Thyroid Hormone Receptor-β (TRβ) Mediates Runt-Related Transcription Factor 2 (Runx2) Expression in Thyroid Cancer Cells: A Novel Signaling Pathway in Thyroid Cancer

Abstract

Dysregulation of the thyroid hormone receptor (TR)β is common in human cancers. Restoration of functional TRβ delays tumor progression in models of thyroid and breast cancers implicating TRβ as a tumor suppressor. Conversely, aberrant expression of the runt-related transcription factor 2 (Runx2) is established in the progression and metastasis of thyroid, breast, and other cancers. Silencing of Runx2 diminishes tumor invasive characteristics. With TRβ as a tumor suppressor and Runx2 as a tumor promoter, a compelling question is whether there is a functional relationship between these regulatory factors in thyroid tumorigenesis. Here, we demonstrated that these proteins are reciprocally expressed in normal and malignant thyroid cells; TRβ is high in normal cells, and Runx2 is high in malignant cells. T₃ induced a time- and concentration-dependent decrease in Runx2 expression. Silencing of TRβ by small interfering RNA knockdown resulted in a corresponding increase in Runx2 and Runx2-regulated genes, indicating that TRβ levels directly impact Runx2 expression and associated epithelial to mesenchymal transition molecules. TRβ specifically bound to 3 putative thyroid hormone-response element motifs within the Runx2-P1 promoter (⁻105/⁺133) as detected by EMSA and chromatin immunoprecipitation. TRβ suppressed Runx2 transcriptional activities, thus confirming TRβ regulation of Runx2 at functional thyroid hormone-response elements. Significantly, these findings indicate that a ratio of the tumor-suppressor TRβ and tumor-promoting Runx2 may reflect tumor aggression and serve as biomarkers in biopsy tissues. The discovery of this TRβ-Runx2 signaling supports the emerging role of TRβ as a tumor suppressor and reveals a novel pathway for intervention.

Thyroid cancer, the most common endocrine malignancy, includes differentiated papillary thyroid cancer (PTC), follicular thyroid cancer (FTC), as well as the most invasive and lethal undifferentiated or anaplastic thyroid cancer (ATC). These nonmedullary cancers, derived from follicular epithelial cells, reflect a spectrum of cumulative cellular dysregulation, distinct molecular profiles, altered nuclear architecture, and epigenetic dysregulation that correlates with dedifferentiation, tumor aggression, and metastases (1). Distant metastases are infrequent in well-differentiated thyroid cancer; however, when present, approximately 50% are in lung and 25% in bone (2–5). Distinct mutational events (6–8) and disrupted intracellular signaling characterize thyroid tumorigenesis (9, 10), yet regulatory genes in the epithelial to mesenchymal transition (EMT) and tumor invasion are not well delineated. Moreover, there are limited detection and treatment options for aggressive microcarcinomas, undifferentiated thyroid cancers as well as recurrent disease. Thus, there is a critical need for therapies targeted at the underlying tumor biology.

Transcription factors that mediate developmental processes may be central to thyroid tumorigenesis. Runt-related transcription factors 1–3 (Runx1–Runx3) are key regulators in mesenchymal cell differentiation in osteogenic lineage and of proliferation and differentiation during embryonic development and tissue remodeling (11). In particular, Runx2 is a key regulator of tumorigenesis and functionally linked with tumor progression in breast (12–14), ovarian epithelial (15), and prostate cancers (16). Activation or increased expression of Runx2, noted particularly in metastatic tumors targeting bone (17–20), induces an array of genes associated with cell adhesion, invasion, survival as well as increased expression of markers of metastases (21, 22). Inhibition of Runx2 expression is associated with either a reversal or a less aggressive phenotype (15, 18, 19, 22). Runx2 promotes cancer progression and invasion in part through up-regulation of matrix metalloproteinases (MMPs), angiogenic factors as well as EMT-related proteins, whereas inhibition of Runx2 silences or dampens these factors and associated invasive characteristics (23, 24). Similar findings are recently noted in thyroid cancer (23–28). The Runx2 gene is characterized by 2 distinct promoters, P1 and P2, that give rise to distinct isoforms (29, 30). Tissue-specific expression of these isoforms is well characterized in osteogenesis yet the specific roles in tumorigenesis have not been established. Elevated Runx2-P2 transcript levels in PTC cells compared with nonmalignant thyroid cells have been reported; an effect not observed with Runx2-P1 transcripts (24). Thus, the Runx2 isoforms may have distinct roles in tumorigenesis.

Like Runx2, the thyroid hormone receptors (TRs), liganded and unliganded, are extensively characterized key regulators of growth, differentiation, and development. Functional loss or modifications of these transcription factors are increasingly recognized in human cancers. Dysregulation of TRs through point and frameshift mutations as well as epigenetic silencing is established in a variety of solid tumors (31–40). The propensity for reduced expression of TRs, notably TRβ1, as tumorigenesis progresses is consonant with a suppressor role for TR (41–43). Reexpression of TRβ in breast cancer cell lines (44) or thyroid tumors (45, 46) decreases tumor growth, invasion, and metastases.

Our microarray analyses of human thyroid cell lines revealed differential expression of Runx2 and TRβ in human normal and PTC, FTC, and ATC cells (47). Based on these findings, an emergent question is whether there is a molecular signaling pathway linking TRβ and Runx2 expression revealed in thyroid tumorigenesis. In the present study, we first demonstrated reciprocal expression of TRβ and Runx2 in progressive stages of thyroid cancer. We then found that knockdown of TRβ resulted in an increase in Runx2 expression and corresponding tumorigenic genes. Finally, we have shown that TRβ directly modulated Runx2 expression through binding to thyroid hormone-response element (TRE) motifs in the Runx2-promoter (P1). Thus, we can conclude that reduced function of TRβ contributes to the increase in Runx2 and associated thyroid cancer progression. Our results mechanistically reveal a novel signaling pathway in thyroid cancer and a possible mode of action for TRβ suppressor effects in an epithelial-derived cancer.

