Glucocorticoid signaling synchronizes the liver circadian transcriptome


  • Potential conflict of interest: Nothing to report.


Circadian control of physiology is mediated by local, tissue-based clocks, synchronized to each other and to solar time by signals from the suprachiasmatic nuclei (SCN), the master oscillator in the hypothalamus. These local clocks coordinate the transcription of key pathways to establish tissue-specific daily metabolic programs. How local transcriptomes are synchronized across the organism and their relative contribution to circadian output remain unclear. In the present study we showed that glucocorticoids alone are able to synchronize expression of about 60% of the circadian transcriptome. We propose that synchronization occurs directly by the action of glucocorticoids on a diverse range of downstream targets and indirectly by regulating the core clock genes mPer1, Bmal1, mCry1, and Dbp. We have identified the pivotal liver transcription factor, HNF4α, as a mediator of circadian and glucocorticoid-regulated transcription, showing that it is a key conduit for downstream targeting. Conclusion: We have demonstrated that by orchestrating transcriptional cascades, glucocorticoids are able to direct synchronization of a diverse range of functionally important circadian genes. (HEPATOLOGY 2007;45:1478–1488.)

There is growing recognition of the critical contribution of circadian timing to health and disease.1, 2 Circadian control over physiology is underpinned by the localized but orchestrated activity of tissue-based clocks. These clocks drive local metabolic programs to ensure the appropriately phased temporal functioning of the organ.3 Studies with tissue explants and fibroblast cultures have shown that peripheral cellular oscillators rapidly lose synchrony in the absence of entraining cues.4, 5 The principal neural oscillator contained in the suprachiasmatic nuclei (SCN) of the hypothalamus is necessary for this circadian coordination in vivo,4, 6 but the avenues by which the SCN maintains global circadian synchrony have not been uncharacterized.

At their core, the molecular clocks of the SCN and peripheral tissues consist of an autoregulatory feedback loop involving transcriptional and posttranslational processes. Genes necessary for the core feedback loop include those from the Period (Per) and cryptochrome (Cry) families, expression of which is driven by heterodimeric complexes of basic helix–loop–helix transcription factors encoded by Clock and Bmal1 that interact with E-box elements in the regulatory regions of Per and Cry.7, 8 The periodic expression of Per, Cry, and Bmal1 maintained by the core loop is further able to impose circadian regulation on other “clock-controlled” genes that carry regulatory sequences sensitive to these circadian factors. Indeed, recent studies have identified extensive and divergent “circadian transcriptomes” of major organ systems such as the liver, kidney, and heart9–15 that are a product of this cyclical transcriptional output. Furthermore, recent work has highlighted that key rate-limiting metabolic enzymes are under clock control at the protein level, extending these findings to a “circadian proteome.”3 The starting point for synchronization of metabolic programs therefore lies in temporally orchestrated programs of gene expression. A variety of biochemical factors including glucocorticoids have been shown to have the potential to alter transcription of core clock genes and a limited number of clock-controlled factors when tested in vitro. However, the role played by any one factor has not been established in vivo. In the present study we specifically examined the effects of glucocorticoids in synchronizing circadian gene expression in the mouse liver in vivo.

Glucocorticoids are potent transcriptional regulators of the major organ systems. They are secreted by the adrenal glands in a pronounced circadian profile and therefore well placed to mediate internal synchronization. A single exposure of fibroblast cultures to a glucocorticoid agonist resets rhythmic but asynchronous cells to a common circadian phase,5 whereas acute activation of glucocorticoid signaling in vivo induces the expression of the core circadian gene mPer1 in the liver. This induction is direct as it is dependent on local expression of the glucocorticoid receptor (GR).16 However, it is transient because in the intact animal other SCN-dependent synchronizing cues rapidly restore circadian gene expression to its characteristic pattern. This interplay between cues means that the in vivo synchronizing action of glucocorticoids (or any other single factor) can only be characterized if all other rhythmic cues are absent.

We therefore examined the effects of activation of glucocorticoid receptors on the expression of the circadian transcriptome in mice subjected to surgical ablation of the SCN, the master clock in mammals. This was designed to remove all other potentially confounding circadian cues, such as feeding and body temperature rhythms, in order to examine, for the first time, the circadian actions of glucocorticoid signaling in isolation.16–19 This surgical approach has major advantages over the use of mutant mice because even though the central timekeeping machinery is disrupted following ablation, peripheral tissues still have fully competent circadian and noncircadian genotypes not subject to the potential developmental disturbance associated with clock mutations.11, 20, 21 Therefore, screening for dexamethasone (DEX)-regulated genes on this arrhythmic background would reveal those components of the circadian transcriptome for which a single glucocorticoid stimulus was sufficient to trigger synchronized expression in the absence of other temporal cues. Our approach thus complements studies that have used glucocorticoid receptor-null and adrenalectomized mice to identify transcripts for which glucocorticoid-mediated signaling is necessary for synchronized expression in the liver in the presence of other rhythmic cues.19, 22


CT, circadian time; DEX, dexamethasone; GR, glucocorticoid receptor; IVT in vitro transcription; SCN, suprachiasmatic nuclei.

