Transcriptional Control of Circadian Rhythms and Metabolism: A Matter of Time and Space

Abstract

All biological processes, living organisms, and ecosystems have evolved with the Sun that confers a 24-hour periodicity to life on Earth. Circadian rhythms arose from evolutionary needs to maximize daily organismal fitness by enabling organisms to mount anticipatory and adaptive responses to recurrent light-dark cycles and associated environmental changes. The clock is a conserved feature in nearly all forms of life, ranging from prokaryotes to virtually every cell of multicellular eukaryotes. The mammalian clock comprises transcription factors interlocked in negative feedback loops, which generate circadian expression of genes that coordinate rhythmic physiology. In this review, we highlight previous and recent studies that have advanced our understanding of the transcriptional architecture of the mammalian clock, with a specific focus on epigenetic mechanisms, transcriptomics, and 3-dimensional chromatin architecture. In addition, we discuss reciprocal ways in which the clock and metabolism regulate each other to generate metabolic rhythms. We also highlight implications of circadian biology in human health, ranging from genetic and environment disruptions of the clock to novel therapeutic opportunities for circadian medicine. Finally, we explore remaining fundamental questions and future challenges to advancing the field forward.

Graphical Abstract

Graphical Abstract.

Essential Points.

• Circadian rhythms synchronize mammalian physiology with daily environmental variations.

• Molecular clocks comprised of transcriptional-translational feedback loops generate oscillating rhythms that orchestrate rhythmic physiology.

• Clock transcription factors modulate epigenomic rhythms, including histone modification and chromatin architectural oscillation.

• Reciprocal crosstalk between the central clock and peripheral clocks aligns feeding behavior with circadian metabolic processes.

• The clock serves a dual role as a driver and an integrator of metabolism.

• Circadian medicine has the potential to leverage circadian physiology to add a new dimension to pharmacological treatments of disease.

The daily rotation of Earth around its axis creates an inherently dynamic ecosystem with recurrent light-dark cycles. In this geophysical framework, life emerged inextricably with the Sun, which delivers light to Earth’s surface with a periodicity of 24 hours. As a source of energy and genotoxic stress, light exerts a powerful evolutionary pressure to all photosensitive forms of life. Although light provides photic energy necessary for photosynthesis, it also poses formidable environmental insults in the form of damaging radiation and drastic oscillations in temperature through its light-dark cycles (1, 2). As a result, a myriad of life forms and their biological processes arose and evolved with circadian rhythms, which in Latin translates to “rhythms around (circa-) a day (-diem).”

From an evolutionary prospective, circadian rhythms confer distinct survival advantages. While irreversible environmental perturbations drive the selection of one or a few favorable phenotypes over generations of species, circadian variations of the environment select not only favorable phenotypes but also the time at which they should appear during the day. This evolutionary pressure gave rise to intrinsic timekeeping mechanisms that enabled circadian organization of phenotypic variations, which are temporally programmed to anticipate and adapt to the daily environmental cycle in order to maximize organismal fitness on a 24-hour time scale. It is on this evolutionary basis that nearly all fundamental biological processes are coupled with the rising and setting of the Sun (1, 2).

The notion that intrinsic timekeeping mechanisms coordinate circadian rhythms began with de Mairan’s observation in the 18th century that daily leaf movements of the Mimosa pudica plant occurred autonomously in darkness (3). A seminal study by Konopka and Benzer two centuries later demonstrated the genetic basis of rhythmic locomotor activity in fruit flies (4). This work inspired positional cloning experiments in the laboratories of Jeffrey Hall, Michael Rosbash, and Michael Young that ultimately identified the first clock gene period, for which they shared the 2017 Nobel Prize in Physiology or Medicine (5–8). Since the discovery of the Period gene in fruit flies, many clock genes encoding transcription factors (TFs) that compose the molecular clock have been cloned and characterized in mammalian species. However, even as the complexity of the clock field has grown, fundamental questions remain as to how these molecular cogs function together to keep the clock ticking at the systems level, ranging from transcriptional oscillations to cell autonomous rhythms and ultimately organismal physiology.

In the past decade, the advent of genome-wide techniques and genome-editing tools has offered a tangible opportunity to systemically untangle the clock at the magnitude of speed and complexity that has not been afforded in the past. In this review, we will focus on recent advances in our understanding of the transcriptional architecture of the mammalian clock and highlight its role in human physiology and pathology, with a specific emphasis on circadian metabolic homeostasis.

Mechanisms of the Mammalian Clock

Transcription factor network

The mammalian clock is orchestrated by activator and repressor TFs interlocked in cell-autonomous transcription-translational feedback loops with a delay, a fundamental governing principle of the molecular clock (9, 10). Nearly all clock TFs are embedded in negative feedback loops that lead to transcriptional repression of their own genes. With declining production, self-repression is abated, and transcription begins again, thereby generating oscillating levels of transcription and translation that recur every 24 hours. At the core, the mammalian clock consists of a forward limb that drives 3 regulatory limbs, which confer circadian rhythmicity to the forward limb in an interdependent manner (Fig. 1A).

Figure 1.

Transcriptional network of core clock transcription factors (TFs). (A) A composite illustration of all major regulatory limbs. (B) PER/CRY regulate circadian BMAL1 transcriptional activity. (C) ROR and REV-ERB regulate circadian transcription of BMAL1. (D) D-box TFs enhance circadian amplitude of BMAL1 transcription. (E) Different regulatory limbs regulate one another to further fine tune the clock.

The forward limb of the mammalian clock is mainly driven by 2 basic helix-loop-helix PER-ARNT-SIM (PAS) domain-containing TFs, BMAL1 (brain and muscle ARNT-like 1, also referred to as MOP3, encoded by ARNTL) and CLOCK (circadian locomotor output cycles kaput, encoded by CLOCK) (11–14). BMAL1 and CLOCK heterodimerize to form a transcriptional activator complex, which binds E-box motifs at promoters and enhancers, including those of core clock genes that constitute the negative regulatory limbs. Although BMAL1 is an indispensable component of the mammalian clock, genetics studies demonstrated functional redundancy of CLOCK with its paralogue NPAS2 (neuronal PAS domain protein 2; encoded by NPAS2) in certain tissues, such as the forebrain and suprachiasmatic nucleus (15–18).

The first regulatory limb consists of PER1/2/3 (Period, encoded by PER1, PER2, PER3) and CRY1/2 (Cryptochrome, encoded by CRY1, CRY2). They are repressive TFs whose orthologues in fruit flies were first found to have clock functions, albeit working through divergent molecular mechanisms in the fruit fly and mammalian systems. (19–22). In the mammalian clock, the forward limb induces the expression of PER and CRY, which heterodimerize in the cytoplasm and translocate to the nucleus upon phosphorylation by casein kinase I (23–26). In the nucleus, PER-CRY heterodimers bind BMAL1-CLOCK heterodimers to form quaternary complexes that act as transcriptional repressors at BMAL1-target genes, including their own. As a result, this negative feedback loop confers a circadian rhythm to BMAL1/CLOCK transcriptional activity (27–30) (Fig. 1B). Of note, similar to CLOCK and NPAS2, PER and CRY paralogues share redundant and nonredundant functions in regulation of the clock because of their differential expression patterns across various tissues. This subject has recently been reviewed in greater detail (31).

