Figures
Abstract
There are two homologous thyroid hormone (TH) receptors (TRs α and β), which are members of the nuclear hormone receptor (NR) family. While TRs regulate different processes in vivo and other highly related NRs regulate distinct gene sets, initial studies of TR action revealed near complete overlaps in their actions at the level of individual genes. Here, we assessed the extent that TRα and TRβ differ in target gene regulation by comparing effects of equal levels of stably expressed exogenous TRs +/− T3 in two cell backgrounds (HepG2 and HeLa). We find that hundreds of genes respond to T3 or to unliganded TRs in both cell types, but were not able to detect verifiable examples of completely TR subtype-specific gene regulation. TR actions are, however, far from identical and we detect TR subtype-specific effects on global T3 response kinetics in HepG2 cells and many examples of TR subtype specificity at the level of individual genes, including effects on magnitude of response to TR +/− T3, TR regulation patterns and T3 dose response. Cycloheximide (CHX) treatment confirms that at least some differential effects involve verifiable direct TR target genes. TR subtype/gene-specific effects emerge in the context of widespread variation in target gene response and we suggest that gene-selective effects on mechanism of TR action highlight differences in TR subtype function that emerge in the environment of specific genes. We propose that differential TR actions could influence physiologic and pharmacologic responses to THs and selective TR modulators (STRMs).
Citation: Lin JZ, Sieglaff DH, Yuan C, Su J, Arumanayagam AS, Firouzbakht S, et al. (2013) Gene Specific Actions of Thyroid Hormone Receptor Subtypes. PLoS ONE 8(1): e52407. https://doi.org/10.1371/journal.pone.0052407
Editor: Wendong Huang, Beckman Research Institute of City of Hope, United States of America
Received: August 17, 2012; Accepted: November 14, 2012; Published: January 3, 2013
Copyright: © 2013 Lin et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Thyroid hormone (TH) receptors (TRs α and β) are highly homologous transcription factors which transduce signals of active forms of TH (predominantly tri-iodothyronine, T3) [1], [2]. Like other nuclear receptors (NRs), TRs bind specific DNA response elements (TREs) comprised of degenerate repeats of the sequence AGGTCA, usually as heterodimers with retinoid X receptors (RXRs). From these locations, the TRs recruit coregulator complexes that influence gene expression and T3 modulates transcription by inducing conformational changes in the receptor C-terminal ligand binding domain which, in turn, alters the complement of TR associated coregulators [3]–[5]. Despite similarities in structure and function, analysis of TR gene knockout mice and human patients with TRα or TRβ mutations has revealed that the two TRs display unique activities in vivo [6]–[11]. TRα plays major roles in regulation of heart rate and muscle whereas TRβ modulates serum cholesterol levels via actions in liver and feedback inhibition of TH production through the hypothalamic-pituitary-thyroid axis. TRs also exhibit subtype-specific effects in regulation of basal metabolic rate, bone development and other processes.
Differential effects of TRs have commonly been attributed to variations in TR expression levels in target tissues. Liver, for example, expresses TRβ and TRα in a 5∶2 ratio and this could explain the predominant role of TRβ in cholesterol metabolism [12], [13]. Indeed, comparison of T3 effects in hypothyroid wild type, TRα, TRβ and double TR knockout mouse livers failed to identify truly TR subtype selective genes [14]. It is noteworthy, however, that other closely related NRs, including estrogen receptors (ERs) α and β, regulate quite different gene sets [15]. Given that current approaches to selectively modulate TH signaling pathways have often focused on development of TR subtype selective modulators (STRMs) [12], it is important to assess the extent of TR subtype-specific effects on gene regulation.
Several pieces of evidence suggest that TRα and TRβ actions may not be absolutely identical at the level of individual target genes. First, stably transfected TRα and TRβ regulate the same set of around 40 target genes in HepG2 liver cells, but with significant subtype specificity in magnitude of T3 response at individual genes [16]. Second, TRα and TRβ exhibit similar effects at transfected reporters with the most common form of TRE (direct repeat spaced by 4 nucleotides, DR-4), but TRβ is more active at inverted palindromic (IP) TREs [17]. Third, detailed qPCR analysis of expression of liver genes in TR knockout mice has revealed apparently exclusively TRβ-dependent T3 regulated genes, including thrsp and angptl3 [18]. Fourth, TR subtypes exhibit differential effects on ABCD2 gene via indirect differential effects on expression of the transcription factor SREBP1c [19]. Fifth, a constitutively unliganded TRα mutant represses transcription of the liver C/EBPβ gene more efficiently than an equivalent TRβ mutant, via differential corepressor recruitment [20]. Finally, TRβ selectively represses the thyroid stimulating hormone (TSH) gene in cultured pituitary cells even though both TRs are present and TRα is functional when expression of TRβ is knocked down [21].
In addition to TR subtype specificity, there is evidence that activity of both TRs is highly gene-context dependent [1], [14]. T3 can either activate or repress transcription of target genes but, within this framework, there are differences in contributions of unliganded and liganded TRs. Commonly, unliganded TRs suppress T3-inducible genes and agonists induce transcription by both reversing the inhibitory effects of unliganded TRs and eliciting further transcriptional activation. There are, however, cases in which unliganded TRs fail to suppress T3 induced genes or where T3 activates genes solely by relieving unliganded TR-dependent basal repression [14], [16]. Similar gene-specific variations in the balance of unliganded and liganded TR actions also occur at negatively regulated genes and TRs can even exhibit completely ligand-independent actions [14], [16]. T3 concentration dependence also varies; euthyroid T3 levels are sufficient for optimal induction of some genes whereas others need higher (hyperthyroid) T3 levels [14], [22].
Here, we set out to define the extent of TR subtype and gene-specific variations in transcriptional response by creating cells with similar levels of TRα or TRβ. While our data suggests that there are no completely TR subtype specific genes, we observed TR-subtype and cell-specific effects on the kinetics and magnitude of transcriptional response, patterns of TR-dependent gene expression and T3 concentration dependence that verify and extend conclusions of previous groups. We discuss possible mechanisms of these differential effects and their impacts upon physiological responses to THs and actions of STRMs.
