Thyroid hormone receptors mutated in liver cancer function as distorted antimorphs

Abstract

Aberrant thyroid hormone receptors (TRs) are found in over 70% of the human hepatocellular carcinomas (HCCs) analysed. To better understand the role(s) of these TR mutants in this neoplasia, we analysed a panel of HCC mutant receptors for their molecular properties. Virtually all HCC-associated TR mutants tested retained the ability to repress target genes in the absence of T3, yet were impaired in T3-driven gene activation and functioned as dominant-negative inhibitors of wild-type TR activity. Intriguingly, the HCC TRα1 mutants exerted dominant-negative interference at all T3 concentrations tested, whereas the HCC TRβ1 mutants were dominant-negatives only at low and intermediate T3 concentrations, reverting to transcriptional activators at higher hormone levels. The relative affinity for the SMRT versus N-CoR corepressors was detectably altered for several of the HCC mutant TRs, suggesting changes in corepressor preference and recruitment compared to wild type. Several of the TRα HCC mutations also altered the DNA recognition properties of the encoded receptors, indicating that these HCC TR mutants may regulate a distinct set of target genes from those regulated by wild-type TRs. Finally, whereas wild-type TRs interfere with c-Jun/AP-1 function in a T3-dependent fashion and suppress anchorage-independent growth when ectopically expressed in HepG2 cells, at least certain of the HCC mutants did not exert these inhibitory properties. These alterations in transcriptional regulation and DNA recognition appear likely to contribute to oncogenesis by reprogramming the differentiation and proliferative properties of the hepatocytes in which the mutant TRs are expressed.

Introduction

Thyroid hormone receptors (TRs) play multiple roles in normal differentiation, proliferation, and homeostasis, whereas aberrant TR activity results in endocrine and neoplastic disease (Tenbaum and Baniahmad, 1997; Apriletti et al., 1998; Zhang and Lazar, 2000; Ando et al., 2001; Winter and Signorino, 2001; Cheng, 2003; Gonzalez-Sancho et al., 2003; Yen and Cheng, 2003). Wild-type TRs operate as hormone-regulated transcription factors that bind to specific DNA sequences (denoted thyroid response elements, or TREs) and regulate transcription of adjacent target genes (Mangelsdorf et al., 1995; Glass, 1996; Apriletti et al., 1998; Ribeiro et al., 1998; Beato and Klug, 2000; Zhang and Lazar, 2000). TRs bind to TREs, recruit corepressors (such as N-CoR and SMRT), and typically repress transcription in the absence of hormone; conversely, the binding of T3 hormone induces a conformational change in TRs that results in corepressor release, the recruitment of coactivators (such as SRC-1), and transcriptional activation (Glass, 1996; Horwitz et al., 1996; Koenig, 1998; Ordentlich et al., 2001; Lee and Kang, 2002; Privalsky, 2004). TRs can also regulate target genes indirectly through protein–protein contacts with other transcription factors. TRs, for example, enhance AP-1 activity in the absence of T3, and suppress AP-1 function in the presence of T3 (Desbois et al., 1991; Zhang et al., 1991; Sharif and Privalsky, 1992; Pfahl, 1993; Schmidt et al., 1993; Moehren et al., 2004). TRs are encoded at two distinct loci, α and β, and by alternative mRNA splicing so as to generate three major isoforms: TRα1, TRβ1, and TRβ2 (Lazar, 1993; Murata, 1998; Zhang and Lazar, 2000).

Inherited defects in TR function result in a human endocrine disorder, denoted Resistance to Thyroid Hormone (RTH) Syndrome (Refetoff, 1993; DeGroot, 1996; Kopp et al., 1996; Chaterjee, 1997; Nagaya and Seo, 1998). RTH Syndrome is associated with mutations at the TRβ locus that disrupt the triggering mechanism by which hormone binding induces corepressor release and coactivator acquisition (Yoh et al., 1997; Liu et al., 1998; Nagaya et al., 1998; Safer et al., 1998; Tagami and Jameson, 1998; Matsushita et al., 2000; Yoh and Privalsky, 2000). RTH-mutant TRs remain inappropriately bound to corepressors at physiological T3 concentrations and operate as dominant-negative inhibitors of wild-type TR function. The resulting defects in T3 sensing produce a variety of endocrine defects and, rarely, pituitary neoplasias, but do not typically manifest as overt peripheral neoplasia (Ando et al., 2001; Kamiya et al., 2003). Conversely, somatic mutations in human TRs are found at high frequency in human hepatocellular carcinoma (HCC), renal clear cell cancer (RCCC), and certain thyroid and gastric neoplasia (Lin et al., 1997b, 1999, 2001; Kamiya et al., 2002; Puzianowska-Kuznicka et al., 2002; Wang et al., 2002). Whereas RTH-Syndrome mutations map almost exclusively to the TRβ locus, mutations in both TRα and TRβ are detected in human HCC and RCCC (Lin et al., 1997b, 1999, 2001; Kamiya et al., 2002). The precise contributions of these mutant TRs to oncogenesis remain incompletely understood, but their high frequency in these neoplasias (65% HCC samples displayed TRα1 mutations and 76% displayed TRβ1 mutations, with both loci mutated in some samples) is highly suggestive of a causal or contributory role. Strongly supporting this hypothesis, an additional mutant TRα allele, first identified as the v-erb A oncogene in an avian retrovirus, can cause HCC when expressed as a transgene in mice (Barlow et al., 1994).

We wished to better understand the molecular events that distinguish the TR mutants associated with neoplasias such as HCC from the endocrine-disruptive TR mutants associated with RTH Syndrome. We report here an analysis of a panel of TR mutants isolated from human HCC samples. Several common features were observed. Virtually all HCC-associated TR mutants tested retained the ability to repress TRE target genes in the absence of T3, yet were impaired in T3-driven gene activation and functioned as dominant-negative inhibitors of wild-type TR activity. Intriguingly, the HCC TRα1 mutants exerted dominant-negative interference at all the T3 concentrations tested, whereas the HCC TRβ1 mutants inhibited wild-type function at only low T3 concentrations, reverting to transcriptional activators and augmenting wild-type function at higher hormone levels. Conversely, whereas wild-type TRs interfere with AP-1 function in a T3-dependent fashion, the HCC mutations were impaired in these inhibitory properties. Notably, several of the TRα and TRβ HCC mutations also altered the DNA recognition properties of the encoded receptors, suggesting that these HCC TR mutants may recognize a distinct set of target genes from those regulated by wild-type or RTH-Syndrome TRs. Supporting a biological consequence for these changes in TR function, wild-type TRα suppressed anchorage-independent growth when expressed in HepG2 cells, whereas a HCC mutant did not. Our results suggest that the TR mutations associated with HCC are likely to contribute to oncogenesis by reprogramming the differentiation and proliferative properties of the hepatocytes in which they are expressed.

