Isoform-Specific Transcriptional Activity of Overlapping Target Genes that Respond to Thyroid Hormone Receptors α1 and β1

Abstract

Thyroid hormone receptors (TRs) are hormone-regulated transcription factors that control multiple aspects of physiology and development. TRs are expressed in vertebrates as a series of distinct isoforms that exert distinct biological roles. We wished to determine whether the two most widely expressed isoforms, TRα1 and TRβ1, exert their different biological effects by regulating different sets of target genes. Using stably transformed HepG2 cells and a microarray analysis, we were able to demonstrate that TRα1 and TRβ1 regulate a largely overlapping repertoire of target genes in response to T₃ hormone. However, these two isoforms display very different transcriptional properties on each individual target gene, ranging from a much greater T₃-mediated regulation by TRα1 than by TRβ1, to near equal regulation by both isoforms. We also identified TRα1 and TRβ1 target genes that were regulated by these receptors in a hormone-independent fashion. We suggest that it is this gene-specific, isoform-specific amplitude of transcriptional regulation that is the likely basis for the appearance and maintenance of TRα1 and TRβ1 over evolutionary time. In essence, TRα1 and TRβ1 adjust the magnitude of the transcriptional response at different target genes to different levels; by altering the ratio of these isoforms in different tissues or at different developmental times, the intensity of T₃ response can be individually tailored to different physiological and developmental requirements.

Thyroid hormone receptors α1 and β1 regulate overlapping panels of target genes, yet mediate distinct levels of transcriptional activity on any given target gene.

Thyroid hormone receptors (TRs) are ligand-regulated transcription factors that play multiple roles in biology, including the control of nervous system development, cell proliferation, cardiac function, color vision, basal metabolism, lipid utilization, hearing, and metamorphosis (1,2,3,4,5). TRs typically modulate gene regulation by binding to specific DNA sequences (known as thyroid hormone response elements, or TREs) (4,6,7,8,9,10,11); consensus TREs are comprised of repeats of AGGTCA sequences in various orientations and spacings (12,13,14,15,16,17,18,19,20,21). TRs can also regulate gene transcription indirectly through protein-protein interactions with other transcription factors, as well as exert nongenomic effects (7,22,23,24).

TRs have bimodal transcription properties and can either repress or activate their target genes depending on whether hormone (T₃) is bound or not bound to the TR (3,4,6,11,25,26,27,28,29,30,31,32). Positive regulatory targets of TRs refers to those genes that are repressed in the absence and activated in the presence of T₃, whereas negative regulatory targets of TRs refers to the opposite: genes that are activated in the absence and repressed in the presence of T₃.

TRs are encoded at two different loci, α and β, that give rise to four major isoforms by alternative mRNA splicing: TRα1, TRα2, TRβ1, and TRβ2 (1,3,11,33,34,35). Only TRα1, TRβ1, and TRβ2 are able to bind to thyroid hormone and constitute the classic thyroid hormone receptors; TRα2 may have hormone-independent function (3,11,33,34,35). TRα1 is expressed very early in development and is found in most adult tissues, whereas TRβ1 is expressed later in fetal development and is more restricted as to location in the adult (3,11,33,34,35,36). TRβ2 expression is most restrictive of all and is primarily confined to the pituitary, hypothalamus, cone cells of the retina, and the auditory cells of the cochlear (3,11,33,34,35,36,37). TRα1, TRβ1, and TRβ2 are highly homologous within their DNA- and hormone-binding domains; both bind T₃ with similar affinities and can recognize common artificial consensus TREs in vitro.

TRα and TRβ appear to be the result of an ancestral gene duplication event that occurred approximately 550 million yr ago (38,39,40). The retention of these two separate loci during a half-billion years of subsequent evolution strongly suggests that the encoded TRs must exert distinct biological functions. Gene knockout mouse studies of the various TR isoforms have indeed revealed divergent functions for the different isoforms (33,35,36,41). For example, TRα knockout mice exhibit growth retardation due to delayed bone formation, as well as cardiac defects, whereas TRβ knockout mice display elevated circulating thyroid hormone levels, deafness, and defects in retina development (1,33,35,37,42,43,44,45). However these distinctions are not absolute: the two different TR loci display a significant degree of functional redundancy, and TRα and TRβ double-knockout mice exhibit more severe phenotypes than do the single-knockout mice (43,46).

Do TRα and TRβ mediate their different biological roles by regulating distinct sets of genes? To address this question, we examined global transcriptional regulation in HepG2 human liver cells engineered to highly express either TRα1 or TRβ1. Using microarray analysis, we determined that TRα1 and TRβ1 regulate a mostly shared panel of genes in response to T₃. Despite this general overlap in target gene recognition, it was nonetheless apparent that the ability of TRα1 and TRβ1 to activate any given target gene in response to T₃ displayed substantial isoform specificity. Although TRα1 displayed stronger transcriptional activation than did TRβ1 on several genes in these cells, the relative ability of TRβ1 to activate a given target gene in response to T₃ ranged from virtually undetectable to slightly greater than that of TRα1. We propose that it is this TR isoform-specific control of the amplitude of transcriptional regulation that is the likely basis for the appearance and maintenance of TRα1 and TRβ1 over evolutionary time. By altering the ratio of TRα1 to TRβ1 in different tissues or at different developmental times, the magnitude of the T₃ response can be individually tailored for different target genes, allowing the organism to customize the hormone response under different conditions to different physiological needs. We also identified an additional set of target genes that are repressed or activated by TRα1 and TRβ1 irrespective of T₃ status, and which also appear likely to contribute to the divergent biological roles mediated by TRα1 and TRβ1.

Results

Stable ectopic expression of TRα1 or TRβ1 in HepG2 cells can dictate the T₃ response in HepG2 cells

We wished to identify TRα and TRβ target genes in a cell type, such as hepatocytes, naturally responsive to thyroid hormone (5,10,47). We chose for our studies HepG2 cells, a well-characterized cell line derived from a human hepatocellular carcinoma that retains many liver-specific functions. HepG2 cells display a detectable response to thyroid hormone (e.g. Refs. 48,49,50,51,52), confirming that the necessary auxiliary machinery exists in these cells (53). Using quantitative RT-PCR (qRT-PCR) and suitable primers, we determined that TRα1 and TRβ1 are both expressed endogenously in these cells at modest, but near equal, levels (288 ± 15 vs. 294 ± 10 copies per ng RNA), consistent with their near equal expression in intact liver (54). To identify the specific target genes regulated by TRα1 or TRβ1, we chose a strategy of overexpressing one or the other TR isoform to create a TRα1- or TRβ1-dominant environment. We used a neomysin-resistance cotransformation protocol to isolate HepG2 cell lines containing an integrated ectopic human TRα1 construct, an integrated ectopic human TRβ1 construct, or an integrated empty expression plasmid (as a negative control).

Multiple independent HepG2 cell transformant clones were obtained for each construct and were analyzed for TR integration and expression, after which appropriate positive clones were pooled (at least six for each construct). We employed this strategy to minimize any possible clone to clone variation in the T₃ response among the transformant population. We also confirmed the legitimacy of this pooling strategy by analysis of individual clones (described below).

Using qRT-PCR, we determined that overall TRα1 expression was elevated more than 4-fold in the TRα1 transformants relative to the empty vector control cells (Fig. 1A). Similarly, overall TRβ1 expression was elevated approximately 5-fold in the TRβ1 transformant pool (Fig. 1A). Ectopic overexpression of a given isoform did not detectably alter expression of the other, endogenous isoform (Fig. 1A). Other than the actual receptor-coding sequences, our TRα1 and TRβ1 expression constructs possess identical backbones, 5′- and 3′-untranslated regions, and Kozak sequences, and are expected to yield comparable translational efficiency/protein expression.

Figure 1.

Stable ectopic expression of TRα1 and TRβ1 in HepG2 cells results in enhanced T₃-dependent gene activation. A, TRα1 and TRβ1 mRNA levels in HepG2 cells stably transformed with TRα1, TRβ1, or empty [no receptor (NR)] expression plasmids. Total RNA from each pool of stable HepG2 cells was isolated, and TR mRNA levels were determined by quantitative RT-PCR. Relative copy number was calculated as a ratio of TR to GAPDH expression used as an internal normalization control (mean + sem; n = 3). B, Enhanced T₃-dependent reporter activation in TRα1 and TRβ1 transformants. HepG2 cells stably integrated with TRα1, TRβ1, or empty plasmid control (NR) were transiently transfected with a DR4-TK-luciferase reporter and a pCH110 construct as an internal control. The cells were treated with or without T₃ 24 h later as indicated and harvested 48 h after transfection, and the luciferase activity was determined relative to β-galactosidase activity (mean + sem; n = 3).

