Oncogene (2009) 28, 4162–4174 & 2009 Macmillan Publishers Limited All rights reserved 0950-9232/09 $32.00 www.nature.com/onc ORIGINAL ARTICLE Thyroid hormone receptor mutants implicated in human hepatocellular carcinoma display an altered target repertoire

IH Chan and ML Privalsky

Department of Microbiology, College of Biological Sciences, University of California at Davis, Davis, CA, USA

Thyroid hormone receptors (TRs) are hormone-regulated lapping, biological roles (Lazar, 1993; Murata, 1998; transcription factors that control multiple aspects of Yen, 2001; Harvey and Williams, 2002). Both TRa1 and normal physiology and development. Mutations in TRs TRb1 bind to specific DNA sequences (response have been identified at high frequency in certain cancers, elements), and regulate the expression of adjacent target including human hepatocellular carcinomas (HCCs). The up or down (Katz and Koenig, 1993; Lazar, 1993, majority of HCC–TR mutants bear lesions within their 2003; Yen, 2001). On ‘positively regulated’ target genes, DNA recognition domains, and we have hypothesized that TRs recruit corepressors and repress gene transcription these lesions change the mutant receptors’ target gene in the absence of T3, but release corepressors, recruit repertoire in a way crucial to their function as oncoproteins. coactivators and activate gene transcription in the Using stable cell transformants and expression array presence of T3 (Cheng, 2000; Zhang and Lazar, 2000; analysis, we determined that mutant TRs isolated from Harvey and Williams, 2002; Tsai and Fondell, 2004). two different HCCs do, as hypothesized, display a target There is less understanding of ‘negatively regulated’ gene repertoire distinct from that of their normal TR target genes that are activated by TRs in the absence, progenitors. Only a subset of genes regulated by wild-type but repressed in the presence, of hormone; this inverted TRs was regulated by the corresponding HCC–TR T3 response presumably arises because of alterations in mutants. More surprisingly, the HCC–TR mutants also coregulator recruitment or function, perhaps reflecting gained the ability to regulate additional target genes not combinatorial interactions with other transcription recognized by the wild-type receptors, and were not simply factors also arrayed on the same promoter (Meyer restricted to repression, but could also activate a subset of et al., 1997; Tagami et al., 1997; Berghagen et al., 2002; their target genes. We conclude that the TR mutants Nygard et al., 2003; Matsushita et al., 2007). isolated from HCC have sustained multiple alterations Disruptions of TR function lead to disease (Yen and from their normal progenitors that include not only changes Cheng, 2003; Cheng, 2005). Inherited mutations in the in their transcriptional outputs, but also changes in the TRb result in human resistance to thyroid genes they target; both are likely to contribute to neoplasia. hormone syndrome, an endocrine disorder (Refetoff, Oncogene (2009) 28, 4162–4174; doi:10.1038/onc.2009.265; 1993; Refetoff et al., 1993; DeGroot, 1996; Kopp et al., published online 14 September 2009 1996; Yen, 2003). A virally transduced mutant form of TRa contributes to leukemogenesis by the avian Keywords: hepatocellular carcinoma; thyroid hormone erythroblastosis retrovirus (Graf and Beug, 1983; Sap receptors; ; mutant; dominant negative et al., 1986; Weinberger et al., 1986; Damm et al., 1989; Privalsky, 1992). Spontaneous mutations in TRa and TRb are found at high frequency in human hepatocel- lular carcinomas (HCCs), renal clear cell carcinomas, Introduction and certain thyroid malignancies and are believed to participate in the initiation or progression of these Thyroid hormone receptors (TRs) have key roles in malignancies (Lin et al., 1996, 1999; Kamiya et al., 2002; normal physiology and development (Brent, 2000; Yen, Puzianowska-Kuznicka et al., 2002; Cheng, 2003; 2001; Harvey and Williams, 2002; Buchholz et al., 2006; Gonzalez-Sancho et al., 2003; Chan and Privalsky, Flamant et al., 2006). Two distinct genetic loci encode 2006). Virtually, all of these TR mutants are impaired in TRs, denoted a and b, each of which is alternatively the ability to exchange coactivator for corepressor in spliced to generate additional receptor diversity. TRa1 response to physiological concentrations of T3. As a and TRb1 are among the most abundant of these result, the mutant receptors retain corepressor inappro- receptor isoforms, and exert distinct, if partially over- priately and can function as dominant-negative inhibi- tors of wild-type receptor function (Lin et al., 1996, Correspondence: Professor ML Privalsky, Department of Microbio- 1997; Chan and Privalsky, 2006; Privalsky, 2008). logy, One Shields Avenues, University of California at Davis, Davis, Given this commonality, why do certain dominant- CA 95616, USA. negative TRs produce primarily endocrine disease, E-mail: [email protected] Received 13 April 2009; revised 3 July 2009; accepted 31 July 2009; whereas others are closely associated with neoplasia? published online 14 September 2009 Notably, the mutant TRs found in resistance to thyroid Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4163 hormone syndrome represent single mutational events, whereas the mutant TRs found in neoplasia are typically aggregates of multiple genetic lesions, often including mutations in the DNA recognition domain of these receptors (Refetoff, 1993; Lin et al., 1999; Kamiya et al., 2002; Puzianowska-Kuznicka et al., 2002; Yen and Cheng, 2003). We have proposed that single TR mutations give rise to simple dominant-negative recep- tors that produce endocrine disorders, whereas addi- tional mutations further modify the dominant-negative phenotype, at least in part by altering target gene recognition, to generate the neoplastic phenotype (Chan and Privalsky, 2006; Privalsky, 2008). To test this hypothesis, we used a microarray analysis to compare the target gene profiles of wild-type and HCC mutant TRs. We report that the HCC–TR mutants tested showed widespread alterations in their target gene repertoire compared with the corresponding wild-type controls, and we discuss how these changes may contribute to oncogenesis.

