The Nuclear Receptor Rev-erbα Regulates Adipose Tissue-specific FGF21 Signaling*

FGF21 is an atypical member of the FGF family that functions as a hormone to regulate carbohydrate and lipid metabolism. Here we demonstrate that the actions of FGF21 in mouse adipose tissue, but not in liver, are modulated by the nuclear receptor Rev-erbα, a potent transcriptional repressor. Interrogation of genes induced in the absence of Rev-erbα for Rev-erbα-binding sites identified βKlotho, an essential coreceptor for FGF21, as a direct target gene of Rev-erbα in white adipose tissue but not liver. Rev-erbα ablation led to the robust elevated expression of βKlotho. Consequently, the effects of FGF21 were markedly enhanced in the white adipose tissue of mice lacking Rev-erbα. A major Rev-erbα-controlled enhancer at the Klb locus was also bound by the adipocytic transcription factor peroxisome proliferator-activated receptor (PPAR) γ, which regulates its activity in the opposite direction. These findings establish Rev-erbα as a specific modulator of FGF21 signaling in adipose tissue.

Adipose tissue is an important fat storage and endocrine organ whose dysfunction is closely associated with the development of obesity, diabetes mellitus, and cardiovascular disease (1)(2)(3)(4)(5). Nuclear receptors are DNA-binding proteins that directly regulate gene expression in response to ligands derived from endocrine glands, metabolism, diet, and the environment (6). They are widely believed to act as key regulators in various physiological processes, such as circadian rhythm, development, reproduction, energy homeostasis, and metabolism (7,8).
The nuclear receptor Rev-erb␣ differs from other members of the nuclear receptor superfamily because it lacks a classical activation domain and thus functions as a constitutive repressor of transcription (9 -11). Acting in this repressive manner, Rev-erb␣ has been described as a core component of the mammalian biological clock (12) that links circadian rhythms to metabolism in diverse tissues. Indeed, studies from different groups have revealed Rev-erb␣ as a key regulator of multiple biological processes in various metabolic tissues, including liver (13)(14)(15), macrophages (16), muscle (17), brown fat (18), and brain (19). However, little is known about potential role in white adipose tissue (WAT). 3 FGF21 is an atypical member of the FGF superfamily that, because of a lack of a heparin binding domain, is able to escape into the circulation, functioning as a hormone to regulate carbohydrate and lipid metabolism (20,21). In the metabolic context, FGF21 was first discovered to induce glucose uptake in 3T3L1 adipocytes (22). Subsequently it was demonstrated that FGF21 administration to obese rodents and non-human primates improves hyperglycemia, lowers elevated triglyceride levels, and reduces body weight (22)(23)(24)(25). In rodents, the mechanisms underlying FGF21 actions include improving whole-body insulin sensitivity and ␤ cell function, reducing hepatic lipogenesis, and enhancing brown fat thermogenic activity (22,23,(25)(26)(27). These effects identify FGF21 as an attractive therapeutic agent for the treatment of metabolic disease. Mechanistically, FGF21 interacts directly with the extracellular domain of the membrane bound co-factor ␤Klotho (KLB) in the FGF21-KLB-FGF receptor (FGFR) complex to activate FGF receptor substrate 2␣ and ERK1/2 phosphorylation (28 -30). Although FGFRs are expressed in most tissues, KLB expression is restricted to a few, including liver, WAT, brown adipose tissue (BAT), and hypothalamus (31). Mice lacking KLB are resistant to both acute and chronic effects of FGF21. However, the acute insulin-sensitizing effects of FGF21 are also absent in mice with specific deletion of adipose KLB (32) or FGFR1 (33), consistent with the notion that direct adipose tissue activation is required for FGF21 action (34).
Here, based on integrated analysis of transcriptomes and cistromes in Rev-erb␣ KO mice, we found that KLB mRNA and protein are markedly induced in WAT, but not in BAT or liver, of mice lacking Rev-erb␣. Moreover, mice lacking Rev-erb␣ were hyperresponsive to FGF21 in WAT. These results reveal a tissue-specific modulation of FGF21 signaling by Rev-erb␣ via the repression of ␤Klotho.