To examine the overall T₃ response in our stable HepG2 transformants, each was transiently transfected with a thymidine kinase promoter-luciferase reporter bearing a generic DR-4 TR response element/binding site. The empty plasmid control transformants displayed a low basal level of luciferase expression that was induced 2- to 3-fold higher in the presence of 1 nm T₃; no further activation was observed at 100 nm treatment (Fig. 1B). Both the TRα1 and the TRβ1 transformants displayed levels of basal luciferase expression approximately equal to that of the empty plasmid control (repression below basal level is not always observed with this reporter), but a much greater response to T₃ (Fig. 1B). Notably the activation of luciferase expression by T₃ in the TRβ1 transformants (∼6- to 8-fold over that seen in the absence of T₃) was less than that observed for the TRα1 transformants (∼24- to 28-fold over that seen in the absence of T₃) (Fig. 1B). This appears to reflect a generally somewhat higher transcriptional activity of the TRα1 isoform over the TRβ1 isoform (see below and Discussion). Our results, taken collectively, indicate that ectopic expression of TRα1 or TRβ1 significantly enhances the overall T₃ response of our HepG2 transformants beyond that observed for the endogenous TRs.

TRα1 and TRβ1 regulate primarily overlapping sets of T₃-responsive genes

To identify the specific cellular target genes regulated by TRα1 or TRβ1, our stable transformants expressing TRα1, TRβ1, or the empty plasmid control were treated with either ethanol vehicle alone, or ethanol plus 100 nm T₃ for 6 h (a time period designed to primarily focus on direct, rather than secondary, targets of T₃ regulation). RNAs were isolated from each transformant after the treatment period, were converted to cDNA probes, and were analyzed on Affymetrix GeneChip Human Gene 1.0 ST microarrays. Using R or dChip statistics packages and an unadjusted P value <0.05, 59–100 genes were regulated by T₃ in the control transformants (presumably reflecting the activity of the endogenous TRs in these cells), and more than double that number in the ectopic TRα1 or TRβ1 transformants (data not shown). To minimize false positives and also focus on the genes that reflect the actions of the ectopically introduced TR isoforms, we further applied a Benjamini-Hochberg analysis to yield an adjusted P value of < 0.05. The genes that displayed a statistically significant induction by T₃ in each cell population by these criteria are listed in Table 1 and are presented schematically in Fig. 2.

Table 1.

List of genes regulated by TRα1, TRβ1, and/or control after treatment with 100 nm T₃ for 6 h compared with the absence of T₃

Regulated by	Symbol	Gene	Biological process	T3 responsea (fold, adj. P value)
				TRα1		TRβ1
TRα1	ANGPTL2	Angiopoietin-like 2	Signal transduction	1.6 (0.036)		1.4 (0.437)
	CLMN	Calmin		1.7 (0.020)		1.4 (1.000)
	FLJ21963		Metabolic process	1.4 (0.043)		1.3 (0.165)
	GBE1	Glucan (1,4-α-), branching enzyme 1	Carbohydrate metabolic process	2.5 (0.043)		2.0 (0.468)
	GRHPR	Glyoxylate reductase/hydroxypyruvate reductase	Electron transport/excretion	1.5 (0.043)		1.2 (1.000)
	HERPUD1	Homocysteine-inducible, ER stress-inducible, ubiquitin-like domain member 1	Protein modification process	1.3 (0.029)		1.3 (0.107)
	IRF1	Interferon-regulatory factor 1	Transcription/immune response/cell cycle	2.1 (0.003)		1.4 (1.000)
	KCNJ10	Potassium inwardly rectifying channel, subfamily j, member 10	Potassium ion transport	2.7 (1E-4)		1.5 (0.787)
	MMP11	Matrix metallopeptidase 11	Proteolysis	1.7 (0.043)		1.5 (0.486)
	MPV17Lb	mpv17-like protein type 2		2.1 (0.002)		1.8 (0.075)
	RHPN2	Rhophilin, ρ gtpase binding protein 2	Signal transduction	1.6 (0.029)		1.5 (0.141)
	SALL2	sal-like 2	Transcription	1.7 (0.032)		1.4 (0.914)
	SLC38A3	Solute carrier family 38, member 3	Sodium ion transport/amino acid transport	1.5 (0.029)		1.4 (0.157)
	TMEFF1	Transmembrane protein with EGF-like and two FSH-like domains 1	Multicellular organismal development	1.5 (0.029)		1.2 (1.000)
	ZNF165	Zinc finger protein 165	Transcription	1.8 (0.047)		1.7 (0.156)
TRβ1	ACSM2Bb	Hypothetical protein loc123876	Metabolic process	2.1 (0.081)		2.2 (0.047)
	HIF1Ab	Hypoxia-inducible factor 1, α subunit	Response to hypoxia/signal transduction	1.5 (0.068)		1.5 (0.049)
	MASP1b	Mannan-binding lectin serine peptidase 1	Complement activation/proteolysis	1.5 (0.090)		1.7 (0.022)
	PCK1	Phosphoenolpyruvate carboxykinase 1	Gluconeogenesis/lipid metabolic process	2.0 (0.255)		3.1 (0.004)
	ZMIZ1b	Retinoic acid induced 17	Transcription	1.4 (0.084)		1.5 (0.043)
TRα1/TRβ1	ARSB	Arylsulfatase b	Lysosomal transport/glycosaminoglycan metabolic process	1.7 (0.029)		1.7 (0.023)
	C18orf17	Chromosome 18 open reading frame 17		2.0 (1E-4)		1.9 (0.001)
	C1orf115	Chromosome 1 open reading frame 115		1.5 (0.032)		1.5 (0.023)
	CABLES2	cdk5 and abl enzyme substrate 2	Cell cycle/regulation of cell division	2.0 (0.002)		1.8 (0.022)
	CD14	cd14 antigen	Apoptosis/immune response/signal transduction	1.9 (0.002)		1.7 (0.019)
	CPT1A	Carnitine palmitoyltransferase 1a	Lipid/fatty acid metabolic process	2.1 (0.003)		2.0 (0.014)
	CYP24A1	Cytochrome p450, family 24, subfamily a, polypeptide 1	Electron transport	3.6 (0.017)		3.6 (0.023)
(Continued)

Table 1A.

Continued

Regulated by	Symbol	Gene	Biological process	T3 responsea (fold, adj. P value)
				TRα1		TRβ1
	ENG	Endoglin	Cell adhesion/regulation of transcription	2.6 (6E-5)		1.7 (0.041)
	FYN	fyn oncogene related to src, fgr, yes	T cell receptor signaling pathway	3.5 (3E-8)		2.6 (3E-6)
	HIVEP2	HIV type 1 enhancer binding protein 2	Transcription	2.3 (4E-5)		2.4 (3E-5)
	LDLRAD1	Low-density lipoprotein receptor class a domain containing 1		1.7 (0.005)		1.8 (0.003)
	PPL	Periplakin	Keratinization	1.9 (0.029)		2.0 (0.020)
	PTAFR	Platelet-activating factor receptor	Chemotaxis/immune response/signal transduction	2.6 (2E-5)		2.0 (0.002)
	RAB11FIP4	rab11 family interacting protein 4 (class ii)		1.8 (2E-4)		1.6 (0.004)
	RASSF4	ras association (ralgds/af-6) domain family 4	Cell cycle/signal transduction	1.7 (0.009)		1.7 (0.004)
	SCNN1A	Sodium channel, nonvoltage-gated 1 α	Sodium ion transport/excretion	3.4 (4E-4)		2.5 (0.022)
	SERPINB9	Serpin peptidase inhibitor, clade b, member 9	Antiapoptosis/signal transduction	2.1 (0.001)		2.1 (0.002)
	SGK	Serum/glucocorticoid regulated kinase 1	Apoptosis/sodium ion transport	1.8 (0.020)		2.0 (0.004)
	SLC16A6	Solute carrier family 16, member 6	Monocarboxylic acid transport	2.5 (0.012)		2.4 (0.023)
	TNS4	Tensin 4	Apoptosis/intracellular signaling cascade	2.8 (9E-5)		2.0 (0.020)
	TP53INP2	Tumor protein p53-inducible nuclear protein 2		3.1 (0.001)		2.9 (0.003)
	USP30	Ubiquitin-specific peptidase 30	Ubiquitin cycle	1.4 (0.016)		1.4 (0.041)
	SLC2A1	Solute carrier family 2, member 1	Glucose transport	0.7 (0.012)		0.6 (0.004)
				TRα1	TRβ1	NR
TRα1/TRβ1/NR	COQ10A	Coenzyme Q10 homolog A		3.6 (3E-8)	3.2 (3E-7)	2.6 (1E-5)

Genes were identified as T₃-responsive at a Benjamini-Hochberg (BH) adjusted P value <0.05. NR, No receptor. 

