REV-ERBα Participates in Circadian SREBP Signaling and Bile Acid Homeostasis

To gain insight into the physiological functions of REV-ERBα, we identified REV-ERBα target genes in a genome-wide manner, using Affymetrix microarray hybridization analysis of liver RNA from Rev-erbα knock-out (Rev-KO) and wild-type (WT) animals humanely killed at ZT12 (ZT, Zeitgeber time; ZT0, lights on; ZT12: lights off). At this time point the differences in the expression of direct REV-ERBα target genes can be expected to be maximal in Rev-KO versus control animals (see [8] and Figure S1). Because REV-ERBα acts as a transcriptional repressor, transcripts specified by direct target genes should be overrepresented in the Rev-KO animals. Transgenic mice that overexpress hepatic REV-ERBα throughout the day [11] were also examined. In these animals, referred to as TgRev, direct REV-ERBα target genes should be expressed at low levels irrespective of the time of day. Accordingly, indirect target genes with reduced expression in Rev-KO mice should display constant high levels in TgRev animals. The differential expression of genes in the livers of WT, Rev-KO, and TgRev mice is shown in the genome-wide transcriptome profiling experiments displayed in Figure 1A (see lists in Table S1 and S2). As expected, most genes found to be up- and down-regulated in Rev-KO animals were down- and up-regulated, respectively, in TgRev mice (Figure 1A, bracket I and III). Only a subset of genes did not follow this simple scenario and showed equal regulation in both Rev-KO and TgRev animals (Figure 1A, bracket II and IV). Mechanistic models potentially accounting for such less intuitive accumulation profiles are depicted in Figure S2.

REV-ERBα oscillates with a sharp expression peak in WT animals. We thus expected that most of the 76 genes (represented by 97 probe sets) that we identified as up-regulated in Rev-KO animals at ZT12 (Table S1) would show circadian mRNA accumulation profiles in WT animals. For the 60 genes (represented by 81 probe sets) down-regulated in Rev-KO (Table S2) this is not necessarily expected as they probably represent indirect REV-ERBα target genes. So far REV-ERBα has only been described as a transcriptional repressor. Obviously, it cannot be excluded that REV-ERBα might also function as an activator on certain target genes, for example by recruiting co-activators. Indeed, recruitment of both co-activators and co-repressors has been reported for the related nuclear receptor Retinoic acid receptor-related Orphan Receptor α (RORα) [33]. We thus extracted the circadian accumulation profiles of the identified genes from a comprehensive study of genome-wide circadian hepatic gene expression at a 1-h resolution that was recently performed by the Hogenesch laboratory [10] (see also Gene Expression Omnibus GSE11923↗ [http://www.ncbi.nlm.nih.gov/geo↗] and http://bioinf.itmat.upenn.edu/circa↗). Using stringent algorithms, around 1% of the liver transcriptome was circadian in this analysis, whereas 54% (41 transcripts) of the up-regulated and 43% (26 transcripts) of the down-regulated fractions identified in our microarray analysis corresponded to circadian genes (Figure S3). This result confirms that REV-ERBα is predominantly a regulator of circadian gene expression. The circadian phases of most of the up-regulated transcripts corresponded to that expected for direct REV-ERBα targets. The phases of the down-regulated transcripts were less uniform, possibly reflecting several indirect mechanisms.

Analysis of the microarray data showed that Rev-erbα disruption led to the misregulation of several genes involved in lipid metabolism. In particular, known target genes of the cholesterol-sensing transcription factor SREBP [34] were expressed at lower levels in Rev-KO animals (Table S2). This suggested that the INSIG-SCAP-SREBP pathway was temporally misregulated. To examine this conjecture, we assessed the nuclear accumulation of SREBP1c in liver at 4-h intervals around the clock. In WT animals, nuclear SREBP1c protein was highly abundant at ZT12, but barely detectable at ZT0 and ZT4 (Figure 1B). In Rev-KO animals, however, the accumulation of SREBP1c in the nucleus was shifted by 8 h (Figure 1B). This phase-shifted nuclear SREBP1c protein accumulation in Rev-KO mice correlated with the phase-shifted expression of SREBP1c target genes, such as Fas, Elovl6, Fads2, Insig1, or Srebp1c itself (Figure 1C, left panel). In TgRev mice, these genes were up-regulated at ZT0 and virtually unchanged at ZT12, with the exception of Fads2 which was up-regulated at both time points (Figure 1C). These findings substantiated the hypothesis that REV-ERBα controlled the temporal expression of SREBP1c target genes. The altered expression of Hmgcr, Low density lipoprotein receptor (Ldlr), and Acetoacetyl-CoA synthetase (Aacs) in Rev-KO and TgRev animals suggested that the activity of SREBP2 was affected in these animals as well (Figure 1D), in spite of the apparently normal Srebp2 mRNA accumulation (Figure S4A).

