Genetics and Epigenetics of Circadian Rhythms and Their Potential Roles in Neuropsychiatric Disorders

Abstract

Circadian rhythms alterations have been implicated in multiple neuropsychiatric disorders, particularly sleep-wake disorders, addiction, and anxiety and mood disorders. Circadian rhythms are known to be maintained by a set of classic clock genes that form complex mutual and self regulatory loops. While many other genes showing rhythmic expression have been identified through genome-wide studies, their roles in the circadian regulation remain largely unknown. In attempts to directly connect circadian rhythms and neuropsychiatric disorders, genetic studies have identified genes with mutations associated with several rare forms of sleep disorders or sleep-related traits. Other than that, genetic studies of circadian genes in psychiatric disorders have yielded limited success. As important mediator of environmental factors and regulators of circadian rhythms, the epigenetic system may hold the key to the etiology or pathology of psychiatric disorders, their subtypes or endophenotypes. Epigenomic regulations of the circadian system and its related changes have not been thoroughly explored in the context of neuropsychiatric disorders. We argue for systematic investigation of the circadian system, particularly epigenetic regulation, and its involvement in neuropsychiatric disorders, to improve our understanding of human behavior and disease etiology.

Introduction

Circadian rhythms are endogenous biological cycles that are approximately 24 hours in length. They are found in most living organisms, and can be adjusted by factors called zeitgebers, or “time-givers”, including light^[1], temperature^[2], diet^[3], odor^[4], and gravity^[5], with light being the dominant cue. Maintaining a rhythmic daily life is critical for a living organism to survive the recurrent environmental changes on Earth. These rhythms can be easily observed in behaviors such as sleeping and eating, but also, less visibly, affect crucial biological systems such as metabolism^[6;7] and the cardiovascular^[7] system.

Multiple evidence have suggested the potential roles of circadian rhythms in neuropsychiatric disorders such as sleep disorders, anxiety, mood disorders and addiction, Meanwhile, studies in animal models have identified several regulators and effectors of the endogenous clock. These core clock genes are known to exhibit transcriptional-translational auto-regulatory complexes. However, these remain insufficient to explain all our observations, especially the contribution to human behavior traits and disorders. Further identification of molecular components of circadian systems and their regulatory relationships is an important step for understanding neuropsychiatric disorders for better diagnostics and treatment.

This review will describe the current findings of genetic and epigenetic determinants of the circadian system in the context of neuropsychiatric disorders. Through reviewing literature, we will highlight the complexity of circadian regulation beyond the classic core clock genes. Such complexity involves many genetic and epigenetic factors. Since epigenetic mechanisms are important mediators of environmental factors and regulators of rhythmic gene expression, we therefore propose that developing comprehensive genome-wide and epigenome-wide data from multiple sample sources will improve our understanding of circadian regulatory system and its role in neuropsychiatric disorders.

Clock genes, rhythmic expression, and regulatory networks

Circadian rhythms in vertebrates are controlled by a conserved brain region at the anterior hypothalamus called the suprachiasmatic nucleus (SCN), made up of about 20,000 neurons. The SCN serves as a central regulator of circadian rhythms throughout the rest of the brain^[8] and the body^[9]. At the same time, peripheral tissues, even cultured cells^[10;11] have their own local, autonomous clocks that can be self-sustaining, but they may be synchronized by the signals from the SCN^[12].

Clock genes, underlying circadian rhythms, can be broadly defined as genes that show diurnal variation of activity or function, typically showing rhythmic changes of transcript abundance, as such measure is currently more accessible than other molecular phenotypes, such as protein levels and activities. Although an increasing number of genes have been found to demonstrate circadian characteristics of clock-controlled genes (or CCGs), a small set of genes is denoted here as core “classic clock genes (CG)”. The CGs include Period (PER), Timeless (TIM), Clock (CLK), Cycle (CYC, a Drosophila gene, with mammal homolog ARNTL or BMAL1), Cryptochrome (CRY), REV-ERBalpha, retinoic acid related-orphan receptor alpha (RORalpha), D-box-binding protein (DBP), thyrotroph embryonic factor (TEF), hepatic leukemia factor (HLF), E4BP4 (also known as NFIL3), deleted in esophageal cancer 1 (DEC1), DEC2, Neuronal PAS domain-containing protein 2 (NPAS2), and Double Time (DBT, a Drosophila gene, with mammal homolog casein kinase 1e, CSNK1E). These genes have been mostly identified through screening of mutants of fruit flies, mice and hamsters^[13]. These few CGs make up a group of auto-regulatory loops and present rhythmic expression of their own and their regulatory target transcripts. The CGs have been frequently called clock genes in the literature but as will be discussed in this review, these CGs only represents a small set of a much broader network of clock genes.

Most of the CGs encode proteins that function as transcription factors to drive rhythmic expression of their target genes. Some of the CG proteins form heterodimer complexes, such as PER-CRY, CLK- BMAL1, and TIM-PER. They not only regulate expressions of many other genes that carry E-box promoters, but also their own expressions. In contrast, DBP, HLF, TEF and E4BP4 regulate through D-box promoters^[14;15]; while the REV-ERBalpha and ROR family members bind to the REV-ERB/ROR response element (RRE)^[16]. cAMP response elements (CREs) is another central regulatory motif that mediates rhythmic expression^[17]. These regulatory systems have been thoroughly reviewed by several papers^[13;18–20].

