The Thyroid Hormone Receptor Is a Suppressor of ras-Mediated Transcription, Proliferation, and Transformation

Abstract

The thyroid hormone triiodothyronine (T3) has a profound effect on growth, differentiation, and metabolism in higher organisms. Here we demonstrate that T3 inhibits ras-induced proliferation in neuroblastoma cells and blocks induction of cyclin D1 expression by the oncogene. The hormone, at physiological concentrations, strongly antagonizes the transcriptional response mediated by the Ras/mitogen-activated protein kinase/ribosomal-S6 subunit kinase (Rsk) signaling pathway in cells expressing thyroid hormone receptors (TRs). T3 blocks the response to the oncogenic forms of the three ras isoforms (H-, K-, and N-ras) and both TRα and TRβ can mediate this action. The main target for induction of cyclin D1 transcription by oncogenic ras in neuroblastoma cells is a cyclic AMP response element (CRE) located in proximal promoter sequences, and T3 represses the transcriptional activity of b-Zip transcription factors such as CREB (CRE-binding protein) or ATF-2 (activation transcription factor 2) that are direct targets of Rsk2 and bind to this sequence. The hormone also blocks fibroblast transformation by oncogenic ras when TR is expressed. Furthermore, TRs act as suppressors of tumor formation by the oncogene in vivo in nude mice. The TRβ isoform has stronger antitransforming properties than the α isoform and can inhibit tumorigenesis even in hypothyroid mice. These results show the existence of a previously unrecognized transcriptional cross talk between the TRs and the ras oncogene which influences relevant processes such as cell proliferation, transformation, or tumorigenesis.

The ras protooncogenes encode 21-kDa GTP-binding proteins which act as pivotal mediators of signals acting at the membrane by transferring information from this cellular compartment to the nucleus. Activating mutations in ras are present in at least 30% of human tumors, and oncogenic ras efficiently transforms most immortalized rodent cell lines (3, 23). Several downstream pathways are initiated after Ras activation. The best studied are those involved in cell survival, the phosphatidylinositol-3-OH (PI3) kinase pathway, and in mitogenic signaling, the Ras/mitogen-activated protein kinase (MAPK) signaling pathway (5, 46). In the latter, activation of the MAPK extracellular signal-regulated kinase 1/2 (Erk1/2) permits its translocation to the nucleus, where it may modulate gene expression via the direct phosphorylation of transcription factors or the activation of downstream kinases such as Rsk (51), which then phosphorylate, among other substrates, b-Zip transcription factors of the cyclic AMP (cAMP) response element-binding protein (CREB)/activation transcription factor 2 (ATF-2) family (48).

Cyclin D1 plays an important role on cell cycle progression and is one of the main targets for the proliferative and transforming effects of ras oncogene (8, 22). It has been shown that ras-induced tumorigenesis depends on signaling pathways that act preferentially through cyclin D1. Thus, analysis in cyclin D1 knockout mice reveals that this protein is required for ras-dependent malignant transformation of the mammary glands (50). Similarly, skin tumorigenesis mediated by oncogenic ras is strongly reduced in mice deficient in cyclin D1 (35). Ras regulates the activity of the cyclin D1 promoter in various cellular systems (1), and multiple effector pathways and promoter elements can contribute to cyclin D1 expression (9, 12).

The thyroid hormones are important regulators of growth, development, and metabolism in higher animals and humans. The actions of the thyroid hormone triiodothyronine (T3) are initiated by binding to nuclear thyroid receptors (TRs), the cellular counterparts of the retroviral v-erbA oncogene, encoded by two genes, α and β, which give rise to different receptor isoforms (49). TRs are widely distributed in mammalian tissues, but transformed or immortalized cells in general express very low levels of TR. In addition, there is increasing evidence that alterations in TRs are common events in cancer. These alterations, which include loss of heterozygosity, gene rearrangements, promoter methylation, aberrant splicing, point mutations, or changes in the level of expression, suggest that TR genes may function as tumor suppressors (7, 10, 21, 24), although the role of these receptors in the pathogenesis and progression of neoplasic processes is currently unclear.

TRs act as ligand-inducible transcription factors by binding to DNA response elements (TREs) located in regulatory regions of target genes. Nuclear receptors can also modulate gene expression by mechanisms that are independent of binding to DNA. Thus, they can alter expression of genes that do not contain a hormone response element through positive or negative interference with the activity of other transcription factors and signaling pathways, a mechanism generally referred to as transcriptional cross talk. For example, some nuclear receptors can negatively regulate target gene promoters that carry AP-1, CRE (for cAMP response element), or NF-κB sites without binding to these DNA elements themselves (11, 17, 32, 38). The receptors do not bind to these elements in vitro, but in vivo the liganded receptors can be tethered to the promoter through protein-protein interactions (25, 28, 36).

In the present study we analyzed the existence of a potential cross talk between the TR and Ras signaling pathways. For this purpose, one of the models used was N2a neuroblastoma cells, which express the TR β1 isoform (N2a-β cells). In these cells T3 blocks proliferation and induces morphological differentiation by an arrest in G₀/G₁ (18). We have previously demonstrated that T3, in the presence of serum growth factors, coordinately regulates the expression of several genes, among them cyclin D1, that play a key role in cell cycle control (30, 31). Our results show that T3 blocks oncogenic ras-mediated proliferation and transcriptional induction of cyclin D1 in neuroblastoma cells by interfering specifically with the activity of the Ras/Erk/Rsk pathway and CRE-mediated transcription. The hormone also blocks fibroblast transformation by oncogenic ras when TR is expressed. Furthermore, TRs act as suppressors of tumor formation by the oncogene in vivo in nude mice. These results demonstrate the existence of a previously unrecognized transcriptional cross talk between the TRs and the ras oncogene and show that these receptors could play a relevant role as suppressors of ras-dependent tumors.

MATERIALS AND METHODS

Transfections.