The altered nuclear accumulation of SREBP1c in livers of Rev-KO animals could have been due to a misregulation of its mRNA (and consequently protein) expression (Figure 1C) and/or a delay in the processing and transport of SREBP1c to the nucleus. To investigate the latter possibility, we analyzed whether in Rev-KO mice there were any peculiarities to the SREBP processing machinery that would explain the delay in nuclear SREBP1c accumulation. Whereas Insig2a mRNA (coding for the major INSIG isoform in liver) showed a 7-fold circadian amplitude in WT livers, it was indeed expressed at constantly high and low levels in Rev-KO and TgRev mice, respectively (Figure 1E). This suggested that Insig2a was a bona fide REV-ERBα target gene. We were unfortunately unable to examine whether corresponding changes also exist for INSIG2a on the protein level, since the two commercial anti-INSIG2a antibodies we purchased did not yield satisfactory results in immunoblot experiments. However, since INSIG2a is known to be a short-lived protein [35], the oscillation of Insig2a mRNA very likely also translated to the rhythmic accumulation of INSIG2a protein in ER membranes. The expression of Scap mRNA (Figure S4B) and the non-circadian minor Insig2b isoform (Figure 1E) were not affected in Rev-KO mice.

As shown above, REV-ERBα controlled the nuclear accumulation of SREBPs and, thus, the expression of SREBP targets. However, it has also been shown that these processes are regulated by fasting–feeding [36], fatty acids [37]–[39], and insulin [40]. Therefore, we wished to determine if the temporal misregulation of the SREBP pathway in Rev-KO mice merely reflected altered feeding rhythms and/or differences in plasma insulin. To this end, we recorded the time of food consumption in WT and Rev-KO animals by an infrared detection system; there were no major differences between the genotypes (Figure 2A).

We then fasted WT and Rev-KO mice for 24 h and measured the expression of selected SREBP target genes that we had found to be regulated in Rev-KO mice fed ad libitum at ZT12 (see Figure 1C). As shown in Figure 2B, fasting-induced repression of SREBP target genes such as Srebp1c, Fas, Elovl6, and Aacs, also occurred in the Rev-KO background. Also the fasting-induced up-regulation of Insig2a[41] reached similar expression levels in Rev-KO and control animals (Figure 2B). The fold-up-regulation was obviously less dramatic in the Rev-KO because basal expression levels in the fed state were already 10-fold higher than in control animals.

We next fed mice with a high carbohydrate–no fat diet that is known to induce the expression of lipogenic SREBP1c target genes such as Srebp1c, Fas, and Elovl6. Again, we measured the effects at ZT12 in Rev-KO and control mice, and found that diet-dependent regulation was independent of the genotype (Figure 2C). In addition, we did not notice any differences in plasma insulin concentrations between the genotypes at ZT12 (when SREBP target genes are differentially regulated in WT and Rev-KO mice), at ZT4 (the beginning of the day phase), or at ZT16 (the beginning of the night phase) (Figure 2D).

These findings suggest that the regulation of the SREBP pathway is governed by at least two independent processes. In addition to its regulation by fasting and feeding that has been described before, the circadian clock, through rhythmic REV-ERBα accumulation, regulates the SREBP pathway independently of the feeding regimen.

We wished to determine whether the temporal misregulation of SREBP1c also translated to corresponding alterations in the physiology of Rev-KO mice. SREBP1c target genes are implicated in the control of triglyceride synthesis in vivo, and Rev-KO mice indeed showed an altered accumulation of hepatic triglycerides (Figure 3A). Moreover, AACS and HMGCR are key enzymes in cholesterol synthesis, and LDLR is responsible for the uptake of low-density lipoprotein (LDL)-cholesterol from the blood. We hence measured the hepatic cholesterol and the plasma LDL-cholesterol in Rev-KO and WT animals. As shown in Figure 3B, Rev-KO animals had only a modest (albeit statistically significant) decrease in hepatic cholesterol at ZT12. By contrast, plasma LDL-cholesterol levels were highly increased in Rev-KO (Figure 3C), possibly because of a defective LDL-cholesterol uptake by the LDL-receptor (Figure 1D). We also noticed that HDL-cholesterol concentrations were augmented in the plasma of Rev-KO animals specifically at ZT12 (Figure 3D and 3E). Conceivably, increased HDL-cholesterol levels in Rev-KO mice may have been caused by down- and up-regulation of the genes encoding endothelial lipase and apolipoprotein A1 (APOA1), respectively (Figure S5A and S5B). APOA1 is the major lipoprotein in HDL particles [42], and endothelial lipase is an enzyme that catabolizes HDL-cholesterol [43]. However, in contrast to what had been reported previously [44], we did not notice changes in plasma-VLDL between both genotypes, and ApocIII mRNA expression in liver was not altered in our Rev-KO or TgRev mouse models (Figure S5C).