Materials and Methods
Construction of HepG2-TR Cells
HepG2 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 U/mL of penicillin, 0.1 g/L of streptomycin and 4 mmol/L glutamine, under 95% air and 5% CO2 at 37°C. Lentiviral vector pSicoR containing GFP was a gift from the McManus laboratory (UCSF, http://mcmanuslab.ucsf.edu/). Sequences for appropriate tags were added to TR cDNAs derived from previously described eukaryotic expression vectors by standard PCR amplification and resulting hybrid cDNAs were cloned into pSicoR vector at a location C-terminal of GFP. An oligonucleotide encoding a T2A peptide that will mediate co-translational cleavage of protein was inserted between GFP and TRs coding sequences to facilitate independent expression of GFP and TRs. pSicoR vectors containing tagged TRs were transfected into HEK 293 cells to create packaged virus particles by the Diabetes center core lab at UCSF. The virus was titrated by GFP fluorescence. HepG2 cells grown on a 6-well plate were transduced with lentivirus particles containing tagged TR sequences. After incubation with virus particles for 72 hours, cell culture media were replaced with regular growth media and GFP expression was verified by fluorescent microscopy. Cells were split upon confluence and sorted by fluorescence-activated cell sorting in UCSF core facility after two passages. Clones of GFP positive cells were collected and grown in 10 cm plates. When cells reached confluence, TR expression was analyzed with Western blots using anti-tag antibodies and transient transfection of TRE driven luciferase reporters, described below. Cells were maintained in regular growth medium as above.
Construction of HeLa-TR cells
HeLa cells were grown in similar conditions to HepG2 cells. HeLa cells stably expressing tagged TRα and TRβ were generated using the Tet-Off gene expression system (Clontech). Stable clones were selected by hygromycin-resistance (400 µg/ml) and screened for TR protein expression. HeLa cells expressing TRβ and TRα were maintained in media supplemented with doxycycline (20 ng/ml).
Western Analysis
HepG2 or HeLa cells were lysed using Triton X-100 lysis buffer. Cell extracts (10 µg of total protein) were separated by SDS-PAGE and transfer onto a PVDF membrane. Transfer membranes were then incubated with anti-flag M2 antibody (Sigma-Aldrich) or c-Myc antibody (Clontech) at a 1∶1000 dilution at 4°C for 16 h followed by goat anti-mouse IgG-horseradish peroxidase antibody (Santa Cruz Biotechnology sc-2004) at a 1∶10000 dilution for 45 min at room temperature. Blots were visualized by applying ECL Plus (GE Healthcare).
Hormone Binding
Hormone binding assays were carried out as described in [23]. The Kd values were calculated using the Graph-Pad Prism computer program (Graph-Pad Software Inc).
Transfection
Cells were co-transfected with a DR-4 or IP-6 TRE-driven luciferase reporter and constitutive renilla luciferase reporter (Promega) using Transfectin Reagent (BioRad) and plated in 12-well plates in growth medium (DMEM with 10% hormone-depleted FBS) [23]. After 16 h of incubation, T3 (100 nM) or vehicle (DMSO) was added in triplicate. After an additional 24 h of incubation, cells were harvested and assayed for luciferase activity using the Promega Dual Luciferase Reporter Assay (Promega). Data were normalized to the renilla luciferase activity.
mRNA and cDNA Preparation
For HepG2, total RNA was prepared using the Aurum Total RNA kit (Bio-Rad). Reverse transcription reactions in these samples were performed using 1 µg of total RNA with an iScript cDNA Synthesis kit (Bio-Rad). Total RNA concentrations were measured using NanoDrop ND-1000 spectrophometer. For HeLa, total RNA was extracted from cells with Qiazol Lysis Reagent (Invitrogen) and purified with RNeasy® Mini kit (Qiagen) following manufacturer's instructions. mRNA was reverse transcribed into cDNA with a mixture Oligo(dT)20 and Random Hexamers (1∶1 ratio) using SuperScript™ III First-Strand Synthesis System for RT-PCR kit (Invitrogen).
Microarray hybridization
Human whole genome expression arrays were purchased from Illumina (Human WG-6v2 and Human WG-6v3). cRNA synthesis and labeling were performed using Illumina® TotalPrep™-96 RNA Amplification Kit (Ambion). Labeling in vitro transcription reaction was performed at 37°C for 14 h. Biotinylated cRNA samples were hybridized to arrays at 58°C for 18 h according to manufacturer's protocol. Arrays were scanned using BeadArray Reader.
Statistical analysis
Unmodified microarray data obtained from GenomeStudio was background-subtracted and quantile-normalized using the lumi package [24] and analyzed with the limma package [25] within R [26]. To determine the effect of TRα and TRβ over-expression in the absence of ligand (“unliganded-effect”), cell lines were analyzed separately by LIMMA (“parental”, with no exogenous TRs, TRα, and TRβ), followed by contrast analysis. To better determine TR isoform effects, factorial LIMMA analysis was conducted comparing ligand (T3) with over-expression of the TRα or TRβ (“TR over-expression with ligand effect”; interaction between T3 and over-expression of TRα or TRβ), followed by contrast analysis. All analysis was corrected for multiple hypothesis testing [27], and effects determined to be significant when ≥2-fold with an adjusted p-value ≤ 0.05 when compared to their respective parental cell line. To facilitate comparisons among the various datasets, all data was uploaded into a SQLite3 database [28]. Heatmaps were produced and clustered using multiarray viewer [29].
qPCR
Real-time qPCR in HepG2 samples was performed with the Roche LightCycler 480 RT PCR Instrument using SYBR Green Mastermix (Roche). The sequences of the primers are listed in Table S1. The data were collected and analyzed using the comparative threshold cycle method. Experiments were performed at least three times, and the mean ± SE was calculated using the Prism curve-fitting program (GraphPad Software, version 3.03; GraphPad). For HeLa, qPCR was performed using SYBR Green PCR Master Mix (Applied Biosystems) on ABI 7900HT RT-PCR system (Applied Biosystems) with default two-step QRT-PCR program. Amplification curves were evaluated by the comparative Ct analyses.