In addition to CGs, hundreds of non-CG genes have been shown to transcribe rhythmically. They are part of the broadly defined clock genes. In fact, one would expect that genes carrying E-box, D-box, RRE and CRE promoter could be potential clock genes^[21;22]. Certainly, these genes could include both drivers and passengers of a large circadian regulatory system, though most of the cause-effect relationships still remain to be discovered. Increasing number of CGs have been identified through genome-wide expression profiling studies, mostly in mice, as summarized in Table 1. Just to name a few, 2 to 10% of genes were expressed in a circadian manner in various mouse tissues^[23–28]; in mouse SCN, 337 genes were found to be expressed cyclically, while 335 were in the liver^[26]. Another mouse study detected 575 genes in liver and 462 genes in heart with circadian expressions^[27].

Table 1.

list of genome-wide study of genetics, epigenetics of circadian rhythms

Author, Year	Species	Tissues	Measure	Strains	Microarray/Methods
McDonald MJ et al. 2001[153]	Drosophila	Head	Expression	clock mutant (Clk) flies	Affymetrix array
Hughes ME et al. 2012[135]	Drosophila	Brain	Expression	Wild-type Canton-S flies, y w, per0, and per0 in a Canton-S background	RNA-Seq, 100-bp paired-end reads. On average, each sample more than 40 million reads
Kornmann B et al. 2001[24]	Mouse	Liver	Expression	C57BL/6 X 129/SV	ADDER (Amplification of Double-stranded cDNA End Restriction fragments) differential display
Ueda et al. 2002[154]	Mouse	SCN and liver	Expression	Balb/c	Affymetrix U74Av2 and U74A oligonucleotide array
Akhtar RA et al. 2002[23]	Mouse	Liver	Expression		Custom-made cDNA microarray
Panda S et al. 2002[26]	Mouse	SCN and liver	Expression	male C57BL/6J and Clock mutant	Affymetrix mouse (U74A) high-density arrays
Storch KF et al. 2002[27]	Mouse	liver and heart	Expression	C57/Bl6	Affymetrix U74Av2 oligonucleotide array
Zvonic et al. 2006[155]	Mouse	Liver and brown and white adipose tissue	Expression	AKR/J	Mouse430a2
Oster et al. 2006[156]	Mouse	adrenal gland	Expression	C57BL/6J	Mouse4302
Lemos et al. 2006[157]	Monkey (macaca mulatta)	adrenal gland	Expression		Affymetrix U133-A GeneChips.
Zvonic et al. 2007[158]	Mouse	calvarial bone	Expression	AKR/J	Mouse430a2
Yang et al. 2007[159]	Mouse	prefrontal Cortex	Expression	C57BL/6J	Mouse4302
Maret et al. 2007[160]	Mouse	whole brain	Expression	C57BL/6J, AKR/J, DBA/2J	Mouse4302
Bray et al. 2007[161]	Mouse	atrium and ventricle	Expression	cardiomyocyte-specific circadian clock mutant	IlluminaMousev1
Miller et al. 2007[25]	Mouse	Liver, Skeletal Muscle, gastrocnemius	Expression	male C57BL/6J and Clock mutant	Custom-made genome arrays
Storch KF et al. 2007[28]	Mouse	Retina	Expression	male CBA/CaJ	Affymetrix mouse 430.2 arrays
Yan J et al. 2008[32]	Mouse, human and monkey	14 tissue	Expression	22 datasets
Hughes ME et al. 2012[12]	Mouse	Liver	Expression	ClockΔ19 mouse	Affymetrix Mouse Exon 1.0 ST Arrays
Zambon et al. 2003[162]	Human	Skeletal muscle	Expression		Affymetrix U95a
Leonardson AS et al. 2010[29]	Human	Blood	Expression		custom array
Li JZ et al. 2013[30]	Human	6 different brains areas	Expression		Affymetrix U133-A or U133Plus-v2 GeneChips.
Lim AS et al. 2014[31]	human	dorsolateral prefrontal cortex	Gene Expression and DNA methylation		Illumina Infinium HumanMethylation450k Bead Chip; Illumina HiSeq with 101 bp paired-end reads and 150M reads for the first 12 samples. The remaining samples 50M reads

Human studies have also revealed time-dependent expression in blood and brain. A 2010 study on gene expression induced by food intake identified expression changes associated with biopsy time for 8,197 genes in blood (false discovery rate (FDR) 2.6%)^[29]. This is the largest human circadian study thus far, with 40 individuals sampled for 14 time points each, for a total of 560 blood samples. In this study, Li et al. (2013) reported circadian expression in human postmortem brain. They used time of death (TOD) to represent time-points in the 24 hour cycle, turning individual differences into differences of expression at different time points. They analyzed 12,000 transcripts of six different brain areas (dorsolateral prefrontal cortex, anterior cingulated cortex, hippocampus, amygdale, nucleus accumbens, and cerebellum) from 55 controls and 34 major depressive disorder (MDD) patients^[30]. Among the healthy controls, more than 417 transcripts in each brain region showed 24-hour oscillations (nominal P < 0.05), while 169 genes have a FDR of <0.5 for combined P values across brain regions. These 169 genes were considered as common circadian genes in the brain^[30]. Another study on human postmortem dorsolateral prefrontal cortex samples of 536 subjects used RNA-Seq and global statistics to show rhythmic expression without naming specific clock genes^[31].