In summary, the misregulation of key enzymes involved in cholesterol metabolism in the livers of Rev-KO mice correlate well with the cholesterol phenotype of these animals. We nevertheless want to point out that extra-hepatic tissues may also have contributed to the phenotype (see). Discussion

Cyp7a1 expression has long been known to be circadian [15], and feeding–fasting rhythms cannot fully account for the daily oscillations in bile acid synthesis [45]. Thus, the circadian clock probably contributes to cyclic Cyp7a1 expression directly, thereby enabling the anticipation of the need for bile acids. As shown in Figure 4A, both phase and magnitude of Cyp7a1 expression were severely altered in Rev-KO mice. Moreover, Cyp7a1 mRNA was constitutively high in TgRev mice (Figure 4B). In Rev-KO animals, the amount of Cyp7a1 mRNA, integrated over a day, only represented around 60% of the level seen in the WT animals. To investigate whether the 40% decrease has a consequence for bile acid metabolism, we determined the bile acid content in the gallbladder of Rev-KO animals and found a clear decrease in the most abundant primary and secondary bile acids cholate and deoxycholate (Figure 4C). Furthermore, we also noted a general tendency for other bile acid species to accumulate to lower levels in the gallbladders of Rev-KO as compared to control animals (Figure 4C).

Although it has been known for more than 30 years that Cyp7a1 is expressed in a circadian fashion, the transcription factors responsible for this regulation are still subject to controversy. Our observation that diurnal Cyp7a1 was dampened in Rev-KO mice and constitutively high in TgRev mice strongly indicated that Cyp7a1 expression was linked to the circadian oscillator by REV-ERBα. Since REV-ERBα operates as a transcriptional repressor, the underlying mechanism was most likely indirect, and could have been caused by the down-regulation of a repressor that acts on the Cyp7a1 gene. Our transcriptome profiling studies revealed the basic leucine zipper protein E4BP4/NFIL3 as a plausible candidate for such a repressor (Table S1). As expected for a direct REV-ERBα target gene, E4bp4/Nfil3 mRNA accumulation oscillated with the expected phase in WT mice (Figure S6A), and was constitutively high and low in Rev-KO mice and TgRev mice, respectively (Figure S6A and S6B). Temporal E4BP4/NFIL3 protein accumulation was found to be altered accordingly in WT and Rev-KO mice (Figure S6C). Since E4BP4/NFIL3 has been shown to repress transcription through the D-element consensus sequence 5′-RTTAGTAAY-3′ (Figure S6E) [46], a sequence also bound by the three PAR bZip transcription factors DBP, HLF, and TEF, it was tempting to speculate that the E4BP4/NFIL3 repressor and PAR bZip activators could have antagonistic effects on Cyp7a1 transcription. In cotransfection experiments, DBP [47]–[49] and E4BP4/NFIL3 (Figure S6D) indeed activated and repressed, respectively, transcription from the Cyp7a1 promoter in a dose- and D-element-dependent manner. However, the implication of PAR bZip proteins or E4BP4 in circadian Cyp7a1 regulation did not withstand scrutiny by loss-of-function experiments. As shown in Figure 5A, Dbp/Tef/Hlf triple knock-out (TKO) mice still exhibited circadian Cyp7a1 accumulation levels similar to those of WT animals (Figure 5A). If E4BP4/NFIL3 directly repressed Cyp7a1 transcription, Cyp7a1 mRNA should accumulate to higher levels in E4bp4/Nfil3-KO mice at time ZT0, when Cyp7a1 expression is low and E4BP4/NFIL3 protein highly abundant (Figures 4A and S6C). However, we found that Cyp7a1 mRNA accumulated to similar levels at ZT0 and was dampened at ZT12 in E4bp4/Nfil3-deficient animals (Figure 5B). These in vivo data argue against a putative role of E4BP4/NFIL3 as a direct Cyp7a1 repressor. An indirect role of E4BP4/NFIL3 in regulating Cyp7a1 appears to exist nevertheless, as judged by the down-regulation in the E4bp4/Nfil3-KO at ZT12 (a time point when E4BP4/NFIL3 is barely expressed; see Figure S6C).