Materials and Methods

Animals and Procedures.

All animal experimentation was licensed by the Home Office under the Animals (Scientific Procedures) Act, 1987. Adult male CD1 mice (Harlan-Olac) were housed and subject to SCN ablation or sham surgery as previously described.11 Dexamethasone 21-phosphate (D-1159; Sigma), 300 μg/ml dissolved in phosphate-buffered saline (PBS), was delivered intraperitoneally (2 mg/kg) at predicted CT6 under dim red light.16 PBS vehicle was used as the control with 0.15% ethanol. Liver and kidney tissue was harvested on the second cycle after transfer from 12 hours light (L):12 hours dim red light (DR) to DR:DR and immediately frozen prior to RNA extraction.

Real-Time Polymerase Chain Reaction.

RNA from tissue was extracted using TRIzol reagent (Invitrogen) and purified using RNAeasy Mini Kits (Qiagen) and DNAse I treatment to remove subgenomic DNA contamination. Total RNA (25-100 ng) was used in 25 μL of RT-PCR reaction according to the manufacturer's protocol (TaqMan EZ RT-PCR Kit, Applera) with appropriate primers and probes. For control reactions, mouse β-actin mRNA was amplified from the same RNA samples. Real-time polymerase chain reaction (PCR) was performed on ABI 7700 Sequence Detection System (Applera) and iCycler (Bio-Rad) systems. PCR conditions were optimized for probe and primer concentrations, and standard thermal cycling parameters were used. The relative level of each mRNA was calculated by the 2−ΔΔCt method (Ct stands for the cycle number at which the signal reaches the threshold of detection) and normalized to the corresponding β-actin mRNA level. Each Ct value used for these calculations was the mean of at least 3 duplicates of the same reaction. Relative mRNA was determined as the percentage of the maximum value obtained for each experiment. Oligonucleotide primers and TaqMan probe sets were designed using ABI Primer Express software (Applera) and synthesized by standard methods. All probes were 5′-labeled with FAM fluorophore and 3′-labeled with the quencher TAMRA. Where possible, probe sequences were designed to cross exon boundaries in the corresponding cDNA sequences, eliminating the possibility of contaminating genomic DNA amplification. Table 2 shows the TaqMan primer and probe sequences used.

Table 1. Summary of Temporally Restricted Circadian Gene Regulation by Dexamethasone (DEX)
Gene ResponseTime after Injection of DEXN
5 Hours (CT11)17 Hours (CT23)29 Hours (CT35)
Activation, n (%)44 (52)7 (8)33 (39)84
Suppression, n (%)28 (22)78 (63)17 (13)123
Table 2. TaqMan Primer and Probe Sequences Used
Sequence definitionβ-Actin
Sequence definitionBmal1
Sequence definitionBnip3
Sequence definitionChd1l
Sequence definitionDbp
Sequence definitionDdit3
Sequence definitionFgf18
Sequence definitionHNF4α
Sequence definitionJam3
Sequence definitionmCry1
Sequence definitionmPer1
Sequence definitionmPer2
Sequence definitionOtc
Sequence definitionSgk

Affymetrix GeneChip Studies.

Modified Affymetrix protocols were used. Briefly, RNA from tissue was extract using TRIzol reagent (Invitrogen) and purified using RNAeasy Mini Kits (Qiagen) and DNAse I treatment to remove subgenomic DNA contamination. RNA integrity and purity were further analyzed with an Agilent 2100 Bioanalyzer. Double-stranded complementary DNA (cDNA) was synthesized using a SuperScript Kit (Invitrogen). The first strand was synthesized using SuperScript II reverse transcriptase, primed with a T7-(dT)24 primer. Second-strand synthesis was carried out by DNA polymerase I, using the first strand as a template. Double-stranded cDNA was then purified by phenol-chloroform extraction, ethanol precipitation, and pellet resuspension. Purified cDNA was then used as a template for in vitro transcription (IVT) using a BioArray RNA Transcript labeling kit (ENZO). The labeled cRNA was purified using RNAeasy Kits (Qiagen) before hybridization to Affymetrix Mu74Av2 GeneChips (i.e., approximately 12,500 transcripts sampled). An Affymetrix Hybridisation Oven 6400 was used to perform hybridizations overnight and then a GeneChip Fluidics Station 400 (Affymetrix) was used to automate posthybridization washing and staining procedures. Array cartridges were subsequently analyzed using an Agilent GeneArray Scanner and the Affymetrix analysis software suite. Raw data (average difference values) from the GeneChip were imported into GeneSpring software (Silicon Genetics) for subsequent processing and analysis. Statistical analyses were performed using Microsoft Excel and SPSS as appropriate. The microarray data were deposited in the GEO Database under accession numbers GSM8589-GSM8615.