Materials and Methods

Cell culture

Thyroid cell lines corresponding to benign (THJ, ORI), papillary (LAM-1, BHP2-7), follicular (KAT-1, WRO), and anaplastic carcinomas (BHT101, KTC-2, FRO, SW1736) were cultured in RPMI 1640 growth media with L-glutamine (300 mg/L), sodium pyruvate and nonessential amino acids (1%) (Cellgro/Mediatech), supplemented with 10% fetal bovine serum (Life Technologies) and penicillin-streptomycin (200 IU/L) (Cellgro/Mediatech). For transfection experiments and hormone treatments, cells were maintained in 10% charcoal stripped fetal bovine serum (Life Technologies). Cells were maintained at 37°C, 5% CO₂, and 100% humidity. Thyroid tumor cell lines were generously provided by Dr John Copland III (Mayo Clinic). Normal thyroid cell lines (ORI) were obtained from European Collection of Cell Cultures (ECACC). All cell lines were authenticated (8/16/13) using short tandem repeat profiles and then Promega GenePrint10 System in the Advanced Genome Technologies Core at the University of Vermont.

Western blot analysis

Proteins were isolated from whole cells in lysis buffer (20mM Tris-HCl [pH 8], 137mM NaCl, 10% glycerol, 1% Triton X-100, and 2mM EDTA) containing Protease Inhibitor Cocktail 78410 (Thermo Scientific) agitated at 4°C for 30 minutes, and then sonicated for a total of 20 seconds (4 × 5 s). Lysates were centrifuged at 12 000 rpm at 4°C for 20 minutes and supernatants frozen at −80°C. Nuclear and cytoplasmic proteins were prepared with N-PER Nuclear and Cytoplasmic extraction reagents 78833 (Thermo Scientific) per manufacturer's protocol and frozen at −80°C. Proteins were resolved by polyacrylamide gel electrophoresis on 10% sodium dodecyl sulfate gels EC60752 (Life Technologies) and immobilized onto nitrocellulose membranes (GE Healthcare) by electroblot (Bio-Rad Laboratories). Specific proteins were detected by immunoblotting with the indicated antibodies (Table 1). Immunoreactive proteins were detected by enhanced chemiluminesence (GE Healthcare), visualized by VersaDoc MP3000 (Bio-Rad Laboratories), and intensities were quantified by Quantity One (Bio-Rad Laboratories).

Table 1.

Antibody Table

Peptide/Protein Target	Antigen Sequence (if Known)	Name of Antibody	Manufacturer, Catalog Number	Species Raised in; Monoclonal or Polyclonal	Dilution Used	Technique
TRβ-1	Human TRβ-1	TRβ-1 antibody (J52)	Thermo Scientific, MA1-216	Mouse; monoclonal	1:500	WB
TRβ	Synthetic peptide derived from human TRβ	Anti-TRβ antibody	Abcam, AB53170	Rabbit; polyclonal	1:500	WB
Runx2	Amino acids 294–363 of Runx2 of mouse origin	RUNX2 (M70) antibody	Santa Cruz Biotechnology, Inc, sc-10758	Rabbit; polyclonal	1:750	WB
β-Actin		Affinity-purified anti-β-actin	Rockland Immunochemicals, 600-401-886	Rabbit; monoclonal	1:20 000	WB
TRβ	GMSEACLHRKSHSERRSTLK	Anti-TRβ antibody	Abcam, AB15545	Rabbit; polyclonal	1:500	IF
Runx2	RUNX2 partial recombinant protein with GST tag	Anti-RUNX2 (251–350) mAb	Abnova, H00000860-M06	Mouse; monoclonal	1:500 (IF), 1:100 (IHC)	IF, IHC
TRβ	Synthetic peptide corresponding to amino acids 31–50 of human THRB	Anti-THRB pAb	Abnova, PAB11824	Rabbit; polyclonal	1:100	IHC
Runx1	Synthetic peptide corresponding to amino acids near the amino terminus of human AML1	AML1 antibody	Cell Signaling, CS4334	Rabbit; polyclonal	1:100	IHC
TRβ	Recombinant fragment corresponding to a region within amino acids 1 and 199 of human TRβ	TRβ antibody	Thermo Scientific, PA5-29684	Rabbit; polyclonal		ChIP
Runx2	KLH-conjugated synthetic peptide between 445–474 amino acids surrounding S533 of human RUNX2	RUNX2 antibody	Thermo Scientific, PA5-14-816	Rabbit; polyclonal		ChIP
Rabbit-γ globulin			Pierce, 31887	Rabbit; polyclonal		ChIP
TRβ	Purified fragment of human TRβ-1 corresponding to residues 201–456	TR antibody (C3)	Thermo Scientific, MA1-215	Mouse; monoclonal		EMSA
Goat antimouse HRP		Goat antimouse IgG, H&L chain specific, peroxidase conjugate	Calbiochem EMD, 401253	Goat; polyclonal	1:5000	WB
Mouse antirabbit HRP		Peroxidase-conjugated IgG fraction monoclonal mouse antirabbit IgG; light chain specific	Jackson ImmunoResearch, 211-032-171	Mouse; polyclonal	1:10 000	WB
Goat antirabbit		Goat antirabbit Alexa Fluor 555	Invitrogen/Life Technologies, A21429	Goat; polyclonal	1:500 (IF), 1:1000 (IHC)	IF, IHC
Goat antirabbit		Goat antirabbit Alexa Fluor 488	Invitrogen/Life Technologies, A11001	Goat; polyclonal	1:500 (IF), 1:1000 (IHC)	IF, IHC