^aT₃ response values are presented as fold change, and the BH adjusted (adj.) P values are shown in parentheses. 

^bGenes T₃ regulated by both TRα1 and TRβ1 at a BH adjusted P < 0.1. 

Figure 2.

TRα1 and TRβ1 regulate overlapping sets of T₃-responsive genes. A, Schematic of pairwise comparisons employed: HepG2 cells stably transformed by TRα1, by TRβ1, or by the empty plasmid control were treated with vehicle only or with T₃ for 6 h. RNA was isolated, and the genes in each transformant were up- or down-regulated by T₃ were identified in the following pairwise comparisons: TRα1 transformants minus vs. plus T₃, TRβ1 transformants minus vs. plus T₃, and empty plasmid (NR) transformants minus vs. plus T₃. B, Venn diagram of genes regulated by T₃ in TRα1 transformants, by T₃ in TRβ1 transformants, or by T₃ in the empty plasmid control (NR), identified at a Benjamini-Hochberg (BH) adjusted P value of <0.05. The number of T₃-regulated genes shared, or not, among the three different transformants are enumerated and are shown schematically as overlapping circles. C, The same genes identified as T₃-up-regulated in panel B are presented again, but at a BH-adjusted P value of <0.1. NR, No receptor.

Under these stringent statistical criteria, only one gene, COQ10A, was flagged as up-regulated by T₃ in the empty plasmid control transformants; examination of the microarray intensity values illustrates this more quantitatively (Fig. 3A), and this result was confirmed by RT-PCR assay (Fig. 4A). By the same criteria, 37 additional genes (plus COQ10A) were flagged as up-regulated by T₃ in the TRα1 transformants, whereas 27 additional genes (plus COQ10A) were flagged as up-regulated by T₃ in the TRβ1 transformants. Of the genes flagged as up-regulated by TRα1, 23 were independently flagged as up-regulated by TRβ1 (Fig. 2B) by the adjusted P < 0.05 criterion. One gene, SLC2A1, was found to be down-regulated by T₃ in the TRα1 transformants (Table 1).

Figure 3.

T₃ response of representative target genes. A–J, Microarray intensity signal values of mRNAs, minus and plus T₃, for genes flagged as T₃-regulated in the TRα1, TRβ1, or empty plasmid transformants (adjusted P < 0.05; mean + sd; n = 3). Asterisk indicates that the difference between minus and plus T₃ was significant at a P value ≤0.05. NR, No receptor.

Figure 4.

Validation of microarray data by RT-PCR analysis and heat map of genes regulated by T₃ in TRα1 or TRβ1 transformants. A–D, The mRNAs from TRα1, TRβ1, or empty plasmid (NR) transformants, minus or plus 6 h of T₃ treatment, were subjected to RT-PCR using the gene-specific primers described in Materials and Methods. The resulting PCR products were stained with SYBR Green, and the gel band intensity values were quantified using an 8900 gel documentation system (Alpha Innotech, San Leandro, CA). The level of expression of each gene product was defined as 1 for the empty-vector control in the absence of T₃ (mean + sem; n = 3). E, Heat map clustering of genes up-regulated by T₃ in the HepG2 TRα1, TRβ1, or empty plasmid transformants (P <0.05). Dark blue indicates low expression, dark red indicates high expression, with intermediate values represented by lighter shades. *, The difference between minus and plus T₃ was significant at a P value ≤0.05; #, the difference between minus and plus T₃ was significant at a P value ≤0.1. NR, No receptor.

To examine whether these distinct sets of genes were truly unique, the same gene list in Table 1 was analyzed at a less stringent adjusted P value cutoff of P < 0.1. A Venn diagram at this P value shows that four of the five genes originally identified as being regulated by only TRβ1 at P < 0.05, are moved to the shared group of target genes under this criterion, indicating that they are also likely regulated by TRα1 (Fig. 2C). One gene, originally identified as only TRα1 regulated, also moved to the shared group of target genes under the adjusted P < 0.1 (Fig. 2C). We conclude that a majority of the genes regulated by T₃ in these assays were shared targets of both TRα1 and TRβ1. Further individual analysis of these T₃-regulated genes, below, elaborated on this concept.

Despite their ability to recognize a shared set of target genes, TRα1 and TRβ1 differ dramatically in the extent to which they can activate each target gene in response to T₃

We next examined specific representatives of the T₃-regulated genes characteristic of the three major categories: 1) shared, 2) TRα1 specific, or 3) TRβ1 specific. Two target genes, PTAFR and PPL, flagged as “shared” at an adjusted P < 0.05, were, as anticipated, induced by T₃ in both the TRα1 and TRβ1 transformants, but not in the empty plasmid control cells; this was observed by microarray intensity measurement in three biological repeats and, for PTAFR, confirmed by RT-PCR (Fig. 3, B and C, and Fig. 4B). Two representative target genes, PCK1 and MASP1, flagged as “TRβ1-specific” at an adjusted P value <0.05, were induced by T₃ in the TRβ1 transformants, as expected, but also were detectably induced by T₃ in the TRα1 transformants (Fig. 3, D and E, and Fig. 4C). In fact, MASP1 was among the TRβ1-specific genes that joined the shared category when the stringency of the array analysis was decreased to an adjusted P < 0.1. We conclude that our initial exclusion of MASP1 and PCK1 from the shared group more likely reflected our initial choice of a statistical cutoff in our overall array analysis, rather than a true, TRβ1-specific regulation of these genes.

Similarly, a more detailed analysis of two target genes, KCNJ10 and MPV17L, flagged as “TRα1-specific” at an adjusted P < 0.05, showed that these genes were indeed activated by T₃ in the TRα1 transformants; however, it appeared that these genes were also activated, although to a lesser extent, in the TRβ1 transformants (Fig. 3, F and G); indeed MPV17L was one of the TRα1-specific genes that became regrouped as “shared” when the adjusted P value stringency was lowered to P < 0.1 (Fig. 2C). KCNJ10 and MPV17L were not regulated by T₃, or were regulated to a lesser degree, in transformants bearing the empty plasmid (Fig. 3, F and G).

Our results indicate that TRα1 and TRβ1 regulate a largely overlapping set of target genes in response to T₃; however, what is also clear is that these two different isoforms display very different transcriptional properties on each of these different targets. TRα1 was a stronger activator in response to T₃ than is TRβ1 on several genes (as was also observed using the artificial DR-4-tk-luciferase reporter; Fig. 1B). Nonetheless, there was a wide range of relative activities of TRα1 to TRβ1 on the individual target genes, with certain target genes exhibiting equal or slightly greater T₃ activation in the TRβ1 transformants than in the TRα1 transformants (e.g. PCK1), other target genes displaying a somewhat lower T₃ response in the TRβ1 transformants than in the TRα1 transformants (e.g. PTAFR or KCNJ10), and a few target genes (e.g. IRF1, CLMN, and TMEFF1) displaying a T₃ response statistically distinct from the empty plasmid controls in the TRα1 transformants, but not in the TRβ1 transformants (Table 1). Although the final category may truly represent TRα1-specific target genes, individual analysis of these genes (Fig. 3, H–J, and Fig. 4D) suggest they are also potentially, if weakly, TRβ1 responsive. We therefore cannot exclude the possibility that the lack of a clearly enhanced T₃ response in the TRβ1 transformants in these cases (compared with empty plasmid controls) simply reflects a much weaker transcriptional response of TRβ1, rather than an absolute isoform specificity.

The range of transcriptional responses for all the genes flagged by the microarray analysis as regulated by T₃ is displayed in a heat map in Fig. 4E. Many of these genes displayed little or no detectable regulation by T₃ in the absence of an ectopically introduced TR, whereas other genes were regulated by T₃ in the empty plasmid control, but were regulated more strongly in one or both TR transformants. Parallel results were observed when a variety of selected genes from the different categories were validated by RT-PCR (data not shown).

