8) (n = 5) compared with levels for wild-type mice (lanes 1 to 3, upper) (n = 3). The loading controls using α-tubulin are shown in the lower panel. (Ab) Reduced abundances of NCoR proteins (55% reduction) in cultured tumor cells of TRβ^PV/PV mice (lanes 3 and 4) compared with those in cultured primary thyrocytes of wild-type mice (lanes 1 and 2). The loading controls using α-tubulin are shown in the lower panel. The band intensities were quantified with NIH IMAGE software (ImageJ 1.34s; Wayne Rashband, NIH) (http://rsb.info.nih.gov/ij), and statistical analysis was performed as described in Materials and Methods. The P values are indicated. (B) Distribution of NCoR in the cytosolic and nuclear compartments of wild-type (WT) primary thyrocytes (lanes 1 and 2) and tumor cells of TRβ^PV/PV mice (lanes 3 and 4). Cellular extracts were separated into nuclear and cytosolic fractions as described in Materials and Methods. Western blot analysis was carried out using 30 μg of total protein and anti-NCoR antibody. PARP and α-tubulin were used as markers for nuclear and cytosolic fractions.

Figure 1C shows that NCoR was also localized in the cytoplasm of thyrocytes and tumor cells. To further confirm the distribution of NCoR in the extranuclear compartment, we fractionated primary thyroid cells into nuclear (Fig. 3B, lanes 1 and 3) and cytosolic (Fig. 3B, lanes 2 and 4) fractions and showed that the fractions were not contaminated with each other by using PARP as a nuclear marker and α-tubulin as a cytosolic marker (Fig. 3B, lower two panels). Figure 3B shows that NCoR was mainly localized in the nuclear fractions of wild-type thyrocytes (Fig. 3B, compare lane 1 to lane 2). By contrast, in tumor cells, NCoR was predominantly localized in the cytosol (compare lane 3 to lane 4). These results support the observations by fluorescence confocal microscopy that in addition to nuclear distribution, NCoR was also localized in extranuclear compartments of wild-type thyrocytes and tumor cells. However, in PV-expressing cells, cytoplasmic distribution was higher in tumor cells than in wild-type thyrocytes.

That the expression of NCoR was lower in thyroid tumor cells of TRβ^PV/PV mice suggests that the expression of NCoR could be positively regulated by T3/TRβ and that the mutation of TRβ led to the loss of this up-regulation. Previously, we have shown that TRβ is the major TR isoform in the thyroid. To test this hypothesis, we treated the primary thyrocytes isolated from wild-type mice and tumor cells from TRβ^PV/PV mice with or without T3. Figure 4 shows that in the wild-type thyrocytes, T3 activated the expression of mRNA (compare bar 2 to bar 1). In tumor cells, the expression of NCoR mRNA was significantly lower (compare bar 4 to bar 1 and bar 5 to bar 2). Furthermore, there was no T3 activation in the presence of T3 in tumor cells (compare bar 5 to bar 4). These results indicate that mutation of TRβ led to the obliteration of T3 induction. To ascertain whether the T3-activated expression of NCoR mRNA was a direct or indirect effect, we treated the cells with cycloheximide. As shown in bar 3, cycloheximide treatment led to the loss of T3-activated expression (compare bar 3 with bar 2). These data demonstrate that the T3-induced increase in NCoR mRNA is an indirect effect. No significant differences caused by cycloheximide treatment were observed in tumor cells, as the T3-induced effect of NCoR mRNA was lost.

FIG. 4.

T3 regulates the expression of NCoR mRNA indirectly. Primary thyroid cells isolated from wild-type mice and tumor cells isolated from TRβ^PV/PV mice were preincubated with T3-depleted medium for 24 h and then treated with T3 (100 nM) in the absence or presence of cycloheximide (100 μg/ml) for 24 h. NCoR mRNA was determined by quantitative real-time RT-PCR using 100 ng of mRNA (n = 6). Relative quantification of target mRNA was determined by arbitrarily setting the control value for wild-type primary thyroid cells to 1. All data are expressed as means ± SD (n = 5). Statistical differences were significant, as indicated by P values from the analysis by ANOVA.

NCoR negatively regulates PI3K downstream signaling.