Experimental Procedures
Animals-The Rev-erb␣ KO mice were obtained from B. Vennström and backcrossed for more than seven generations with C57/Bl6 mice. They were maintained on a standard diet under 12 h-light/12 h-dark cycles. The experiments were performed on 12-to 15-week-old males and females. The tissues were harvested at zeitgeber time (ZT) 10 (5 p.m.) when Rev-erb␣ protein level peaks in WT mice. Recombinant human FGF21 was provided by Lilly Research Laboratories. Animal care and use procedures followed the guidelines of the Institutional Animal Care and Use Committee of the University of Pennsylvania in accordance with the guidelines of the National Institutes of Health.
Reverse Transcription and Quantitative Real-time PCR (RT-qPCR)-Total RNA was isolated from tissues using TRIzol (Invitrogen), followed by purification with the RNEasy mini kit (Qiagen). 1 g of purified RNA was used to generate cDNA (Applied Biosystems), and RT-qPCR analysis was performed. Amplicons were detected with Power SYBR Green Master Mix (Applied Biosystems). Relative gene expression levels were determined by the standard curve method, followed by normalization to the housekeeping gene 36B4.
Microarray-Each RNA sample was processed with the Ambion WT expression kit and the GeneChip WT terminal labeling and controls kit (Affymetrix), and hybridized to the Mouse Gene 1.0 ST Array (Affymetrix). Differential gene expression was determined using a one-way analysis of variance with cutoff of a p value of 0.05/0.01 and a -fold change of 1.3/1.5. The EWAT microarray data are available in the GEO under accession number GSE79166. The microarray data of PPAR␥ knockdown in 3T3-L1 adipocytes are from GEO accession number GSE14004 (35).
ChIP-Mice were euthanized, and tissue was harvested immediately, quickly minced, and cross-linked in 1% formaldehyde for 20 min, followed by quenching with 1/20 volume of 2.5 M glycine solution and two washes with PBS. Cell lysates with fragmented chromatin were prepared by probe sonication in ChIP dilution buffer (50 mM HEPES, 155 mM NaCl, 1.1% Triton X-100, 0.11% sodium deoxycholate, 0.1% SDS, 1 mM PMSF, and a complete protease inhibitor tablet (pH 7.5). ChIP was performed using 2-10 g of antibodies. For ChIP-seq, material from three to four mice was pooled prior to library generation.
ChIP-seq and Cistromic Analysis-ChIP DNA was prepared for sequencing according to the amplification protocol pro-vided by Illumina. Deep sequencing was performed using Illumina Genome AnalyzerIIx, and sequencing reads were obtained using the Solexa analysis pipeline and mapped to the mouse genome (UCSC Genome Browser mm9) using Bowtie software (36). Peak calling was carried out by the HOMER software suite (37). Location analysis of ChIP-seq peaks was performed using the Cistrome platform (38). Rev-erb␣ ChIP-seq in liver and BAT are from GEO accession numbers GSE26345 (39) and GSE79167 (18), respectively. Rev-erb␣ ChIP-seq in EWAT is available under GEO accession number GSE79166. PPAR␥ ChIP-seq in 3T3-L1 is from GEO accession number GSE27450 (40), PPAR␥ ChIP-seq in EWAT is from GEO accession number GSE64458 (41), and PPAR␣ ChIP-seq in liver is from accession number GSE61817 (42). Histone 3 lysine 27 acetylation ChIP-seq in EWAT is from GEO accession number GSE63964 (43).