To confirm that these patterns of gene regulation were not distorted by our strategy of analyzing pooled transformants, we repeated RT-PCR validations on several target genes using individual clonal populations of our TRα1 and TRβ2 transformants (choosing clones displaying comparable levels of TR expression). Notably, very similar patterns of target gene expression were obtained on these individual clones as were seen for the pooled analysis (e.g. supplemental Figs. S1A and S1B published on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org). We conclude that TRα1 and TRβ1 recognize a large panel of shared target genes, but that these two isoforms differ in the transcriptional response they can initiate once arrayed on each individual target gene.

TRα1 and TRβ1 display additional regulatory effects on target genes that are observable irrespective of the presence or absence of T₃

Our results to this point focused on genes that displayed a statistically significant change in expression in response to T₃ in our HepG2 transformants (Fig. 2A). However, this focus on T₃ response could potentially ignore genes that are regulated by TRs irrespective of the absence or presence of T₃, i.e. in a constitutive fashion. To explore this aspect, the microarray data for our various stable transformants were reanalyzed in four pairwise comparisons (Fig. 5A) to identify: 1) genes that are down-regulated by introducing TRα1 relative to the empty plasmid, regardless of T₃ status; 2) genes that are down-regulated by introducing TRβ1 relative to the empty plasmid, regardless of T₃ status; 3) genes that are up-regulated by introducing TRα1 relative to the empty plasmid, regardless of T₃ status; and 4) genes that are up-regulated by introducing TRβ1 relative to the empty plasmid, regardless of T₃ status.

Figure 5.

A subset of genes are down- or up-regulated by TRα1 or TRβ1 irrespective of T₃ status. A, Schematic of pairwise comparisons employed: HepG2 cells stably transformed by TRα1, by TRβ1, or by the empty plasmid control were treated with vehicle only or with T₃ for 6 h. RNA was isolated, and the genes down- or up-regulated by the presence of the ectopic TRα1 or TRβ1 were identified in the following pairwise comparisons: TRα1 transformants vs. empty plasmid (NR) transformants in the absence of T₃, TRβ1 transformants vs. empty plasmid (NR) transformants in the absence of T₃, TRα1 transformants vs. empty plasmid (NR) transformants in the presence of T₃, TRβ1 transformants vs. empty plasmid (NR) transformants in the presence of T₃. B, Venn diagram of genes down-regulated in TRα1 transformants compared with the empty plasmid (NR) control, or down-regulated in TRβ1 transformants compared with the empty plasmid (NR) control regardless of T₃ concentration. All were flagged using a Benjamini-Hochberg (BH) adjusted P value of <0.05. The number of down-regulated genes shared, or not, between the TRα1 and TRβ1 transformants is enumerated and is shown schematically as overlapping circles. C, Venn diagram of genes up-regulated in TRα1 transformants compared with the empty plasmid (NR) control, or up-regulated in TRβ1 transformants compared with the empty plasmid (NR) control regardless of T₃ concentration. All were flagged using a Benjamini-Hochberg (BH) adjusted P value <0.05. The number of up-regulated genes shared, or unshared, between the TRα1 and TRβ1 transformants is enumerated and is shown schematically as overlapping circles. NR, No receptor.

Using an adjusted P value <0.05, relative to the empty plasmid control, the introduction of TRα1 repressed 33 genes, and the introduction of TRβ1 repressed 12 genes in both the absence and presence of T₃ (Fig. 5B and Table 2). Nine of these genes were repressed in both the TRα1 and TRβ1 transformants. The PROM1 gene, for example, was repressed by both TR isoforms in either the absence or presence of T₃ (Table 2 and confirmed by RT-PCR; Fig. 6A). Several of the repressed genes flagged as TRα1 or TRβ1 specific in the microarray analysis appeared to be truly isoform specific on additional analysis. DOCK9, for example, was significantly repressed only in the TRα1 transformants, whereas CDCA7L was significantly repressed only in the TRβ1 transformants (Table 2 and Fig. 6, B and C). Several of these genes were repressed constitutively and displayed no detectable response to T₃ (e.g. DOCK9 in the TRα1 transformant), whereas others were marginally repressed further in the presence of T₃ than in its absence (e.g. PROM1 for TRβ1) (Fig. 6); these changes in response to T₃ were too modest to be detected as T₃ responsive in our initial analysis. Additional specific examples were confirmed by RT-PCR validation (e.g. supplemental Fig. S1C), and a complete list of genes repressed by the introduction of TRα1 or TRβ1 in either the absence or presence of T₃ is found in Table 2.

Table 2.

List of genes down-regulated by TRα1 and/or TRβ1 compared with control

Regulated by	Symbol	Gene	Biological process	Fold change
				−T3		+T3
TRα1	AFP	α-Fetoprotein	Transport/immune response	NS		0.8
	ATP10A	ATPase, class v, type 10c	Phospholipid transport	NS		0.6
	C14orf126	Chromosome 14 open reading frame 126	d-Amino acid catabolic process	NS		0.7
	CAP2	Cap, adenylate cyclase-associated protein, 2	Signal transduction	0.5		0.5
	CYP7A1	Cytochrome p450, family 7, subfamily a, polypeptide 1	Electron transport/lipid metabolic process	NS		0.5
	DOCK9	Dedicator of cytokinesis 9		0.4		0.4
	F13B	Coagulation factor xiii, b polypeptide	Blood coagulation	NS		0.4
	FKBP14	fk506 Binding protein 14, 22 kDa	Protein folding	NS		0.7
	FRMD3	FERM domain containing 3		NS		0.6
	GDA	Guanine deaminase	Nucleotide and nucleic acid metabolic process	0.5		0.5
	LYZ	Lysozyme	Inflammatory response/metabolic process	NS		0.7
	NR0B2	Nuclear receptor subfamily 0, group b, member 2	Transcription/cholesterol metabolic process	0.6		NS
	NUBPL	Nucleotide-binding protein-like		NS		0.8
	OXA1 liter	Oxidase (cytochrome c) assembly 1-like	Electron transport	0.8		NS
	PAWR	prkc, apoptosis, wt1, regulator	Transcription/apoptosis	NS		0.7
	PCSK5	Proprotein convertase subtilisin/kexin type 5	Proteolysis/cell-cell signaling	0.7		NS
	PELI1	Pellino homolog 1		0.7		NS
	SCN1A	Sodium channel, voltage-gated, type i, α	Sodium ion transport	NS		0.5
	SGEF	src Homology 3 domain-containing guanine nucleotide exchange factor	Regulation of ρ protein signal transduction	0.6		NS
	SLC2A1	Solute carrier family 2, member 1	Glucose transport	NS		0.7
	SNX6	Sorting nexin 6	Intracellular protein transport/cell communication	NS		0.7
	TFPI	Tissue factor pathway inhibitor	Blood coagulation	NS		0.6
	TTR	Transthyretin	Thyroid hormone generation/transport	0.7		0.6
	VPS54	Vacuolar protein sorting 54	Protein transport	0.7		0.7
				TRα1		TRβ1
				−T3	+T3	−T3	+T3
TRα1/TRβ1	ANKRD1	Ankyrin repeat domain 1	Signal transduction	0.5	0.4	NS	0.4
	BICC1	Bicaudal c homolog 1	Visual perception/response to stimulus	0.4	0.3	0.4	0.4
	DPYD	Dihydropyrimidine dehydrogenase	Multicellular organismal development	0.6	0.6	0.6	0.6
	ERRFI1	erbb receptor feedback inhibitor 1	Electron transport/metabolic process	0.7	NS	0.7	NS
	GPC3	Glypican 3	Cell proliferation/growth	0.6	0.7	0.7	0.7
	HDAC1	Histone deacetylase 1	Transcription/chromatin modification	0.7	0.7	NS	0.7
	LTB4DH	Leukotriene b4 12-hydroxydehydrogenase	Response to stress	0.5	0.5	0.5	0.5
	PROM1	Prominin 1	Leukotriene metabolic process	0.3	0.2	0.5	0.4
	SCHIP1	Schwannomin interacting protein 1		NS	0.6	NS	0.5
					−T3	+T3
TRβ1	BHMT2	Betaine-homocysteine methyltransferase 2			0.6	NS
	C3orf57	Chromosome 3 open reading frame 57			0.6	NS
	CDCA7 liter	Cell division cycle associated 7-like	Transcription		0.5	0.5

Genes were identified at a Benjamini-Hochberg adjusted P value <0.05. NS, Not significant (difference from control did not meet a P value cutoff of 0.05). 

Figure 6.