Most of these rhythmic expressions are tissue-specific, suggesting tissue-specific regulatory network. Only 28 cycling genes were found to be shared between mouse SCN and liver^[26], while 37 were shared between mouse liver and heart^[27]. Another analysis of 21 microarray data sets from 14 mouse tissues found that the expressions of 41 genes out of 19,168 genes showed consistent circadian oscillation across multiple tissues^[32]. This alerts us that study of circadian regulation system should take tissue-specificity into account.

It should be noted that the number of cyclically expressed genes reported is related to the experimental design, statistical method and significance cutoff. With the liberal significance criteria (P < 0.05) used in Li et al.’s human brain study, some of CGs still did not show rhythmicity^[30]. However, these findings are not definitive, since many pre- and post-mortem factors could have destroyed rhythmic expression patterns and produced false negatives. Moreover, as transcription factors are typically expressed at low level, major circadian regulators, which are often transcription factors, could be missed by some techniques used in genome-wide studies when sensitivity is not sufficient.

After considering experimental artifacts, the fact that many genes outside of the CGs are rhythmically expressed and that some CGs do not show cycling expression in genome-wide studies could have many different implications, including the possibility that our current list of clock genes may not be exhaustive. Given the complexity of the system, it is likely that we have not yet recovered the complete set of genes regulating and responding to circadian rhythms. Several studies have proposed novel genes as important central regulator genes. For example, a 2008 study of mutant mice proposed that NR3C1 and FKBP/HSP90 complexes are central to the control of circadian gene expression by environmental cues^[32]. CHRONO^[33;34] and UBE3A^[35] were found to be essential in regulating circadian rhythms in mice in three 2014 studies. Genes involving protein translation, including rRNA, also showed rhythmic expression^[36]. Furthermore, an siRNA screening on human osteosarcoma cell line, targeting 17,631 known and 4,837 predicted human genes, discovered about 343 clock genes or modulators^[37]. These data suggests that the underlying organization of circadian rhythms has not yet been completely described.

Rhythmic expression is certainly not the only aspect of circadian rhythms. Protein abundance and post-translational modifications, such as phosphorylation and ubiquitination, have also been shown to have daily oscillations in Neurospora crassa, fruit fly, and multiple tissues in mouse, rat, and hamster, as reviewed by others^[38–41]. Similar circadian molecular mechanisms may exist in human, but remain to be explored. From chromatin to transcripts, mRNA to proteins and to protein modifications, circadian rhythms encompasses a complex regulatory system, including the epigenomic components that will be discussed below.

Genetics of sleep-related traits

While circadian patterns can be observed from the molecular level all the way up to the organismal behavior level, the sleep-wake cycle and other sleep-related traits are probably the most salient outputs of the circadian clocks. These traits have been shown to be heritable^[44], and underlying genes can be identified directly through genetic methods using human population data, without relying on knowledge of specific known clock genes. The heritability of sleep measures, including timing, duration, and quality, vary between 12.4% and 29.4%^[42]. A study of 410 normal adults identified a polymorphism in CLK associated with morningness-eveningness preferences^[43]. Furthermore, in a GWAS of 4251 subjects, Allebrandt et al. (2013) identified an intronic variant (rs11046205) in the ABCC9 gene associated with sleep duration (P=3.99e-8)^[44].

On the other hand, the circadian system is not the only regulator of sleep. Energy homeostasis and its interactions with circadian also contribute to the maintenance of sleep-wake cycle^[45;46]. While homeostasis and other factors play significant roles in sleep regulation, sleep has been used as the major model to study genetics and regulation of circadian rhythms, although it should be noted that genetics of sleep-wake cycle is not necessarily all about circadian rhythms.

Healthy people vary in their preferences of sleep timing and length; some people are often classified as morning lark or evening owl chronotypes. If these variations do not impair quality of life, they are considered normal. Sleep-wake behaviors in human and animal models offer opportunities to understand circadian regulation. In humans, variable sleep traits or disorders provided subjects for studying the molecular bases of sleep regulation, and, by extension, circadian rhythms. In animal models, one can take advantage of better-controlled environmental factors to study their contribution to circadian regulation. For example, manipulating lighting environment and feeding pattern have been shown to induce circadian and related genomic and epigenomic changes^[47].

Circadian disruptions are implicated in multiple neuropsychiatric disorders

Disruption of circadian rhythms have been associated or implicated in many traits or diseases, including metabolic syndrome^[48], obesity^[49], diabetes^[50], inflammatory diseases and autoimmune disorders^[51], cancer^[52], drug efficacy and toxicity^[53], cardiovascular disorders^[54], and mental disorders^[55;56].