Calculation for categorizations of expression patterns
Custom Python scripts were used to organize the expression patterns into induced and repressed effects, through the calculation of minimal Euclidean distances between nine hypothetical patterns (2-by-3; unliganded and liganded effect expressed as induced (e.g., 2-fold), repressed (e.g., 0.5-fold) or no effect (e.g., 1)) and that experimentally derived. The hypothetical patterns (unliganded, ligand) were: pattern RR = 0.5, 0.5; pattern RO = 0.5, 1; pattern RI = 0.5, 2; pattern OR = 1, 0.5; pattern OO = 1, 1; pattern OI = 1, 2; pattern IR = 2, 0.5; pattern IO; 2, 1; pattern II = 2, 2. Note, hypothetical pattern OO translates to no substantial effect with TR over-expression (unliganded effect) or T3-treatment (ligand effect) and was not included in the final table. All probes with a BH adjusted p-value< = 0.05 within the 3 hr HepG2 and 24 hr Hela treatments were analyzed. To lessen the effect of extremes, fold-change values were first transformed to 2 or 0.5 if their fold-change values were >2 or <0.5, respectively. Euclidean distances were then calculated, and the probe/gene transcript grouped into the hypothetical pattern that delivered the minimal Euclidean distance between the experimental and/or the transformed vector and the hypothetical pattern vector. The genes were subsequently translated from Probe_ID to official gene name using an SQLite database.
Results
Cells that Express Comparable Levels of TRα1 or TRβ1
We created two sets of TR-expressing cells to compare actions of major TR subtypes at endogenous genes. For HepG2 liver cells, we used a retroviral infection to express epitope-tagged (Flag) TRα1 or TRβ1. For HeLa, we used stable transfection to express epitope-tagged (myc) TRs under tetracycline control (tet-off system, Fig. S1). We screened multiple clones of both cell types by western, using antibodies against respective epitope tags to facilitate direct comparisons of protein levels, and identified pairs of cell lines with comparable TRα and TRβ expression (Fig. 1A). We confirmed that exogenous TRs were recognized by antibodies against TR primary sequences, that these TRs were expressed at higher levels than endogenous TRs, which were either present at very low levels (TRβ in HepG2 [22]) or undetectable (TRα in HepG2 and both TRs in HeLa) and that expressed TRs were of correct molecular weight (Fig. S2 and not shown). We also verified that TR expression was stable over several passages (not shown).
A. Equal expression of exogenously expressed TRs. Upper panel, western blot of extracts of HepG2 parental cells (1), HepG2 cells infected with control lentivirus (2) or cells infected with lentivirus expressing TRα (3) or TRβ (4) and blotted with anti-flag antibody. Inset beneath shows the same extracts blotted with a β-actin antibody as a loading control. Lower panel; western blot of HeLa-TR extracts after +/− doxycycline withdrawal to induce TRs and blotted with anti-myc. B. Results of T3 binding assays performed on extracts of HeLa-TR cells after 24 hrs doxycycline withdrawal; figures in panels represent deduced affinities of expressed TRs for T3. C. Results of luciferase assays performed upon HeLa-TR cells transfected with standard TRE-driven reporters, DR-4 Luc and IP-6 (F2)-Luc after doxyclycline withdrawal to induce TR expression. D. Western blot of HepG2-TR extracts at various times after initial T3 treatment.
Exogenously expressed TRs are functional. We verified that the presence of the tag did not affect TR function in transient transfections, in which we compared native and tagged TRs (Fig. S3 and not shown). We were also unable to detect any major differences between the function of exogenously expressed TRs versus previous assessments of endogenous TR function. Hormone binding analysis confirmed that TRs exhibit affinities for T3 that are consistent with previously reported values (Fig. 1B and not shown). This data also allowed us to estimate numbers of T3 binding sites per cell (i.e. TRs), which were between 7-10,000 in transfected cells versus essentially undetectable in parental cells [22] and within physiological range (<10,000 receptors per cell) [30]. Exogenous TRs conferred T3 responses on standard TRE-dependent reporters in both cell types (Fig. 1C and not shown). Further, T3 elicited similar levels of activation with TRα and TRβ at a DR-4 reporter but larger levels of activation with TRβ at an IP-6 reporter, in accordance with previous results [17]. Finally, TR steady state levels were diminished after T3 treatment (Fig. 1D); this phenomenon is common to many NRs and a consequence of ubiquitin-dependent turnover of activated receptors [31]. Interestingly, however, TRα levels were rapidly reduced (within 1 hour of T3 addition) whereas TRβ levels only obviously became diminished after extended ligand treatments.
TRα and TRβ Regulate Similar Gene Sets in HepG2 with Different Kinetics
We performed transcriptome wide analysis of TR target genes in our HepG2-TR cells and parental HepG2 controls after 3, 6 and 24 hr induction with saturating (100 nM) T3. Since HepG2 expresses low levels of endogenous TRβ1, we examined interaction between treatment and cell line (i.e. T3 + TR over-expression) to determine TR-specific effects.
Most T3 responses required exogenous TRs (Fig. 2A–D, Fig. 3). As we previously reported [22], a few genes responded to T3 in parental HepG2 cells, with around 17 meeting cut-offs (>2.0 fold, BH-adjusted P value 0.05) at 24 hrs. This is due to vanishingly low levels of functional TRβ present in HepG2 [22]. By contrast, hundreds of genes responded to T3 in cells that express either of the two TRs (Fig. 2A–D). Of these, the majority (more than 70%) were induced by T3 with the remainder repressed. Additionally, most genes that exhibited T3 responses in parental HepG2 cells exhibited amplified responses in the presence of exogenous TRs. The sole exception was that we discovered expression of the highly T3 responsive ANGPTL4 gene, a verified direct TRβ target in parental HepG2 cells [22], was silenced by exogenous TRs when we performed qPCR analysis (Fig. S4). We confirmed that enhanced T3 responses seen in the presence of transfected TRs in HepG2 were dependent upon exogenous TR expression using an siRNA specific to the 5′ portion of the dual EGFP/TR transcript to inhibit exogenous TR expression (Fig. S5).
A–D. Numbers of genes that meet cut-offs for fold induction and statistical significance in parental HepG2, HepG2-TRα or HepG2-TRβ at each time point, A, 3 hr, B, 6 hr, C, 24 hr, D, all three time points combined. T3 induced genes are represented in upper panels (red) and T3 repressed genes in lower panels (blue), note the difference in scale which means that many more genes are positively regulated than negatively regulated. E–H. Plots of fold induction/repression by T3 in the presence of TRβ (y-axis) versus TRα (x-axis). E, 3 hr blue, F, 6 hr green, G, 24 hr red, H, all three time points.