Figure 1 The human hepatocellular carcinoma–thyroid hormone receptor (HCC–TR) mutants are impaired in transcriptional Results activation. (a) Schematic of wild-type and mutant TRs. The different TRs are depicted with the locations of the DNA binding The HCC–TR mutants are impaired in transcriptional (DBD), hormone-binding domain (HBD) and HCC mutations activation and act as dominant-negative inhibitors of indicated. (b) Impaired reporter activation in HCC mutant TR transformants. HepG2 cells stably integrated with TRa1-WT, wild-type TRs in reporter assays TRb1-WT, TRa1-I, TRb1-N or an empty-plasmid control (no- HepG2 are a well-characterized human HCC cell line receptor (NR)) were transiently transfected with a DR4-TK- that lacks known TR mutations and expresses modest luciferase reporter and pCH110 as an internal control. The cells levels of wild-type TRa1andTRb1 (Chamba et al., were treated 24 h later with T3 or vehicle alone, harvested 48 h after 1996; Lin et al., 1996). We generated stable HepG2 transfection, and the luciferase activity was determined relative to b-galactosidase activity (mean þ s.e.m., n ¼ 3). transformants expressing a representative HCC–TRa1 mutant (denoted TRa1-I), a representative HCC-TRb1 mutant (denoted TR-b1-N), wild-type TRa1 (TRa1- WT), wild-type TRb1 (TRb1-WT), or an empty expres- microarray analysis. The TRa1-I and TRb1-N mutants sion vector control (Figure 1a). We first examined the have been suggested to operate in cancer by constitu- overall T3 response in these cells using a luciferase tively mimicking the repressive functions mediated by reporter containing an archetypal DR4 positive- the wild-type receptors in the absence of T3 (Chan and response element (Figure 1b). The HepG2 transformants Privalsky, 2006). Therefore, we first identified the containing the empty expression plasmid showed a low cellular genes repressed by the mutant TRs relative to level T3-mediated regulation of this reporter mediated the empty-vector transformants, and compared these by the endogenous TRs present in these cells (Darby genes to those repressed by the wild-type TRs relative to et al., 1991; Theriault et al., 1992). Transformants the empty-vector transformants, all in the absence of T3 bearing the TRa1-WT constructs showed a significantly (symbolized as kTR–T3) (Figures 2a and b). Representa- enhanced T3 response (Figure 1b); a similar, although tive target genes were confirmed by reverse transcrip- more modest T3 response was also observed for TRb1- tase–PCR (RT–PCR) (Figure 2c and data not shown). WT (Figure 1b). In contrast, HepG2 transformants Notably, several genes, such as BICC1, were strongly expressing the TRa1-I or TRb1-N mutants showed repressed by all TRs tested (Figures 2b and c). However, reduced levels of reporter expression in the absence of other genes were repressed by the wild-type TR, but not T3 compared with the empty-vector controls, and a by the corresponding mutant under these conditions. severely attenuated reporter gene activation in response For example, DOCK9 was repressed by TRa1-WT but to T3. These results support previous findings that these not TRa1-I, and PROM1 was repressed by both HCC–TR mutants are impaired for T3-induced tran- TRa1-WT and TRb1-WT, but not by either mutant scriptional activation (Chan and Privalsky, 2006). (Figures 2b and c). Reciprocally, certain genes were selectively repressed by the mutant version of a given The HCC–TR mutants repress only a subset of the genes receptor: for example, CDCA7L was repressed only by repressed by wild-type TRs in the absence of T3, and have the TRa1-I mutant, and not by TRa1-WT, whereas gained a novel ability to repress additional targets SLC2A2 was preferentially repressed by the TRb1-N not recognized by wild-type TRs mutant compared with TRb1-WT (Figures 2b and c). A We next compared the patterns of cellular gene more comprehensive list of all genes described in this expression in our HepG2 transformants by use of a study is provided in Supplementary Table S2.

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4164 These results support our hypothesis that the mutant We next examined the effect of T3 on the same panel receptors are selective dominant-negatives able to of genes that were repressed by TRs in the absence of repress some, but not all the genes repressed by their this hormone. As anticipated from our reporter gene wild-type progenitors. Significantly, our data also assays, the TRa1-I and TRb1-N mutants behaved as showed an additional, if limited, capability of the hormone-independent, constitutive repressors (kTR–T3 mutant receptors to repress target genes that were not remained kTR þ T3) on most of their cellular target genes repressed by the corresponding wild-type receptors, (for example, CDC7AL for TRa1-I and TRb1-N, or indicating that the mutations in these receptors did not BHMT2 for TRa1-I) (Figures 3c–e, Supplementary simply narrow, but actually redirected their target gene Table S2). In contrast, repression by the wild-type repertoire. receptors was at least partially reversed by T3 for certain target genes (for example, ERRFI1 for both TRa1-WT and TRb1-WT, and, to a lesser extent, DOCK9 for TRa1-WT; Figures 3a and b, Supplementary Table S2). However, other wild-type TR target genes behaved constitutively and were equally repressed by the wild- type receptor in both the absence and presence of T3 (for example, CAP2 for TRa1-WT or BHMT2 for TRb1-WT; Figures 3d and e). These results were confirmed for representative genes by RT–PCR (data not shown). Target genes that are repressed more strongly in response to T3 than in its absence (that is, negatively regulated genes; kTR–T3 changes to kkTR þ T3) are discussed separately, below. We conclude that wild-type TRs possess a mix of hormone-indepen- dent and hormone-reversible repressive properties, with the latter largely absent for the HCC mutants.

The HCC mutant TRs are not exclusively transcriptional repressors, but are also able to activate a limited set of target genes in the presence of T3 Although several of the genes characterized above were derepressed by T3, most were not activated by T3 above the empty-vector controls. We therefore next searched for TR gene targets that were upregulated above basal levels by T3 (that is, mTR þ T3) (Figures 4a and b, and validated by RT–PCR, Figure 4c). As expected, TRa1-I and TRb1-N failed to induce many of the cellular genes induced by the corresponding wild-type receptors under these conditions (for example, PTAFR, Figures 4b and c).

Figure 2 In the absence of T3, the human hepatocellular carcinoma–thyroid hormone receptor (HCC–TR) mutants repress a distinct set of genes compared with wild-type TRs. (a) Venn diagram of gene transcripts downregulated in each TR transformant compared with the empty-vector (no-receptor (NR)) control, all in the absence of T3. HepG2 cells transformed with the TR alleles indicated, or with the empty-plasmid control, were incubated in the absence of T3 (vehicle only) for 6 h; RNA was isolated, and used to probe the arrays. Transcripts downregulated in each TR transformant compared with the empty-vector control (NR) were identified using a Benjamini–Hochberg adjusted P-value of o0.05. (b) Heat map clustering of the genes from panel (a). For each gene, dark blue indicates lowest expression, dark red indicates highest expression, with intermediate values represented by lighter shades. These comparisons were –T3; asterisks indicate genes that were also downregulated in the presence of T3 (see Figure 6). (c) Reverse transcriptase–PCR (RT–PCR) on representative genes identified in panel (b), using gene-specific primers. The level of expression of each gene product was defined as ‘1’ for the empty-vector control in the absence of T3 (mean þ s.e.m., n ¼ 3). An ‘*’ indicates that the difference between the TR transformant and the empty-vector control was significant at a P-value p0.05.