Among neuropsychiatric disorders, sleep-wake disorders, anxiety, mood disorders and addiction have the strongest connections to altered circadian rhythms. While the connection between circadian rhythms and sleep-wake disorder is self-evident, circadian rhythms have been implicated in other non-sleep psychiatric disorders based on biological and clinical observations. Specifically, many of these disorders exhibit co-morbidity with sleep disturbance and their treatments often elicit responses that are related to clock candidate genes, behavioral or clock gene expression changes in animal models. Additional link between circadian rhythms and neuropsychiatric disorders can be found in several candidate gene association studies. Major evidence are summarized in Table 2. A few examples of those indirect evidence linking circadian rhythms to non-sleep neuropsychiatric disorders can be provided as following: abnormal sleep is co-morbid with multiple disorders^[57]. Persistent sleep disturbances have been found to increase risk of developing anxiety^[58] and depression^[59;60]. Insomnia and substance abuse disorders promote risk of each other^[60;61]. Melatonin is important in synchronizing circadian rhythms^[62], and agomelatine that targets the melatonergic system is an antidepressant ^[63]; agomelatine has been shown to increase the relative amplitude of an individual’s rest-activity cycles^[64]. Ketamine, a drug with rapid-acting antidepressive effects, can influence the recruitment of the CLK-BMAL1 complex to E-box promoters and alter the expressions of CGs^[65]. Another antidepressant, escitalopram, was reported to restore the disrupted rhythmic expression of several CGs in a study of blood samples from 12 MDD patients and 12 controls^[66]. Animal models with disrupted clock genes show behavioral changes similar to mood disorders^[67] or schizophrenia^[68]. Therefore, circadian rhythms have been a system of interest in the study of these disorders. However, the question of causation largely remains to be addressed.

Table 2.

Studies That Implicate Circadian Rhythms in Psychiatric Disorders

	Mood Disorder (MDD and bipolar disorder)	Anxiety	Addiction
Sleep disturbance comorbidity, or as risk factor	Risk factor for developing MDD[59;60]. Abnormal circadian rhythms in hormone levels, body temperature, sleep, and behavioral patterns reported in patients with MDD[163–165]; The degree of circadian misalignment correlated with severity of depression[166]; Sleep disturbance featured in manic and depressive phases, even possibly euthymic phases[167;168].	Risk factor for developing anxiety[58]; children daily regularity predict anxiety levels more than a decade later[169]	Bi-directional relationship: addiction disrupt circadian rhythms, sleep and mood problem increase chance of addiction[60;170]
Therapy	Seasonal affective disorder (SAD) patients are responsive to light therapy[171]; Sleep deprivation treats depression[172]. Sleep deprivation may switch a depressed patient into hypomania or mania[173]. Agomelatine increase the relative amplitude rest-activity cycles[64]. Lithium lengthen the circadian period in treated hamsters[174].	Agomelatine affects circadian rhythms[64], is also effective in treating anxiety disorders[175;176].
Chronotype	Mood disorders associated with chronotype[177;178]		Evening chronotype associated with substance abuse[179;180]
Social rhythms	disrupted social rhythms[181–183]	Lower daily regularity in anxiety patients[184]
Seasonal pattern	Seasonal changes in bipolar[185;186]		A case report for SAD with cyclical cocaine craving[187]
Hormone	Rhythmic melatonin level[188;189]	Cortisol levels is rhythmic[190]	Cortisol secretion patterns linked to addiction[191]
Clock gene genetic association	17 studies of candidate genes[192]	One candidate gene study[193]	Six genetic studies[194–199]
Clock gene in animal model	Knockout of REV-ERBalpha was found to increase midbrain dopamine production and induce mania-like behavior through regulating tyrosine hydroxylase (TH) gene expression in a mouse study[200]. Clk mutant shows mania[67]	Per1 and Per2 expression levels in the nucleus accumbens were reported to regulate anxiety levels in knockout mouse models and mice experiencing chronic social defeat stress[201]. Clk mutant less anxious[67]	Fourteen studies of Clock, per1 and per2 mutant mice showed behavior changes in response to drug[202]
Clock gene expression in response to drug treatment	Ketamine influence the expressions of clock genes[65]. Escitalopram restore the disrupted rhythmic expression of several clock genes[203]. Lithium’s effects on expression of multiple clock genes were reported[204–207]. Valproic acid changes the phase and amplitude of PER2 expression in cultured cells[117]	Anxiolytic medications reduce mPer1 expression in mice[208]	chronic methamphetamine treatment desynchronize clock gene expression between straitum and SCN[209]; ethanol and drugs of abuse alter clock gene expression in SCN and other brain regions[202]
Clock gene expression changes in patients	Expression changes of clock genes reported in MDD brains[30] and blood[210].
Review	Gonzalez 2014[192]	Philip Gorwood, 2012[69]	Logan RW et al. 2014[202]

A multi-system hypothesis has been formulated to explain the connections between circadian rhythms disturbance and addiction, as well as anxiety. Based on literature review, Gorwood highlighted the cortisol-melatonin-vasopressin interaction for anxiety, as this interaction nicely bridges stress response and circadian systems^[69]. Drugs of abuse may influence the interwoven molecular networks of circadian rhythms, stress-response, reward circuitry, neuroplasticity and memory, and ultimately lead to the development of addiction, as well as withdrawal symptoms. The paraventricular nucleus (PVN) in the hypothalamus has been proposed to be the location where circadian and stress signals converge, and where multiple clock genes, neuropeptides, and stress response genes interact^[70]. Such interaction between circadian systems and stress response systems may play an important role in multiple psychiatric disorders, including but not limited to addiction^[71].