Global Run-on Sequencing (GRO-Seq)-The nuclear run-on assay was performed as described previously (44). GRO-seq reads were mapped to the mouse reference genome (mm9) and extended to 150-bp fragments. Gene body quantification was computed and normalized to reads per kb per ten million reads for each Refseq-annotated transcript. ChIP-seq and GRO-seq data were visualized using Integrative Genomics Viewer (IGV) (45). GRO-seq in 3T3-L1 adipocytes is from GEO accession number GSE56747 (46). GRO-seq performed on liver is from GEO accession number GSE59486 (44). GRO-seq performed on EWAT, IWAT, and BAT is available in GEO accession number GSE79168. Statistics-Comparisons between two groups were performed using Student's t test. Multiple-group comparisons were performed by a two-way analysis of variance. Statistical significance was defined as p Ͻ 0.05. All data are presented as mean Ϯ S.E.

Results
␤Klotho (Klb) Is a Direct Target Gene of the Nuclear Receptor Rev-erb␣ in Adipose Tissue-To investigate the role of Rev-erb␣ in adipose tissue, epididymal WAT (EWAT) was collected from WT and Rev-erb␣ KO mice at ZT 10 (5 p.m.), when the Rev-erb␣ protein level peaks in WT mice, and subjected to microarray analysis of gene expression. Consistent with the transcriptional repression function of Rev-erb␣, about twice as many genes were induced than decreased (with cut-off of -fold change Ͼ1.5 and p Ͻ 0.01) in Rev-erb␣ KO EWAT compared with the WT (Fig. 1). To identify induced genes that were direct targets of Rev-erb␣, we performed chromatin immunoprecipitation using antibodies against Rev-erb␣, followed by deep sequencing (ChIP-seq) of EWAT at ZT 10 ( Fig. 1). More than 5000 Rev-erb␣ peaks were detected, and the integrated analysis of the genome-wide studies identified 40 genes up-regulated in the Rev-erb␣ KO EWAT microarray (with cut-off of -fold change Ͼ1.5 and p Ͻ 0.01) and with nearby Rev-erb␣ binding sites (with cut-off of binding strength Ͼ1.5 rpm). These 40 genes were considered to be potential direct target genes of Rev-erb␣ in EWAT. In this study we focused on ␤Klotho (Klb) because it was one of the most induced potential target genes of Rev-erb␣ in EWAT and is a critical regulator of FGF21 signaling and intermediate metabolism in vitro (28,29) and in vivo (32,47).
Klb mRNA and Protein Levels Are Increased in White Adipose Tissue of Rev-erb␣ KO Mice Compared with the WT-Consistent with our microarray result, the mRNA levels of Klb were robustly increased in both EWAT and inguinal WAT (IWAT)  of Rev-erb␣ KO mice compared with WT mice (Fig. 2A). In contrast, no significant change in Klb mRNA level was observed in liver, hypothalamus, and BAT ( Fig. 2A). KLB protein levels were also markedly elevated in EWAT and IWAT, but not liver, of Rev-erb␣ KO mice (Fig. 2, B and C). These results suggest that Rev-erb␣ regulates the expression of Klb specifically in WAT.
Rev-erb␣ Controls Enhancer RNA Expression at the Klb locus Specifically in Mouse White Adipose Tissue-Inspection of the ChIP-seq data revealed three robust Rev-erb␣ binding sites within the first intron of the Klb locus in EWAT (48) but also in liver (39) and BAT (18) (Fig. 3A, green tracks, yellow boxes). To address the mechanism of the tissue-specific regulation of Klb transcription by Rev-erb␣, we performed global run-on followed by high-throughput sequencing (GRO-seq) to measure nascent transcription in WAT, liver, and BAT of WT and Rev-erb␣ KO mice (Fig. 3A, the positive and negative strands are illustrated in blue and red, respectively). This revealed increased transcription of the Klb gene body in WAT, but not in BAT or liver, of Rev-erb␣ KO mice relative to the WT (Fig. 3A).