Validation of microarray data by RT-PCR analysis: genes down- or up-regulated by TRα1 or TRβ1 regardless of T₃ status. A–F, The mRNAs from TRα1, TRβ1, or empty plasmid (NR) transformants, minus or plus 6 h of T₃ treatment, were subjected to RT-PCR using the gene-specific primers described in Materials and Methods. The resulting PCR products were stained with SYBR Green, and the gel band intensity values were quantified using an 8900 gel documentation system (Alpha Innotech). The level of expression of each gene product was defined as 1 for the empty-vector control in the absence of T₃ (mean + sem; n = 3). *, The difference between vector only and plus the TR isoform was significant at a P value ≤ 0.05; #, the difference was significant at a P value ≤0.1. NR, No receptor.

Finally we flagged the genes that were activated in response to expression of TRα1 or TRβ1 in both the absence and presence of hormone. Compared with the empty plasmid control, introduction of TRα1 up-regulated 105 target genes in both the absence and presence of T₃, whereas introduction of TRβ1 up-regulated 38 genes in both the absence and presence of T₃; 30 of these up-regulated genes were shared between both isoforms (Fig. 5C and Table 3). HEPACAM, for example, was up-regulated by the ectopic expression of either TR isoform (Fig. 6D). SLC6A12 was strongly up-regulated in the TRα1 transformants both minus or plus T₃ (although more activation was detected in the presence of T₃), but was only slightly up-regulated in the TRβ1 transformants (Fig. 6E). Conversely, ZNF185 was more strongly up-regulated in the TRβ1 transformants than in the TRα1 transformants (Fig. 6F). Additional specific examples were confirmed by RT-PCR validation (e.g. supplemental Fig. S1D), and a complete list of genes up-regulated by TRα1 or TRβ1 in both the presence and absence of T₃ is displayed in Table 3.

Table 3.

List of genes up-regulated by TRα1 and/or TRβ1 compared with control

Regulated by	Symbol	Gene	Biological process	Fold change
				−T3		+T3
TRα1	ALDH8A1	Aldehyde dehydrogenase 8 family, member a1	Retinoic acid metabolic process	2.3		1.8
	ANGPTL2	Angiopoietin-like 2	Signal transduction	NS		1.7
	APCDD1	Adenomatosis polyposis coli down-regulated 1		2.0		1.7
	ARSI	Arylsulfatase i	Metabolic process	2.7		NS
	BHMT	Betaine-homocysteine methyltransferase	Protein amino acid methylation	2.0		NS
	C12orf40	Chromosome 12 open reading frame 40		2.1		NS
	C1S	Complement component 1, s subcomponent	Proteolysis/complement activation/immune response	NS		2.3
	C1orf115	Chromosome 1 open reading frame 115		NS		1.5
	C2orf54	Hypothetical protein flj22671		NS		1.5
	CCL16	Chemokine (c-c motif) ligand 16	Chemotaxis/immune response/cell-cell signaling	NS		2.2
	CD14	cd14 antigen	Apoptosis/immune response/signal transduction	NS		1.5
	CENTD1	Centaurin, δ 1	Signal transduction	1.8		1.5
	CES4	Carboxylesterase 4-like	Xenobiotic metabolic process	1.9		2.0
	CLDN6	Claudin 6	Calcium-independent cell-cell adhesion	3.0		3.3
	CLMN	Calmin		NS		2.1
	CORIN	Corin, serine peptidase	Proteolysis/lipid metabolic process	NS		2.1
	CYBRD1	Cytochrome b reductase 1	Electron transport	2.8		3.1
	DSC3	Desmocollin 3	Cell adhesion	NS		2.6
	FGF2	Fibroblast growth factor 2 (basic)	Angiogenesis/chemotaxis/signal transduction/cell signaling/proliferation	1.7		2.0
	G6PD	Glucose-6-phosphate dehydrogenase	Glucose metabolic process	1.3		1.4
	GAL3ST1	Galactose-3-o-sulfotransferase 1	Galactosylceramide biosynthetic process/myelination	1.5		NS
	H2AFY2	h2a histone family, member y2	Nucleosome assembly/dosage compensation/chromatin modification	1.7		1.6
	HMOX1	Heme oxygenase (decycling) 1	Heme oxidation/regulation of I-κB kinase/NF-κB cascade	NS		2.1
	HPX	Hemopexin	Cellular iron ion homeostasis/heme transport	2.6		NS
	HSPB1	Heat shock 27 kDa protein 1	Regulation of translational initiation/antiapoptosis/cell motility	1.4		NS
	IL1R1	IL-1 receptor, type I	Immune response	1.7		1.5
	ITGB7	Integrin, β 7	Cell adhesion/integrin-mediated signaling	NS		1.4
	ITIH4	inter-α (globulin) inhibitor h4	Hyaluronan metabolic process	1.5		NS
	KCNJ10	Potassium inwardly rectifying channel, subfamily j, member 10	Potassium ion transport	NS		2.5
	KLF12	Kruppel-like factor 12	Transcription	NS		1.9
	MAGEA3	Melanoma antigen family A, 3		2.8		3.2
	MASP1	Mannan-binding lectin serine peptidase 1	Complement activation/proteolysis	NS		1.5
	MCC	Mutated in colorectal cancers	Cell cycle/signal transduction	1.7		NS
	MEP1A	Meprin a, α (PABA peptide hydrolase)	Proteolysis	NS		1.9
	MKRN3	Makorin, ring finger protein, 3		1.6		1.5
	MNS1	Meiosis-specific nuclear structural 1	Meiosis	2.0		2.0
	MOGAT1	Monoacylglycerol o-acyltransferase 1	Lipid metabolic/biosynthetic process	1.5		NS
	MSRB3	Methionine sulfoxide reductase b3	Protein repair	NS		1.6
	MTMR8	Myotubularin related protein 8	Phospholipid dephosphorylation	1.6		NS
	MYL9	Myosin, light polypeptide 9, regulatory	Regulation of muscle contraction	1.9		2.1
	NOTCH2NL	Notch homolog 2 (Drosophila) N-terminal like	Notch signaling pathway/cell differentiation	1.4		NS
	NPNT	Nephronectin	Cell adhesion/cell differentiation	2.2		2.2
	PAK1	p21/cdc42/rac1-activated kinase 1	Apoptosis/ER-nuclear signaling pathway/cytoskeleton organization	2.2		2.0
	PLAGL1	Pleiomorphic adenoma gene-like 1	Transcription/induction of apoptosis/cell cycle arrest	1.8		2.1
	PLCE1	Phospholipase c, ε 1	Lipid metabolic process/signal transduction/cell proliferation	1.9		1.6
	PLP1	Proteolipid protein 1	Synaptic transmission	1.5		1.7
	PPL	Periplakin	Keratinization	NS		1.8
	PRKCE	Protein kinase c, ε	Apoptosis/intracellular signaling cascade	NS		1.6
	PRRG4	Praline-rich gla (g-carboxyglutamic acid) 4		NS		1.8
	PTPRJ	Protein tyrosine phosphatase, receptor type, j	Cell-cell signaling	1.7		2.0
	RCN1	Reticulocalbin 1, ef-hand calcium binding domain		1.7		1.6
(Continued)

Table 3A.

Cont.