A. Representation of changes in all T3-dependent genes that meet cutoffs for fold induction/statistical significance in HepG2 parental cells and HepG2-TRα or HepG2-TRβ cells at each time point. Red, upregulated, green, downregulated, black, no change. Genes were clustered according to similarities in response patterns as described in Methods. B. As for Fig. 3A, with a section of the heatmap expanded to reveal one of the clusters of late emerging TRβ-preferential T3 responses.
Unexpectedly, different numbers of genes met cutoffs for fold induction and statistical significance with TRα and TRβ at each of the three times (Fig. 2A–C). More T3 responsive genes appeared with TRα at 3 and 6 hrs (Fig. 2A, B), whereas TRβ responses predominated at 24 hrs (Fig. 2C). Overall, similar numbers of genes exhibited T3 -responses when all three times were considered together (Fig. 2D). Closer analysis revealed no completely TR subtype-specific genes within the datasets, there was a high degree of overlap between TRα and TRβ responsive genes and nearly all genes that responded to T3 with either TRα or TRβ at any of the three time points exhibited qualitatively similar responses with the other TR in at least one time point (Fig. 3A and not shown).
To better understand differential kinetics of T3 response in HepG2-TRα1 cells and HepG2-TRβ1 cells, we compared fold T3 induction/repression of each gene (Probe_ID) in the presence of the two TRs (Fig. 2E–H). Although there were more TRα selective genes at 3 and 6 hrs, there was a strong apparent correlation between fold induction/repression when the two TRs are compared (Fig. 2E, F). Visual inspection (Fig. 3A, not shown) and statistical correlation analysis suggested that many of the apparently TRα-selective genes responded in a similar fashion to TRβ, but that T3 response sometimes failed to meet cutoffs for fold induction and/or statistical significance resulting in the discrepancies between numbers of regulated genes. While there was also apparent correlation between overall TRα and TRβ responses at 24 hrs (Figs. 2G, H; Fig. 3A, B), we observed a shift in slope that reflected an increase in the number of genes (probes) responding to T3 in the presence of TRβ versus TRα, resulting in a deviation from the straight line relationship at the earlier time points. The latter phenomenon reflected emergence of a subset of genes with preferential TRβ responses, although all members of this gene class exhibited qualitatively similar regulation by TRα in at least one earlier time point (see Fig. 3B). We used qRT-PCR analysis to confirm that members of this strongly TRβ-dependent late responding gene set (G6Pc, GSTA1) retained preferential TRβ responses that persisted over multiple T3 incubation times (Fig. 4).
qPCR analysis of T3 induction of two genes (A, G6Pc and B, GSTA1) identified as preferential late TRβ responders. Note that TRα-responses are weak and that TRβ preference persists across all timepoints.
Together, our data suggests that TRs regulate similar gene sets but with different kinetics in HepG2; T3 responses emerge earlier with TRα versus TRβ. Further, a strongly TRβ-dependent subset appears after prolonged T3 treatment.
T3-induced Genes are Direct TR Targets
We determined the proportion of T3 responsive genes that were direct TR targets (Fig. 5). To do this, we examined effects of pretreatment with the protein synthesis inhibitor cycloheximide (CHX) upon T3 response at 3 h and defined direct targets by persistence (at least 100%) of gene expression obtained with T3 after CHX treatment relative to levels obtained with T3 alone. Most (>80%) of the genes that are positively regulated by T3 are scored as direct targets by these criteria (Fig. 5A upper panel). Investigation of remaining positively regulated genes revealed that some T3 activation persisted in the presence of CHX for many of the remaining 20% of genes, suggesting that representation of direct TR targets within this dataset may be even larger than this analysis suggests (Fig. 5B, not shown). Interestingly, a much smaller proportion of negative T3 responses persisted after CHX treatment versus positive responses (Fig. 5A, lower panel). This suggests that novel protein synthesis is needed for T3 repression in this cellular context. In total, CHX treatment only completely abolished T3 response of a small subset of genes (Fig. S6); responses of ≈11% of all genes that displayed 2- or greater fold responses to T3 at 3 hrs were completely inhibited by CHX. These are likely to represent secondary responses to T3-dependent changes in protein levels.
A. Bar graph representing numbers of T3 induced (upper panels) and repressed (lower panels) genes at 3 hrs in HepG2-TRα and HepG2-TRβ cells that persist with CHX pre-treatment (upper panel, red, lower panel, blue). B. Heat map representing gene expression changes at 3 hrs timepoint in HepG2-TRα and HepG2-TRβ cells with T3, CHX and T3 + CHX. Note that most target genes retain their T3 responses with CHX. Examples of genes with unusual responses are marked by lower case letters: a = stronger T3 responses with TRα, b = amplification of weak T3 responses in the presence of TRβ with CHX, c = selective CHX-dependent gene induction in the presence of TRα, d = selective CHX-dependent gene induction in the presence of TRβ.
As seen with the complete dataset, we observed little TR subtype selectivity among direct TR targets (Fig. 5B). Some T3 responsive genes display stronger TRα responses and qualitatively similar but weaker responses with TRβ; this is evident from comparison of columns 1 and 4 in the heat map (some examples of this set of genes marked “a”). However, T3 responses mostly persisted with CHX in the presence of both TRs and we even detected cases of amplification of weak T3 response with TRβ in the presence of CHX (examples of this set of genes marked “b”). There were also gene-specific interactions of TRs and CHX; some genes were selectively de-repressed by CHX treatment alone in the presence of one of the two TRs (TRα selective de-repression is marked “c” and TRβ selective de-repression marked “d”). In general, however, most direct target genes appear similarly regulated by both TRs.
TR/Gene-Selectivity in T3-response
Within broad TR response patterns outlined above, there was gene-specific variability of T3 regulation patterns and we verified some of these observations using RT- qPCR (Fig. 6). Many genes displayed similar time courses of T3 induction with both TRs (PCK1, Fig. 6A), but others exhibited differential responses to the at individual time points (SLC16A6, Fig. 6B) and yet others displayed sustained preferential responses to TRβ (HIF2A; Fig. 6C) or to TRα (MYH6 Fig. 6D). Thus, differences in magnitude and kinetics of T3 response with the two TRs are reflected at the level of the global T3-dependent gene expression program (Figs. 2 and 3) and at individual gene-specific responses (Fig. 6).