Because GRO-seq measures transcription wherever it occurs in the genome, it also detects bidirectional transcripts at enhancers, called eRNAs, whose regulation often correlates with that of nearby gene expression and is thus a useful measure of enhancer activity (49 -51). Robust eRNA transcription was detected in the first intron of the Klb locus (Fig. 3A, three yellow  boxes) and was greatest at the Rev-erb␣ binding site closest to the transcriptional start site (Fig. 3A, left yellow boxes). Indeed, the sites of strongest Rev-erb␣ binding were highly enriched for H3K27 acetylation in EWAT (Fig. 3A) (43), a well recognized epigenomic feature of active enhancers (52).
Changes in eRNA upon deletion of a transcription factor that binds at the enhancer can separate functional from non-functional binding sites (44). In this context, it is remarkable that deletion of Rev-erb␣ markedly increased the transcription of the most abundant eRNA in EWAT and IWAT (Fig. 3A, left yellow boxes), whereas expression of the eRNA was constitutive in liver and BAT (Fig. 3A). When we zoomed in on the first peak, we found that it did not overlap in EWAT and liver (Fig.  3B, blue boxes). Indeed, the peaks were ϳ250 bp apart (Fig. 3B). Furthermore, the EWAT peak contained a DR2 motif and not the liver peak, explaining the different effects of Rev-erb␣ deletion on Klb transcription in EWAT and liver. In BAT, although Rev-erb␣ bound exactly the same site as in EWAT, that site was not at the center of the bidirectional eRNA locus (Fig. 3B, blue  boxes), suggesting that it was not a part of the functional enhancer. The specific role of Rev-erb␣ at this enhancer in WATs was confirmed quantitatively by RT-qPCR in the different tissues of Rev-erb␣ KO mice compared with the WT (Fig.  3C). These results suggest that Rev-erb␣ is functionally active at this site in WAT but not BAT or liver, which would explain the tissue-specific regulation of Klb by Rev-erb␣.
Rev-erb␣ Controls the Circadian Transcription of Klb in WAT-As expected, Rev-erb␣ expression was circadian, peaking at ZT 6 -10 (1-5 p.m.), with a robust amplitude in the WAT and liver of WT mice (Fig. 4A). We hypothesized that Rev-erb␣ would repress the transcription of Klb in a circadian manner in WAT but not in liver. Indeed, Klb mRNA expression was circadian and antiphase to Rev-erb␣ in EWAT and IWAT, but not in liver, of WT mice (Fig. 4B). This circadian pattern was greatly attenuated in EWAT and IWAT of Rev-erb␣ KO mice (Fig. 4B) because of increased Klb expression at its normal nadir, indicating that Rev-erb␣ was responsible for the antiphase rhythm. RT-qPCR analysis demonstrated a similar Rev-erb␣-dependent circadian rhythm of the eRNAs at the functional Klb enhancer in EWAT and IWAT, strongly suggesting direct transcriptional control of the circadian expression of Klb mRNA (Fig. 4C). Surprisingly, although the overall KLB protein expression was dramatically increased in Rev-erb␣ KO IWAT compared with the WT, the expression of KLB protein was not nearly as circadian as its mRNA, and the hint of a rhythm was not in synch with the mRNA (Fig. 4D), suggesting that posttranscriptional factors dampen the endogenous rhythm of KLB protein.
PPAR␥ Directly Modulates the Rev-erb␣-regulated Enhancer at the Klb Locus-Given that Rev-erb␣ often binds to the genome near lineage determination factors (48), we hypothesized that the Rev-erb␣-controlled enhancer might also be bound and modulated by PPAR␥, the master regulator of adipocyte development and function (53)(54)(55). Indeed, PPAR␥ bound at the major Rev-erb␣ binding sites at the Klb locus in EWAT (41) (Fig. 5A, purple track). Interestingly, we found that PPAR␣ bound at the major Rev-erb␣ binding sites at the Klb locus in liver (42) (Fig. 5A, orange track). Furthermore, we noted binding of PPAR␥ at the major Rev-erb␣ binding sites at the Klb locus in 3T3-L1 mouse adipocytes (40) (Fig. 5B). Moreover, analysis of GRO-seq in 3T3-L1 adipocytes (46) revealed that transcription at both the Klb gene body and the most transcription start site-proximal enhancer was increased as early as 10 min after addition of rosiglitazone (rosi), a potent synthetic agonist ligand of PPAR␥ (Fig. 5C). Moreover, Klb gene expression was down-regulated 14.6-fold after silencing PPAR␥ in 3T3-L1 adipocytes (35) (Fig. 5D), indicating that PPAR␥ was required for Klb transcription. These results show that PPAR␥ regulates the enhancer activity at the Klb locus in the opposite direction of Rev-erb␣.