The Dopamine D2 receptor (DRD2) is another interesting candidate linking the circadian system and the reward pathway, as it mediates photic response to regulate circadian rhythms, and the most important reward pathway is dopaminergic ^[72]. Multiple candidate gene studies have found significant associations between DRD2 variants and different kinds of addiction (alcohol, cocaine, heroin, and nicotine)^[73;74]. However, a meta-analysis^[75] and a genome-wide association study (GWAS)^[76] of alcoholism reported inconsistent results, suggesting that the DRD2 contribution may be small, if at all. DRD2 variant was also reported to be associated with anxiety disorders with co-morbid alcohol use disorder^[77]. But the finding is also weak and requires replication.

Circadian genetics of neuropsychiatric disorders

Genetic studies may capture some direct evidence that specific clock genes are involved in neuropsychiatric disorders if mutations or variants of clock genes are associated with risk of disorders. Genetic association is an important venue leading to translation of animal model clues to clinical relevance.

Sleep-wake disorders

Sleep-wake disorders impair quality of life, affect learning, memory, and mood. These disorders have a clear genetic basis. Family and twin studies of insomnia reported heritability ranging between 21% and 58%^[78–80] (see Palagini et al. (2014) for a thorough review^[81]). Travel across time zones, sleep deprivation, and shiftwork all disturb sleep patterns. Insomnia, hypersomnia and narcolepsy are major sleep-wake disorders^[82]. Some of these are common, like insomnia (about 10% adults have severe insomnia that cause daytime consequences^[83]), and some are rare, like narcolepsy (affecting about 1 in 3,000)^[84]. Jet lag and shiftwork-related sleep problems are common in specific occupations, like flight attendants, nurses, and soldiers.

Mutations in genes responsible for some specific, mostly rare forms of sleep disorders have been discovered. A mutation (R192H) in GABRB3 has been found in patients with chronic insomnia. Since GABRB3 encodes a subunit of a chloride channel that serves as the receptor for gamma-aminobutyric acid (GABA), a major inhibitory neurotransmitter of the mammalian nervous system, this suggests that a decrease in GABAergic inhibition may contribute to insomnia^[85]. Interestingly, hypersomnia was also recently linked to GABA(A) receptor regulation^[86]. Another excessive sleeping disorder, narcolepsy, is mostly caused by a deficiency in hypocretin (HCRT), an excitatory neuropeptide^[87].

Mutations in PER2 and CSNK1D have been reported in Advanced Sleep Phase Syndrome (ASPS) patients. A rare autosomal dominant mutation of PER2 has been found to be responsible for ASPS in members of a Utah family^[88]. CKIdelta (CSNK1D) was found to have a missense mutation responsible for ASPS^[89].

Knockout or mutation of many other genes, including IA2^[90], have been found to change sleep-related behaviors in mice, and have been reviewed elsewhere^[91]. These genes could be candidates for human sleep disorders, but mutations have not been detected in humans so far.

The search for genes of common forms of sleep disorders has produced positive and negative results. A GWAS of insomnia with 2,267 samples did not detect any genome-wide significant association^[92]. However, several other sleep disorders have yielded genome-wide significant signals in GWASs with hundreds of cases, including restless legs syndrome (MEIS1, BTBD9, PTPRD, MAP2K5, SKOR1, TOX3, BC034767, MAP2K5, LBXCOR1)^[93–96] and narcolepsy (TRA-alpha, TRAJ10)^[97].

It should be noted that circadian defect is not the only cause of sleep-wake disorders. Cardiovascular, neurological, and pulmonary diseases, use of substance and medication, irregular metabolism, and bad habits all could disturb sleep. One can certainly argue that some of those genes associated with sleep disorders may not be involved in circadian regulation at all. In fact, among the classic clock genes, CGs, only PER2 has been found to carry a mutation (c.1984A>G (p.Ser662Gly)) responsible for a sleep disorder^[88]; all the other associated genes are outside of the CGs. Most of them do not have any known connection to circadian regulation, or have not been studied for rhythmic expression. How sleep disorder-associated genes are related to circadian remains to be investigated. It may turn out that some of those non-CG genes are also actual clock genes, participating in circadian regulation.

Non-sleep neuropsychiatric disorders

Genetic variants of candidate genes of CGs, have been tested for association with bipolar disorder^[98;99], depression^[100], seasonal affective disorder^[101;102], or anxiety disorders^[77], alcohol use^[103], heroin addiction^[104], bipolar and schizophrenia^[105], major depression and bipolar^[106], depression and sleep disorder^[107], Though some positive associations were reported, most findings were weak and not replicated.

Genetic associations of about 360 selected clock genes were systematically assessed in 14 psychiatric GWAS data sets based on relaxed thresholds for significance, by McCarthy et al. (2013)^[108]. Bipolar disorder, schizophrenia, attention deficit hyperactivity disorder, and MDD as a group of disorders and lithium-responsiveness have been shown to have association signals enriched in 18 core clock genes and genes reported to be rhythmically expressed in more than six mouse tissues. This is the first GWAS evidence supporting potential genetic contributions of circadian system in neuropsychiatric disorders, although the selection of clock gene in this study may be debatable; and replication is warranted.

It is important to note that the CGs did not make any top GWAS signals of these psychiatric disorders, despite the fact that circadian disruption has been strongly implicated in neuropsychiatric disorders. Genetic studies of depression, anxiety, and addiction, have yielded largely negative results^[109–113]. The study by McCarthy et al. suggested collective weak contributions from clock genes for susceptibility of various disorders. This has multiple implications: First, it is possible that some other non-CG clock genes associated with disease are also part of the circadian system, but have not yet been identified as such. Secondly, clock genes may be more relevant to specific subtypes or endophenotypes of those diseases; therefore, subgroups of those disorders may provide better association with clock genes. Lastly, the circadian system may contribute to disease risk more through a non-genetic route, such as epigenetics.