Regulatory Sequence Analysis.

The mouse genomic sequence was extracted from the UCSC Genome Bioinformatics server ( Examined were 5 kbp of 5′-flanking sequence and all intronic sequence using MotifScanner v3.0, applying appropriate Transfac position weight matrices for binding sites of the glucocorticoid receptor, HNF4α and custom-made matrices for canonical E-box (5′-CACGTG-3′), E′-box (5′-CACGTT-3′), D-box (TTA[T/C]GTAA) and RORE ([A/T]A[A/T]NT[A/G]GGTCA) sequences. A control group of genes (genes assayed by the GeneChip there were not found to be cyclical or induced by dexamethasone) was simultaneously scanned for motifs so that the frequency in this set could be compared with that in the experimental gene sets.

Luciferase Reporter Gene Assays.

Plasmids were constructed as follows. The pGL3-promoter vector (Promega) was digested with KpnI and BglII and then gel-purified with a QiaQuick gel extraction kit (Qiagen). Oligonucleotides designed with KpnI- and BglII-compatible overhangs were manufactured by standard procedures and engineered with 5′ phosphorylated ends for easier cloning. These oligonucleotides are listed in Table 3 and encompass the relevant HNF4α regulatory element and its native surrounding sequence in the promoter (wild-type). In addition, mutant elements were constructed in which the regulatory element (but not the surrounding native sequence) was altered by substituting purines for pyramidines and vice versa. Double-stranded annealed oligonucleotides (formed by annealing the forward and reverse oligonucleotides) were then cloned into the digested reporter vector by ligation. A SV40-driven rat HNF4α expression vector was kindly provided by Dr. F. Gonzalez (pSG5-rHNF4a). Table 3 shows the oligonucleotides used to construct the plasmids (all 5′-3′).

Table 3. Oligonucleotides Used to Construct Plasmids (All 5′–3′)
Wild Type

NIH3T3 cells were grown in a humidified incubator (at 37°C in 5% CO2) in Dulbecco's Modified Eagle Medium supplemented with 10% fetal calf serum (Invitrogen). Cells (3 × 105) were seeded into 35-mm wells and grown to 60% confluence before they were transfected with plasmid DNA using GeneJuice reagent (Novagen). Cell lysates were harvested 24 hours after transfection. The cells were transfected with 20 ng of the luciferase reporter vector and 500 ng of the HNF4α expression construct; if the latter was not used, an empty pcDNA3.1 vector (Invitrogen) was used instead. In addition, 25 ng of pCMVβ plasmid (Clontech) for expressing the β-galactosidase gene was transfected to measure (and correct for) differences in transfection efficiency. The total amount of DNA transfected was adjusted to 1 μg with pcDNA3.1. Cell lysates were harvested 24 hours after transfection using Glo Lysis Buffer (Promega). Protein assays (Bradford method) were then performed using protein assay reagent (Bio-Rad) and concentrations of total protein adjusted so that all samples were at the same concentration. Luciferase assays were performed on lysates using the Bright-Glo system (Promega), and fluorescence assays for β-galactosidase activity were performed with a FluoReporter lacZ/Galactosidase quantification kit (Molecular Probes). Luciferase activity was expressed relative to β-galactosidase levels. Three replicates of each assay were performed on samples from 3 independent experiments.

Results and Discussion

Definition of the Circadian Transcriptase and Its Dependence on the SCN.

As a prerequisite to mapping the global actions of glucocorticoids on circadian gene expression, we first sought to rigorously define the liver circadian transcriptome and confirm our previous findings that synchronized circadian transcription in the liver is dependent on the SCN.11 Based on our synopsis of the results of other studies using the Affymetrix Genechip™ to define circadian genes in the liver, we focused initially on an amalgamated gene set.13–15 Unexpectedly, only 68 of the aggregate total of 981 transcripts (7%) previously reported as circadian were common to all 3 initial studies. This presumably reflects differences in the criteria used to define circadian rhythmicity in each independent study, as well as the use of mouse strains with differing genetic backgrounds. Our preliminary list of circadian genes thus consisted of the initial 981 transcripts reported to be rhythmically expressed in at least 1 of these previous studies.