The second regulatory limb is controlled by the nuclear hormone receptor TFs, RORα/β/γ (RAR-related orphan receptor, encoded by RORA, RORB, RORC), and REV-ERBα/β (reverse strand of c-erbα, encoded by NR1D1, NR1D2) (32–36). ROR and REV-ERB recognize and compete for RORE and RevDR2 motifs at promoters and enhancers, where they have opposing effects on transcription: ROR activates transcription, whereas REV-ERB represses transcription (37, 38). The forward limb drives the expression of REV-ERB, which then competes with ROR at cis-regulatory elements throughout the genome, including the BMAL1 promoter (39-42). As a result, rhythmic competition between REV-ERB and ROR establishes a negative feedback loop at the level of BMAL1 transcription, thereby augmenting circadian transcriptional activity of the forward limb (Fig. 1C). Of note, although both REV-ERBα and β show circadian expression in most tissues, expression patterns of RORα, β, and γ are considerably more complex, highly tissue-restricted, and not necessarily circadian in certain tissues (36, 43-48). For instance, RORβ is expressed in select regions of the brain and retina, whereas a specific γ-isoform RORγT is expressed during differentiation of naïve CD4⁺ T cells into T helper 17 cells where it serves as a critical lineage-determining factor (45, 46, 49–51). Such observations underscore the diverse biological roles of RORs in addition to its clock function that entail complex tissue-specific regulation [reviewed in (52)].

The third regulatory limb comprises the proline and acidic amino acid-rich basic leucine zipper TFs, including DBP (D-binding protein, encoded by DBP), TEF (thyrotroph embryonic factor, encoded by TEF), HLF (hepatic leukemic factor, encoded by HLF), and NFIL3/ E4BP4 (nuclear factor, interleukin-3 regulated/E4 promoter binding factor 4, encoded by NFIL3) (53-55). All of these factors recognize and compete for D-box motifs at promoters and enhancers, where they stimulate transcription in a redundant manner, except for NFIL3, which represses transcription (56, 57). The forward limb drives the circadian expression of DBP in the opposite phase of NFIL3, which is separately driven by the ROR-REV-ERB axis (58-63). The temporal opposition of DBP and NFIL3 activity generates circadian expression of their target genes, including RORγ. This cascade thus coordinates the peak expression of RORγ at the trough of REV-ERB expression, thereby enhancing the amplitude of BMAL1 transcription (62) (Fig. 1D).

Through their cognate genetic elements and combinations thereof, these regulatory limbs coordinate temporal expression patterns of clock genes as well as numerous clock-controlled genes that carry out circadian functions of the cell. Although this overview aims to simplify and highlight several key aspects of rhythm generation with an emphasis on transcriptional regulation of the forward limb, it is important to consider key questions that remain in the field. For instance, there is no clear consensus on what is required to be a core clock member because not all clock TFs have the same magnitude of effect on the forward loop. For example, mice lacking both REV-ERB α/β exhibit arrhythmic behavior and disrupted expression of many clock genes, including BMAL1 (39, 41). However, another study reported that inhibition of both REV-ERB α/β does not significantly impair BMAL1 transcriptional activity, despite the lack of rhythmic BMAL1 expression (64). Similarly, constitutive expression of BMAL1 in mice lacking endogenous BMAL1 has been shown to restore circadian behavior and transcriptional rhythms of the forward limb, which are likely generated by the intact PER-CRY axis (65). This finding that rhythmic BMAL1 expression may not be required for the forward limb suggests that the ROR-REV-ERB axis likely serves as an axillary limb that fine tunes the clock. Despite its mild effect on the forward limb, REV-ERB plays an indispensable role in the clock transcriptional network as a primary clock TF that is directly controlled by BMAL1 and intimately regulates expression of other clock genes and specific clock-controlled genes that coordinate cellular and organismal rhythms, as evidenced by the α/β knockout studies.

Extensive crosstalk among the regulatory limbs is also critical for orchestrating the clock. REV-ERB represses both BMAL1 and NPAS2, 2 positive regulators of the forward limb, while simultaneously repressing the negative regulator CRY1 by binding to its intronic enhancer (66-68) (Fig. 1E). Genetic deletion of this enhancer results in reduced rhythmicity of CRY1 expression and a shorter period in locomotor activity (68, 69). Adding more complexity to the simplified model of a single cognate genetic element controlling a clock gene, several clock TFs have been found to share overlapping genomic binding sites near clock genes, suggesting potential coregulation by more than one clock TF (39, 60). In addition, not all core clock genes are driven by the forward limb. In mice constitutively expressing REV-ERBα that abrogates BMAL1 expression, PER2 maintains robust circadian expression in the liver, which is abolished in ex vivo explants, implicating a role of systemic cues in direct regulation of the cell-autonomous clock (70).

Recent studies also identified the novel circadian TF CHRONO (computationally highlighted or chromatin immunoprecipitation-derived repressor of the network oscillator, encoded by GM129) (71–73). CHRONO is expressed in a circadian manner and interacts with BMAL1-CLOCK and PER to repress BMAL1-target genes, although it remains unknown how CHRONO complements the PER-CRY axis at the biochemical and transcriptional levels, which will require further investigation for a detailed functional characterization (Fig. 1E).

Epigenetic regulation of the clock

The observation that posttranslational modifications of histone tails regulate transcription provided the first molecular basis of epigenetics, which refers to heritable changes in gene expression without changes in genotype (74). In past decades, cloning of various epigenetic regulators spurred the rapid growth of epigenetics research dedicated to elucidating the biochemical basis of epigenetic modifications and understanding the epigenetic control of gene expression in numerous biological systems, including circadian rhythms (75, 76).

Broadly, epigenetic regulators are classified as “writers or erasers” and “readers.” Epigenetic writers and erasers modify histone tails with an array of posttranslational modifications that produce distinct chromatin states, which are recognized by epigenetic readers to generate specific transcriptional outputs. DNA methylation is another major branch of epigenetic regulation (77). Current evidence suggests DNA methylation is stable throughout circadian phases and likely not directly regulated by the clock (78, 79). Thus, this review focuses on histone modifications.

Histone acetylation is regulated by opposing actions of histone acetyl transferases (HAT) and histone deacetylases (HDAC). Acetylation of H3 lysine 9 and 27 (H3K9Ac and H3K27Ac) promotes an open and easily accessible euchromatin state for TFs to bind the genome to activate transcription (80, 81). Histone acetylations can also be recognized by a group of bromodomain and extraterminal motif proteins that participate in various aspects of transcription, such as recruitment of nucleosome remodeling complexes and elongation factors (82, 83).

Histone methylation is regulated by a large number of histone methyltransferases (HMT) and histone demethylases (84). Unlike acetylation, which is generally indicative of active transcription, histone methylation can be associated with either active or inactive transcription. Basic residues on histones can be mono-, di-, or tri-methylated, each with a discrete effect on transcription. For instance, H3 lysine 4 tri-methylation (H3K4me3) is associated with active promoters, whereas H3 lysine 4 mono-methylation (H3K4me1) is associated with active enhancers (85–87). In contrast, H3 lysine 27 tri-methylation (H3K27me3) catalyzed by polycomb repressive complex 2 is a repressive marker that serves as a scaffold for heterochromatin protein 1 (HP1), which causes local nucleosome condensation to promote a compact and transcriptionally silenced heterochromatin state (88, 89).

BMAL1-CLOCK.

BMAL1-CLOCK/NPAS2 interact with the transcriptional co-activators p300, CREB-binding protein (CBP), and p300/CBP-associated factor (pCAF), all of which have HAT activity (90, 91) (Fig. 2A). In addition, CLOCK itself has been reported to have intrinsic HAT activity, which is enhanced by BMAL1 and functionally required to stimulate transcription at the PER and DBP promoters (92). BMAL-CLOCK also interact with the HMT mixed lineage leukemia 1 (MLL1) that catalyzes H3K4me3 at active promoters (93) (Fig. 2A). Interestingly, other members of the MLL family, MLL3 and MLL4, have also been shown to be rhythmically recruited to active promoters, although TFs mediating their recruitment remain unknown (94, 95). BMAL-CLOCK interact with lysine-specific demethylase 1 (LSD1) and another lysine demethylase JARID1A (Fig. 2A). Circadian phosphorylation of LSD1 by protein kinase Cα licenses LSD1 to bind BMAL1-CLOCK to activate transcription (96). Similarly, JARID1A is rhythmically recruited by BMAL-CLOCK to activate transcription (97). Intriguingly, both studies revealed that their catalytic activities are dispensable for transcriptional activation. Although the mechanisms of their catalytic-dependent and catalytic-independent functions remain unknown, an increasing number of epigenetic regulators have also been found to have catalytic-independent functions, with some studies demonstrating that certain epigenetic regulators, such as MLL3 and 4, coordinate transcription predominantly through their interactions with collaborative partners independently of their catalytic activity (98–102).