It is known that in rodents changes in CYP7A1 activity parallel those of HMGCR activity, the enzyme that catalyzes the rate-limiting step in de novo cholesterol synthesis [50]. The specific inhibition of HMGCR by compounds such as lovastatin results in decreased Cyp7a1 mRNA accumulation [51],[52]. This decrease is prevented if mice receive the HMGCR inhibitor together with an infusion of mevalonate, the product of the reaction catalysed by HMGCR [51],[52]. In addition, inhibition of the last step in cholesterol synthesis by the compound AY9944 also decreases Cyp7a1 expression [53]. Thus circadian endogenous cholesterol synthesis [17],[54] seems to be linked to the regulation of the Cyp7a1 transcription.

In Rev-KO and TgRev mice the down- and up-regulation of Hmgcr expression observed at ZT12 and ZT0, respectively (Figure 1D), can be expected to engender corresponding changes in oxysterol production [55]. This, in turn, might have an impact on Cyp7a1 transcription by LXR, which is known to be regulated by oxysterol availability [56]. It is thus conceivable, if not likely, that this pathway can impart on circadian Cyp7a1 transcription via the activation of LXR receptors in spite of unchanged Lxrα and Lxrβ mRNA levels in Rev-KO mice (Figure 4SC). To test if LXRs could be at least in part responsible for circadian Cyp7a1 transcription, we measured Cyp7a1 transcript levels in Lxrα/Lxrβ double knock-out mice (LXRα/β) at ZT0 and ZT12. As shown in Figure 5C, control mice showed a 4-fold amplitude in Cyp7a1 mRNA levels between ZT0 and ZT12, which was absent in LXRα/β-KO mice. This effect was not due to changes in Rev-erbα levels (unpublished data). Thus, a lack of LXR activation by oxysterols at ZT12 seemed to be the most likely scenario to explain the dampened Cyp7a1 mRNA accumulation seen in Rev-KO mice. Similarly, the up-regulation of Hmgcr seen at ZT0 in TgRev animals could have led to increased levels of oxysterols and thus an increase of Cyp7a1 transcription by LXR receptors.

All animal studies were conducted in accordance with the regulations of the veterinary office of the State of Geneva and Dutch and Austrian regulations according to criteria outlined in the Guide for the Care and Use of Laboratory Animals prepared by the National Academy of Sciences, as published by the National Institutes of Health. Rev-KO mice have been previously described [8]. Transgenic mice that overexpress hepatic REV-ERBα contain two transgenes [11]. Briefly, the first transgene encodes an HA-epitope-tagged REV-ERBα cDNA controlled by a tetracycline responsive element and the second one expresses a hepatocyte-specific tetracycline-dependent transactivator. In the presence of doxycycline (Dox), transactivator binding is inhibited and thus the Rev-erbα transgene remains silent. Dox-fed mice or mice containing only one of the two transgenes were used as control animals. Lxrα/Lxrβ double knock-out and Dbp/Tef/Hlf triple knock-out have been previously described [69],[70]. The E4bp4/Nfil3 knock-out mice have been generated by the laboratory of Andrew Thomas Look (Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, Massachusetts). Briefly, exon 2 of the E4bp4/Nfil3 gene that contains the full coding sequence has been replaced by a neomycin cassette (A.T. Look, personal communication). As expected, no E4bp4/Nfil3 mRNA could be detected in these mice, see Figure S11. Except for Figure 2B and 2C where mice were fasted for 24 h or were fed a high carbohydrate–no fat diet (Harlan Teklad TD03314), animals received a normal chow diet. Mice were maintained under standard animal housing conditions (12-h light/12-h dark cycle), with free access to food and water.

Whole-cell liver RNAs from 18 WT and 18 Rev-KO animals (3–4 mo old) humanely killed at ZT12 were used. Three RNA pools of six mice per pool were assembled by mixing equal amounts of RNA. For liver-specific transgenic Rev-erbα mice, six animals each for Dox-treated and untreated mice were humanely killed at ZT12 and whole-cell liver RNA was extracted from each animal. Two RNA pools of three animals were assembled by mixing equal amounts of RNA. For Rev-KO, TgRev and WT control animals, 5 µg of pooled RNA were employed for the synthesis of biotinylated cRNA, and 8.75 µg of this cRNA were hybridized to Affymetrix Mouse Genome 430 2.0 array (according to the Affymetrix protocol). To identify differentially expressed transcripts, pairwise comparisons were carried out using Affymetrix GCOS 1.2 software. In the Rev-KO versus WT mice experiment, each of the three experimental samples was compared to each of the three reference samples, resulting in nine pairwise comparisons. This approach is based on the Mann-Whitney pairwise comparison test and allows the ranking of results by concordance, as well as the calculation of significance (p-value) for each identified change in gene expression [71],[72]. Genes for which the concordance in the pairwise comparisons exceeded the imposed threshold of 77% (seven out of nine comparisons) were considered to be statistically significant. In addition, we only selected transcripts whose accumulation had an average change of at least 1.5-fold. For these selected probe sets microarray data for the transgenic Rev-erbα mice was then extracted (Figure 1A) from the study reported in Kornmann et al. [11].