To directly confirm circadian rhythmicity of these transcripts and to demonstrate the effect of the SCN lesion on liver gene expression, we collected livers from individual mice under constant conditions (on the second cycle of continuous dim red light). Extracted RNA from livers taken at circadian time (CT) 11, CT23, and CT35 was processed and applied to individual microarray chips, rather than pooled, in order that we could use ANOVA to apply a rigorous statistical test for rhythmic expression. These time points were chosen for direct comparison with our subsequent study of glucocorticoids. Of the 981 putative circadian transcripts, 472 (48%) were also significantly rhythmic in our study and, based on our 3-point analysis, could be assigned to 1 of 2 phase clusters: mPer1-like transcripts peaking at CT11 and CT35 (Fig. 1A) and Bmal1-like transcripts peaking at CT23 (Fig. 1B). Given that our samples covered this limited set of time points, we consider 48% concurrence with the published data to be acceptable, especially considering that the overlap between the previous studies was only 7%.13–15 The effect of the SCN lesion (SCNX) on synchronized circadian expression in the livers was dramatic. A total of 366 of the 472 validated circadian transcripts (78%), including canonical clock genes, were nonrhythmic by ANOVA, extending our previous findings with a different microarray platform (Fig. 1A,B).11

Figure 1.

Synchronized circadian expression of known and novel circadian genes in the liver was abolished in SCN-lesioned mice. Circadian rhythmicity of previously identified circadian transcripts falling into (A) mPer1-like (n = 246) and (B) Bmal1-like (n = 120) clusters in livers of intact mice (CON) was lost following SCN ablation (SCNX). (C) Comparison between gene expression profiles revealed by microarray analysis (Array) and real-time PCR (qPCR) for the canonical clock genes mPer1, mPer2, mCry1, and Bmal1 for intact mice (black line) Each value is the mean ± SEM of measurement of 3 animals. ANOVA did not reveal any temporal variation in expression of any of the 4 genes in SCNX animals (red line).

Glucocorticoid Signaling Regulates Core Clock Genes in the Liver.

We therefore established a “circadian set” of 366 genes that were rhythmic in this and at least 1 other microarray study and were also sensitive to SCN ablation. Although this represents only 3% of the transcriptome we sampled (and our previous estimate for circadian-regulated genes was ≈9%11), this must inevitably represent an underestimate because our sampling strategy was optimized for our glucocorticoid study to only focus on 2 circadian phases 12 hours apart. Quantitative real-time PCR (qPCR) was used to cross-validate our microarray data for 4 canonical clock genes, mPer1, mPer2, mCry1, and mBmal1. Although all 4 showed synchronized, highly rhythmic expression in livers from intact mice, this was lost after SCN ablation (Fig. 1C), confirming the dependence of the liver clock on the central oscillator in our samples.

Having robustly defined a population of 366 hepatic circadian transcripts with confirmed loss of rhythmicity in SCNX mice treated with vehicle (VEH), we next examined the effect of glucocorticoid receptor activation, by injecting SCN-lesioned mice with the receptor analogue dexamethasone (DEX) at predicted CT6. As for the intact and VEH-injected mice, liver tissue was collected on the second cycle under constant conditions at CT11, CT23, and CT35, which corresponded to 5, 17, and 29 hours after injection of VEH or DEX. Our rationale was that when applied against the arrhythmic background of SCN-ablated mice, an effective stimulus would resynchronize multiple cellular clocks in the tissue, driving them to a unique phase and thereby establishing a common temporal pattern in groups of animals sampled at different times following injection.