Figure 2.

Epigenetic regulators of the molecular clock. (A) Multiple epigenetic regulators associate with BMAL1/CLOCK and PER/CRY to facilitate transcriptional activation and repression, respectively. (B) ROR and REV-ERB recruit nuclear receptor coactivators and corepressors via mechanisms conserved in nuclear receptors. REV-ERB can also tether to tissue-specific transcription factors to regulate transcription in a DNA-binding domain (DBD)-independent manner.

PER-CRY.

During the repressive phase, PER-CRY interact with the Mi-2/nucleosome remodeling deacetylase (NuRD) and SIN3A corepressor complexes, both of which recruit HDAC1/2 to deacetylate histones and repress transcription (90, 103–108) (Fig. 2A). Approximately 4 hours later, PER recruits the HP1γ-Suv39h HMT complex to catalyze di- and tri-methylations on unmodified H3K9 residues to promote a local repressive chromatin state (105) (Fig. 2A). Sequential modifications of H3K9 in 2 consecutive circadian phases are mediated by 2 distinct PER complexes containing either HDAC1 or HP1γ, which are formed in a temporally discrete manner via an unknown mechanism. A separate study also demonstrated that HP1γ is recruited to these sites marked with H3K9me2, where it facilitates transcriptional repression by increasing local nucleosome density (63). In addition, PER interacts WD repeat-containing protein 5 (WDR5), a subunit of methyltransferase complexes, to augment PER-mediated transcriptional repression (107). Interestingly, WDR5 has been shown to complex with MLL1 that participates in the active phase with BMAL1, suggesting that the formation of epigenetic complexes is likely regulated by local chromatin and temporal contexts, as illustrated by the PER examples described previously (109, 110).

ROR and REV-ERB.

ROR and REV-ERB belong to the nuclear receptor (NR) superfamily consisting of TFs that share a conserved structural organization containing specific modular domains, such as DNA-binding domains (DBD) and ligand-binding domains (LBD) (111, 112). DBDs consist of 2 zinc-fingers that enable the recognition of and binding to specific DNA sequences (113, 114). LBDs consist of 12 α-helices that bind ligands with high affinity and specificity (115-117). In the classical model of NRs, ligand binding induces a conformational switch in LBDs that exposes a C-terminal helix 12 (H12) domain to recruit coactivators, whereas unliganded LBDs recruit corepressors (118, 119). Together, DBDs and LBDs serve as modular determinants of NR transcriptional activity, such that DBDs specify target genes and LBDs determine transcriptional outcomes through their interactions with ligands and coregulators.

ROR interacts with nuclear receptor coactivator (NCOA) 1 and 2 complexes (also known as steroid receptor complex or SRC1 and 2, among others names), which are members of the p160 coactivator family (120–123) (Fig. 2B). The C-terminal H12 of ROR binds short peptide leucine-x-x-leucine-leucine motifs within conserved NR interaction domains of the coactivators, termed NR boxes (124-126). This coordinated interaction recruits NCOA and its associated HAT CBP, p300, and pCAF to transactivate target genes (124, 127–130) (Fig. 2B). One interacting partner of SRC2 is the polybromo-associated BRG1/BRM-associated factor (PBAF) complex, which belongs to the ATP-dependent switch/sucrose non-fermentable chromatin remodeling complex family (121, 131–134) (Fig. 2B). Rhythmic recruitment of PBAF by ROR/SRC2 promotes chromatin accessibility, which is thought to facilitate loading REV-ERB onto chromatin during the repressive phase. In addition, another coactivator PGC1α has been shown to be rhythmically expressed in the liver and muscle tissues where it potentiates ROR transcriptional activity to coordinate the clock, although the generalizability of this interaction has not been established in other tissues (135). NCOA also interacts with coactivator-associated arginine methyltransferase 1 (also known as protein arginine N-methyltransferase 4) that has HMT activity, although its role in the clock has not been studied.

ROR is designated as an orphan NR without known endogenous ligands, although sterol derivatives have been proposed as possible candidates (136–139). Studies have yielded inconclusive results regarding ligand-dependent activation of ROR because ROR has been described as a constitutive activator that does not require ligands and also as a transcriptional repressor in certain contexts (138, 140, 141). Therefore, future studies are warranted to determine to what extent circadian ROR activity is regulated by endogenous ligands that are yet to be confirmed.

Among NRs, REV-ERB is unique in that it lacks a C-terminal H12 and thus acts as a constitutive transcriptional repressor that interacts with the nuclear receptor corepressor (NCOR) 1 and 2 complexes (NCOR2 is also known as silencing mediator of retinoid and thyroid receptor or SMRT) (37, 142–144) (Fig. 2B). For REV-ERB and other unliganded NRs, LBDs bind leucine-x-x-leucine-leucine motifs in short hydrophobic clefts of NCOR called corepressor nuclear receptor boxes or CoRNR boxes. The CoRNR boxes are analogous to the NR boxes of coactivators in that they both enable binding to NRs in a structurally similar mechanism (124). NCOR interacts with a cohort of factors that mediate transcriptional repression, including HDAC3 (119) (Fig. 2B). HDAC3 recognizes highly conserved deacetylase activating domains (DAD) of NCOR 1 and 2, which are also necessary for HDAC3 catalytic activity (145–147). Together, genomic recruitment of NCOR and HDAC3 by REV-ERB promotes histone deacetylation and clock gene repression (148-150). In support of this model, liver-specific deletion of either NCOR1 or HDAC3 de-represses BMAL1 expression (98, 148, 149). In addition, mice harboring point mutations that inactivate the DAD of NCOR1 or SMRT/NCOR2 (named N or S-DADm, respectively) exhibit a reduction in HDAC3 genomic recruitment and deacetylase activity, which results in perturbed clock gene expression and aberrant circadian behavior (151, 152). Interestingly, reconstitution of HDAC3 deletion with a catalytically inactive form of HDAC3 largely rescues repression of metabolic genes, as well as alterations in hepatic lipid metabolism, suggesting a deacetylase-independent role of HDAC3 (98).

Although ROR and REV-ERB are classically known to compete for their cognate motifs because of their highly similar DBDs, a hypomorphic form of REV-ERBα lacking its DBD has been found to be recruited to the genome in a DBD-independent manner by tethering to lineage-determining TFs (Fig. 2B). While DBD-dependent sites include clock genes coregulated by ROR and REV-ERB, DBD-independent sites are enriched for tissue-specific genes, revealing 2 discrete modes of action for REV-ERB in regulation of the clock and tissue-specific rhythms (149, 153).

Finally, heme has been identified as an endogenous ligand that potentiates RER-ERB repressive action by stabilizing its interaction with NCOR (154, 155). However, the extent to which heme dynamically controls REV-ERB activity in physiological contexts remains to be determined.

The complex epigenetic regulation of the clock revealed by these studies illuminate an extensive interplay between clock TFs and their coregulators while also raising a number of fundamental questions, particularly regarding the role of histone modifications and their respective epigenetic regulators in general transcription mechanisms. The finding that certain epigenetic regulators do not necessarily require catalytic activity challenges us to reconsider the notion that histone modifications have direct effects on transcription, a subject that remains to be intensely investigated (100–102, 156). Also, despite the growing list of known histone modifications, the biological significance and transcriptional function of these modifications have yet to withstand the test of time (157). Therefore, as our understanding of epigenetics continues to advance, biochemical and structural studies will become more imperative in unraveling the elusive details of catalytic-dependent and independent functions of epigenetic regulators that together coordinate their actions in various aspects of transcription.