Results

TRβ1 HCC mutants repress gene expression in the absence of hormone, but are defective in T3-induced transcriptional activation

We first examined a panel of TRβ1 HCC mutants (Figure 1a) for the ability to regulate a prototypic TRE reporter (a DR4-TK-luciferase construct) when introduced into cultured cells (Figure 1b). CV-1 cells lack detectable endogenous TR activity, and the DR4-TK-luciferase reporter displayed little or no response to T3 in the absence of ectopic TR expression (Figure 1b, ‘empty pSG5’ control). Introduction of wild-type TRβ1 into these cells repressed reporter expression in the absence of T3, and activated reporter expression in the presence of T3 (Figure 1b; basal reporter activity in the absence of TRs was defined as 1). Half-maximal reporter activation was observed at approximately 1−2 nm hormone. Deleting the TRE element from the TK-luciferase reporter eliminated the T3 response (data not shown). Notably, three different TRβ1 mutants derived from HCCs (denoted TRβ1- I, J, and N) retained the ability to repress in the absence of T3, but were defective in their ability to activate the DR4-TK-luciferase reporter in response to T3. As a result, these HCC mutants strongly repressed DR4-TK-luciferase expression over a range of intermediate T3 concentrations sufficient to induce reporter gene activation by the wild-type TRβ (Figure 1b). Significantly, however, all three β1 mutants recovered activity and strongly induced the DR4-TK-luciferase reporter at the highest T3 concentrations (Figure 1b). We conclude that the TRβ1 HCC mutants are specifically defective in the hormone-driven conversion from transcriptional repression to transcriptional activation.

Figure 1.

TR mutants associated with HCC display altered transcriptional properties. (a) Schematic of wild-type and mutant TRs. A representation of TRα1 and TRβ1, and the mutants under study, is shown. The locations of the DNA-binding and hormone-ligand-binding domains (DBD and LBD) are indicated, as are the sites of the genetic lesions found in each mutant allele. (b) Altered transcriptional regulatory properties of TRβ1 mutants. An empty pSG5 vector (dotted line), pSG5 expressing wild-type TRβ1 (solid lines), or pSG5 expressing the HCC mutant TRs indicated (dashed lines) were introduced into CV-1 cells together with a DR4-TK-luciferase reporter and a pCH110 β-galactosidase construct (employed as an internal normalization control). The cells were subsequently treated with T3, or not, as indicated, harvested 24 h later, and the luciferase activity, relative to β-galactosidase activity, was determined. ‘Fold activation’ represents the relative luciferase units observed in the presence of the given TR divided by the relative luciferase activity observed in its absence (i.e. the empty pSG5 vector). A fold activation of one therefore indicates basal level transcription. Results are also presented for an empty pSG5 vector control (dotted line). Average and s.e.m. values are shown from at least three independent experiments; wild-type receptor results are re-iterated in each panel for comparison.

The aberrant transcriptional properties of the TRβ1 HCC mutants reflect changes in their affinity and specificity for corepressors, and in the ability of T3 to drive corepressor release and coactivator acquisition

We examined if the transcriptional defects observed for the TRβ1 HCC mutants were because of alterations in corepressor or coactivator interaction. As expected, wild-type TRβ1 bound to a GST-N-CoR corepressor construct in the absence of hormone and was released by addition of T3, with half-maximal release observed at approximately 3 nm hormone (Figure 2a, solid line); little or no receptor binding was observed to a non-recombinant GST construct employed as a negative control (data not shown). Intriguingly, all three HCC TRβ1 mutants displayed altered interactions with N-CoR. The TRβ1-J and TRβ1-N mutants interacted much more strongly with N-CoR in the absence of hormone than did wild-type TRβ1, whereas the TRβ1-I mutant interacted more weakly than wild type under the same conditions (Figure 2a compare each mutant to the wild type at zero hormone). In addition, all three mutants required much higher than normal T3 levels to be released from the corepressor than did wild-type TRβ1 (Figure 2a).

Figure 2.

TRβ1 mutants display defective coregulator interactions and defective T3 binding. (a) Altered interaction and impaired release of N-CoR by the TRβ1 mutants. Radiolabeled mutant or wild-type TRβ1s were incubated with an immobilized GST-N-CoR construct at the T3 concentrations indicated. After washing, the percentage of receptor bound to GST-N-CoR, relative to input was determined. The average and range of two experiments are shown. Data for wild-type TRβ1 is re-iterated in each panel for comparison. (b) Altered interaction and impaired release of SMRT by the TRβ1 mutants. The protein–protein interaction assay was performed as described in panel a, but using a GST-SMRT construct. The average and s.e.m. values of at least three or more independent experiments are shown. Data for wild-type TRβ1 is re-iterated in each panel for comparison. (c) Impaired T3-dependent acquisition of SRC by the TRβ1 mutants. The protein–protein interaction assay was performed as described in panel a, but using a GST-SRC-1 coactivator construct. The average and s.e.m. values of at least three independent experiments are shown. Data for wild-type TRβ1 is re-iterated in each panel for comparison. (d). Reduced T3 binding by TRβ1 mutants. T3 binding by wild-type and mutant TRβ1s was determined using a hormone-mediated protease resistance assay. The percentage of TR resistant to elastase over a range of T3 concentrations (relative to input) was assayed; the average and s.e.m. values of three independent experiments are presented. Data for wild-type TRβ1 is re-iterated in each panel for comparison.