Dexamethasone had a variety of effects, both early and late, involving core elements of the circadian clockwork. Induction of mPer1 has previously been implicated in peripheral clock resetting by glucocorticoids,16 and we identified a prominent induction of mPer1 by both microarray and independent qPCR (Fig. 2A). Similarly, in the oppositely phased cluster, we found that transcription of Bmal1 and mCry1 was up-regulated by DEX (Fig. 2A). The presence of Glucocorticoid Response Elements (GREs) in the regulatory regions of these core clock components suggests a likely mechanism for these effects (Fig. 2B). qPCR for representative genes, all expressed in phase with mPer1, further cross-validated the microarray data (Fig. 2C-E). Serum/glucocorticoid-induced kinase (Sgk), reported previously as glucocorticoid-responsive,23 exhibited circadian expression in the liver. This rhythm was lost in tissues from SCNX mice, with expression at basal levels, but DEX caused a pronounced and acute induction in the livers of SCN ablated mice (Fig. 2C). A comparable pattern of rhythmicity in intact mice, basal expression in SCNX, and acute induction by DEX was also observed for chromodomain helicase DNA binding protein (Chd1l; Fig. 2D). In contrast, Dbp, an archetypal circadian output gene, also expressed rhythmically in phase with mPer1 and arrhythmically in tissues from SCN-lesioned animals, which first showed acute suppression in response to DEX, followed by delayed induction 24 hours later (Fig. 2E). Thus, in addition to mPer1, we found that numerous additional core clock elements are regulated by glucocorticoid signaling, illustrating multiple entry routes into the core oscillator by which glucocorticoids are able to bring about peripheral circadian synchronization.

Figure 2.

Glucocorticoid receptor activation synchronized expression of known circadian genes in the livers of SCN-lesioned mice. (A) Induction of the canonical clock genes mPer1, mCry1, and Bmal1 by dexamethasone administered at CT6, as shown by real-time PCR. Each value is the mean ± SEM of 3 animals (*P < 0.01). (B) Comparison of the 5′-flanking regions of mouse and human mPer1, mCry1, and Bmal1 genes, showing the locations of GREs. (C) Circadian profiles of mRNA expression for serum/glucocorticoid regulated kinase (Sgk) as determined by real-time PCR. (D) Circadian profiles of mRNA expression for chromodomain-helicase DNA-binding protein-1 like (Chd1l) as determined by real-time PCR. (E) Circadian profiles of mRNA expression for albumin D-element binding protein (Dbp) as determined by real-time PCR. Each assay was repeated 3 times on pooled RNA from 6 animals. Each value is the mean ± SEM of 3 assays. Induction of Sgk and Chdl and suppression of Dbp transcription by dexamethasone is also shown (mean ± SEM, n = 3 mice; *P < 0.01).

Glucocorticoid Signaling Regulates a Large Proportion of the Hepatic Circadian Transcriptome.

Globally, across the circadian transcriptome, genes in both mPer1- and Bmal1-like clusters were induced or suppressed by glucocorticoid receptor activation at CT6 (Fig. 3). Regulation of transcription proceeded in a wave of activation and inhibition of genes, some being induced/suppressed acutely, at CT11, and others regulated more chronically, at CT23 and CT35 (Fig. 3). Of the 366 target circadian genes, 20% (n = 72) were acutely resynchronized to their appropriate phase within 5 hours of treatment, and 57% (n = 207) showed temporally specific induction or suppression in the 29 hours following receptor activation (see Table 1). Moreover, the inductive effect of DEX on circadian genes was antiphasic to the suppressive effect. Induced genes responded most obviously either soon after glucocorticoid receptor activation, at CT11, or at the start of the following circadian cycle, at CT35 (see Table 1). In contrast, genes that were suppressed were most apparent in the intervening sample (CT23), 12 hours after the acute surge. This statistically significant (χ2= 62.9, P < 0.0001) differential time course of induction and suppression indicates that a single episode of receptor activation is succeeded first by waves of transcriptional activation, then suppression, and then back to activation across the circadian transcriptome. Thus, in the absence of other rhythmic cues, glucocorticoids are sufficient to exert a rapid and extensive effect on the major part of the liver circadian transcriptome, confirming their role as principal synchronizing factors in vivo.

Figure 3.

Global synchronization of hepatic circadian gene expression by glucocorticoids. The effects of the administration of vehicle (VEH) or dexamethasone (DEX) at CT6 in livers of SCN-lesioned mice was assayed. Circadian genes in (A) mPer1-like or (B) Bmal1-like clusters were induced by DEX but not by VEH in SCN-lesioned mice. Conversely, transcription of circadian genes in (C) mPer1-like or (D) Bmal1-like clusters was suppressed by DEX but not by VEH in the same mice. The arrow shows the time of injection in VEH and DEX animals (CT6). Animals were subsequently sampled for 5 hours (CT11), 17 hours (CT23), and 29 hours (CT35) following injection. Each value is the mean of measurements of 3 animals and was normalized to the mean average difference (a measure of gene expression) in control mice (unlesioned, untreated). ANOVA failed to reveal any temporal variation in gene expression in vehicle-treated mice, but there was a significant effect of administration of dexamethasone in animals compared with animals treated with vehicle at CT11 (P < 0.05). Affymetrix Probe IDs are given for each transcript.