Results of qPCR analysis of representative gene expression changes at various times after T3 induction in HepG2-TRα and HepG2-TRβ cells. A, PCK1, similar with both TRs, B, SLCA16A, similar with both TRs at most times, C, HIF2A, TRβ preference at both early and late times, D, Myh6, TRα preference at early and late times.
Similar TRα and TRβ Responses in HeLa cells
To extend comparative analysis of TR subtype effects, we examined T3 responses in HeLa cells that express exogenous TRs. Here, TR expression was induced by doxycyclin withdrawal for 24 hrs; this regimen elicited optimal TR mRNA induction (Fig. S7). We then treated cells +/− saturating T3 (100 nM) for a further 24 hrs.
More genes responded to T3 in the presence of TRs in HeLa cells than HepG2 cells (Fig. 7A). All responses were dependent on exogenous TRs; unlike HepG2 cells, our HeLa cells lack detectable TR protein and transcripts (not shown). Like HepG2 cells, however, the majority of genes were induced by T3 and there was near complete overlap between TRα and TRβ target genes; plots of T3 responses with TRβ versus TRα again revealed apparent correlation between induction/repression for most genes (Fig. 7B). This extends our conclusion that TRα and TRβ responses are broadly similar and also implies that late TRβ-specific effects observed in HepG2 are a feature of the latter cell type.
A. Numbers of genes that meet cutoffs for fold T3 activation (upper panel, blue) or repression (lower panel, red) in HeLa-TRα and Hela-TRβ cells at 24 hrs treatment, as in Fig. 2. B. Plots of fold induction/repression by T3 in the presence of TRβ (y-axis) versus TRα (x-axis) in HeLa cells. C–D. Representative qPCR analysis showing examples of different gene regulation patterns with the two TRs. C, pck1, D, thrsp.
Also as seen in HepG2, we detected genes which exhibit preferential responses to TRα or TRβ and verified some effects with qRT-PCR. PCK1 was strongly induced by T3 with TRβ but not TRα, although both TR subtypes enhanced transcript abundance without ligand (Fig. 7C). PCK1 was strongly T3-dependent in HepG2-TRα and HepG2-TRβ cells (see Fig. 4B) implying that this is a cell-specific effect. More commonly, and similar to HepG2, we observed TR subtype specificity in magnitude of T3 response; for example, the THRSP gene displayed stronger T3 induction with TRβ versus TRα (Fig. 7D).
Finally, there was only limited overlap between T3-regulated gene sets in HepG2 and HeLa (Fig. 8). These differences were not explained by failure to meet cut-offs for fold induction and TRs did regulate different gene sets in the two cell types. This means that strong overlap between TRα and TRβ target genes occurs with two largely distinct gene sets of T3 regulated genes in two cell types.
Venn diagrams of numbers of T3 induced and repressed genes identified in each cell type with TRα and TRβ and overlaps.
Unliganded TRα and TRβ Regulate Similar Gene Sets
Since TRs are transcriptionally active without hormone [3], [4], we compared effects of unliganded TRα and TRβ in both cell types (Fig. 9A–D). To do this, we assessed differences in gene expression in HepG2-TR cells versus HepG2 parental cells and parental HeLa cells versus HeLa-TR cells after 24 hrs doxycyclin withdrawal.
A. Numbers of genes that meet fold cutoffs for activation/repression and statistical significance in response to unliganded TR expression in HepG2 cells, TRα and TRβ expressing cells were compared to parental. B. HeLa cells, TRα and TRβ expressing cells after doxycycline withdrawal to induce TRs versus doxycyclin treated cells. Similar results were obtained in comparisons with parental HeLa cells (not shown). C. Plots of fold induction/repression by TRβ (y-axis) versus TRα (x-axis) in HepG2 cells. D. Plots of fold induction/repression by TRβ (y-axis) versus TRα (x-axis) in HeLa cells.
Unliganded TRs influenced many genes in both experimental systems (Fig. 9A, B). Expression of more than two thousand genes was altered by the presence of unliganded TRs relative to parental controls in HepG2 cells; with similar numbers up- and down-regulated. Large numbers of genes (≈1000) also responded to short term TR induction in HeLa cells and, again, similar numbers of genes were up and down-regulated. More TR-dependent genes met fold cutoff and statistical significance with unliganded TRα versus TRβ in both cell types. As seen with T3 regulation, however, these differences were generally qualitative and not absolute and we observed an essentially linear relationship between induction/repression with unliganded TRα and TRβ in both cell types even though some probe sets suggested preferential response to one of the two TR subtypes (Fig. 9C, D).
Gene-Specific Variations in Pattern of Response to TRs +/− T3
Next, we examined specific patterns of target gene regulation by TRs [14]. To do this, we grouped genes with statistically significant responses to unliganded TR or T3 into categories according to whether they are repressed (R), unaffected (O) or induced (I) relative to basal gene expression levels in parental cells (Methods and Fig. 10/Table 1/Table S2). In this way, TR and T3-dependent genes could be placed into one of eight response patterns shown in the heat map at left. Another category, “OO”, included genes that displayed small changes in response to TR or T3 that reached statistical significance, but was assigned to the non-responsive pattern and are not shown (see Methods).
Genes with statistically significant responses to T3 or unliganded TRs were assigned into categories according to net repression (R), induction (I) or no change (O) represented in the schematic heat map. Numbers of genes in each category and overlaps between genes that respond to TRα or TRβ in this manner are shown in Table 1.
We detected examples of all classes of predicted responses to TR +/− T3 (Fig. 10/Table 1/Table S2). As seen in previous study [16], a small percentage of genes were constitutively repressed or induced by TRs (patterns RR and II), and we verified some observations with qRT-PCR (Fig. 11, MST1, hel308 and also see Fig. S4). Other genes were activated or repressed by T3, with a majority displaying one of several possible patterns of positive T3 response (R0, RI, 0I) and a large minority exhibiting one of several patterns of negative regulation in response to T3 (0R,0I, IR). Distributions of genes between different categories of positive and negative response varied with cell type. In HepG2, large majorities of positively and negatively regulated genes grouped into pattern R0 and pattern I0, respectively. Interestingly, these response patterns were mirror images of each other, with unliganded TR repressing positively regulated genes or activating negatively regulated genes and T3 reversing these effects. By contrast, the response patterns were more evenly distributed in HeLa.
qPCR verification of genes that display hormone-independent repression by both TRs in HepG2. A, mst1, B, hel308.