N-CoR shares extensive sequence and functional overlap with a second corepressor, SMRT, although the affinities of SMRT and N-CoR differ for different nuclear receptors (Privalsky, 2004). We determined if the HCC mutations altered the relative affinity of TRβ1 for SMRT versus N-CoR. As shown in prior studies (Zamir et al., 1997; Goodson et al., 2005), wild-type TRβ1 interacted more weakly with GST-SMRT than with GST-N-CoR in the absence of hormone, but released from both corepressors at comparable T3 concentrations (Figure 2b, solid line). Interestingly, the affinities of the TRβ1 HCC mutants for SMRT versus N-CoR differed. Although the TRβ1-N and -J mutants bound to N-CoR more efficiently than did wild-type TRβ1, these same mutants bound to SMRT less well than did wild-type receptor (compare middle and bottom panels of Figure 2b with those of Figure 2a). Conversely, the TRβ1-I mutant, which bound N-CoR relatively weakly, bound to SMRT at wild-type levels (compare top panels of Figure 2b and a). As was seen with N-CoR, all three of the TRβ1 HCC mutants required significantly higher levels of T3 to dissociate from SMRT than did wild-type TRβ1 (Figure 2b).

Binding of T3 to TRs not only releases corepressors but also recruits coactivators (Glass and Rosenfeld, 2000; Privalsky, 2004). Repeating our protocol with a GST-SRC-1 coactivator construct, we observed the behavior expected for wild-type TRβ1: a low basal level of binding of the receptor to GST-SRC-1 minus hormone that increased substantially in the presence of T3 (Figure 2c, solid line). Half-maximal binding of wild-type TRβ1 was observed at approximately 1−2 nm hormone. Significantly, each of the mutants interacted more weakly with GST-SRC-1 in the absence of T3 than wild-type, and required substantially more T3 than did wild-type to induce half-maximal binding (Figure 2c, dashed lines). Nonetheless, when exposed to saturating T3 concentrations, all three HCC TRβ1 mutants were able to bind to SRC-1 at levels comparable to that seen for wild-type TRβ1 (Figure 2c).

The ability of supraphysiological T3 concentrations to induce exchange of corepressor for coactivator, and transcriptional activation, by the TRβ1 HCC mutants suggested that their primary defect might be in hormone-binding affinity. We employed a protease protection assay by which T3 binding is measured through its ability to generate a protease-resistant conformation in the receptor (Leng et al., 1993; Lin et al., 1997a; Ikeda et al., 1996). This methodology has been validated previously, and although not a true equilibrium binding assay, provides a reproducible measure of relative hormone avidity sufficient to show differences between different receptor mutants. In the absence of T3, both wild-type TRβ1 and the HCC TRβ1 mutants were extensively degraded by protease (Figure 2d and data not shown). Incubation of wild-type TRβ1 in increasing concentrations of T3 resulted in formation of a protease-resistant polypeptide, with half-maximal protection obtained at approximately 3 nm (Figure 2d, solid line). Notably, all of the HCC TRβ1 mutants required significantly higher than wild-type T3 concentrations to acquire comparable protease resistance (Figure 2d, dashed lines). We conclude that the primary molecular defect in these HCC TRβ1 mutants is likely to be an impairment in hormone binding, which in turn manifests as defects in T3-mediated corepressor release, coactivator acquisition, and transcriptional activation.

The TRβ1 HCC mutants are dominant-negative inhibitors of wild-type receptor function

The impaired hormone binding, corepressor release, and coactivator acquisition by the HCC mutants resembled the TR defects seen with RTH Syndrome (Yoh et al., 1997; Liu et al., 1998; Nagaya et al., 1998; Safer et al., 1998; Tagami and Jameson, 1998; Matsushita et al., 2000; Yoh and Privalsky, 2000). We next tested if our HCC mutants also exhibited dominant-negative properties when co-expressed with wild-type TR (Figure 3). Reporter gene activation by the wild-type receptor was significantly blunted by co-introduction of the TRβ1 mutants at low to intermediate T3 concentrations (e.g. 1−10 nm), whereas wild-type receptor function was restored at 100−1000 nm T3 (Figure 3, compare the T3 response of the reporter in the presence of wild-type TRβ1 only, solid line, to the response in the presence of both wild-type and HCC-mutant receptors, dashed lines). Results shown are for 5:1 ratios of mutant to wild-type receptor; only weak or undetectable dominant-negative effects were observed at 1:1 ratios (data not shown). The dominant-negative activity was specific to the TR mutants, and not the result of ‘transcriptional squelching’; co-introduction of estrogen receptor-α had relatively little effect on wild-type TR activity (Figure 3). We conclude that all three HCC TRβ1 mutants can function as dominant-negative inhibitors of wild-type TR function over a defined range of T3 concentrations, but this is reversed at levels of T3 sufficient to overcome their impaired T3-binding affinity.

Figure 3.

TRβ1 mutants are dominant-negative inhibitors of wild-type TRβ1 function. The same protocol as in Figure 1 was repeated, except each well was cotransfected with both 5 ng of wild-type receptor and 25 ng of the mutant receptor indicated (dashed lines). Average and s.e.m. values are shown from at least three independent experiments. The corresponding fold activation for the wild-type TR when cotransfected with 25 ng of an empty pSG5 vector is re-iterated in each panel for comparison purposes (solid line).

TRα1 mutants are also defective in transcriptional activation and function as dominant negatives, but through mechanisms sharply distinct from those observed for the TRβ1 mutants

We extended our studies to wild-type TRα1 and to two HCC mutant versions of this isoform, TRα1-I and TRα1-M. When introduced into CV-1 cells, wild-type TRα1 repressed expression of the DR4-TK-luciferase reporter in the absence of hormone, and activated it in the presence of T3 (Figure 4a, solid line). In superficial agreement with the HCC TRβ1 mutants, both TRα1-I and TRα1-M repressed reporter expression in the absence of hormone, and were impaired in the ability to activate in response to T3 (Figure 4a, dashed lines). Closer inspection, however, revealed that the nature of the transcriptional defect was distinct for the TRα1 and TRβ1 HCC mutants. The TRα1-I mutant failed completely to respond to T3 and functioned as a constitutive repressor at all hormone concentrations tested (Figure 4a, top panel); the TRα1-M mutant repressed in the absence of hormone, but was only weakly induced as a transcriptional activator at wild-type T3 concentrations and was not further induced at still higher T3 levels (Figure 4a, bottom panel).