N2a-β cells, which express in a stable manner the TRβ1 isoform (18), were grown in medium containing 10% bovine fetal serum depleted of thyroid hormones by treatment with resin AG-1-X8 (Bio-Rad). Cells were transfected with 1.5 μg of constructs containing 5′-flanking deletions of the cyclin D1 promoter cloned in pXP2-luc (13). In plasmid extending to nucleotide −269, the CRE sequence GTAACGTCA was mutated to GTACCCCCA. A palindrome of the AGGTCA sequence inserted into the mouse mammary tumor virus promoter was used as a TRE-containing construct. Luciferase plasmids containing 2.3 kb of the p21^Cip1 promoter cloned in pGL3 (Promega) and 1.9 kb of the p27^kip1 promoter cloned in pGL2 (Promega) were also used. Reporter plasmids were cotransfected with expression vectors for constitutively active Val12 mutants of Ha-, K, or N-ras (50 ng) or for the dominant-negative mutants pCEFL-Ha-ras^asn17, pEF-myc-Raf301, or pcDNA-dnMEK (a gift from P. Crespo). Constitutively active forms of Raf (pCEFL-RafBXB), Erk (pCEFL-MEK-ERK), Rsk2 (pH3K-Rsk2 Y707A), or viral Src (pSV-Src) were also used when indicated. The expression vectors for cTRα1 and hTRβ1 were pSG5-TRα and pSG5-TRβ. Murine cyclin D1 cloned in pLPC, a gift from Manuel Serrano, was used as an expression vector for this protein. Fusions of CREB, ATF2, and ELK1 with the DNA-binding domain of GAL4 (250 ng) were cotransfected with 1 μg of a reporter plasmid containing four binding sites for GAL4 (pE1b 4xUAS-luc) to determine the activity of these transcription factors. Cells were transfected by incubation with a mixture of cationic liposomes in 35-mm wells for 6 h. Cells were then treated for 36 h in the presence or absence of 5 nM T3 in medium containing 0.1% of T3-depleted serum. Luciferase or chloramphenicol acetyltransferase (CAT) activity were determined in 10 μg of cell protein. The transfection efficiency was determined by cotransfection with a cytomegalovirus-β-galactosidase vector. Each experiment was performed in triplicate and was repeated at least three times. The data are means ± the standard deviations unless otherwise indicated and are expressed as the level of induction relative to the values obtained in the control cells transfected with an empty vector.

BrdU incorporation.

N2a-β cells were plated in 24-well plates and cotransfected by using Transfact transfection reagent (Promega) with 10 ng of a plasmid encoding enhanced green fluorescence protein (p-EGFP-C1; Clontech) and 40 ng of pCEFL-Ha-ras^val12 or the corresponding empty vector. After 48 h of incubation in the presence or absence of 50 nM T3, cells were pulsed with bromodeoxyuridine (BrdU). Labeling was performed as indicated in the Boehringer Mannheim Biochemica manual for the 5-bromo-2′-deoxyuridine labeling and detection kit I by using the anti-BrdU BMC 9318 antibody and the rhodamine-conjugated anti-immunoglobulin G antibody 115-025-003 (Jackson Immunoresearch). Both antibodies were used at a 1/300 dilution. Cells were visualized under fluorescence microscopy, and data are expressed as the percentage of BrdU-positive cells with respect to total EGFP-expressing cells for each experimental group. Values were obtained from analysis of three independent cultures, in which at least 300 EGFP-labeled cells were scored.

Flow cytometry.

Triplicate cultures of N2a-β cells grown in 90-mm petri dishes were transfected with 2 μg of pEGFP-C1 and 8 μg of pCEFL-Ha-ras^val12 in the presence or absence of 0.5 μg of pLPC-cyclin D1. After an overnight incubation, cells were shifted to medium containing 0.1% of T3-depleted serum and incubated for an additional 24 h in the presence or absence of 5 nM T3 before being sorted in a FACSVantage (Becton Dickinson) cell sorter. Fluorescent cells were harvested, centrifuged, resuspended in phosphate-buffered saline containing 0.1% NP-40 and 0.2 μg of RNase/ml, and incubated for 30 min before the addition of propidium iodine (0.05 mg/ml). Cell cycle distributions of cells were determined by measurement of DNA content with the use of an EpicsXL flow cytometer (Coulter).

Western blot analysis.

Whole extracts (15 μg) from N2a-β cells treated with 5 nM T3 for 16 h were used for immunodetection of cyclin D1, p27^Kip, and p21^Cip with antibodies sc-718, sc-776, and sc-397-G, respectively, from Santa Cruz Biotechnology at a 1/2,000 dilution as described previously (30). Monoclonal antibody Ab-4 (Calbiochem) was used at a 1/1,000 dilution for the detection of Ras in NIH 3T3 fibroblasts. This antibody recognizes all Ras isoforms. Antibodies for TRα and TRβ (sc-772 and sc-738, respectively; Santa Cruz Biotechnology) and for ERK2 and Rsk2 (sc-153 and sc-1430, respectively; Santa Cruz Biotechnology) were used at a 1/2,000 dilution.

Determination of Rsk2 and Erk activity.

For determination of Rsk2 activity, 10⁶ N2a-β cells were transfected by using Transfact with 250 ng of hemagglutinin (HA)-tagged Rsk2 in combination with the same amount of pCEFL-Ha-Ras^val12 or pCEFL. Extracts from untreated cells or cells treated with T3 for 24 h were immunoprecipitated with 10 μl of HA hybridoma. Kinase assays were performed in immunoprecipitates in the presence of 5 μCi of [³²P]ATP and 100 μM unlabeled ATP with 0.5 μg of glutathione S-transferase-Myt as substrate. In the case of Erk, cells were transfected with 2 μg of HA-tagged Erk2 and 500 ng of pCEFL-Ha-Ras^val12 or the empty vector, and the kinase assays were performed with 1 μCi of [³²P]ATP and 20 μM unlabeled ATP with 1.5 mg of myelin basic protein/ml as substrate.

Gel retardation assays.

Electrophoretic mobility shift assays (EMSAs) were performed as described elsewhere (31) with 2 μg of nuclear extracts from N2a-β cells and a ³²P-labeled oligonucleotide probe, comprising the −75/−48 region of the cyclin D1 promoter. Specific antibodies for CREB (sc-186X; Santa Cruz Biotechnology), ATF-2 (sc-187X; Santa Cruz Biotechnology), or PPARα (sc-1984X; Santa Cruz Biotechnology) at 1 μg were used in supershift assays.