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4165

Figure 3 Certain target genes that are downregulated by TRs in the absence of T3 are derepressed by T3; others are not. (a–e) Expression levels, minus or plus T3 treatment, of representative gene transcripts from the panel of genes identified in Figure 2. Microarray intensity signal values (mean þ s.d., n ¼ 3) are presented for HepG2 transformants bearing each of the thyroid hormone receptor (TR) expression vectors indicated. The ‘*’ indicates that the difference between the TR transformant and the empty-vector control was significant at a P-value p0.05.

However, these mutants retained an unanticipated mTR þ T3), such as ANGPTL2 and PPL (Figures 5b and ability to activate several of the same genes activated c). Yet other target genes were activated by the wild-type by the wild-type TRs (for example, HEPACAM, TRs essentially independent of T3 status (that is, mTRÀT3 Figure 4b, and by RT–PCR, Figure 4c), and actually remains mTR þ T3), such as CYBD1 (Figure 5d). gained an ability to activate a novel set of target genes Although most of the targets activated by the HCC– not regulated by the wild-type TRs (for example, TR mutants fell into this constitutively activated GNG12; Figure 4b, and by RT–PCR, Figure 4c). category, a subset of HCC–TR target genes retained a Further comparison of this same panel of genes in the residual hormone response and were more strongly absence as well as in the presence of T3 revealed more activated plus T3 versus minus T3 (mTRÀT3 became information (Figure 5). As expected, many of the genes mmTR þ T3; for example, FYN and SGK, Figures 5e and regulated by the wild-type receptors were expressed at f). Representative genes were confirmed by RT–PCR higher levels in the presence of T3 than in its absence. (data not shown). Genes that were more strongly This grouping included genes activated by receptor in the repressed in the presence of T3 (that is, negatively absence of T3, but expressed at still higher levels in the regulated genes) are described below. We conclude that presence of T3 (that is, mTRÀT3 changes to mmTR þ T3), the HCC mutant TRs unexpectedly retain an ability to such as ENPP2 (Figure 5a). It also included genes activate a subset of the genes induced by the correspond- expressed at basal levels (that is, equal to empty-vector ing wild-type receptor, and have also gained an ability to controls) in the absence of T3, and induced above basal activate a novel, although limited, repertoire of addi- only in the presence of T3 (that is, basalTRÀT3 changes to tional genes not activated by the wild-type TRs.

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4166

Figure 4 The human hepatocellular carcinoma (HCC) mutant thyroid hormone receptors (TRs) are able to upregulate a distinct subset of target genes in the presence of T3. (a) Venn diagram of gene transcripts upregulated in each TR transformant compared with the empty-vector (no-receptor (NR)) control, all in the presence of T3. HepG2 cells transformed with the TR alleles indicated, or with the empty-plasmid control, were incubated with T3 for 6 h; RNA was isolated, and used to probe the arrays. Transcripts upregulated in each TR transformant compared with the empty-vector control (NR) were identified using a Benjamini–Hochberg adjusted P-value of o0.05. (b) Heat map clustering of the genes from panel (a). For each gene, dark blue indicates lowest expression, dark red indicates highest expression, with intermediate values represented by lighter shades. These comparisons were þ T3; asterisks indicate genes that were also upregulated in the absence of T3 (see Figure 7). (c) Reverse transcriptase–PCR (RT–PCR) on representative genes identified in panel (b), using gene-specific primers. The level of expression of each gene product was defined as ‘1’ for the empty-vector control in the absence of T3 (mean þ s.e.m., n ¼ 3). An ‘*’ indicates that the difference between the TR transformant and the empty-vector control was significant at a P-value p0.05; ‘#’ indicates a P-value p0.1.

The HCC–TR mutants regulate a narrow subset of the of T3, or preferentially activated in the absence of T3. T3-repressed, negative-response genes that are targeted To capture these additional TR targets, our next step by the wild-type TRs was to identify the genes repressed by mutant or wild- Genes possessing ‘negative-response elements’ operate type receptors in the presence of T3 (kTR þ T3). This reciprocally to the positive-response genes described parsing found both overlap and divergence between the above: they are preferentially repressed in the presence genes targeted by the wild-type versus the mutant TRs

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4167

Figure 5 Certain target genes are constitutively upregulated by thyroid hormone receptors (TRs) independent of hormone status; others are upregulated only in response to T3. (a–f) Expression levels, –T3 or þ T3 treatment, of representative gene transcripts from the panel of genes identified in Figure 4. Microarray intensity signal values are presented (mean þ s.d., n ¼ 3) are presented for HepG2 transformants bearing each of the TR expression vectors indicated. An ‘*’ indicates that the difference between the TR transformant and the empty-vector control was significant at a P-value p0.05.

(Figure 6). Notably, most of the target genes repressed T3 (mTRÀT3) (Figure 7). A significant number of the by TRs in the presence of T3 were identified earlier in targets identified earlier in Figure 4 as induced by the Figure 2 as repressed by TRs in the absence of T3 mutant or wild-type receptors in the presence of T3 were (identified by asterisks in Figures 2b and 6b); the also induced in its absence (Figure 7 and Supplementary majority of these are hormone non-responsive genes Table S2). Many of these targets are not negative- (kTRÀT3 remains kTR þ T3), rather than negative-re- response genes per se, but rather fall into the two sponse genes per se (Figure 6 and Supplementary Table subclasses of positive-response genes already discussed: S2) and have already been discussed. However, several those that are constitutively upregulated in a hormone- genes (such as ANKRD1 and PROM1, Figure 6c) were independent manner (mTRÀT3 remain mTR þ T3), and reproducibly more strongly repressed in the presence those that are upregulated in the absence of hormone, than in the absence of T3 (kTRÀT3 changed to and further upregulated in the presence of T3 (mTRÀT3 kkTR þ T3). The mutant TRs showed an attenuated, change to mmTR þ T3); these are indicated by asterisks in although often still detectable, negative response to T3 Figures 4b and 7b. Although much more rare, genes that on several of these same genes (Figure 6c, Supplemen- do fit our definition of negative regulation, preferentially tary Table S2). upregulated in the absence, relative to the presence, of Our final comparison was to identify the genes that T3 (mmTRÀT3 changes to mTR þ T3 or mTRÀT3 changes to are selectively activated by receptors in the absence of basalTR þ T3), were also identified. These included genes

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4168 We conclude that the HCC TRa1-I and TRb1-N mutants are generally restricted to a narrower panel of target genes and a narrower dynamic range of tran- scriptional output than are the wild-type TRs. None- theless, both TRa1-I and TRb1-N have also acquired an ability to target novel genes not regulated by the wild- type receptors, and can regulate a subset of their target genes up or down at least as strongly as can their wild- type TR progenitors.