Having demonstrated the pervasive effect of glucocorticoid signaling on circadian gene expression, we investigated the mechanisms by which circadian synchronization might be established. The acute responses of the core clock genes, such as Per1, Bmal1, and Cry1, likely represent the basic mechanism underlying circadian entrainment by glucocorticoids, their synchronization most likely mediated via glucocorticoid response elements (GREs) in the regulatory regions of their human and mouse orthologues (Fig. 2B). Circadian coordination may then proceed along 2 main routes. First, receptor activation may act directly on additional GRE-bearing circadian genes, and second, synchronization may proceed indirectly by activation of the core clock loop.

Transcriptional Mechanisms Underlying Glucocorticoid Regulation Circadian Genes.

To investigate this further, we scanned 5 kbp of a 5′-flanking regulatory sequence and intronic sequence of glucocorticoid-induced circadian genes for GREs as well as binding sites for various transcription factors that have recently been shown to be important for clock output: E-boxes, E′-boxes, D-boxes, and ROREs.24 A total of 70 such genes had sequence available for analysis. Almost two-thirds (46 genes) carried GREs, implying that direct regulation of circadian genes by glucocorticoids is a common mechanism of synchronization. However, genomic analysis revealed an additional level of regulatory complexity, as a significant proportion of GRE-containing genes (33 of 46) also possess elements responsive to alternative factors. Thus, only 13 genes are potentially regulated exclusively through GREs (see Supplementary Table 1; Supplementary material for this article can be found on the HEPATOLOGY website: A possible way to differentiate glucocorticoid-mediated circadian regulation from that involving the clock loop would be to study circadian expression in glucocorticoid receptor knockout mice or adrenalectomized animals. Although analysis of these systems may reveal some circadian targets for glucocorticoids, circadian regulation via alternative regulatory sequences would mask other receptor targets. Therefore, it is difficult to disentangle any direct transcriptional effects downstream of glucocorticoid signaling from indirect effects via the core loop. Our analysis of a recent study looking at circadian transcripts in adrenalectomized animals illustrates this. Of the 13 genes we identified as containing only GREs, only 3 genes (573003B10Rik, Fkbp5, and Herpud1) showed adrenal-dependent circadian expression.22

Thus, glucocorticoid regulation of circadian genes lacking GREs implies secondary activation via the core circadian loop, with contributions from transcriptional networks involving E-boxes, E′-boxes, D-boxes, and ROREs.25, 26Dbp is an example of this, as it does not carry GRE sequences yet exhibits potent suppression after administration of dexamethasone. This regulation likely occurs through functional intronic E-boxes,27 and accordingly, its rhythmic expression was lost in the livers of Clock mutant mice (data not shown). Overall, however, few of the glucocorticoid-regulated circadian genes would be subject to exclusively indirect regulation via the core loop, although those genes with both GREs and E-boxes/D-boxes/RORES are potentially convergent targets for systemic (glucocorticoid-mediated) and intracellular (E-box/D-box/RORE-mediated) synchronization (Fig. 4A and Supplementary Table 1).

Figure 4.

Hepatocyte nuclear factor 4α (HNF4α) is an interface between glucocorticoids and gene expression in the liver. (A) Venn diagram showing glucocorticoid-regulated circadian genes (n = 72) grouped according to possession of glucocorticoid response elements (GREs), E-boxes, and HNF4α regulatory sequences. Genes that do not possess any of these regulatory elements are shown under “other.” (B) High-resolution circadian profile of HNF4α in the liver (n = 6 animals per time point). (C) Effects of vehicle or dexamethasone administration at CT6 on expression of HNF4α in SCN-lesioned mice. (D) Schematic representation of mouse HNF4α gene showing the positions of glucocorticoid regulatory elements (GREs) and canonical E-boxes within 5′-flanking and intronic regions of the gene. Gray bars represent exons. An arrow represents the translation start site. (E) HNF4α mRNA expression in wild-type (WT, black) and Clock −/− mutant mice (Clock −/−, red) determined by real-time PCR (n = 6 animals per group). (F) High-resolution circadian profile for ornithine transcarbamylase (Otc) and its expression in liver-specific HNF4α-null mice assayed by real-time PCR. FLOX indicates control mice and KO indicates HNF4α-null mice (n = 6 animals per group per time point). (G) Persistent but low-amplitude variation of mPer2 and Bmal1 in HNF4α-null mice.