Epigenetic factors regulate circadian rhythms

When genetics have yielded very limited success in providing direct evidence to link circadian rhythms to neuropsychiatric disorders, epigenetics naturally get our attention as the critical regulators (for gene expression) and mediators (for environmental factors). Then the first questions are whether epigenetic factors regulate circadian rhythms as they should be in theory, and what are those circadian epigenetic factors.

The circadian system is dynamic and flexible, and is tightly regulated by the interactions between internal molecular systems and environmental cues. The environmental factors, including light, food, temperature, stress, hormones, drugs, and age, act through epigenetic factors to shape the phenotypes. Gene expression, as the molecular representatives of circadian rhythms, is known to be regulated by genetic variants and epigenetic factors. Epigenetic factors include DNA methylation, histone modifications (methylation, acetylation, phosphorylation, and citrullination, biotinylation, ribosylation, ubiquitination, palmitoylation etc.), and non-coding RNAs. Studies on histone acetylation, DNA methylation, non-coding RNA, and RNA modification have shed light on their roles in regulating expression of clock genes, and ultimately, circadian phenotypes. With such studies, we could look for circadian epigenetic factors, and further study their contribution to neuropsychiatric disorders.

Acetylation and deacetylation

The epigenetic mechanism that CLK uses to regulate circadian rhythms is histone acetylation and deacetylation. Etchegaray et al (2003) showed in a mouse liver study that histone acetyltransferase (HAT) p300 works with Clock/Bmal1 complex to regulate histone H3 acetylation at the promoters of Cry and Per genes to influence their expressions.^[114] Doi et al (2006) further showed that CLK itself possesses HAT activity, which can be enhanced by its partner BMAL1, when bound to E-box ^[115]. CLK is also involved in acetylating other non-histone substrates including BMAL1. Acetylated BMAL1 recruits CRY1 to CLK-BMAL1 complex and represses transcription^[116].

Histone deacetylase (HDAC) has a function opposite to that of HAT, and is also an important regulator of circadian rhythms and memory formation, as well as metabolism. It removes acetyl groups from an ε-N-acetyl lysine on histones, allowing the histones to wrap the DNA more tightly. HDAC inhibitor, Valproic acid (VPA) and trichostatin A (TSA) were found to increased H3 acetylation and affect Per2 expression in an in vitro study^[117].

A mouse model has shown that Hdac3, one of the Hdac subtypes, is recruited by nuclear receptor corepressor 1 (Ncor1) and is involved in repressing Bmal1 expression, thus affecting circadian rhythms and metabolism^[118]. Hdac3 recruitment also fluctuates rhythmically in the mouse liver, in conjunction with Rev-erb-alpha and Ncor, to form a Hdac3/Rev-erb-alpha/Ncor complex^[119]. It is expected that transcriptions of many genes will oscillate with the fluctuation of HDAC3-related histone modification, or Rev-Erb-alpha/NCoR1-related signaling pathways.

Another member of the HDAC family, an NAD(+)-dependent protein deacetylase, SIRT1, also works directly with clock genes. SIRT1 binds CLK-BMAL1 and promotes deacetylation and degradation of the PER2 protein in mice^[120]. SIRT1 is also a metabolic sensor, as it requires binding of its coenzyme NAD+ for its HDAC enzymatic activity. Thus, through SIRT1, metabolic states are linked to the circadian system. In addition, SIRT1 has been implicated in aging and neurodegeneration^[121;122], synaptic plasticity and memory formation in mouse studies^[123;124].

The lysine-specific demethylase, JumonjiC (JmjC) and ARID domain-containing histone lysine demethylase 1a (JARID1a) are also major binding partners of CLK-BMAL1. It can inhibit HDAC1 function and enhance transcription by CLK-BMAL1 in a demethylase-independent manner. CLK-BMAL1 complex plays a conserved circadian regulatory role across insect and mammalian species^[125].

DNA methylation

The role of DNA methylation in circadian regulation is supported by a human study in which plasma homocysteine levels and global DNA methylation level showed 24 hour variation in blood of 15 males and 15 females^[126]. Homocysteine level has been linked to DNA methylation in multiple studies^[127]. An epigenome-wide study by methyl-DNA immunoprecipitation (MeDIP-chip) in mice showed that altered day length changed gene expression profiles and promoter DNA methylation in SCN, suggesting that DNA methylation regulates the circadian clock in SCN^[128]. Moreover, a study in mice showed that sleep deprivation can change DNA methylation and hydroxymethylation of hundreds to thousands of CpG sites near genes involving neuritogenesis and synaptic plasticity, cytoskeleton, signaling and neurotransmission^[129]. The direct evidence supporting the roles of DNA methylation in regulating circadian rhythms came from a human study, which used global statistics to show evidence of significant 24-hour rhythmicity of DNA methylation, as well as its correlation with rhythmic gene expression in human dorsolateral prefrontal cortex^[31].