The perturbed gene expression mediated by the HCC–TR mutants parallel known alterations of the transcriptosome found in primary HCC tumors At least seven of the genes we identify as expressed at higher levels in the HCC–TR mutant transformants compared with wild-type (that is, either activated by the mutant TRs but not by the wild-type TRs, or repressed by the wild-type TRs but not by the mutant TRs) have been identified earlier as upregulated in primary HCC tumors (Table 1). At least eight of the genes we identify as expressed at lower levels in the TR–HCC mutants compared with wild-type TRs (that is, genes activated by wild-type TRs but not by the mutants) have been identified earlier as downregulated in primary HCC tumors (Table 1). A yet-additional panel of the target genes aberrantly regulated by the TR mutants studied here, although not previously reported to be associated with HCC, have been implicated in other forms of neoplasia (Table 2 and Supplementary Table S3). These comparisons support the relevance of our cell culture analysis to the tumor transcriptosome in vivo, and suggest that the altered gene repertoire of the TR mutants likely contributes to the overall changes in gene expression characteristic of HCC.

Figure 6 The mutant and wild-type thyroid hormone receptors (TRs) negatively regulate distinct sets of target genes in response to T3. (a) Venn diagram of gene transcripts downregulated in each Discussion TR transformant compared with the empty-vector (no-receptor (NR)) control, all in the presence of T3. HepG2 cells transformed with the TR alleles indicated, or with the empty-plasmid control, Thyroid hormone receptor mutants implicated in human were incubated with T3 for 6 h; RNA was isolated, and used to HCC are selective antimorphs, repressing only a subset probe the arrays. Transcripts downregulated in each TR transfor- of the genes targeted by wild-type TRs mant compared with the empty-vector control (NR) were identified Dominant-negative mutant TRs have roles in resistance using a Benjamini–Hochberg adjusted P-value of o0.05. (b) Heat map clustering of the genes from panel (a). For each gene, dark to thyroid hormone syndrome, avian erythroleukemia, blue indicates lowest expression, dark red indicates highest HCC, RCCC and thyroid cancers (Graf and Beug, 1983; expression, with intermediate values represented by lighter shades. Zenke et al., 1990; Lin et al., 1999; Rietveld et al., 2001; These comparisons were þ T3; asterisks indicate genes that were Kamiya et al., 2002; Puzianowska-Kuznicka et al., 2002; also downregulated in the absence of T3 (see Figure 2). (c) Expression levels, minus or plus T3 treatment, of representative Yen and Cheng, 2003; Chan and Privalsky, 2006). What gene transcripts from the genes identified in panel (b). Microarray accounts for the ability of certain of these TR mutants intensity signal values are presented (mean þ s.d., n ¼ 3). An ‘*’ to induce neoplasia, whereas others are exclusively indicates that the difference between the TR transformant and the associated with endocrine disorders? Notably, the empty-vector control was significant at a P-value p0.05. mutant TRs isolated from human HCCs function not only as dominant-negative inhibitors on certain reporter genes but often have sustained additional mutations that regulated preferentially by the wild-type receptors alter their DNA recognition properties (Chan and (for example, APCDD1, Figure 7c) or preferentially Privalsky, 2006). A similar phenomenon was observed by the mutant receptors (for example, GLT25D2 and for the oncogenic v-Erb A allele and for certain RCCC– CX3CR1, Figure 7c; TRa1-WT showed little or no TR mutants (Bonde and Privalsky, 1990; Sharif and regulation of these genes, whereas TRb1-WT showed an Privalsky, 1991; Chen et al., 1993; Ventura-Holman attenuated response compared with the two TR et al., 2008; Rosen and Privalsky, 2009). In fact, v-Erb mutants). A, when expressed in transgenic mice, induces HCC

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4169

Figure 7 The mutant and wild-type thyroid hormone receptors (TRs) positively regulate distinct sets of target genes in the absence of T3. (a) Venn diagram of gene transcripts upregulated in each TR transformant compared with the empty-vector (no-receptor (NR)) control, all in the absence of T3. HepG2 cells transformed with the TR alleles indicated, or with the empty-plasmid control, were incubated in the absence of T3 (vehicle only) for 6 h; RNA was isolated, and used to probe the arrays. Transcripts upregulated in each TR transformant compared with the empty-vector control (NR) were identified using a Benjamini–Hochberg adjusted P-value of o0.05. (b) Heat map clustering of the genes from panel (a). For each gene, dark blue indicates lowest expression, dark red indicates highest expression, with intermediate values represented by lighter shades. These comparisons were –T3; asterisks indicate genes that were also upregulated in the presence of T3 (see Figure 4). (c) Expression levels, –T3 or þ T3 treatment, of representative gene transcripts from the genes identified in panel (b). Microarray intensity signal values are presented (mean þ s.d., n ¼ 3). An ‘*’ indicates that the difference between the TR transformant and the empty-vector control was significant at a P-value p0.05.

(Barlow et al., 1994). We hypothesized that these result is consistent with our reporter experiments and changes might confer an altered target gene repertoire with in vitro data that the TRa1-I mutant binds to a that contributes to the ability of these receptors to more narrow set of natural and artificial DNA-binding induce neoplasia. elements than does WT-TRa1 (Supplementary Figure We report here an analysis of target gene expression S1 and Chan and Privalsky, 2006). Notably, our in HepG2 cells expressing wild-type or mutant TRs. approach also identified an additional set of target This approach confirmed a key aspect of our working genes repressed by the TRa1-I or TRb1-N mutants hypothesis: the TRa1-I and TRb1-N mutants each but not by the wild-type receptors. TRa1-I binds better repressed only a subset of the cell genes repressed by the to at least one artificial DNA sequence than does corresponding wild-type receptors, and repression by TRa1-WT (Supplementary Figure S1 and Chan and the mutants was typically independent of hormone. This Privalsky, 2006), and it is likely these HCC–TR cellular

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4170 Table 1 Aberrantly expressed genes earlier identified in HCC that are also aberrantly regulated by the HCC–TR mutants Gene symbol Gene Exp. in HCC Misreg. by

Repressed by WT, not mutant TR, or activated by mutant, not WT TR CAP2 CAP, adenylate cyclase-associated , 2 m aI CSF1 Colony stimulating factor 1 (macrophage) m aI GPC3 Glypican 3 m aI, bN HDAC1 Histone deacetylase 1 m aI HDAC5 Histone deacetylase 5 m aI PROM1 Prominin 1 m aI, bN TTR Transthyretin (prealbumin, amyloidosis type I) m aI

Activated by WT, not mutant TR C1S Complement component 1, s subcomponent k aI CES1 Carboxylesterase 1 (monocyte/macrophage serine esterase 1) k aI HMOX1 Heme oxygenase (decycling) 1 k aI PCK1 Phosphoenolpyruvate carboxykinase 1 k bN PLAGL1 Pleiomorphic adenoma gene-like 1 k aI PTAFR Platelet-activating factor receptor k aI, bN SERPINF2 Serpin peptidase inhibitor, clade F, member 2 k aI TIMP3 TIMP metallopeptidase inhibitor 3 k aI

Abbreviations: HCC, human hepatocellular carcinoma; TR, thyroid hormone receptor; WT, wild type.