FGF21 Action Is Enhanced in WAT of Rev-erb␣ KO Mice-Previous work has demonstrated that KLB is essential for FGF21 activity both in vitro (28,29) and in vivo (32,47) and that KLB in adipose tissue contributes to the beneficial metabolic actions of FGF21 (32). Therefore, we tested whether the upregulation of KLB in the WAT of Rev-erb␣ KO mice sensitized the mice to the effect of FGF21 by treating WT and Rev-erb␣ KO mice with either vehicle or recombinant human FGF21 protein and then measuring the expression of FGF21 target genes in WAT. As expected, FGF21 significantly induced the EWAT expression of cFos, Egr1, and Glut1 (by 2.1-, 14.4-, and 1.7-fold, respectively) in WT mice (Fig. 6, A-C). However, the  Fig. 3A) (39,48) and PPAR␥ (purple track) (41) or PPAR␣ (orange track) (42) binding at the Klb locus in, respectively, EWAT or liver of WT mice. The major Rev-erb␣ peak at the Klb locus is highlighted in pink. The y axis scale refers to the normalized tag count per million reads. B, ChIP-seq profiles of PPAR␥ binding at the Klb locus in 3T3-L1 adipocytes (40). The y axis scale refers to the normalized tag count per million reads. C, GRO-seq was performed in 3T3-L1 adipocytes treated with rosiglitazone (Rosi) for 10 min or left untreated (46). Genome browser views of nascent transcripts at the Klb locus are shown. GRO-seq signals on the ϩ and Ϫ strand are illustrated in blue and red, respectively. The y axis scale refers to the normalized tag count per million reads. D, Klb gene expression in 3T3-L1 adipocytes treated with siRNA against scrambled (siSCR) or PPAR␥ (siPPAR␥) mRNA (35). Data are expressed as the mean Ϯ S.E. and normalized to the WT (Student's t test; *, p Ͻ 0.001 versus the WT; n ϭ 3).
induction of these three genes by FGF21 treatment was much more dramatic in Rev-erb␣ KO mice (3.9-fold for cFos, 34.8fold for Egr1, and 4-fold for Glut1) (Fig. 6, A-C). Similarly, a more robust FGF21 response was also observed in IWAT of Rev-erb␣ KO mice compared with WT mice (Fig. 6, D-F). As described previously, mice lacking Rev-erb␣ displayed a similar body weight (Fig. 6G) and increased adiposity (Fig. 6H) compared with wild-type mice on a chow diet (56). Moreover, Rev-erb␣ KO mice exhibited mild hyperglycemia without insulin resistance compared with wild-type mice on a chow diet (56). Of note, the modest hyperglycemia exhibited by Rev-erb␣ KO mice (Fig. 6I) was abrogated by FGF21 treatment (Fig. 6I). Together, these results indicate that Rev-erb␣ plays an important role in restraining FGF21 signaling in WAT, most likely through the direct repression of its co-receptor, KLB.

Discussion
We have demonstrated that the deletion of Rev-erb␣ leads to the up-regulation of ␤Klotho, an essential co-receptor for FGF21, specifically in WAT, where it modulates the downstream effects of the FGF21 pathway. These findings reveal an important and previously unrecognized role for Rev-erb␣ in adipose function.