Additional levels of transcriptional synchronization arise from clock-regulated transcription factors such as Dbp that direct further cascades of temporally regulated gene expression through their specific target sequences (e.g., D-boxes). Indeed, glucocorticoid-mediated circadian synchronization, apparently independent of GREs and E-boxes, was observed for about one quarter of the genes in our sample (15 of 70). We therefore screened our set of 70 glucocorticoid-sensitive circadian genes for transcription factor binding sites and identified a high frequency of targets (39%)27 of hepatocyte nuclear factor 4α (HNF4α, see Fig. 4A). Notably, 4 of the induced genes carried HNF4α sequences but not GREs, whereas the remaining 23 genes possessed GREs in addition to HNF4α sites, suggesting a convergent regulatory action (Fig. 4A and Supplementary Table 1). Importantly, when we concurrently analyzed a set of 69 highly expressed control genes (which had noncircadian profiles statistically and did not respond to DEX at any time), we found that only 7% of these possessed HNF4α regulatory sequences, clearly illustrating a significant enrichment in glucocorticoid-regulated circadian genes (χ2= 16.7, P < 0.0001).

HNFα Is a Novel Regulator of Circadian Gene Expression in the Liver.

HNF4α is a pivotal liver transcription factor that regulates numerous metabolic pathways, including ureagenesis, the production of serum proteins, and the activity of cytochrome P450 genes.28, 29 The results of a recent work suggest that a large proportion of the liver transcriptome is under the control of HNF4α, making it an important orchestrator of hepatic gene expression in general. This vital role explains why its dysfunction contributes to, for example, the pathogenesis of type 2 diabetes mellitus.29 The identification of HNF4α as a circadian regulator from our bioinformatic screen was supported by expression data in vivo. Both microarray and qPCR showed expression of HNF4α in intact mouse livers was highly circadian, peaking at CT16, in phase with mPer2 (Fig. 4B). In vehicle-treated, SCN-lesioned animals, synchronized circadian expression of HNF4α was absent, and expression remained at basal levels (Fig. 4C). Glucocorticoid activation, however, resulted in a profile that was almost identical to that seen in SCN-intact control animals (Fig. 4C). Thus, HNF4α is a glucocorticoid-sensitive circadian gene.

Circadian and glucocorticoid control over HNF4α looks to be exercised by 2 convergent routes. First, as with Dbp,27 it carries several intronic canonical E-boxes that are targets of Clock:Bmal1 complexes (Fig. 4D). Consistent with this, circadian expression of HNF4α was abolished in Clock mutant mice (Fig. 4E). Second, HNF4α also carries several GRE sequences through which DEX can act (Fig. 4D). To characterize the downstream effects of circadian and glucocorticoid-mediated HNF4α induction, we initially examined ornithine transcarbamylase (Otc), which is a definitive target of HNF4α, and a central factor in nitrogen metabolism and ureagenesis.28 It is expressed in liver with a high-amplitude circadian cycle (Fig. 4F), and the circadian pattern of Otc expression was lost in the livers of HNF4α-null mice, showing that HNF4α is necessary for its circadian expression. Importantly, this was not a result of a global effect on the circadian clock, because HNF4α-null mice retained circadian gene expression in the liver, as evidenced by higher levels of mPer2 at ZT16 versus ZT04 and reciprocal changes in another clock gene, Bmal1 (Fig. 4G).

To further delineate targets for indirect glucocorticoid regulation via HNF4α, we focused on 4 glucocorticoid-responsive genes (Bnip3, Nfil3, Pklr, and Ahcy) that possess HNF4α regulatory sequences but lack GREs (see Supplementary Table 1). Expression was highly rhythmic in liver from intact mice (Fig. 5A) and this was lost following SCN ablation (Supplementary Fig. 1B). Despite the absence of GREs, all 4 of these genes were up-regulated by DEX in SCN-ablated mice, indicating a role for HNF4α in their circadian coordination (Fig. 5B and Supplementary Fig. 1A,B). To determine the in vivo contribution of HNF4α to circadian coordination, we examined expression of several of these targets in both wild-type and liver-specific HNF4α-deficient mice.28 Rhythmic expression of Bnip3, for example, was lost in the livers of both Clock mutant and HNF4α-null mice, with levels being basal (Fig. 5B). Thus, these results are wholly consistent with the presence of E-boxes and HNF4α sequences in the regulatory regions of this apoptosis regulator and demonstrate the necessity of both the clock complex and HNF4α for circadian expression of this important factor in vivo. We additionally tested the function of putative HNF4α regulatory sequences in the promoters of the 4 glucocorticoid-responsive genes using luciferase reporter assays in order to validate their predicted role. All 4 regulatory elements were responsive to HNF4α in vitro, and this effect was clearly abolished by mutating each of the regulatory elements (Fig. 5C). This substantiates our view that circadian transcription of these genes is regulated by HNF4α, as predicted by our bioinformatics analysis.