Table 2 Partial list of cancer-implicated genes aberrantly regulated by the HCC–TR mutants Gene symbol Gene Implicated role

AGR2 Anterior gradient 2 homolog Cancer cell survival APOL6 Apolipoprotein l, 6 Cancer cell apoptosis CEACAM1 Carcinoembryonic antigen-related molecule 1 Tumor suppressor CSF1 Colony stimulating factor 1 Cancer progression CX3CR1 Chemokine (C–X3–C motif) receptor 1 Cell migration/survival DKK1 Dickkopf homolog 1 Tumor suppressor ERRFI1 erbb receptor feedback inhibitor 1 Tumor suppressor FGF2 Fibroblast growth factor 2 Cell survival/ FYN fyn oncogene related to src, fgr, yes Oncogene GPC3 Glypican 3 Tumor suppressor MCC Mutated in colorectal cancers Tumor suppressor NR0B2 Nuclear receptor subfamily 0, group b, member 2 Tumor suppressor NRCAM Neuronal cell adhesion molecule Cancer progression PAK1 p21/cdc42/rac1-activated kinase 1 Oncogene PAWR prkc, apoptosis, wt1, regulator Tumor suppressor PLAGL1 Pleiomorphic adenoma gene-like 1 tumor suppressor PTPRJ Protein tyrosine phosphatase, receptor type, j Tumor suppressor SGK Serum/glucocorticoid-regulated kinase 1 Oncogene TIMP3 TIMP metallopeptidase inhibitor 3 Tumor suppressor TMEFF1 Transmembrane protein w/ egf-like and two follistatin-like domains 1 Tumor suppressor VCAN Chondroitin sulfate proteoglycan 2 (versican) Metastasis WT1 Wilms tumor 1 Tumor suppressor

Abbreviations: HCC, human hepatocellular carcinoma; TR, thyroid hormone receptor.

target genes possess related, mutant-specific response Unexpectedly, the HCC–TR mutants were able to elements. activate transcription of a subset of the target genes Therefore, the mutations in the HCC–TR mutants induced by the wild-type receptors, plus an additional set have not simply narrowed their gene recognition of mutant-specific target genes properties, but have also shifted them to encompass Our study also identified genes whose expression was novel targets. We favor the model that this altered target increased by the introduction of a given TR. A subset of gene repertoire arises primarily from the altered DNA these genes were induced by the wild-type TRs more sequence recognition properties of these HCC–TR strongly in the presence than in the absence of T3, mutants; however, we cannot exclude the possibility presumably reflecting the actions of the T3-dependent that alterations in transcriptional regulation after DNA ‘AF-2’ activation domain within the receptor hormone- binding may also contribute. For example, a coactivator binding domain (Yen, 2001). Interestingly, a second required for activation of a specific subset of target panel of target genes were constitutively upregulated genes may be recruited by the wild-type but not by the by the wild-type TRs independent of T3 status; mutant TRs. this category may represent the actions of the TR

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4171 N-terminal ‘AF-1’ domain, which is known to mediate protein–protein interactions with other transcription hormone-independent transcriptional activation (Yen, factors, rather than by direct DNA binding (Desbois 2001). Our results support previous studies indicating et al., 1991; Sharif and Privalsky, 1992). Notably, TRa1-I that wild-type TRs exert a spectrum of possible has a substitution in a highly conserved lysine (codon responses to hormone ranging from derepression to 74). K74 is located in the receptor DNA recognition activation (Yen, 2001). helix and makes base-specific contacts in the major Unexpectedly, a panel of genes was also induced groove of the DNA-binding site (Rastinejad et al., above basal levels by introduction of the HCC mutant 1995). As a result, it is not surprising that the TRa1-I receptors. These mutant-activated genes were generally mutant shows an altered recognition of genes regulated a subset of those activated by the wild-type TRs, through direct receptor/DNA contacts (presumably the although a few were unique to a given mutant receptor. bulk of the positive-response genes identified here). Most were activated by the mutant receptors in a However, K74 has been shown to also exert unusual T3-independent manner, perhaps through the actions of functions on negative-response genes. An artificial lysine the AF-1 domain, which is untouched by the mutations to alanine substitution in TRa1 (K74A) converts the in TRa1-I and TRb1-N. However, a narrow subset of mutant TR from a negative-response regulator of the these target genes was activated more strongly by the collagenase promoter into a positive-response regulator HCC mutant receptors in the presence versus the (Uht et al., 2004). It has been suggested that the wild- absence of hormone. This is a particularly unexpected type lysine at codon 74 may contribute to the direction result for the TRa1-I mutant, which is largely incapable of the hormone response by sensing whether nuclear of activation in artificial reporter gene assays (Chan and receptors are bound to DNA (a positive-response gene) Privalsky, 2006). However, the TRa1-I mutant retains or not (a negative-response gene) (Starr et al., 1996; the ability to bind hormone in vitro, and it is possible Meyer et al., 1997). At least some of the alterations in that the remnants of the AF-2 domain in this mutant are cellular gene regulation observed in our current study sufficient to mediate a T3 response on certain promo- may reflect the dual roles of this lysine both as a sensor ters. Alternatively, endogenous TRs are present at low and as a mediator of DNA binding. levels in the HepG2 cells (Chamba et al., 1996); it is possible that dimers may form between mutant and endogenous TRs, with the endogenous TR partner The TRa1-I and TRb1-N mutants have both convergent conferring a T3-response on target genes determined by and divergent target repertories the mutant TR partner. Mutations in TRa1 and in TRb1 occur separately in We conclude that the HCC–TR mutants can behave certain HCC tumors, indicating that aberrant regulation both as constitutive repressors on certain target genes, by either isoform may contribute to neoplasia (Lin et al., and as constitutive activators on others. In much more 1999). In fact, certain genes identified here, such as restricted circumstances, these same mutants also show PROM1, were regulated by both wild-type receptor a residual T3 response on a specific subset of their target isoforms, but not by either mutant; conversely other gene repertoire. This opens the possibility that certain genes were targeted by both mutants, but by neither aspects of the HCC neoplastic phenotype may be wild-type isoform, such as GNG12. These shared gene responsive to treatment with TR agonists or antago- targets define a common mechanism through which nists, an issue that should be examined further in future either mutant might independently contribute to estab- studies. lishment of the oncogenic phenotype. Nonetheless, approximately 47% of HCCs bear mutations in both TR isoforms, suggesting that mutations in TRa1and The altered gene repertoire of the HCC–TR mutants TRb1 can also work together to exert complementary extends to negative-response genes and reveals divergent roles in a single tumor (Lin et al., 1999). In fact, other modes of target gene recognition sets of the target genes characterized here were regulated An additional category of TR targets, denoted negative- by one or the other mutant but not by both, perhaps response genes, are expressed at lower levels in the helping to explain how two mutant TRs operating presence versus the absence of T3, and have been together might be better than either one operating implicated in the ability of wild-type TRs to suppress separately in tumorigenesis. cell proliferation in response to T3. Although relatively What roles do these genes have in cancer initiation or few of these negative-response genes were detected by progression? Wild-type TRs can act as inhibitors of our array approach, we did observe examples in all three tumor invasiveness, metastasis and proliferation (Gar- possible categories: targets negatively regulated by T3 cia-Silva and Aranda, 2004; Martinez-Iglesias et al., only in the wild-type TR transformants, only in the 2009); one role of these TR mutants, therefore, may be mutant TR transformants or in both types of transfor- to interfere in a dominant-negative manner with the mants. These results extend our previous observations antiproliferative properties of the wild-type TRs. Con- using a prototypic negative-response reporter derived sistent with this proposal, our TRa1-WT transformants from the collagenase promoter (Chan and Privalsky, were inhibited for anchorage-independent growth, 2006). whereas the TRa1-I transformants were not (Chan Thyroid hormone receptors are believed to be and Privalsky, 2006). Perhaps also reflecting the recruited to many negative-response genes through same phenomenon, we noted a slight bias against the