From epigenomics to transcriptomics

Upon binding to the genome, clock TFs commence a temporal cascade of coregulator recruitment and epigenetic modifications that help initiate transcription. Several studies have used chromatin-immunopreciptation followed by sequencing (ChIP-seq) to delineate circadian dynamics of transcription at the genome-wide level. The first study that sought to explore this process performed ChIP-seq of REV-ERBα, which revealed that REV-ERBα binds thousands of cis-regulatory elements throughout the genome, collectively called cistrome (148). At the peak of REV-ERBα level, NCOR and HDAC3 are rhythmically recruited to REV-ERBα cistrome, which results in histone deacetylation and decreased RNA Pol II occupancy, demonstrating that REV-ERBα orchestrates an epigenomic rhythm of histone deacetylation and gene repression through NCOR and HDAC3.

A similar genome-wide study that focused on BMAL1/CLOCK and CRY/PER uncovered 3 distinct transcriptional phases of the forward limb: a poised state, a transcriptional activation state, and a repressed state (158). During the poised state, BMAL1/CLOCK and CRY1 are cobound without active transcription. The transcriptional activation state begins with declining CRY1 level, which is likely caused by decreased production or degradation mediated by the E3 ubiquitin ligase F-box/LRR-repeat protein 3 (FBXL3) (159–162). Subsequently, BMAL1/CLOCK occupancy rises, which recruits p300 and CBP and is accompanied by active marks H3K4me1 and H3K9Ac. Next, RNA Pol II is loaded into promoters to transcribe genes, which is accompanied by other active marks H3K4me3 and H3K27Ac. The active phase transitions to the repressive phase when PER1/2 and CRY2 co-occupy these sites to repress transcription, which is followed by delayed CRY1 binding that resets the cycle. Another study that measured circadian changes in histone modifications reported a related finding that oscillating levels of H3K4me3 at promoters and H3K27Ac at enhancers are correlated with rhythmic expression of nearby genes genome-wide (78).

The same study that identified the distinct transcriptional phases of the clock also examined posttranscriptional regulation of circadian genes (158). Using whole transcriptomic RNA-seq, they identified 2 groups of genes, intron cycling versus exon cycling, that show circadian expression at the nascent transcription versus steady-state mRNA levels, respectively. Surprisingly, the vast majority of exon cycling genes do not have underlying transcriptional rhythms, leaving only 22% of exon cycling genes that are driven by de novo nascent transcription. These genes are enriched for the core clock genes and their known target genes with high rhythmic amplitudes. In a separate study, 15% of all detected genes are found to be transcribed rhythmically, 42% of which maintain comparable rhythms at the steady-state mRNA level (163). On the other hand, 22% of all detected genes show circadian rhythms at the steady-state mRNA level, but 70% of these genes do not exhibit underlying transcriptional rhythms. These studies demonstrated that nascent transcript rhythms do not necessarily correlate with steady-state mRNA rhythms and vice versa, revealing pervasive posttranscriptional regulation that likely plays a major role in shaping the rhythmic mRNA landscape (Fig. 3A).

Figure 3.

Epigenomics and transcriptomics of the molecular clock. (A) A majority of rhythmic mRNA transcripts do not exhibit underlying transcriptional rhythms. (B) Enhancers with circadian activity based on rhythmic enhancer RNA (eRNA) transcription show phase-specific enrichment of DNA binding motifs bound by both known and unknown clock transcription factors.

In another study, Global Run-On sequencing (GRO-seq) was used to not only measure nascent transcription genome-wide, but also to identify cis-regulatory elements that are marked by enhancer RNAs (eRNA), short noncoding and often bidirectional RNA species transcribed at active enhancers (60) (Fig. 3B). eRNA has been widely used as a robust marker of enhancer activity because its transcription level highly correlates with target gene expression (164). Although the functional significance of eRNA is debated, inhibition of eRNA synthesis by REV-ERBs has been shown to represses gene transcription in macrophages (165). Unbiased examination of eRNA expression in mouse livers collected over the course of 24 hours identified >5000 enhancers with circadian activity, as well as nearby target genes with oscillating expression in the same phases. In addition, de novo motif analyses of these enhancers revealed enrichment of DNA motifs, such as E-Box, RORE/RevDR2, and D-Box, in phases that correspond to maximal enhancer activity coordinated by the core clock TFs that recognize these motifs. Interestingly, E-twenty six (ETS) motifs were enriched in circadian enhancers with peak activity at zeitgeber time 0 to 3, implying a potential role of ETS TFs in regulating this phase of circadian transcription. Furthermore, correlation of eRNA expression levels with REV-ERBα activity enabled identification of a small fraction of its cistrome where REV-ERBα functionally represses transcription, which orthogonally confirmed the previous finding that NFIL3 is a direct target of REV-ERB (166, 167).

Despite such significant scientific strides made by the recent genome-wide studies, fundamental questions remain with regard to how clock TFs coordinate epigenomic rhythms in a tissue-specific manner. Although most core clock TFs are similarly expressed in nearly all cells except for embryonic stem cells that appear to lack a functional clock, it is unclear how clock TFs recognize and occupy different cistromes in different cell types (168, 169). One possible mechanism by which this is achieved is through the tethering of REV-ERB to tissue-specific, lineage-determining TFs in a DBD-independent manner, although it is unknown whether other clock TFs share a similar mechanism (149, 153). It is also possible that clock TFs can only bind enhancers that have been made accessible or “pioneered” by tissue-specific TFs that bind closed chromatin to loosen local nucleosome density. Interestingly, BMAL1/CLOCK have been reported to have a pioneering activity (170, 171). Another possibility is that genomic binding of clock TFs is determined by collaborative interactions with other TFs, though it is unclear what combinations of TFs are required to stably maintain enhancer activity (172).

While these genome-wide studies re-demonstrated the key importance of enhancer action in organizing epigenomic and transcriptional waves across the genome, they also unveiled the widespread prevalence of posttranscriptional regulation in setting transcriptomic rhythms. The finding that most rhythmic mRNA transcripts do not have underlying transcriptional rhythms alludes to possible circadian mechanisms that control mRNA stability and degradation. One regulator of this process is short noncoding microRNA, some of which show circadian expression and regulate the stability of their target transcripts, including several clock genes (173–177). Other aspects of mRNA processing, such as polyadenylation, can also regulate circadian accumulation and degradation of transcripts (178–180). Moreover, genome-wide ribosome profiling and proteomic approaches have uncovered new regulatory layers of the transcription-translational feedback loop that extend beyond transcription (181-184). Therefore, although this review focuses on transcriptional mechanisms, it is important to recognize that posttranscriptional and posttranslational mechanisms also help generate circadian rhythms, both in conjunction with and independently of rhythmic transcription.

Circadian gene expression in time and space

The observation that chromosomes occupy certain territories in the nucleus led to the first recognition that 3-dimensional (3D) organization of the genome is nonrandom (185–187). A century after this rudimentary microscopic finding was made, our understanding of 3D genome architecture has become much more sophisticated with development of cutting-edge chromatin conformation capture techniques that provided a magnified view into how the genome folds in 3D space to control gene transcription and genome function (188–191).

One major breakthrough was made by 3 independent laboratories that demonstrated that the genome is partitioned into highly conserved units called topologically associating domains (TAD) (192–194). TAD is on average 1 megabase pairs long and conserved in different tissues, as well as in certain syntenic regions shared between species (192). TAD boundaries are anchored together by a pair of CCCTC-binding factor (CTCF) that homodimerize in a convergent orientation to form a long-range chromatin loop, which is further stabilized by a ring-like protein complex called Cohesin (192, 195, 196). By spatially restraining a linear genomic region into a physically isolated loop, TAD insulates enhancers and genes to promote their interactions within the same TAD. As such, genetic disruption of TAD boundaries results in loss of insulation, leading to ectopic enhancer-promoter rewiring between neighboring TADs and aberrant gene expression (197, 198). TAD also demarcates regions of euchromatin and heterochromatin (190, 194, 195). TADs containing either euchromatin or heterochromatin segregate together in distinct higher order structures called compartment A or B, respectively. Compartment B is also enriched with lamin-associated domains, which contain repressed genes and associated with repressive histone mark H3K27me3 (199).