Although TRs regulated similar numbers of genes in the same way, there was only limited overlap between genes that grouped into the same pattern with TRα and TRβ (Fig. 10). This suggests that relatively subtle differences in magnitude of response to TRs +/− T3 can translate into different response patterns. We confirmed this impression at a limited set of target genes with qRT-PCR (Fig. 12A–D). For example, Myh6, Furin, ALPI and HIF2A are all induced by T3 but Myh6 and furin exhibit the same basic regulation pattern (Fig. 12A, B, pattern 0I), but ALPI is induced by unliganded TRβ (pattern II) and not TRα (pattern 0I) whereas HIF2A is induced by unliganded TRα (Fig. 12C, pattern II)) and not TRβ (Fig. 12D, pattern 0I).
A, Myh6 B, furin. Both genes display the same pattern of response to unliganded TRs and T3, despite preferential T3 induction of Myh6 with TRα. C, ALPI, display exclusively ligand-dependent induction with TRα and ligand-dependent induction with TRβ coupled to a strong ligand-independent component. D, HIF2A, displays the opposite profile to ALPI in Fig. 12C.
TR Subtype and Gene Selectivity in T3 Dose Response
Finally, we performed a limited survey of relationships between T3 dose response, TR subtype and gene (Table 2 and Fig. S8). To do this, we examined effects of varying doses of T3 upon selected TR targets in HepG2 cells. Most genes exhibited EC50 values in the low nM range (≈1 nM), but TRα exhibited some gene-specific requirements for lower T3 concentrations (klf9, pck1) whereas TRβ exhibited gene-specific requirements for higher T3 levels (furin, alpi, myh6, hr). The combination of these effects means that many genes exhibit similar T3 concentration-dependence in the presence of both TRs whereas others exhibited differential responses to T3 with TRα and TRβ, with EC50 values varying up to 30-fold when different genes and response to TR subtype are considered. It was noteworthy that higher levels of T3 were needed for optimal induction with TRβ in most cases of differential concentration dependence.
Discussion
In this study, we compared effects of equal levels of exogenous TRs upon endogenous genes, +/− T3, in different cell backgrounds (HepG2 and HeLa) to determine the prevalence of TR subtype specific genes. While hundreds of genes respond to T3 or to unliganded TRs in both cell types, and there are differences in degree of response +/− T3, we were unable to identify truly TR subtype-specific genes in either cell background; i.e. genes which respond only to one TR subtype and not the other. We conclude that TRα and TRβ regulate the same genes, different from some homologous NR pairs such as the ERs and classic steroid receptors. Our results agree with previous studies of other groups who found that TRs regulate the same genes in mouse liver [14] and in a prior study conducted in HepG2 cells [16].
While TRα and TRβ may regulate the same genes, their actions are far from identical. Part of our analysis overlaps a previous study of similar design by Chan and Privalsky [16], who also stably expressed exogenous TRs in HepG2 cells and found strongly overlapping responses to TRα and TRβ after 6 hrs T3 induction but with gene-specific differences in magnitude of response to the two TR subtypes. While our results do not precisely reproduce previous findings, we regard differences as minor and emphasize that there is remarkable concordance between the key points of the two studies. Chan and Privalsky observed fewer genes that responded to T3 (≈40 versus 150–300) or unliganded TRs (≈100 versus 1500–2000) and detected a stronger bias towards TRα versus TRβ responses. Both discrepancies are probably explained by differences in TR expression levels. Our comparisons of levels of TR transcripts relative to parental cells in both datasets (not shown) suggest that our cells express more TRs, explaining detection of more TR target genes in our stably transfected HepG2 cells. Moreover, evidence described in the Chan and Privalsky study suggests that their cells may express more TRα than TRβ, explaining observed TRα bias. More importantly, however, both groups conclude that TRs regulate the same genes, all genes flagged as T3 regulated in the Chan study appear in our dataset at the 6 hr time point and regulation patterns appear very similar in both analyses (not shown). Our analysis therefore confirms conclusions of this study and, because it also includes extra time points and two cell types, confirms and extends the conclusion that TRs regulate the same genes with differences in precise magnitude of response. Additionally, the fact that we have confirmed that many early responding genes are direct TR targets with CHX treatment indicates that many differential effects must be related to primary differences in direct TR actions.
Our study has also uncovered other ways that TRα and TRβ differ. First, most T3 responses tend to be stronger with TRα at earlier times and with TRβ at 24 hrs, although there are numerous counter-examples of individual genes which deviate from this pattern. Second, a large set of late responding highly TRβ-specific genes appears in HepG2 cells. Third, we find that TR subtype preferences in magnitude of T3 response may appear at selected time points or persist across several time points. Fourth, differences in magnitude of response +/− T3 means that the two TR subtypes exhibit different regulation patterns at individual genes. Finally, we detect variations in T3 dose response in a limited survey of target genes and this effect displays a TR subtype-selective component.
Our experiments do not address mechanisms of differential effects, but do permit speculation about possible causes:
- We suspect that there are fundamental differences in kinetics of T3 activation processes in the HepG2 experimental system. We note that: i) faster transcriptional responses to TRα are paralleled by faster T3-dependent reductions in steady state TRα levels versus TRβ; this often reflects ubiquitin-dependent turnover of transcriptionally active complexes [31] and ii) some verified early direct TRα targets display similar but slower T3 responses with TRβ (see heat map in Figure 5).
- Early T3-dependent changes in gene expression seem to foster an intracellular environment that enhances TRβ actions at some genes; the late HepG2 TRβ-dependent gene set also responds weakly to TRα suggesting that prolonged T3 treatment selectively augments TRβ action in these contexts. One possible explanation for this effect is that TRs may induce differential expression or activity of transcription factors that regulate downstream genes and possibly cooperate with TRs in some contexts.