Figure 4.

TRα1 mutants exhibit altered transcriptional activation and dominant-negative function. (a). Impaired transcriptional activation by TRα1 mutants. Experimental procedures and data analysis were performed as described in Figure 1 using TRα1 wild-type and mutant receptors in place of TRβ1. (b). Dominant-negative activity by TRα1 mutants. Experimental procedures and data analysis were as in Figure 3, using TRα1 wild type and mutant receptors in place of TRβ1.

We investigated the molecular basis for these defects in transcriptional activation by the TRα1 mutants. Both the TRα1-I and TRα1-M mutants bound at wild-type levels to N-CoR and SMRT in GST-pulldown experiments in the absence of T3 (Figure 5a and b, compare solid to dashed lines). Both HCC TRα1 mutants released from these corepressors in response to T3, although the TRα1-I mutant required higher than wild-type levels of T3 to do so (Figure 5a and b). Reciprocally, both mutants bound to the SRC-1 coactivator in a hormone-dependent manner, with the curve for the TRα1-I mutant again displaced to somewhat higher T3 levels than that of wild-type TRα1 (Figure 5c, compare solid and dashed lines). The ability of these receptor mutants to bind to T3 in the protease-resistance assay was normal or near normal (Figure 5d). These hormone-binding and coregulator exchange properties of the HCC TRα1 mutants in vitro therefore contrasts with their hormone-independent defects in transcriptional activation in transfected cells.

Figure 5.

Altered transcriptional properties of TRα1 mutants do not parallel their coregulator and T3 binding properties in vitro. (a) Binding and T3-dependent release of N-CoR. Experimental procedures and data analysis were as in Figure 2a, except for the use of TRα1 wild type and mutants. The average and s.e.m. values of at least three independent experiments are shown. (b) Binding and T3-dependent release of SMRT. Experimental procedures and data analysis were as in Figure 2b, except for the use of TRα1 wild type and mutants. (c) T3-dependent binding of SRC-1. Experimental procedures and data analysis were as in Figure 2c, except for the use of TRα1 wild type and mutants. (d). T3 binding by wild-type and mutant TRα1s. Experimental procedures and data analysis were as in Figure 2d, except for the use of TRα1 wild type and mutants. Data for wild-type TRα1 is re-iterated in each panel to simplify the relevant comparisons.

We also examined the ability of the TRα1 mutants to function as dominant negatives. Consistent with their impaired activation properties when introduced individually, both HCC TRα1 mutants strongly attenuated reporter gene activation when co-introduced with wild-type TRα1 (Figure 4b, compare solid to dashed lines). Unlike the HCC TRβ1 mutants, however, the dominant-negative effects of the TRα1 mutants spanned all hormone concentrations, extending to the highest nontoxic levels of T3 that could be achieved (Figure 4b). Collectively, these results establish that both the TRα1 and TRβ1 HCC mutants can interfere in a dominant-negative fashion with the corresponding wild-type TRs, yet the molecular basis behind and hormone-dependent manifestation of this interference differs for the two different isoforms.

The TR HCC mutants are defective in regulation of AP-1-mediated transcription

Nuclear receptors can also modulate gene expression indirectly through protein–protein contacts with non-receptor transcription factors. TRs are known to enhance the function of the AP-1 protein c-Jun in the absence of hormone and repress in the presence of T3 (Desbois et al., 1991; Zhang et al., 1991; Sharif and Privalsky, 1992; Pfahl, 1993; Schmidt et al., 1993; Moehren et al., 2004). Given the important role of AP-1 in the control of cell proliferation, we examined whether this TR-mediated regulation of AP-1 might be altered or lost in the HCC TR mutants. In the absence of T3, both wild-type TRα1 and wild-type TRβ1 further activated expression of an AP-1 reporter gene greater than introduction of c-Jun alone (Figure 6a and b, solid lines, and data not shown). Addition of T3 reversed this activation into a repression, with the transition occurring at approximately 1 nm T3 (Figure 6a and b, solid lines). All three TRβ1 mutants shared the ability to enhance AP-1 activation in the absence of T3 to levels near equal to that of wild-type TRβ1; however, all three TRβ1 mutants required substantially higher than normal concentrations of T3 to reverse this activation into repression (Figure 6a, dashed lines). All three TRβ1 mutants also blunted the suppressive effects of T3 on AP-1 when co-introduced with wild-type TRβ1 (data not shown). In contrast, the HCC TRα1 mutants were defective both in the ability to enhance AP-1 function in the absence of hormone and in the ability to suppress in the presence of hormone (Figure 6b, dashed lines). TRα1-I displayed little or no activity on AP-1 function in either the absence or presence of T3, whereas TRα1-M displayed a weak enhancement of AP-1 function in the absence of T3, which was suppressed in the presence of hormone (Figure 6b). We conclude that the TRβ1 mutants require higher than normal T3 concentrations to switch from positive regulators of AP-1 function to suppressors, whereas the TRα1 mutants are impaired in both positive and negative AP-1 regulation through a mechanism distinct from changes in affinity for T3 hormone.

Figure 6.

Both TRβ1 and TRα1 mutants demonstrate defective inhibition of c-Jun/AP-1 transcriptional activity. (a) Increased T3 is required for inhibition of AP-1 by TRβ1 mutants. CV-1 cells were transfected with expression vectors for the various TR alleles and for c-Jun, together with a c-Jun responsive AP1-TK-luciferase reporter and the pCH110 internal normalization control, were treated for 24 h with the T3 concentrations indicated, were harvested, and the relative luciferase levels were determined as for Figure 1. A fold activation of one, denoted with an arrowhead, indicates basal level transcription. Average and s.e.m. values are shown from at least three independent experiments. Data for the corresponding wild-type TR is re-iterated in each panel for comparison. (b) Defects in c-Jun enhancement and suppression by the TRα1 mutants. Experimental procedures and data analysis were performed as described in panel a, except for the use of TRα1 wild type and mutants.