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4172 wild-type TRs when generating our stable HepG2 ethanol carrier alone for 6 h in DMEM containing 10% transformants: 31% of the G418 clones were TRa1-WT hormone-stripped fetal bovine serum (this time period was positive and 34% were TRb1-WT positive, whereas 45% chosen to focus primarily on direct-response genes). The cells were TRa1-I positive and 42% were TRb1-N positive. were harvested and RNA was isolated using an RNeasy kit Prior studies on thyroid and pituitary neoplasia (Qiagen Inc., Valencia, CA, USA). The purified RNA was associated with mutant TR expression have implicated submitted to the Cancer Center Gene Expression Resource, University of California, Davis, for subsequent complemen- several specific pathways that contribute to tumor tary DNA probe generation, hybridization to and scanning of suppression by wild-type TRs and to tumor promotion Affymetrix GeneChip Human Gene 1.0 ST microarrays by TR mutants, such as alterations in Wnt signaling, (Affymetrix Inc., Santa Clara, CA, USA). Raw microarray cyclin D1 regulation and phosphatidylinositol 3-kinase data were normalized by the Robust Multichip Array method (PI3 K) activity (Miller et al., 2001; Furumoto et al., and P-value adjustments were performed by the Benjamini– 2005; Furuya et al., 2006). Our current study is Hochberg method using R software and the affylmGUI consistent with this concept that the HCC–TR mutants package (available at http://www.bioconductor.org/ and function, in part, by counteracting the growth suppres- http://www.r-project.org/; Irizarry et al., 2003; Smyth, 2004). sive properties of the wild-type TRs. However, we also Heat maps were generated with GenePattern (Reich et al., show that the HCC–TR mutants can regulate novel 2006). Three independent biological repeats were analysed for each of the five transformant pools. Luciferase reporter assays genes not recognized by the wild-type receptors, and were performed by transient transfection of these stable may also operate through novel, pro-oncogenic path- transformants (Chan and Privalsky, 2006). ways. A partial list of the genes we identify here as aberrantly regulated by the HCC–TR mutants, and which have been implicated earlier in various aspects of Reverse transcriptase–PCR oncogenic signaling, is provided in Tables 1 and 2. The RNA was isolated from cells using Qiagen’s RNeasy kit Among these are tumor suppressors in the Wnt using the manufacturer’s protocols (Qiagen Inc.). Complementary DNA was synthesized using Qiagen’s QuantiTect kit and 1 mg signaling pathway (DKK1 and WT1), known oncogenes RNA. Semi-quantitative RT–PCR was performed using GoTaq (PAK1 and SGK) and genes involved in other features DNA Polymerase (Promega, Madison, WI, USA) and the primer of cancer such as metastasis (VCAN) and survival sequences listed in Supplementary Table S1. PCR products were (AGR2) (Vadlamudi and Kumar, 2003; Lee et al., 2004; analysed on 8% Tris-acetate-EDTA polyacrylamide gels, stained Vogelstein and Kinzler, 2004; Zhang et al., 2005; with SYBR green, and quantified with a Flurochem8900 Imager Mikheev et al., 2008; Ramachandran et al., 2008; (Alpha Innotech, San Leandro, CA, USA). Ricciardelli et al., 2009). More analysis will be required to determine the precise role of these targets in HCC, their relationship to other cancers associated with mutant TRs, and their significance relative to the Conflict of interest growth regulatory properties of the wild-type TRs. The authors declare no conflict of interest.

Materials and methods Acknowledgements Isolation of stable transformants, microarray analysis, and reporter assay We are extremely grateful to Liming Liu for excellent technical The HepG2 transformants bearing pCIneo constructs of assistance, Michael L Goodson for expert assistance in human wild-type TRa1, wild-type TRb1, TRa1-I, TRb1-N analysis of the array data, and Elsie L Campbell for helpful or an empty-plasmid control were generated using a G418 co- discussions. This work was supported by the Public Health selection methodology (Chan and Privalsky, 2006). Individual Service/National Cancer Institute award R37-CA53394 and by cell clones were propagated and screened for integration and the UC Davis Cancer Center Gene Expression Resource (NCI expression of the TRs by PCR/RT–PCR. In all, 6 to 12 P30-CA93373). IHC was supported in part by a PHS pre- independent cell clones positive for the expression of each doctoral training award, T32-GM007377, from the National receptor were pooled and were treated with 100 nM T3 or with Institute of General Medical Sciences.