Immunofluorescence

Proteins were detected in whole cells by immunofluorescence. Cells, seeded in 6-well dishes on coverslips coated with 0.5% gelatin (Sigma-Aldrich), were grown overnight in phenol red-free RPMI 1640 (Cellgro) at 37°C, 100% humidity, and 5% CO₂. The cells, packed on ice, were washed with PBS (Cellgro); fixed with 3.7% formalin in PBS for 10 minutes. After 2 washes in PBS, cells were permeabilized in 0.025% Triton X-100 (Sigma-Aldrich) for 20 minutes, then washed in PBS with 3% BSA (PBSA) (Sigma-Aldrich). Cells were blocked with 10% normal goat serum (Jackson ImmunoResearch) in PBSA with 0.3% Triton X-100 (PBSAT), then washed in PBSAT. Primary antibodies, diluted in PBSAT with 10% normal goat serum, were added and cells incubated overnight at 4°C at 100% humidity. Cells were washed with PBSAT and incubated with secondary antibodies diluted in PBSAT at room temperature for one hour. After washes in PBSA, cells were incubated at room temperature for 15 minutes in 1× DAPI nucleic acid stain (Life Technologies) in PBSA. After washes in PBSA, cells were mounted on slides using Prolong Gold (Life Technologies). Specific proteins were detected by immunofluorescence with the indicated antibodies (Table 1). Images (×63) were captured using a Zeiss LSM 510 confocal microscope.

Immunohistochemistry

Proteins were detected in tissue microarrays (TMAs). The TMA slides (TH641; US Biomax, Inc) were deparaffinized, immersed in 1× DAKO (Agilent Technologies), then diluted in 50% glycerol (Fisher Scientific) for 20 minutes at 95°C for antigen retrieval. Cooled slides were washed in H₂O, then 0.2% Tween 20 in H₂O (Sigma-Aldrich) and blocked for 1 hour at room temperature in 10% normal goat serum (Jackson ImmunoResearch) diluted in 5% BSA in PBS-0.3% Triton X-100. Primary antibodies, diluted in PBS containing 10% normal goat serum, 5.0% BSA, and 0.3% Triton X-100, were added and incubated overnight at room temperature. Slides were washed with 5.0% BSA in PBS and secondary antibodies, diluted in 5.0% BSA in PBS, were applied for 1 hour at room temperature. The TMA slides were washed, stained with 10-μg/mL DAPI (Life Technologies) for 15 minutes, and then washed in 5.0% BSA in PBS. Proteins were detected using antibodies optimized for immunohistochemistry (Table 1). Images were captured using a Zeiss LSM510 META confocal microscope (3 images/sample, ×63) and quantified by MetaMorph Image Analysis Software. Statistical significance was assessed by ANOVA and Bonferroni's Test using GraphPad Prism (GraphPad Software).

Chromatin immunoprecipitation (ChIP) and EMSA

TRβ binding to Runx2 chromatin was determined by ChIP-PCR. Cultured human thyroid cells were cross-linked with 1% formaldehyde for 10 minutes, neutralized with 125mM glycine, rinsed twice with PBS, pelleted, and frozen. Cells were lysed in the presence of protease inhibitors (Roche cOmplete) and chromatin was extracted and sonicated to 200–500 bp in size using a Covaris S220 Focused-ultrasonicator. Sonicated cell lysate was incubated with 2 μg of indicated antibody (Table 1) and rotated overnight at 4°C in 250-μL lysis buffer (50mM HEPES [pH 7.5], 140mM NaCl, 1mM EDTA, 1% Triton X-100, 0.1% Na-deoxycholate, and 0.1% sodium dodecyl sulfate). Twenty microliters of Protein G Dynabeads (Invitrogen) were blocked in 0.5% BSA and resuspended in lysis buffer, added to the lysate/antibody mix, and rotated at 4°C for 3 hours. Immune complexes were washed extensively, eluted, and incubated at 65°C overnight to reverse cross-links. Samples were treated with ribonuclease A, incubated with Proteinase K, phenol/chloroform extracted, and ethanol precipitated. Pellets were resuspended in 70-μL 10mM Tris-HCl, pH 8. Three microliters of IP or total input DNA was amplified with Q5 High-Fidelity DNA Polymerase (New England Biolabs) according to manufacturer's recommendations. Analysis of P1 and P2 transcripts of the Runx2 gene used promoter-specific primers (Table 2).

Table 2.

List of PCR Primers

Gene Name	Forward	Reverse	Amplicon (bp)	Sequence (Accession Number/Reference)
Human Runx2 Promoter 1	GGACAGCAAGAAGTCTCTGGTT	GATGCCATAGTCCCTCCTTTTT	358	Vega: OTTHUMG00000014774
Human Runx2 Promoter 2	CGGGTGTTCCAAAGACTCC	AAGTTTGATGAGGCCGACTG	129	Vega: OTTHUMG00000014774
Human TRβ	AGCTGCAGAAGTCCATCGG	GCATTGACTATTGGTGCTTG	166	Vega: OTTHUMG00000130478
Human TRβ	GCTAGGTACCTGGTATACATGAAACATTACATTTAATCTTTATT	GCTAGAATTCAGTCCCTCCTTTTTTTTTCAGATAGAA	1201	Vega: OTTHUMG00000130478
Human GAPDH	TGGCACCGTCAAGGCTGAGAA	TGCTAAGCAGTTGGTGGTGC	615	Vega: OTTHUMG00000137379
Human GAPDH	ATTGGGCGCCTGGTCAC	AAGATGTAAACCATGTAGTTGAGGTCA	91	Vega: OTTHUMG00000137379

Primers' sequence (5′ to 3′).