The above-described in vivo findings predicted that a decrease in the cellular NCoR would lead to an activation of PI3K activity and its downstream signaling. We therefore used siRNA approaches to ascertain the effect of knocking down the expression of NCoR on the PI3K activities of primary thyrocytes of wild-type mice and tumor cells of TRβ^PV/PV mice. Figure 5Aa shows that NCoR expression was knocked down after treatment of primary thyrocytes (compare lane 2 with lane 1) and tumor cells (compare lane 4 with lane 3) with specific siRNA as determined by Western blot analysis. The reduced NCoR protein abundance led to significantly increased PI3K kinase activities in primary thyrocytes (Fig. 5Ab, compare bar 2 with bar 1) (2.5-fold increase) as well as in tumor cells (Fig. 5Ab, compare bar 4 with bar 3) (1.4-fold increase). Consistent with the increased kinase activity of PI3K caused by reduced NCoR protein levels, the phosphorylation of its immediate downstream effector, AKT (pAKT), was increased in primary thyrocytes (Fig. 5Ac, upper, compare lane 3 with lane 1) and tumor cells (Fig. 5Ac, upper, compare lane 4 with lane 2). The increased p-AKT level was not due to an increase in total AKT, because total AKT virtually was not altered by treatment of cells with siRNA (Fig. 5Bc, middle). The lower panel of Fig. 5Bc shows the loading controls using α-tubulin. Taken together, these results indicate that the reduced NCoR protein levels would favor the interaction of TRβ1 or PV with p85α to activate PI3K/AKT signaling.

FIG. 5.

Effects of siRNA knockdown (A) or overexpression (B and C) of cellular NCoR on PI3K downstream signaling. (Aa) NCoR expression was knocked down in primary thyrocytes of wild-type mice (lane 2) and tumor cells of TRβ^PV/PV mice (lane 4) by siNCoR compared with what was found for controls (lanes 1 and 3, respectively). Cellular extracts were analyzed by Western blot analysis for NCoR (Aa, upper) and α-tubulin for loading controls (Aa, lower). In panel Ab, the PI3K kinase activities of primary thyrocytes of wild-type mice (bars 1 and 2) and tumor cells of TRβ^PV/PV mice (bars 3 and 4) treated with siNCoR or with control RNA were determined as described in Materials and Methods. The differences were statistically significant as analyzed by ANOVA (n = 6). The P values are marked. (Ac) Western blot analysis was carried out for determination of p-AKT (anti-phosphorylated S473 antibody) (upper), total AKT (middle), and α-tubulin as a control for protein loading (lower). WT, wild type. (B) Wild-type primary thyroid cells or PV tumor cells were transfected with NCoR expression vector (pCMX-NCoR; 8 μg). Western blot analysis was carried out to determine the effects of NCoR on the activity of p-AKT. Protein abundance of cellular NCoR (Ba), p-AKT (anti-phosphorylated S473 antibody) (Bb), total AKT (Bc), and α-tubulin as a control for protein loading (Bd). (C) Activation of p-AKT in kidney 293 cells stably expressing TRβ1 or PV in the presence (lanes 2, 4, 6, and 8) or absence (lanes 1, 3, 5, and 7) of IGF-1 (100 nM) with or without overexpression of NCoR as described in panel B. Western blot analysis was carried out as described in Materials and Methods to determine the abundance of p-AKT (top). No changes were observed in total AKT (middle). The loading control using α-tubulin is shown in the lower panel.

To provide additional evidence to support the conclusion that NCoR competed with TRβ1 or PV for binding to p85α, we overexpressed NCoR by transfection into primary thyrocytes of wild-type mice and tumor cells of TRβ^PV/PV mice and determined the effect of its overexpression on PI3K signaling (Fig. 5B). As shown by Western blot analysis, NCoR was overexpressed in the transfected primary thyrocytes of wild-type mice (Fig. 5Ba, compare lane 3 to lane 1) and tumor cells of TRβ^PV/PV mice (Fig. 5Ba, compare lane 4 to lane 2). In the absence of overexpressed NCoR, consistent with previous findings (5), more p-AKT was observed in tumor cells of TRβ^PV/PV mice than in primary thyrocytes of wild-type mice (Fig. 5Bb, compare lane 2 to lane 1). When NCoR was overexpressed, a concurrent reduced p-AKT level was found in tumor cells as well as in thyrocytes of wild-type mice (Fig. 5Bb, compare lane 4 to lane 2 and lane 3 to lane 1). The observation that total AKT protein levels were not altered by NCoR overexpression (Fig. 5Bc; the loading controls are shown in panel Bd) further supports the conclusion that NCoR regulated the interaction of p85α with PV or TRβ1, thus modulating the PI3K downstream signaling.