Although these studies provided an unprecedented insight into new mechanisms of gene regulation at the level of TAD and compartments, how enhancer and promoter (E-P) loops are formed and regulated continues to be an active field of investigation (200–205). In general, activator TF are thought to bind enhancers and recruit coactivators, such as the Mediator complex, which then interact with the preinitiation complex at promoters, thereby bringing enhancers to promoters to activate transcription (206, 207). Cohesin has been shown to be recruited to enhancers, particularly a long cluster of enhancers called super enhancers, but other studies also showed that not all enhancers are bound by Cohesin, thus leaving the exact role of Cohesin in regulation of E-P loops unclear (208-210).

Consistent with the growing evidence that 3D chromatin architecture controls genome function, several studies have linked circadian transcription to spatiotemporal reorganization of the genome. For instance, the DBP promoter has been shown to undergo circadian interchromosomal interactions in a BMAL1-dependent manner, although the functional significance of trans-interactions in gene regulation remains controversial (211) (Fig. 4A). Depletion of Cohesin in cultured cells alters the local long-range topology of circadian genes and causes aberrant expression of several circadian genes (212). Nuclear positioning of genes has also been shown to impact transcription. Circadian interactions between CTCF and poly ADP-ribose polymerase 1 (PARP1) rhythmically bring CTCF-mediated interchromosomal hubs to the nuclear periphery containing lamin-associated domains, where genes acquire repressive H3K9me2 marks and become transcriptionally repressed (213) (Fig. 4B).

Figure 4.

Circadian regulation of 3-dimensional chromatin architecture. (A) The DBP promoter undergoes circadian interchromosomal interactions in a BMAL1-dependent manner. (B) Chromatin hubs organized by CCCTC-binding factors (CTCFs) are repositioned to the repressive nuclear laminar environment through circadian interaction with poly ADP-ribose polymerase-1 (PARP-1). (C) Enhancer-promoter loops undergo circadian reorganization. REV-ERB opposes functional enhancer-promoter formation to repress transcription.

In addition, comparison of 3D genome architecture at 2 opposite circadian phases revealed that although TAD boundaries are stable, interactions within TADs that contain circadian genes are dynamic and positively correlated with gene expression (214). One major type of such interactions is E-P loops, which are controlled by coordinated actions of TFs. In support of this, an intronic enhancer at the CRY1 locus has been shown to rhythmically loop to the promoter to drive circadian expression of CRY1 (69, 214) (Fig. 4C). Interestingly, formation of this E-P loop is opposed by REV-ERBα binding at the enhancer. Moreover, integrative analysis of REV-ERBα cistrome, nascent transcriptome, and 3D genome architecture data identified 2 groups of E-P loops, engaged versus passive, where REV-ERBα engages in transcriptional repression or passively binds without affecting transcription, respectively. At engaged sites, E-P loops are attenuated upon REV-ERBα binding, whereas those at passive sites are unaffected. Furthermore, engaged sites show stronger recruitment of NCOR1/HDAC3 and concordant histone deacetylation, which leads to eviction of the histone acetylation reader BRD4 and its interacting partner MED1, a component of the Mediator complex known to promote looping between enhancers and promoters (215–219) (Fig. 4C). This study revealed genome-wide 3D organizational plasticity that occurs in normal mammalian physiology and also demonstrated a novel role of transcriptional repressors in opposing E-P loops as a mechanism to repress gene transcription. To complement this finding, another study showed that BMAL1 rhythmically promotes chromatin interactions between enhancers that regulate the same genes, illustrating that dynamic chromatin loops underlie circadian gene transcription (220).

The governing principles of 3D genome organization unveiled a new regulatory layer of gene regulation, which generated new exciting questions and helped us revisit the unresolved from a renewed perspective. For instance, it remains unclear how enhancers specifically loop to their distal target promoters as opposed to interacting with the nearest promoters within the same TAD. Future studies that systemically compare molecular and topological characteristics of E-P loops may help identify emergent properties of these loops as well as regulatory logics behind enhancer-promoter pairing.

In addition to genome folding, mechanisms of nuclear positioning will improve our understanding of the spatiotemporal regulation of transcription. Interestingly, HDAC3 has been shown to interact with the nuclear envelope protein and transcriptional repressor LAP2β to repress transcription by tethering chromatin to the repressive nuclear lamina in a deacetylase-independent manner (221–223). Therefore, it will be important to examine whether circadian genes are regulated via this mechanism, and also whether clock TFs and their associated coregulators participate in this process.

Another important question to consider is why only a small subset binding sites are functional for any given TF. The observation that the repressive action of REV-ERBα is permitted at engaged sites but somehow buffered against at passive sites is perplexing (224). Although there is no definitive answer to this conundrum that has been observed with other TFs in numerous ChIP-seq studies, the composition of local chromatin landscapes at enhancers may play a role in determining transcriptional permissiveness.

Finally, there is an emerging appreciation that chromatin is biophysically compartmentalized into membraneless liquid states by phase separation (225). Several studies have reported that many TFs and coregulators contain a structural domain called intrinsically disordered regions that enable the formation of phase-separated compartments (226–229). Several groups have further extrapolated this finding to propose that phase separation drives and stabilizes the formation of E-P loops to activate transcription (230). Intriguingly, BRD4 and the Mediator complex have been implicated as critical components of phase-separated transcriptional compartments. It will thus be informative to examine how clock TFs and their associated coregulators control phase separation, which may have overarching implications in general mechanisms of transcriptional regulation.

Circadian Mammalian Physiology

Hierarchical structure of the organismal clock

Self-sustaining cellular rhythms generated by complex transcriptional processes give rise to organismal rhythms. In the mammalian system, organismal rhythms are hierarchically compartmentalized into the central clock of the brain and the peripheral clocks of various organs. The central clock is largely controlled by the suprachiasmatic nucleus (SCN) of the hypothalamus, which is entrained by light detected by melanopsin-expressing retinal cells of the retinohypothalamic track (231, 232). The SCN projects to various output pathways in the brain and to peripheral organs via neuronal and humoral signals (233–236). Interestingly, a recent study demonstrated that the astrocytic clock in the SCN is necessary for rhythm generation, revealing an unexpected role of glial cells in supporting the central clock (237).

By coupling the circadian neurosensory circuit with the peripheral clocks, the SCN ensures proper timing of daily physiological processes and behavioral patterns, such as sleep-wake and fasting-feeding cycles. Indeed, mice lacking clock genes often exhibit damped circadian variations of food intake (238–242). The SCN also controls the synthesis and secretion of a number of hormones to entrain organismal rhythms. Diurnal secretion of glucocorticoids facilitates synchronization of the liver and kidney clocks with the central clock, and that timed administration of exogenous glucocorticoids in SCN-lesioned mice can restore 60% of circadian transcriptome of the liver (243, 244). In diurnal mammals such as humans, melatonin is produced from the pineal gland at night to help initiate sleep, among its many functions (245). Despite the wide use of melatonin as a supplemental sleep aid, its role in circadian rhythms and sleep has been studied only in limited contexts. This is largely because certain inbred rodent species do not produce melatonin, which poses a major experimental limitation and also restricts the generalizability of its function across mammalian species (246).

Through the hypothalamic control of behavior and endocrine hormones that synchronize the peripheral clocks, the central clock serves as the master clock that entrains organismal rhythms with the environmental cycle. However, although this whole-body entrainment is crucial for setting organismal rhythms, every tissue and organ can also be entrained by signals related to their functions (247–249). For instance, the liver clock is predominantly entrained by feeding and nutrient availability, such that feeding restricted to only certain hours of the day can rapidly shift the liver clock independently of the SCN clock that remains locked to the prevailing light-dark cycle (250, 251). The ability of the peripheral clocks to override entrainment by the central clock confers functional plasticity to peripheral organs so that they can rapidly adapt to stimuli and mount appropriate physiological responses at any time of the day.