Non-coding RNA

MicroRNAs (miRNAs), probably the most intensively studied class of non-coding RNAs (ncRNAs) so far, may contribute to the regulation of circadian rhythms. Dicer is the major enzyme in miRNA biogenesis, and Dicer-deficient mice and cells show shorter circadian cycles due to faster translation of PER1 and PER2 proteins. It has been proposed that microRNAs miR-24, miR-29a, and miR-30a may specifically target PER1 and PER2, thus determine the period of the cycle^[130].

Studies in mice also implicated miR-134 and miR-132 as two other miRNAs in circadian regulation. miR-134 is brain-specific, and regulated by SIRT1^[124]. It is involved in regulation of CREB and BDNF levels, two proteins that are important in many neuronal functions and activities^[124]. miR-132 is a direct link between light and chromatin remodeling: it is induced by photic entrainment cues via the mitogen-activated protein kinase (MAPK)–CREB signaling pathway^[131] and regulates chromatin remodeling and translation^[132].

Other ncRNAs have strong potential in regulating circadian rhythms too. Rhythmic expression was reported for 112 long non-coding RNAs (lncRNAs) in the rat pineal gland, which is the source of melatonin^[133], while melatonin is an important hormone timing circadian rhythms. A study of Neurospora gene frequency (frq) proved that lncRNAs could regulate circadian rhythms through anti-sense expression^[134]. RNA-Seq of period-null Drosophila identified several ncRNAs with diurnal expression, including a family of small nucleolar RNAs (snoRNA)^[135]. It should be noted that some non-coding RNAs, particularly long non-coding RNAs, evolved fast and are species-specific^[136;137]. These non-human findings only suggest some possible epigenetic mechanisms that may occur in human. The actual genes in human remain to be discovered.

RNA modification

Post-transcriptional RNA processing and modifications may be relevant to clock function. A recent study in mice and cultured human cells showed that n⁶-methyladenosine (m⁶A) RNA-methylation, one of the most common RNA modifications, was involved in circadian clock regulation^[138].

Studies also showed that diet could affect epigenetic regulation of circadian function. In a study of Japanese macaques, maternal high-fat diet in utero can disrupt expression regulation, and increase individual variations of fetal hepatic Npas2, one of the CGs. Such disruption was associated with altered histone acetylation (H3K14ac) but not DNA methylation at the Npas2 promoter region. The changes of gene expression and histone modification can be reversed by postnatal diet^[139]. Mouse exposed to different length of light a day changed SCN and neuronal Per1 gene expression and behavior after birth^[140], suggesting possible epigenetic modification induced by early-life environmental effects, although epigenetics was not part of the study.

The epigenetics of circadian systems is a new emerging research field, leaving a lot to be investigated. Most studies have been performed in mouse models, and only a few in humans. Since differences in epigenetic regulation between mice and humans during pre-implantation development have been observed^[141], findings from mice and other species may not translate to humans directly. Moreover, many epigenetic factors, such as hydroxymethylation, long non-coding RNAs, and most of the histone modifications other than acetylation, have not been studied in the context of circadian regulation in humans. Even for those factors studied, the findings are still fragmentary, and do not form one coherent picture of the regulatory system. It is not known whether those factors work independently or interactively to regulate each of the clock genes, or the circadian system as a whole, and how. For these reasons, we are advocating a more comprehensive epigenomic study of the circadian system in humans.

Circadian epigenetics in neuropsychiatric disorders

Although plenty of data has implicated circadian in the risk of neuropsyhicatric disorders, and that epigenetic factors are important regulators of circadian, only very limited studies have been performed to explore the epigenetics changes in neuopsychiatric disorders.

Sleep-related disorders

A few studies have been published on gene expression changes in disturbed sleep. The epigenetic regulation of those changes remains largely unknown as only one candidate gene study exists for DNA methylation.

Möller-Levet et al. studied gene expression profiles and reported that 711 genes were up- or down-regulated in the blood of people suffering from insufficient sleep. The number of genes with a circadian expression profile also reduced from 1,855 to 1,481^[142]. This same research group also studied blood transcriptome in desynchrony of sleep-wake timing and circadian rhythms, and identified a dramatic reduction of rhythmic transcripts (6.4% drop to 1.0%) caused by the desynchrony^[143]. The chromatin modification and expression regulation pathways were consistently implicated by the differentially expressed genes in these two sleep studies.

Bollatti et al. (2010) studied the effects of daytime and nighttime shiftwork and its effects on global DNA methylation and the methylation of the promoters of three candidate genes (glucocorticoid receptor (GCR), tumor necrosis factor alpha (TNF-alpha), and interferon-gamma (IFN-gamma)) using peripheral blood DNA from 100 shiftworkers and 50 day workers in Northern Italy. A small but significant difference in methylation was detected between morning and evening type shiftworkers in TNF-alpha promoter. However, no significant methylation difference was detected when comparing shiftworkers to day workers^[144]. It is noted that all the reported associations or correlations were weak.

Non-sleep neuropsychiatric disorders

Prader-Willi syndrome (PWS) is the first neuropsychiatry disorder with evidence of disrupted circadian epigenomics. PWS is a genetic disorder featured by obesity, intellectual disability and sleep abnormalities. This disorder frequently co-morbid with psychiatric problems^[145]. It is caused by a deletion on the paternal chromosome 15q11-q13, considered to be caused by loss of snoRNAs^[146], which are processed products of a lncRNA gene, 116HG. Study of mice lacking 116HG showed altered expression of several clock genes and energy use in brain^[147].