The DNA-binding properties of the HCC-TR mutants are distorted from those of wild type

Many HCC TR mutants have sustained multiple amino-acid substitutions, including changes in regions known to play a role in DNA recognition. TRα1-I and TRα1-M are both mutated at lysine 74, which in the wild-type receptor makes discriminatory contacts with the first guanine in the AGGTCA consensus half-site (Luisi et al., 1991; Rastinejad et al., 1995; Starr et al., 1996). We examined these mutants, as well as those in TRβ1, for possible alterations in their DNA recognition properties, by using an electrophoretic mobility shift assay (EMSA) and a panel of different TREs. We assayed both receptor homodimers, and heterodimers with RXR (RXRs can heterodimerize with and enhance TR binding to certain DNA sites). Wild-type TRβ1 and the HCC TRβ1-I and TRβ1-J mutants bound to TREs containing AGGTCA consensus half-sites efficiently, whether these half-sites were displayed in a DR4, DR5, or INV0 configuration, and whether these receptors were tested as homodimers or as heterodimers with RXR (Figure 7a). The TRβ1-N mutant was slightly reduced in its ability to bind to DNA in the same contexts (Figure 7a). The four TRβ1 receptors tested, mutant or wild type, displayed a reduced ability to bind to DNAs bearing substitutions at the ‘G’ in the second position of the prototypic half-site (Figure 7a).

Figure 7.

TR mutants display alterations in DNA binding and in anchorage-independent growth. (a). Binding of TRβ1 wild type and mutants to DNA as homodimers and heterodimers. The relevant wild-type and mutant TRs were incubated, alone or with RXRα, together with the radiolabeled DNA oligonucleotide probes indicated; the resulting DNA/receptor complexes were resolved by non-denaturing gel electrophoresis and were visualized by phosphor imager analysis. The identity of each DNA/protein complexes is indicated to the left or right of each panel. A quantification relative to wild-type TR/RXR heterodimers (defined as 100) for each probe is provided at the base of each lane. A representative experiment is presented; comparable results were obtained in repeated experiments. (b) Binding of TRα1 wild type and mutants to DNA as homodimers and heterodimers. Experimental procedures and data analysis were otherwise as performed as described in panel a. (c) Suppression of anchorage-independent growth by wild type, but not mutant TRα1. HepG2 cells stably transformed by wild-type TRα1, the HCC-TRα1-I mutant, or by an empty pCI-neo vector were tested for colony formation in soft agar; two independent transformant lines (denoted 1 and 2) were tested, each in duplicate (average and range are indicated).

The HCC TRα1 mutants yielded different results. Whereas wild-type TRα1 efficiently bound to DNA containing an AGGTCA DR4 consensus element as a receptor homodimer, neither TRα1-I nor TRα1-M was able to do so (Figure 7b). Formation of RXR heterodimers further enhanced wild-type TRα1 binding to the AGGTCA DR4 element, and, intriguingly, conferred partial or substantial binding by TRα1-I and TRα1-M, respectively (Figure 7b). Conversely, the TRα-I mutant, and to a lesser extent the TRα1-M mutant, bound to an ACGTCA DR4 element better than did the wild-type TRα1; this was particularly evident for the TRα1/RXRα heterodimers but could also be observed for TRα1-I homodimers (Figure 7b). Differences in DNA recognition by wild-type and the two TRα1 mutants were also noted on DR4s bearing AAGTCA and ATGTCA elements (Figure 7b). Analogous differences in DNA recognition by these different TR mutants were observed using naturally occurring T3-response elements (data not shown). We conclude that, in addition to operating as dominant negatives on positive-acting TREs, the TRα1-M and -I mutants are also altered in their DNA recognition properties compared to wild-type TRα1.

Wild-type TRα inhibits anchorage-independent growth of HepG2 cells, whereas the TRα1-I mutant does not

We next determined if ectopic expression of these receptors altered the growth properties of HepG2 cells. HepG2 cells are derived from a human HCC, and form anchorage-independent colonies in soft agar. Introducing wild-type TRα into HepG2 reproducibly suppressed anchorage-independent colony formation relative to the empty expression vector (Figure 7c; two independent stably transformed cell lines were analysed for each). In contrast, HepG2 transformants ectopically expressing the HCC-TRα-I mutant (chosen because of its strong dominant-negative properties and altered DNA recognition) displayed little or no effect on soft agar colony formation (Figure 7c). Wild type and HCC-TRα-I were expressed in the stable transformants at comparable levels, and although the efficiency of soft agar colony formation varied between independent assays, the same overall results were obtained in multiple experiments (data not shown). These experiments were performed using non-stripped serum, and thus contained measurable levels of T3; further addition of T3 to 100 nm reduced the size of the soft agar colonies for both wild-type and TRα-I transformants, although the latter always displayed more colonies than the former (data not shown).

The TRβ1-E HCC mutant displays a unique mix of transcriptional and DNA recognition properties suggestive of neoplastic progression

A fourth HCC TRβ1 mutant we analysed, denoted TRβ1-E, exhibited a mix of properties. In common with the other three TRβ1 HCC mutants, the TRβ1-E mutant required higher than wild-type levels of T3 to release from N-CoR or SMRT corepressor, and to bind to SRC-1 coactivator in vitro (Figure 8a, b and c); this faulty coregulator exchange by TRβ1-E was partially overcome at high hormone concentrations (Figure 8). In marked contrast to the other receptor mutants tested, however, the TRβ1-E mutant displayed little or no ability to regulate, either up or down, the DR4-TK-luciferase reporter and exhibited only very weak dominant-negative activity when co-introduced together with the wild-type TRβ1 (Figure 8d and e). TRβ1-E possesses three amino-acid substitutions, one each in the N-terminal, DNA-binding, and hormone-binding domains. The amino-acid substitution in the TRβ1-E hormone-binding domain is likely responsible for the observed defect in cofactor exchange. Notably, the substitution in the TRβ1-E DNA-binding domain disrupts a highly conserved cysteine crucial for proper folding and function of this domain in other nuclear receptors (Luisi et al., 1991; Rastinejad et al., 1995). Consistent with the location of this Cys substitution, the TRβ1-E mutant displayed a complete loss of DNA binding in our EMSA assay on all TREs tested, minus or plus RXR (data not shown), and it is likely that the loss of target gene regulation in vivo by this mutant reflects this loss of DNA binding. The TRβ1-E mutant is also relatively inactive in the AP-1 regulation assay (data not shown). We speculate that TRβ1-E may represent a ‘genetic fossil’ that contributed to oncogenesis at some point in HCC genesis or progression, but which subsequently underwent the additional inactivating Cys mutation and became dysfunctional as an oncoprotein.