Focus formation assays.

NIH 3T3 fibroblasts were plated in 90-mm dishes and transfected with standard calcium phosphate procedures with 50 ng of pCEFL-Ha-Ras^val12 and 4 μg of expression vectors for TRα1 or TRβ1 or with the same amount of empty vector. Cultures were fed with fresh medium containing 10% nonstripped donor calf serum every 2 days in the presence or absence of 5 nM T3. Alternatively, the cells were transfected with 50 ng of pCEFL-Ha-Ras^val12, 0.5 μg of TRβ1, and/or 0.5 μg of an expression vector for cyclin D1 (pLPC-cyclin D1), and the cells were grown in thyroid hormone-depleted medium. In both cases, at 14 days after transfection the foci were stained with Giemsa and scored visually.

Tumorigenesis in nude mice.

Stable transformants of NIH 3T3 fibroblasts were obtained by transfection with 100 ng of pCEFL-Ha-Ras^val12 or pCEFL alone, followed by selection with G418. Receptor-expressing fibroblasts were obtained by transfection of these cells with 1 μg of pLPCX-cTRα1 or pLPCX-hTRβ1 and selection with puromycin. Pools of resistant cells were used in all cases. For tumor formation in mice (athymic nude), 10⁶ cells were injected subcutaneously into each flank of four mice. Similar injections were performed in parallel in normal mice and in mice made hypothyroid by treatment with 0.02% methymazole and 0.1% sodium perchlorate in the drinking water. Treatment started 12 days before inoculation and was continued for the duration of the experiment. This treatment reduces by more than 80% the levels of circulating thyroid hormones (T3 and thyroxine). The lag time for tumor formation was defined as the time period comprised between inoculation and the day at which tumors reached 1 cm in diameter. Mice were sacrificed when tumors reached 2 cm. Experiments were performed in compliance with European Community law 86/609/EEC and were approved by the Consejo Superior de Investigaciones Científicas committee. Histological analysis was performed after hematoxylin-eosin staining.

RESULTS

T3 blocks ras-induced proliferation and ras-mediated transcriptional responses.

To assess the influence of T3 on ras-induced proliferation, N2a-β cells were transfected with Ha-ras^val12 and treated with the hormone for 48 h. Expression of oncogenic ras increased cell growth, and T3 reduced basal growth and blocked oncogene-dependent proliferation (Fig. 1A).

FIG. 1.

T3 blocks ras-mediated proliferation and cyclin D1 expression. (A) BrdU incorporation was determined as indicated in Materials and Methods in N2a-β cells cotransfected with EGFP and Ha-ras^val12 and incubated in the presence or absence of 5 nM T3 for 48 h. (B) Representative Western blot of cyclin D1, p27^Kip, and p21^Cip in cells transfected with 50 ng of Ha-ras^val12 or with the same amount of an empty vector. After transfection cells were treated for 16 h in the presence or absence of 5 nM T3. (C) Flow cytometry analysis of cells transfected with EGFP and Ha-ras^val12 alone or in combination with an expression vector for cyclin D1. The number of cells in S+G₂/M phases was determined in control cells and in cells incubated with T3, as indicated in Materials and Methods, and the percentage of inhibition by the hormone is shown for each experimental condition.

Since proliferative responses to ras appear to depend on induction of cyclin D1, the levels of this protein were determined in neuroblastoma cells expressing Ha-ras^val12. The oncogene caused a significant increase of cyclin D1 levels (Fig. 1B), and incubation with T3 markedly inhibited induction of cyclin D1 expression by oncogenic ras. To establish that repression of cyclin D1 levels by T3 indeed plays a role in the blockage of ras-dependent proliferation, we analyzed the effect of cyclin D1 overexpression on cell cycle control by T3 (Fig. 1C). Expression of cyclin D1 did not affect cell cycle progression in Ha-ras^val12-transfected cells, but it reversed to a significant extent the inhibitory effect of T3, demonstrating that the reduction of cyclin D1 levels by T3 is an important component of the mechanism by which the hormone antagonizes ras-mediated proliferation.

Other proteins, such as the cyclin kinase inhibitors (CKIs), could also be targets of T3 for repression of Ha-ras^val12-mediated proliferation. Since we have previously demonstrated that T3 increases p27^Kip levels in N2a-β cells (30) and p21^Cip is a well-known target of oncogenic ras, the level of these proteins was also examined. As shown in Fig. 1B, T3 produced a strong increase in the expression of both CKIs, whereas the oncogene increased specifically p21^Cip levels. However, the effect of T3 and Ha-ras^val12 was not additive, and p21^Cip levels were even reduced when both agents were present together. Therefore, T3 can also antagonize the effect of the oncogene on CKI expression.

To analyze whether changes in transcription of the cyclin D1 gene are involved regulation by T3, we performed transient-transfection assays with a reporter plasmid containing the 5′-flanking region of the cyclin D1 gene. Ha-ras^val12 stimulated transcription of the cyclin D1 gene, and incubation with T3 for 36 h strongly antagonized the response (Fig. 2A). Maximal inhibition by T3 was found at 1 nM, demonstrating that the hormone blocks the transcriptional response to oncogenic ras at physiological concentrations, an effect compatible with a TR-mediated response. To dismiss the possibility that the inhibitory effect of T3 on cyclin D1 promoter activation by ras could be secondary to the reduction on cell proliferation, cyclin D1 promoter activity was determined after different periods of incubation with the hormone. As illustrated in Fig. 2B, T3 was able to repress the response to Ha-ras^val12 as soon as the effect of the oncogene was detected in the transfected cells, i.e., 12 h. Not only Ha-ras^val12 but also the oncogenic forms of N- and K-ras activated the cyclin D1 promoter, and T3 antagonized the response to the different oncogenes (Fig. 2C). This was not due to a decrease in the expression of Ras proteins that was similar in the absence and presence of T3. Furthermore, the inhibition of cyclin D1 levels is not due to a general repressive effect of T3 on protein expression, since the levels of Erk2 used as a loading control were unaltered upon incubation with the hormone (Fig. 2C). To analyze whether the repressive effect of T3 is a physiological process or is due to overexpression of the TRβ isoform in N2a-β cells, TR expression levels, as well as the influence of T3 on induction of cyclin D1 promoter activity by ras, was also examined in pituitary GH4C1 cells, which express endogenous TRs. Levels of TRβ were very low in parental N2a cells but were similar in N2a-β and in the pituitary cells where TRs exist at natural levels and are not overexpressed (Fig. 2D). As shown in Fig. 2E, T3 also very potently antagonized induction of cyclin D1 promoter activity by oncogenic ras in GH4C1 cells. These results show that antagonism of ras-induced cyclin D1 transcriptional induction is observed at physiological receptor levels and that T3-dependent repression is not a specific effect on neuroblastoma cells and can be extended to other cell types.