Regulated by	Symbol	Gene	Biological process	Fold change
				−T3		+T3
	RNF157	Ring finger protein 157		NS		1.7
	SALL2	sal-like 2	Transcription	NS		2.0
	SERPIND1	Serpin peptidase inhibitor, clade d, member 1	Chemotaxis/blood coagulation	1.3		1.3
	SERPINF2	Serpin peptidase inhibitor, clade f, member 2	Acute-phase response	1.4		1.4
	SERPING1	Serpin peptidase inhibitor, clade g, member 1	Complement activation	2.2		2.3
	SIRPA	Protein tyrosine phosphatase, nonreceptor type substrate 1	Cell adhesion	NS		2.0
	SLC22A7	Solute carrier family 22, member 7	Ion transport	2.2		2.1
	SLC6A12	Solute carrier family 6, member 12	Neurotransmitter transport	1.9		2.4
	SLCO2B1	Solute carrier organic anion transporter family, member 2b1	Ion transport	NS		2.4
	SPARC	Secreted protein, acidic, cysteine-rich	Transmembrane receptor protein tyrosine kinase signaling pathway	2.4		2.3
	TEAD2	TEA domain family member 2	Transcription	1.5		1.5
	TGM4	Transglutaminase 4	Protein amino acid polyamination	1.5		NS
	THRA	TRα	Transcription	NS		1.9
	TIMP3	TIMP metallopeptidase inhibitor 3	Visual perception/apoptosis	NS		1.6
	TMEFF1	Transmembrane protein with EGF-like and two FSH-like domains 1	Multicellular organismal development	1.9		2.1
	TNS1	Tensin 1	Intracellular signaling cascade	NS		1.9
	TRIM2	Tripartite motif-containing 2		1.8		1.7
	TRIM73	Tripartite motif-containing 50b		2.6		NS
	TRIM74	Tripartite motif-containing 50c		NS		1.4
	TSPAN7	Tetraspanin 7	Protein amino acid N-linked glycosylation	1.7		2.1
	TTC7B	Tetratricopeptide repeat domain 7b		NS		1.8
	WNT11	Wingless-type MMTV integration site family, member 11	Signal transduction/Wnt receptor signaling	NS		2.1
	WT1	Wilms tumor 1	Transcription/cell cycle	1.6		1.7
	ZAK	Sterile α motif and leucine zipper containing kinase azk	Cell cycle/DNA damage checkpoint/cell death/proliferation/differentiation	1.6		1.5
				TRα1		TRβ1
				−T3	+T3	−T3	+T3
TRα1/TR1	ARHGEF9	cdc42 guanine nucleotide exchange factor (gef) 9	Regulation of ρ protein signal transduction	2.1	1.7	1.7	NS
	AYTL1	Acyltransferase like 1	Phospholipid biosynthetic process	1.4	1.5	1.8	1.8
	CES1	Carboxylesterase 1	Metabolic process/response to toxin	1.9	2.2	NS	1.6
	CHN1	Chimerin 1	Intracellular signaling cascade	3.9	4.8	2.2	2.2
	CNNM1	Cyclin m1	ion transport	2.0	1.9	1.5	1.5
	ENG	Endoglin	Cell adhesion/regulation of transcription	NS	3.3	NS	1.9
	ENPP2	Ectonucleotide pyrophosphatase/phosphodiesterase 2	Cell motility/chemotaxis/lipid catabolic process	3.6	5.0	2.0	2.3
	FLJ21986			2.2	2.1	2.0	1.9
	FYN	fyn Oncogene related to src, fgr, yes	T cell receptor signaling pathway	1.8	4.0	NS	1.8
	HEPACAM	Hepatocyte cell adhesion molecule	Cell adhesion/cell cycle/cell growth	2.0	1.7	2.1	1.8
	HIST1H2AH	Histone cluster 1, H2ah	Nucleosome assembly	3.5	2.8	2.2	2.1
	HIST1H2BK	Histone 1, h2bk	Nucleosome assembly	6.6	9.7	NS	6.8
	HTR1D	5-Hydroxytryptamine (serotonin) receptor 1d	Signal transduction	1.9	2.3	1.9	2.1
	INS	Insulin	Glucose metabolic process/cell-cell signaling/cell death	NS	1.4	NS	1.4
	KLHL13	Kelch-like 13		2.0	2.1	2.3	2.0
	MUC15	Mucin 15		1.9	2.0	1.9	1.9
	MX1	Myxovirus resistance 1, interferon-inducible protein p78	Immune response/signal transduction	1.4	1.5	1.5	NS
	MX2	Myxovirus resistance 2	Immune response	1.6	1.6	1.6	1.6
	NAT8 liter	Hypothetical protein loc339983	Metabolic process	1.8	2.1	1.6	1.8
	OLFML1	Olfactomedin-like 1		3.3	2.6	2.5	NS
	PRKCH	Protein kinase c, η	Intracellular signaling cascade	2.4	2.2	2.1	1.7
	PTAFR	Platelet-activating factor receptor	Chemotaxis/immune response/signal transduction	NS	2.7	NS	1.8
	RNASE2	Ribonuclease, rnase a family, 2	RNA catabolic process/chemotaxis	NS	1.5	NS	1.6
(Continued)

Table 3B.

Cont.

Regulated by	Symbol	Gene	Biological process	Fold change
				−T3		+T3
	SERPINB9	Serpin peptidase inhibitor, clade b, member 9	Antiapoptosis/signal transduction	NS	2.5	NS	2.0
	SLAIN1	Hypothetical protein flj30046		2.7	2.7	2.0	NS
	SLC37A2	Solute carrier family 37, member 2	Transport	NS	3.1	NS	2.1
	SULT1C2	Sulfotransferase family, cytosolic, 1c, member 1	Amine metabolic process	3.2	2.9	2.3	NS
	TRIB2	Tribbles homolog 2	Regulation of MAPK activity	2.2	2.9	1.9	2.3
	UGT2B4	UDP glucuronosyltransferase 2 family, polypeptide b4	Estrogen/xenobiotic metabolic process	3.6	4.1	2.5	2.9
	VCAN	Chondroitin sulfate proteoglycan 2	Cell adhesion/cell recognition	5.5	5.3	3.6	2.9
					−T3	+T3
TRβ1	BBS5	Bardet-Biedl syndrome 5	Visual perception/response to stimulus		1.6	NS
	ENPP4	Ectonucleotide pyrophosphatase/phosphodiesterase 4	Nucleotide metabolic process		NS	1.6
	HERC5	HECT domain and rld 5	Regulation of cyclin-dependent protein kinase activity/ubiquitin cycle		1.7	NS
	NEGR1	Neuronal growth regulator 1	Cell adhesion		NS	1.5
	NRCAM	Neuronal cell adhesion molecule	Central nervous system development/cell-cell adhesion		1.8	1.8
	PCK1	Phosphoenolpyruvate carboxykinase 1	Gluconeogenesis/lipid metabolic process		NS	3.0
	SGK	Serum/glucocorticoid-regulated kinase 1	Apoptosis/sodium ion transport		NS	2.0
	ZNF185	Zinc finger protein 185			1.8	1.7

Genes were identified at a Benjamini-Hochberg adjusted P value <0.05. NS, Not significant (difference from control did not meet a P value cutoff of < 0.05). 

These data suggest that TRα1 and TRβ1 can not only regulate genes in a T₃-responsive manner but also in a hormone-independent fashion. This hormone-independent regulation occurs upon introduction of the TR irrespective of the presence of T₃; in certain cases this regulation by receptor is truly neutral to hormone, whereas in other cases the effect of adding the receptor in the absence of T₃ is further modified by the addition of T₃. Intriguingly, somewhat greater isoform-specific regulation was observed for these hormone-independent target genes (Fig. 5, B and C, and Fig. 6) than for the target genes identified through their T₃ responsiveness (Figs. 2, 3, and 6).

Discussion

TRα1 and TRβ1 regulate a shared panel of target genes

TRα1 and TRβ1 represent the products of an ancestral gene duplication and divergence that occurred early in the vertebrate lineage approximately 550 million years ago (38,39); their maintenance as separate loci in all extant gnathostomes (including all cartilaginous and bony fish, amphibians, reptiles, and mammals examined) powerfully suggests that they play nonredundant functions in evolutionary terms. In fact, there is much stronger sequence conservation between the same isoform compared among different species than between the TRα1 and TRβ1 isoforms compared within the same species (1,3,11,33,34,35). Their patterns of expression are also indicative of distinct biological functions: TRα1 is expressed very early in embryonic development and is found in virtually all tissues, whereas TRβ1 expression arises later in fetal development, cotemporal with the appearance of T₃ hormone, and exhibits a more organ-specific pattern and distinct diurnal cycle than does TRα1 (1,3,11,33,34,35,55). Parsing these characteristics together with those of other nuclear receptors suggests that TRα1 maps to a receptor clade primarily involved in regulation of central nervous system, circadian and basal metabolic functions, whereas TRβ1 maps to a distinct clade responsible for control of lipid metabolism and energy utilization (56,57).

Our goal in the current study was to determine whether TRα1 and TRβ1 play their different biological roles primarily by regulating different panels of target genes. We explored this question by using a hepatocyte-derived cell line, HepG2, that is innately T₃ responsive, yet which, through overexpression of one or the other isoform, could be converted into either a TRα1-dominant or TRβ1-dominant background. We chose these cells, rather than a cell line entirely lacking an endogenous T₃ response, to ensure that all the components required for appropriate TR-mediated regulation were present. Confirming the validity of this approach, our HepG2 transformants behaved as expected, expressing the ectopic TR isoform at levels 4 to 5 times that of the endogenous receptor, regulated an artificial reporter gene in an appropriate manner, and displayed a target gene response readily distinguishable from that observed in the absence of an ectopic receptor.

We first identified genes that demonstrated a statistically relevant response, up or down, to T₃, intentionally choosing a relatively stringent, adjusted P value cutoff of <0.05. Notably, even at this first pass, the majority of T₃-regulated genes were flagged by the array analysis as regulated by both TRα1 and TRβ1 (23 of 43). Detailed further analysis indicated that 23 was, in fact, an underestimate and that most of the genes initially identified as specific for a given isoform were actually regulated by both TRα1 and TRβ1, but to different extents. These results were confirmed for representative target gene candidates using RT-PCR.