Figure 8.

The TRβ1-E mutant displays mixed transcriptional and molecular properties. (a) Impaired release of N-CoR by TRβ1-E. (b) Impaired release of SMRT by TRβ1-E. (c) Altered binding and release of SRC by TRβ1-E. (d) Loss of transcriptional regulation by TRβ1-E. (e) Weak dominant-negative ability of TRβ1-E 9. For all panels, general experimental procedures and data analysis were as in Figures 2 and 3.

Discussion

The mutant TRs found in human HCCs interfere with wild-type receptor function

Major risk factors for HCC include viral infection, aflatoxin exposure, and cirrhosis; however, the molecular basis of HCC development and progression is not fully understood and likely involves multiple mechanisms (Suriawinata and Xu, 2004). Mutations in TRs occur at extremely high prevalence in human HCC, suggestive of a role for these receptors in oncogenic causation and/or progression (Lin et al., 1997b, 1999, 2001). To examine the molecular consequences of these mutations on the encoded receptors, we characterized the transcriptional and DNA recognition properties of a panel of human HCC TR mutants. Notably, virtually all of the HCC TR mutants examined retained the wild-type receptor's ability to repress transcription in the absence of hormone, yet were defective in their ability to activate in response to T3 hormone. As a consequence, the HCC mutant receptors functioned as dominant-negative inhibitors, and attenuated target gene activation when co-expressed with a corresponding wild-type receptor.

Interestingly, the molecular basis for this dominant-negative phenotype differed for our HCC TRα1 mutants versus our HCC TRβ1 mutants. The HCC TRβ1 mutants required higher than normal T3 concentrations to release from corepressor, to bind coactivator, and to switch from a transcriptional repressor to an activator, yet displayed virtually wild-type activity when tested at sufficiently high T3 concentrations. The same TRβ1 mutants displayed a correspondingly reduced ability to bind T3, indicating that their primary molecular defect is in T3 binding, rather than in transcriptional activity per se. This is consistent with the location of the corresponding amino-acid substitutions (M313, T329, and C446), which map proximal to the T3-binding pocket of the receptor (Wagner et al., 1995). Paralleling this hormone affinity defect, the TRβ1 mutants functioned as dominant-negative inhibitors of wild-type receptor action at low to moderate hormone concentrations, but not at higher T3 levels.

All three TRβ1 mutants also displayed alterations in their affinity for corepressor even in the absence of T3. Two of the HCC TRβ1 mutants exhibited a stronger affinity for N-CoR than did wild-type TRβ1. Several TRβ1 mutants associated with RTH Syndrome similarly display elevated corepressor binding affinities that manifest as enhanced transcriptional repression/dominant-negative function in vivo (Yoh et al., 1997). The precise molecular basis underlying these alterations in corepressor affinity is not fully clear, but it is intriguing that TRβ1-I is mutated in the pivot (C446) for the helix 12 toggle switch, whereas the relevant RTH-Syndrome mutants have alterations in nearly helix 11. The third HCC TRβ1 mutant tested here, TRβ1-I, interacted more weakly with N-CoR, but more strongly with SMRT than wild-type TRβ1. TRβ1-I displays dominant-negative properties in vivo, suggesting that the loss in N-CoR interaction may be compensated by the gain in SMRT interaction. A reduced interaction with N-CoR has been previously reported for several other HCC TR mutants (Lin et al., 2001); it will be important to determine the interactions of these mutants with SMRT and with other corepressors.

In contrast to our TRβ1 mutants, our TRα mutants were defective for transcriptional activation over a wide range of hormone concentrations, and displayed dominant-negative properties at even the highest T3 levels examined. The molecular basis behind the transcriptional defects in the TRα mutants, unlike that for the TRβ1 mutants, could not be fully ascertained in our in vitro assays. The TRα1-M mutant appeared to be virtually wild type in corepressor release, coactivator binding, and hormone affinity. The TRα1-I mutant required moderately higher T3 than wild type to release corepressor and to bind coactivator in vitro; this is consistent with the location of this mutation near the ligand-binding cavity in the receptor (Wagner et al., 1995). However, concentrations of T3 able to fully induce this coregulator exchange in vitro failed to significantly induce transcriptional activity in vivo. The TRα1-M mutant includes two amino-acid substitutions that map between the hormone- and DNA-binding domains, but are of unclear impact on the receptor structure. Presumably, these TRα1 lesions impair a receptor function necessary for activation in vivo that is not detectable in the in vitro assays employed here. It also should be noted that the differences we observe in the properties of the TRα1 mutants versus TRβ1 mutants may be due, in part, to the nature and location of the mutations themselves, rather than the isoform.

A previous analysis of a separate set of HCC TR mutants identified several that either failed to recruit N-CoR and SMRT, or interacted only weakly relative to wild type (Lin et al., 1997b, 2001). Notably, these other mutants displayed dominant-negative properties, although often weaker than those observed here, and a complete range of T3 concentrations was not tested (Lin et al., 1997b). It is likely that the strong dominant-negative properties displayed by the mutants analysed here are mediated through SMRT or N-CoR, whereas the weaker dominant-negative effects reported for these other mutants may result instead from impaired interactions with coactivators, or other unknown factors.