FIG. 2.

T3 represses activation of the cyclin D1 promoter by oncogenic ras. (A) Transient-transfection assays with a reporter plasmid containing the cyclin D1 promoter (sequences −1720/+141) and 50 ng of Ha-ras^val12 or the empty vector. The luciferase activity was determined in cells incubated for 36 h with the concentrations of T3 indicated and are expressed as the fold induction over the values obtained in the untreated cells transfected with the empty vector. (B) The reporter activity was determined after different periods of incubation in the presence or absence of 5 nM T3. (C) Cells were transfected with activated forms of Ha-, K-, and N-ras^val12 (50 ng) and incubated with 5 nM T3 for 36 h before determination of luciferase activity. In the lower panels, the levels of Ras proteins were determined by Western blotting with a pan-Ras antibody, and the levels of Erk2 were determined as a loading control. (D) TRβ expression was determined by Western blotting with 20 μg of proteins from parental N2a, N2a-β, and pituitary GH4C1 cells. (E) Cyclin D1 promoter activity was determined in pituitary GH4C1 cells transiently transfected with 50 ng of Ha-ras^val12 and treated for 36 h in the absence and presence of 5 nM T3.

The requirement of TRs for transcriptional antagonism of ras-mediated responses by T3 was analyzed in cells, such as N2a cells, expressing low TR levels. The hormone did not repress cyclin D1 promoter activity in parental N2a cells (Fig. 3A). However, expression of both TR isoforms caused a paradoxical increase of basal promoter activity in the absence of ligand, which was reversed by T3, and conferred a strong T3-dependent antagonism to promoter activation by oncogenic ras. In other cell types, such as MCF-7 (Fig. 3B) or HeLa (not illustrated) cells, which express low TR levels, incubation with T3 did not affect the response to ras oncogene. In contrast to the findings in neuroblastoma cells, a significant activation by the unoccupied TRα or TRβ was not observed in these cells, but again a strong repressive effect of the hormone on stimulation by Ha-ras^val12 was found after expression of both receptor isoforms. Therefore, although in neuroblastoma cells both basal and induced levels are reduced by T3, the hormone can repress specifically the response to the oncogene independently of basal reporter levels in different cell types.

FIG. 3.

TRs antagonize transcription by oncogenic ras and src. (A) Parental N2a cells were transfected with the construct containing the cyclin D1 promoter and expression vectors for activated Ha-ras^val12 (50 ng) and TRα1 (100 ng) or TRβ1 (100 ng) as indicated. The luciferase activity was determined after 36 h of incubation in the presence or absence of 5 nM T3. (B) Similar experiments were performed in MCF-7 breast cancer cells. (C) Activity of p21^Cip (left) and p27^Kip (right) promoters in N2a-β cells expressing Ha-ras^val12. The luciferase activity was determined in control and T3-treated cells. (D) N2a-β cells were transfected with a CAT construct containing a consensus palindromic TRE fused to the mouse mammary tumor virus promoter and 50 ng of Ha-ras^val12. CAT activity was determined after 48 h of incubation in the presence or absence of 5 nM T3. (E) Cyclin D1 promoter activity in N2a-β cells transfected with v-src (250 ng) in the presence or absence of 25 ng of the dominant-negative N-ras^asn17 mutant (dnRas). The luciferase activity was determined in control cells and in cells treated with 5 nM T3 for 36 h.

The influence of T3 on regulation of the p21^Cip and p27^Kip promoters was also examined in N2a-β cells. Although T3 induced the levels of both CKIs (see Fig. 1B), the hormone did not increase the activity of promoter constructs containing the 5′-flanking regions of these genes. This is in agreement with our previous results and with the finding that these inhibitors are mainly regulated at a posttranscriptional level by stabilization of the protein half-life (30). The p21^Cip and p27^Kip promoter constructs were, however, stimulated by Ha-ras^val12, and incubation with T3 also strongly reduced this response (Fig. 3C). Therefore, TRs mediate a strong repression of ras-mediated transcriptional responses on the different promoters and cell types assayed. In order to examine whether the transcriptional antagonism between oncogenic ras and TRs is mutual, we also analyzed the effect of Ha-ras oncogene on the activity of a reporter plasmid containing a consensus TRE. T3 stimulated the activity of this construct in N2a-β cells (Fig. 3D), and transfection of Ha-ras oncogene significantly enhanced T3-mediated transactivation. Therefore, antagonism is not reciprocal and oncogenic ras could potentiate, rather than repress, TR-mediated transcriptional responses.

Ras is an important effector of the src oncogene (33, 43), and cyclin D1 also appears to be a target for the effects of this tyrosine kinase (34). For that reason, we also tested the influence of T3 on the transcriptional response to v-src. Cyclin D1 promoter activity was stimulated by v-src in N2a-β cells by a Ras-dependent mechanism, since this stimulation was blocked by a dominant-negative form of Ha-Ras (Fig. 3E). Accordingly, T3 also repressed induction of cyclin D1 promoter activity by v-src.

T3 antagonizes the Ras/Erk/Rsk pathway.

Different effector pathways can be initiated after Ras activation. To analyze the pathways involved in cyclin D1 promoter activation by ras in N2a-β cells, we first used different inhibitors. As shown in Fig. 4A, the PI3-kinase inhibitor LY294002 did not affect induction by Ha-ras^val12, and the same occurred with SB203580, an inhibitor of the MAPK p38. In contrast, incubation with the MEK inhibitor U0126 strongly reduced basal promoter activity and also blocked the response to the oncogene, showing that the Ras/Erk pathway is its main effector to stimulate cyclin D1 transcription in neuroblastoma cells. Confirming a key role for this pathway, expression of dominant-negative mutants of Ras and its downstream effectors Raf and MEK totally blocked promoter activation by Ha-ras^val12 (Fig. 4B). This implies that T3 should antagonize activation by the Ras/Erk pathway. In agreement with this hypothesis, T3 did not affect stimulation of cyclin D1 promoter activity by a constitutively active mutant of the catalytic subunit of PI3-kinase (not illustrated) but strongly repressed stimulation by the activated forms of Ras, Raf, and Erk (Fig. 5A). From the finding that T3 is still able to block stimulation by Erk, it can be deduced that the antagonism is exerted downstream this kinase. However, we analyzed the possibility that T3 could repress the promoter response by inhibiting Erk activity. For this purpose, Erk kinase activity was determined in Ha-ras^val12-expressing cells in the presence or absence of T3. As shown in Fig. 5B, Erk activity, which was undetectable in the absence of Ras, was strongly induced by the oncogene, but T3 did not significantly alter this response.

FIG. 4.

The Ras/Erk pathway is the main effector for Ras-mediated stimulation of the cyclin D1 promoter in neuroblastoma cells. (A) N2a-β cells were transfected with the cyclin D1 promoter construct and 50 ng of Ha-ras^val12. Reporter activity was determined in control cells and in cells incubated for 36 h with 5 nM T3, 10 μM U0126, 10 μM SB203580, or 20 μM LY294002. (B) The Ras vector was cotransfected with the dominant-negative mutants pCEFL-N-ras^asn17 (dnRas; 100 ng), pEF-myc-Raf301 (dnRaf; 500 ng), and pcDNA-dnMEK (dnMEK; 100 ng) or with the same amounts of the corresponding empty vectors. The luciferase activity was determined 36 h after transfection.

FIG. 5.

T3 antagonizes transcriptional responses mediated by the Ras/Erk/Rsk2 signaling pathway. (A) N2a-β cells were transfected with the cyclin D1 promoter construct and the indicated constitutively active forms of Ha-ras (Ras^val12; 50 ng), raf (RafBXB; 250 ng), or erk (MEK-Erk; 25 ng). The luciferase activity was determined after 36 h in control and T3-treated cells. (B) Erk kinase activity was determined as indicated in Materials and Methods in N2a-β cells transfected with HA-tagged erk2 and Ha-ras^val12, followed by incubation in the presence or absence of 5 nM T3 for 24 h, with myelin basic protein as the substrate. The lower panels show the autoradiography of a representative kinase assays, as well as the total amount of Erk protein measured by Western blotting. The kinase blots were quantitated, and the upper panel represents the mean kinase activity obtained in three independent assays. (C) N2a-β cells were transfected with constitutively active forms of erk (100 ng) and rsk2 (500 ng), and luciferase activity was determined in control cells and in cells treated with 5 nM T3 for 36 h. (D) Rsk2 kinase activity was determined in cells transfected with HA-tagged rsk2 and Ha-ras^val12 and incubated in the presence or absence of T3 for 24 h with GST-Myt as acceptor. The lower panels show a representative blot and the total cellular levels of Rsk2, and the upper panel illustrates the mean values of kinase activity obtained from three independent experiments.

Since Rsks are important targets of Erks and mediate many of the effects of these kinases, the influence of T3 on the promoter response to Rsk was also examined. As shown in Fig. 5C, a constitutively active Rsk2 mutant was as potent as the constitutive active Erk to activate the cyclin D1 promoter, but T3 was unable to repress the response to Rsk2 (Fig. 5C). This result suggests that T3 could antagonize Ras-mediated transcription by preventing Rsk2 stimulation. Therefore, we also investigated the effect of T3 on Rsk2 kinase activity. As illustrated in Fig. 5D, expression of oncogenic ras induced Rsk2 activity and, in contrast to the lack of effect of T3 on Erk activity, the hormone strongly inhibited Rsk2 kinase activity. The total amount of Rsk2 was not affected by either Ha-ras^val12 expression or incubation with T3. These results again demonstrate that Rsks rather than Erks could be a target for the repressive effect of the hormone on oncogenic ras-induced transcription.

CRE-mediated transcription is the main target for repression by T3.

We next analyzed the DNA elements responsible for activation by ras oncogene and repression by T3 in N2a-β cells by using transient-transfection assays with reporter plasmids containing successive 5′ deletions of the cyclin D1 promoter. Deletion of sequences comprised between −1720 and −91 did not greatly affect these responses (Fig. 6A). However, deletion to nucleotide −29 abolished stimulation by the oncogene and consequently inhibition by T3. The region between −91 and −43 does not contain a recognizable TRE but contains a CRE. To analyze contribution of this promoter element to regulation by oncogenic ras and T3, the CRE was mutated in the context of the −269 reporter plasmid. Mutation of the CRE abolished both responses (Fig. 6B), showing that this element mediates cyclin D1 regulation of transcription by these signals in neuroblastoma cells. Gel retardation assays with nuclear extracts from N2a-β cells revealed the formation of retarded complexes with the CRE that contain CREB and ATF-2, as shown by incubation with antibodies specific for these transcription factors. Figure 6C shows that a CREB antibody induced a supershift of the retarded band and that the ATF-2 antibody blocked complex formation that was, however, not affected by a nonspecific antibody. Furthermore, cotransfection of dominant-negative forms of CREB and ATF-2 blocked the promoter response to ras oncogene and T3 (data not shown), suggesting again their implication. On the other hand, expression of oncogenic ras and treatment with T3 did not alter the abundance of proteins that bind the CRE (Fig. 6C). This indicates that changes in activity rather than in the levels of these factors are responsible for transcriptional regulation. To prove this point, GAL4-CREB and GAL4-ATF-2 fusions were cotransfected into N2a-β cells with a reporter plasmid containing binding motifs for GAL4. In parallel with the observed changes in Rsk2 activity, expression of Ha-ras oncogene strongly increased CREB- and ATF-2-dependent gene expression, and this response was inhibited in T3-treated cells (Fig. 6D). Therefore, T3 directly repressed ATF-2 and CREB transcriptional activity, in agreement with our previous finding that TR antagonizes CRE-mediated transcription (25). Although T3 reduced activity of these transcription factors that are activated by Rsks, this was not observed with a GAL-ELK1 construct, a direct target of Erk, demonstrating again that antagonism by T3 is exerted downstream of this kinase.