Rev-erb␣ bound to the Klb locus in different metabolic tissues, including WAT, BAT, and liver, but the induction of Klb transcription upon Rev-erb␣ deletion was restricted to WAT. This is consistent with previous findings that Rev-erb␣ is not transcriptionally active at all of its cistromic binding sites in the liver (44,48). The Rev-erb␣ dependence of eRNA transcription in WAT but not in other tissues may not only reflect the activity of the Klb enhancers but could also be playing a direct role in transcriptional regulation, as reported in macrophages (57).
Some tissue-specific functions of Rev-erb␣ are explained by its selective binding near lineage determination factors such as hepatocyte nuclear factor 6 (HNF6) in liver (48). In this context, it is noteworthy that we found the Rev-erb␣ binding regions of the Klb locus to also be occupied by the nuclear receptor PPAR␥, the master lineage determination factor for adipose tissue (53)(54)(55). Ligand activation of PPAR␥ acted in the opposite direction, inducing the transcription of Klb and its eRNAs. KLB was also found to be up-regulated in EWAT of mice treated with PPAR␥ ligands (58,59). Moreover, knockdown of PPAR␥ led to a decrease in KLB mRNA level in 3T3-L1 adipocytes. However, PPAR␥ is not likely to be specifically required for Rev-erb␣ binding in WAT because we observed comparable binding in liver, which expresses very low levels of PPAR␥. Therefore it is likely that WAT-specific coregulator recruitment or chromatin remodeling plays a role in the tissue specificity of Rev-erb␣ regulation of Klb.
Previous studies have identified a clear link between the molecular clock and hepatic FGF21 signaling (60,61), including suppression of Fgf21 transcription by E4BP4, which is a direct Rev-erb␣ target in liver (60). In addition, Rev-erb␣ has been shown to modulate FGF21 expression in liver (62). Given the circadian expression of Rev-erb␣ in WAT, we hypothesized that the repression by Rev-erb␣ would also be rhythmic. Indeed, we demonstrated circadian mRNA expression of Klb and its enhancer activity in WAT, and this was antiphase to, and dependent upon, the expression of Rev-erb␣. Surprisingly, although KLB protein levels were dramatically elevated at all times of day in Rev-erb␣ KO WAT compared with the WT, its normal protein expression was hardly, if at all, circadian and did not follow the pattern of circadian Klb gene expression of the Klb mRNA. This phenomenon may be caused by posttranscriptional mechanisms and suggests that KLB protein has a long half-life that reduces the impact of circadian gene expression on KLB protein levels. The physiological significance of the robust, Rev-erb␣-dependent circadian rhythm of Klb gene expression is thus unclear. It is possible that the Klb mRNA half-life is regulated so that the Rev-erb␣-driven rhythm becomes significant in some physiological states, but this has yet to be shown. Because KLB protein levels were not found to be normally circadian but were markedly and constitutively elevated in WAT of Rev-erb␣ KO mice, we focused mainly on the effects of FGF21 in this system.
The beneficial effects of FGF21 on glucose metabolism and body weight have identified it as a central player in regulating metabolic processes, and adipose tissue is required for its antidiabetic actions (32,33,63). These findings show that FGF21 signaling is modulated by Rev-erb␣ in WAT, not by altering FGF21 levels but through the regulation of its co-receptor KLB in WAT but not in liver. Consistent with this, FGF21 target genes, including early response genes and GLUT1, were superinduced in WAT after treatment of Rev-erb␣ KO mice with recombinant FGF21. Moreover, FGF21 had a modest effect on glycemia in Rev-erb␣ KO mice. It is tempting to speculate that this is related to the increase in GLUT1 expression, but there are many other potential mechanisms, including systemic effects of FGF21, beyond its direct actions on WAT.
In summary, this study extends our knowledge about Rev-erb␣ functions in mouse adipose tissue and provides further sights into the mechanisms underlying the FGF21 actions in fat. The development of appropriate synthetic Rev-erb␣ antagonists could be beneficial in facilitating FGF21-related therapeutic approaches for various metabolic diseases.