The peripheral clocks also influence the central clock through hormones and metabolic signals. A high-fat diet (HFD) lengthens free-running periods and attenuates feeding and locomotor rhythms in mice (252). In addition, a variety of hormones that are tightly regulated by fasting and feeding, such as fibroblast growth factor 21 (FGF21), ghrelin, leptin, insulin, and glucagon-like peptide 1, can signal to the central clock and other food-entrainable clocks in the brain [reviewed in (253)]. For instance, FGF21 directly acts on the SCN to suppress physical activity and alter circadian behavior during the times of starvation (254). Such bidirectional crosstalk allows the central clock to entrain the peripheral clocks while simultaneously enabling the peripheral clocks to shift their own phases and feedback to the central clock to optimally meet tissue-specific demands (Fig. 5).

Figure 5.

Hierarchical structure of the organismal clock. The suprachiasmatic nucleus (SCN) of the hypothalamus serves as the central clock that entrains the organismal clock by synchronizing the peripheral clocks with the environment. The peripheral organs also have tissue-specific cues related to their functions that can entrain their own clocks independently of the central clock. Reciprocally, the peripheral clocks can feedback to the central clock to further fine tune the organismal clock.

Circadian regulation of metabolism

Metabolic rhythms arise within the framework of feeding and neuroendocrine rhythms coordinated by the central and peripheral clocks. Metabolism is regulated at multiple levels of organismal functions, ranging from cellular anabolism and catabolism of nutrients to systemic energy balance. In this review, we will specifically focus on the transcriptional control of metabolism, with an emphasis on how the clock transcriptional network generates metabolic output pathways, as well as reciprocal ways in which metabolic processes reprogram the clock. In support of the notion that metabolism is under circadian control, the vast majority of circadian transcriptome of the liver has been shown to belong to metabolic pathways (255). Moreover, many of cycling genes control key rate-limiting steps in their respective pathways, such as ALAS1 for heme biosynthesis and HMGCL for ketogenesis.

Mouse genetic studies provided early evidence that the core clock TFs play a critical role in circadian metabolism. Mice homozygous for a dominant-negative mutation in the CLOCK gene are hyperphagic and obese and exhibit profound metabolic derangements, including hyperleptinemia, hyperlipidemia, hepatic steatosis, hyperglycemia, and hypoinsulinemia (242, 256). Both CLOCK mutant mice and BMAL1 knockout mice exhibit loss of diurnal variations in triglyceride and glucose levels, reduced insulin secretion resulting from impaired gluconeogenesis and defective pancreatic islets development (257, 258). Further corroborating this finding, conditional ablation of BMAL1 in pancreatic β cells also causes hyperglycemia because of disrupted nutrient-responsive insulin secretion (258, 259). In addition, liver-specific deletion of BMAL1 causes hepatosteatosis and suppression of de novo lipogenesis (DNL) and fatty acid oxidation (260). In support of the fact that the positive regulators of the forward limb share overlapping transcriptional activity, NPAS2 knockout mice also show similar dysregulation of hepatic lipid metabolism genes (261). However, these knockout mice also display distinct physiological and metabolic phenotypes, implying TF-specific roles in organ function and development.

Consistent with cistromic and transcriptomic studies demonstrating that REV-ERB directly controls genes involved in lipid metabolism, mice lacking REV-ERBα/β have been shown to develop hepatosteatosis, along with elevated serum glucose and triglycerides, and reduced serum fatty acids (39, 41, 60, 148). Interestingly, another study reported a contrary finding that liver triglyceride levels are reduced in REV-ERBα knockout mice (262). This discrepancy could be due to different gene deletion strategies used in these studies, strain-specific genetic variations, or animal husbandry conditions affecting the microbiome, among other possibilities. Regulation of hepatic lipid metabolism by REV-ERB is in part mediated through the tethering mechanism, in which REV-ERB binds lineage-determining TFs, such as hepatocyte nuclear factor 6, and recruits the NCOR/HDAC3 complex to repress metabolic genes (149, 153). Indeed, mice with liver-specific deletion of NCOR or HDAC3 also develop hepatosteatosis, albeit to a greater extent than seen in REV-ERB knockout mice, which could be explained by the fact that NCOR and HDAC3 regulate other NRs in the liver, including hepatocyte nuclear factor 4α (98, 263–265).

Another means by which REV-ERB controls hepatic lipid metabolism is through its regulation of the NR peroxisome proliferator-activated receptor δ (PPARδ) (60, 214). At the trough of REV-ERB level, PPARδ drives the DNL biosynthetic pathway that produces the lipid species phosphatidylcholine 18:0/18:1 (266). This metabolite is secreted into the bloodstream and taken up by the muscle, where it serves as a ligand for another NR, PPARα, to promote fatty acid oxidation, thus simultaneously driving anabolism and catabolism in the 2 distant metabolic organs. Although ROR has also been implicated in hepatic lipid metabolism, studies have yielded conflicting results regarding the effects of ROR on hepatic triglyceride levels and lipogenic gene expression, which were speculated to be due to different experimental conditions, such as timing of feeding (267–271).

Although many of these studies focused on the role of clock TFs in major metabolic tissues, such as the liver and the pancreas, all metabolic organs act in concert to coordinate rhythmic metabolism and physiology. The diverse biology of REV-ERBα best illustrates how a single clock TF can orchestrate multiple metabolic processes in distant organs. In thermogenic brown adipose tissue, REV-ERBα directly represses the expression of uncoupling protein 1 to generate a circadian rhythm of body temperature (272, 273). In white adipose tissues, REV-ERBα represses the expression of β-KLOTHO, an essential coreceptor for FGF21, to modulate responsiveness to FGF21 signaling important for carbohydrate and lipid metabolism (274). In oxidative muscle tissues, REV-ERBα improves exercise capacity by promoting mitochondrial biogenesis through rhythmic activation of the PGC1α pathway (275). Beyond these organs, REV-ERBα also influences the gut microbiome through circadian regulation of Toll-like receptors in the intestinal epithelial cells, which have indirect effects on the intestinal clock, corticosterone synthesis, and systemic metabolism (276).

These in vivo studies demonstrate bona fide physiological significance of the clock in metabolic regulation. However, caution should be taken in interpreting metabolic studies involving whole-body knockout mice. For instance, whole-body knockout mice may exhibit arrhythmic behavior causing dyssynchrony between the central and peripheral clocks that may be responsible for the observed metabolic phenotypes. Tissue-specific and conditional knockout mice serve a particularly useful purpose in this regard because they provide an amendable experimental tool that can help dissociate primary metabolic derangements from confounding behavioral effects. Another challenge in interpreting metabolic studies is that different primary metabolic changes may ultimately converge into similar metabolic outcomes, which may occur through different combinations of disrupted anabolic and catabolic pathways or systemic decompensation. For example, despite the finding that CRY opposes the transcriptional activity of BMAL1 and CLOCK, liver-specific deletion of CRY1/2 also results in hyperglycemia, as seen in BMAL1 knockout and CLOCK mutant mice (277). Future in vivo studies could combine unbiased cistromic and transcriptomic analyses with quantitative metabolic tracing in an integrative approach to identify and link transcriptional changes to tissue-specific metabolic alterations (278–281).

Metabolic regulation of the clock

Mirroring the key role of the clock in metabolic homeostasis, numerous metabolic pathways have been shown to have reciprocal feedback mechanisms to reprogram the clock. For example, clock TFs can directly interact with signaling factors of metabolic pathways. AMP-activated protein kinase (AMPK) is a nutrient-responsive modulator of metabolic signals (282, 283). AMPK is activated by phosphorylation from upstream signal factors during nutrient restriction that produces high AMP to ATP ratios (284). Activated AMPK phosphorylates a number of target proteins, including CRY1 in the liver. Phosphorylation of CRY1 causes protein destabilization that leads to rapid degradation and a phase shift of the liver clock (285) (Fig. 6). AMPK also activates casein kinase I via phosphorylation, which then phosphorylates PER2 for degradation to rewire the peripheral clock (286).