TRβ binding in the Runx2-P1 promoter was further characterized by EMSA. Complementary oligonucleotide sequences containing putative TREs (Supplemental Table 1) were biotin labeled at the 3′-OH terminus (Enzo). Annealed DNA was purified with Agencourt AMPure XP beads per manufacturer's protocol (Beckman Coulter). Biotinylated DNA (15 fmol) was incubated with either 10-μg nuclear protein lysate from thyroid cells or 100-ng recombinant TRβ protein (Active Motif) in binding buffer (Invitrogen/LifeTechnologies) for 30 minutes at room temperature in the absence or presence of anti-TRβ antibody (Table 1). The resulting complexes were resolved on 6% native polyacrylamide gels (Invitrogen/LifeTechnologies) and transferred to Biodyne B modified nylon membranes (Thermo Scientific) by electroblot (Bio-Rad Laboratories). After UV cross-link for 15 minutes, complexes were detected by enhanced chemiluminesence LightShift Chemiluminescent kit according to the manufacturer's protocol (Thermo Scientific) and imaged with VersaDoc MP3000 (Bio-Rad Laboratories).

Small interfering RNA (siRNA) transfection

Loss of function was assayed by siRNA using On Target Plus siRNAs targeting human TRβ (L-003447-00) with nonsilencing scrambled siRNA (D-001810-10-20; GE Healthcare Dharmacon) as control. Cells at a density of 2.5 × 10⁵ per well, 30%–50% confluence, were using transfected with 25nM–200nM siRNA with Oligofectamine per manufacturer's protocol (Invitrogen/Life Technologies). The effect of knockdown of TRβ on endogenous Runx2 and metastatic markers was determined after 24 hours. Negative control cells were treated with Oligofectamine without siRNA.

Reverse transcription-polymerase chain reaction

Total RNA extracted from cells using an RNeasy Plus Micro kit 74034 (QIAGEN Sciences) was used for cDNA synthesis using a cDNA synthesis kit 4368814 (Applied Biosystems). The TaqMan primers (Applied Biosystems) used for quantitative reverse transcription PCR (qRT-PCR) are summarized in Supplemental Table 1. For qRT-PCR, the relative copy number was calculated based on the target gene/actin as indicated using comparative threshold cycle method. For semiquantitative RT-PCR, transcripts were visualized on 1.8% agarose gels stained with ethidium bromide. Relative expression was determined by image capture using ImageJ 149 software, and data were summarized as target gene/glyceraldehyde 3-phosphate dehydrogenase (GAPDH). Primer sequences are shown in Table 2 and Supplemental Table 1.

Reporter constructs

The human Runx2-P1 1.2-kb (−763 to +408) and 0.7-kb (−763 to −16) luciferase constructs were designed by PCR amplification from genomic DNA extracted from H9 human embryonic stem cells (ATCC) and conventional cloning methods to insert fragments into the pGL3 luciferase reporter plasmid (Promega) (Supplemental Figure 1). The primers used are summarized in Supplemental Table 1. The −763 to −16 construct was generated by cloning the KpnI and HindIII digest fragment of the −763 to +408 pGL3 construct into the pGL3 plasmid. Mutations were introduced by QuikChange II Site-Directed Mutagenesis kit as indicated (Agilent Technologies). The human TRβ expression plasmid, pDEST-515 Flag, was a generous gift from Dr David Armstong (Laboratory of Neurobiology; NIEHS) (Supplemental Figure 2). Each construct was validated by Sanger sequencing.

Cotransfections and transient expression assays

Candidate TREs were functionally assessed in COS-7 cells transfected (Lipofectamine 2000) with human TRβ expression vectors and human Runx2 promoter luciferase reporter plasmids. The SV40-Renilla construct (Promega) was used to control for slight variability in transfection efficiency between plates. A total of 2.3 μg of plasmid DNA (0.25 μg:2.0 μg:0.05 μg; test:reporter:reference) was transfected per 1 × 10⁵ cells. After 2 days, cultures were harvested and reporter activities measured using the Dual-Luciferase Reporter Assay System (Promega) on a VICTOR X4 Multilabel Plate Reader (PerkinElmer) according to the protocol of the manufacturer. Statistical significance values were assessed by ANOVA and Bonferroni's Test using GraphPad Prism (GraphPad Software).

Results

Runx2 and TRβ are reciprocally expressed in thyroid benign and malignant cells

Increased Runx2 expression has been linked to tumor development, progression, and metastatic invasion in a number of cancers. Conversely, TRβ has been associated with tumor suppression and delayed tumor progression. Thus, we interrogated the underlying relationship and cellular processes of these transcription factors in thyroid cancer. In a microarray analysis, we had previously noted differential expression of Runx2 and TRβ (47). Thus, we first sought to establish Runx2 and TRβ expression in thyroid cell lines representing a spectrum of thyroid cancers as well as in thyroid tissue biopsy samples. Western blot analyses (Figure 1A) revealed reciprocal expression of Runx2 and TRβ proteins (low Runx2:high TRβ in normal cells, and high Runx2:low TRβ in malignant cells). The difference in expression corresponded to the degree of differentiation in PTC, FTC, and ATC with the greatest differences observed between normal and ATC cells. Because Runx1, another member of the Runx family, has also been implicated in breast (48), ovarian (49), prostate (50), and solid tumors, we also determined its relative expression. Runx 1 protein levels were not correlated with thyroid cancer cell types (Figure 1A).

Figure 1.

Runx2 and TRβ expression are inversely correlated in thyroid cells and tissues. A, Representative immunoblot of TRβ, Runx2, Runx1, and β-actin in thyroid cells is illustrated (upper panel). The histogram shows quantified levels of 3 independent experiments, mean ± SD (lower panel). Significances at P < .01 are indicated compared with levels in normal cells (ORI); +, TRβ; ++, Runx2; +++, Runx1. Levels are standardized to β-actin. Data are arbitrary densitometric units. B, Immunohistochemistry demonstrates TRβ, Runx2, and Runx1 proteins in TMAs corresponding to normal (N), benign (B), and specific stages (I–IV) of PTC, FTC, and ATC cancers (upper panel). Graph summarizes quantitation, mean ± SD (n = 6 per stage) of 3 separate experiments. Significances at P < .05 compared with normal tissue are indicated for each protein +, TRβ; ++, Runx2; +++, Runx1 (lower panel). C, Immunofluorescence illustrates detection of Runx2 and TRβ proteins in whole-thyroid cells.

In thyroid biopsy tissues, Runx2 was strongly expressed in late tumor stages (III, IV) of all thyroid cancer types, with the highest expression in ATC (Figure 1B), whereas TRβ expression was highest in normal and benign tissues. For comparison, Runx1 protein levels were also evaluated, but no significant correlation in expression across the spectrum of thyroid cancer tissues was noted. Thus, we did not further investigate Runx1 in these studies. Immunofluorescence confirmed that Runx2 protein was strongly expressed in malignant compared with normal thyroid cells and is predominantly located in the nucleus. In striking contrast, TRβ protein expression was highest in normal thyroid cells and primarily nuclear. In both differentiated and undifferentiated thyroid cancer cells, TRβ was located in the cytoplasm, perinuclear membrane and nucleus albeit at lower levels than in normal cells (Figure 1C).

The Runx2 gene is characterized by 2 promoters (P1 and P2), which generate distinct mRNAs that encode proteins with distinct N-termini. The P1 transcript corresponds to the Runx2 protein isoform with the N-terminal pentapeptide MASNS and the P2 transcript encodes the Runx2 protein isoform with N-terminal pentapeptide MRIPV (51). Using primers specific for each variant (Table 2), both Runx2-P1 and Runx2-P2 mRNAs were detected (Figure 2). Runx2-P1 transcript was minimally detectable in normal cells and 2- to 4-fold higher in differentiated PTC and FTC thyroid cells. Similar results were noted for Runx2-P2 mRNA. However, in the anaplastic cells (SW1736), Runx2-P1 transcript was dramatically increased (8-fold), similar to protein levels noted. These results indicate that changes in Runx2 expression in the malignant thyroid cells are mediated mainly by the Runx2-P1 promoter.

Figure 2.

Runx2-P1 (isoform 2) expression is increased in malignant thyroid cells. Representative semiquantitative RT-PCR shows RNA levels in triplicate corresponding to Runx2-P1 (isoform 2), Runx2–P2 (isoform 1), and TRβ in in normal (ORI), PTC (LAM-1), FTC (WRO), and ATC (SW1736) (upper panel). Significances at P < .01 are indicated compared with levels in normal cells (ORI); +, TRβ; ++, Runx2-P1; +++, Runx2–P2. Levels are standardized to GAPDH (lower panel). Data are arbitrary densitometric units.

Runx2 expression is decreased by T₃

Given the significant loss of TRβ expression in thyroid tumor cells that corresponds to an increase in Runx2 expression, we determined whether T₃ mediated cellular events could alter Runx2 levels. We observed that T₃ caused a time- and concentration-dependent decrease in Runx2 protein levels in both normal (ORI) and malignant (SW1736) thyroid cells (Figure 3A). We further found that T₃ caused a concentration-dependent decline in Runx2 mRNA levels at 6 hours (Figure 3B, upper panel) that was not significant for normal cells but was significant for anaplastic cells. By 24 hours, T₃ significantly suppressed Runx2 mRNA in both ORI and SW1736 cells (Figure 3B, lower panel). It is noteworthy that Runx2 levels increase in the first days of culture similar to what has been described in osteogenic cells. The rapid decline may reflect a “switch point” as Runx2 levels are sensitive to growth potential in a cell type-specific manner (52). Nevertheless, these results implicate a ligand-meditated TRβ suppression of Runx2 levels with greater responsivity in malignant cells.

Figure 3.

T₃ modulates Runx2 expression in normal (ORI) and malignant (SW1736) thyroid cells. A, T₃ (10⁻⁷M) induced time (upper panel)- and concentration (10⁻¹¹M to 10⁻⁷M, 24 h) (lower panel)-dependent changes in Runx2 protein levels. Graphs of Western blot analyses summarize quantitation of protein levels normalized to β-actin. Data are mean ± SD, n = 3. Significances at P < .05 compared with untreated control are indicated (+). B, Graphs summarize the quantitation of the effect of T₃ (10⁻¹¹M to 10⁻⁷M) for 6 hours (upper panel) or 24 hours (lower panel) on Runx2 mRNA. Results were standardized to β-actin and normalized to untreated cells. Data are mean ± SD, n = 3. Significances at P < .05 compared with untreated control are indicated (+).

Loss of TRβ increases Runx2 expression and tumorigenic genes

To determine whether TRβ plays a direct role in modulating Runx2 expression, we next determined the effect of knockdown of TRβ by siRNA transfection. With increasing siRNA, TRβ levels are reduced and concomitantly, Runx2 protein levels are increased (Figure 4A). Similar results were observed for TRβ and Runx2 RNA levels (Figure 4B). These data indicated that TRβ could directly alter Runx2 levels. Consistent with the increase in Runx2 levels, we observed an increase in Runx2-regulated genes associated with tumorigenesis (Figure 4C). These findings support the hypothesis that the reciprocal expression of TRβ and Runx2 in thyroid tumorigenesis reflects a novel regulatory pathway.

Figure 4.

Loss of function of TRβ increases Runx2 and Runx2 target genes. A, Western blot analysis illustrates that siRNA (TRβ) transfection into normal thyroid cells (ORI) results in a concentration-dependent decrease in TRβ protein and increase in Runx2 protein compared with scrambled nonspecific siRNA (ns) (upper panel). Graph summarizes quantitation, mean ± SD, of 3 separate experiments. Protein levels were standardized to β-actin and compared with respective controls. Significances at P < .01 are indicated (+). B, qRT-PCR illustrates that siRNA (TRβ) transfection results in corresponding changes in mRNA. Graph summarizes quantitation, mean ± SD of 3 separate experiments, of TRβ and Runx2 mRNA levels standardized to β-actin. Significance at P ≤ .01 compared with respective control is indicated (+). C, Knockdown of TRβ impacts Runx2 target genes; MMP13, MMP2, cyclin D1, osteopontin (OPN), and cadherin 6. qRT-PCR levels of target genes were standardized to β-actin. Data are the averaged fold change in transfected cells normalized to mock-transfected cells. Graph summarizes quantitation, mean ± SD of 3 separate experiments. Significances at P ≤ .01 compared with respective control are indicated (+).

TRβ directly binds to 3 TRE motifs in Runx2-P1

Given that Runx2 total expression and Runx2-P1 transcript levels appeared to directly correlate with TRβ levels, we next examined the possible direct binding of TRβ to the Runx2-P1 promoter by ChIP and EMSA. We identified 3 putative TRE motifs in the proximal Runx2-P1 promoter region (Figure 5). This region of the Runx2-P1 promoter is highly conserved across species and contains cis activation and repressor domains (53). Site A (−105 to −49) is located in the region with well-characterized regulatory elements including a vitamin D-response element (54) and Runx2-response elements (RREs) (55, 56). Sites B (+21 to +66) and C (+88 to +133) are located within the 5′-untranslated region (UTR). Site B contains multiple contiguous RREs and a single half-site TRE that corresponds to a negative TRE (57). Site C contains only a TRE motif and no RRE. Individual TRE motifs and organization are highly variable with diverse half-site spacing and orientation indicating that cellular context is a critical factor in TRβ-TRE interaction (58, 59). Thus, we first determined whether TRβ could directly bind to the DNA.

Figure 5.

TREs motifs are located in the Runx2-P1 promoter. Within the proximal region of the P1 promoter (⁻140/⁺140), several regulatory response elements have been identified, including vitamin D (VDRE) and Runx2 (RRE). TRE motifs located in this region are indicated as site A (⁻105/⁺49), site B (⁺21/⁺66), and site C (⁺88/⁺133). Nucleotide mutations are highlighted and mutated TRE sequences are indicated as AM, BM, and CM, respectively.

TRβ binding to the Runx2-P1 promoter was validated by ChIP-PCR (Figure 6A). As expected, Runx2 binding was also detected in the well-characterized promoter region. Of note, TRβ and Runx2 protein interaction with the P2 promoter was also detected although at lower levels in the anaplastic SW1736 cells (Figure 6A) as well as normal (THJ) and papillary cancer (BHP2-7) thyroid cells (Supplemental Figure 3.) Using EMSA, we demonstrated that recombinant TRβ specifically binds to each of the biotinylated Runx2 P1 native promoter sequences TRE-A, TRE-B, and TRE-C (Figure 6B). TRβ protein in normal (ORI) nuclear lysates specifically bound to each site as indicated (Figure 6C, lane 2). Addition of anti-TRβ antibody to the incubation resulted in a supershift of the TRβ protein complex as indicated by the upper arrow (Figure 6C, lane 3). Competition with 10-fold excess unlabeled homologous dsDNA resulted in a diminished binding (Figure 6B, lane 4). Further, introduction of selected mutations into each of the putative TRE sites (summarized in Figure 5) resulted in a loss of specific TRβ binding with both recombinant TRβ protein (Figure 6B) and nuclear lysate protein (Figure 6C).

Figure 6.

TRβ protein binds to Runx2 promoter sequences. A, TRβ binds to Runx2 promoters P1 and P2. Sonicated cross-linked chromatin extracts from human SW1736 anaplastic thyroid cells were IP with antibodies to TRβ, Runx2, or rabbit γ-globulin control (rIgG). Positive control input DNA (not IP) and purified IP DNA amplified with Runx2 P1 primers produce a 358-bp band for both TRβ and Runx2, but not for rIgG. Input DNA, TRβ, and Runx2-IPs, also show enriched amplification of the expected 129-bp Runx2 P2 product as compared with rIgG. Input DNA produces a 91-bp fragment when amplified with GAPDH primers, whereas TRβ, as expected, does not. B, EMSA illustrates recombinant TRβ protein bound to biotinylated Runx2 double-strand native oligomer sequences denoted as TRE-A (⁻105/⁻49), TRE-B (⁺21/⁺66), and TRE-C (⁺88/⁺133) (left panels). Lane 1 is DNA alone; lane 2 includes TRβ protein; arrow denotes complex. Mutations of TRE sequences as denoted in Figure 5, ablated the detection of the complexes (right panel). C, EMSA demonstrates the formation of nuclear protein complexes (indicated by lower arrows) with normal (ORI) cell nuclear extracts incubated with the TRE oligomer sequences (left panels). Incubation with antibody to TRβ protein resulted in a supershift indicated by the upper arrow. Mutation of the TRE sites resulted in a loss of detectable TRβ and associated nuclear protein complexes (right panels).

TRβ suppresses Runx2 promoter activity

We next examined the possible functionality of these TREs using a transient expression assay system that has been extensively used to characterize TRβ-TRE interactions. We tested the effect of TRβ expression on a Runx2-P1 promoter (−763 to +408), which contained the 3 TRE sites (A, B, and C) or a Runx2-P1 promoter without the 5′-UTR and site A alone (−763 to −16) linked to a luciferase (Luc) reporter. Of particular note, we found that basal expression of the 2 Runx2-P1-Luc expression plasmids is significantly different; Runx2-P1 (−763 to −16) Luc is 2- to 3-fold greater than Runx2-P1 (−763 to +408) (Figure 7, A and B). Cotransfection with human TRβ resulted in a decrease in Runx2 promoter activity in both Runx2 reporter plasmids (Figure 7, A and B). Surprisingly, the suppression was observed in the absence of added T₃. Furthermore, the addition of T₃ did not cause a further decrease in expression (Figure 7A).

Figure 7.

TRβ suppresses Runx2 promoter activity. A, Runx2-P1 promoter activity in the absence and presence of cotransfected TRβ expression plasmid is summarized as mean ± SD of 3 independent transient expression assays. Runx2-P1 ABC-Luc denotes ⁻763/⁺408 Runx2 sequence containing TRE sites A–C, Runx2-P1 A-Luc corresponds to ⁻763/⁻16 Runx2 sequence, which contains only site A. Significances at P < .01 are indicated (+). Differences in basal expression of the Runx2-P1 promoters are indicated; *, P < .01. The effect of added T₃ (10⁻⁷M, 48 h) is illustrated. B, Mutations of the TREs in the Runx2-P1 promoter resulted in differential basal promoter activity and loss of TRβ effects. Mutations in the specific TREs, summarized in the schematic in Figure 5, are indicated by lower case. Basal expression of Runx2-P1-Luc in the absence and presence of TRβ expression plasmid is summarized as mean ± SD of 3 independent transient expression assays. Differences in basal Runx2-P1 promoter activity in the absence and presence of TRβ expression plasmid are indicated; +, P < .01. Differences in native promoter sequences are noted: Runx2-P1 ABC; *, P < .01 and Runx2-P1 A; **, P < .01.

To further characterize the TREs, we determined the effect of mutations of the TREs comparable with those used for the EMSA analysis. In the Runx2-P1 (−763 to +408) Luc construct, mutation of site A (TRE aBC) resulted in loss of TRβ suppressive effect and no change in basal expression (Figure 7B). Mutations of site B (TRE AbC) and site C (TRE ABc) resulted in an increase in basal Runx2 promoter activity and a loss of TRβ responsiveness as well. In the Runx2-P1 (−763 to −16) Luc, mutation of site A (TRE a), resulted in a decrease in basal expression as well as a loss of TRβ responsiveness. Together, these results indicate that TRβ can bind directly to TRE motifs in the Runx2 P1 promoter to decrease gene expression.

Discussion

In the present study, we identified a novel signaling pathway that may be an important mediator of thyroid tumorigenesis; TRβ suppression of Runx2 expression. TRβ dysregulation has been reported in a number of cancers (10, 34, 36, 39, 45) and restoration of TRβ function linked with a reduction in tumor progression (34–36). However, the mechanism(s) by which this TRβ tumor suppressor effect occurs is not yet delineated. Further, Runx2, a recognized tumor promoter in epithelial cancers (12–14), including thyroid (23–28), has been shown to enhance tumor progression, in part by increasing expression of various MMPs, angiogenic factors such as vascular endothelial growth factor, and other EMT-related proteins (23, 24, 60). Our hypothesis that a molecular signaling pathway might link these thyroid tumor regulators was strengthened with our observation that TRβ and Runx2 proteins are reciprocally expressed in benign and malignant thyroid cell lines as well as in TMAs. The lowest expression of TRβ occurred in undifferentiated thyroid cancer cells, and in stage III and IV of tissue biopsy arrays. Our finding of diminished expression of TRβ in thyroid cancer types is consistent with previous reports of TRβ silencing in thyroid as well as other cancers (10, 34, 36, 39, 45).

The ratio of TRβ to Runx2 levels may be an important marker of tumor aggression. Consonant with characteristics of a tumor promoter, Runx2 protein and mRNA levels were low in normal and benign cells, and high in undifferentiated cancer cells particularly stage III and IV tumors, regardless of thyroid cancer type. Increased expression of Runx2 and its oncogenic functions have been identified in the progression of PTC, most notably associated with invasion and metastases, and correlated with increased levels of MMPs, angiogenic factors such as vascular endothelial growth factor, and other EMT-related proteins (23, 24, 60). Our findings support the concept of Runx2 as a tumor promoter in papillary thyroid tumors and extend these observations to include notably aggressive follicular, anaplastic, and undifferentiated thyroid tumors.

The discovery of TRβ regulation of Runx2 is clinically relevant elucidating multiple hormonal influences on this transcription factor critical to metastases to bone. This regulation was evidenced by T₃ induction of a time and concentration-dependent decrease in Runx2 protein and RNA levels. Moreover, the loss of TRβ by siRNA resulted in an increase in Runx2 expression as well as Runx2-regulated genes associated with tumor progression and metastases. These results implicated a T₃-TRβ-mediated regulation of Runx2 levels.

TRβ regulation of Runx2 transcriptional activity is mediated through the P1 promoter. The putative TREs, comprised of degenerate AGGTCA motifs (58, 59), were identified in the Runx2-P1 promoter (−105 to +133). The first of the 3 TRE motifs, TRE-A, located close to the transcriptional start site, resides within a domain of only 56 residues with regulatory elements that include the vitamin D receptor, activator protein 1, and Runx2 itself, RRE (54–56). TRE-B, similar to a well-characterized negative TRE (57), is located in the 5′-UTR near 3 RREs. TRE-C, which includes the sequence AGGTCA, is also located in the 5′-UTR without a proximal RRE. Each of these TRE motifs, A–C, mediated TRβ suppression of the Runx2 transcriptional activity. Together, these findings indicate that the TREs are a significant component of repressor regulation of Runx2 in normal thyroid cells that become deregulated in thyroid cancer.

Additional evidence for this regulation was provided by the mutations and deletions of the TREs. Deletion of the 5′-UTR (−15 to +408) encompassing sites B and C resulted in a higher basal level of expression. This observation is consistent with previous studies which revealed autoregulation of Runx2 mediated through 3 contiguous RREs in the 5′-UTR-mediating suppression by Runx2 (55, 56). However, mutation of TRE-B adjacent to the RREs as well as TRE-C both in the 5′-UTR resulted in an increase in basal expression consistent with TRβ suppression of Runx2 mediated through these native TRE motifs. The mutation of TRE-A in the context of TRE-B and TRE-C does not show a significant change in basal expression indicating that the TRE sites in the 5′-UTR are potent regulators of basal Runx2 transcriptional activity. Mutation of each of the TREs eliminated the TRβ suppressive effect further indicating that TRβ could suppress Runx2-P1 promoter activity through the direct interaction with the TRE motifs. We also noted that mutation of TRE-A in the Runx2-P1 resulted in a decrease in basal expression as well as a loss of TRβ induced effects. This observation is consistent with an activation of Runx2 mediated through this sequence. The Runx2-P1 promoter is characterized by repression and activation regions (56, 61, 62), as well as distinct regulatory DNA motifs and their cognate transcription factors (54–56, 63, 64). Given the proximity of regulatory elements within this region near the TRE, it is possible that the mutation in TRE-A may alter additional nuclear protein complexes (56). Given that TRs may also exert effects through “transcriptional cross talk,” protein-protein interaction with other DNA-modulating proteins rather than direct DNA interaction may also be contributing to the complex regulatory mechanisms (65–67).

Given that T₃ caused a suppression of Runx2 levels in thyroid cells, we had anticipated that TRβ genomic effects would be ligand-mediated. To our surprise, addition of T₃ in the transient expression assay did not amplify the TRβ suppression of Runx2 expression; T₃ instead blunted the effect. Unliganded TRs can act as strong repressors, in part, through the demonstrated interaction with corepressor regulatory molecular complexes such as nuclear receptor corepressor or silencing mediator of retinoic and thyroid receptor and recruitment of epigenetic modifiers such as histone deacetylase (65–69). Furthermore, molecular complexes that regulate transcription are localized in tissue-specific chromatin structures within distinct nuclear domains.

Our studies show that unliganded TRβ specifically binds to chromatin as well as TRE motifs of the Runx2 promoter and directly represses Runx2 promoter activity. That T₃ did not amplify the TRβ suppression of Runx2 promoter activity but did decrease Runx2 protein and RNA levels may reflect the limitations of a transient expression assay to entirely replicate intact cell systems and/or the possibility that T₃ may also alter additional signaling pathways to alter Runx2 expression and activity. The mechanism(s) by which TRβ induces inhibition of tumorigenesis have not yet been clarified although TRβ suppression of activation of phosphoinositide 3-kinase and extracellular signal-regulated kinase signaling pathways (44, 46, 70–73) as well as inhibition of β-catenin (45) are clearly implicated. The mitogen activated protein kinase inhibitor U0126 has been shown to suppress Runx2, implicating an involvement of the mitogen activated protein kinase/extracellular signal-regulated kinase pathway (23). Further, a reciprocal activation of Runx2 and phosphoinositide 3-kinase/protein kinase B has been reported in thyroid cancer with thyroid tumor progression. Future studies will seek to reveal possible involvement of intracellular signaling pathways that might additionally mediate TRβ modulation of Runx2 in cell systems and in xenograft models.

Our findings demonstrate that TRβ modulates Runx2 in thyroid cells at least in part through the regulation of gene expression. We can conclude that the loss of functional TRβ contributes to Runx2 oncogenic actions in thyroid tumorigenesis. Understanding the relationship between dysregulation of these transcription factors and associated changes in the nuclear microenvironment, correlated with progression, is critical for development of novel interventions and diagnostics. We have functionally identified a novel signaling pathway in thyroid cancer and a possible mechanism for TRβ tumor suppressor effects in thyroid cancer which provides an opportunity for new targeted studies.