That overexpressed NCoR could reduce PI3K/AKT activity was also shown in kidney HEK 293 cells stably expressing TRβ1 or PV (Fig. 5C). Lane 3 of Fig. 5C shows that in PV-expressing cells, PI3K/AKT was activated (p-AKT) compared with what was observed in cells expressing TRβ1 (lane 1) and that this activation was attenuated when NCoR was overexpressed by transfection (top, compare p-AKT in lane 7 to p-AKT in lane 3). To ascertain whether NCoR-regulated PI3K/AKT activity could be further modulated by upstream signals, we treated the cells with IGF-1. Indeed, treatment of IGF-1 led to further activation of p-AKT (top, compare lane 2 to lane 1 and lane 4 to lane 3). The overexpression of NCoR attenuated IGF-1-induced activation of PI3K/AKT, as shown by the reduction of p-AKT (top, compare lane 6 to lane 2 and lane 8 to lane 4). The changes in p-AKT were not due to the alteration in the total AKT levels, as shown in the middle panel of Fig. 5C. The lower panel shows the corresponding loading controls. These findings suggest that the interaction of p85α with NCoR was weakened by treatment of cells with IGF-1, thus facilitating the interaction of p85α with TRβ1 or PV.

To ascertain how NCoR could modulate thyroid carcinogenesis via regulation of PI3K/AKT signaling, we further evaluated the effect of NCoR knockdown on the activity of one of the PI3K/AKT downstream effectors, MMP2, known to be critically involved in the degradation of the extracellular matrix (3, 17, 19) and in cell motility. Previously, we showed that the activation of AKT and MMP2 is accompanied by an increase in tumor cell motility in TRβ^PV/PV mice (5, 10). Figure 6A shows that when NCoR was knocked down (Fig. 6A, upper, compare lane 4 with lane 3), MMP2 cellular abundance was further increased in thyrocytes of wild-type mice (Fig. 6A, middle, compare lane 2 with lane 1) (∼2.5-fold) and tumor cells (Fig. 6A, middle, compare lane 4 to lane 3) (1.5-fold). These increases were accompanied by increases in cell motility (Fig. 6B, compare bar 4 with bar 3 for tumor cells and bar 2 with bar 1 for wild-type thyrocytes). These results indicate that NCoR via competition with TRβ1 or PV for interaction with p85α could contribute to progression of thyroid cancer.

FIG. 6.

NCoR modulates thyroid cell motility via the PI3K/AKT/MMP2 pathways. (A) NCoR expression was knocked down in primary thyrocytes of wild-type mice (upper, lane 2) and tumor cells of TRβ^PV/PV mice (upper, bar 4) by siNCoR compared with the siRNA of controls (upper, lane 1 and 3, respectively). Cellular extracts were analyzed by Western blot analysis for NCoR (upper), MMP2 (middle), and α-tubulin for loading controls. (B) Primary thyrocytes of wild-type mice (bar 2) and tumor cells of TRβ^PV/PV mice (bar 4) were treated with siNCoR or with control siRNA (bar 1 or 3, respectively) and were assayed for cell motility as described in Materials and Methods. Data are expressed as means ± SD (n = 6).

We have previously shown that the activation of PI3K/AKT increases thyroid cell growth (5). We therefore further determined the effect of NCoR knockdown on thyrocyte proliferation. As shown in Fig. 7A, knockdown of NCoR increased proliferation of thyrocytes isolated from wild-type mice. In tumor cells (Fig. 7B), additional reduction of cellular NCoR by knockdown further increased cell proliferation, although the extent of the increases was smaller than that in wild-type thyrocytes because NCoR expression was lower in the tumor cells (Fig. 4). Taken together, these results indicate that NCoR via competition with TRβ1 or PV for interaction with p85α could contribute to thyroid carcinogenesis through increasing cell proliferation.

FIG. 7.

Inhibition of NCoR expression increases thyrocyte proliferation. Thyrocytes isolated from wild-type mice (A) and tumor cells from TRβ^PV/PV mice were transfected with siRNA (open circles, control using irrelevant RNA; solid circles, siRNA for NCoR) as described in Materials and Methods. Cell numbers were counted daily using a Coulter cell counter. All data are expressed as means ± SD (n = 3).