- We note that TRα and TRβ subtype- and gene-selective actions emerge within the context of wide gene-specific variations in TR action that have also been observed by other groups [14], [16]. We observed that: a) some genes respond to low levels of endogenous TRs in HepG2 parental cells whereas others require exogenous TR expression to mount a detectable response and one gene (ANGPTL4) that is a verified direct TRβ target in parental HepG2 cells [22] is even silenced by TR overexpression, b) magnitude and direction of response to T3 and unliganded TRs varies widely, c) there are gene-specific interactions between CHX and T3, see Figure 5 d) there are variations in response patterns of both activated and repressed genes +/−T3 and this effect displays a cell-specific component and e) T3 dose response is gene-specific. We suggest that these gene-specific variations in response reflect gene-context specific variations in mechanisms of TR action and that some of these, in turn, highlight differences in TRα and TRβ function that are not always apparent from standard reporter assays alone. Elucidation of mechanisms of these effects will require better understanding of gene architecture and TR influences upon transcription factor and cofactor recruitment and we propose that systems described within this paper will help us to dissect influences of gene context upon precise mechanisms of TR action.
Finally, it is important to consider whether gene-specific differential actions of TRα and TRβ also occur in vivo and possible physiologic impacts of such effects. We know from previously published studies of wild type and TR gene knockout mice that some gene-specific TR behaviors that we have been able to document in culture have direct parallels in vivo; these include variations in patterns of TR regulation +/− T3 and in T3 concentration dependence [14]. This implies that at least some of the gene-specific differential effects seen in our study will also be observed in vivo. Indeed, our initial survey of effects of TRβ knockout upon endogenous liver genes in mice revealed that T3 response of all genes is diminished but that there are more severe effects at some genes than others (not shown), implying different contributions of TRα and TRβ to T3 response.
We can imagine two situations in which gene-specific differential TR actions would be physiologically important. First, TRα and TRβ expression is highly influenced by diurnal rhythm [32]. Thus, variations in TR protein levels could affect the TR target repertoire based upon whether a particular target gene responds preferentially to TRα or TRβ; such differential effects would be observed even without changes in T3 levels. Second, TRβ and liver selective thyromimetics (STRMs) such as GC-1 (sobetirome) and KB2115 (eproterome) have been developed to selectively lower serum cholesterol without deleterious effects on heart and combat other aspects of metabolic disease [12]. Some natural TR ligands such as TRIAC are also TRβ selective [23]. Doses of such ligands that activate TRβ but not TRα would be expected to alter gene expression in a TRβ biased manner that would differ from T3, which binds the two TRs with similar affinity. It will be interesting to determine whether any such TR subtype and gene-specific effects occur in vivo and whether their impact is physiologically or pharmacologically important.
Supporting Information
Figure S1.
TR induction by doxycyclin withdrawal. Western blot showing TRβ expression levels (anti-myc, as in Fig. 1A) after 24 hrs treatment with increasing doxycyclin concentrations up to 20 ng/ml.
https://doi.org/10.1371/journal.pone.0052407.s001
(PPTX)
Figure S2.
Exogenous TRs are recognized by antibodies against TR primary sequences. We compared TRβ and TRα levels in HepG2-TR cells versus HepG2 cells +/−1 hr T3 treatment. The panel shows representative western blots of HepG2-TR and HepG2 cell extracts probed with TR antibodies. Antibodies were TRβ (TRβ72–93; BabCO, Berkeley Antibody Co., now Covance, Richmond, CA, see reference [22]) and TRα (abcam: ab53729). Secondary antibody was goat anti-mouse IgG-HRP conjugate from Santa Cruz Antibodies.
https://doi.org/10.1371/journal.pone.0052407.s002
(PPTX)
Figure S3.
The TR expression tag does not affect T3 response in transient transfection assays. The panel shows luciferase activities at a standard DR-4 reporter with equivalent (optimal) levels of transfected wild type TRβ expression vector or similar flag-TRβ vector.
https://doi.org/10.1371/journal.pone.0052407.s003
(PPTX)
Figure S4.
Silencing of ANGPTL4 expression by TR overexpression. Results of qPCR analysis to show ANGPTL4 expression levels +/_ T3 in parental HepG2 cells versus HepG2-TR cells. Note the silencing of T3 response and strong suppression of basal expression levels by unliganded TRs.
https://doi.org/10.1371/journal.pone.0052407.s004
(PPTX)
Figure S5.
SiRNA directed against the upstream EGFP coding sequences selectively inhibits responses to exogenous TRs. HepG2-TRβ cells were transfected with 5 nM Qiagen negative control siRNA (siRNA_Qiagen-NC) “UUCUCCGAACGUGUCACGU” or and siRNA-EGFP “GCCACAACGUCUAUAUCAUGG”, treated with DMSO (vehicle) or 100 nM T3 24 hr later post siRNA transfection, and RNA was isolated 24 hr later. The left panel shows relative expression of TRβ +/− T3 in the presence of control or EGFP siRNA confirming efficient knockdown of TRβ transcripts in HepG2-TRβ cells (similar knockdown was observed for EGFP). The right panel shows inhibition of T3 response at the highly induced C10orf10 gene with the EGFP siRNA treatment and not Qiagen negative control (QNC) siRNA. Similar results were obtained with other T3 induced genes in the HepG2-TRβ cells, including anxa1, pck1, slc16a6, and scnn1a. Data presented represents 3 biological replicates, and bars with the same letters are not statistically different (adjusted p-value>0.05; ANOVA, Tukey-HSD on (C10orf10_Cp) - (RPS27A_Cp)).
https://doi.org/10.1371/journal.pone.0052407.s005
(PPTX)
Figure S6.
Heatmap showing a subset of “indirect” T3 target genes, implied by a disruption of the T3-response in the presence of CHX. Seventeen probes (genes) of the 158 probes (156 genes) that displayed a≥2-fold fold-change and Benjamini-Hochberg adjusted p-value< = 0.1 (i.e., T3 vs. Ctrl treatment) in either HepG2-TRα or HepG2-TRβ cells were determined to be indirect targets based on the compromise of the T3-response within the presence of CHX.
https://doi.org/10.1371/journal.pone.0052407.s006
(PPTX)
Figure S7.
Time course of TR induction in HeLa cells. A. Results of qPCR analysis showing optimal induction of TRβ transcripts after 24 hrs DOX withdrawal. B. Induced TR is functional as observed in the known TR gene target thrsp. Note that expression of this gene is strongly suppressed in response to TR induction within the 24 hr time period of DOX withdrawal.
https://doi.org/10.1371/journal.pone.0052407.s007
(PPTX)
Figure S8.
Representative concentration dependence curves. qPCR analysis of T3 response. 24 hrs treatment with varying amounts of T3 shown on the x-axis. Note differences in response curve for KLF9 in the presence of TRα and TRβ.
https://doi.org/10.1371/journal.pone.0052407.s008
(PPTX)
Table S2.