It is interesting that most of the genes identified as T₃ responsive in our array experiments were up-regulated by this hormone; SLC2A1 was the only gene flagged by our approach as down-regulated (using the Benjamini-Hochberg adjusted P-value <0.05) by T₃. However, use of a less stringent statistical approach flagged additional candidate genes as down-regulated by T₃ in our TRα1 or TRβ1 transformants; the absence of these genes from our original analysis appears to reflect a lesser dynamic range for down-regulation compared with up-regulation. It should be noted that we purposely chose conditions to minimize false positives and to focus on genes that were specifically regulated as a consequence of introduction of the ectopic TR isoforms; therefore, the list of T₃-regulated genes presented in this study is not meant to be comprehensive either for up- or down-regulated genes.

A significant number of the TR target genes identified in our analysis study have been previously found to be T₃ regulated in prior microarray or gene-specific studies, although many are newly identified by our current characterization, and some previously identified were not flagged in our own study (5,58,59,60,61,62,63,64,65,66,67,68). Differences between our array analysis and previously published array studies on T₃-responsive genes are likely to be due to differences in the time course chosen (we employed 6 h of T₃ treatment to focus on primary response genes, but at the risk of missing additional target genes only detectable at later time points), differences in the statistical analysis (we used a highly stringent approach), and/or differences in the specific target cells examined (e.g. primary liver vs. cultured HepG2 cells).

Despite regulating a largely overlapping set of target genes, TRα1 and TRβ1 mediate isoform-specific, individually tailored, levels of transcriptional activity once recruited to a given target gene

Although regulating a mostly shared set of target genes, it was nonetheless apparent that the ability of TRα1 and TRβ1 to activate any given target gene in response to T₃ did display isoform specificity. TRα1 was a stronger transcriptional activator in the HepG2 cells than was TRβ1 on a consensus DR4 reporter and on several endogenous target genes. However, the TRα1 ≫ TRβ1 response clearly was not universally the case, and the relative ability of TRβ1 to activate a given target gene in response to T₃ ranged from virtually undetectable to slightly greater than that of TRα1. Although modest differences in the expression level of TRα1 and TRβ1 at the protein level might alter the precise ratio of the response at each target gene, it is clear that each isoform displays a unique ability to regulate each target gene. We suggest that it is this gene-specific, isoform-specific amplitude of transcriptional regulation that is the likely basis for the appearance and maintenance of TRα1 and TRβ1 over evolutionary time. In essence, TRα1 and TRβ1 can adjust the magnitude of the transcriptional response at different genes to different levels in much the same way as an audiographic equalizer permits the volume to be individually adjusted at different frequencies. By altering the ratio of TRα1 to TRβ1 in different tissues or at different developmental times, the magnitude of the T₃ response can be individually tailored for different target genes, customizing the hormone response under different conditions to different physiological needs.

This proposal is consistent with observations that although the DNA-binding domain is strongly conserved between TRα1 and TRβ1, the transcriptional regulatory domains (denoted AF-1 and AF-2 (activator proteins 1 and 2) located in the receptor N- and C-terminal regions, respectively] are much more highly divergent between the two isoforms (4,6,7,8,9,10,11). It is likely that by recruiting distinct sets of coregulators to TRα1 and β1, and by working combinatorially with other transcription factors arrayed on each promoter, these divergent AF-1 and AF-2 domains confer the target gene-specific regulatory functions observed in our experiments. As a result, the magnitude of each transcriptional response by TRα1 and TRβ1 may also vary in different cell types that express different repertoires of coregulator proteins.

It is worth noting that our analysis of TRα1 and TRβ1 revealed less isoform specificity in their target gene repertoire than has been found in otherwise similar studies of several steroid receptors. Estrogen receptor (ER) α and β, for example, up-regulated 34 shared genes, 33 ERα-unique genes, and 87 ERβ-unique genes in transfected U2OS cells (69). Of 94 genes regulated in response to progesterone in transfected T47D cells, 25 were jointly regulated by both progesterone receptor (PR) A and B, four were uniquely regulated by PR-A, and 65 were uniquely regulated by PR-B (70). The basis for this difference between the TRs and the reproductive steroid receptors is not clear, but it may be relevant that ERs and PRs have larger, more complex A/B domains than do TRs, providing more versatile platforms for target- and isoform-specific coregulator recruitment (and regulation) by the former. ERs and PRs also are known to participate in an extensive network of protein-protein interactions that contribute to their target gene specificity (e.g. Ref. 71); isoform-specific differences in these protein-protein interaction sites may result in a more strict isoform-specific access to or regulation of genes by ERs or PRs compared with TRs.

TRα1 and TRβ1 can also repress or activate a subset of target genes regardless of hormone status

The TRs have traditionally been viewed as repressing transcription below basal levels in the absence of T₃ and activating above basal levels in the presence of T₃. However, this bipolar mode of regulation is not the only one possible (e.g. Refs. 72,73,74). By parsing our analysis to compare genes regulated by the introduction of the ectopic receptors independent of hormone status, we were able to further identify target genes that were repressed in both the absence and presence of hormone. We also were able to identify other target genes that were activated in both the absence and presence of hormone. The regulation of certain of these genes appeared to be truly hormone independent; for others, T₃ further modified the transcriptional response but did not change its fundamental direction relative to that seen in the absence of ectopic receptor. Presumably these distinct transcriptional regulatory responses reflect the differing contributions of the TRα1 and TRβ1 AF-1 (hormone-independent) and AF-2 (hormone-dependent) domains when recruited to these different target genes, and the interactions of these TR activation domains with coregulators, with other transcription factors, and with epigenetic marks present at each target gene location.

Many of the genes that were regulated by TRs independent of hormone displayed a somewhat stronger isoform specificity than did the genes flagged by virtue of their T₃ response. For several of the target genes repressed by receptor regardless of T₃ status, this isoform specificity was near absolute (e.g. CDCA7L). Repression by nuclear receptors is often mediated by inhibitory interactions with other transcription factors that are conferred by protein-protein interactions, rather than by direct DNA binding. This protein-protein mode of inhibition is likely to be particularly dependent on, and therefore to reflect the sequence differences between, the AF-1 and AF-2 domains of TRα1 and TRβ1.

The isoform specificity observed in the TR transcriptional response, rather than in the TR target gene repertoire per se, helps explain the partial functional redundancy observed for TRα1 and TRβ1

Targeted disruptions of murine TRα vs. TRβ clearly produce different phenotypes: TRα knockouts display defects in linear growth, bone deposition, and cardiac function, whereas TRβ knockouts present with a distinct phenotype of elevated circulating T₃/T₄ (due to loss of a negative feedback regulation in the hypothalamus/pituitary/thyroid gland axis), defects in inner ear and cone cell differentiation, and altered hepatic metabolism (1,33,35,37,42,43,44,45). Despite this clear division of biological function between TRα and TRβ, it is notable that neither knockout is lethal, and that many of the physiological and developmental roles attributed to a given TR isoform remain detectable, if impaired, in the knockout of the corresponding isoform. Similarly, naturally occurring dominant-negative mutations in TRα or TRβ are found in several human diseases (75); these mutant receptors inhibit in trans not only their own, cognate isoform, but also the corresponding paralog, and the resulting neoplastic and endocrine disorders present clinically with defects that are a blend of both isoform-specific and nonspecific characteristics.

This pattern of partial redundancy is not readily explained if TRα1 and TRβ1 regulate rigidly separate sets of target genes, but is fully compatible with the differentially programmed graphic equalizers concept we suggest here. In this model, knocking out either TRα or TRβ would not halt the overall concert of gene regulation involved in differentiation and physiology: most genes regulated by one isoform would still be recognized and regulated by the other. Nonetheless, the resulting music will be distorted by being filtered through only one, instead of both, TRα and TRβ equalizers, and the resulting organism, although viable, will display a range of developmental and regulatory pathological impairments.

Our results also complement and may reconcile two prior elegant studies examining the effects of TR isoform-specific knockouts on T₃-regulated liver gene expression. In Yen et al. (74), the authors reported an overall reduction in the T₃ transcriptional response in TRβ knockout mice vs. wild-type or TRα knockout animals, but observed little or no target gene-specific effects resulting from disruption of either isoform. In contrast, Flores-Morales et al. (60) reported that a specific subset of genes regulated by T₃ in wild-type mice were significantly impaired in their regulation in TRβ knockout animals. These results foreshadow our own conclusions that TRα1 and TRβ1 regulate largely overlapping sets of target genes, but to a significantly different extent. It is likely that both prior studies were observing the same fundamental phenomenon but obtained incompletely congruent results due to technical differences in their approaches. We note also that compensatory changes known to occur in genetically manipulated animals may have attenuated in the knockout mice the isoform-specific differences that we observe in the transcriptional activities of TRα1 and TRβ1 in cultured cells.