DISCUSSION

Since the discovery of NCoR as a corepressor of TR transcription (7), studies on the functions of NCoR have mainly focused on its role in transcription regulation. Nuclear NCoR associates with HDACs and other proteins to form large complexes to repress the transcription activity of unliganded or antagonist-bound nuclear receptors (15). Recent reports of its cytoplasm-nucleus redistribution further expand its regulatory role as a transcription repressor. Thus, the interleukin-1β-induced nuclear export of the NCoR/TAB2/HDAC3 complex to the cytoplasm results in the derepression of the tetraspanin KAI1/CD82 target (2). In line with this observation, upon stimulation by NRG1 and presenilin, cytoplasmic NCoR forms complexes with TAB2 and the cleavage product of ErbB4, E41CD, which is translocated to the nucleus to repress the expression of astrocytic genes (16). The present study, however, demonstrates a novel role for NCoR independent of transcription regulation. We found that NCoR physically interacted with PI3K via direct binding to the regulatory subunit p85α but not the catalytic subunit (p110α). This binding competitively inhibited the association of PV or TRβ with p85α. Lowering cellular NCoR by siRNA knockdown in primary thyrocytes of wild-type mice and thyroid tumor cells of TRβ^PV/PV mice led to an activation of the PI3K downstream effector p-AKT (Fig. 5A). Conversely, overexpression of NCoR by transfection resulted in decreased activation of PI3K, as indicated by reduced p-AKT in primary thyrocytes and thyroid tumor cells (Fig. 5B). These results indicate that NCoR is a novel regulator of PI3K signaling.

The importance of this novel regulatory role of NCoR is evident in that the motility and proliferation of tumor cells were significantly increased when the expression of NCoR was repressed by siRNA knockdown (Fig. 6B and 7). Previous studies of human thyroid cancer specimens by several groups have shown AKT overexpression and overactivation in primary thyroid cancers (12, 20). Consistent with these observations, activated AKT was also demonstrated during thyroid carcinogenesis of TRβ^PV/PV mice (10). Subsequently, it was found that PV interacts strongly with p85α to increase PI3K activity to activate AKT, resulting in increased tumor cell motility (5, 10). Taken together, our data support the model shown in Fig. 8, indicating that PV and NCoR compete for binding to the CSH2 domain of p85α. In tumor cells, the cellular levels of NCoR are low, thereby facilitating the binding of PV to p85α to activate PI3K/AKT signaling (5). One of the activated PI3K/AKT downstream pathways, MMP2, is activated, resulting in increasing cell motility (5, 10). In this context, NCoR could be considered a newly identified tumor suppressor in a mouse model of thyroid cancer.

FIG. 8.

A proposed molecular model for the regulatory role of NCoR in PI3K signaling. NCoR and PV compete for binding to the CSH2 domain of p85α. In thyroid tumors, NCoR protein abundance is low, thereby facilitating the association of PV with p85α to activate PI3K/AKT signaling, contributing to thyroid carcinogenesis in TRβ^PV/PV mice. The major domains of p85α are indicated. PRD, proline-rich domain.

The present study shows that NCoR was a T3-positively regulated gene in the thyroid, as treatment of primary thyrocytes isolated from wild-type mice increased the NCoR mRNA (Fig. 4). This T3 activation was an indirect effect, however, as in the presence of cycloheximide, the T3 induction was lost. Furthermore, in tumor cells isolated from TRβ^PV/PV mice, the expression of NCoR mRNA was markedly decreased, suggesting that mutation of TRβ resulted in the loss of this activation. The expression of NCoR was recently shown to be reduced by 90% in the cerebella of TRβ^−/− mice compared with what was found for wild-type mice, whereas no significant changes were detected in TRα^0/0 mice (14, 21). These results suggest that TRβ could play a role in regulating the expression of NCoR. That notion is consistent with the current findings that PV could interfere with the transcriptional activity of wild-type TRs to repress the expression of NCoR in the thyroids of TRβ^PV/PV mice, as has been shown in the repression of other T3-positively regulated genes, such as the malic enzyme, S14, and deiodinase 1 genes in the liver and the growth hormone gene in the pituitary, by PV (8). However, the verification of this hypothesis awaits future studies.