Figure 6.

Metabolic regulation of the clock. Activated AMP-activated protein kinase (AMPK) phosphorylates CRY1 to induce its degradation, which results in reprogramming of the liver clock. Many metabolic pathways such as heme biosynthesis and NAD⁺ production, are regulated in a circadian manner at the transcriptional level. NAD⁺ regulates activity of sirtuin 1 (SIRT1) that deacetylates clock TFs and their associated factors and histones. Heme serves as an endogenous ligand that potentiates REV-ERB activity. The hexosamine biosynthesis pathway regulates O-GlcNAc transferase (OGT) activity, which controls the stability and transcriptional function of clock TFs.

The intracellular NAD⁺/NADH redox state also has a broad impact on the clock. Reduced forms of the redox cofactors, NAD(H) and NADP(H), can directly bind BMAL1/CLOCK and NPAS2 heterodimers to enhance DNA binding. In contrast, their oxidized forms, NAD⁺ and NADP⁺, counteract this effect (287). In addition, NAD⁺ can serve as a cofactor for sirtuin 1 (SIRT1), which is an HDAC that has been shown to rhythmically bind CLOCK-BMAL1 and deacetylate PER2 and BMAL1 (288, 289) (Fig. 6). Deacetylation of PER2 promotes its degradation, whereas deacetylation of BMAL1 reverses CLOCK-mediated acetylation that promotes interaction with CRY1 (289, 290). SIRT1 can also deacetylate MLL1 bound to BMAL1/CLOCK, which results in suppression of MLL1 catalytic activity and transcriptional activity (291) (Fig. 6). Interestingly, BMAL1/CLOCK directly controls circadian expression of nicotinamide phosphoribosyltransferase (NAMPT), a rate-limiting enzyme in the NAD⁺ biosynthetic pathway (292, 293). This generates an oscillating level of NAD⁺ that drives circadian SIRT1 activity, thereby establishing a metabolic feedback loop tied to the clock transcriptional network. Furthermore, AMPK increases the intracellular NAD⁺/NADH ratio during fasting, which activates SIRT1 to deacetylate PGC1α and turn on its transcriptional activity, suggesting a possible convergence of the two metabolic pathways in regulation of the clock (294, 295).

The intracellular glucose state can affect clock TF activity via addition of O-linked β-N-acetylglucosamine (O-GlcNAcylation), a nutrient-driven posttranslational modification regulated by the hexosamine biosynthesis pathway (296). PER reversibly undergoes O-GlcNAcylation catalyzed by O-GlcNAc transferase at serine residues, which competitively prevents phosphorylation and potentiates its repressive activity (297) (Fig. 6). Moreover, BMAL1/CLOCK also rhythmically undergo O-GlcNAcylation, which augments their transcriptional activity by increasing protein stability through competitive inhibition of ubiquitin-mediated degradation (298) (Fig. 6).

Other metabolites also have direct effects on the clock. Many small lipophilic metabolites induce transcriptional changes by serving as ligands for NRs, which regulate metabolic pathways related to their respective ligands. Heme is a known ligand for REV-ERB that potentiates its repressive activity (299) (Fig. 6). The intracellular redox state has been shown to regulate the interaction between REV-ERB and heme by multiple mechanisms. Reduction of a thiol-disulfide redox switch in the LBD of REV-ERBβ to a dithiol state increases heme binding in the LBD, thus linking oxidative stress to circadian rhythms and metabolic imbalance (300). Similarly, the REV-ERBβ-heme complex can be regulated by nitric oxide and carbon monoxide gases, which induce at least 6 distinct heme-binding conformational states, though it remains unclear whether these states modulate REV-ERB transcriptional activity in vivo (301).

NPAS2 has also been shown to bind heme via its 2 PAS domains, where heme functions as a redox sensor. Carbon monoxide can bind to this redox sensor to hinder heterodimerization with BMAL1 and reduce DNA binding in vitro (302). Interestingly, heme metabolism has been shown to be regulated by both REV-ERB and NPAS2. REV-ERB inhibits heme biosynthesis by repressing the expression of PGC1α, a potent inducer of ALAS1 expression, whereas BMAL1/NPAS2 promote heme biosynthesis by driving circadian expression of ALAS1 (303, 304). Interestingly, heme is synthesized from glycine and succinyl-CoA, raising the possibility that amino acid biosynthesis and the citric acid cycle may indirectly influence the clock through heme and its modulatory effects on these clock TFs.

Although much less is known about ligand-mediated regulation of ROR, both ROR and REV-ERB have been shown to regulate steroid and bile acid metabolism in the liver, which may produce endogenous ligands that can form a transcriptional feedback loop with ROR (48, 167, 262, 270, 305). In support of this speculation, many NRs show circadian expression across multiple metabolic tissues, suggesting that NRs and their ligands could broadly link rhythmic metabolism to the clock, and vice versa (44).

Although these studies highlight the intricate reciprocal relationship between metabolism and the clock, metabolic processes can also generate transcriptional rhythms that are functionally uncoupled from the clock. A combined transcriptomic and metabolomic study revealed HFD causes both phase shifts and abolition of metabolite and transcriptional rhythms in mice (306). Interestingly, HFD also induces new oscillations of metabolites and transcripts in the liver. Motif analysis of promoter regions of the genes that gained new rhythms revealed enrichment of DNA sequences recognized by 2 major regulators of lipid metabolism, PPARγ and sterol regulatory element-binding protein (SREBP). Consistent with this analysis, the study found that circadian expression of PPARγ was induced by HFD.

Another study revisited the same question with an unbiased approach using GRO-seq to identify HFD-specific circadian enhancers based on eRNA transcription (307). Motif analysis confirmed that DNA binding motifs for PPARγ and SREBP are indeed enriched in circadian enhancers induced by HFD. However, liver-specific deletion of PPARγ failed to abrogate activity of these enhancers, which led to the identification of PPARα as the culprit factor with the same DNA binding motif that orchestrated the HFD-induced transcriptional rhythm. Furthermore, a genetic loss of function experiment showed that SREBP-dependent DNL drives rhythmic PPARα activity likely through circadian production of endogenous PPARα ligands, unraveling a new mechanism by which metabolic processes generate transcriptional rhythms without redirecting the core clock TFs.

Although these examples focus how metabolites and energetic states are integrated through regulation of TF activity, there are also indirect ways in which metabolism can influence circadian transcription. For instance, various metabolites serve as substrates for histone modifications catalyzed by metabolic enzymes [reviewed in (308)]. Global histone acetylation levels are under the control of glucose availability and the intracellular acetyl-CoA level (309, 310). Despite the growing evidence that many metabolite-derived modifications are present on chromatin and facilitate gene regulation, it remains unknown whether these modifications regulate specific metabolic or clock genes in a transcriptional feedback loop to control their respective metabolic pathways. In the future, it will be exciting to explore how metabolic processes affect other aspects of gene regulation, some of which were discussed in this review. For instance, PARP1 activity is highly dependent on NAD⁺. It is possible that nuclear positioning mediated by PARP1 and CTCF is controlled by the intracellular redox state. PARP1 has also been shown to participate in phase separation, suggesting possible redox-dependent regulation of E-P loops by PARP1 (311–314). Such studies will not only yield a new biological insight into the complex interplay between metabolism and circadian rhythms, but also open up novel therapeutic opportunities for treating metabolic and circadian disorders.

Implications of circadian rhythms in human health

The recognition that circadian rhythms play a central role in human physiology is founded upon numerous studies demonstrating that both genetic and environmental perturbations of the clock result in human pathologies. In particular, human genetics studies have significantly advanced our understanding of circadian rhythms and sleep. These seminal studies not only identified causal mutations in clock genes underlying different chronotypes, such as familial advanced sleep-phase syndrome and delayed sleep phase disorder, but they also illuminated fundamental genetic and biochemical mechanisms of the mammalian clock (23, 315–323).