References

Barlow C, Meister B, Lardelli M, Lendahl U, Vennstrom B. (1994). Brent GA. (2000). Tissue-specific actions of thyroid hormone: insights Thyroid abnormalities and hepatocellular carcinoma in mice from animal models. Rev Endocr Metab Disord 1: 27–33. transgenic for v-erbA. EMBO J 13: 4241–4250. Buchholz DR, Paul BD, Fu L, Shi YB. (2006). Molecular and Berghagen H, Ragnhildstveit E, Krogsrud K, Thuestad G, Apriletti J, developmental analyses of thyroid hormone receptor function in Saatcioglu F. (2002). Corepressor SMRT functions as a coactivator Xenopus laevis, the . Gen Comp Endocrinol 145: for thyroid hormone receptor T3Ralpha from a negative hormone 1–19. response element. J Biol Chem 277: 49517–49522. Chamba A, Neuberger J, Strain A, Hopkins J, Sheppard MC, Bonde BG, Privalsky ML. (1990). Sequence-specific DNA binding by Franklyn JA. (1996). Expression and function of thyroid hormone the v-erbA oncogene protein of avian erythroblastosis virus. J Virol receptor variants in normal and chronically diseased human liver. 64: 1314–1320. J Clin Endocrinol Metab 81: 360–367.

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4173 Chan IH, Privalsky ML. (2006). Thyroid hormone receptors mutated Lee AY, He B, You L, Xu Z, Mazieres J, Reguart N et al. (2004). in liver cancer function as distorted antimorphs. Oncogene 25: Dickkopf-1 antagonizes Wnt signaling independent of beta-catenin 3576–3588. in human mesothelioma. Biochem Biophys Res Commun 323: Chen H, Smit-McBride Z, Lewis S, Sharif M, Privalsky ML. 1246–1250. (1993). Nuclear hormone receptors involved in neoplasia: erb A Lin KH, Shieh HY, Chen SL, Hsu HC. (1999). Expression of mutant exhibits a novel DNA sequence specificity determined by thyroid hormone nuclear receptors in human hepatocellular amino acids outside of the zinc-finger domain. Mol Cell Biol 13: carcinoma cells. Mol Carcinog 26: 53–61. 2366–2376. Lin KH, Zhu XG, Hsu HC, Chen SL, Shieh HY, Chen ST et al. Cheng SY. (2000). Multiple mechanisms for regulation of the (1997). Dominant negative activity of mutant thyroid hormone transcriptional activity of thyroid hormone receptors. Rev Endocr alpha1 receptors from patients with hepatocellular carcinoma. Metab Disord 1: 9–18. Endocrinology 138: 5308–5315. Cheng SY. (2003). Thyroid hormone receptor mutations in cancer. Lin KH, Zhu XG, Shieh HY, Hsu HC, Chen ST, McPhie P et al. Mol Cell Endocrinol 213: 23–30. (1996). Identification of naturally occurring dominant negative Cheng SY. (2005). Thyroid hormone receptor mutations and disease: mutants of thyroid hormone alpha 1 and beta 1 receptors in a beyond thyroid hormone resistance. Trends Endocrinol Metab 16: human hepatocellular carcinoma cell line. Endocrinology 137: 176–182. 4073–4081. Damm K, Thompson CC, Evans RM. (1989). Protein encoded by Martinez-Iglesias O, Garcia-Silva S, Tenbaum SP, Regadera J, v-erbA functions as a thyroid-hormone receptor antagonist. Nature Larcher F, Paramio JM et al. (2009). Thyroid hormone receptor 339: 593–597. beta1 acts as a potent suppressor of tumor invasiveness and Darby IA, Bouhnik J, Coezy ED, Corvol P. (1991). Thyroid hormone metastasis. Cancer Res 69: 501–509. receptors and stimulation of angiotensinogen production in HepG2 Matsushita A, Sasaki S, Kashiwabara Y, Nagayama K, Ohba K, cells. in vitro Cell Dev Biol 27: 21–24. Iwaki H et al. (2007). Essential role of GATA2 in the negative DeGroot LJ. (1996). Resistance to thyroid hormone. Ann Intern Med regulation of thyrotropin beta gene by thyroid hormone and its 125: 623. receptors. Mol Endocrinol 21: 865–884. Desbois C, Aubert D, Legrand C, Pain B, Samarut J. (1991). A novel Meyer T, Starr DB, Carlstedt-Duke J. (1997). The rat glucocorticoid mechanism of action for v-ErbA: abrogation of the inactivation of receptor mutant K461A differentiates between two different transcription factor AP-1 by retinoic acid and thyroid hormone mechanisms of transrepression. J Biol Chem 272: 21090–21095. receptors. Cell 67: 731–740. Mikheev AM, Mikheeva SA, Maxwell JP, Rivo JV, Rostomily R, Flamant F, Baxter JD, Forrest D, Refetoff S, Samuels H, Scanlan TS Swisshelm K et al. (2008). Dickkopf-1 mediated tumor suppression et al. (2006). International Union of Pharmacology. LIX. in human breast carcinoma cells. Breast Cancer Res Treat 112: The pharmacology and classification of the nuclear receptor 263–273. superfamily: thyroid hormone receptors. Pharmacol Rev 58: Miller LD, Park KS, Guo QM, Alkharouf NW, Malek RL, Lee NH 705–711. et al. (2001). Silencing of Wnt signaling and activation of multiple Furumoto H, Ying H, Chandramouli GV, Zhao L, Walker RL, metabolic pathways in response to thyroid hormone-stimulated cell Meltzer PS et al. (2005). An unliganded thyroid hormone beta proliferation. Mol Cell Biol 21: 6626–6639. receptor activates the cyclin D1/cyclin-dependent kinase/retinoblas- Murata Y. (1998). Multiple isoforms of thyroid hormone receptor: an toma/E2F pathway and induces pituitary tumorigenesis. Mol Cell analysis of their relative contribution in mediating thyroid hormone Biol 25: 124–135. action. Nagoya J Med Sci 61: 103–115. Furuya F, Hanover JA, Cheng SY. (2006). Activation of phosphati- Nygard M, Wahlstrom GM, Gustafsson MV, Tokumoto YM, dylinositol 3-kinase signaling by a mutant thyroid hormone beta Bondesson M. (2003). Hormone-dependent repression of the E2F- receptor. Proc Natl Acad Sci USA 103: 1780–1785. 1 gene by thyroid hormone receptors. Mol Endocrinol 17: 79–92. Garcia-Silva S, Aranda A. (2004). The thyroid hormone receptor is a Privalsky ML. (1992). v-erb A, nuclear hormone receptors, and suppressor of ras-mediated transcription, proliferation, and trans- oncogenesis. Biochim Biophys Acta 1114: 51–62. formation. Mol Cell Biol 24: 7514–7523. Privalsky ML. (2008). Thryoid Hormone Receptors, Coregulators, and Gonzalez-Sancho JM, Garcia V, Bonilla F, Munoz A. (2003). Thyroid Disease. World Scientific Publishing, Ltd.: London. hormone receptors/THR genes in human cancer. Cancer Lett 192: Puzianowska-Kuznicka M, Krystyniak A, Madej A, Cheng SY, 121–132. Nauman J. (2002). Functionally impaired TR mutants are Graf T, Beug H. (1983). Role of the v-erbA and v-erbB oncogenes of present in thyroid papillary cancer. J Clin Endocrinol Metab 87: avian erythroblastosis virus in erythroid cell transformation. Cell 1120–1128. 34: 7–9. Ramachandran V, Arumugam T, Wang H, Logsdon CD. (2008). Harvey CB, Williams GR. (2002). Mechanism of thyroid hormone Anterior gradient 2 is expressed and secreted during the develop- action. Thyroid 12: 441–446. ment of pancreatic cancer and promotes cancer cell survival. Cancer Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Res 68: 7811–7818. Scherf U et al. (2003). Exploration, normalization, and summaries Rastinejad F, Perlmann T, Evans RM, Sigler PB. (1995). Structural of high density oligonucleotide array probe level data. Biostatistics determinants of nuclear receptor assembly on DNA direct repeats. 4: 249–264. Nature 375: 203–211. Kamiya Y, Puzianowska-Kuznicka M, McPhie P, Nauman J, Cheng Refetoff S. (1993). Resistance to thyroid hormone. Clin Lab Med 13: SY, Nauman A. (2002). Expression of mutant thyroid hormone 563–581. nuclear receptors is associated with human renal clear cell Refetoff S, Weiss RE, Usala SJ. (1993). The syndromes of resistance to carcinoma. Carcinogenesis 23: 25–33. thyroid hormone. Endocr Rev 14: 348–399. Katz RW, Koenig RJ. (1993). Nonbiased identification of DNA Reich M, Liefeld T, Gould J, Lerner J, Tamayo P, Mesirov JP. (2006). sequences that bind thyroid hormone receptor alpha 1 with high GenePattern 2.0. Nat Genet 38: 500–501. affinity. J Biol Chem 268: 19392–19397. Ricciardelli C, Sakko AJ, Ween MP, Russell DL, Horsfall DJ. (2009). Kopp P, Kitajima K, Jameson JL. (1996). Syndrome of resistance to The biological role and regulation of versican levels in cancer. thyroid hormone: insights into thyroid hormone action. Proc Soc Cancer Metastasis Rev 28: 233–245. Exp Biol Med 211: 49–61. Rietveld LE, Caldenhoven E, Stunnenberg HG. (2001). Avian Lazar MA. (1993). Thyroid hormone receptors: multiple forms, erythroleukemia: a model for corepressor function in cancer. multiple possibilities. Endocr Rev 14: 184–193. Oncogene 20: 3100–3109. Lazar MA. (2003). Thyroid hormone action: a binding contract. J Clin Rosen MD, Privalsky ML. (2009). Thyroid hormone receptors Invest 112: 497–499. involved in renal clear cell carcinoma alter corepressor release and