The TRα mutants display an alteration in DNA-binding specificity

The dominant-negative activity exhibited by the HCC TR mutants studied here is likely to contribute to, but cannot fully explain, how these mutants act in HCC. RTH-Syndrome TR mutants exhibit dominant-negative properties virtually identical to those described here for the HCC TR mutants, are expressed in a wide range of tissues, including hepatocytes, yet RTH Syndrome is not known to increase the incidence of HCC (Refetoff, 1993; DeGroot, 1996; Kopp et al., 1996; Chaterjee, 1997; Nagaya and Seo, 1998). Notably, RTH-Syndrome mutations are characteristically single substitutions that map to the hormone-binding domain, whereas many of the HCC mutants have sustained multiple substitutions that encompass lesions in the DNA recognition domain of the receptor (Refetoff, 1993; DeGroot, 1996; Kopp et al., 1996; Chaterjee, 1997; Nagaya and Seo, 1998; Lin et al., 1999). This suggests the possibility that the altered disease proclivities of these HCC TR mutants might be because of an alteration in their DNA sequence recognition specificity. Consistent with this concept, the two TRα1 mutations studied here display an altered DNA specificity when assayed on a range of half-site variant sequences. The TRα1-I and M mutants, in particular, when assayed as homodimers exhibited a sharply reduced affinity for the AGGTCA consensus sequence, but an increased affinity for ATGTCA or ACGTCA half-sites.

Both TRα1-I and TRα1-M have substitutions at position 74, a lysine in the wild-type receptor that contacts the second position of the half-site (Luisi et al., 1991; Rastinejad et al., 1995; Starr et al., 1996); this is consistent with the changes in their half-site specificity observed here. Intriguingly, a dominant-negative mutant of TRα1 implicated in avian leukemogenesis, v-Erb A, has also sustained substitutions in its DNA recognition domain that alter its half-site specificity and that contribute to its leukemogenicity (Chen et al., 1993; Sande et al., 1993; Judelson and Privalsky, 1996). We suggest that analogous alterations in the DNA recognition domain in the TRα1 HCC mutants result in alterations in their target gene repertoire that contribute to their oncogenic properties and that these changes in target gene specificity may help to distinguish the neoplastic TR alleles from the ‘simple’ dominant-negative mutations associated with inherited RTH Syndrome. The HCC TR mutants that possess wild-type DNA recognition domains may display more subtle changes in DNA recognition, or may play a distinct role in neoplasia from the HCC TR mutants that do display altered target gene specificities.

Altered combinatorial regulation by the TR HCC mutants may impair the antiproliferative properties of T3

TRs can regulate gene transcription through combinatorial interactions with other transcription factors, such as c-Jun/AP-1. AP-1 complexes play a central role in the control of normal cell proliferation and survival, and unrestrained c-Jun function leads to oncogenesis (Eferl and Wagner, 2003). Wild-type TRs generally enhance c-Jun/AP-1 activity in the absence of T3, but inhibit c-Jun in the presence of T3 (Desbois et al., 1991; Zhang et al., 1991; Sharif and Privalsky, 1992; Pfahl, 1993; Schmidt et al., 1993; Moehren et al., 2004). We determined that all of the HCC TR mutants studied here are altered in AP-1 regulation. The three HCC TRβ1 mutants retained the ability to enhance c-jun/AP-1 function in the absence of T3, but required higher than normal T3 to suppress AP-1 activation; if co-introduced with wild-type TRβ1, these HCC mutants interfered with T3 suppression of AP-1 activity by the wild-type receptor. In contrast, the HCC TRα1 mutants have lost much of the ability of the wild-type receptor to stimulate AP-1 activity in the absence of hormone, yet have also lost the ability to suppress AP-1 function in the presence of T3. Therefore, many of the HCC mutants can interfere with the antiproliferative effects of the corresponding liganded receptors, and may allow genes involved in cell proliferation and survival to be expressed under T3 conditions that would normally be growth suppressive. This may operate through AP-1, or through other TR targets (e.g. Garcia-Silva and Aranda, 2004; Furumoto et al., 2005).

Ectopic expression of wild-type TRα in HepG2 cells inhibited anchorage-independent growth by this HCC-derived cell line. Conversely, the TRα-I mutant, although expressed at comparable levels, had little or no suppressive effect on anchorage-independent growth in the same context. These experiments were necessarily performed with media containing intact serum (and therefore containing T3) because charcoal-stripped serum (to remove T3) did not support the growth of colonies. Adding still more T3 reduced the size of the soft agar colonies, with wild-type TRα1 again yielding fewer colonies than did empty vector or the TRα1-I mutant (IHC and MLP, unpublished results). We are currently exploring the growth modulatory properties of wild-type TRβ1 and the other HCC mutants.

TR mutants in HCC: innocent bystanders, criminals, or ex-felons?

It is conceivable that the high incidence of TR mutations in HCC might be an irrelevant byproduct of the genetic instability of these cancers and have no role in tumor establishment or progression. However, the HCC TR mutations are not the null or neutral genetic lesions characteristic of a random mutagenesis event, but rather generate a gain-of-function, dominant-negative phenotype; this commonality of phenotype over a panel of mutant TRs isolated from multiple independent tumors is strongly suggestive of a contributory role. This possibility is further supported by the observation that a mutant TR first isolated as a causal agent in avian erythroleukemia, v-ErbA, induces HCC when expressed in a transgenic mouse model (Barlow et al., 1994) and prior evidence that wild-type TRs may have tumor suppressor-like properties (Sisley et al., 1993; Huber-Gieseke et al., 1997; Lee et al., 2002; Kato et al., 2004). More difficult to address is whether the HCC TR mutants play a continuing role in maintenance of the oncogenic phenotype, or if they instead function at an restricted stage of tumor evolution and are subsequently superceded by additional cell changes associated with tumor progression. Our observations on the TRβ1-E mutation may be relevant in this regard. This mutant was unusual in that it was completely dysfunction in transcriptional assays and failed to display dominant-negative activity. Intriguingly, however, this same mutant closely resembled the other TRβ-1 HCC mutants by displaying defects in corepressor release and coactivator acquisition in vitro, presumably a consequence of the amino-acid substitution sustained within its hormone-binding domain.