Epidemiological studies linked environmental circadian disruptions to overall poor health outcomes. The Nurses’ Health Study, a large prospective cohort study of the long-term effect of lifestyle and environmental factors, demonstrated association between night shift work and an increased risk of type 2 diabetes, a finding that has been consistently reproduced by other studies (324–331). Modern lifestyle is a common cause of circadian misalignment because of altering bedtime on weekends and frequent exposure to disruptive blue light from electronics at night. This condition, commonly referred to as social jetlag, has been associated with detrimental effects on sleep and metabolism (332–334). To mitigate such adverse effects, dynamic lighting that simulates circadian variations of light has been implemented in nursing homes and intensive care units as a preventive strategy to improve sleep parameters and well-being, though its benefits have been limited (335–337). In parallel with these epidemiological observations, laboratory studies demonstrated a causal role of circadian misalignment in metabolic dysregulation. Healthy individuals who were subjected to controlled circadian disruption exhibited signs of prediabetes, characterized by poor glucose homeostasis, attenuated insulin function, and an overall decrease in energy expenditure (338–340). A genome-wide association study found a genetic link between metabolic syndrome and single nucleotide polymorphisms in the clock genes NPAS2 and PER2 (341). In addition, a single nucleotide polymorphism in the human melatonin receptor B1 gene (MTNR1B) has been associated with type 2 diabetes and obesity, suggesting complex genetic interactions among the clock, sleep, and metabolism (245, 342–344).

The growing appreciation of circadian rhythms in human health spearheaded clinical efforts to translate circadian biology into “chronotherapies,” novel therapeutic interventions that target the clock or leverage circadian physiology to improve pharmacodynamics and efficacy of existing medications (Fig. 7). The field of cardiovascular medicine in particular has longstanding interests in implementing circadian medicine. The observation that myocardial infarction (MI) occurs more frequently in the morning provided early evidence that circadian rhythms are important for cardiovascular health, which is further corroborated by a related finding that the incidence of MI increases around shifts to and from daylight savings time (345, 346). The higher incidence of MI in the morning is thought to be caused by rhythmic expression of the major fibrolytic inhibitor plasminogen activator inhibitor type 1, whose activity generates a nadir in net fibrinolysis in the early morning that coincides with the period of elevated MI risk (347, 348). A recent study demonstrated that, like plasminogen activator inhibitor type 1, many targets of common cardiovascular medications, such as beta blockers, calcium channel blockers, and nonsteroidal anti-inflammatory drugs, are expressed in a circadian manner along with other proteins involved in drug transport and metabolism, supporting the notion that timing is an important determinant of therapeutic efficacy (349). Indeed, small randomized trials suggested that antihypertensive medications, such as angiotensin-converting enzyme inhibitors and angiotensin-receptor blockers, are more effective when administered in the evening, and this time-dependent effect was observed with other pharmacologic agents as well (350–354).

Figure 7.

Promise of circadian medicine. Understanding circadian physiology will enable development of novel therapeutic approaches that will help improve human health in the future.

Timing of operation has also been shown to impact perioperative surgical risks. A single-centered randomized study demonstrated that patients who underwent aortic valve replacement in the afternoon were less likely to experience major adverse cardiac events compared with the morning group (355). Ex vivo analysis of human myocardial biopsy samples and animal experiments implicated that high REV-ERBα expression is associated with an elevated risk for hypoxia–reoxygenation injury, likely resulting from REV-ERBα-mediated repression of the ischemia–reperfusion injury modulator CDKN1α/p21. Moreover, inhibition of REV-ERBα, either by genetic deletion or pharmacologic intervention with a putative REV-ERBα antagonist SR8278, ameliorated the hypoxia–reoxygenation injury noted at the peak of REV-ERBα expression, although this observation should be interpreted with caution in the light of a recent study suggesting that REV-ERBα modulators may have nonspecific effects (356).

Last, intermittent fasting has recently garnered significant clinical interest for its broad benefits in metabolic health (357, 358). One form of intermittent fasting is time-restricted feeding (TRF), which refers to limitation of daily food intake to a consistent interval of less than 12 hours without reduction in caloric intake. Several studies in mice demonstrated that TRF provides protection against a hypernutritive challenge and reverses metabolic imbalances, resulting in improved glycemic control, decreased hepatosteatosis, improved serum lipid profiles, and attenuated inflammatory responses (359–361). In support of these rodent studies, small human cohort studies reported similar trends (362, 363). These beneficial effects are presumed to be in part mediated by harmonization of circadian rhythms and feeding patterns that favorably optimizes energy utilization and storage (364). However, another study suggested that rhythmic food intake alone can significantly remodel rhythmic transcriptome of the mouse liver without affecting the core clock genes (365). Similarly, mice lacking the liver clock still show improved metabolic parameters on TRF, suggesting that the liver clock may be dispensable for the therapeutic benefits of TRF (366). Therefore, future studies are required to not only gain a mechanistic insight into how TRF improves metabolic health, but also to rigorously investigate its promising benefits in humans in large controlled cohorts.

Concluding Remarks

In this review, we have explored various transcriptional and epigenomic mechanisms by which the core clock TFs orchestrate circadian gene expression. We revisited the complex biology of epigenetic regulation that challenged our previous thinking, and discussed the emerging role of 3D chromatin architecture in genome function. We highlighted recent studies that examined transcriptional regulation in circadian time and genome space, which have not only advanced our understanding of the transcriptional architecture of the clock, but also generated new exciting questions and hypotheses aforementioned. In addition, we focused on how the clock transcriptionally regulates metabolism through reciprocal crosstalk mechanisms that allow the clock to serve a dual role as a driver and an integrator of metabolism.

Implications of circadian biology in mammalian physiology informed by these animal studies were further advanced by human genetics, epidemiological, and clinical studies that linked circadian misalignment to various human diseases, including metabolic syndrome and cardiovascular diseases. Indeed, time is certainly ripe for circadian medicine that will begin to translate our accumulated knowledge into novel therapeutic opportunities. However, equally imperative are rigorous clinical investigations in humans and parallel mechanistic studies in animals that will help discover the molecular underpinning of circadian pathophysiology at the systems level spanning from transcriptional rhythms to circadian behavior, which will unravel more dimensions of the molecular timekeeper that has existed within us as an integral component of life on Earth.

Glossary

Abbreviations

3D

3-dimensional

AMPK

AMP-activated protein kinase

BMAL

brain and muscle ARNT-like 1

CBP

CREB-binding protein

ChIP

chromatin immunoprecipitation

CHRONO

computationally highlighted or chromatin immunoprecipitation-derived repressor of the network oscillator

CLOCK

circadian locomotor output cycles kaput

CRY1/2

cryptochrome 1/2

CTCF

CCCTC-binding factor

DAD

deacetylase activating domain

DBD

DNA-binding domain

DBP

D-binding protein

DNL

de novo lipogenesis

E-P

enhancer and promoter

eRNA

enhancer RNA

FBXL

F-box/LRR-repeat protein

FGF

fibroblast growth factor

HAT

histone acetyl transferase

HDAC

histone deacetylase

HFD

high-fat diet

HMT

histone methyltransferase

HP1

heterochromatin protein 1

LBD

ligand-binding domain

LSD

lysine-specific demethylase

MLL

mixed lineage leukemia

NCOA

nuclear receptor coactivator

NCOR

nuclear receptor corepressor

NFIL

nuclear factor, interleukin

NPAS2

neuronal PAS domain protein 2

NR

nuclear receptor

MI

myocardial infarction

PARP

poly ADP-ribose polymerase

PPAR

peroxisome proliferator-activated receptor

REV-ERB

reverse strand of c-erbα

ROR

RAR-related orphan receptor

SCN

suprachiasmatic nucleus

SIRT

sirtuin

SREBP

sterol regulatory element-binding protein

TAD

topologically associating domain

TF

transcription factor

TRF

time-restricted feeding

Additional Information

Disclosure Summary: M.A.L. is consultant to Pfizer, Novartis, Madrigal Pharmaceuticals, and Calico, and receives research support from Pfizer.

Data Availability: Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.