Oncogene Altered gene recognition by TR mutants in HCC IH Chan and ML Privalsky 4174 reveal helix 12 as a key determinant of corepressor specificity. Mol binding domain regulates ligand activation profiles at AP-1 sites, Endocrinol 23: 1183–1192. possibly by controlling interactions with a modulating repressor. Nucl Sap J, Munoz A, Damm K, Goldberg Y, Ghysdael J, Leutz A et al. Recept 2:2. (1986). The c-erb-A protein is a high-affinity receptor for thyroid Vadlamudi RK, Kumar R. (2003). P21-activated kinases in human hormone. Nature 324: 635–640. cancer. Cancer Metastasis Rev 22: 385–393. Sharif M, Privalsky ML. (1991). v-erbA oncogene function in Ventura-Holman T, Mamoon A, Subauste JS. (2008). Modulation of neoplasia correlates with its ability to repress retinoic acid receptor expression of RA-regulated genes by the oncoprotein v-erbA. Gene action. Cell 66: 885–893. 425: 23–27. Sharif M, Privalsky ML. (1992). V-erbA and c-erbA enhance Vogelstein B, Kinzler KW. (2004). Cancer genes and the pathways they transcriptional activation by c-jun. Oncogene 7: 953–960. control. Nat Med 10: 789–799. Smyth GK. (2004). Linear models and empirical bayes methods for Weinberger C, Thompson CC, Ong ES, Lebo R, Gruol DJ, Evans assessing differential expression in microarray experiments. Stat RM. (1986). The c-erb-A gene encodes a thyroid hormone receptor. Appl Genet Mol Biol 3: Article3. Nature 324: 641–646. Starr DB, Matsui W, Thomas JR, Yamamoto KR. (1996). Intracel- Yen PM. (2001). Physiological and molecular basis of thyroid lular receptors use a common mechanism to interpret signaling hormone action. Physiol Rev 81: 1097–1142. information at response elements. Genes Dev 10: 1271–1283. Yen PM. (2003). Molecular basis of resistance to thyroid hormone. Tagami T, Madison LD, Nagaya T, Jameson JL. (1997). Nuclear Trends Endocrinol Metab 14: 327–333. receptor corepressors activate rather than suppress basal transcrip- Yen PM, Cheng SY. (2003). Germline and somatic thyroid hormone tion of genes that are negatively regulated by thyroid hormone. Mol receptor mutations in man. J Endocrinol Invest 26: 780–787. Cell Biol 17: 2642–2648. Zenke M, Munoz A, Sap J, Vennstrom B, Beug H. (1990). v-erbA Theriault A, Ogbonna G, Adeli K. (1992). Thyroid hormone oncogene activation entails the loss of hormone-dependent modulates apolipoprotein B gene expression in HepG2 cells. regulator activity of c-erbA. Cell 61: 1035–1049. Biochem Biophys Res Commun 186: 617–623. Zhang J, Lazar MA. (2000). The mechanism of action of thyroid Tsai CC, Fondell JD. (2004). Nuclear receptor recruitment of histone- hormones. Annu Rev Physiol 62: 439–466. modifying enzymes to target gene promoters. Vitam Horm 68: 93–122. Zhang L, Cui R, Cheng X, Du J. (2005). Antiapoptotic effect of serum Uht RM, Webb P, Nguyen P, Price Jr RH, Valentine C, Favre H and glucocorticoid inducible protein kinase is mediated by a novel et al. (2004). A conserved lysine in the DNA mechanism activating IkappaB. Cancer Res 65: 457–464.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene