Genetics of Circadian Rhythms

Introduction: The Many Levels of Circadian Genetics

The topic, “Genetics of Circadian Rhythms” could be a consideration of a number of questions. Are differences between individuals in their circadian rhythmicity a result of differences in their genes? If so, what genes? What genes encode the “clock” mechanism? How can genes form a clock? Does the circadian clock mechanism affect gene activity? If so, how?

Here, we describe the developing appreciation of the multiple interconnections of circadian rhythmicity with genetics in mammals. As illustrated in Figure 1, a core set of genes function as a clock in every cell as a feedback loop of gene transcription and translation-- providing a means for circadian rhythms to persist independently in each cell. However, the clocks within the cell can be influenced by a variety of interacting pathways and influences ranging from the intracellular (e.g., redox state) to nutritional, endocrine and behavioral. Similarly, a variety of genetic polymorphisms in other genes appear to influence the expression of circadian rhythms in individuals. The genetic circadian clock regulates rhythms of gene expression via direct transcriptional regulation, via histone acetylation/deacetylation. Despite the many genes and mechanisms, such results point to a very central role of circadian clock genes in regulating biochemical, metabolic and physiological processes at many different levels of organization.

Figure 1. Multidimensional genetic clockwork.

The core clock of animals has the proteins CLOCK and BMAL1 functioning as positive regulators, binding to E-box regulatory elements and transactivating the transcription of the per and Cry genes, as well as multiple other rhythmically expressed genes. Phosphorylation of PER by CSNK1 can lead to its degradation. PER and CRY proteins form dimers and enter the nucleus to repress the CLOCK:BMAL1 complex. Rhythmic transcription leads to rhythms in key regulator genes and cellular functions that can feedback to alter core clock function. In addition, CLOCK affects histone acetylation, another mechanism by which gene activity can be modulated.

Circadian Rhythms are Heritable

One of the first questions to consider is one of “Nature or Nurture”. To what extent are attributes of circadian rhythms determined by one’s genes (Nature) as opposed to one’s environment (Nurture)? A variety of experimental approaches to this question have been employed over the years, all supporting the role of genetics. For example, attempts to alter the endogenous period of mice by gestation and rearing under non-24 hour days (either 20hr or 28hr light-dark cycles) failed to have lasting effects on free-running period in the mice as adults¹. Mice of an inbred strain, which are effectively genetically identical to one another, but genetically distinct from mice of another strain, can be compared. Significant differences between strains in circadian rhythm properties have been noted^2–6, that are above and beyond the differences between individuals of the same strain.

Similarly, studies of human twins have provided a means to estimate the “heritability” (proportion of phenotypic variance attributable to genetic variance) of circadian rhythms. The extent to which monozygotic (identical) twins are more similar than dizygotic twins can be used to estimate heritability. Hence, significant heritability has been described for the timing and pattern of serum cortisol rhythms^7–8 and of chronotype (morningness vs. eveningness)^9–10 by twin comparisons.

Evidence of heritability does not provide an indication of the number or identity of the genes at play, however. Genes contributing to one aspect of circadian rhythmicity (the endogenous, or “free-running” periodicity, for example) may be different from those involved in another (such as morning vs. evening preference). Further, variants in genes involved in critical functions may be deleterious and thus eliminated by natural selection-- meaning that the allelic polymorphisms naturally present in populations (and represented in strain or twin comparisons) may be in different genes from those essential for generation of circadian timing. Hence, there are many reasons to expect the existence of multiple types of circadian rhythm genes.

The Core Circadian Clock

Early evidence suggested a genetic mechanism lay at the core generator of self-sustained circadian rhythms. In 1961, Colin Pittendrigh¹¹ noted that the clock mechanism could be contained in single cells and was considered unlikely to be simply a chemical reaction since it was relatively insensitive to being sped up or slowed down by changes in temperature. Further evidence that gene transcription and translation was involved included demonstration that inhibition of protein synthesis could reset, or temporarily stop, the circadian clock^12–14. To identify what the genes and proteins comprising the circadian clock were, induction of mutations and screening for abnormal circadian rhythms proved the most effective approach-- in both Drosophila melanogaster¹⁵ and in mice¹⁶. Mutagenesis and phenotypic screening has also been successful in identifying circadian clock genes in other organisms such as the fungus Neurospora crassa¹⁷, plants¹⁸ and cyanobacteria¹⁹.

Described here are the identification of the transcription-translation feedback genes now known to form the core circadian clock. These are the “positive” elements CLOCK and BMAL1 that drive the transcription of the “negative” elements PER and CRY that in turn feedback to inhibit CLOCK:BMAL1.

Period

The first identified clock component gene, period (denoted per) was discovered in 1971 in Drosophila by a “forward genetic” approach consisting of chemically inducing random mutations in the genome, and detecting those mutations that affect circadian rhythms by screening the progeny of the mutagenized individuals for altered rhythmicity¹⁵. This approach has the advantage that no assumptions are made about the nature of the genes or gene products involved, but is based on the presumption that there exist single genes that, when mutated, will alter rhythms in a detectable manner. At the time, this presumption of the existence of single genes that could regulate a complex behavior was considered radical or absurd, but now appears to have been foresight.

Initially, three alleles of the per gene were identified by the process of mutagenesis and screening. Flies carrying these alleles had either no apparent rhythm in eclosion (emergence from the pupal case) or locomotion, or had either long (approximately 29 hours) or short periods (approximately 19 hours) for the rhythms of eclosion and locomotor activity¹⁵.

Confirmation of the importance of the per gene as a central circadian clock component was possible upon its molecular identification, for which the 2017 Nobel Prize was awarded to Hall, Rosbash, and Young. The gene’s function was demonstrated by rescue of the arrhythmic mutant phenotype after introduction of the wild-type allele of the per gene into mutant flies^20,21. The level of the mRNA transcript encoded by the per gene was shown to oscillate in a circadian fashion [40] as a result of transcriptional regulation²³, and the levels of the PER protein were shown to lag the per mRNA levels²⁴. In fact, shifts in the circadian phase can be evoked by the induction of PER protein under the control of a non-circadian promoter²⁵. Thus, many lines of evidence indicate that the per gene encodes a protein that is a clock component. Three orthologs of the per gene, Per1, Per2, and Per3, have now been identified in mammals and the levels of their mRNA have also been shown to oscillate with a circadian period^26–30.

Clock

In the early 1990s, no genes in mammals had been identified as even possible candidate circadian clock genes, leading to an effort to again apply chemical mutagenesis and phenotypic screening, this time in mice. In a screen of over 300 first-generation progeny of mutagen-treated mice (potential heterozygous mutants), Vitaterna and colleagues found one animal that had a free- running period of about 24.8 hours, more than six standard deviations longer than the mean of ~24.7 hours¹⁶. In homozygotes, this mutation results in a dramatic lengthening of the period to about 28 hours which is usually followed by the eventual loss of circadian rhythmicity (i.e., arrhythmicity) after about 1–3 weeks in DD. The affected gene was named Clock (an acronym for Circadian locomotor output cycles kaput¹⁶) and mapped to mouse chromosome 5^16,31. The Clock gene was cloned by a combination of genetic rescue and positional cloning. Clock/Clock mutant mice were phenotypically rescued by a bacterial artificial chromosome (BAC) transgene that contained the Clock gene, allowing for functional identification of the gene³². The Clock gene encodes a transcriptional regulatory protein having a basic helix-loop-helix DNA-binding domain, a PAS dimerization domain and a Q-rich transactivation domain. The mutant form of the CLOCK protein (CLOCK-∆19) lacks a portion of the activation domain found in wild-type protein, and thus, while it is capable of protein dimerization, transcriptional activation is diminished or lost. The protein dimerization domain is named PAS because the genes originally identified with it were, per, ARNT and sim. Clock mRNA is expressed in the SCN as well as other tissues, but it does not oscillate in a circadian manner in all tissues³³.

Bmal1

The presence of the PAS dimerization domain in CLOCK protein led to a screen for potential partners for the CLOCK protein. A protein of unknown function, BMAL1 (Brain and Muscle ARNT-Like 1, now denoted ARNT-Like or Arntl), was able to dimerize with the CLOCK protein³⁴. Creation of mice with a null allele of Bmal1 demonstrated the critical role of this gene in circadian rhythm generation. These mutant mice, while displaying light-dark responsive differences in activity level, become arrhythmic immediately upon release in constant darkness.

Additional actions of the CLOCK:BMAL1 heterodimer have become clear. While Clock mRNA does not oscillate, its protein’s nuclear vs. cytoplasmic localization does³⁵. By studying the intracellular localization of CLOCK and BMAL1 in fibroblasts of mouse embryos with mutations in different clock genes, and ectopically expressing the proteins, it was found that nuclear accumulation of CLOCK was dependent on formation of the CLOCK:BMAL1 dimer, as was phosphorylation of the complex and its degradation³⁵. Other PAS domain containing proteins failed to affect the localization of CLOCK, indicating that these posttranslational events are specific to the CLOCK:BMAL1 dimer.

Following the identification of CLOCK:BMAL1 dimerization, the ability of this heterodimer to regulate transcription was tested using a reporter construct based on the upstream regulatory elements of the per gene. The per gene of Drosophila contains an upstream regulatory element, the “clock control region”, within which is contained a sequence needed for positive regulation of transcription, the E-box element (CACGTG)³⁶. CLOCK-BMAL1 heterodimers were found to activate transcription of the mPer gene in a process that requires binding to the E- box element³⁴. However, CLOCK-∆19 mutant protein was not able to activate transcription, consistent with the finding that exon 19, which is skipped in Clock mutants³³, is necessary for transactivation. Thus, CLOCK protein interacts with the regulatory regions of the per gene to allow transcription of the per mRNA and eventual translation of PER protein. However, this positive regulation alone will not produce an oscillation in per mRNA levels, which is known to be responsible for the oscillation in PER protein levels²³. Findings that the Clock mutation dramatically decreases per genes’ expression also confirms the positive regulation of CLOCK:BMAL1 on per transcription in situ^37,38. Mice with null mutations of Per1, Per2, or Per3 alone display altered circadian periods^39,40, while mice with both Per1 and Per2 null mutations lose rhythmicity. Per3 null mutant mice exhibit only a subtle alteration in rhythmicity and Per1/Per3 or Per2/Per3 double mutants are not substantially distinct from the Per1 or Per2 single mutants. These findings suggest there may be some functional redundancy among the different mammalian per genes, but raise the question of the role of Per3 in mammalian circadian rhythms.

Cryptochromes

Cryptochromes are blue light responsive flavoprotein photopigments related to photolyases, so named because their function was cryptic when first identified. In mammals, two cryptochrome genes, Cry1 and Cry2 have been identified, and were found to be highly expressed in the ganglion cells and inner nuclear layer of the retina as well as the SCN⁴¹, and their mRNA expression levels oscillate in these tissues. Targeted mutant mice lacking Cry2 exhibit a lengthened circadian period, while mice lacking Cry1 have shortened circadian period; mice with both mutations have immediate loss of rhythmicity upon transfer to constant darkness^42–45. Thus, like the mammalian period genes, the cryptochrome genes appear to have both distinct (given their opposite effects on circadian period) and compensatory (given that either gene can sustain rhythmicity in the absence of the other) functions.

Further evidence for a central clock function is the finding that the cryptochromes appear to share a number of regulatory features with the period genes. In Clock mutant mice, the mRNA levels of Cry1 and Cry2 are reduced in the SCN and in skeletal muscle⁴⁶, suggesting that the cryptochromes also are induced by CLOCK:BMAL1 transactivation. Using mammalian cell lines, CRY1 and CRY2 were found by coimmunoprecipitation to interact with PER1, PER2, and PER3, leading to nuclear localization of the CRY:PER dimer as indicated by co-transfection assays with epitope-tagged proteins⁴⁶. Either CRY:CRY or CRY:PER complexes were capable of inhibiting CLOCK:BMAL1 transactivation of Per1 or vasopressin transcription⁴⁶. Thus, the CRYs as well as the PERs are capable of a negative feedback function, inhibiting CLOCK:BMAL1-induced transcription.

Casein Kinase 1

The tau mutation of the hamster arose spontaneously in a laboratory stock⁴⁷. The mutation is semi-dominant and shortens the period from 24 to 22 hours in heterozygotes, and 20 hours in homozygotes. The mutation pre-dated the Clock mutation and demonstrated that single gene mutations could profoundly alter⁴⁸ the circadian clock in mammals, just as in flies and Neurospora. Unfortunately, the genetic tools needed for cloning this important and interesting gene were not available for the hamster, and thus its molecular identity could not be determined by conventional genetic mapping/positional cloning approaches.

Lowrey and colleagues were able to identify a genomic region of conserved synteny (a grouping of genes together on a chromosome) in hamsters, mice, and humans, that encompassed the tau mutation⁴⁹. Tau was thus identified as being a mutation in the Casein Kinase 1 epsilon (CK1e) gene, the mammalian ortholog of the Drosophila doubletime gene. Sequencing of the gene identified a point mutation, which leads to altered enzyme dynamics and autophosphorylation state. CK1e can phosphorylate PER proteins, and the tau mutant enzyme is deficient in this ability. Thus, CK1e may lead to degradation of PERs, slowing the accumulation of PER in the nucleus and thus repression of CLOCK:BMAL1. Casein Kinase 1 delta (CK1d) has also been implicated in mammalian circadian rhythmicity^50–53.

Interacting Clock Pathways

Although not depicted in figure 1, a number of genes have been identified as key regulators of the core circadian clock genes. These genes form addition feedback loops to the genetic clock, and are described here.

Rev-erb alpha and ROR

While the negative feedback of PER and CRY proteins on their own CLOCK:BMAL1-induced transcription constitutes a form of negative feedback, and may be sufficient to explain the oscillations in expression of Per and Cry genes, the rhythmic expression of Bmal1 with an opposite phase was not explained by this feedback. Rev-erbα, an orphan nuclear receptor (now denoted Nr1d1), was the missing link. Its promoter region contains 3 E-boxes and transcription is thus positively regulated by CLOCK and BMAL1⁵⁴. Its transcription is negatively regulated by PER and CRYs and is at a minimum when mPER2 is at a maximum, and it is constitutively expressed at intermediate levels in Cry1/Cry2 or Per1/Per2 double knockouts. REV-ERBα protein appears to drive the circadian oscillation in Bmal1 transcription: the Bmal1 promoter includes two RORE sequences (enhancer sequences that recognize members of the REV-ERB and ROR orphan nuclear receptor families) and Bmal1 expression is drastically reduced in Rev- erba null mutants⁵⁴. Thus Rev-erbα may act to link the positive and negative regulatory signals of other clock genes to the transcription of Bmal1 and to cellular metabolism. CLOCK:BMAL1 activity is suppressed by a NAD⁺ dependent deacetylase SIRT1, which is in turn regulated by the circadian control of NAD⁺ biosynthesis^55–58.

Rev-erba also contributes to the differences between the phase of Cry1 mRNA rhythms relative to other clock genes whose transcription is enhanced by CLOCK:BMAL binding to E- boxes. The Cry1 gene has three candidate REV-ERB/ROR binding sites⁵⁹; in vitro assays indicate that REV-ERBα binds to two of these sites. Luciferase reporter assays indicate that REV-ERBα protein can inhibit transcription of Cry1 through binding at these two sites. REV- ERBβ also appears to share some functional redundancy with REV-ERBα⁶⁰.

Fbxl3 and Fbxl21

In addition to Rev-erb, other genes modulate the Cryptochromes. The Overtime mutation in mice was identified in a mutagenesis screen based on a lengthened free-running circadian period⁶¹. The responsible mutation was ultimately identified as being in a known gene encoding the F-box protein Fbxl3, but a gene previously unknown to be involved in circadian rhythmicity. Fbxl3^OVTM mutant appear to be functionally comparable to null mutants. FBXL3 protein leads to degradation of CRY1, while the FBXL3-OVTM mutant protein is less effective in this capacity. Thus, the period lengthening may be a direct result of a delay in degradation of CRY, effectively preventing the core cycle from restarting. The closely related protein FBXL21⁶² has been found to function in a similar manner^63,64.

NPAS2

NPAS2 (neuronal PAS family member 2) shares the closest homology with CLOCK of all identified bHLH-PAS family members. Null mutants of this gene have altered circadian activity patterns, noteably the absence of a “siesta” in later subjective night, but no dramatic alterations in circadian free-running period or persistence⁶⁵. However, when null mutants of the Clock had less dramatic phenotypes than the dominant-negative Δ19 mutant^31,66, the role of NPAS2 was re- examined. In the absence of functioning CLOCK, NPAS2 appears to be able to partially compensate⁶⁷.

Dec1 and Dec2

Like other clock genes, Dec1 and Dec2 are basic helix-loop-helix transcription factors that bind to E-boxes. DEC1 and DEC2 have been found to inhibit transactivation of Per by CLOCK and BMAL1⁶⁸. DEC1 and DEC2 form dimers⁶⁹. The inhibition of CLOCK and BMAL1 transactivation may be related to interactions with BMAL1, but can also be attributed to binding to (and thus possibly competition for) E-boxes⁷⁰. A human allelic variant in Dec2 has been linked with total sleep time⁷¹. This functional relationship has been confirmed in transgenic mice expressing the human Dec2 allele.

Polygenic Circadian Traits

A few spontaneous allelic variants in core clock genes have been identified from a familial pattern that suggests a single gene pattern of inheritance. For example, a familial form of Delayed Sleep Wake Phase Disorder (DSWPD, the most common circadian sleep-wake disorder) has been identified, associated with a mutation in the CRY1 gene⁷². Individuals with Advanced Sleep Wake Phase Disorder (ASWPD) fall asleep several hours earlier than the general population, but again if allowed to sleep during their desired time window, sleep quality and duration are normal for age⁷³. There is a strong familial component to ASWPD, with mutations demonstrated in casein kinase Id⁵¹, and the casein kinase Ie (CKIe) binding region of PER2⁷⁴ and of CRY2⁷⁵. These mutations result in a more rapid progression through the transcription/ translation feedback loop, leading to an overall shorter intrinsic circadian period⁷⁶. However, in the majority of cases, traits such as intrinsic circadian period are continuous or quantitative (rather than discrete or qualitative), indicative of multiple genetic loci influencing the trait.

Two general approaches to the identification of genes underlying heritability of circadian rhythms in mammals have been taken. In mice, the approach of identification of Quantitative Trait Loci (QTLs) uses linkage mapping to find genes or genomic regions that contribute to quantitative differences in circadian traits. In humans, the approach of Genome-Wide Association Studies uses statistical association of circadian phenotypic traits to genetic polymorphisms in populations.

Quantitative Trait Loci

Genetic heterogeneity underlies many phenotypic variations observed in circadian rhythmicity. The complex nature of circadian behavior is reflected in the phenotypic variability observed in mammalian species; this has been described particularly well in mice. Significant differences in circadian behavior have been demonstrated when comparing multiple inbred strains of mice^2–5. In particular, when endogenous period has been measured, differences of nearly a full hour have been observed among the most divergent strains. Strain differences are also present in other aspects of circadian behavior, including entrainment to light-dark cycles and the robustness of the rhythm. Crosses among such inbred strains have indicated a polygenic basis for these differences in circadian behavior.

The first comprehensive quantitative trait locus (QTL) analysis of circadian behavior in mammals (mouse) and the first to use genetic interaction analysis to identify QTLs was conducted in 2001⁷⁷. In this work, 14 QTLs in the mouse genome associated with five different phenotypic traits underlying circadian behavior. Importantly, all but one of these QTLs did not overlap with known circadian clock genes, demonstrating that many allelic variants in genes beyond the canonical circadian clock mechanism must exist. Although severe disruption of circadian rhythms may be caused by mutations in core clock genes, it is possible that the broad variety of circadian behavior observed in mammalian species, including humans, is the result of polymorphisms in multiple, interacting loci.

Molecular identification of interacting QTL

Quantitative genetic approaches were used to identify modifiers of the original Clock mutation, Clock^∆19, in the genetic backgrounds of C3H/HeJ and BALB/cJ inbred mouse strains. In C3H/HeJ, genetic suppression of circadian phenotype in Clock^∆19 mutant mice is mediated via the melatonin biosynthetic pathway⁷⁸. Two major QTLs were mapped on chromosomes 11 and X in which genes encoding melatonin biosynthetic enzymes, Aanat (aralkylamine N- acetyltransferase) and Asmt (acetylserotonin O-methyltransferase) are mapped, respectively. It has been known that most laboratory mouse strains, except C3H/HeJ and CBA/J, do not make detectable level of melatonin due to the mutations on Aanat and Asmt^79,80. Melatonin and the selective melatonin agonist, ramelteon, indeed can phenocopy the genetic suppression of the Clock^∆19 mutation in the suprachiasmatic nucleus (SCN). The Clock^∆19 mutation has also been shown to be associated with metabolic syndrome in mice⁸¹. It is therefore interesting that the melatonin receptor 1B (MTNR1B) gene has been strongly associated with fasting serum glucose levels in humans^82–84. Perhaps, the suppressive effects of melatonin on circadian rhythm may also have relevance to other phenotypes such as metabolism.

On the other hand, the gene responsible for suppression of the Clock^∆19 mutant phenotypic in the BALB/cJ strain background is a transcription factor, Usf1, occurring as a consequence of a single-nucleotide change in the promoter region that increases mRNA and protein expression in BALB/cJ mice⁸⁵. USF1 and CLOCK:BMAL1 compete for the same E-box regulatory sites in circadian genes, such as Per1,Per2 or Dbp. The Clock^∆19 mutation significantly reduces the affinity of CLOCK:BMAL1 binding to E-box sites. The reduced affinity of the mutant CLOCK:BMAL1 complex permits USF1 to compete more effectively for the same E-box sites, and the transcriptional activation potential of USF1 can then rescue the reduced transcriptional activation of the CLOCK∆19 mutant protein. USF1 and CLOCK:BMAL1 DNA binding were examined on a genome-wide basis using ChIP-Seq technology. This revealed that the Clock^∆19 mutant causes dramatic increases in USF1 binding genome-wide and also reveals that the USF1 and CLOCK:BMAL1 transcriptional networks interact extensively. This has ramifications for understanding the shared role of USF1 and CLOCK:BMAL1 in the regulation of metabolism and lipid biosynthetic pathways.

Genome-Wide Association Studies

Genome wide association studies (GWAS) are used to identify associations between chromosomal regions (loci) and trait of interest in human. Since genomic regions of linkage disequilibrium are much smaller in human populations (<100kb) than in experimental mouse cross (40–80Mb), the mapping resolution in GWAS is much higher than mouse QTL so that very limited number of candidate genes are going to be identified. Several large scale GWAS were conducted for circadian amplitude and for chronotype (such as morningness and eveningness) recently^88–88. These studies identified a large number of loci, most of which are not canonical clock genes.

These human findings coincide with the previous QTL mapping in mouse crosses⁷⁷ and supports the notion that the broad variety of circadian behavior observed in humans in phase and in amplitude is the result of polymorphisms in multiple, interacting loci. The number of loci detected in these GWAS was much higher than the mouse study. This may be because the number of samples for GWAS are 2–3 log unit higher than mouse study, so that detection of loci associating with circadian phenotype is much more sensitive. Natural variations in human circadian rhythms are subject to complex, polygenic regulation, with allelic variants at large numbers of loci affecting the circadian phenotypes. With the possibility of a hundred or more loci influencing the phenotype, the relative contribution of each genetic locus may be very small.

The Circadian Transcriptome

As described above, the molecular circadian clock, at its core, is a transcriptional and translational feedback loop, a direct cellular output of which is waves of oscillatory gene expression driven by the rhythmic transcriptional activity of the clock. It has been a central question to understand the program and functional importance of the circadian transcriptome in gene expression profile studies, in which gene expression levels in the SCN and peripheral tissues are accessed at regular time-points under entrained (e.g., 12hr Light:12hr Dark) or free- running (e.g., constant dark) conditions. Early studies have estimated that roughly 30% of the genome is under circadian control^89–92. This number has been increasing since due to technical advances in gene expression profiling techniques and larger sample sizes used in subsequent studies. A comprehensive study in mice sampled 12 tissues and found that collectively the expression of 43% of all protein-coding genes showed a circadian rhythmicity⁹³. A recent study found an even higher number of genes whose expression were rhythmic in at least one tissue of a diurnal primate⁹⁴. These gene expression-profiling studies have revealed a complex circadian transcriptome. First, the circadian transcriptome is highly tissue-specific, and only a small fraction of clock-regulated genes are rhythmic across different tissues^91–95. This tissue specificity appears to be associated with the particular function and metabolic need of the tissue^91–96. For example, genes rhythmic in the SCN are involved in neuropeptide metabolism and are known for regulating the circadian locomotor activity, while genes cycling in the liver are important for nutrient metabolisms and regulation of metabolic intermediates⁹¹. In addition, the phases (e.g., peak time) of the circadian gene expression in a tissue distribute across the entire 24-hour day, as opposed to be restricted to the peak of transcriptional activity of the clock in the tissue, although tissue-specific “rush hours” (times when a large proportion of rhythmic genes have peak transcription) can be observed⁹³. Furthermore, even for a gene that are cycling across multiple tissues, the peak of the oscillation can vary by hours in one tissue compared to another⁹³. The complexity of circadian transcriptomic oscillations across multiple tissues raises an intriguing implication that the circadian transcriptomic program must be under intricate orchestration throughout the body for an overall optimal functional output and health outcome.

A number of mechanisms may contribute to this complexity of circadian transcriptome. First, the tissue-specific circadian program in the transcriptome is also likely to be associated with tissue-specific configurations of the epigenome, the chemical modifications to the DNA and histone proteins that regulate the binding of transcription factors to regulatory regions of genes and thus the transcriptional activity. Tissue-specific epigenome configurations are critical to establishing cell identity. The epigenome also exhibits clock-regulated circadian oscillations, which correlate with the rhythms in the transcriptome^97–100. Interestingly, CLOCK shows histone acetyltransferase activity, which modifies chromatin to allow for the accessibility of regulatory elements of genes by transcription factors¹⁰¹, a required step for clock-driven circadian transcription activation¹⁰². Secondly, in addition to interacting with each other to regulate gene expression, each of the clock proteins also interacts with other transcriptional regulators with distinctive profiles of transcriptional target genes¹⁰³. For example, CLOCK in flies interacts with a range of partner transcription factors and binds to cis-regulatory motifs of genes in a tissue- specific manner, which is associated with the tissue-specific program of circadian transcriptome¹⁰⁴. Furthermore, by comparing oscillatory patterns of total mRNA to pre-mRNA or nascent mRNA, recent studies suggest that rhythmic transcription activation may only contribute to a portion or even a small portion (e.g., 20%−30%) of the rhythms observed in the transcriptome and that post-transcriptional mechanisms are involved^103,105,106. Consistent with these observations, the transcript levels of hundreds of non-coding regulatory RNAs are also rhythmic^93,99.

In addition to these cellular mechanisms, the circadian transcriptome is also regulated by systemic signals either directly from the central brain clock in the suprachiasmatic nuclei (SCN) or indirectly from SCN output rhythms. When the clock machinery was arrested in the liver by tissue-specific inhibition of Bmal1 expression, 10% of circadian transcriptome remained cycling, suggested a strong contribution from the central clock⁹⁶. Time of feeding, an output rhythm driven by the SCN clock, together with the SCN clock itself, has been shown to drive the rhythms in the hepatic transcriptome in mice^107,108. Similarly, when mice were sleep deprived at different times of the day so that their sleep debt no longer varied across the circadian cycle, the rhythmic pattern of gene expression in the brain are largely diminished, indicating circadian rhythmic gene expression in the brain are largely dependent on the sleep/wake rhythms¹⁰⁹. Consistent observations have also been reported in human peripheral blood^110,111. Finally, body temperature rhythms also reset the phase of the circadian clock in peripheral tissues¹¹².

The combination of cellular and systemic mechanisms in shaping the peripheral circadian transcriptomic profiles has important health implications. Under conditions such as jet lag and shift work, the timing signals driven by the SCN clock, behavioral activities, and the local clock are no longer in concert, leading to a state known as the “internal desynchrony”. Internal desynchrony is associated with disrupted circadian transcriptome, altered functional outputs, and often disorders. As an example, feeding at a “wrong” time of the day, a condition associated with disrupted metabolic output and deregulation of body weight¹¹³, creates desynchrony between the transcriptional output of the local clock and systemic signals from the master clock and a reprogrammed circadian transcriptome¹⁰⁷. Another important health implication of the circadian transcriptome can be drawn from the finding that many genes encoding drug targets or metabolizing enzymes are expressed rhythmically⁹³. Thus, the efficacy and toxicity of a drug, particularly one with a short half-life, may be influenced by the time when the drug is taken, a phenomenon that has been reported in circadian pharmacokinetics studies^114,115.

A challenge of circadian-based drug treatment is the lack of biomarkers of internal timing, which is characteristic to individuals and is highly variable in human populations¹¹⁶. Although in principle, internal timing can be inferred using transcriptomic and metabolomic data at only one time point^117,118, or with only a few markers in combination with advanced statistical methods¹¹⁹, its utility remains to be tested under disease conditions, which typically disrupts circadian transcriptomic patterns, such as seen in the brain of depression patients¹²⁰. In addition, given the complexity of the tissue-specific circadian program, methodologies capable of inferring tissue-specific circadian timing and internal desynchrony are needed. Finally, future studies are needed to demonstrate whether the circadian transcriptome can be used as a disease signature to aid the disease diagnosis, or even as a reference point to treat the disease by restoring the “optimal” cycling pattern. Indeed, these are very challenging tasks at present, particularly given the circadian transcriptome is also plastic to environmental inputs, such as diet¹²¹, and are modulated by aging¹²². Nevertheless, with the advances in circadian metabolomics¹²³ and proteomics^124–128, we are now beginning to gain an even more detailed understanding of the tissue-specific circadian program and its role in health and disease, building a knowledge base that will eventually enable circadian-based medicine.

Conclusions: Clock genes are everywhere

The core circadian oscillator can be found in individual cells. The basis for this oscillation in mammals, as in other organisms, lies in rhythmic feedback regulation of transcription of clock genes. The levels of the PER and CRY proteins alter the rate of transcription of their own genes. This alteration is achieved by inhibition of the enhancement of transcription that results from binding of the CLOCK-BMAL1 heterodimer to the E-box element of the promoter region of the Per and Cry genes. Additional interactions between circadian clock proteins may slow the time course of this feedback, achieving the near-24-hour interval: the phosphorylation of PER by CKIε may lead to its degradation, and the association with BMAL1 appears needed for CLOCK to be present in the nucleus. Mutations in core clock genes are involved in the pathogenesis of some familial circadian rhythm sleep-wake disorders. Beyond these core clock genes, rhythmic transcription programs regulate gene expression and gene activity throughout the genome. Finally, it appears that much of the heritability of circadian rhythm traits in mammals stems from polymorphisms in numerous loci beyond the set of core clock genes, with implications for a wide range of neurologic disorders. As computational and molecular techniques mature, future studies are expected to synthesize an understanding of the circadian genetic program at a multi-scale level, with important implications for circadian medicine.

Key Points.

• Circadian rhythms are generated by a genetic mechanism: a feedback loop of gene transcription and translation of core “clock genes”. This genetic clock mechanism functions in diverse cell and tissue types throughout the body.

• Gene and polymorphisms in both core clock genes and interacting genes contribute to individual differences in the expression and properties of circadian rhythms.

• The circadian clock profoundly influences the patterns of gene activity in nearly all tissues, indicating that genetic alterations lead to expression and development of disease.