Although we focused this study on a comparison of target genes regulation by TRα1 and TRβ1 (the most abundant and widely expressed TR isoform), the TRα and TRβ loci generate additional isoforms through alternative mRNA splicing. TRα2, for example, contains the TRα1 N-terminal and DNA-binding domains, but possesses a unique C-terminal domain that cannot bind T₃. TRα2 appears to operate in most studies as a general inhibitor of TR function (e.g. Ref. 76). However, our observations that TRα1 and TRβ1 can exert hormone-independent, target gene-specific transcriptional properties raise the possibility that TRα2 may similarly display hormone-independent activating functions on a subset of target genes (perhaps mediated by the A/B domain that is shared by both TRα1 and TRα2). Conversely, the alternative splice product of the TRβ locus, TRβ2, contains the same DNA-binding and hormone-binding domains of TRβ1 but possesses a distinct N-terminal A/B domain. We have already shown that TRβ2 displays very different transcriptional properties compared with either TRα1 or TRβ1 (77,78,79); our current study suggests that the TRβ2 transcriptional program is likely to also be gene specific, a question we are currently exploring.

Materials and Methods

Molecular clones

Molecular clones of human TRα1 and TRβ1, described elsewhere (80), were first subjected to standard PCR using oligonucleotide primers designed to incorporate the restriction sites NotI and SalI at 5′- and 3′-ends, respectively. The amplified products were subsequently subcloned into the pBI-L vector within the NotI and SalI multiple cloning site plasmid (CLONTECH Laboratories, Inc., Mountain View, CA). The TRs were then subcloned from pBI-L into the pCI-neo vector using NheI (5′) and SalI (3′) restriction sites (Promega Corp., Madison WI). The thymidine kinase promoter (TK)-luciferase reporter bearing a direct-repeat 4 TRE (DR4-tk-luciferase) was previously described (80). Molecular clones of the human TRα1 and TRβ1 were also subjected to PCR to incorporate the restriction sites SacII (5′) and EcoRI (3′) and subcloned into the pUHD10–3 vector (81) for use as qRT-PCR standards. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH)-pOTB7 (IMAGE clone 2989348) was obtained from American Type Culture Collection (Manassas, VA).

Stable transformants

HepG2 cells were maintained at 37 C in DMEM supplemented with 5% fetal bovine serum using bicarbonate buffer and a 5% CO₂ atmosphere. HepG2 transformants expressing ectopic TRα1, ectopic TRβ1, or an empty plasmid control were generated as previously described using a G418 coselection methodology (80). Cell clones arising from individual transformation events were propagated and screened for integration and expression of the expression plasmid by PCR/RT-PCR. To minimize any clone-to-clone variation, six, 12, and 11 independent cell clones, scored as positive for the presence/expression of the introduced construct and (excluding the vector-only transformants) for TR expression, were pooled for the TRα1, TRβ1, and empty plasmid control cell populations, respectively, and used in the expression analysis. Analysis of individual clones revealed some clone-to-clone variation in the overall strength of the T₃ responsive, but confirmed that the pooled populations accurately recapitulated the patterns of specific target gene regulation observed in individual clones.

Transient transfection assays

The pooled HepG2 stable transformant populations were subsequently analyzed for T₃ responsiveness by introduction of a luciferase reporter using an Effectene transient transfection protocol (80). Briefly, HepG2 cells expressing TRα1, TRβ1, or the empty plasmid control were switched to DMEM containing 10% serum-stripped fetal bovine serum. Following the manufacturer’s recommendations (QIAGEN, Valencia CA), 50 ng of the DR4-tk-luciferase reporter, 50 ng of pCH110-lacZ (as an internal transfection control), and sufficient pUC18 to bring the total DNA to 250 ng were mixed with 2 μl of Enhancer and 2.5 μl Effectene and were added to 7 × 10⁴ HepG2 cells. The transfection medium was replaced 24 h later with fresh medium containing either T₃ or ethanol carrier alone, as indicated. Luciferase and β-galactosidase activities were determined 24 h later, and relative luciferase activity was calculated.

Microarrays

The HepG2 transformant pools expressing ectopic TRα1, ectopic TRβ1, or the empty plasmid control were treated with 100 nm T₃ or with ethanol carrier alone for 6 h in DMEM containing 10% hormone-stripped fetal bovine serum. Separate biological replicates were performed on different days to generate independent samples for analysis. The cells were harvested and the RNA was isolated using an RNeasy kit (QIAGEN). The purified RNA was provided to the University of California, Davis, Cancer Center Gene Expression Resource, for subsequent cDNA probe generation and hybridization to Affymetrix GeneChip Human Gene 1.0 ST microarrays (Affymetrix, Inc., Santa Clara, CA). After scanning, the raw microarray data were normalized by the Robust Multichip Array method, with P value adjustments performed by the Benjamini-Hochberg method using R software and the affylmGUI package (82,83). Heat maps were generated with GenePattern (84). Three independent biological repeats were analyzed for each of the three transformant pools (empty plasmid control, TRα1, and TRβ1).

RT-PCR

RNA was isolated from cells using either TRIzol (Invitrogen, Carlsbad, CA) or RNeasy kit (QIAGEN) following the manufacturer’s protocols. cDNA was synthesized from 1 μg of each RNA preparation using a QuantiTect kit (QIAGEN) and quantitative real-time PCR was performed using 0.5 μl of the resulting cDNA and the ABsolute QPCR SYBR Green Mix kit (ABgene/Thermo Scientific, Epsom Kent, UK) with the following primers. TRα: sense, 5′-CAAGGTGGAGTGTGGGTCA-3′; antisense, 5′-CCGTTCTTTCTTTTTCGCTTT-3′; TRβ1: sense, 5′-ATGACTCCCAACAGTATGACAG-3′; antisense, 5′-TCCAGTCGTGTTCTCGGTCT-3′; and GAPDH: sense, 5′-GAAGGTGAAGGTCGGAGTC-3′; antisense, 5′-GAAGATGGTGATGGGATTTC-3′. Serial dilutions of TRα1 and TRβ1 plasmid clones served as copy number standards. Semiquantitative PCR was performed using 0.5 μl cDNA, GoTaq DNA Polymerase and run for 25 cycles with the following primers. COQ10A: sense, 5′-TGTGCCCTGGTGTAAGAAG-3′; antisense, 5′-TCAGTACAAACAGCCTTGACC-3′; PTAFR: sense, 5′-CAGAGACACACGGTCACTGC-3′; antisense, 5′-GTCCATGTGGGAGGAGTCAT-3′; IRF1: sense, 5′-GACCCTGGCTAGAGATGCAG-3′; antisense, 5′-AGGCATCCTTGTTGATGTCC-3′; PROM1: sense, 5′-ACTCCCATAAAGCTGGACCC-3′; antisense, 5′-TTTGGATTCATATGCCTTCTGT-3′; PCK1: sense, 5′-GAGAAAGCGTTCAATGCCAG-3′; antisense, 5′-ATGCCGATCTTTGACAGAGG-3′; SLC6A12: sense, 5′-CAATGGACCAACAAGATGGA-3′; antisense: 5′-GCTCCACCTCCGTTTTTGTA-3′; DOCK9: sense, 5′-GGAGACTCGGAAGTTCACCC-3′; antisense, 5′-GTGGCTCAATTAGCTTTGGC-3′; CDCA7L: sense, 5′-TAGGAAGAATGGAGTTGGCG-3′; antisense, 5′-TCTGACGAGAGGGTTTCCAT-3′; HEPACAM: sense, 5′-CGCCTTGCTCCTTTTGTCTA-3′; antisense, 5′-ACTGCACAGAAAGCAGAGCC-3′; ZNF185: sense, 5′-AGCGGATCTGAGCAACTTGT-3′; antisense, 5′-GTACAGCTGGGTCTCCTTGG-3′. PCR products were subsequently analyzed on 8% Tris-acetate-EDTA polyacrylamide gels and visualized with SYBR Green. Bands were quantified with a Flurochem900 Imager (Alpha Innotech, San Leandro, CA). Very good matches were observed between the microarray and RT-PCR methods, with the former often, in fact, the more precise of the two approaches.