FIG. 6.

CRE mediates stimulation by ras oncogene and repression by T3. (A) Schematic representation of the cyclin D1 promoter showing the position of binding sites for different transcription factors. In the lower panel, the effect of 36 h of incubation with 5 nM T3 on luciferase activity was determined in N2a-β cells cotransfected with Ha-ras^val12 and the indicated 5′ cyclin D1 promoter deletions. (B) The CRE at nucleotide −58 in the promoter was mutated in the plasmid extending to −269, and the reporter activity was determined in the native (−269) and mutated (−269CREm) constructs after transfection with Ha-ras^val12 and incubation in the presence or absence of T3. (C) CREB and ATF-2 bind the CRE of the cyclin D1 promoter. For the top panel, nuclear extracts from N2a-β cells were incubated in the absence (lane 2) or presence of antibodies for CREB, ATF-2, or a nonrelated protein (lanes 3, 4, and 5, respectively). Free probe (lane 1). For the bottom panel, EMSAs were performed with extracts from N2a-β cells transfected with 50 ng of Ha-ras^val12 or the corresponding empty vector. Cells were incubated for 16 h in the presence or absence of 5 nM T3. (D) Cells were transfected with an UAS reporter plasmid and 250 ng of the indicated GAL4 fusion constructs or the GAL4 DBD alone. These constructs were cotransfected with Ha-ras^val12 (50 ng) as indicated, and the luciferase activity was determined after 36 h in control and T3-treated cells.

TRs inhibit ras-mediated transformation.

Since TRs display a strong repressive activity of ras-mediated proliferation and transcriptional responses, we also explored the possibility that the receptor could inhibit ras-mediated cellular transformation. To prove this hypothesis, we performed focus formation assays with NIH 3T3 fibroblasts transfected with 50 ng of H-ras^val12 in the presence or absence of 4 μg of TRs. The transforming ability of the ras oncogene was greatly reduced in cells coexpressing TRs, although TRβ1 appears to have a stronger antitransforming effect than the α1 isoform (Fig. 7A). TRα1 reduced focus formation in the absence of exogenously added T3, and this reduction was stronger in T3-treated cells, inhibiting by >60% the number of transformation foci. On the other hand, expression of TRβ1 was sufficient to strongly reduce ras-mediated transformation even in the absence of exogenously added hormone, and transformation was essentially abolished in cells expressing the β1 isoform upon incubation with T3. In similar assays, v-src also induced fibroblast transformation, and focus formation was repressed in cells expressing TRs (data not shown).

FIG. 7.

TR represses Ha-ras^val12-induced transformation of NIH 3T3 fibroblasts. (A) Representative foci formation in fibroblasts transfected with 50 ng of Ha-ras^val12 and 4 μg of TRα1 or TRβ1. After transfection, cells were grown in the presence or absence of T3 for 14 days. In the right panel, the foci formed per dish were counted in three independent experiments. (B) Fibroblasts were transfected with 50 ng of Ha-ras^val12 and 0.5 μg of TRβ1 in the presence or absence of 0.5 μg of a cyclin D1 expression vector. The cells were grown in hormone-depleted medium for the same time period, and the number of foci was scored in the different groups in control and T3-treated cultures.

To analyze whether cyclin D1 levels can modulate the antitransforming activity of the receptor, the influence of overexpression of this protein on focus formation by cells transfected with 50 ng of H-ras^val12 in the presence or absence of 0.5 μg of TRβ1 was also examined (Fig. 7B). Incubation with T3 caused a 60% reduction in the number of transformation foci in cells expressing this amount of TRβ1, but this effect was lost in cells cotransfected with cyclin D1 that, however, did not affect focus formation by the oncogene. These data, together with those shown in Fig. 1C, indicate a key role of cyclin D1 in the antiproliferative and antitransforming effects mediated by the receptor.

TRs suppress in vivo tumorigenesis by the Ha-ras oncogene.

We next analyzed whether TRs also have tumor suppressor activity in vivo. For this purpose we prepared NIH 3T3 fibroblasts expressing in a stable manner oncogenic Ha-ras^val12 alone or in combination with TRα1 or TRβ1. These fibroblasts express similar levels of the oncoprotein, as assessed by Western blotting (Fig. 8A), and incubation with T3 did not alter these levels (Fig. 8B). On the other hand, cyclin D1 levels were not affected by T3 in cells expressing Ha-ras^val12 alone, but the hormone significantly reduced cyclin D1 in cells transfected with the oncogene in combination with TRα1 or TRβ1 (Fig. 8B). In addition, T3 did not increase TRE-dependent transcription in parental NIH 3T3 cells, but expression of both TR isoforms led to comparable ligand-dependent transactivation (Fig. 8C). The different transfectants were injected into the flanks of immunodeficient mice, and tumor growth was monitored for 2 months. Whereas large tumors developed in mice injected with fibroblasts expressing Ha-ras^val12 alone, no tumors were detected in mice injected with fibroblasts coexpressing the oncoprotein and either TRα1 or TRβ1 at 25 days postinjection (Fig. 9A). All injections of fibroblasts expressing Ha-ras^val12 alone gave rise to tumors with a short latency. In contrast, only 50% of injections of fibroblasts expressing oncogenic ras and TRα1 caused tumor growth, and this occurred with a substantially delayed appearance (Fig. 9B). Moreover, although all tumors were aggressive fibrosarcomas, histological analysis demonstrated that the presence of TRα1 conferred a relatively lower degree of tumor dedifferentiation, as shown by an increased presence of collagen and a more fusiform morphology (Fig. 9C). Consistent with the results obtained in focus formation assays, the β1 receptor isoform had even stronger antitumorigenic effects, since we did not observe tumor generation even after 2 months of inoculation with fibroblasts expressing both Ha-ras^val12 and TRβ1. The stronger antitransforming ability of TRβ1 does not appear to be due to a higher level of expression of this receptor isoform (see Fig. 8), demonstrating that TRβ1 is more potent than TRα1 in suppressing ras-mediated transformation, whereas both isoforms can mediate similar T3-dependent transcriptional stimulation.

FIG. 8.

Generation of NIH 3T3 cells expressing Ha-ras^val12 and TRs in a stable manner. (A) Ras, TRα, and TRβ proteins were measured by Western blotting in NIH 3T3 fibroblasts stably transfected with empty vector (Neo) or expression vectors for Ha-ras^val12, TRα1, and TRβ1 (B) NIH 3T3 cells expressing Ha-ras^val12 alone or in combination with TRα1 or TRβ1 were incubated in the presence or absence of 5 nM T3 for 36 h, and the levels of cyclin D1 and Ras were determined by Western blotting. (C) The indicated groups of NIH 3T3 cells were transfected with the plasmid containing the palindromic TRE, and the reporter activity was measured after 36 h of treatment with or without the hormone.

FIG. 9.

Antitumorigenic effects of TRs. (A) Tumor formation in athymic nude mice injected 25 days before into both flanks with the NIH 3T3 cells expressing in a stable manner Ha-ras^val12 alone or in combination with TRα1 or TRβ1, as indicated. (B) Tumorigenesis was monitored for 60 days both in euthyroid and in hypothyroid animals injected with the groups of cells indicated at the right. (C) Hematoxylin-eosin staining of fibrosarcomas formed in euthyroid mice injected with fibroblasts expressing Ha-ras^val12 (left panel), as well as in euthyroid (middle panel) and hypothyroid mice (right panel) inoculated with NIH 3T3 cells expressing both the oncogene and TRα1.

To analyze the effect of thyroidal status on tumor growth, nude mice were treated with antithyroidal drugs (Fig. 9B). All hypothyroid animals injected with ras-expressing fibroblasts developed tumors, although, unexpectedly, the lag time for tumor formation was somewhat increased. Expression of TRα1 produced a further delay in tumor appearance; however, in contrast to results obtained in euthyroid animals, 100% of the injections generated tumors in hypothyroid mice. This is in agreement with the finding that TRα1 caused a stronger inhibition of transformation by ras in T3-treated cells. Tumors induced by Ha-ras^val12- and TRα1-expressing fibroblasts in hypothyroid animals had a morphology similar to that obtained in euthyroid animals expressing ras oncogene alone (Fig. 9C). In addition, TRβ1 was able to totally suppress ras-mediated tumorigenesis in hypothyroid animals (Fig. 9B). This is again consistent with the effects of this receptor isoform on ras-mediated transformation shown in Fig. 7 and indicates again the stronger tumor suppressor activity of TRβ1 compared to TRα1 that can occur even at subphysiological concentrations of T3.

DISCUSSION

Although increasing evidence suggests that aberrant expression and mutations in TR genes could be associated with carcinogenesis (10), the role of these receptors in tumor generation or progression is unclear. We provide evidence here that TRs are strong suppressors of the actions of the ras oncogene. The thyroid hormone antagonizes ras-mediated transcriptional responses in TR-expressing cells. This regulation is observed in different cell types and occurs with promoters of several genes that play an important role in cell proliferation. In agreement with previous biochemical evidence of signaling between ras oncogene and cyclin D1 (29, 47), we demonstrate here that activated Ras proteins induces cyclin D1 expression and that T3 blocks this induction. Repression of cyclin D1 levels by T3 appears to be an important component of the mechanism by which the hormone blocks Ras-dependent proliferation. Thus, expression of oncogenic ras increases the number of cells that leave G₁ and progress through the cell cycle, and T3 blocks this response. However, repression by T3 is reversed to a significant extent when cyclin D1 is overexpressed. In addition to cyclin D1, the CKIs p21^Cip and p27^Kip may also be targets for T3-dependent repression of proliferation. The hormone increases the levels of these proteins and, although the response of the p21^Cip and p27^Kip promoters to ras is repressed, the levels of these proteins are high in T3-treated cells.

Transcriptional activation of the cyclin D1 gene by mitogenic signals and oncogenes such as ras and v-src can be mediated by multiple cis elements, including AP-1 (1), Sp-1 (26), and CRE (4, 19) sites. Our results show that the CRE is the main acceptor for cyclin D1 induction by oncogenic ras in neuroblastoma cells. This element binds constitutively the b-Zip transcription factors CREB and ATF-2, and we have shown that their transcriptional activity is markedly enhanced upon expression of Ha-ras^val12. These transcription factors contain a kinase-inducible domain necessary for activation in response to external stimuli. A serine residue in this domain can be phosphorylated in response to multiple kinases, which are activated in response to different signaling pathways (41). One of the main pathways stimulated by Ras is the Raf/Erk pathway. Although these b-Zip factors are not directly phosphorylated by Erk1/2, they are phosphorylated and transactivated by the Erk1/2-activated Rsk (15, 48). In agreement with previous results obtained in a different cell type (19), our results show that this is the main pathway used by the oncogene to increase cyclin D1 transcription in neuroblastoma cells.

T3 represses expression of the cyclin D1 gene in response to ras oncogene through promoter sequences that do not contain a TRE by interference with the activity of the Ras/Erk/Rsk pathway. Src also stimulates transcription of this gene in a Ras-dependent manner, and T3 also antagonizes the transcriptional response to this oncoprotein. In transient-transfection assays in neuroblastoma cells the hormone blocks almost totally the induction by oncogenic ras and also causes a weaker inhibitory effect on basal cyclin D1 promoter activity. This activity appears to reflect a certain level of stimulation of the Ras pathway, as demonstrated by the finding that a MEK inhibitor represses significantly basal reporter levels, and the reduction by T3 most likely reflects the antagonistic action on this endogenous Ras activity. In contrast, T3 represses the response to oncogenic ras without affecting basal cyclin D1 promoter levels in other cell type, such as pituitary GH4C1 cells. In these cells, where T3 did not decrease basal levels, incubation with the inhibitor did not affect basal cyclin D1 promoter activity (not illustrated). These findings demonstrate that TRs antagonize specifically ras-dependent transcriptional responses.

Antagonism of the Ras pathway by T3 is exerted downstream of Erk and appears to involve inhibition of Rsk2 kinase activity. Indeed, we have found that T3 represses ras-induced transcriptional activity of the Rsks targets CREB and ATF-2 that bind the CRE. This is in agreement with our recent demonstration that TRs antagonize CRE-mediated transcription without binding to this motif. A direct interaction of TR with CREB (25), as well as the inhibition of Rsk activity observed here, appears to underlie repression of the transcriptional responses to oncogenic ras. Rsks have not only transcription factors of the CREB family but also histones and transcriptional coactivators as downstream targets (27, 40) and can then deeply influence chromatin structure and transcription of different genes crucial for cell proliferation and transformation. The mechanism by which T3 represses activation of Rsk2 by oncogenic ras is currently unknown, but the possibility that could be a consequence of inhibition of Erk activity that is immediately upstream can be dismissed because T3 did not significantly reduce stimulation of this kinase by the oncogene. We have previously demonstrated that the interaction of TR with CREB strongly blocks the ability of protein kinase A to phosphorylate CREB (25). It is not unlikely that phosphorylation of CREB or other b-Zip factors by other kinases such as Rsk2 could be also affected. In contrast, we have discarded the possibility of the existence of a direct interaction between TRs and Rsk2 that could affect Rsk2 activity (unpublished results).

Transient-transfection assays with the cyclin D1 promoter in cells devoid of TRs prove that both TRα1 and TRβ1 are similarly effective in mediating repression of ras-mediated transcriptional responses by the hormone. In addition, we have observed that in neuroblastoma cells the unoccupied receptors can cause a significant induction of promoter activity that is reversed by T3. This paradoxical ligand-independent activation is a common finding in genes regulated negatively by T3, although the mechanisms responsible for this regulation are not yet understood (2). Stimulation by the unoccupied receptor is not observed in other cell types, where a strong repression by T3 of ras-dependent stimulation was also observed. Therefore, T3 appears to repress specifically the response to ras independently of basal reporter levels in different cell types.

Antagonism between TR and ras oncogene is not a reciprocal phenomenon because expression of Ha-ras^val12 enhances instead of reducing TR-dependent transactivation. This effect could be mediated by phosphorylation of TR (6) or coregulators (14, 37), which are targets of MAPKs. Independently of the mechanism responsible for this action of the oncogene, our results suggest the existence of a regulatory loop in which oncogenic ras would strengthen transcriptional responses to thyroid hormones, one of them being the inhibition of the effects of the oncogene.

TRs not only reduce transcriptional responses to ras and antagonize ras-induced proliferation but, more importantly, they are able to block cellular transformation by this oncogene as well as by v-src. Repression of cyclin D1 expression also appears to play an important role in the antitransforming effect of these receptors, since overexpression of this protein blocks T3-dependent reduction of the number of transformation foci produced by oncogenic ras in NIH 3T3 fibroblasts.

The antitransforming activity of these receptors agrees with the properties of the retroviral v-erbA oncogene. TRα is the cellular counterpart of v-ErbA, a mutated oncoprotein that no longer binds hormone and acts as a constitutive repressor of T3-regulated gene expression (39). v-ErbA potentiates transformation by oncoproteins derived from tyrosine kinases that constitutively activate Ras (45) and could have dominant-negative effects on the antitransforming actions of the native TRs. Furthermore, studies with mice expressing a dominant-negative TRβ mutant present in some patients with thyroid hormone resistance syndrome spontaneously develop metastastic thyroid carcinoma (44), suggesting again an important role of TRs in suppressing tumorigenesis.

The present study also shows that both TRα1 and TRβ1 can play a role as tumor suppressors in vivo. Although both receptor isoforms mediate a similar T3-dependent transactivation in transient-transfection assays, our results show that the β-isoform appears to have stronger antitumorigenic effects in vivo. This suggests that TRs could use distinct mechanisms to control those processes. Expression of TRβ1 abolishes totally tumor formation by ras-transformed cells in nude mice, even under hypothyroid conditions. These results are also consistent with those found in foci formation assays in which TRβ1 can block to a significant extent fibroblast transformation by ras in the absence of exogenously added ligand. On the other hand, tumor formation is reduced in euthyroid mice inoculated with cells expressing ras oncogene and TRα1, but all hypothyroid animals develop tumors, although tumor appearance is significantly delayed. These results are consistent with the ligand-dependent antitransforming activity of this receptor isoform in focus formation assays and indicate that TRα1 could only have tumor suppressor activity under euthyroid conditions.

In contrast with the increasing evidence that inactivation of TRs by mutation or by promoter methylation is a common event in cancer (16, 20, 21, 42), there is not a clear relationship of human neoplasias with thyroid diseases or circulating thyroid hormone levels. Our finding that expression of TRβ1 is sufficient to prevent the transforming effects of oncogenic ras even in hypothyroid animals could explain why a higher incidence of ras-dependent tumors is not generally found in hypothyroid patients. Furthermore, we have found that thyroidal status has some effect on tumorigenesis by ras-transformed fibroblasts in the absence of ectopic TR expression since, paradoxically, tumor development is slightly retarded in hypothyroid animals. The existence of low levels of TRs in these cells that are insufficient to confer ligand-dependent transcriptional stimulation suggests that the metabolic changes associated with hypothyroidism could also influence tumor growth.

It is important to point out that although many mammalian tissues express TRs, except for a few exceptions no models of transformed or immortalized cells expressing high levels of TRs exist. Our findings, together with the alterations of TRs found in cancer, suggest that loss of expression and/or function of these receptors could result in a selective advantage for cell transformation and tumor development. It is clear that further studies are needed to establish the role of these receptors in human cancer, and our observations open the interesting possibility of reexpression of these nuclear receptors as a novel therapeutic strategy.