Figure 5.

Analysis of circadian HNF4α targets. (A) Liver expression profiles for 3 putative HNF4α-regulated circadian genes: nuclear factor, interleukin 3, regulated (Nfil3), pyruvate kinase liver and red blood cell (Pklr), and S-adenosylhomocysteine hydrolase (Ahcy). Our own microarray analysis and Storch et al. (2002) both found that these 3 genes, containing HNF4 regulatory sequences, show circadian expression. (B) Circadian profile for liver expression of BCL2/adenovirus E1B 19-kDa-interacting protein (Bnip3). Rhythmic expression was lost in SCNX mice and acutely induced by DEX. Rhythmic expression of Bnip3 was also lost in livers of Clock mutant and HNF4α-null mice. (C) Luciferase reporter gene assays testing the functioning of putative HNF4α regulatory sequences in the regulatory regions of Bnip3, Nfil3, Pklr, and Ahcy (WT, wild-type element; Mut, mutated element). A luciferase positive control (pGL3-Control; Luc control) was used for comparison with the reporter assays. Each value is the mean ± SEM of 3 replicates from a single assay. The results are representative of 3 independent experiments (*P < 0.001). All high-resolution circadian profiles are based on the mean ± SEM of 3 independent real-time PCR assays using pooled RNA from 6 animals. Data from SCNX mice treated with VEH or DEX are mean ± SEM for 3 mice. For Clock and HNF4α mutant mice, each value is the mean ± SEM for 6 animals in each group (*P < 0.01).

In addition to exclusively HNF4α-regulated genes, we also examined regulation of Fgf18, a convergent glucocorticoid-responsive target containing both HNF4α sequences and GREs in its 5′-regulatory region. qPCR confirmed circadian expression of Fgf18 in the livers of wild-type mice and that this rhythm was lost in SCN-lesioned mice and in HNF4α-null mice (Supplementary Fig. 1C). Fgf18 was up-regulated 12 hours after treatment of SCN-lesioned mice with DEX, by direct GRE-dependent action and/or by mediation by HNF4α (Supplementary Fig. 1C). In contrast, Ddit3 (also known as Chop-10/Gadd143), which encodes a product involved in apoptosis,30–32 contains neither HNF4α sequences nor GREs. Nevertheless, it was highly rhythmic in liver of intact animals, a rhythm that was lost on SCN ablation (Supplementary Fig. 1D). Expression was induced by DEX, and rhythmic expression was lost in HNF4α-null mice, indicating that intermediate factors sensitive to HNF4α must sustain its normal rhythmic expression. Finally, we screened all noncircadian transcripts that were significantly up-regulated by DEX for HNF4α regulatory sequences in their 5′-flanking and/or intronic regions and identified 45 glucocorticoid-regulated genes possessing HNF4α sequences but lacking GREs (Supplementary Fig. 1A). Jam3 is such a gene. It is not expressed rhythmically in intact animals, but its hepatic expression was severely compromised in HNF4α-null mice (Supplementary Fig. 1E). However, it was strongly induced by DEX, presumably indirectly via HNF4α. Therefore, taken as a whole, these data reveal an important contribution of HNF4α to glucocorticoid-regulated hepatic gene expression, both circadian and noncircadian.


Overall, our results illustrate the effectiveness of glucocorticoid signaling as a global circadian synchronizer in vivo. In the absence of other SCN-dependent rhythmic cues, a single activation of the glucocorticoid receptor is sufficient to synchronize almost 60% of the circadian transcriptome, manifested as a wave of gene induction, suppression, and then induction. We have shown that glucocorticoid-dependent synchronization works through a hierarchy of interlinked mechanisms, critical among which is direct regulation of the core circadian oscillator, with contributions from direct and indirect regulation of clock-controlled transcription factors such as DBP. In this regard, we have identified HNF4α, a pivotal transcriptional regulator in the liver, as a novel link between the SCN, glucocorticoid signaling, and liver-specific gene expression, both circadian and noncircadian. Transcriptional cascades downstream of such regulators and the intricate interactions between their attendant regulatory networks direct synchronization of genes involved in the diverse clock-controlled functions of the liver, including metabolism and ureagenesis. Further delineation of such complex networks and their connectivity is challenging and will involve characterizing additional clock-controlled transcription factors using a range of systems-biological approaches.


The authors thank S. Reppert and H. Okamura. for kindly providing plasmids, J. Takahashi for providing founder Clock mutant mice, and M. Llenicek-Allen and J. McBride (MRC Geneservice) for assistance with microarrays.