The inactivity of the TRβ1-E mutant in transfection experiments was because of an additional amino-acid substitution in its zinc-finger domain that disrupts DNA binding and reporter gene recognition. We suggest that the TRβ1-E mutant represents an ‘ex-felon’ that, at some point in tumor progression, contained the hormone-binding domain mutation, but lacked the zinc-finger lesion, and functioned in oncogenesis in a dominant-negative fashion similar to the other TR HCC mutants studied here. We further suggest that subsequent accumulation of genetic or epigenetic lesions in this tumor rendered the TRβ1-E progenitor unnecessary for the oncogenic phenotype, and in the absence of selective pressure the TRβ1-E acquired the inactivating zinc-finger domain mutation found in the current triple mutant. Although we favor this ‘ex-felon’ view of TRβ1-E, we of course cannot rule out that this mutant may continue to play a role in oncogenesis by yet-unknown mechanisms.

Materials and methods

Molecular clones

Human TRα1 (IMAGE clone 4137980) was amplified by standard polymerase chain reaction (PCR) methodologies to introduce a Kozak sequence and BamHI (5′) and SmaI (3′) sites for subsequent ligation into pGEM7z (Promega Corp., Madison, WI, USA) using the following primers: sense, 5′-GGATC CACCA TGGAA CAGAA GCCAA GCAAG G-3′ and antisense, 5′-GTCGA CTTAG CTAGC GACTT CCTGA TCCTC AAAGA CC-3′. The mutations in TRs previously identified in human HCCs (Lin et al., 1999), TRα1-I (K74E, A264V), TRα1-M (K74R, M150T, E159K), TRβ1-E (M32I, C107R, T368N), TRβ1-I (S43L, C446R), TRβ1-J (M313I), and TRβ1-N (K113N, T329P) were created by a QuikChange protocol (Stratagene, La Jolla, CA, USA). Wild-type TRα1, TRα1-I, and TRα1-M were cloned into pSG5.2 using EcoRI (5′) and HindIII (3′). Wild-type and mutant TRs were also cloned into pFastBac1 (Invitrogen, Carlsbad, CA, USA) using EcoRI restriction sites for the TRβ1 clones and EcoRI (5′) and HindIII (3′) for the TRα1 clones, and wild-type TRα1 and TRα1-I were cloned into pCI-neo using EcoRI (5′) and NotI (3′). The pSG5-wild-type TRβ1, pSG5.2 expression vector, DR4-thymidine kinase promoter (TK)-luciferase reporter, AP-1 collagenase luciferase reporter, pRSV-c-jun, pGEX-MPa-SRC1, pGEX-MPc-SMRT and pGEX-N-CoR constructs were previously described (Weinberger et al., 1986; Sharif and Privalsky, 1992; Wong and Privalsky, 1998; Jonas and Privalsky, 2004; Goodson et al., 2005).

Transient transfection assays

CV-1 cell transfections were performed using 2 μl of Enhancer, 2.5 μl. Effectene, 50 ng of pCH110-lacZ, sufficient pUC18 to bring the total DNA to 250 ng, and additional plasmids as follows: for DR4 reporter assays, 50 ng of DR4-TK-luciferase reporter and 5 ng of the appropriate pSG5-TR; for dominant-negative assays, 50 ng of DR4-TK-luciferase reporter, 5 ng of wild-type pSG5-TR, and 25 ng of the appropriate mutant pSG5-TR; for c-Jun reporter assays, 50 ng of 3 × collagenase promoter-luciferase reporter, 10 ng of the appropriate pSG5-TR, and 10 ng of the pRSV-c-jun construct. Medium was replaced 24 h post-transfection with fresh medium containing either the T3 thyronine indicated or equivalent ethanol carrier. Luciferase and β-galactosidase activities were determined 24 h later (Yoh et al., 1997).

Stable transformed cell lines and soft agar assays

HepG2 cells (3.3 × 10⁵) were seeded into six-well plates and transfected with 8 μl of Enhancer, 10 μl of Effectene, 200 ng of the appropriate pCI-neo-TR construct or pCI-neo vector control, and 800 ng of pUC18. The next day, cells were trypsinized and diluted 1 to 2 and plated in −10-cm dishes. Selective media containing 0.9 mg/ml G418 (Invitrogen) were applied 24 h after the split. Resistant colonies were isolated after 2 weeks and integration of TR was confirmed by RT–PCR. For soft agar assays, cells (5 × 10³) were suspended in 1.5 ml of medium (DMEM with 10% FBS supplemented with 20% HepG2-conditioned medium) containing 0.35% Bacto-Agar (Difco, Becton Dickinson, Sparks, MD, USA). The agar-cell mixtures were then plated on top of 1.5 ml of bottom agar medium (1.2% Bacto-Agar in DMEM with 10% FBS) already solidified in the bottom of a 60-mm plate and allowed to grow for 2 weeks. Colonies were quantified with an Alpha Innotech imager.

In vitro protein–protein interaction assays

GST-N-CoR, GST-SMRT, and GST-SRC1 constructs were expressed in Escherichia coli strain using appropriate pGEX vectors (Guan and Dixon, 1991). Typically, 2−4 μl of ³⁵S-methionine-radiolabeled TRs (synthesized using a Promega TnT reticulocyte system and representing from 200 000−700 000 phosphorimager units) were incubated with the immobilized GST fusion protein of interest (5 μl of agarose matrix per reaction) in a total volume of 120 μl of HEMG buffer containing the appropriate T3 concentration with mixing for 3 h (Farboud and Privalsky, 2004). Radiolabeled TR bound to the GST-fusion after repeated washings was eluted, resolved by SDS–PAGE, and quantified by phosphor imager analysis (Farboud and Privalsky, 2004).

Protease resistance assay

Hormone-mediated protease protection was determined as previously described, except using TRs and T3 in place of retinoic acid receptors and all-trans retinoic acid, and using 0.05 U of elastase (Sigma Chemical Co., St Louis, MO, USA) in place of carboxypeptidase Y (Farboud and Privalsky, 2004).

EMSAs

Oligonucleotides representing two AGGTCA half-sites arranged as a direct repeat with a four-base spacer (a DR4), as a direct repeat with a five-base spacer (a DR5), or as an inverted repeat with a zero spacer (an INV0) were previously described (Lee and Privalsky, 2005). Radiolabeled probes were prepared, incubated with TRs produced in baculovirus-infected Sf9 cells, and the electrophoretic mobility shift/supershift assays performed as previously noted (Lee and Privalsky, 2005). Bound and free probe complexes were visualized and quantified using a Storm phosphor imager.