TR regulated gene pattern information.
https://doi.org/10.1371/journal.pone.0052407.s010
(XLSX)
Acknowledgments
We wish to honor the memory of John D. Baxter (deceased) whose insights into thyroid hormone action inspired this study.
Author Contributions
Conceived and designed the experiments: JZL DHS AC PW. Performed the experiments: JZL DHS CY ASA SF JJCP FDR AC. Analyzed the data: JZL DHS JS XZ AC PW. Wrote the paper: JZL DHS AC PW.
References
- 1. Yen PM (2001) Physiological and molecular basis of thyroid hormone action. Physiol Rev 81: 1097–142.
- 2. Cheng SY, Leonard JL, Davis PJ (2010) Molecular aspects of thyroid hormone actions. Endocr Rev 31: 139–70.
- 3. Glass CK, Rosenfeld MG (2000) The coregulator exchange in transcriptional functions of nuclear receptors. Genes Dev 14: 121–41.
- 4. Astapova I, Hollenberg AN (2012) The in vivo role of nuclear receptor corepressors in thyroid hormone action. Biochim Biophys Acta
- 5. Fondell JD (2012) The Mediator complex in thyroid hormone receptor action. Biochim Biophys Acta
- 6. Forrest D, Vennstrom B (2000) Functions of thyroid hormone receptors in mice. Thyroid 10: 41–52.
- 7. O'Shea PJ, Williams GR (2002) Insight into the physiological actions of thyroid hormone receptors from genetically modified mice. J Endocrinol 175: 553–70.
- 8. Flamant F, Gauthier K (2012) Thyroid hormone receptors: The challenge of elucidating isotype-specific functions and cell-specific response. Biochim Biophys Acta
- 9. Bassett JH, Williams GR (2009) The skeletal phenotypes of TRalpha and TRbeta mutant mice. J Mol Endocrinol 42: 269–82.
- 10. Weiss RE, Refetoff S (2000) Resistance to thyroid hormone. Rev Endocr Metab Disord 1: 97–108.
- 11. Bochukova E, Schoenmakers N, Agostini M, Schoenmakers E, Rajanayagam O, et al. (2012) A mutation in the thyroid hormone receptor alpha gene. N Engl J Med 366: 243–9.
- 12. Baxter JD, Webb P (2009) Thyroid hormone mimetics: potential applications in atherosclerosis, obesity and type 2 diabetes. Nat Rev Drug Discov 8: 308–20.
- 13. Feng X, Jiang Y, Meltzer P, Yen PM (2000) Thyroid hormone regulation of hepatic genes in vivo detected by complementary DNA microarray. Mol Endocrinol 14: 947–55.
- 14. Yen PM, Feng X, Flamant F, Chen Y, Walker RL, et al. (2003) Effects of ligand and thyroid hormone receptor isoforms on hepatic gene expression profiles of thyroid hormone receptor knockout mice. EMBO Rep 4: 581–7.
- 15. Thomas C, Gustafsson JA (2011) The different roles of ER subtypes in cancer biology and therapy. Nat Rev Cancer 11: 597–608.
- 16. Chan IH, Privalsky ML (2009) Isoform-specific transcriptional activity of overlapping target genes that respond to thyroid hormone receptors alpha1 and beta1. Mol Endocrinol 23: 1758–75.
- 17. Velasco LF, Togashi M, Walfish PG, Pessanha RP, Barra GB, et al. (2207) Thyroid hormone response element organization dictates the composition of active receptor. J Biol Chem 282: 12458–66.
- 18. Fugier C, Tousaint JJ, Prieur X, Plateroni M, Samarut J, et al. (2006) The lipoprotein lipase inhibitor ANGPTL3 is negatively regulated by thyroid hormone. J Biol Chem 281: 11553–9.
- 19. Weinhofer I, Kunze M, Rampler H, Forss-Petter S, Samarut J, et al. (2008) Distinct modulatory roles for thyroid hormone receptors TRalpha and TRbeta in SREBP1-activated ABCD2 expression. Eur J Cell Biol 87: 933–45.
- 20. Fozzatti L, Lu C, Kim DW, Cheng SY (2011) Differential recruitment of nuclear coregulators directs the isoform-dependent action of mutant thyroid hormone receptors. Mol Endocrinol 25: 908–21.
- 21. Chiamolera MI, Sidhave AR, Matsumoto S, He Q, Hashimoto K, et al. (2012) Fundamentally distinct roles of thyroid hormone receptor isoforms in a thyrotroph cell line are due to differential DNA binding. Mol Endocrinol 26: 926–39.
- 22. Yuan C, Lin JZ, Sieglaff DH, Ayers SD, Denoto Reynolds, et al. (2012) Identical gene regulation patterns of T3 and selective thyroid hormone receptor modulator GC-1. Endocrinology 153: 501–11.
- 23. Martinez L, Nascimento AS, Nunes FM, Philips K, Aparacio R, et al. (2009) Gaining ligand selectivity in thyroid hormone receptors via entropy.. Proc Natl Acad Sci U S A 106: 20717–22.
- 24. Du P, Kibbe WA, Lin SM (2008) lumi: a pipeline for processing Illumina microarray. Bioinformatics 24: 1547–8.
- 25. Smyth GK (2004) Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Statistical Applications in Genetics and Molecular Biology 3 (1)
- 26. Team RDC (2009) R: A language and environment for statistical computing.. R Foundation for Statistical Computing: Vienna
- 27. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological) 57: 289–300.
- 28.
Hipp DR, SQLite.
- 29. Saeed AI, Sharov V, White J, Li J, Liang W, et al. (2003) TM4: a free, open-source system for microarray data management and analysis. Biotechniques 34: 374–8.
- 30. Baxter JD, Eberhardt NL, Apriletti JW, Johnson LK, Ivarie RD, et al. (1979) Thyroid hormone receptors and responses. Recent Prog Horm Res 35: 97–153.
- 31. Lonard DM, O'Malley BW (2009) Emerging roles of the ubiquitin proteasome system in nuclear hormone receptor signaling. Prog Mol Biol Transl Sci 87: 117–35.
- 32. Yang X, Downes M, Yu RT, Bookout AL, He W, et al. (2006) Nuclear receptor expression links the circadian clock to metabolism. Cell 126: 801–10.