cDNA was synthesized from 1 µg of liver whole-cell RNA using random hexamers and Superscript II reverse transcriptase (RT) (Invitrogen) following the supplier's instructions. Five percent of this cDNA was PCR amplified (7900HT Sequence Detection Systems, Applied Biosystems) by using the Sybr Green master mix (Applied Biosystems), and raw threshold-cycle (Ct) values were calculated with SDS 2.0 software (Applied Biosystems). Mean values were calculated from triplicate PCR assays for each sample and normalized to those obtained for Cyclophillin and Eef1a1 transcripts, which served as internal controls [73]. For a list of primers used in this study, see Table S3.

Nuclear extracts were prepared by the NUN procedure as described previously [47], and Western blotting was performed according to standard protocols using a mouse anti-SREBP1 monoclonal antibody (BD Biosciences 557036).

Blood samples were collected at Zeitgeber times 12 and 20 (ZT12 and ZT20) in lithium heparin containing tubes (BD Microtainer) and centrifuged for 10 min at 4,500 rpm. Plasma supernatants were kept frozen at −20°C. The levels of HDL- and LDL-cholesterol were measured using a Roche-diagnostics Enzymatic kit for the Hitachi 902 robot according to the manufacturer's instruction. For Rev-KO mice and WT animals (humanely killed at ZT12 and 20) plasma samples were pooled (n = 6) and lipoproteins separated by fast protein liquid chromatography (FPLC) on a Superose 6B 10/30 column (Amersham Bioscience). The level of insulin was measured using an ELISA kit (Mercodia 10-1150-01) according to the manufacturer's instruction, from blood samples collected at ZT4, ZT12, and ZT16.

Mice were anesthetized by intraperitoneal injection of Hypnorm (fentanyl/fluanisone 1 ml/kg) and diazepam (10 mg/kg). Bile was collected during 30 min by cannulation of the gallbladder, and total bile acids were measured enzymatically [74]. Total gallbladder cholesterol was measured according to Kruth et al [75]. For the determination of specific bile acid species, bile acids extracted from bile [76] were analyzed by gas chromatography mass spectrometry (GC-MS).

Livers were homogenized and lipids extracted as described by Bligh and Dyer [76]. Hepatic triglyceride and cholesterol concentrations were measured using an enzymatic kit (Roche Molecular Biochemicals, Mannheim, Germany) according to the manufacturer's instruction.

The GenBank (http://www.ncbi.nlm.nih.gov/Genbank↗) accession numbers for the genes and gene products discussed in this paper are Aacs (NM_030210↗); Acot3 (NM_134246↗); Apoa1 (NM_009692↗); ApoCIII (NM_023114↗); Bmal1 (NM_007489↗); Cyp7a1 (NM_007824↗); Dbp (NM_016974↗); E4bp4/Nfil3 (NM_017373↗); Elovl3 (NM_007703↗); Elovl5 (NM_134255↗); Elovl6 (NM_130450↗); Fabp5 (NM_010634↗); Fads2 (NM_019699↗); Fas (NM_007988↗); Fgf15 (NM_008003↗); G6pc (NM_008061↗); Glucokinase (NM_010292↗); Hlf (NM_172563↗); Hmgcr (NM_008255↗); Insig1 (NM_153526↗); Insig2 (NM_133748↗); Ldlr (NM_010700↗); Lipase endothelial (NM_010720↗); Lpl (NM_008509↗); Lxrα (NM_013839↗); Lxrβ (NM_009473↗); Pctp (NM_008796↗); Pepck (NM_011044↗); Rev-erbα (NM_145434↗); Scap (NM_001001144↗); Shp (NM_011850↗); Srebp1 (NM_011480↗); Srebp2 (NM_033218↗); Tef (NM_017376↗).

The ArrayExpress repository (http://www.ebi.ac.uk/arrayexpress↗) accession number for the microarray data is E-TABM-726.