Another pilot study connecting genetics and epigenetics of clock genes to psychiatric disorders is on miRNA. The precursor of miR-182 was found to carry a SNP rs76481776 that was associated with late insomnia in MDD patients (corrected p <0.00625), in a study of 359 MDD patients and 341 control individuals. CLK is one predicted target of miR-182 and regulation relationship was validated through in vitro assays^[148]. This relationship between an SNP, the expression of a miRNA and its targets warrant further investigation. However, it does suggest that we should pay more attention to subtypes or endophenotypes rather than diagnostic classification, when studying genetics and epigenetics of those disorders.

Clearly, circadian epigenomics has not received sufficient attention in study of neuropsychiatric disorders although the circadian rhythms have been one of the major phenotypes to study in these disorders.

Future perspectives

Based on studies reviewed above, we see that knowledge of regulatory systems of circadian rhythms may provide opportunity to understand psychiatric disorders. However, we still have limited understanding of the broader network context of such circadian regulatory systems, particularly the epigenetic aspects of such regulation. Genes involved in circadian rhythms remain to be discovered and organized into network systems. Epigenomics need to be integrated in order to complete the expression regulatory circuitry, and connect with environmental factors. More importantly, considering such regulation and networks in the context of neuropsychiatric disorders would provide new perspectives on the link between circadian and human behaviors, therefore allowing better understanding of disorders.

Understanding the complex biological systems that produce them is necessary to decipher complex traits such as neuropsychiatric disorders^[149]. Circadian regulation is a complex biological system. Environmental factors and internal biological infrastructure work together: light acting through photoreceptors, food working through SIRT1-related pathways, and psychological stress through the HPA-neuroendocrine system--all regulate an organism’s internal clock. Serotonergic, dopaminergic, and maybe other neurotransmitters systems interact with circadian regulatory networks to influence human behavior and disorders.

Our current knowledge about circadian rhythms is largely derived from studies of candidate genes, for their biochemistry, genomics and epigenomic regulation. There are very limited genome-wide, systematic studies in human (Table 1). Genome-wide study is critical for obtaining unbiased understanding of biological systems. It is important to put existing knowledge into a biological network, to re-assess all the interactions and signaling connections. Novel components of circadian controls, from its environmental cues to the downstream effectors will be discovered. Several papers have been published to advocate the use of systems biology to study circadian rhythms, and to construct regulatory system of circadian rhythms through integration of multiple -omics^[150–152].

A complete circadian regulation system should contain every member of the circadian regulation cascades, from the core clock modulators to the downstream effector genes. It should also represent regulators at different levels, from environmental cues to epigenetic factors, RNA and protein modifications. Relationships among these nodes, genes and their interactions are critical parts of the systems. The use of such systems holds the key to understanding of circadian rhythms and its role in neuropsychiatric disorders.

Regulatory systems are spatiotemporally-specific. This suggests that tissue selection is important for the study of circadian rhythms and psychiatric disorders. Brain is the critical organ/tissue for understanding neuropsychiatric traits or disorders, including those that are circadian-related, particularly as the circadian regulation has tissue-specificity. However, other than imaging studies, it is almost impossible to study live human brain for circadian dynamics. Excessive assumptions have to be made in the analyses of human postmortem data. Different individuals could differ by variables other than the time of death. Model animal brains, cultured or induced neuronal cells derived from stem cells, and human blood are a few alternatives that could provide multiple time-point data around clock. However, each model system has its own limitations. Complementary use of these different models may help us build a comprehensive understanding of the regulatory network and its relevance to human circadian-related traits and disorders.

Genomic and epigenomic studies of patient and control samples should also take into account the diurnal variations. Hundreds or even more genes have variable gene expression or epigenetic markers within 24 hours in a day. In the past, the time when data or material was collected has rarely been recorded and incorporated into analyses. As a result, artifacts may have been introduced in some published data unless the sample collection was done at a similar time of a day. Circadian studies in healthy humans will provide critical baseline information for other studies when time of day data are not available.

The findings from genetic and epigenetic studies could lead to novel drug targets. Belsomra, a hypocretin receptor antagonist, was recently approved by FDA as a new drug to treat insomnia. Hypocretin has been connected to sleep-wake cycle since the discovery of a mutation responsible for narcolepsy, though it is not considered part of the CGs. We may have many other drug targets buried in the list of rhythmic expression or epigenetic regulation of circadian rhythms, the complete circadian regulation systems. Epigenetic drugs have big potentials in treating neuropsychiatric disorders. Theoretically, it will be much easier to use drugs to modify the epigenome than to correct mutated genes.

Understanding the genetics and epigenetics of circadian-related traits and diseases will lead to better and more precise diagnosis of circadian-related disorders. Ultimately, it will improve the quality of life for people suffering from disorders, such as jet lag or shiftwork, when we are able to develop epigenetic intervention to ease the pains and discomfort.

Through our review, it is clear that epigenetics may play important roles in regulating circadian rhythms and associated neuropsychiatric disorders, but related studies are lacking today. The circadian cycle is a highly environment-dependent biological process. The circadian cycle is one of the best models to study environmental impact and gene-environment interactions. Much effort should be placed into this interesting research field. Using the circadian-related phenotypes and biomarkers, we may have an exceptional opportunity to access the dark kernel of psychiatric disorders.